University of Southern California Dissertations and Theses

Identifying neural markers of centralized pain

(USC Thesis Other)

Identifying neural markers of centralized pain

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content IDENTIFYING NEURAL MARKERS OF CENTRALIZED PAIN
by
Natalie Jo McLain
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(BIOKINESIOLOGY)
August 2024

ii
Dedication
To my family, given and chosen

iii
Acknowledgements
This work is the culmination of five years of learning and exploring, and it would not have been
possible without several people to whom I owe a great deal of thanks.
I, of course, have so much gratitude for the support and guidance provided by my advisor, Dr.
Jason Kutch. You took a chance on a student with no computational know-how and no master’s
degree, and I’m so grateful for all you’ve taught me. You helped me make the scary leap from
wetlab to neuroimaging, and I’ve grown into a better scientist and (marginally) better coder
under your guidance. You’ve always had my back, made time for me, and facilitated a PhD
experience with a healthy balance between work, passion projects, and a life outside of the lab.
I hope I can one day replicate a space like AMPL.
I owe so much of the development of this manuscript to my committee members. Thank you to
Dr. Kay Jann for his patience in helping me parse neuroimaging papers and giving me his
thoughts on random pieces of data sent via Slack. To Dr. Beth Smith for her input on my EEG
pipelines but also for the many conversations that extended beyond science and helped support
my career goals. To Dr. Nicolas Schweighofer for always challenging me, making my analyses
more rigorous, and deepening my understanding of statistics. And to Dr. Andrew Schrepf for his
guidance in understanding the mechanisms of pain and generally being a source of constant
positivity. I am so appreciative of your collective support.
Thank you to the Division of Biokinesiology and Physical Therapy for providing funding support
through Teaching Assistantships as well as travel grants and funding for socials to engage with
all of my classmates. Thank you to my advisor Dr. Jason Kutch for funding my Research
Assistantships and facilitating conference attendance early in the program. These were
invaluable resources for putting together my scientific aims.

iv
To all the members of AMPL, past and present, you have helped and supported me in so many
ways. Thank you to Jayati Upadhyay, Jason Cherin, and Matt Heindel for support in the lab, but
also for providing a space to vent and socialize. I’m so lucky to have a lab that is also a group of
friends. Thank you to Dr. Eileen Johnson for sharing your clinical knowledge and kind spirit. I
will miss our thresholding team. To Dr. Moheb Yani, who took so much of his own time to teach
me when I started at USC, let me test new collection protocols on him, and is still a source of
inspiration and encouragement. Giselle Garcia, data collection extraordinaire, for all of her in-lab
help, after-hours texts, and walk-and-talks. To the undergraduate interns: Arike Coker, Olivia
Means, Mia Montel, and Lauren Tomita (the self-dubbed “fan girls”), you taught me about
mentorship and provided so much support to everyone in the lab. From tedious literature
searches to data collection, you were on top of it. I am so excited to see what you all do next; I
know it will be something impressive.
A huge, sweeping thank you to the Cline Lab, without which I probably wouldn’t have a research
career. Particularly to my mentors, Dr. Holly Cline, Dr. Caroline McKeown, Dr. Haiyan He, and
Dr. Abby Gambrill. You are an amazing group of women that prepared me with a strong sense
of curiosity, organization, and boundaries that made grad school much easier to manage. Thank
you for teaching me to always ask questions and make space for myself.
To the educators who fostered creativity and excitement about my place in the natural world
early and with enthusiasm, and especially those who helped me through uncertainty during
important transition periods in my academic trajectory: Mr. Ring from South Pasadena High
School, Dr. John Serences from UC San Diego, and Dr. Dave Singer from San Diego City
College. Thank you also to Dr. Kristan Leech and Dr. Carly Lochala for supporting my interest in
teaching and facilitating my fellowship through the Future Faculty Institute. I learned so much
from you both.

v
To the USC Figure Skating team, the Pasadena lunch bunch crew, and my whole figure skating
community. Thank you for giving me a place to be artistic and active. To Jovicarole Raya, my
friend and babberzino. You probably know best what a difficult journey it has been to get to this
point, and I couldn’t have done it without you. You are talented, knowledgeable, so dedicated to
other people’s success, and will probably be my boss someday. To Shreya Jain and Sarah
Kettlety, thank you for being my unwavering support system during our time at USC. Yes, those
three are always together. By the end of this summer, those three will all officially be doctors! I
can’t think of a better pair of people to have shared these five years with. Your knowledge of
your fields and willingness to speak up inspires me.
To my family, and especially my mom, who has been so supportive of every step in my
academic journey. Thank you for introducing me to the most important principles of science: our
nature walks looking at bugs and picking flowers in childhood taught me to notice the little things
and be inquisitive, skills I use and treasure every day. To Roy, who started the program as my
partner and ended it as my husband. You challenge me to consider new perspectives and
believe in my own competence, something I often need help with. You push me forward, both
metaphorically and literally, because you let me take your scooter to campus. And finally, to my
late grandmother, whose phone calls were the soundtrack of my walk to the car in the first year
of grad school. That stretch of Eastlake still summons your voice, and I think so often of how
you encouraged me to follow the sunshine and make myself heard. Even in your passing, you
taught me a final lesson about compassion, service to others, and unconditional love. I miss you
every day, but I know you are still with me.
You all have been the village that made this journey possible, the et al. to my McLain 2019-
2024. Thank you.

vi
Table of Contents
Dedication________________________________________________________________________ ii
Acknowledgements _______________________________________________________________iii
List of Tables ___________________________________________________________________ viii
List of Figures_____________________________________________________________________ix
Abstract __________________________________________________________________________xi
Chapter 1. General Overview ______________________________________________________ 1
Chapter 2. Background____________________________________________________________ 6
2.1. A spectrum of drivers: what do we mean by centralized pain? ________________________ 6
2.2. The nociceptive pathway____________________________________________________________ 7
2.3. Clinical questionnaires and centralized pain________________________________________ 13
2.4. Measures of neural activity ________________________________________________________ 16
Chapter 3. Analytic stability and neural correlates of peak alpha frequency ________ 21
3.1. Introduction_______________________________________________________________________ 22
3.2. Methods __________________________________________________________________________ 24
3.3. Results____________________________________________________________________________ 39
3.4. Discussion ________________________________________________________________________ 53
Chapter 4. Brain connectivity associated with chronic pain intensity within
individuals: a 3-year longitudinal study of the MAPP Research Network ____________ 63
4.1. Introduction_______________________________________________________________________ 64
4.2. Materials and Methods ____________________________________________________________ 66
4.3. Results____________________________________________________________________________ 78
4.4. Discussion _______________________________________________________________________103
Chapter 5. Peak alpha frequency is associated with chronic pain diagnosis
but not pain intensity ___________________________________________________________109
5.1. Introduction______________________________________________________________________110
5.2. Materials and Methods ___________________________________________________________112
5.3. Results___________________________________________________________________________120
5.4. Discussion _______________________________________________________________________126
Chapter 6. Discussion___________________________________________________________131

vii
6.1. Findings from our work ___________________________________________________________131
6.2. Impact of Dissertation ____________________________________________________________134
6.3. Limitations and future work _______________________________________________________135
References _____________________________________________________________________137

viii
List of Tables
Table 3.1Processing pipeline data from the 17 articles reviewed. .............................................................36
Table 3.2 Participant data from the 17 papers reviewed............................................................................38
Table 3.3. Summary statistics and matrix of Pearson’s correlation coefficient for electrode ROIs
(four most common ROIs in the PAF-pain literature reviewed). .........................................................46
Table 3.4 Cluster maxima coordinates for significant clusters from the fMRI-PAF analysis. .....................50
Table 3.5 Cluster maxima coordinates for significant clusters from the slow-5 fALFF with EEG ROI
analysis...............................................................................................................................................52
Table 4.1 Demographic information for the 492 SPS subjects included in the analysis.............................78
Table 4.2 Summary scores across all visits for the 5 clinical measures entered into Models
2 and 5, mean and SD reported for males, females, and across the whole population (total)...........80
Table 4.3 Correlation matrix for the 5 clinical measures at baseline visit...................................................80
Table 4.4 MNI information for the node pairs with a significant relationship between f~ pain
from Model 4.......................................................................................................................................89
Table 4.5 Matrix representing the percent of connections significant for pain in Model 4 in each
pair of networks. More information on rank and total number of connections can be found in
Table 4.6.............................................................................................................................................90
Table 4.6. Ten blocks with the highest percentage of connections significant for f~pain in Model 4. ........90
Table 4.7 MNI information for the node pairs with a significant relationship between
f~ multisite score*pain from Model 5. .................................................................................................96
Table 4.8 Matrix representing the percent of connections significant for pain*multisite score in
Model 5 in each pair of networks. More information on rank and total number of connections
can be found in Table 9. .....................................................................................................................97
Table 4.9 Ten blocks with the highest percentage of connections significant for
f~pain*multisite score in Model 5........................................................................................................97
Table 4.10 Ten blocks with the highest percentage of connections significant for f~pain in Model
4. node pairs that were significant for f~pain in Model 4 and f~ multisite score in Model 5..............101
Table 5.1 Demographic information and clinical characteristics for three groups in dataset one:
chronic widespread pain, chronic back pain, and healthy controls. NC=not collected .....................119
Table 5.2 Regression table for the models run on the global PAF values................................................123
Table 5.3 Demographic information and clinical characteristics for the two groups in dataset
two: patients with widespread pain (3 or more painful body sites) and patients with
localized pain (2 or fewer painful body sites)....................................................................................124
Table 5.4. Regression table for the models run on global PAF values.....................................................125

ix
List of Figures
Figure 1.1 Recording centralized pain..........................................................................................................2
Figure 2.1 Basic description and illustration of the four categories of pain proposed by IASP ....................6
Figure 2.2 Ascending (red) and descending (blue) pain pathways. .............................................................8
Figure 2.3 The dynamic pain connectome model.......................................................................................12
Figure 3.1 PRISMA flow diagram outlining the steps of the literature search for Chapter 3. ....................34
Figure 3.2 Sixty-four channel EEG data used to examine differences in individuals with high and
low global-average PAF values. .........................................................................................................39
Figure 3.3 Comparison of EEG data after being run through each of the three pre-processing
pipelines. ............................................................................................................................................41
Figure 3.4 Comparison of grand-average PAF values calculated with seven different epoch lengths.......43
Figure 3.5 Comparison of grand-average PAF calculated with three different alpha band bounds. ..........45
Figure 3.6 Comparison of grand-average PAF calculated from four ROIs.................................................47
Figure 3.7 Association of grand-average PAF varied across alpha band and calculation method
with fMRI data.....................................................................................................................................48
Figure 3.8 Association of ROI average PAF with fMRI data.......................................................................51
Figure 4.1 Mean (left) and max (right) framewise displacement for all scans. ...........................................69
Figure 4.2 In line with previous results, SPS participants experienced fluctuations in
average self-reported pain across the four visits (36 month timeline). ...............................................79
Figure 4.3 Results from Model 1: functional connectivity within the pain network in bladder
full and bladder empty state associated with average pain in the week preceding the visit
(PainRecent) and pain at the time of scan (PainNow)........................................................................81
Figure 4.4 Results from Model 2: the relationship between functional connectivity and pain
within the pain network is differentially modified by clinical variables based on bladder
state (empty versus full) and pain metric (PainNow versus PainRecent)...........................................82
Figure 4.5. Results from Model 5: when the relationship between functional connectivity and
pain was assessed across the whole brain (Schaefer atlas), only the full bladder scans
produced significant results.. ..............................................................................................................83
Figure 4.6 Changes in connectivity associated with pain fluctuations within participants
across four visits.................................................................................................................................85
Figure 4.7 Changes in connectivity associated with pain fluctuations within participants
across four visits in the Brainnetome atlas. ........................................................................................86
Figure 4.8 Changes in connectivity associated with the interaction between multisite pain and
pain fluctuations within participants across four visits. .......................................................................92
Figure 4.9 Changes in connectivity associated with the interaction between multisite pain and
pain fluctuations within participants across four visits in the Brainnetome atlas. ...............................93
Figure 4.10 Unscaled version of Figure 6a for f~pain and Figure 8b for f~pain*multisite
score. Networks are color-coded and all spheres are the same size, regardless of the
number of significant node pairs they are a part of. ...........................................................................98
Figure 4.11 Areas of connectivity with significant relationships to both pain (Model 4) and
pain*multisite score (Model 2). ...........................................................................................................99
Figure 4.12 Significant areas unique to the pain only model (a and b) versus the those unique
to the pain*multisite score model (c and d) ......................................................................................100
Figure 4.13 Impact of pain on physical and mental functioning as assessed by the SF12. .....................102
Figure 4.14 Pain ratings were well predicted by whether the patient reported being in a flare
(yes or no) (p<<0.01)........................................................................................................................102
Figure 5.1 Group differences in global average PAF:back pain versus controls versus
widespread pain. ..............................................................................................................................120

x
Figure 5.2 Averaged power spectrum for the three groups (chronic back pain, chronic
widespread pain, and controls) across four ROIs.............................................................................122
Figure 5.3 Distribution of the group differences in PAF at each channel: a. Back pain - controls,
b. Widespread pain - controls, c. widespread pain - back pain. .......................................................122
Figure 5.4 β values widespread pain versus localized pain as a predictor of PAF values in the
UCPPS dataset after controlling for age, sex, and depression. .......................................................126

xi
Abstract
Chronic pain is an impactful condition with highly variable presentation and etiology. It is
additionally characterized by no widely effective treatments and a subset of patients who never
fully recover. Understanding the origins of pain is critical to informing effective treatments. One
relevant metric by which to divide pain origin is peripheral versus centralized: peripheral pain
can be defined as direct activation of the nociceptors (damage or threat in the periphery) while
centralized pain is compounded by dysregulation of the central nervous system. Past work
suggests peripheral and centralized pain are mediated by different neural circuitry. Additionally,
symptom pattern studies indicate that centralized pain conditions have distinct symptom profiles
from conditions that are driven primarily by peripheral nociceptive input. Attempts to classify or
assign treatment using clinical questionnaire scores alone, however, have been unreliable. This
work is focused on the combination of clinical phenotypes with two complementary modalities
for measuring brain function: functional magnetic resonance imaging (fMRI) and
electroencephalography (EEG). This project aims to use EEG, a neural recording technique that
is clinically accessible, to identify a neural marker of centralized pain through association with
putative clinical measures of centralized pain (widespread pain, heightened pain
unpleasantness, worsened affect). Additionally, we aim to understand what this EEG marker
indicates about differences in brain function that distinguish centralized from peripheral pain
types using fMRI. The overall goal of our proposed work is to determine how resting state peak
alpha frequency (rs-PAF), a promising EEG marker, is associated with features of centralized
pain and what it reflects about chronic pain-related changes in brain function. This project will
answer important questions about patients experiencing pain with heightened involvement of
their central nervous system in pain amplification and chronification, allow for targeted
assignment of existing therapies, and provide the basis for future development of targeted
treatments based on the neural circuits identified. Chapter 3 will test the analytic stability of rsPAF and then associate it with rs-fMRI measures to elucidate what individual differences in rs-

xii
PAF indicate about brain function. Chapter 4 will longitudinally assess the relationship between
rs-fMRI measures and fluctuations in pain as modified by clinical questionnaires thought to
reflect centralized pain types. Chapter 5 will associate rs-PAF with clinical questionnaires
thought to reflect centralized pain types and explore diagnosis-specific differences. The
proposed work will have specific implications for the clinical understanding and treatment of
chronic pain, as well as more broadly for the understanding of neural activity markers that can
be identified and utilized in interventions for centralized chronic pain.
Chronic pain is an impactful condition characterized by no widely effective treatments
and a subset of patients who never fully recover. Understanding the origins of pain is critical to
informing effective treatments. One relevant metric by which to divide pain origin is peripheral
versus centralized: peripheral pain can be defined as direct activation of the nociceptors
(damage or threat in the periphery) while centralized pain is compounded by dysregulation of
the central nervous system. Past work suggests peripheral and centralized pain are mediated
by different neural circuitry. Additionally, symptom pattern studies indicate that centralized pain
conditions have distinct symptom profiles from conditions that are driven primarily by peripheral
nociceptive input. This work is focused on the combination of clinical phenotypes with two
complementary modalities for measuring brain function: functional magnetic resonance imaging
(fMRI) and electroencephalography (EEG). The overall goal of our proposed work is to
determine how resting state peak alpha frequency (rs-PAF), a promising EEG marker, is
associated with phenotypic features of centralized pain and what it reflects about changes in
brain function using fMRI. This project will answer important questions about pain patients with
heightened involvement of their central nervous system in pain amplification and chronification,
allow for targeted assignment of existing therapies, and provide the basis for future
development of targeted treatments based on the neural circuits identified. The proposed work
will have specific implications for the clinical understanding and treatment of chronic pain, as

xiii
well as more broadly for the understanding of neural activity markers that can be identified and
utilized in interventions for centralized chronic pain.

1
Chapter 1. General Overview
Chronic pain has been identified as a global research priority, with an estimated
worldwide prevalence of 10-25% (Goldberg and McGee, 2011). It is an impactful condition with
highly variable presentation and etiology, additionally characterized by no widely effective
treatments and a subset of patients who never fully recover (Dahlhamer et al., 2018). There is a
large body of literature dedicated to parsing the different mechanisms and factors that can
amplify risk for chronic pain or increased pain sensitivity, but subtyping patients with pain in a
clinically meaningful way can be difficult (Edwards et al., 2016). Our lab’s primary population of
interest is Urologic Chronic Pelvic Pain Syndrome (UCPPS). There are currently no effective
treatments and no established biomarkers for UCPPS (Clemens et al., 2014; Hosier et al.,
2018). The majority of well-conducted randomized trials evaluating current treatment options
have not demonstrated efficacy over placebo (Clemens et al., 2014). UCPPS fits into the larger
category of pain and sensory sensitivity disorders where patients may have mixed etiologies
despite a shared diagnosis. We propose that being able to subtype pain patients (including
those with UCPPS) based on pain source would present a way forward in personalizing
treatment plans and improving patient outcomes. Our proposed metric for this stratification is
centralized versus peripheral pain types based on activity in the brain. While there are important
mechanisms of centralized pain at the level of the spinal cord, medulla, and midbrain (Figure
1.1), this project will focus on mechanisms that are possible to investigate using fMRI and EEG
as in previous Multidisciplinary Approach to the Study of Chronic Pelvic Pain (MAPP) Research
Network study projects assessing centralized pain symptom profiles (Kutch et al., 2017).
While a variety of self-reported, questionnaire-based clinical features (e.g. widespread
body pain, mood disturbances, increased pain unpleasantness) have been associated with
centralized pain, there is a strong need for an objective marker that is accessible to capture and

2
calculate. Past attempts to use clinical features alone for classification or treatment response
have been unreliable (Mors⊘ et al., 2021; Rifbjerg-Madsen et al., 2018; Soni et al., 2019;
Trainor & Miner, 2008), and even successful classification using the synthesis of multiple clinical
features leave us without an understanding of the underlying mechanisms separating pain types
(Kratz et al., 2021; Reimer et al., 2017). An understanding of the network changes
distinguishing centralized from peripheral pain opens the door for developing new, targeted
treatments. Our lab is leading the way in using neuroimaging to design brain-targeted therapies
for generalized markers for pain (Yani et al., 2019), but has yet to address the centralized
versus peripheral distinction. Additionally, resting-state functional magnetic resonance imaging
(rs-fMRI), the primary method used in our past publications, can be clinically inaccessible.
Electroencephalography (EEG), by contrast, is fast and relatively affordable to capture, in
addition to being simpler to integrate with concurrent neuromodulation. Therefore, we aim to
identify an EEG marker that distinguishes centralized from peripheral pain, and interpret
what that EEG marker indicates about changes in brain function using fMRI.
Figure 1.1 Recording centralized pain. Recording centralized pain. Amplification of peripheral pain signals can occur
at several places within the central nervous system: DRG=dorsal root ganglia, RVM=Rostral ventromedial medulla,
PAG=periaqueductal gray, DLPFC= dorsolateral prefrontal cortex

3
The perception of pain intrinsically requires altered patterns of brain activity: chronic pain
models in animals have consistently shown central activity markers that can identify chronic and
acute pain states (da Silva & Seminowicz, 2019), and human research has identified differences
in neural activity between chronic pain populations and healthy controls that could be the basis
for diagnostic biomarkers (van der Miesen et al., 2019). However, central amplification of pain
and peripheral nociceptive input are mediated by different pathways and areas of the brain
(Martucci & Mackey, 2018; Phillips & Clauw, 2011). We, therefore, expect the biomarkers for
pain conditions that are more driven by peripheral, nociceptive input to be distinct from markers
for conditions driven by central factors. A potential marker of interest that is non-invasive to
capture is peak alpha frequency (PAF). This EEG measure refers to the most prominent
frequency within the alpha band of brain wave activity, which typically ranges between 7 and 13
Hz (McLain et al., 2022), and slow PAF is established as a marker of acute pain sensitivity.
Further work in chronic pain is needed, but variability across the literature seems to suggest
slowed rs-PAF may not be a generalized marker for chronic pain states, and instead be specific
to certain types of chronic pain such as neuropathic or centralized (Boord et al., 2008; Furman
et al., 2019; Krupina et al., 2020; Llinas et al., 1999; Nir et al., 2010; Schmidt et al., 2012).
Understanding the origins of pain is critical to informing effective treatments. One
relevant metric by which to divide pain origin is peripheral versus centralized: peripheral pain
can be defined as direct activation of the nociceptors (damage or threat in the periphery) while
centralized pain is compounded by dysregulation of the central nervous system. While
peripheral pain will likely respond best to peripherally directed treatments, centralized conditions
may respond better to the addition of centrally acting interventions. Therefore, being able to
distinguish peripheral from centralized pain types is of great clinical significance.
There are very few biomarkers developed/validated to the point of use in clinical practice
for pain (C.-W. Woo et al., 2017), and none yet employed in the treatment or diagnosis of

4
UCPPS patients. Using fMRI data in UCPPS, our lab is developing targeted brain stimulation
approaches but hasn’t yet addressed peripheral versus centralized pain types. Several of the
large datasets used to develop these stimulation approaches specific to UCPPS include EEG,
fMRI, and clinical questionnaire data. Advantaged by these multimodal datasets, I am uniquely
positioned to make critical steps in identifying and validating a marker for centralized pain and
understanding what that marker reflects about changes in brain function.
This work will leverage four datasets:
a. LAMPP dataset: observational EEG and fMRI data collected as part of “Sensorimotor
impairments in men with Chronic Prostatitis/Chronic Pelvic Pain Syndrome.” Includes
resting-state EEG and fMRI data as well as demographic data for 47 healthy men and
42 men with chronic prostatitis
b. MAPP dataset: fMRI and demographic/questionnaire data set from Multidisciplinary
Approach to The Study of Chronic Pelvic Pain (MAPP) Research Network’s Symptom
Patterns Study (SPS) with 492 males and females with UCPPS
c. Ploner dataset: A large, publicly available dataset with resting-state EEG and
questionnaire data from 101 chronic pain patients of variable diagnosis along with 87
age and sex matched, healthy controls
d. IC/BPS dataset: EEG and clinical questionnaire data from 28 women with interstitial
cystitis/bladder pain syndrome (IC/BPS). This data was collected at the baseline visit for
a clinical trial investigating repetitive transcranial magnetic stimulation as an intervention
for pelvic pain. The data included in this project was collected before the start of the
intervention
Using these datasets, I will first investigate whether PAF is stable against common
processing decisions made in the pain literature (LAMPP dataset) and if PAF is associated with
local activity in any regions that are related to the experience of chronic pain. In a large

5
population of chronic pelvic pain patients with variable clinical presentation (MAPP dataset), I
will then investigate what regions of functional connectivity underlie fluctuations in reported pain
in the chronic pain state, and whether clinical features of centralized pain (i.e. widespread pain,
depression, anxiety) affect this relationship between brain activity and reported pain level.
Finally, I will use a large dataset of chronic pain patients with variable diagnoses to compare
patient groups against one another and determine whether a potential marker for chronic pain,
peak alpha frequency (PAF), varies by diagnosis type. This analysis will shift the focus from
patient control differences to within-patient and between-diagnosis differences. We aim to bring
much-needed understanding to the different subtypes of chronic pain patients and the potential,
unique differences in brain activity and clinical profiles that distinguish them.

6
Chapter 2. Background
2.1. A spectrum of drivers: what do we mean by centralized pain?
Understanding the origins of pain is critical to informing effective treatments. One relevant
spectrum along which to categorize pain origin is peripherally-generated versus centralized:
peripherally-generated pain can be defined as direct activation of the nociceptors (damage or
threat in the periphery) while centralized pain is compounded by dysregulation of the central
nervous system. However, centrally maintained pain can be further subdivided based on the
mechanisms underlying pain chronification/sensitivity.
Figure 2.1 Basic description and illustration of the four categories of pain proposed by IASP (Note that central and
peripheral neuropathic pain are included in a single inset). From Cohen et al., 2021.

7
Previously published literature and the International Association for the Study of Pain’s
definitions of pain (“News, Announcements, and IASP,” 2017) separate centralized pain into two
general categories: central neuropathic pain and central amplification (also referred to as
“nociplastic”). Central neuropathic pain is a condition with some discernible damage to the
central nervous system. Based on previous literature, this type of central pain would likely
include disorders such as multiple sclerosis and spinal cord injury. In central amplification, there
is no clear evidence of damage to the central nervous system. Central amplification would,
based on previous literature, include disorders such as fibromyalgia and bladder pain syndrome
(Figure 2.1). Importantly, individuals can (and often do) have overlapping pain types (Cioffi,
2018). Here I discuss some of the specific mechanisms of centralized pain.
2.2. The nociceptive pathway
Before describing how pain signals can be amplified or modulated, it is helpful to first define
some of the basic components of the nociceptive pathway. Sensory afferents are made up of
neurons that respond to stimuli in the environment or within the body and transmit this
information along axons to a synapse in the dorsal horn of the spinal cord. Different classes of
sensory afferents have different levels of myelination along their axons (fibers), affecting the
speed at which they transmit information. The three primary fiber types relevant to this work are
Aβ, Aδ, and C fibers. Aβ fibers transmit non-noxious stimuli, primarily information about touch
from the skin. They are well myelinated and large in diameter, meaning they transmit
information rapidly. Aδ fibers carry multiple types of nociceptive information (mechanical,
thermal, and chemical) and are smaller in diameter. While still myelinated, they transmit slower
than the Aβ fibers. Finally, C fibers carry what is often called “second pain” information. They
are the smallest in diameter and unmyelinated, which means they transmit information
extremely slowly. C fibers are associated with the dull ache that might follow the initial, sharp
pain from an acute pain stimulus, but they are also associated with the dull ache from visceral

8
pain and pain from burns/heat (Kandel & Professor of Biochemistry and Molecular Biophysics
Thomas M Jessell, M D, 2000; Nolte et al., 2016).
2.2.1. Ascending pain amplification
Ascending pain amplification starts as early as the synapses of pain fibers in the dorsal horn of
the spinal cord. At this level, we can consider pain gating as somewhere at the crossroads of
ascending pain amplification and deficiency in pain modulation. The primary mechanism of
spinal cord pain gating is through the modulation of ascending signals at the intersection of the
central and peripheral systems. In short, activation of inhibitory interneurons is thought to “close
the gate” and prevent pain signals from C fibers, known to transmit the slow, “second pain”
signal, from traveling up to the brain (Basbaum & Fields, 1978; Mendell, 2014). One of the
earliest identified forms of pain gating is the mediation of inhibitory interneurons onto wide
dynamic range cells (Figure 2.2). Wide dynamic range cells receive both tactile and painful
stimuli from aβ and C fibers, respectively. Activation of the aβ through tactile stimulation, as you
would get from massage or rubbing an area that
was injured, excites the inhibitory interneuron,
which then inhibits the wide dynamic range
neuron and decreases the chances it will transmit
nociceptive signals from the C fiber (Melzack,
1965). However, any disinhibition of the system
(disruption of glycinergic and GABAergic signaling
in the spinal cord) could result in “opening the
gate” for nociceptive signals from the periphery to
higher cortical centers in a similar way (Cioffi,
2018). Recent studies have shown that, in fact,
similar circuits in the dorsal horn can experience Figure 2.2 Ascending (red) and descending (blue) pain pathways.
Taken from Cioffi, 2018.

9
both short- and long-term disinhibition by genetic and chemical factors (Takazawa &
MacDermott, 2010; Taylor, 2009; Todd, 2015; Zeilhofer, 2005) and contribute to pathological
pain signaling (Takazawa & MacDermott, 2010). This may be of particular importance for the
development of allodynia, the perception of pain at what is typically a non-painful stimulus (e.g.,
light touch), and hyperalgesia, increased sensitivity to painful stimuli (Cioffi, 2018), but may also
suggest certain individuals with lower endogenous levels of inhibitory control in their spinal
cords may have a lower baseline sensitivity to pain.
The pain circuitry in the spinal cord is subject to synaptic plasticity just like pain circuits
in the brain: any repeated or salient noxious input from the periphery can potentially contribute
to sensitization and maladaptive pain signaling in the spinal cord (Ferguson et al., 2012).
Repeated stimulation of C fibers, for example, leads to “wind up” through synaptic plasticity in
the spinal cord (Mendell, 1966; Mendell & Wall, 1965) and nerve injury can result in disinhibition
of the spinal cord interneurons, particularly at the dorsal horn (Castro-Lopes et al., 1993; Lever
et al., 2003; Moore et al., 2002). This may, in part, explain how peripheral neuropathic pain
patients transition to central neuropathic pain patients.
Several areas of the brain have also been associated with the amplification of ascending
pain. In an fMRI experiment, the thalamus, sensory cortices, anterior and posterior insula, and
anterior cingulate cortex (ACC) were all associated with thermal, temporal summation of pain
(Staud et al., 2007). Therefore, ascending pain amplification likely takes place at the level of the
spinal cord, brain stem, and in various brain regions.
2.2.2. Deficiencies in pain modulation
While most of the spinal cord mechanisms mentioned in ascending pain amplification
can be driven by peripheral stimuli, there is less evidence for impact of psychological factors on
mechanisms such as wind-up or spinal cord pain gating (unless through an indirect mechanism
like immune system activity or descending disinhibition). Deficiencies in pain modulation, by

10
contrast, are more readily connected to psychological factors such as acute stress (Linton &
Shaw, 2011). This may be important for the discussion of putative psychological markers of
centrally maintained pain such as depression and pain catastrophizing.
The periaqueductal gray (PAG) is one of the best-researched centers for descending
pain modulation. PAG directly receives nociceptive and thermal information via the
spinomesencephalic tract and can send descending pain inhibition signals (Nolte et al., 2016) in
two primary ways. Firstly, activity in the PAG promotes the secretion of serotonin (5-HT) through
the raphe nuclei which, through a series of connections, facilitates the block of substance P
release, an inflammatory neuropeptide (Suvas, 2017). It additionally blocks ascending pain
signals that would otherwise pass through the thalamus. The PAG also projects to the rostral
ventromedial medulla (RVM) which sends both excitatory and inhibitory fibers to the dorsal horn
of the spinal cord (Kandel & Professor of Biochemistry and Molecular Biophysics Thomas M
Jessell, M D, 2000; Nolte et al., 2016). Descending pain facilitation from the RVM has been
proposed as a mechanism for maintenance (though not initiation) of hyperalgesia and allodynia
following nerve injury in rats (Vera-Portocarrero et al., 2006). Note that this descending
inhibition by the PAG is also considered part of the gate theory of pain (Basbaum & Fields,
1978) previously mentioned in the ascending pain amplification section.
The connectivity between the PAG and the thalamus is also important for directing
attention to and away from painful stimuli (Valet et al., 2004). The thalamus receives pressure,
pain, temperature, itch, and touch information from the spinothalamic tract, and is an important
center for sensory integration (Nolte et al., 2016). The thalamus has long been suspected to
play a role in the chronification of pain through thalamocortical dysrhythmia: this theory posits
that high threshold bursting in a subset of thalamocortical neurons slows as a result of reduced
modulatory corticothalamic feedback and contributes to pathologies such as chronic pain
(Hughes & Crunelli, 2005; da Silva et al., 1980). This, however, has yet to be substantiated in

11
human models (Hughes & Crunelli, 2005) and recent work in human intracranial recordings
(Halgren et al., 2019) and electroencephalography (EEG power) analysis in source space (Ta
Dinh et al., 2019) indeed casts doubt on the theory.
In addition to lower-level centers such as the PAG, medulla, and thalamus, descending
pain modulation also includes modulation from higher-order centers such as the dorsolateral
prefrontal cortex (DLPFC). Activity in the DLPFC during experimentally induced heat allodynia
has an inverse relationship with the affective component of pain (perceived intensity and
unpleasantness) and is thought to regulate other pathways responsible for the experience of
pain (Lorenz et al., 2003). It is additionally suggested that modulation by the DLPFC is what
allows some individuals to better ignore pain and attend to other tasks, while others struggle to
focus on stimuli other than pain (Kucyi et al., 2013; Kucyi & Davis, 2015). This draw on
attentional resources could contribute to the documented response slowing on cognitive tasks in
chronic pain patients (Lee et al., 2010; Nadar et al., 2016).
An organized introduction to some of the brain regions involved in attention to pain are
outlined in the dynamic pain connectome model (Kucyi & Davis, 2015). This model proposes
three networks as being responsible for attention to or away from painful stimuli: the salience
network, the default mode network, and the antinociceptive system. While the networks outlined
in this model (Figure 2.1) do not completely account for all the regions implicated in pain
modulation, it is a good overview of attentional control in particular. Individuals with high intrinsic
attention to pain (IAP) have lower structural connectivity and less flexible functional connectivity
between the PAG and and nodes of the default mode network (DMN) (particularly the medial
prefrontal cortex (mPFC)). As attention to pain increases, so does activity in the salience
network, while activity in the DMN is suppressed. By contrast, when individuals are able to
divert their attention away from painful stimuli, activity in the DMN is higher than activity in the
salience network (Kucyi et al., 2013; Kucyi & Davis, 2015).

12
Figure 2.3 The dynamic pain connectome model. Taken from Kuyci & Davis, 2015. This model proposes the
involvement of three neural networks in the attention to pain. High intrinsic attention to pain (IAP) individuals have
higher structural connectivity between PAG and DMN.Review of the literature surrounding the salience
network suggests it as a potential mediator between bottom-up and top-down signals (Menon &
Uddin, 2010). The anterior insula (AI) in particular has been shown to be involved in mediating
attention, detecting salient stimuli, initiating control signals, and focusing attention on external
stimuli. Together, the AI and anterior cingulate cortex (ACC) integrate bottom–up attention
switching with top–down control and biasing of sensory input (Menon & Uddin, 2010). Though
not specific to pain, these findings indicate that, generally, high levels of activity in the AI and
ACC may indicate high levels of inhibitory control over ascending sensory signals (including
pain). In particular, the insula has been found to exert inhibitory control over the parabrachial
nuclei (PBN). PBN send signals to the amygdala and have been implicated in some of the
emotional and aversive aspects of pain (Sun et al., 2020; Wang & Xu, 2021). A study in rats
found that increased glutamatergic activity in the parabrachial nuclei resulted in neuropathic-like

13
pain phenotypes (generalized hypersensitivity) (Sun et al., 2020). Thalamic inhibitory
neurotransmitter content is also significantly reduced in patients with neuropathic pain following
trigeminal nerve damage, suggesting a similar possible relationship between activity in the
thalamus and the level of inhibitory control over pain signals (Henderson et al., 2013). Multiple
nodes of the salience network (ACC, AI, and thalamus) play an important role in modulating
attention to pain signals and there is significant evidence that attention to pain may underlie or
result from chronic pain states (Kucyi & Davis, 2015).
Alterations in network connectivity of the dynamic pain connectome model have been
identified with functional magnetic resonance imaging (fMRI) in chronic pain patients and are
associated with certain clinical characteristics such as widespread pain, depression, pain
sensitivity, and pain catastrophizing (Ellingsen et al., 2021). The salience network connectivity
to the DMN has been directly implicated in the degree of pain spread in fibromyalgia patients, as
well as the relationship between widespread pain and other amplifying factors such as pain
catastrophizing (Ellingsen et al., 2021). Increased connectivity in the salience network has also
been associated with increased pain sensitivity (but not pain duration) in chronic widespread
pain patients (van Ettinger-Veenstra et al., 2019).
In this series of analyses, we will focus specifically on changes in brain activity that may
be associated with centralized pain, as these are more readily captured than spinal cord
mechanisms and have a better-established relationship with our second metric of interest,
clinical questionnaires.
2.3. Clinical questionnaires and centralized pain
While there are several measures from clinical questionnaires thought to indicate
centralized versus peripheral pain types, there is no true gold standard. Literature has shown
that chronic pain conditions thought to have a stronger, centralized component tend to have
different symptom profiles from those with stronger peripheral components, (Alter et al., 2021;

14
Bergbom et al., 2011; Edwards et al., 2016; Finnern et al., 2021; Kutch et al., 2017) even within
diagnostic categories. The clinical questionnaires of interest in this project are those that assess
pain unpleasantness/sensory quality (painDETECT), affect (depression, anxiety), and pain
widespreadness (body map scores). See Figure 1.1. Each of these markers has been
previously associated with centralized pain (Williams, 2018).
Patients with centralized pain often exhibit higher levels of depression and anxiety as
compared to other pain patients (Borchers & Gershwin, 2015; van Ettinger-Veenstra et al.,
2020). Pain that radiates or spreads far beyond a localized area is a hallmark of centralized pain
conditions and often associated with other central amplifying factors such as pain
catastrophizing, symptoms that are included in the painDETECT and MPQ (allodynia,
hyperalgesia, burning, numbness, tingling) (Dydyk & Givler, 2022; Ji et al., 2018), and increased
depression scores (Hah et al., 2022). Pelvic pain patients with widespread pain are more likely
to have overlapping conditions such as fibromyalgia as well as higher depression scores (Kutch
et al., 2017). While originally developed and validated for use in low back pain patients
(Freynhagen et al., 2006), painDETECT has since been applied to determine the likelihood of
neuropathic pain in musculoskeletal conditions (Berthelot et al., 2019; Liu et al., 2017) as well
as in conditions with unclear or mixed pain origins such as neuromyelitis optica spectrum
disorder, multiple sclerosis, (Bosma et al., 2018; Hyun et al., 2020) burning mouth syndrome,
(Lopez-Jornet et al., 2017) and ankylosing spondylitis (Kisler et al., 2020). A recent study of
multiple sclerosis patients found that centralized and neuropathic pain types, when measured
with the fibromyalgia index and painDETECT, respectively, often co-occur (Kratz et al., 2021),
suggesting a unified, centrally-driven pain type. The MPQ captures both the sensory qualities of
the painDETECT as well as the affective component contained within anxiety and depression
scores (Melzack, 1975).

15
A recent study found that painDETECT, Hospital Anxiety and Depression Scale (HADS),
and gender all had some power to predict response of chronic low back pain to a centrally
acting analgesic, tapentadol (Reimer et al., 2017). Similarly, a study of multiple sclerosis using
painDETECT and the American College of Rheumatology (ACR) 2011 Fibromyalgia (FM)
Survey Criteria (a questionnaire thought to measure centralized pain characteristics) found that
patients with centrally maintained pain experienced greater pain relief from cannabinoids while
those with nociceptive pain experienced greater relief from the use of NSAIDs (Kratz et al.,
2021). These findings indicate there is clinically relevant information in the combination of
clinical measures, and that painDETECT, anxiety/depression, and gender specifically may be
important for categorizing patients that respond to centrally directed treatments.
The salience network connectivity to the DMN has been directly implicated in the degree
of pain spread in fibromyalgia patients, as well as the relationship between widespread pain and
other amplifying factors such as pain catastrophizing (Ellingsen et al., 2021). Increased
connectivity in the salience network has also been associated with increased pain sensitivity
(but not pain duration) in chronic widespread pain patients (van Ettinger-Veenstra et al., 2019).
This lines up with findings in healthy adults injected with nerve growth factor to produce a
progressive muscle pain model, where reduced PAF was associated with greater average pain
experienced (Furman et al., 2018, 2019). Another study in chronic pelvic pain and fibromyalgia
showed an association between widespread pain and altered connectivity between the salience
network and motor cortex.
The relationship between reduced PAF and depression scores is not as well studied and
slightly less consistent than other measures. While one study looking at PAF and depression
scores found no relationship (Jiang et al., 2016), another that separated participants by sex
found that the relationship between PAF and depression scores was positive in males and
negative in females (Tement et al., 2016). This raises the possibility that the paper finding no

16
relationship may have washed out any association by combining two groups (male and female)
with inverse relationships between PAF and depression in a single analysis. This highlights the
need for us to include gender in our model of the relationship between PAF and the identified
centrally amplifying factors.
Taken together, the literature suggests a certain symptomatology associated with
centralized pain, but it is unclear whether these features are useful markers for determining the
level of brain involvement in maintaining the chronic pain state. We will address this issue in
Chapter 4.
2.4. Measures of neural activity
2.4.1. Electroencephalography and peak alpha frequency
EEG is a technique that measures brain activity through recordings of electrical
potentials at the scalp produced by the pyramidal neurons in the cortex. Therefore, it cannot
directly capture activity from deep brain structures. Because it estimates neural activity through
electrical potentials, it has excellent temporal resolution and, compared to a modality like fMRI,
EEG is relatively inexpensive, portable and can be used (to some extent) in naturalistic settings
(Louis et al., 2016). For the study of pain, it is particularly beneficial because recordings can be
taken while the patient is positioned comfortably, minimizing the risk of exacerbating pain
symptoms or triggering claustrophobia when restricted to lying supine in an MRI machine
(Murphy & Brunberg, 1997). Although patients do report some amount of discomfort from the
cap, it is relatively unobtrusive. EEG has great clinical potential, as already evidenced by its use
in diagnosing sleep disorders and types of epilepsy (Tatum et al., 2021). While EEG is very
susceptible to noise (artifacts can easily be introduced from electrical, ocular, muscular,
movement of equipment, respiration, etc), the most common of these can be remedied with the
right data cleaning/preprocessing (Winkler et al., 2011).

17
There is recent interest in resting state peak alpha frequency (rs-PAF), an EEG
measure, as a neural marker of pain conditions with central amplifying factors. rs-PAF has been
shown to be highly heritable through the use of twin studies (Posthuma et al., 2001; van
Beijsterveldt & van Baal, 2002), and generally stable over time in healthy adults (Grandy et al.,
2013). These attributes paired with the relative ease of collection for EEG data make PAF an
attractive neural marker candidate. In the study of pain, slow rs-PAF has been associated with
pain in spinal cord injury (Ngernyam et al., 2015; Sato et al., 2017) and sensitivity to heat pain in
healthy individuals, (Nir et al., 2010; Raghuraman et al., 2019) a potential precursor for pain
amplification. Recent literature has further suggested that individual PAF may be a stable
biomarker of who is susceptible to developing a chronic pain condition, with lower rs-PAF
predicting increased average pain experienced in an induced, progressive muscle pain model
through nerve growth factor injection (Furman et al., 2019). The induced pain findings in healthy
individuals are particularly relevant to centralized pain as pain sensitivity may be a centrally
amplifying factor: a recent meta-analysis found that quantitative sensory testing was predictive
of both pain and disability at a post-intervention follow-up visit across multiple musculoskeletal
conditions and anatomical sites (Georgopoulos et al., 2019). While there is some ambiguity
concerning what rs-PAF indicates about chronic pain (Krupina et al., 2020; Schmidt et al., 2012;
Ta Dinh et al., 2019; van den Broeke et al., 2013), it is thought that this may arise from small
sample sizes and heterogeneous patient groups with inadequate phenotyping (Boord et al.,
2008; Krupina et al., 2020; Schmidt et al., 2012). We will address these prior limitations in
Chapters 3 and 5.
There are some limitations of EEG. One potential drawback is the sheer number of
different ways one can process EEG data, an issue we will explore in Chapter 3. When
comparing across papers for any given EEG marker, careful consideration should be given to
the parameters set for preprocessing (bandpass filtering, notch filtering, rereferencing) as well

18
as post processing (visual versus automated artifact removal, type of transformation if looking at
power spectra, window size, tapers, etc). Additionally, while EEG is well established for use in
diagnosing sleep disorders and epilepsy, its history in the study of chronic pain is less clear cut.
Despite decades of research, in fact, there are very few EEG or fMRI biomarkers
developed/validated to the point of use in clinical practice for pain at all (C.-W. Woo et al.,
2017), and none yet employed in the treatment or diagnosis of UCPPS patients. The previously
mentioned patient-control difference for PAF, for example, was first documented in patients with
pain from spinal cord injury in 2008 (Boord et al., 2008). And yet in 2022, there is still no
consensus as to what this difference reflects about the pain experience or which populations
with pain will exhibit PAF slowing (Zebhauser et al., 2023). Some of the lack of consensus can
be attributed to the wide variability of processing pipelines across labs. Although there has
recently been a push towards standardizing pipelines (Bigdely-Shamlo et al., 2015; Fló et al.,
2022; Gabard-Durnam et al., 2018), most labs process data using parameters specific to their
lab. This work will aim to disentangle whether differences across the literature are more likely
attributable to the aforementioned processing decisions or differences in the etiology of the pain
conditions being studied (Chapters 3 and 5).
A final limitation of EEG as a whole is the inverse problem: for any given EEG signal
there are infinitely many sources or combinations of sources that could have produced that
pattern of activity. This is part of the larger issue of low spatial resolution in EEG. The inverse
problem can be addressed in part by the use of source localization models such as low
resolution electrical activity tomography (LORETA) which statistically estimates the current
sources of a given signal (Pascual-Marqui, 2002). However, the extra time and computing
power required for this type of processing make it less appealing for eventual translation into
clinical settings. While using a global measure such as PAF greatly cuts down on the need for
processing, it is also a nonspecific signal that may be subject to fluctuations from attributes

19
associated with but not causal or indicative of specific pain types (Ta Dinh et al., 2019). This
necessitates understanding of the specific brain changes underlying differences in PAF and
putative markers of centralized pain. A modality that can help in this regard is fMRI.
2.4.2. fMRI, fALFF, and functional connectivity
fMRI is a form of neural recording that indirectly measures brain activity through changes
in blood flow (specifically the ratio of deoxyhemoglobin to oxyhemoglobin). One of the major
strengths of fMRI as a modality is its spatial precision and depth of the signal acquired: unlike
EEG, fMRI can record signal from deep brain structures as well as the more superficial cortex
(Menon & Crottaz-Herbette, 2005; Wirsich et al., 2021). Despite these benefits, fMRI is
extremely costly when compared to the other modalities discussed here. In a study of the
UnitedHealthcare network, researchers found that insured patients typically paid between $112-
$374 for an EEG while paying $241-$875 for an MRI. This is on top of any additional amount
that the UnitedHealth care paid to the providers (Hill et al. 2021). Unlike EEG, it also requires a
dedicated, immobile space and specialized equipment due the powerful magnetic field. This can
be problematic for researchers who aren’t in close proximity to an imaging center, but even
more so for the eventual translation of the measure of centralized pain into the clinic, as it
necessitates that the clinical space has imaging capabilities or a center they can easily refer
patients to. Additionally, while the set-up time is much shorter than an EEG, it requires a
certified technician to operate the machine, and has a long list of contraindications
(pacemakers, implantable neurostimulation systems, drug infusion pumps, non-removable
piercings, etc) (Ghadimi & Sapra, 2023). It can also be an uncomfortable experience for patients
with anxiety, which is often comorbid with chronic pain (Kim et al., 2022), or claustrophobia
(Ghadimi & Sapra, 2023). In these ways, the strengths and weaknesses of EEG and fMRI
balance nicely.

20
In this work, we use fMRI as a research tool to better understand the changes in brain
activity underlying PAF values (Chapter 3) and clinical markers of centralized pain (Chapter 4).
This is a first step towards validating and elucidating the mechanisms behind measures that
would be feasible to capture in a clinical setting. We use a measure of local activity as well as
functional connectivity in this set of analyses. We assessed local activity in discrete brain
regions using the Amplitude of Low Frequency Fluctuations (ALFF). ALFF is a local measure of
spontaneous brain activity, which is computed from the power spectrum of the BOLD signal.
Fractional ALFF (fALFF) accounts for physiological confounds and individual differences by
examining power in the frequency range of interest compared to power in the total frequency
range. Here we focused on slow-5 oscillations (0.01–0.027 Hz) (Mawla et al., 2020) - decreased
relative power in this frequency band has been associated with increases in neural activity
(Kilpatrick et al., 2014; Mawla et al., 2020; Yani et al., 2019). Functional connectivity is a
measure of the statistical relationship between the signal at any two nodes in the brain. In
Chapter 4, we use functions under nilearn.connectome (e.g., ‘ConnectivityMeasure’) to compute
Fisher z-transformed bivariate correlation (Pearson's r) matrices, giving us a single connectivity
value ‘f’ for every pair of nodes in the scan.

21
Chapter 3. Analytic stability and neural correlates of peak alpha
frequency
Published as:
McLain NJ, Yani MS, Kutch JJ. Analytic consistency and neural correlates of peak alpha frequency in the study of
pain. J. Neurosci. Methods. 2022; 368(): 109460.
Abstract Several studies have found evidence of reduced resting-state peak alpha frequency
(PAF) in populations with pain. However, the stability of PAF from different analytic pipelines
used to study pain has not been determined and underlying neural correlates of PAF have not
been validated in humans. For the first time we compare analytic pipelines and the relationship
of PAF to activity in the whole brain and thalamus, a hypothesized generator of PAF. We
collected resting-state functional magnetic resonance imaging (rs-fMRI) data and subsequently
64 channel resting-state electroencephalographic (EEG) from 47 healthy men, controls from an
ongoing study of chronic prostatitis (a pain condition affecting men). We identified important
variations in EEG processing for PAF from a review of 17 papers investigating the relationship
between pain and PAF. We tested three progressively complex pre-processing pipelines and
varied four postprocessing variables (epoch length, alpha band, calculation method, and regionof-interest [ROI]) that were inconsistent across the literature. We found a single principal
component, well-represented by the average PAF across all electrodes (grand-average PAF),
explained > 95% of the variance across participants. We also found the grand-average PAF
was highly correlated among the pre-processing pipelines and primarily impacted by calculation
method and ROI. Across methods, interindividual differences in PAF were correlated with rsfMRI-estimated activity in the thalamus, insula, cingulate, and sensory cortices. These results
suggest PAF is a relatively stable marker with respect to common pre and post-processing
methods used in pain research and reflects interindividual differences in thalamic and salience
network function.

22
3.1. INTRODUCTION
Chronic pain is very impactful. A challenge of studying chronic pain is that it is an
inherently subjective experience. Therefore, identifying objective markers is a pressing need.
Markers of particular interest would be those that could measure the predisposition to
developing chronic pain. Resting-state peak alpha frequency (PAF) measured from
electroencephalography (EEG) has been proposed as such a marker (Furman et al., 2019).
Previous work has also shown that resting-state PAF is reduced in populations with neuropathic
pain in persistent abdominal pain as a result of chronic pancreatitis (de Vries et al., 2013),
neuropathic pain as a result of spinal cord injury (Boord et al., 2007; Sato et al., 2017), and
increased subjective perception of tonic heat pain (Nir et al., 2010; Raghuraman et al., 2019).
Interindividual differences in PAF may relate to awareness and the sampling of sensory inputs
(Angelakis et al., 2004; Mierau et al., 2017), and thus may reflect an individual’s response to
acute pain in a way that may influence the transition to chronic pain.
However, the existing literature is not conclusive in establishing a relationship between
PAF and pain: studies on chronic back pain (Schmidt et al. 2012), central neuropathic pain in
multiple sclerosis patients (Krupina et al. 2020), and persistent pain after breast cancer
treatment (van den Broeke et al. 2013) did not find a relationship between slowed PAF and
pain. Together the positive and negative findings of the existing literature are often interpreted
as suggesting either a differential relationship between PAF and pain of different origins, or as
lack of a strong relationship between PAF and pain altogether. Within the pain field there is also
substantial variation in how PAF is computed and interpreted, which complicates the
comparison of results across studies.
In order to interpret and compare results across the field in a meaningful way, we must
first be certain that differences in findings are not simply a byproduct of differences in
processing pipelines. While previously published papers have focused on developing a method
that best estimates the “true” PAF value (Corcoran et al., 2018), the goal of this paper is instead

23
to analyze previously employed methods and determine whether differences in processing
decisions may impact the final PAF calculation. We reviewed 17 papers with complete available
methods investigating the relationship between PAF and pain and found that there were major
differences in data filtering, whether data were re-referenced, what kind of artifact removal was
performed, epoching, alpha band bounds, and formula for calculating PAF. Furthermore,
comparisons between PAF and other measures of neural activity are infrequent, limiting the
interpretation of PAF in terms of underlying cortical and subcortical activity patterns. The
thalamus is a hypothesized generator of the alpha rhythm and has been implicated in alpha
rhythm alterations, but the association between thalamic activity and PAF in humans has not
been thoroughly explored.
Given these uncertainties, in this study we aimed to assess the robustness of PAF to
three different pre-processing pipelines as well as four processing variables (seven epoch
lengths, two formulas for calculating PAF, three different alpha bands, and five regions-ofinterest [ROIs]: grand-average, frontal, parietal, occipital, sensorimotor) representative of what
is currently being used in the pain literature. Additionally, we aimed to associate PAF with
fractional amplitude of low-frequency fluctuations (fALFF), a resting-state functional magnetic
resonance imaging (rs-fMRI)-derived measure of local activity. To accomplish these aims, we
analyzed data from 47 healthy men, a subset of an ongoing study of urologic chronic pelvic
pain. The dataset included 64-channel resting-state electroencephalographic (EEG) and restingstate functional magnetic resonance imaging (rs-fMRI) from all participants. We first ran three of
the most representative EEG pre-processing pipelines and then varied post-processing
parameters (epoch length, PAF formula, alpha band bounds, and ROI) as determined by our
review of the PAF and pain literature. We then determined whether there was a relationship
between the ROI averages and whole brain as well as thalamic activity in healthy men.

24
3.2. METHODS
3.2.1. Participants
Forty-seven healthy men who were the control group with complete EEG and rs-fMRI data for
an ongoing study of chronic prostatitis were entered into this study. Participants were included if
they were older than 18 years of age, able to participate in the informed consent process, safe
to be scanned by magnetic resonance imaging, had no diagnosis of chronic prostatitis, had no
active urinary, anal, or genital infection, and no severe, urgent, or debilitating medical condition.
All aspects of the study conformed to the principles described in the Declaration of Helsinki and
were approved by our Institutional Review Board. All participants provided informed consent.
The final group of participants was 35.10 ± 13.30 years old (mean ± SD, range, 22.89-63.56
years).
3.2.2. EEG collection and analysis
3.2.2.1. Collection
Continuous EEG data was collected using a 64-channel, ANT Neuro gel-based
electrode cap with sintered Ag/AgCl electrodes. The online reference was placed at the right
mastoid. Signal was acquired with eego sports acquisition software (v1.2.1) from an Ant Neuro
eego Sports amplifier (product number ee-202) at a sampling rate of 2048 Hz. Impedances for
all electrodes were kept below 15 kΩ.
Participants were lying supine and told to keep their head as still as possible, relax, and
not go to sleep. Participants were then told to follow automated, alternating voice commands to
open or close their eyes. The continuous recording was annotated at the beginning of each
eyes-open/eyes-closed epoch: in total, there were ten minutes of continuous EEG data with
non-overlapping, one-minute epochs marked for five eyes-open and five eyes-closed periods.
All offline data processing was performed with EEGLAB v13.6.5.b and MATLAB
(R2018b) scripts. Parameters for the EEG analysis were based on a review of 17 papers
investigating the relationship between chronic pain and PAF (Bjørk et al., 2009; Boord et al.,

25
2007; Corlier et al., 2021; Furman et al., 2020, 2019, 2018; Krupina et al., 2019; Ngernyam et
al., 2015; Nir et al., 2010; Raghuraman et al., 2019; Sato et al., 2017; Schmidt et al., 2012;
Simis et al., 2021; van den Broeke et al., 2013; de Vries et al., 2013; Vuckovic et al., 2018,
2014). Papers were selected from an initial search in Pubmed with the following search terms:
EEG AND (alpha frequency OR peak frequency) AND pain AND (human OR patient) NOT
review. In addition, a filter for only papers available in English was applied. This produced 214
titles and abstracts that were screened. Animal research and research from non-applicable
fields such as sleep and anesthesiology were excluded at this step, as were protocol-only
publications. We also excluded magnetoencephalography (MEG) papers at this step. While the
collection setup and final data set in EEG and MEG are highly similar, the underlying
physiological process recorded is quite different. The two modalities are subject to slightly
different sets of noise/processing issues and we elected to focus exclusively on EEG to ensure
results were not confounded by including processing parameters for two different modalities
(Cohen and Cuffin, 1983; Cuffin and Cohen, 1979; Molins et al., 2008; Muthuraman et al.,
2015). 134 articles remained after this step. Full-text articles were then assessed for eligibility
and further exclusions were made for no PAF calculation (n=87), no pain score and/or pain
population (n=10), no resting-state PAF (n=12), never relating pain to resting-state PAF (n=7),
and no full version available (n=2). At this step, a total of 118 articles were excluded, and one
additional article was identified and added through review of literature references in the
screened papers. The remaining 17 articles are the ones discussed in this paper and
summarized in Table 3.1 and Table 3.2. A flow diagram with the complete information on the
conducted search and selection process can be found in Figure 3.1. As indicated in Table 3.2,
this analysis includes papers of varying clinical and healthy populations, as well as papers that
did and did not find a relationship between PAF and pain.
Analysis of EEG data processing parameters was conducted in two steps: preprocessing (time-series de-noising) and post-processing (data reduction to summary

26
measures). First, three representative pipelines of pre-processing steps/variables were
compared to determine the initial impact on PAF calculation. Second, post-processing variables
(epoch length, PAF formula, and alpha band bounds) were varied for the cleaned data and
compared in a sensitivity analysis to determine their impact on PAF calculation. If the preprocessing pipelines from the first step were found to have a significant effect on PAF, the
second step would be performed individually for each of the three pre-processing pipelines.
3.2.2.2. Pre-processing comparison
For all three pipelines, signal for all eyes open/closed epochs was bandpass filtered from
1-100 Hz. These low and high pass cutoffs were determined by taking the mode of the low and
high pass cutoffs in the 17 articles reviewed. During collection, each non-overlapping, oneminute epoch was marked as eyes-open or eyes-closed by the experimenter. During
processing, recorded data was inspected to verify eyes-open/eyes-closed epochs were in the
correct order/marked correctly. Incorrectly marked epochs were removed from the data set
before further analysis. The first and last 10% of each 60-second epoch was removed, leaving
five 48-second, eyes-closed epochs for further analysis. Each epoch was then processed
through three different, progressively complex pre-processing pipelines:
1. Notch: A notch filter was applied from 58-62 Hz to attenuate electrical noise in
the bandpassed EEG signal. A version of the data preprocessed up to this step
was stored for later analysis.
2. ReRef+Notch: The notch pre-processed data was further pre-processed by rereferencing the EEG data to the common average of each participant’s
electrodes. A version of the data preprocessed up to this step was stored for later
analysis. Of the 17 papers reviewed, only seven re-referenced their EEG data
(Bjørk et al., 2009; Corlier et al., 2019; Furman et al., 2019; Schmidt et al., 2012;
de Vries et al., 2013; Vuckovic et al., 2018, 2014).

27
3. MARA+ReRef+Notch: Notch/bandpass filtered data was further pre-processed
by ICA analysis to prepare data for automated artifact removal via the Multiple
Artifact Rejection Algorithm (MARA): Artifact rejection was performed in an
automated manner using MARA within EEGLAB (Winkler et al., 2014, 2011).
MARA is an ICA-based, optimized linear artifact classifier trained on data that
has been visually inspected and scored manually for artifact rejection. It reliably
detects a range of biological and non-biological artifacts, including eye
movement, muscle, and electrical artifacts. We performed artifact rejection in an
automated manner using the default setting in MARA to reject any component
with artifact probabilities greater than 0.5 (Gabard-Durnam et al., 2018; Winkler
et al., 2014, 2011). After artifact removal, each channel's signal was rereferenced to the common average of all channels for each participant.
Data from all three pipelines was then processed in the same way to calculate PAF, detailed
below.
Bounds for the alpha band vary across sources: initially an alpha band of 7.5-13 Hz was
determined to best cover the distribution represented in Table 3.1. Papers examining the thetaalpha range were not included in the initial alpha band determination (Ngernyam et al., 2015;
Schmidt et al., 2012). Follow-up sensitivity analyses were performed on an additional two alpha
bands (further discussed in the post-processing section below). Only eyes-closed data was
used for the calculation of PAF, in keeping with the majority of the 17 papers reviewed and
previous work investigating the predictive value of individual alpha frequency (Corlier et al.,
2019): only three papers reported results for resting-state PAF from eyes-open data (Boord et
al., 2007; Vuckovic et al., 2018, 2014). The power spectral density (PSD) with normalized units
of the EEG frequencies for each eyes-closed epoch were computed in MATLAB. Epochs with
peak-to-peak amplitude exceeding 80mV were completely excluded from further analysis, in line

28
with the more conservative (Sato et al., 2017) of the two articles that reported using amplitude
cutoffs as part of EEG pre-processing (Sato et al., 2017; de Vries et al., 2013). Initially, center of
gravity (COG), as reported in Furman et al., 2019, was used to determine the peak alpha
frequency (7.5-13 Hz) for each epoch at each electrode. In brief, COG takes the weighted sum
of the alpha spectrum divided by the total power, resulting in the “center” of the spectral power
between 7.5-13 Hz. The following equation was used:
fi is the ith frequency bin including and above 7.5 Hz, n is the number of frequency bins between
7.5-13 Hz, and ai is the spectral amplitude for fi.
Six of the 17 papers reviewed used COG to determine PAF (Furman et al., 2020, 2019,
2018; Raghuraman et al., 2019; van den Broeke et al., 2013; de Vries et al., 2013) and three
included measurements for both the dominant peak of spectral density and COG method
(Corlier et al., 2021; Sato et al., 2017; Schmidt et al., 2012). Eight of the 17 papers reviewed
determined PAF by selecting the frequency of the highest power in the alpha band (also known
as “peak picking”) (Bjørk et al., 2009; Boord et al., 2007; Krupina et al., 2019; Ngernyam et al.,
2015; Nir et al., 2010; Simis et al., 2021; Vuckovic et al., 2018, 2014). See comparisons in
Table 3.1. COG is thought to be a more stable measure, particularly in cases where there are
multiple peaks within the band of interest (Brötzner et al., 2014; Klimesch, 1999, 1997; Klimesch
et al., 1993). Analysis of how peak picking and COG measures of PAF differ within individuals,
especially in combination with other varied post-processing parameters, however, has not yet
been conducted. Therefore, we will revisit and vary the calculation method (peak picking versus
COG) in the post-processing section.

29
PAF was then averaged across all remaining epochs at each electrode for spatial
distribution of PAF values (topographic PAF) and further averaged across all electrodes to
compute the grand-average PAF for each participant. Headmaps for topographic PAF were
visualized using the EEGLAB function topoplot.
3.2.2.3. Post-processing comparison
Because the three pre-processing pipelines showed very little variation in grand-average and
topographic PAF, we selected the most comprehensive pipeline (Notch+ReRef+MARA) to serve
as the “synthesized literature pipeline” (SLP), a baseline analysis process we used to compare
four post-processing parameters that varied the most across the reviewed papers: epoch length,
alpha band range, COG versus peak picking methods for determining PAF, and ROI. Epoch
lengths in the reviewed papers ranged from 2 seconds to 30 seconds (Table 3.1), making the
48 seconds in our synthesized literature pipeline epochs much longer than what is commonly
used in the field. Epoch overlap also varied across papers but compared to epoch length, was
far less variable: only four papers used epochs that overlapped, and all four used epochs that
overlapped by 50% (Sato et al., 2017; Schmidt et al., 2012; Simis et al., 2021; Vuckovic et al.,
2014). Two papers did not report if their epochs were overlapping (Corlier et al., 2021; Vuckovic
et al., 2018). All remaining papers used non-overlapping epochs, as reported in Table 3.1.
We performed a sensitivity analysis representative of the epoch lengths used in the
reviewed papers (Table 3.1), ranging up to the length used in our own calculation. The final set
of values used was 2s, 4s, 5s, 10s, 30s, 45s, and 48s (SLP). We varied these epoch lengths
individually for both the COG and peak picking methods of PAF calculation, keeping the alpha
band at the synthesized literature pipeline range (7.5-13 Hz). Epochs with peak-to-peak
amplitude exceeding 80mV were excluded from further analysis. Therefore, varying the epoch
length also affected the amount of data removed by the 80mV cutoff. Grand-average PAF was
then calculated using both peak picking and COG for each epoch at each electrode before
being averaged at each electrode and then across all electrodes. The range of PAF values

30
across the epoch lengths was calculated for each participant (separately for peak picking and
COG) as well as the range for each participant when comparing peak picking and COG for the
synthesized literature pipeline values (epoch length=48 sec). If either variable at this step (COG
versus peak picking or epoch length) was found to significantly impact the final, grand-average
PAF value, it would be carried forward to the next step: varying alpha band bounds.
Because there was little effect of epoch length, but an apparent effect of COG versus
peak picking for PAF calculation, we used the synthesized literature pipeline with the original
48-second epochs to vary alpha band range and COG versus peak picking for PAF calculation.
Sensitivity analysis for the alpha range was performed on 7.5-13 Hz (SLP), and 8-12 Hz and 9-
11 Hz to reflect a comparison performed in Furman et al., 2019. Epochs with peak-to-peak
amplitude exceeding 80mV were excluded from further analysis. Grand-average PAF was then
calculated using both peak picking and COG for each of the three alpha bands for each epoch
at each electrode before being averaged at each electrode and then across all electrodes. The
range of PAF values within each subject was calculated for each of the three alpha bands.
Alpha band was also found to impact grand-average PAF calculation. The grand-average PAF
values from COG versus peak picking for all three alpha bands were carried forward to be
analyzed with the fMRI data (described in section 2.3).
Because there is variation across publications in the final PAF measure used (grandaverage versus ROI and channel level, see Table 3.1) we selected four representative
electrode ROIs from the papers reviewed. Of the 17 papers reviewed, 11 used ROI analysis.
The four most common and consistent ROIs can be summarized with four electrode
combinations: frontal (F3, Fz, F4, Fp1, Fp2), parietal (CP3, CPz, CP4, P3, Pz, P4), occipital
(O1, Oz, O2), and sensorimotor (C3, Cz, and C4). We calculated the average across each of
these ROIs using the SLP with both COG and peak picking calculation methods to compare
against the grand-average measure. These four average PAF values were carried forward to be
analyzed in a separate fMRI analysis alongside the grand-average.

31
3.2.3. MRI collection and analysis
MRI Acquisition and pre-processing. On the same day but before EEG collection, we collected
resting-state functional magnetic resonance imaging (rs-fMRI) data using a 3 Tesla MRI
scanner (Siemens Magnetom Prisma). To spatially align the rs-fMRI data, we also acquired a
structural image with a T1-weighted magnetization prepared rapid gradient-echo sequence (MPRAGE): repetition time 2300 milliseconds, echo time 2.98 milliseconds, slice thickness 1 mm,
240 slices, 256 × 256 mm voxel matrices, and 1×1×1 mm voxel size. The resting scans were
acquired with the participant resting in a supine position, eyes closed, for 10 minutes in 36-slice
whole-brain volumes with repetition time 2000 milliseconds, echo time 28 milliseconds, slice
thickness 4 mm, 220 × 220 mm voxel matrices, 3.4 × 3.4 × 4.0 mm voxel size, and flip angle 77
degrees. All neuroimaging pre-processing was performed using fMRIprep 1.5.8 as we have
described previously (Mawla et al., 2020).
Assessment of Local Activity. We assessed local activity in discrete brain regions using
the Amplitude of Low Frequency Fluctuations (ALFF). ALFF is a local measure of spontaneous
brain activity, which is computed from the power spectrum of the BOLD signal. Fractional ALFF
(fALFF) accounts for physiological confounds and individual differences by examining power in
the frequency range of interest compared to power in the total frequency range. Here we
focused on slow-5 oscillations (0.01–0.027 Hz) (Mawla et al., 2020) - decreased relative power
in this frequency band has been associated with increases in neural activity (Kilpatrick et al.,
2014; Mawla et al., 2020; Yani et al., 2019). In order to generate slow-5 fALFF images, the
3dRSFC function in the Analysis of Functional NeuroImages (AFNI) software package was used
to compute the fractional power in the slow-5 band while spatially smoothing to 6mm FWHM
and removing the effect of 12 parameters with linear regression (six head motion time series
and six aCompCor time series) (Mawla et al., 2020).
fMRI motion assessment. We assessed head motion using Framewise Displacement
(FD), and planned to exclude any participants with gross head motion defined as average FD

32
exceeding 0.55 mm based on recent recommendation (Parkes et al., 2018; Van Dijk et al.,
2012). Of the 47 participants, none had gross head motion exceeding this limit. As an additional
precaution, average FD from each participant was included as a confounding variable in the
statistical analysis below.
3.2.4. Statistics
We ran a principal components analysis (PCA) on the topographic PAF at each electrode for
every participant using the pca function in MATLAB to determine whether or not PAF is low
dimensional and, if so, whether or not grand-average is a good, low dimensional representation
of the data. We examined the percentage of the total variance explained by each principal
component to assess the independence of topographic PAF across all electrodes. We then
examined the principal components scores, to determine if the component explaining the most
variance aligned with grand-average PAF. We then tested the hypothesis that grand-average
PAF would be consistent across the above-described three pre-processing pipelines by
calculating the Spearman’s ranked correlation coefficient between each of the three possible
pairs of pipelines.
fMRI statistics were carried out using FEAT (FMRI Expert Analysis Tool) Version 6.00,
part of FSL (FMRIB's Software Library, www.fmrib.ox.ac.uk/fsl). We tested the hypothesis that
PAF is associated with thalamic activity by masking slow-5 fALFF data to the entire thalamus as
defined by the Harvard Oxford Atlas in FSL (HarvardOxford-sub-prob-2mm_thalamus) and
performed linear regression analyses to assess the association between the masked slow-5
fALFF images and grand-average PAF values. We additionally performed this analysis on the
unmasked/whole-brain slow-5 fALFF images. If grand-average PAFs from the three preprocessing pipelines tested were significantly different, we planned to run this analysis for each
of the three pipelines. If they were found to be similar, we planned to only use grand-average
PAF from the most comprehensive pipeline (MARA+ReRef+Notch). Likewise, we planned to run

33
the analysis for only the post-processing variables (epoch length, alpha bounds, COG versus
peak picking, ROI) that proved to have a significant impact on the final grand-average PAF
calculation. Therefore, the final analysis included grand-average PAF values from the
MARA+ReRef+Notch pipeline for all three alpha bands examined, using both peak picking and
COG to calculate the PAF. ROI for both peak picking and COG were examined in a second
fMRI analysis. Using OLS (ordinary least squares) simple mixed effects, grand-average PAF
was included as a regressor with average FD from each participant included in the model as a
regressor of no interest. Z-statistic images were thresholded using cluster correction determined
by Z>2.3 and a corrected cluster significance threshold of P=0.05 (Worsley, 2001). If the fMRI
analysis with the grand-average PAF values showed a significant impact of any of the postprocessing variables on identified neural correlates, those post-processing variables would also
be carried forward to the EEG ROI fMRI analysis.
The EEG ROI fMRI analysis was an OLS fMRI analysis similar to the grand-average
analysis described above. Each EEG ROI average PAF was entered as the covariate of the
slow-5 fALFF data to further determine if the EEG ROIs had separate neural correlates.
Because the alpha bands in the first fMRI analysis had increasingly fewer correlates as they
became narrower but otherwise overlapped, while the two calculation methods potentially
highlighted different neural structures, we carried forward only the alpha band from the SLP
(7.5-13 Hz) but both calculation methods (peak picking versus COG) for this analysis.

34
Figure 3.1 PRISMA flow diagram outlining the steps of the literature search for Chapter 3.

35
Not varied in our analysis Varied in Preprocessing Varied in post-processing
Paper
Eye
State Bandpass Notch Re-Reference Noise removal
Epoch
length
Epoch
overlap
Alpha
band
Peak or
COG
GA, ROI,
channel
Boord et al.,
2007 Both 0-500 Hz no No ICA 2s No 8-13 Hz peak
GA,
channel
Bjørk et al., 2009 Closed 0.5–70 Hz 50 Hz CA VI 4s no 8-13 Hz peak channel
Nir et al., 2010 Closed
0.15 -100
Hz 50 Hz no
"artifact rejection program"
(max voltage step 50uV) 1s No
7.5-12
Hz peak channel
Schmidt et al.,
2012 Closed 1-30 Hz no CA GC, VI 4s 2s
4-12
Hz** both ROI, GA
van den Broeke
et al., 2013 Closed 0.1-30 Hz no no GC, VI 4s No 7-13 Hz COG ROI
de Vries et al.,
2013 Closed 1-120 Hz no
Mean of
mastoids
GC, VI, rejected if >200 uV
amp or voltage steps>50 uV 10s No
7.5-13
Hz COG ROI, GA
Vuckovic et al.,
2014 Both 1 Hz
48-52
Hz CA
VI, amplitude ≥100 µV or
present across all electrodes
removed; ICA 2s 1s 8-12 Hz peak
GA,
channel
Ngernyama et
al., 2015 Closed 0.1-70 Hz no no VI, ICA 2s No
4-13
Hz** peak ROI
Sato et al., 2017 Closed .5-70 Hz no no Amplitudes >80 μV rejected 2s 1s 7-14 Hz both ROI
Furman et al,
2018 Closed 5-16 Hz no no gamma band-based ICA 5s No 9-11 Hz COG ROI
Vuckovic et al.,
2018 Both 0.5-60 Hz 50 Hz CA
VI, amplitude ≥100 µV or
present across all electrodes
removed; ICA NS NS 8-13 Hz peak channel
Furman et al.,
2019 Closed 2-100Hz 50 Hz CA PCA, VI 5s No 8-12 Hz COG ROI
Raghuraman et
al., 2019 Closed 0.5-100Hz no
NA; surface
Laplacian ICA 5s No
7.4-12
Hz COG channel

36
Furman et al.,
2020 Closed
0.2-100
Hz no no VI, PCA 5s No
9-11
AND 8-
12 Hz COG ROI
Krupina et al.,
2020 Closed 0.5–30 Hz 50 Hz no
VI, BRAINSYS algorithm (4-
5 SD=artifact) 4s No 8-13 Hz peak ROI
Corlier et al.,
2021 Closed NS no CA
semi-automated FASTER
toolbox, VI 4s NS 7-13 Hz COG ROI
Simis et al., 2021 Closed 1-40Hz no no VI 5s 2.5s
8-12.9
Hz peak ROI
Current paper Closed 1-100 Hz 60 Hz CA
MARA, Amplitudes >80 µV
rejected
2,4,5, 10,
30, 45,48s No
7.5-13
Hz
8-12 Hz
9-11 Hz Both GA
Table 3.1Processing pipeline data from the 17 articles reviewed. General abbreviations: NS=not stated, NA=not applicable; preprocessing abbreviations:
CA=common average, GC=Gratton & Coles, VI= visual inspection; post processing abbreviations: COG=center of gravity, GA=grand-average, ROI=region of
interest. Alpha ranges marked with ** were referred to as “theta-alpha” frequency in the original papers and not included in the initial alpha band determination for
this analysis.

37
Paper
Healthy
(n=)
Clinical
(n=) Population
PAF-pain
Relationship?
If relationship reported, PAF diff
(mean ± SD unless otherwise noted)
Boord et al., 2007 16 16 paraplegia (8 w/and 8 w/out NP) and HC yes
8.93 Hz for HC
8.03 Hz for no pain group
7.64 Hz for pain group (*VI Fig 1)
Bjørk et al., 2009 32 41
migraine patients (33 w/ and 8 w/out aura)
and HC yes
Nir et al., 2010 18 NA healthy yes
Schmidt et al., 2012 37 37 back pain patients and HC; y/n NP no
van den Broeke et al., 2013 11 11
women w/ and w/out persistent pain after
breast cancer treatment no
de Vries et al., 2013 16 16
persistent abdominal pain (chronic
pancreatitis), HC yes
9.5 ± 0.5 Hz for pain group
9.9 ± 0.4 Hz for HC
Vuckovic et al., 2014 10 20
SCI w/chronic NP(10), SCI w/out chronic NP
(10), HC yes
9.1 ± .8 Hz for paraplegic w/chronic NP
10.4 ± .9 Hz for paraplegic no pain
10.1 ± .6 Hz for HC
Ngernyama et al., 2015 NA 20 SCI/bilateral NP yes
Sato et al., 2017 10 11 SCI w/pain (11), HC yes
median (IQR)
9.97 Hz (9.41–10.38) for SCI
11.52 Hz (10.56–11.69) for HC
Furman et al, 2018 44 NA healthy yes
Controls: 10.01 Hz
Low pain sensitivity: 9.98 Hz
High pain sensitivity: 9.88 Hz (*VI Fig 3)
Vuckovic et al., 2018 10 31
SCI (11 with NP, 10 without NP, and 10 who
had pain develop within 6 months of EEG
recording), HC yes
9.0 ± 1.4 Hz for SCI NP
8.6 ± 1.0 Hz for SCI eventual pain
9.2 ± 1.0 Hz for SCI w/out NP
10.0 ± 0.6 Hz for HC
Furman et al., 2019 31 NA healthy yes
9.79 ± 0.16 Hz for slow PAF individuals
10.15 ± 0.12 Hz for fast PAF individuals
Raghuraman et al., 2019 31 NA healthy yes
Furman et al., 2020 58 NA healthy yes
Krupina et al., 2020 12 24
MS w/centralized NP (12), MS w/out
centralized NP (12), HC no

38
Corlier et al., 2021 NA 97
MDD patients w/comorbid pain (46) and
w/out (51) yes
Simis et al., 2021 NA 39 SCI yes
8 Hz for pain SCI
10 Hz for no pain SCI (*VI Fig 3)
Current paper 47 NA Healthy NA NA
Table 3.2 Participant data from the 17 papers reviewed. Abbreviations: HC=Healthy Controls, NP=neuropathic pain, SCI=spinal cord injury, MS=multiple sclerosis,
MDD=Major depressive disorder, NA= not applicable. “*VI Fig #”= estimated from figures when means were not reported in the text of the paper.

39
3.3. RESULTS
Figure 3.2 Sixty-four channel EEG data used to examine differences in individuals with high and low global-average
PAF values. Data was run through the most comprehensive pre-processing pipeline, MARA+ReRef+Notch.
Participants were split into three groups based on their grand-average PAF, with distribution shown for the lowest 16
PAF individuals in blue, the highest 16 in orange, and the remaining, middle 15 in black (a). The PAF averaged
across all epochs at each electrode was then further averaged across the lowest and highest 16 participants and
plotted topographically on headmaps (b). Averaged power spectra across all epochs and electrodes for the lowest
and highest 16 participants were then plotted to illustrate the differing peaks (c), which match up with PAF
distributions in (a). An inset shows the data zoomed into just the alpha band, 7.5-13 Hz, with standard error shaded
(c).
All processing variables of interest are reported in Table 3.1, and information about subject
populations (sample size, clinical versus healthy, pathology if applicable) as well as the
relationship between PAF and pain are reported in Table 3.2. Detailed criteria and steps for the
literature review process are reported in Figure 3.1.
The mean number of ICs removed using MARA for all participants was 27.85±9.54
(mean±STD). The total amount of data removed due to the 80mV cutoff for the SLP (alpha 7.5-
13 Hz, epoch length=48s, COG used to calculate PAF) was 1.5%, and no more than half the

40
data was removed for any one subject. Specifically for the purpose of visualizing the distribution
of grand-average and topographic PAF values, we split the MARA+ReRef+Notch data into the
highest 16, lowest 16, and remaining middle 15 grand-average PAF value individuals. We colorcoded the distribution of the grand-average PAF for all participants (Fig 3.2a.), and plotted the
averaged scalp maps (topographic PAF values) and power spectra averaged across all
electrodes for the highest and lowest 16 individuals (Fig 3.2b). The mean grand-average PAF
value was 10.50±0.14 (mean±SD) for the highest 16 participants and 9.67± 0.30 (mean±SD) for
the lowest 16 participants. These distinct peaks are also visible in the averaged power spectrum
data (across all epochs and electrodes) for the highest and lowest 16 participants (Fig 3.2c).
The remaining middle participants had a mean grand-average PAF value of 10.17 ±0.07
(mean±SD). Note that these high, low, and middle groupings were used only for visualization,
and no further statistics were run/no group differences were calculated.
We then split and plotted the grand-average and topographic PAF data in the same way
for each of the three pipelines (Fig 3.3a and 2b.). Note that the top half of the first column of
Figure 3.3 is the same data from Fig 3.2a and 1b. Results for each measure (grand-average
and topographic PAF) appear to be highly similar across all three pre-processing pipelines.
Using the data from all 47 participants, the MARA+ReRef+Notch pipeline produced a mean
grand-average PAF value of 10.11±0.40 Hz (mean±SD) and had a range of 9.24 to 10.79 Hz.
The ReRef+Notch pipeline produced a mean grand-average PAF value for all participants of
10.10±0.38 Hz (mean±SD) and had a range of 9.23 to 10.70 Hz. The Notch pipeline produced a
mean grand-average PAF value for all participants of 10.07±0.33 Hz (mean±SD) and had a
range of 9.31 to 10.61 Hz.

41
Figure 3.3 Comparison of EEG data after being run through each of the three pre-processing pipelines. Participants
were divided based on their grand-average PAF values calculated with the MARA+ReRef+Notch pipeline into the
lowest 16, highest 16, and remaining middle 15. The distributions of PAF values are shown for each group, with the
lowest 16 in blue, the highest 16 in orange, and the middle 15 in black (a). PAF data for the lowest and highest 16
participants was then averaged at each electrode and plotted topographically on headmaps (b). Principal components
analysis was run on all participants (c). It was discovered that over 95% of the variance for the topographic PAF data
from all three pre-processing pipelines was explained by a single principal component, PC1. The weights for PC1
align well with the grand-average PAF for each individual (c). The three possible combinations of pre-processing
pipelines were then compared and found to be highly correlated, with a Spearman’s ranked correlation coefficient of
>0.95 for all three comparisons (d). a-d all indicate a high degree of similarity between the data from each preprocessing pipeline in terms of grand-average PAF, topography, and variance.
Participants were divided based on their grand-average PAF values calculated with the
MARA+ReRef+Notch pipeline into the lowest 16, highest 16, and remaining middle 15. The
distributions of PAF values are shown for each group, with the lowest 16 in blue, the highest 16
in orange, and the middle 15 in black (a). PAF data for the lowest and highest 16 participants
was then averaged at each electrode and plotted topographically on headmaps (b). Principal

42
components analysis was run on all participants (c). It was discovered that over 95% of the
variance for the topographic PAF data from all three pre-processing pipelines was explained by
a single principal component, PC1. The weights for PC1 align well with the grand-average PAF
for each individual (c). The three possible combinations of pre-processing pipelines were then
compared and found to be highly correlated, with a Spearman’s ranked correlation coefficient of
>0.95 for all three comparisons (d). a-d all indicate a high degree of similarity between the data
from each pre-processing pipeline in terms of grand-average PAF, topography, and variance.
The principal components analysis revealed that >95% of the variance for the
topographic PAF values for all three pre-processing pipelines was explained by a single
component (Figure 3.3c). This component aligned well with the grand-average of PAF, as
confirmed by plotting the first principal component score (PC1) against the grand-average PAF
for each participant (Figure 3.3c). Additionally, each participant’s grand-average PAFs for each
of the three pipelines were plotted against one another to ensure that the distribution of
individual scores lined up (Figure 3.3d).

43
Figure 3.4 Comparison of grand-average PAF values calculated with seven different epoch lengths.Cleaned EEG
data from the MARA+ReRef+Notch pipeline was used with epoch lengths representative of what was used in the 17
articles reviewed. The synthesized literature pipeline (SLP) is the grand-average PAF from 48s epoching and
alpha=7.5-13 Hz as done in Figures 1 and 2. Peak picking (a) and COG (b) were used separately across the varied
epoch lengths and SLP to calculate grand-average PAF. Each point along the x-axis is an individual subject, and they
are ascendingly ordered by SLP grand-average PAF value with peak picking in (a) and SLP with COG in (b) and (c).
When epoch length was held constant at 48s (SLP) and calculation method was varied, a substantial difference was
observed in many participants (c). Across the five tested epoch lengths, the mean range of grand-average PAF
values for each participant was 0.28±0.16 Hz when calculated with peak picking (a) and 0.08±0.05 when calculated
with COG (b). When epoch length was held constant at 48 s (SLP), the mean difference of grand-average PAF values
was 0.51±0.40 Hz when calculated using peak picking and COG for each participant (c).

44
All three methods showed strong agreement between participant scores (ρ>0.97 for all three
pre-processing pipeline comparisons: Notch and ReRef+Notch ρ=0.980, Notch and
MARA+ReRef+Notch ρ=0.982, ReRef+Notch and MARA ρ=0.995).
Because of the high degree of similarity between all three pre-processing pipelines, the
sensitivity analysis for epoch length, alpha band bounds, COG versus peak picking, and ROI for
PAF calculation was only performed on preprocessed EEG spectra from the
MARA+ReRef+Notch pipeline.
The total amount of data removed due to the 80mV cutoff was 1.4% for 45s epochs,
1.3% for 30s epochs, 0.58% for 10s epochs, 0.4% for 5s epochs, 0.3% for 4s epochs, and 0.2%
for 2 s epochs. No more than half the data was removed in any one subject, with the max
percentage of data removed for any one subject decreasing with shorter epoch lengths. We
plotted the values of grand-average PAF for all seven epoch lengths (2s, 4s, 5s, 10s, 30s, 45s,
and 48s/synthesized literature pipeline) for both peak picking (Figure 3.4a) and COG PAF
values (Figure 3.4b). Additionally, the grand-average PAF for peak picking versus COG when
keeping the epoch length constant (SLP, epoch length=48s) were plotted (Figure 3.4c).
The mean range of PAF values in a specific subject across epoch lengths was
0.28±0.16 Hz when peak picking was used and 0.08±0.05 when COG was used. The mean
range of values within a subject for peak picking versus COG was 0.51±0.40 Hz. There was
little effect of epoch length, but an apparent effect of COG versus peak picking for grandaverage PAF calculation, being particularly pronounced for participants at either extreme of the
group PAF range (Figure 3.4b). Only peak picking versus COG was carried forward to the
alpha band bound sensitivity analysis.
The comparison of grand-average PAF for the three alpha bands (8-12 Hz, 9-11 Hz, and
7.5-13 Hz/synthesized literature pipeline) was plotted for both peak picking (Figure 3.5 a) and
COG (Fig 3.5b) grand-average PAF. The mean range of grand-average PAF values in a
specific subject across the three alpha bands was 0.38 ±0.42 Hz when peak picking was used

45
and 0.27±0.17 when COG was used (mean±SD). There was a moderate effect of alpha band
bounds on the final grand-average PAF calculation (Figure 3.5). Peak picking versus COG and
the three alpha bands were carried forward to the fMRI analysis.
To obtain a comparison point for the parameters varied above, we reviewed the
difference between healthy populations and pain populations as well as more pain-sensitive
versus less pain-sensitive groups in healthy populations in the 17 papers reviewed. We found
that, when reported, the average difference between slow and fast PAF groups is nearly 1 Hz
(Table 2).
Figure 3.5 Comparison of grand-average PAF calculated with three different alpha band bounds. Cleaned EEG data
from the MARA+ReRef+Notch pipeline was used to calculate grand-average PAF using three alpha bands: 7.5-13 Hz
(synthesized literature pipeline), and 8-12 Hz and 9-11 Hz. The synthesized literature pipeline (SLP) is the grandaverage PAF from 48s epoching and alpha=7.5-13 Hz as done in Figures 1 and 2. Peak picking (a) and COG (b)
were used separately across the varied alpha bands to calculate grand-average PAF. Each point along the x-axis is
an individual subject, and they are ascendingly ordered by SLP grand-average PAF value with peak picking in (a) and
SLP with COG in (b). There is a moderate difference within participants. Across the three alpha bands tested, the
mean range of the grand-average PAF values for each participant was 0.38 ±0.42 Hz when calculated with peak
picking (a) and 0.27±0.17 when calculated with COG (b)

46
The ROI averages and grand-average PAF (referred to collectively as the five ROIs) were highly
correlated within calculation method (Table 3.3): all combinations of averages had correlation
coefficients over 0.89 when calculated with peak picking and over 0.91 when calculated with
COG. As with the other comparisons of peak picking and COG, peak picking values were far
more variable across ROIs within participants. When plotting all five ROI averages across all
participants, there was an apparent difference between sensorimotor and the other four ROIs
for both calculation methods (Figure 3.6). The average range of PAF values in an individual
across the five ROIs was 0.72 ± 0.43 for peak picking and 0.51 ± 0.46 for COG. However, when
the sensorimotor ROI was removed from the analysis, the range dropped to 0.28 ± 0.23 for
peak picking and 0.11 ± 0.06 for COG.
Peak picking
ROI Mean SD 1. 2. 3. 4. 5.
1. Grand avg 9.969 1.002 - - - - -
2. Frontal 9.939 1.045 0.9898 - - - -
3. Parietal 10.030 1.021 0.9913 0.9752 - - -
4. Occipital 10.047 0.987 0.9722 0.9644 0.9507 - -
5. Sensorimotor 10.082 0.337 0.9226 0.9112 0.9215 0.8867 -
COG
ROI Mean SD 1. 2. 3. 4. 5.
1. Grand avg 10.114 0.399 - - - - -
2. Frontal 10.100 0.401 0.9964 - - - -
3. Parietal 10.141 0.430 0.9931 0.9843 - - -
4. Occipital 10.135 0.447 0.9906 0.9902 0.9792 - -
5. Sensorimotor 9.881 0.961 0.9320 0.9240 0.9318 0.9128 -
Table 3.3. Summary statistics and matrix of Pearson’s correlation coefficient for electrode ROIs (four most common
ROIs in the PAF-pain literature reviewed). All correlations were significant at the p<0.001 level.

47
The first slow-5-PAF analysis was run only on the data from the MARA+ReRef+Notch
pipeline, varying the bounds of the alpha band and COG versus peak picking for PAF
calculation. Cluster maxima are reported in Table 3.4. From the masked data, a slow-5 cluster
in the left thalamus was negatively associated with grand-average PAF values from the
MARA+ReRef+Notch pipeline for alpha band=7.5-13 Hz, COG and peak picking, and alpha
band= 8-12 Hz, peak picking only (Z > 2.3; cluster significance: p < 0.05, corrected) (Figure
3.7). Cluster maxima are reported in Table 3.4. From the whole brain data, slow-5 clusters in
the insula, cingulate, and sensory cortices were also negatively associated with grand-average
PAF values for nearly all of the alpha band-PAF calculation combinations. In particular, the
Figure 3.6 Comparison of grand-average PAF calculated from four ROIs. Cleaned EEG data from the
MARA+ReRef+Notch pipeline was used to calculate the average PAF across four representative electrode ROIs.
These values were plotted along with the grand-average PAF values. Processing variables were held at the SLP
(epoch length=48s, alpha band= 7.5-13 Hz), but with separate calculations for both peak picking (a) and COG (b)
across the five measures (four ROIs, one grand-average). Each point along the x-axis is an individual subject, and
they are ascendingly ordered by grand-average PAF value (SLP) with peak picking (a) and COG (b). Summary
statistics and correlation coefficients are reported in Table 4.

48
Figure 3.7 Association of grand-average PAF varied across alpha band and calculation method with fMRI data.
Grand-average PAF values for each participant from the MARA+ReRef+Notch pipeline varied across three alpha
bands (7.5-13 Hz, and 8-12 Hz and 9-11 Hz.) and two calculation methods (peak picking versus COG) were
associated with slow-5 fALFF data using ordinary least squares regression on both whole brain and masked thalamic
data. The area covered by the thalamic mask is indicated in blue. Slow-5 clusters in nodes of the salience network
(insula, cingulate, and sensory cortex) and the left thalamus were negatively associated with grand-average PAF
values (Z > 2.3; cluster significance: p < 0.05, corrected), but shows some variation across the different alpha bands
and calculation methods. Alpha=9-11 Hz paired with COG was the only analysis that produced no neural correlates
with slow-5 fALFF data.
insula (COG) and cingulate (peak picking) for alpha bands 7.5-13 Hz and 8-12 Hz, as well as S2
for all three bands. Alpha band 9-11 Hz paired with the COG calculation was the only set of
grand-average PAF values that produced no neural correlates (Table 3.4, Figure 3.7).
The neural correlates of the five ROIs from the second fMRI analysis heavily overlapped
with one another (Figure 3.8, Table 3.5), with no clear spatial distinction. There was a fairly
consistent negative association between PAF in the five ROIs and slow-5 fALFF in the thalamus
and sensory cortex. There was also a negative association between PAF and slow-5 fALFF in

49
other regions that varied across ROIs. The cingulate stayed specific to the peak picking method
(associated only with the parietal and grand-average ROIs). The insular cluster, however, now
shows up for both COG and peak picking: there is a negative association between PAF and
insula in almost all COG ROIs (besides sensorimotor), but PAF is also negatively associated
with insula in the sensorimotor and occipital peak picking ROIs. Additionally, the peak picking
sensorimotor ROI produced new clusters in the brainstem not found with any of the other ROIs.
Overall, however, the neural correlates for the five averages were highly similar to one another
for both the peak picking and COG calculation methods. The sensorimotor ROI had the least
overlap with the other four ROIs, corresponding to the patterns seen in the average ROI PAF
data (Table 3.3).

50
x (mm) y (mm) z (mm) size (voxels) z-score
7.5-13 Hz, peak picking
Thalamus -6 -16 4 135 4.06
Sensory cortex 44 -8 24 460 3.97
-64 -22 12 274 4.32
Cingulate -6 22 28 255 5.14
7.5-13 Hz, COG
Thalamus -6 -18 4 104 3.86
Insula -40 -12 -12 398 4.26
Sensory cortex -68 -20 16 312 4.3
44 -8 24 302 3.95
Insula 34 0 -10 262 4.26
8-12 Hz, peak picking
Thalamus -6 -16 4 121 3.99
Sensory cortex 44 -8 24 404 3.85
-64 -20 12 271 4.25
Cingulate -6 22 28 251 5.05
8-12 Hz, COG
Sensory cortex -64 -20 12 293 3.99
Insula -40 -12 -12 282 3.94
9-11 Hz, peak picking
Sensory cortex -64 -22 12 207 4.12
Table 3.4 Cluster maxima coordinates for significant clusters from the fMRI-PAF analysis. Includes association with
grand-average PAF values calculated from all three tested alpha bands (7.5-13 Hz, 8-12 Hz, and 9-11 Hz), with both
peak picking and COG. Locations reported as Montreal Neurological Institute (MNI) coordinates, with regions
identified using the Harvard-Oxford Cortical Structural and Juelich Histological Atlases.

51
Figure 3.8 Association of ROI average PAF with fMRI data. Average PAF values for each participant from the
MARA+ReRef+Notch pipeline were calculated for four representative ROIs (frontal, parietal, occipital, sensorimotor)
and the grand-average using two calculation methods (peak picking versus COG). These values were associated
with slow-5 data using ordinary least squares regression on both whole brain and masked thalamic data. The area
covered by the thalamic mask is indicated in blue. Slow-5 clusters in nodes of the salience network (insula, cingulate,
and sensory cortex) and the left thalamus were negatively associated with grand-average PAF values (Z > 2.3;
cluster significance: p < 0.05, corrected) from most of the ROIs. There is some variation across ROIs, but largely
overlapping neural correlates are highlighted

52
x (mm) y (mm) z (mm) size (voxels) z-score
COG
Frontal
Thalamus -6 -18 4 20 3.84
Sensory Cortex -68 -20 16 113 4.24
Insula -40 -12 -12 68 4.23
Parietal
Thalamus -6 -18 4 105 3.8
Insula -40 -12 -12 417 4.31
34 0 -10 214 4.23
Primary auditory cortex 42 -26 8 236 3.79
Sensory Cortex 44 -8 24 205 3.99
-64 -16 22 277 4.3
Occipital
Thalamus -10 -16 4 21 3.88
Sensory Cortex -68 -20 16 120 4.32
Insula -40 -12 -10 84 4.22
Sensorimotor
Thalamus -6 -16 4 31 4.17
Sensory cortex 44 -8 24 336 3.95
-64 -22 12 310 4.4
Peak picking
Frontal
Thalamus -6 -16 4 20 3.78
Sensory Cortex -64 -22 12 70 4.24
Parietal
Sensory Cortex -64 -22 12 87 4.15
Cingulate -6 22 28 63 5.28
Occipital
Thalamus -10 -16 4 21 3.88
Sensory Cortex -68 -20 16 120 4.32
Insula -40 -12 -10 84 4.22
Sensorimotor
Thalamus -6 -18 4 125 3.67
Insula -40 -12 -12 303 4.45
34 -12 10 292 3.88
34 0 -10 265 3.89
Superior parietal lobule 22 -54 64 219 3.41
Brainstem -2 -16 -34 207 4.44
Inferior parietal lobule 62 -32 14 201 3.4
Sensory cortex 44 -8 24 197 4.38
-66 -22 12 282 4.27
Table 3.5 Cluster maxima coordinates for significant clusters from the slow-5 fALFF with EEG ROI analysis. Includes
clusters associated with ROI average PAF values calculated from all four tested regions (frontal, parietal, occipital,
sensorimotor), for both peak picking and COG. Locations reported as Montreal Neurological Institute (MNI)
coordinates, with regions identified using the Harvard-Oxford Cortical Structural and Juelich Histological Atlases.

53
3.4. DISCUSSION
Despite the novelty and promise of PAF, a few key issues complicate the interpretation
of past data and comparison across publications in the pain field. The first issue is a lack of
consistency in the pre-processing pipelines and post-processing variables for EEG data. We
reviewed 17 papers investigating the relationship between PAF and pain and found that there
were major differences in how labs were pre-processing their data, (re-referencing, artifact
removal, etc), epoch length, bounds of the alpha band, COG versus peak picking for PAF
calculation, and ROI versus grand-average summary measures (Table 3.1). Heterogeneity in
the pre-processing pipelines makes it difficult to tease apart whether differences in past findings
are truly attributable to applications in different types of pain populations, or simply an artifact of
differences in data processing. As such, it is necessary to determine how robust PAF is against
different types of processing.
Additionally, there is little accompanying neuroimaging data to support the theoretical
mechanisms behind PAF. While several papers in humans and animals have investigated the
relationship between the alpha rhythm more broadly speaking and potential generators in the
brain through neuroimaging and single-cell recording (Hughes and Crunelli, 2005), there has
been little validation of similar ideas for PAF specifically. One paper identified an association
between cerebral blood flow and PAF in the thalamus, insula, and other cortical and subcortical
structures, but it has yet to be replicated. As such, it is uncertain what physiological changes
underlie altered PAF: proponents of thalamocortical dysrhythmia suggest decreased PAF is a
result of the interaction between the thalamus and cortex, with high threshold bursting in a
subset of thalamocortical neurons slowing as a result of reduced modulatory corticothalamic
feedback (Hughes et al., 2004; Silva et al., 1980). This, however, has yet to be substantiated in
human models (Hughes and Crunelli, 2005) and recent work in human intracranial recordings
(Halgren et al., 2019) and EEG power analysis in source space (Ta Dinh et al., 2019) casts
doubt on the thalamus as the primary alpha rhythm pacemaker.

54
The findings of this study address both primary concerns. Our results indicate that PAF
is indeed robust against the tested EEG data pre-processing pipelines but may be sensitive to
certain post-processing variables (specifically calculation method and ROI). However, across
methods, interindividual differences in PAF were correlated with rs-fMRI-estimated activity in the
thalamus, insula, cingulate, and sensory cortices. The mean grand-average PAF values from
the most comprehensive pipeline in this study (MARA+ReRef+Notch), 10.11±0.34 Hz, as well
as the range, 9.24 to 10.79 Hz, line up well with previous investigations of the stability and
standard values for resting-state PAF in healthy individuals (Furman et al., 2019; Haegens et
al., 2014).
In summary of our analysis, varying pre-processing pipeline had little to no effect on
grand-average PAF values, epoch length had an effect of 0.28 to 0.08 Hz (peak picking and
COG), alpha band had an effect of 0.38 to 0.27 Hz (peak picking and COG), ROIs had an effect
of 0.72 to 0.51 Hz (peak picking and COG) with sensorimotor included and 0.28 to 0.11 Hz
(peak picking and COG) without sensorimotor, and peak picking versus COG had an effect of
0.51 Hz. Reviewing the difference between healthy populations and pain populations as well as
more pain-sensitive versus less pain-sensitive groups in healthy populations, we found that,
when reported, the average difference between slow and fast PAF groups is nearly 1 Hz (Table
3.2). This is substantially larger than the difference between PAF calculations for the factors
varied in this paper, apart from peak picking versus COG, the difference between ROIs when
sensorimotor is included, and, to a lesser extent, alpha band bounds. Together, these findings
provide evidence that grand-average PAF is generally a suitable summary measure for
capturing interindividual differences in PAF, but care should be taken when comparing across
papers that differ on multiple high impact variables (particularly peak picking versus COG and
sensorimotor ROI versus grand average or other examined ROIs).
Given that resting-state PAF is often thought to capture trait rather than state
information, and that our analysis found very little effect of epoch length, longer epochs may be

55
more appropriate to capture a reliable, individual marker. Previous findings indicate test-retest
reliability for PAF generally increases with increased epoch length up to 40-60 seconds
(Gudmundsson et al., 2007; Salinsky et al., 1991). However, groups working with limited data
should consider that with increasing epoch length, more data may be lost depending on the
method of data cleaning. Additionally, it is worth noting that stationarity in our EEG signal is
likely less of a concern than it would be in task-based or otherwise non-resting-state data.
Epoch overlap also varied across papers but compared to epoch length in the 17 papers
reviewed, was far less variable and is likely a smaller concern (Table 3.1). The bounds of the
alpha band used (a concern echoed in Corcoran et al., 2018), the use of peak picking or COG
to calculate the PAF value at each epoch, and some ROIs had a more apparent effect on the
calculation of the grand-average PAF.
In the pain literature, the summary values reported for PAF are not consistent: grandaverage was used in four papers, ROI was used in 11 papers, and channel analysis was used
in six papers (with two using a combination of grand-average and ROI and two using a
combination of grand-average and channel) (Table 3.1). However, the ROIs also used an
average across all the electrodes of interest, papers often only include one or two ROIs, and
two of the ROI papers noted that the PAF group differences were seen across all electrodes.
Our analysis found that over 95% of the variance in the EEG data from all three pre-processing
pipelines was explained by the grand-average PAF of each individual. This lines up well with
previously published data that peak frequency in the alpha and beta band have high interindividual variation compared to intraindividual variation, giving them strong discriminating ability
when classifying individuals based on EEG data (Grosveld et al., 1976). In order to address any
variation between topographic PAF values, in addition to our PCA analysis we analyzed
average PAF values from four representative electrode ROIS: frontal, parietal, occipital, and
sensorimotor. This analysis showed that mean PAF in each of these regions were highly
correlated for both PAF calculation methods (peaking picking and COG). When plotting the

56
averages for the four ROIs and grand-average across participants, however, there did appear to
be a difference between the sensorimotor ROI and other ROIs examined. This is reflected in the
calculation of PAF range across ROIs within each participant: excluding sensorimotor from this
calculation produces a large reduction in range of PAF values.
Our results also indicate a significant relationship (p<0.001) between grand-average
PAF from multiple post-processing variable combinations and interindividual differences in
activity in a cluster in the left thalamus as well as left insula, right cingulate, and bilateral
sensory cortex (coordinates reported in Table 3.4). The findings in Figure 3.7 illustrate that
even with variation in high impact post-processing variables, many of the same neural
correlates are still found, the clusters in the thalamus and sensory cortices being particularly
robust against post-processing variations. The insula, cingulate, and thalamus are also nodes in
the salience network (Hegarty et al., 2020), indicating a positive relationship between salience
network activity and PAF. Our results do, however, indicate that different calculation methods
for PAF (peak picking versus COG) may tap into different structures within the salience network:
the grand-average COG analysis produced clusters in the insula that were not found in the peak
picking analysis, and the grand-average peak picking analysis produced clusters in the
cingulate that were not found in the COG analysis. Previous work using mean regional cerebral
blood flow in healthy adults also found an association between COG PAF and the thalamus and
insula, but not the cingulate (Jann et al., 2010), which aligns well with our findings. The
specificity of these clusters to a particular calculation method, however, is less distinct in the
ROI analysis (discussed below). Further investigation is warranted.
When conducting the fMRI analysis on the PAF averages from the five ROIs (the grandaverage plus four representative ROIs), the results are fairly consistent. There is no apparent
spatial distinction between the neural correlates of the five ROIs. While there are some
differences, the correlates from the ROIs largely overlap, with the clusters in the thalamus and
sensory cortices being the most consistent. The sensorimotor ROI, particularly for the peak

57
picking calculation method, had the most divergent neural correlates. This lines up with the
sensitivity analysis that suggests the mean from the sensorimotor ROI greatly increases the
range of PAF values across all ROIs within participants and is the least correlated with the other
four (Table 3, Figure 3.6).
In the COG analysis, all ROIs besides the sensorimotor were associated with clusters in
the sensory cortex, insula, and the thalamus. The peak picking results, while still overlapping,
were slightly less consistent: the cingulate cluster shows up only for the grand-average and
parietal ROIs, while the occipital and sensorimotor ROIs are associated with insular clusters that
do not appear in the grand-average analysis. The sensorimotor ROI from the peak picking
analysis additionally has some clusters that are not associated with any of the other four ROIs
(most noticeably, one cluster in the brainstem). Despite these differences, there is no clear-cut
indication of divergent neural correlates from ROI analysis alone. Rather, the associated
clusters are highly overlapping across ROIs.
Calculation method in particular seems to have a large impact on the final PAF
calculation that is consistent for all post-processing parameters varied in this paper: COG
measures produce less variability across participants’ grand-average PAF, but also compresses
the range of grand-average PAF values. Given the fMRI findings that additionally suggest a
potential difference in the neural correlates for COG versus peak picking (insula versus
cingulate), care should be taken in future studies to select the most appropriate calculation
method. While COG may be more stable, it is possible that it may also make inter-individual
differences harder to detect.
PAF has great potential as a biomarker for pain sensitivity and/or chronic pain states.
Therefore, it is important to establish whether PAF can be reliably compared across publications
and better understand the neural mechanisms underlying interindividual differences. A
decreased PAF has been associated with neuropathic pain in persistent abdominal pain as a
result of chronic pancreatitis (de Vries et al., 2013), neuropathic pain as a result of spinal cord

58
injury (Boord et al., 2007; Sato et al., 2017), decreases in pain following an anodal tDCS
intervention in spinal cord injury/bilateral neuropathic pain (Ngernyam et al., 2015), and
subjective perception of tonic heat pain (Nir et al., 2010; Raghuraman et al., 2019). Recent
literature has also suggested that individual peak alpha frequency may be a stable biomarker
that indicates who is susceptible to developing chronic pain conditions, with lower pain-free,
PAF predicting increased average pain experienced in an induced, progressive muscle pain
model (Furman et al., 2019). This study in particular makes PAF an attractive marker, as it
suggests PAF may be able to identify those at risk of developing chronic pain conditions before
the onset of pathology. This attribute could be extremely useful in identifying individuals who
need increased attention following events such as surgery and directing brain-based
interventions such as neuromodulation (Corlier et al., 2019).
PAF is additionally attractive as a marker because it has been shown to be highly
heritable through the use of twin studies (Posthuma et al., 2001; van Beijsterveldt and van Baal,
2002) and generally stable over time in healthy adults, displaying high test-retest reliability over
a span of six months and remaining consistent before and after 100 hours of cognitive
intervention/training across those six months (Grandy et al., 2013). Experiments looking at
changes in PAF over the course of several days or weeks show that it is largely stable within
individuals.
However, there is still debate about whether PAF is a marker for pain sensitivity/chronic
pain at all and if it is, whether it might be specific to certain types of pain. Studies on chronic
back pain (Schmidt et al. 2012), central neuropathic pain in multiple sclerosis patients (Krupina
et al. 2020), and persistent pain after breast cancer treatment (van den Broeke et al. 2013) did
not find a relationship between slowed PAF and pain. While it was excluded from the analysis in
this paper because it did not contain detailed methods for calculation of PAF but instead used
peak frequency (defined as being in the 6-14 Hz range), Ta Dinh et al., 2019 has the largest

59
sample size of any peak frequency-pain study discussed in this paper (101 pain patients, 84
healthy controls). This study found no difference in peak frequency between healthy controls
and chronic pain groups. However, it also included a variety of pain conditions with widely
ranging scores for painDETECT, a questionnaire used to assess the likelihood of a neuropathic
pain component (Freynhagen et al., 2006). Much of the discussion around discrepancies in
findings for the PAF-pain relationship relates to the type of pain condition being studied,
particularly neuropathic versus non-neuropathic. The multiple interpretations of Ta Dinh et al.,
2019’s outcomes further support the need to understand whether differences in findings for
PAF-pain relationships stem from true population differences, a lack of a relationship altogether,
or from the processing decisions discussed in this paper.
Our study indicates a negative association between slow-5 band activity in the thalamus
and grand-average PAF. Previous studies have indicated that increases in fALFF across the low
frequency spectrum (0.01–0.1 Hz) have a positive correlation with brain activity measured
through other markers such as glucose metabolism and cerebral blood flow (Aiello et al., 2015;
Wang et al., 2021). However, studies looking at fALFF in the slow-5 band (0.01–0.027 Hz)
specifically indicate that decreased relative power in this frequency band is associated with
increases in neural activity (Kilpatrick et al., 2014; Mawla et al., 2020; Yani et al., 2019).
Interpreting our results based on previous slow-5 data, it potentially indicates a positive
association between grand-average PAF and salience and thalamic activity. This would align
with previous studies that have found a positive relationship between alpha band activity and
thalamic activity (Goldman et al., 2002; Schwab et al., 2015). Additionally, one paper reported
an inverse relationship between alpha power and thalamic glucose metabolism (Lindgren et al.,
1999). The relationship between neural activity and PAF is likely complex and will need further
validation.
While the thalamic mask used in our analysis encompassed both sides of the thalamus
(left and right), our thalamic results were largely lateralized to the left thalamus in both the

60
grand-average and ROI analysis. While this finding is somewhat unusual, the lateralization is
not entirely novel. A previous MEG paper found that, in contrast to our findings, when
comparing neuropathic and non-neuropathic ankylosing spondylitis patients, a significant
increase in normalized alpha power for the NP participants was found only in right thalamus and
right insula after correcting for multiple frequencies (Kisler et al., 2020). There are no obvious
explanations for this lateralization in the literature, therefore more neuroimaging work will be
required to understand these differences.
Though our findings support the relationship between grand-average PAF and thalamic
and salience network activity, there is still uncertainty about the interpretation of PAF and its
neural underpinnings. While PAF is understood to reflect attention and alertness, this
interpretation is largely based on task-based EEG measures and understanding of the alpha
band more generally (Mierau et al., 2017; Saalmann et al., 2012; Stein et al., 2000). It is not yet
clear what resting-state PAF value reflects. In a review of the literature surrounding shifts in PAF
under a host of different testing and intervention conditions, Mierau et al., 2017 suggested that
interindividual differences in PAF may serve as a stable neurophysiological marker, or a brain
“trait” reflecting differences in individual biology, while intraindividual differences in PAF,
especially over shorter timescales, reflect brain “state,” or adaptations in response to different
tasks and conditions (Mierau et al., 2017).
PAF will generally increase as task demands increase, with differing PAF modulation
across electrodes reflecting the cortical networks that are engaged or disengaged with specific
tasks. It follows that PAF may be responsible for or involved in the sampling or processing
frequency of cortical neurons (Mierau et al., 2017). Nir et al., 2010, who collected continuous
EEG during application of a tonic, painful temperature stimulus, found that participants who
experienced more pain had greater increases in PAF in temporal ROIs contralateral to stimulus.
By contrast, those with certain chronic pain conditions or who are more susceptible to
experiencing high levels of pain have lower resting-state PAF than controls. Shifts to lower PAF

61
are also associated with schizophrenia spectrum disorder, obsessive compulsive disorder, and
depressive disorder patient populations (Schulman et al., 2011). Of particular importance, there
is significant evidence to suggest that PAF decreases with increased age (Chiang et al. 2011;
Osaka et al. 1999; Clark et al. 2004). In this paper’s fMRI analysis, we are not associating PAF
with a clinical feature, but instead directly relating two different brain measures. Therefore, while
our age range was large (22.89-63.56 years), we did not include a correction for age. Thus,
while our study supports the theory that PAF is related to thalamic and salience network activity,
further research is needed to understand what differences in PAF indicate about interindividual
differences in brain function.
One potential limitation of our fMRI findings is the z-threshold used. We have used the
z>2.3 threshold in previous work using fALFF to map activity changes related to natural bladder
filling, and derived expected regions based on animal models (Mawla et al., 2020). Threshold is
important to consider and may lead to false-positive results (Eklund et al., 2016), but no zthreshold has been established for fALFF data. Our current findings appear to be plausible
given the association between PAF and pain as well as the association between pain and the
salience network [see next paragraph and Kutch et al., 2017], but the effect of z-threshold
should be carefully considered in future studies.
Review of the literature surrounding the salience network suggests it as a potential
mediator between bottom-up and top-down signals (Menon and Uddin, 2010). The anterior
insula in particular has been shown to be involved in mediating attention, detecting salient
stimuli, initiating control signals, and focusing attention on external stimuli. The thalamus is also
part of the salience network (Hegarty et al., 2020), and is known to be involved in bottom-up
pain processing for both nociceptive and neuropathic pain (Ab Aziz and Ahmad, 2006).
In summary, we provide evidence that PAF is robust against many common differences
in EEG data processing with the second-largest healthy sample of the 17 papers included in the
development of our pipelines and sensitivity analysis. This supports comparing across past

62
studies investigating the relationship between PAF and pain, although special attention should
be given to papers that differ on multiple high-impact variables (alpha bounds, ROI, and COG
versus peak picking). Additionally, we provide further evidence in support of the relationship
between PAF and thalamic activity, as well as a relationship between PAF and the salience
network, which is generally robust across ROIs. We provide support for PAF as a robust neural
activity marker and grand-average PAF as a suitable summary measure for capturing variation
in individuals’ PAF data across the scalp. More information is needed about the nature of the
relationship between PAF and thalamic and salience network activity, as well as the neural
mechanisms reflected in interindividual differences in PAF.

63
Chapter 4. Brain connectivity associated with chronic pain
intensity within individuals: a 3-year longitudinal study of the
MAPP Research Network
Abstract
Fluctuations in pain intensity for individuals with chronic pain conditions are common. These
fluctuations have the potential to help identify neural factors that may process and amplify pain.
We analyzed resting-state functional magnetic resonance imaging (rs-fMRI) and clinical profile
data from 492 human participants (315 female / 177 male) with urologic chronic pelvic pain
syndrome (UCPPS) collected at up to four time points as part of a 3-year longitudinal study
conducted by the Multidisciplinary Approach to the Study of Chronic Pelvic Pain (MAPP)
Research Network. Functional connectivity measures derived from multiple atlases were
analyzed using within-person longitudinal mixed-effect models to determine the relationship
between functional connectivity and pain and how this relationship was modified by different
clinical profiles. We found that the areas of connectivity most likely to vary as a person
experienced more or less pelvic pain were sensorimotor regions, bilateral insula, cingulate, and
ventromedial prefrontal cortex. These areas include nodes in the salience network and
sensorimotor areas related to the painful body region (pelvis). Additionally, we found that of the
five clinical profile markers of interest, only multisite pain significantly modified the relationship
between pain and connectivity at the whole brain level, particularly within the sensorimotor
cortex. Additional analyses of the association between self-reported daily function and pain
flare-up status (reporting ‘yes’ or ‘no’ to currently being in a ‘flare’ state) support pain
fluctuations as an ecologically meaningful metric. These results indicate not only that pain
fluctuations are an important and impactful part of the chronic pain experience, but also that
they are associated with changes in underlying brain connectivity that are related to clinical
characteristics.

64
4.1. Introduction
Pain intensity in patients with chronic pain is often not consistent longitudinally. On the contrary,
pain fluctuations are quite common and occur across many time-scales: days, weeks, and
months (Harris et al., 2005; Tupper et al., 2013; Zakoscielna and Parmelee, 2013). Although
pain variability creates challenging confounds in clinical care and research (Stephens-Shields et
al., 2016), it also presents an opportunity to longitudinally investigate the neural correlates of
pain within individuals. Most previously published literature studying the neural correlates of
pain focused on associations of symptoms with neural data at the group level and often at a
single time point (van der Miesen et al., 2019). This type of analysis has provided important
information about patient/control differences and areas of the brain that may be important for
processing pain (in both normal and chronic pain states). However, these designs lack
sensitivity to individual differences (Davis et al., 2017). Within-subject analysis of brain states
associated with fluctuating pain levels, on the other hand, allows individuals to serve as their
own control, reducing the impact of individual differences on baseline levels of pain severity,
perception, and etiology.
Neuroimaging studies in the last 20 years have made significant progress in
understanding the brain circuitry that encodes pain (Martucci and Mackey 2018). Many studies
have focused on the blood-oxygen-level-dependent (BOLD) response to acutely applied painful
stimuli (Xu et al. 2021; Xu et al. 2020). A large body of research has also considered how
resting-state functional connectivity between BOLD signals differs between healthy individuals
and those with chronic pain (van der Miesen et al., 2019). Recently, models based on restingstate functional connectivity in healthy individuals experiencing sustained, experimentallyinduced pain were able to predict the ecologically valid persistent pain experienced by some
groups of chronic pain patients but not others (Lee et al. 2021). Therefore, several unanswered
questions still limit our understanding of how brain activity in a chronic pain patient experiencing
pain at rest is associated with the networks that encode provoked pain. First, are natural

65
variations in ecologically-valid persistent pain associated with resting-state functional
connectivity patterns in acute pain networks alone, or are different systems involved? Second,
how do the multiple time scales of natural persistent pain (intensity of recently experienced pain
compared to pain in-the-moment) map to resting-state functional connectivity patterns? Third,
how do participant characteristics (e.g. co-morbid pain and mental health) modify the
relationship between resting-state functional connectivity and pain intensity? Finally, how do
these results differ when factoring in longer scans across multiple states? An individual with
chronic pain will likely pass through multiple states in a given day, especially in UCPPS where
pain can be triggered by natural processes such as bladder filling. Does the combined signal
across multiple naturalistic states reflect the same changes identified in distinct states
individually?
Based on the existing literature, we hypothesize that the neural correlates of variations in
ecologically-valid persistent pain will involve areas both within and outside the traditional pain
matrix, vary based on the timescale of the fluctuation, and be influenced by clinical markers of
centralized pain. Direct assessment in humans is sparse, but there is some evidence that the
circuits associated with changes in chronic pain intensity may be dissociable from the neural
correlates of acute pain (Baliki et al. 2006; Shirvalkar et al. 2023). Recalled pain (e.g. average
pain in the past week/s) may be more strongly associated with pain interference than in-themoment pain (Shi et al. 2009; Salovey et al. 1993), providing initial evidence that the time scale
of the fluctuation may reflect different neural processes. Pain interference is also elevated in
patients with hypothesized centralized pain conditions (Suzuki et al. 2021), which have a set of
distinct clinical phenotypes (Bergbom et al., 2011; Edwards et al., 2016; Kutch et al., 2017; Alter
et al., 2021; Finnern et al., 2021). Common clinical features/questionnaires associated with
centralized pain types are anxiety, depression, neuropathic symptoms, pain catastrophizing,
and pain distribution (Borchers and Gershwin, 2015; Kutch et al., 2017; Conti et al., 2020; van
Ettinger-Veenstra et al., 2020; Kratz et al., 2021). Because these features may indicate a

66
heightened level of central nervous system activity in the chronic pain state, it may also be
reflected in the relationship between brain activity and reported pain.
Using a large longitudinal dataset, we first aim to map the neural correlates of withinperson pain intensity fluctuations across a group of individuals with chronic pain for both current
and recent pain and compare this against a published map of regions involved in the perception
of acute, induced pain. We then examine how between-person heterogeneity in clinical features
affects the relationship between brain activity and pain intensity. To accomplish these aims, we
will analyze data from a three-year longitudinal study from the Multidisciplinary Approach to the
Study of Chronic Pelvic Pain (MAPP) Research Network (Clemens et al., 2020). With clinical
questionnaire and fMRI data from 492 subjects across 3 years (four visits), we can, for the first
time, analyze the neural correlates of pain fluctuations within individual pain patients over time
and assess the impact of clinical questionnaires thought to reflect centrally amplifying factors.
4.2. Materials and Methods
4.2.1. EXPERIMENTAL DESIGN
4.2.1.1. Participants
Participant recruitment took place at six sites across the United States as part of the
NIDDK-funded Multidisciplinary Approach to the Study of Chronic Pelvic Pain (MAPP) Research
Network Symptom Patterns Study (SPS), previously described in Clemens et al., 2020.
Recruitment took place at six sites across the US. Clinical questionnaire and imaging data from
492 individuals (315 female / 177 male) with urologic chronic pelvic pain syndrome (UCPPS)
were included in this analysis. Participants came in for an initial baseline visit and longitudinal
follow-ups at 6, 18, and 36 months post-baseline. At each visit, fMRI and clinical questionnaires
were collected. All study protocols were approved by the Institutional Review Boards at the six
study sites. All study protocols were followed according to the Declaration of Helsinki. All
participants provided informed consent for participating in the study.

67
4.2.1.2. Inclusion and Exclusion criteria
Inclusion and exclusion criteria have been previously published (Clemens et al., 2020).
Briefly, UCPPS inclusion criteria were 1) UCPPS symptoms present for a majority of the time
during most recent 3 months; 2) age ≥18 years; and 3) response ≥1 on the bladder/prostate or
pelvic pain/pressure/discomfort scale during past 2 weeks. Exclusion criteria include
symptomatic urethral stricture, neurological disease or disorder affecting the bladder, bladder
fistula, a history of cystitis caused by tuberculosis, radiation therapy or chemotherapy, prior
augmentation cystoplasty or cystectomy, active autoimmune or infectious disorder, history of
pelvic cancer, current major psychiatric disorder, severe cardiac, pulmonary, renal, or hepatic
disease, unilateral orchialgia (without pelvic symptoms), and prior prostate procedures
(transurethral microwave thermotherapy (TUMT), transurethral needle ablation (TUNA), balloon
dilation, prostate cryo-surgery, or laser procedure).
4.2.1.3. MRI Protocol
The SPS MRI acquisition protocol has been described previously (Clemens et al., 2020;
Mawla et al., 2020). Briefly, participants emptied their bladders and then consumed 350cc of
water. After approximately 40 minutes, a resting state fMRI (rs-fMRI) scan called “fuller bladder”
(rs-FB) was performed. Following rs-FB, participants exited the scanner and emptied their
bladders. Returning to the scanner, an “empty bladder” resting state (rs-EB) scan was
performed, followed by T1-weighted structural scan.
4.2.1.4. MRI ACQUISITION
The Neuroimaging Core of the G. Oppenheimer Center for Neurobiology of Stress and
Resilience (CNSR) at UCLA operated as the hub for neuroimaging operations in SPS. Scanning
took place at six collection sites: Northwestern University (NU), Chicago, Illinois; University of
California/University of Southern California, Los Angeles (UCLA/USC); University of Iowa (UI),
Iowa City; University of Michigan (UM), Ann Arbor; University of Washington (UW), Seattle; and
Washington University (WashU), St. Louis, Missouri. The scanning parameters for rs-fMRI and

68
T1 imaging have been described as part of the previous MAPP Epidemiology and Phenotyping
Study (EPS) (Clemens et al., 2020). UCLA-CNSR took several steps throughout the SPS data
acquisition period to ensure multi-site MRI quality as described in Clemens et al, 2020.
Here we briefly describe the nominal rs-fMRI and T1 parameters: rs-fMRI scans were
acquired with a single shot echo planar imaging (EPI) pulse sequences with conventional
rectangular Cartesian sampling. Basic pulse sequence parameters were as follows TR = 2000
ms, TE = 30 ms, Flip angle = 77°, FoV = 220 mm × 220 mm, Resolution = 64 × 64, Phase
encode direction = A > P, Slice thickness = 4.0 mm, Slice gap = 0.5 mm, Slice acquisition =
ascending (not interleaved), Slices per volume = 34–40 to cover entire brain, Phased array
acceleration factor = 2, Bandwidth = maximum to accommodate resolution specifications,
Orientation = axial-oblique parallel to the line between the anterior and posterior commissures,
Number of volumes = 300 (10 min acquisition).
The MP-RAGE pulse sequence was used for high-resolution T1-weighted 3D volume
imaging. The equivalent pulse sequence on a GE scanner was the 3D FSPGR IR. Basic
parameters were as follows: TR = 2300 ms, TE = 2.98 ms, TI = 900 ms, Flip angle = GE: 11°,
Siemens: 9°, Philips: 9°, FoV = 256 mm × 256 mm, Resolution = 256 × 256, Slices per volume =
240 (or maximum available while maintaining all other parameters), Slice thickness = 1 mm,
Inversion = Slice Selective, parallel imaging acceleration factor = 2, Phase encode direction =
left–right and superior-inferior, Orientation = axial-oblique parallel to the line between the
anterior and posterior commissures.
4.2.1.5. RS-FMRI AND T1 PREPROCESSING
rs-fMRI and T1 preprocessing were performed at the Chronic Pain and Fatigue
Research Center, University of Michigan, Ann Arbor, MI. Results included in this manuscript
come from preprocessing performed using fMRIPrep 20.2.0 (Esteban et al., 2019); RRID:SCR
016216), which is based on Nipype 1.5.1 (Gorgolewski et al., 2011; Esteban et al., 2022);
RRID:SCR 002502).

69
Motion assessment
Two criteria were used to exclude high motion runs to ensure the results are not
confounded by motion artifacts: less than 5 mm maximum Framewise Displacement (max-FD)
and less than 0.55 mm average Framewise Displacement. Out of 3108 rs-FB and rs-EB runs
collected from participants, 356 runs were excluded from further analysis (Figure 4.1).
Figure 4.1 Mean (left) and max (right) framewise displacement for all scans. Scans were excluded if the mean
framewise displacement (FD) across the scan exceeded a mean of 0.55 mm (left) or any timepoint within the scan
exceeded a max of 5 mm (right). These movement cutoffs for mean and max are indicated on each graph with a
dashed green line. All scans surpassing the line were removed from further analysis. Out of 3108 total scans, 356
scans were excluded based on these criteria.
Functional data preprocessing
For each of the 2 BOLD runs found per subject (rs-EB and rs-FB scans at each visit), the
following preprocessing was performed. Reference images were co-registered with 9 degreesof-freedom to the T1w reference using boundary-based registration (bbregister, FreeSurfer)
(Greve and Fischl, 2009). Head-motion realignment was performed using mcflirt (FSL 5.0.9)
(Jenkinson et al., 2002). Images were warped to MNI152NLin2009cAsym standard space and
resampled to 2 × 2 × 2 mm voxel dimension to allow for cross-subject comparison. Framewise
Displacement (FD) was calculated using Nipype (Power et al., 2014). Six physiological
regressors were extracted for principal component-based noise correction based on anatomical

70
CSF and WM masks computed in native space (aCompCor) (Behzadi et al., 2007). Following
the fMRIprep minimally preprocessed pipeline, the preproc.nii images were skull-stripped using
a dilated MNI mask (fslmaths). Based on recent recommendations (Lindquist et al., 2019), six
head motion parameters from mcflirt, aforementioned six aCompCor regressors, and high-pass
temporal filtering (0.01 Hz) were done simultaneously using 3dTproject function in AFNI. Finally,
3DBlurToFWHM was used to estimate smoothness of each image followed by iterative
smoothing until the images reached a target smoothness of 6 mm FWHM.
4.2.1.6. Connectivity Matrix Calculation
Connectivity matrices were calculated using three parcellations with differing densities.
The first parcellation was adapted from a large scale metanalysis which identified the
convergent regions activated during the application of multiple modalities of acute pain stimuli in
222 fMRI experiments (Xu et al. 2020). The resulting parcellation included 14 nodes: anterior
cingulate and right ventrolateral prefrontal cortex, as well as bilateral anterior insula, mid insula,
posterior insula, secondary somatosensory cortex, thalamus, operculum, and posterior insula.
This parcellation will be referred to as the pain network going forward. We also included two
whole brain parcelations: (1) Brainnetome (Fan et al., 2016), which consists of 246 nonoverlapping cortical and subcortical nodes, and (2) Schaefer+Brainnetome Subcortex, which
consists of 436 nodes: 400 cortical nodes which have been derived from the Schaefer atlas
(Schaefer et al., 2018) and 36 subcortical nodes that were adopted from the aforementioned
Brainnetome atlas. In the Schaefer+Brainnetome Subcortex atlas, each node is designated to
one of 8 networks: Visual (Vis), somatosensory/motor (SomMot), Dorsal Attention (DorsAttn),
Salience/Ventral Attention (SalVentAttn), Executive Control (Cont), Default Mode (Default), and
Subcortex. In the Brainnetome atlas, each node is designated to one of 7 lobes: frontal, parietal,
insular, limbic, occipital, and subcortical nuclei. Each node also has a corresponding centroid
defined in MNI coordinates used for visualization/approximation of involved structures. The
functional, Schaefer derived atlas is the focus of the analyses described in this paper.

71
Supporting replications of two key analyses were performed in the structural Brainnetome
derived atlas. See section Visualizing Changes in Connectivity.
Post-processed, four-dimensional, rs-FB and rs-EB images were entered into nilearn.
Functions under nilearn.connectome (e.g., ‘ConnectivityMeasure’) was used to compute Fisher
z-transformed bivariate correlation (Pearson's r) matrices of either 14x14 (Xu pain network),
246x246 (Brainnetome), or 436x436 (Schaefer) density matrices. In the resulting density
matrices, each index represents the measure of functional connectivity, f, calculated between a
pair of nodes. The functional connectivity value f will be entered into our models described in the
statistical analysis section, with the 8 networks/7 lobes used to group nodes for
visualization/group analysis purposes in the whole brain models. Much of our whole-brain
analysis will involve the discussion of network-network pairs, meaning the portion of the
connectivity matrix corresponding to the f values/connections between two networks.
4.2.1.7. Questionnaire Data
Participants completed multiple questionnaires at each of their visits as previously
described (Clemens et al., 2020). The two primary pain measures were the pain reported at the
time of the scan, and the average self-reported genitourinary pain over the week prior to visit as
assessed with responses on the Genitourinary Pain Index (GUPI), question number four:
“Which number best describes your [pelvic] AVERAGE pain or discomfort on the days you had
it, over the last week?” Possible scores 0-10, with lower scores indicating less pain (Clemens et
al., 2009). The GUPI contains questions about both pain and urinary symptoms, but past
research has suggested that the pain and urinary components of the GUPI questionnaire should
not be summed across questions as was originally proposed: aggregate scoring obscures
variability in the individual questions and urinary and pain questions capture different aspects of
UCPPS symptomology (Griffith et al., 2015). The other primary pain component of the GUPI,
the pain subscale, is a mixture of binary location values and overall pain intensity at each painful
site, therefore too granular to capture an overall pain state. The portion of the GUPI

72
questionnaire we use here is the most direct measure of the metric of interest: overall pain
intensity. This measure is referred to as “PainRecent.”
The other measures/questionnaires of interest for the models described in this paper
were selected because they are often elevated in patients with centralized pain and are
important components of what is thought to influence the reporting of pain (Borchers and
Gershwin, 2015; Kutch et al., 2017; Conti et al., 2020; van Ettinger-Veenstra et al., 2020; Kratz
et al., 2021). These measures and their score scales (possible scores min-max) are as follows.
Multisite pain was assessed using the collaborative health outcomes information registry
(CHOIR) body map (Scherrer et al., 2021) as previously described in Schrepf et al., 2023.
Patients were asked to select any of 76 body sites where they felt pain and then rate their pain
0-10 at all selected sites. This data was further reduced to 12 nonpelvic regions, and the
multisite pain score was set equal to the number of non-pelvic regions marked with a pain
intensity of 4 or higher: possible scores 0-12. The painDETECT questionnaire was used to
measure the presence of neuropathic-like pain symptoms, with higher scores indicating a higher
likelihood of neuropathic pain (Freynhagen et al., 2006): possible scores 0-35. Anxiety and
depression were both assessed with the Hospital Anxiety and Depression scale (Zigmond and
Snaith, 1983): possible scores for both anxiety and depression scales 0-21, with lower scores
indicating less presence of anxiety or depression. Pain catastrophizing was assessed with the
Catastrophizing subscale of the Coping Strategies Questionnaire (Rosenstiel and Keefe, 1983):
possible scores 0-6, with lower scores indicating less catastrophizing. Additionally, whether the
patient reported being in a flare state, defined as “symptoms much worse than usual,” was
recorded at each visit (yes/no). To assess physical and mental functioning, we used the
physical and mental scales of the 12-Item Short-Form Health Survey (Ware et al., 1996):
possible scores 0-100 for both physical component score (PCS-12) and mental component
score (MCS-12), with lower scores indicating better physical/mental functioning.

73
4.2.2. STATISTICAL ANALYSES
4.2.2.1. Analysis of clinical data
Based on past reports of pain fluctuations in patients with chronic pain, we hypothesized
that our study population would experience fluctuations in their pain ratings across visits but
would not experience net improvement or worsening in their symptoms. To assess this, we
calculated the average maximum change in pain (max-min) and course of change in pain
(current pain – pain at previous visit averaged for each participant) across the four visits for the
entire study population. We calculated separate averages across males and females, and
further subdivided the data into three age groups for visualization purposes.
4.2.2.2. Modeling the relationship between self-reported pain and functional connectivity
We hypothesized that several clinical characteristics previously reported to augment the
experience of pain in chronic pain conditions would influence the relationship between brain
activity and an individual’s self-reported pain. To assess this hypothesis, we needed to look at
the relationship between pain and functional connectivity and whether any clinical variables
significantly modified this relationship. Because of the nature of our scans, this analysis was
performed across multiple combinations of scan types and pain measures. The data was
additionally harmonized across unique scanners using ComBat due to the multisite collectionref-. First, we assessed all combinations of scan types (EB and FB scans) and pain metrics
(PainNow and PainRecent) within the pain network in Models 1-2. For all models, f= the
connectivity value between a pair of nodes in the functional connectivity matrix and pain= either
the GUPI pain measure (PainRecent) or pain at the time of scan (PainNow).
Model 1: To assess the relationship between pain and functional connectivity, we used
a longitudinal mixed effects model with an independent intercept for each subject. The
average framewise displacement (FD_Avg) was included as a covariate of no interest
f ~ pain+ FD_Avg + (1|pid)

74
Model 2: A second longitudinal mixed effects model was used to assess the effect of the
interaction between clinical variables and pain on the relationship between pain and
functional connectivity. Model 2 is equal to Model 1 with the addition of an interaction
term between pain and one of five clinical variables: multisite score, painDETECT,
HADS anxiety, HADS depression, or pain catastrophizing
f ~ pain+ FD_Avg +clinical variable+ pain*clinical variable (1|pid)
Models 1 and 2 were run across all locations in the functional connectivity matrix constructed
from the 14 nodes of the pain network (91 unique pairs). We ran Model 2 in five iterations,
again for every value in the functional connectivity matrix across the available four visits. For
each iteration of Model 2, we included a different clinical variable of interest in the interaction
term (e.g. multisite score*pain, painDETECT*pain, HADS anxiety*pain, HADS depression*pain,
and pain catastrophizing*pain). p-values for the fixed effects of pain in Model 1 and
ClinicalVariable*pain were FDR corrected for multiple comparisons (q<0.05).
The final model run within the pain network was Model 3, which was the same as Model
1, but run across the whole brain Schaefer parcellation instead of within the pain network alone.
P values for the fixed effects of pain were FDR corrected for multiple comparisons (q<0.05).
We additionally ran two models where FB and EB scans were combined: Models 4-5.
These models served two purposes. Firstly, studies suggest that 10 minutes may not be a
sufficiently long resting state (Birn et al. 2013; Laumann et al. 2015) and the acquisition for our
FB and EB scans were 10 minutes each. Secondly, and perhaps more importantly, combining
FB and EB scans allows us to integrate across multiple states to achieve a measure of brain
activity closer to trait than state that would align with the GUPI pain measure used (capturing
real-world pain levels over the past week).
Model 4 was the same as Model 1, with additional covariates added for scan type (EB
or FB), sex, and age. Model 5 was the same as Model 2, with additional covariates added for
scan type (EB or FB), sex, and age. Fixed effects for pain (Model 4) and pain*clinical variable

75
(Model 5) were FDR corrected for multiple comparisons (q<0.05). These analyses were
performed in the Schafer atlas with a supporting replication performed in the Brainnetome atlas.
In summary, the first three models (Models 1-3) kept empty and full brain scans
separate and looked at both PainNow and PainRecent. The last two models (Models 4 and 5)
combined the empty and full brain scans into a single analysis and looked only at PainRecent.
The predictor of interest for Models 1,3, and 4 was pain, and the predictor of interest for
Models 2 and 5 was the interaction between pain and each of the five clinical variables. We will
use these model names going forward.
4.2.3. VISUALIZING CHANGES IN CONNECTIVITY
Stored model information was used to plot both brain maps and connectivity matrix
representations of the node pairs with a significant relationship between functional connectivity
(f) and pain (Models 1, 3, and 4) or ClinicalVariable*pain (Models 2 and 5) after FDR correction
for multiple comparisons.
For the brain maps, a significant node is defined as a location part of at least one node
pair with a significant relationship between f and the variable interest (pain for Models 1,3 and
4; ClinicalVariable*pain for Models 2 and 5). For each significant node, the corresponding MNI
centroid of that node was marked with a sphere color coded by the location (Xu pain network),
network (Schaefer) or lobe (Brainnetome) it belonged to. A single node could be part of several
significant node pairs. To display this effect visually, spheres were scaled up based on the
number of significant pairs, “N”, they belonged to: initial sphere size “S” (shared by all significant
nodes) was multiplied by “N”. Therefore a node that was part of only one significant node pair
would be size “S” while a node that was part of 10 significant node pairs would be size 10xS.
The size of the spheres, however, quickly becomes unwieldy in cases of a high density of
significant connections. Therefore, “S” was adjusted below 1 for some figures to ensure the
results were readable, especially in cases where a high density of significant connections at a

76
node caused spheres to envelope/obscure the entire brain map. By multiplying NxS and varying
the starting value of “S”, it is possible to maintain information about the relative number of
significant connections with a specific node while keeping the figures readable. To further
capture information about which nodes were connected with one another, if any node was a part
of a number of significant pairs surpassing a threshold “P”, pink connecting lines were drawn
between the node and all of its significant node pairs. Similar to scaling factor “S”, “P” was
varied across the figures to ensure information was readable.
Connectivity matrices for Models 4 and 5 were visualized from the same data used for
the brain maps: each node pair with a significant f~pain or f~ClinicalVariable*pain relationship
was color-coded based on the p value for pain in Model 4 and p values for ClinicalVariable*pain
in Model 5. The networks (Schaefer) and lobes (Brainnetome) that each node belonged to were
delineated with grid lines and colored spheres along the axis matching the color coding in the
brain maps. Additionally, we calculated the percent of node pairs with significant f~pain or
f~ClinicalVariable*pain relationships in each “block” or pair of networks/lobes. We then rank
ordered these network pairs highest to lowest for each model based on the percentage of
connections that was significant within that block. For the ten network pairs with the highest
percentage of significant connections, the five node pairs with the lowest p values were pulled
into a separate table: MNI centroids of each node pair were entered into the Harvard Oxford
atlas to determine an approximate label. The result was a table of 50 pairs of MNI coordinates
and their corresponding labels from the 10 networks with the highest percentage of significant
connections.
We then isolated the connections that were identified in both Model 4 and Model 5 and
plotted the brain map and connectivity matrix data as described above: however, the
connectivity matrix color was binarized (red=yes, white=no) instead of coded by p values.
Similarly, we plotted the brain map and connectivity matrix data for the regions unique to
Models 4 and 5.

77
The primary focus of our analyses was the Schaefer atlas; all figures are based on data
from the Schaefer atlas, apart from Figures 4.7 and 4.9, which are the supporting replications of
Figures 4.6 and 4.8 in the Brainnetome atlas.
4.2.4. ASSESSING THE IMPACT OF PAIN ON FUNCTION
Based on a large body of literature, we hypothesized that pain would negatively impact
physical and mental functioning. To assess the impact of pain on physical and mental
functioning, we used the physical and mental functioning scales of the 12-Item Short-Form
Health Survey (SF-12) (Ware et al., 1996). Separately, we regressed pain onto the mental
component score (MCS-12) and the physical component score (PCS-12) from the SF-12. As in
Models 4 and 5, we controlled for sex, age, and collection site and included a random slope
and intercept for each participant in longitudinal mixed effects models. The final model for
mental function was:
MCS-12 ~ sex + age + siteid + pain + (pain|PID)
And for physical functioning:
PCS-12 ~ sex + age + siteid + pain + (pain|PID)
Additionally, we hypothesized that patients reporting a flare would also report higher
levels of pain. To asses this, we looked at the effect of patients reporting a flare on the pain
measure across visits. Participants answered the question, “Are you experiencing a flare now?”
with yes or no. We regressed pain onto this flare status and, as in Models 4 and 5, accounted
for sex, age, and site and included a random slope and intercept for each participant:
pain ~ sex + age + siteid + flare status + (flare status|PID)

78
MAPP-II (N=492) Male (N=177) Female (N=315)
2020 US
Census
(SD) (SD)
Age (years) 47.7 15.2 42.1 15.3
Symptom Duration (years) 10.0 10.1 12.7 12.2
Race
(N) (%) (N) (%) (%)
American Indian 2 1.1% 1 0.3% 1.3%
Asian 1 0.6% 3 1.0% 6.1%
Black 10 5.6% 19 6.0% 13.6%
Native Hawaiian 0 0.0% 1 0.3% 0.3%
White 157 88.7% 276 87.6% 75.8%
Multi Race 4 2.3% 12 3.8% 2.9%
Other 3 1.7% 3 1.0%
Unknown 0 0.0% 0 0.0%
Ethnicity
Hispanic 12 6.8% 20 6.3% 18.9%
Non-Hispanic 165 93.2% 294 93.3% 59.3%
Table 4.1 Demographic information for the 492 SPS subjects included in the analysis. Mean age (±SD) across the
entire study population was 46.0±15.7 years and mean symptom duration across the entire study population (±SD)
was 12.2±11.7 years.
4.3. Results
A total of 492 subjects were entered into this analysis (315 female / 177 male). At baseline,
mean age (mean±SD) across the entire study population was 44.1±15.5 years and mean
symptom duration across the entire study population (mean±SD) was 11.7±11.5 years. Mean
age (mean±SD) was 42.1±14.3 for females and 47.7±15.2 for males. Duration of symptoms for
participants entering the study was, on average (mean±SD) 10±10.1 years for males and
12.7±12.2 years for females. Further demographic data is summarized in Table 4.1.
Symptom Patterns Study (SPS) participants experienced fluctuations in average selfreported pain across the four visits (36-month timeline) with large amounts of variability between
subjects. At the group level, over 50% of both male and female participants experienced a

79
minimum of two points of change in their pain rating (max-min), with this percentage being
slightly higher in females (53.7% versus 62.5%) (Figure 4.2). There was some variation when
the male and female groups were further broken down into the three age groups (Figure 4.2,
top). Patients’ pain changes did not follow (on average) a trend toward improving or worsening
pain ratings: when averaging the pain changes (pain rating at each visit minus the pain rating
from the visit before) for each participant, the distribution across all participants is centered
around 0 (Figure 4.2, bottom).
Figure 4.2 In line with previous results, SPS participants experienced fluctuations in average self-reported pain
across the four visits (36 month timeline). Group level data shows that participants experience fluctuations in their
average self-reported pain across the four visits (36 month timeline). For both sexes, a. over half of participants
experienced max-min>=2 across their four visits (top graphs). We additionally took the mean change over time for
each participant (the average of the difference between each visit and the visit preceding) b. and saw that most
patients scored close to 0, indicating neither an overall worsening nor improvement of symptoms over time (bottom
graphs). This was true for both sexes.

80
MAPP-II (N=492) Scoring Male (N=177) Female (N=315) Total
(SD) (SD) (SD)
Multisite score 0-12 1.1 1.5 2.1 2.6 1.7 2.4
Anxiety 0-21 6.8 4.5 7.5 4.8 7.3 4.7
Depression 0-21 5.6 4.6 5.7 4.5 5.7 4.6
painDETECT 0-35 7.7 6.3 8.9 5.6 8.5 6.1
Pain
Catastrophizing 0-6 1.7 1.3 2.0 1.5
1.9 1.4
Table 4.2 Summary scores across all visits for the 5 clinical measures entered into Models 2 and 5, mean and SD
reported for males, females, and across the whole population (total). Scoring column indicates the range of scores
possible for a given measure. Multisite score was the number of non-pelvic body regions where patients reported
feeling pain greater than 4 on a 0-10 scale. For most measures, apart from depression, scores were slightly higher in
the female portion of the study population.
The average scores for all clinical variables are reported in Table 4.2. The correlation
between clinical variables at baseline was below 0.60 for all pairs and highest for anxiety and
depression at 0.57. The next highest was between anxiety and pain catastrophizing at 0.54. See
Table 4.3 for the correlation matrix between all five variables.
Multisite score Anxiety Depression painDETECT Pain
catastrophizing
Multisite score 1.00
Anxiety 0.25 1.00
Depression 0.29 0.56 1.00
painDETECT 0.39 0.36 0.40 1.00
Pain
catastrophizing
0.15 0.54 0.53 0.28 1.00
Table 4.3 Correlation matrix for the 5 clinical measures at baseline visit. Multisite score was the number of non-pelvic
body regions where patients reported feeling pain greater than 4 on a 0-10 scale. All variable pairs had a correlation
coefficient of <0.6, with the highest correlation between depression and anxiety (0.56).

81
Figure 4.3 Results from Model 1: functional connectivity within the pain network in bladder full and bladder empty
state associated with average pain in the week preceding the visit (PainRecent) and pain at the time of scan
(PainNow). Patients underwent two functional scans, one when their bladder was full (Full), and one after voiding
(Empty). Both pain measures highlight significant connections at most nodes in the pain network, however
PainRecent is more associated with functional connectivity from the empty bladder scan while PainNow is more
associated with functional connectivity in the full bladder scan. The PainNow, full bladder combination is the most
robust of the four analyses.
The results from Model 1 indicate that functional connectivity within the pain network is
significantly associated with fluctuations in genitourinary pain, however PainNow is associated
with signal from the full bladder scan while PainRecent is associated with signal from the Empty
bladder scan. The reverse cases, empty bladder associated with PainNow and full bladder
associated with PainRecent, produce zero and one significant connection, respectively.
Additionally, the association of full bladder signal with the PainNow measure has overall the
most connections and a stronger emphasis of the connection between the vDLPFC and the rest
of the pain network compared to the PainRecent, empty bladder analysis.

82
Figure 4.4 Results from Model 2: the relationship between functional connectivity and pain within the pain network is
differentially modified by clinical variables based on bladder state (empty versus full) and pain metric (PainNow
versus PainRecent). The two pain measures, PainRecent and PainNow, did not share any clinical characteristic that
significantly modified the relationship between functional connectivity and pain. Anxiety was the strongest modifier of
the relationship between empty bladder connectivity and PainRecent and pain distribution/multisite score was the
strongest modifier of the relationship between full bladder connectivity and PainNow.
As seen in Model 1/Figure 4.3, the results from Model 2/Figure 4.4 show that functional
connectivity from Empty bladder produces a stronger association with PainRecent ratings while
functional connectivity from full bladder produces a stronger association with PainNow. The two
pain metrics did not share any clinical feature that significantly modified the relationship
between functional connectivity and pain. In the association of PainRecent with functional
connectivity from the empty bladder scan, anxiety significantly modified the relationship between
functional connectivity and pain at all 14 nodes while depression significantly modified a single
connection between left thalamus and right, anterior insula. There were no significantly modified

83
connections for any clinical variable for PainRecent association with empty bladder scans. In the
association between PainNow and functional connectivity from the full bladder scan, multisite
pain score significantly modified the relationship at all 14 nodes, with a slightly reduced
emphasis on left thalamus. The relationship between PainNow and functional connectivity from
the empty bladder scans was significantly modified at two pairs of nodes: multisite pain score
significantly modified the relationship between left operculum and anterior cingulate, while the
painDETECT score significantly modified the relationship between left posterior insula and right
anterior insula.
Figure 4.5. Results from Model 5: when the relationship between functional connectivity and pain was assessed
across the whole brain (Schaefer atlas), only the full bladder scans produced significant results. Most robust was the
relationship with PainNow, although pain recent also produced a pair of significant connection between left
somatosensory cortex and bilateral cingulate gyrus in the default mode network.

84
The results from Model 4 highlighted several locations where functional connectivity was
significantly associated with the GUPI pain score after FDR correction (q<0.05) (Figure 4.6).
Importantly, these findings were significant while including age, site, sex, scan movement (mean
FD), and scan type (rs-FB and rs-EB) as covariates. Areas of connectivity most likely to vary
with changes in pelvic pain were sensorimotor regions, bilateral insula, posterior and anterior
cingulate, and medial prefrontal cortex (Figure 4.6a, Table 4.4). Findings largely overlapped for
both in both atlases (Figures 4.6 and 4.7). The network pairs with the highest percentage of
significant nodes were within salience/ventral attention (10.41%, 150 significant connections),
within somatosensory/motor (6.65%, 197 significant connections), between
somatosensory/motor and salience/ventral attention, (3.56%, 129 connections) and within the
default mode and visual networks (2.98%, 122 significant connections and 2.35%, 43
connections, respectively) (Table 4.5, Table 4.6, Figure 4.6b).

85
Figure 4.6 Changes in connectivity associated with pain fluctuations within participants across four visits. Areas of
connectivity most likely to vary with changes in pelvic pain were medial sensorimotor regions, bilateral insula,
cingulate, and ventromedial prefrontal cortex after FDR correction (q<0.05). Vis=Visual,
SomMot=somatosensory/motor, SalVentAttn=Salience Ventral Attention, Cont=Executive Control. a) In the brain map
illustration, spheres at each region were scaled by the number of connections with a significant f~pain relationship
(Model 4). Spheres are color coded based on the network they belong to in the Schaefer atlas, displayed on the axes
of b), a connectivity matrix illustrating the nodes with significant f~pain relationships. The connectivity matrix in b) is
color coded by the p value of the relationship between functional connectivity and pain (darker colors indicating
connections with lower p values for pain). Areas of connectivity highlighted are within somatosensory/motor, within
salience/ventral attention, and between somatosensory/motor and salience/ventral attention. MNI coordinates for
significant pairs included in Table 4.4.

86
Figure 4.7 Changes in connectivity associated with pain fluctuations within participants across four visits in the
Brainnetome atlas. Significant connections were seen in several lobes after FDR correction (q<0.05) including limbic,
parietal, insular, and frontal. a) In the brain maps, spheres at each region were scaled by the number of connections
with a significant f~pain relationship (Model 4). Spheres are color coded based on the lobe they belong to in the
Brainnetome atlas, displayed on the axes of b), a connectivity matrix illustrating the nodes with significant f~pain
relationships. The connectivity matrix in b) is color coded by the p value of the relationship between functional
connectivity and pain (darker colors indicating connections with lower p values for pain). Because Schaefer
categorizes by network and Brainnetome categorizes by lobe, a direct comparison of the significant areas can’t be
made, but spatially the distribution of significant connections looks fairly similar (lateral versus medial emphasis

87
MNI centroid Atlas label MNI centroid
X Y Z X Y Z
'SalVentAttn' 40.51 8.74 -2.84 Insular cortex 'SalVentAttn' 10.21 -43.25 53.38 precuneous cortex
-38.74 1.93 -4.62 Insular cortex 40.51 8.74 -2.84 Insular cortex
-43.35 11.93 2.29 Frontal Operculum Cortex 38.38 6.97 10.47 central opercular cortex
40.51 8.74 -2.84 Insular cortex 38.38 6.97 10.47 central opercular cortex
-36.48 4.39 10.69 central opercular cortex 38.38 6.97 10.47 central opercular cortex
'SomMot' -55.94 -21.36 7.70 Planum
Temporale/Heschl's Gyrus
'SomMot' 60.14 -23.69 10.77 Planum Temporale
-48.69 -12.89 14.16 central opercular cortex 40.91 -13.05 18.18 central opercular cortex
40.91 -13.05 18.18 central opercular cortex 11.08 -17.02 41.29 Cingulate Gyrus,
posterior
-48.69 -12.89 14.16 central opercular cortex 49.52 -10.04 12.85 central opercular cortex
-59.55 -2.34 11.27 precentral gyrus 49.52 -10.04 12.85 central opercular cortex
'SomMot' -41.14 -34.95 14.27 Planum Temporale 'SalVentAttn' 38.38 6.97 10.47 central opercular cortex
11.08 -17.02 41.29 Cingulate Gyrus, posterior 37.20 22.63 4.75 Frontal Operculum
cortex
11.08 -17.02 41.29 Cingulate Gyrus, posterior 40.51 8.74 -2.84 Insular cortex
-48.12 -24.15 17.83 Parietal Operculum Cortex 38.38 6.97 10.47 central opercular cortex
60.14 -23.69 10.77 Planum Temporale 38.38 6.97 10.47 central opercular cortex
'Default' 34.98 23.04 -17.23 Frontal orbital cortex 'Default' 4.90 -63.34 31.35 precuneous cortex
-5.33 -60.19 30.06 precuneous cortex 34.98 23.04 -17.23 Frontal orbital cortex
7.42 53.99 12.82 paracingulate gyrus 9.89 -52.61 35.13 precuneous gyrus
34.98 23.04 -17.23 Frontal orbital cortex 9.89 -52.61 35.13 precuneous gyrus
-4.16 -53.35 20.19 Cingulate Gyrus, posterior 5.82 -52.25 22.95 Cingulate Gyrus,
posterior
'Vis' -46.15 -73.07 6.14 lateral occipital cortex,
inferior division
'Vis' 43.78 -78.76 10.42 lateral occipital cortex,
inferior division

88
-46.15 -73.07 6.14 lateral occipital cortex,
inferior division
28.90 -77.75 36.73 lateral occipital cortex,
superior division
-46.15 -73.07 6.14 lateral occipital cortex,
inferior division
-21.24 -78.59 44.69 lateral occipital cortex,
superior division
43.78 -78.76 10.42 lateral occipital cortex,
inferior division
28.90 -77.75 36.73 lateral occipital cortex,
superior division
-21.24 -78.59 44.69 lateral occipital cortex,
superior division
48.62 -65.70 4.28 lateral occipital cortex,
inferior division
'SalVentAttn' -50.48 1.50 4.36 central opercular cortex 'Default' -52.76 6.19 -11.55 temporal pole
59.27 -46.43 6.96 middle temporal gyrus,
temporoccipital part
5.73 24.77 18.24 Cingulate Gyrus,
anterior division
59.27 -46.43 6.96 middle temporal gyrus,
temporoccipital part
-35.64 22.31 -15.49 Frontal orbital cortex
39.18 -1.50 5.55 Insular cortex -52.76 6.19 -11.55 temporal pole
-5.95 21.80 31.14 Cingulate gyrus, anterior
division
-5.69 -40.56 23.26 Cingulate Gyrus,
posterior
'DorsAttn' 45.17 -28.21 42.92 postcentral gyrus 'DorsAttn' 35.76 -35.97 51.54 postcentral gyrus
59.31 -54.65 -2.55 middle temporal gyrus,
temporoccipital part
31.65 -66.54 35.43 lateral occipital cortex,
superior division
-33.11 -46.49 40.87 Superior parietal lobule -49.45 6.04 25.62 precentral gyrus
-46.24 -29.10 43.79 postcentral gyrus 45.17 -28.21 42.92 postcentral gyrus
45.17 -28.21 42.92 postcentral gyrus 26.69 -58.10 60.66 lateral occipital cortex,
superior division
'SomMot' -59.55 -2.34 11.27 precentral gyrus 'Default' -5.94 34.92 -8.76 Paracingulate gyrus
-48.69 -12.89 14.16 central opercular cortex 47.66 16.05 -20.12 temporal pole
-60.92 -17.49 19.46 postcentral gyrus 47.66 16.05 -20.12 temporal pole
-59.55 -2.34 11.27 precentral gyrus 47.66 16.05 -20.12 temporal pole
59.16 0.71 10.36 Precentral gyrus/central
opercular cortex
47.66 16.05 -20.12 temporal pole
'SalVentAttn' -5.95 21.80 31.14 Cingulate gyrus, anterior
division
'Cont' 33.92 21.13 -7.71 Insular cortex/frontal
orbital cortex
-43.35 11.93 2.29 Frontal Operculum Cortex -33.39 16.47 -8.34 Insular cortex
7.33 18.62 35.50 Cingulate gyrus, anterior
division
33.92 21.13 -7.71 Insular cortex/frontal
orbital cortex
37.20 22.63 4.75 Frontal Operculum Cortex -27.05 49.28 -13.74 Frontal pole

89
Table 4.4 MNI information for the node pairs with a significant relationship between f~ pain from Model 4. MNI coordinates for the node pairs with the five lowest pvalues from the network pairs with the ten highest percentage of significant connections. Network pairs are listed in descending order of highest percent of
significant connections. Atlas labels were determined by taking the MNI centroid for each node in the Schaefer atlas (listed as X, Y Z coordinates in the table) and
entering it into the Harvard-Oxford Cortical/Subcortical Atlas.
-32.91 24.57 -0.84 Frontal orbital cortex -9.07 -77.08 45.33 precuneous cortex
'Cont' 33.92 21.13 -7.71 Insular cortex/frontal
orbital cortex
'Cont' 7.49 34.86 25.18 Paracingulate
gyrus/cingulate gyrus,
anterior division
-34.42 -62.10 47.88 lateral occipital cortex,
superior division
-39.17 7.40 33.46 Middle frontal gyrus
-41.61 37.66 21.89 Middle frontal gyrus/frontal
pole
-3.71 28.07 47.12 Superior frontal gyrus
-33.39 16.47 -8.34 Insular cortex 7.49 34.86 25.18 Paracingulate
gyrus/cingulate gyrus,
anterior division
-33.39 16.47 -8.34 Insular cortex 33.92 21.13 -7.71 Insular cortex/frontal
orbital cortex

90
(%)
Vis SomMot DorsAttn Sal/
VentAttn Limbic Cont Default Subcortex
Vis 2.35 0.19 0.46 0.00 0.25 0.06 0.18 0.18
SomMot 6.73 0.25 3.56 0.55 0.05 1.07 0.83
DorsAttn 1.35 0.56 0.00 0.38 0.17 0.06
SalVentAttn 10.64 0.16 0.98 1.57 0.77
Limbic 0.31 0.15 0.30 0.00
Cont 0.98 0.49 0.00
Default 2.98 0.58
Subcortex 0.32
Table 4.5 Matrix representing the percent of connections significant for pain in Model 4 in each pair of networks. More
information on rank and total number of connections can be found in Table 4.6.
Network block
Significant
(%)
Significant/Total
(#)
SalVentAttn SalVentAttn 10.64 115 / 1081
SomMot SomMot 6.73 197 / 2926
SomMot SalVentAttn 3.56 129 / 3619
Default Default 2.98 122 / 4095
Vis Vis 2.35 43 / 1830
SalVentAttn Default 1.57 67 / 4277
DorsAttn DorsAttn 1.35 14 / 1035
SomMot Default 1.07 75 / 7007
SalVentAttn Cont 0.98 24 / 2444
Cont Cont 0.98 13 / 1326
Table 4.6. Ten blocks with the highest percentage of connections significant for f~pain in Model 4. Column one lists
the names of the two networks that make up a given block, column two contains the percentage of connections in
that block that are significant for pain after FDR correction, and column three lists the number of connections that
were significant out of the total number of location pairs in a network-network block. Vis=Visual,
SomMot=somatosensory/motor, SalVentAttn=Salience Ventral Attention, Cont=Executive Control.
The results from Model 5 with the interaction term for pain*multisite score revealed
dispersed and dense node pairs where functional connectivity was significantly associated with
pain* multisite score after FDR correction (q<0.05). Overall, the number of significant pairs
exceeded the number and spread of the relationships found in the original Model 4 (Figure 4.8,

91
Table 4.6, Table 4.9). Importantly, these findings were significant while including age, site, sex,
scan movement (mean FD), and scan type (rs-FB and rs-EB) as covariates. Areas where the
relationship between pain and brain activity was most likely to be modified by individual multisite
pain ratings were sensorimotor regions, the precuneus, and superior frontal gyrus (Figure 4.8,
Table 4.7). There was also significant contribution from the subcortex, largely absent from the
Model 4 findings, with particular emphasis on the caudate, putamen, and thalamus (Table 4.7).
All network pairs, with the sole exception limbic-limbic, had more significant connections than
f~pain in Model 4 (Table 4.6 versus Table 4.9). Anterior cingulate was also absent from the 5
pairs with the lowest p values in salience network blocks (Table 4.7), while insula was still
highlighted. The densest significant connections were found within the somatosensory/motor
network (40.23%, 1177 significant connections), within the subcortex (21.27%, 134 significant
connections), within the salience/ventral attention network (20.17%, 218 significant
connections), between executive control and subcortex (19.34%, 362 significant connections),
and between visual and somatosensory/motor (18.76%, 881 significant connections) (Table 4.8,
Table 4.9). The percentage of connections between somatosensory/motor and salience, denser
in significant connections than f~pain in Model 4, fell in relative rank order with 15.17%
connections significant, 552 significant connections (Table 4.6, Table 4.9). Findings largely
overlapped for both atlases (Schaefer and Brainnetome), but were more robust in Schaefer.
Brainnetome findings are included as Figure 4.9.

92
Figure 4.8 Changes in connectivity associated with the interaction between multisite pain and pain fluctuations within
participants across four visits. Of the five clinical profile markers of interest, only multisite pain significantly modified
the relationship between pain and connectivity: there were widespread increases in significant connections and
particularly dense significant connections with and within the sensorimotor cortex after FDR correction (q<0.05).
Vis=Visual, SomMot=somatosensory/motor, SalVentAttn=Salience Ventral Attention, Cont=Executive Control. a) In
the brain map illustration, spheres at each region were scaled by the number of connections with a significant
f~pain*multisite score relationship (Model 2). Spheres are color coded based on the network they belong to in the
Schaefer atlas, displayed on the axes of b), a connectivity matrix illustrating the nodes with significant f~pain
relationships. The connectivity matrix in b) is color coded by the p value of the relationship between functional
connectivity and pain (darker colors indicating connections with lower p values for pain). MNI coordinates for
significant pairs included in Table 4.7.

93
Figure 4.9 Changes in connectivity associated with the interaction between multisite pain and pain fluctuations within
participants across four visits in the Brainnetome atlas. Of the five clinical profile markers of interest, only multisite
pain significantly modified the relationship between pain and connectivity: there were widespread increases in
connectivity after FDR correction (q<0.05), but the increases don’t seem to be concentrated in any one lobe. a) In the
brain maps, spheres at each region were scaled by the number of connections with a significant f~pain*multisite
score relationship (Model 4). Spheres are color coded based on the lobe they belong to in the Brainnetome atlas,
displayed on the axes of b), a connectivity matrix illustrating the nodes with significant f~pain relationships. The
connectivity matrix in b) is color coded by the p value of the relationship between functional connectivity and pain
(darker colors indicating connections with lower p values for pain multisite score*pain). As in the Schaefer atlas, in the
Brainnetome pain-only model, we see a corresponding shift to more medial areas from the pain-only model. Because
Schaefer categorizes by networks and Brainnetome categorizes by lobe, a direct comparison of the significant areas
can’t be made, but spatially the distribution of significant connections looks quite similar.

94
MNI centroid Atlas label MNI centroid
X Y Z X Y Z
'SomMot' -47.99 -16.29 40.19 postcentral gyrus 'SomMot' 34.12 -27.19 60.96 precentral gyrus
51.55 -6.35 37.16 precentral gyrus 21.21 -24.44 66.03 precentral gyrus
60.14 -23.69 10.77 planum temporale 16.27 -18.42 72.58 precentral gyrus
-8.40 -41.66 70.18 postcentral gyrus 52.28 -12.84 49.20 postcentral gyrus
44.05 -10.85 48.22 precentral gyrus 16.27 -18.42 72.58 precentral gyrus
'Subcortex
'
-22.00 -2.00 4.00 putamen 'Subcorte
x'
12.00 -14.00 1.00 Thalamus
-12.00 14.00 0.00 caudate -14.00 2.00 16.00 caudate
-22.00 -2.00 4.00 putamen 29.00 -3.00 1.00 Putamen
-7.00 -12.00 5.00 thalamus -16.00 -24.00 6.00 Thalamus
-14.00 2.00 16.00 caudate 12.00 -14.00 1.00 Thalamus
'SalVentAt
tn'
51.00 3.41 40.76 precentral gyrus 'SalVent
Attn'
10.21 -43.25 53.38 Precuneous cortex
-36.48 4.39 10.69 central opercular
cortex
-12.57 -41.45 47.31 Precuneous cortex
-5.84 -49.12 57.02 Precuneous Gyrus 52.04 -40.89 12.22 Supramarginal
Gyrus
-32.99 19.16 8.16 frontal operculum -12.57 -41.45 47.31 Precuneous cortex
-12.57 -41.45 47.31 Precuneous cortex 7.33 18.62 35.50 Cingulate Gyrus,
anterior division
'Cont' 46.79 -44.09 46.68 supramarginal
gyrus, posterior
division
Subcorte
x
14.00 5.00 14.00 caudate
4.85 28.27 48.12 superior frontal
gyrus
-14.00 2.00 16.00 caudate

95
-3.71 28.07 47.12 superior frontal
gyrus
-14.00 2.00 16.00 caudate
4.85 28.27 48.12 superior frontal
gyrus
-12.00 14.00 0.00 caudate
4.85 28.27 48.12 superior frontal
gyrus
14.00 5.00 14.00 caudate
Vis -12.03 -80.83 35.95 Cuneal cortex SomMot 16.27 -18.42 72.58 precentral gyrus
-12.03 -80.83 35.95 Cuneal cortex 12.78 -32.86 75.64 postcentral gyrus
-12.03 -80.83 35.95 Cuneal cortex 21.21 -24.44 66.03 precentral gyrus
-12.03 -80.83 35.95 Cuneal cortex 34.12 -27.19 60.96 precentral gyrus
-12.20 -71.04 20.47 Cuneal cortex -19.01 -39.98 71.76 postcentral gyrus
Cont 44.56 18.93 43.49 Middle frontal gyrus Cont 5.74 -63.99 43.64 Precuneous cortex
-42.10 -52.18 48.64 Angular gyrus -3.71 28.07 47.12 superior frontal
gyrus
-27.05 49.28 -
13.74
Frontal pole -3.71 28.07 47.12 superior frontal
gyrus
40.95 -55.14 48.63 Angular gyrus 4.85 28.27 48.12 superior frontal
gyrus
-28.81 -74.19 42.30 Lateral occipital
cortex, superior
division
-3.71 28.07 47.12 superior frontal
gyrus
'SomMot' 16.27 -18.42 72.58 precentral gyrus 'SalVent
Attn'
-5.84 -49.12 57.02 Precuneous Gyrus
-55.94 -21.36 7.70 planum temporale -5.84 -49.12 57.02 Precuneous Gyrus
21.21 -24.44 66.03 precentral gyrus 52.04 -40.89 12.22 Supramarginal
Gyrus
-19.01 -39.98 71.76 postcentral gyrus 52.04 -40.89 12.22 Supramarginal
Gyrus
-8.59 -37.54 53.54 postcentral gyrus -39.79 -14.37 -1.75 Insular cortex
'DorsAttn' -12.94 -50.59 71.58 superior parietal
lobule
'DorsAttn' 24.75 -3.08 63.85 superior frontal
gyrus

96
Table 4.7 MNI information for the node pairs with a significant relationship between f~ multisite score*pain from Model 5. MNI coordinates for the node pairs with
the five lowest p-values from the network pairs with the ten highest percentage of significant connections. Network pairs are listed in descending order of highest
percent of significant connections. Atlas labels were determined by taking the MNI centroid for each node in the Schaefer atlas (listed as X, Y Z coordinates in the
table) and entering it into the Harvard-Oxford Cortical/Subcortical Atlas.
57.16 -22.50 44.04 postcentral gyrus 24.75 -3.08 63.85 superior frontal
gyrus
-30.02 -8.02 51.71 precentral gyrus 8.48
-30.02 -8.02 51.71 precentral gyrus 39.49
-30.02 -8.02 51.71 precentral gyrus 18.58
-70.97 52.43 Precuneous cortex
-3.09 52.88 precentral gyrus
-79.02 50.18 lateral occipital
cortex, superior
division
'Vis' 37.09 -73.33 -
15.43
Occipital fusiform
gyrus
'DorsAttn' 27.63 -3.59 52.19 Middle frontal
gyrus, superior
frontal gyrus
-24.91 -84.65 21.90 Lateral occipital
cortex, superior
division
-30.02 -8.02 51.71 precentral gyrus
-6.73 -75.77 -5.73 Lingual gyrus -30.02 -8.02 51.71 precentral gyrus
28.90 -77.75 36.73 Lateral occipital
cortex, superior
division
-30.02 -8.02 51.71 precentral gyrus
37.09 -73.33 -
15.43
Occipital fusiform
gyrus
-30.02 -8.02 51.71 precentral gyrus
'SomMot' 60.70 6.41 29.25 precentral gyrus 'DorsAttn' 13.50 -63.97 64.75 lateral occipital
cortex, superior
division
-55.94 -21.36 7.70 planum temporale -12.94 -50.59 71.58 superior parietal
lobule
16.27 -18.42 72.58 precentral gyrus 8.48
-30.16 -46.30 62.71 superior parietal
lobule
39.49
16.27 -18.42 72.58 precentral gyrus 20.86
-70.97 52.43 Precuneous cortex
-3.09 52.88 precentral gyrus
-68.57 52.90 superior parietal
lobule

97
(%)
Vis SomMot DorsAttn Sal/
VentAttn Limbic Cont Default Subcortex
Vis 6.56 18.76 12.69 11.51 0.50 3.28 1.13 1.78
SomMot 40.23 12.42 15.25 5.29 5.02 8.65 4.26
DorsAttn 12.95 7.40 1.51 5.73 3.77 10.87
SalVentAttn 20.17 5.24 7.00 5.38 11.70
Limbic 0.31 2.51 0.85 1.50
Cont 18.70 7.23 19.34
Default 6.96 4.73
Subcortex 21.27
Table 4.8 Matrix representing the percent of connections significant for pain*multisite score in Model 5 in each pair of
networks. More information on rank and total number of connections can be found in Table 9.
Network block
Significant
(%)
Significant/Total
(#)
SomMot SomMot 40.23 1177 / 2926
Subcortex Subcortex 21.27 134 / 630
SalVentAtt
n
SalVentAtt
n 20.17 218
/
1081
Cont Subcortex 19.34 362 / 1872
Vis SomMot 18.76 881 / 4697
Cont Cont 18.70 248 / 1326
SomMot
SalVentAtt
n 15.25 552
/
3619
DorsAttn DorsAttn 12.95 134 / 1035
Vis DorsAttn 12.69 356 / 2806
SomMot DorsAttn 12.42 440 / 3542
Table 4.9 Ten blocks with the highest percentage of connections significant for f~pain*multisite score in Model 5.
Column one lists the names of the two networks that make up a given block, column two contains the percentage of
connections in that block that are significant for pain*multisite score after FDR correction, and column three lists the
number of connections that were significant out of the total number of location pairs in a network-network block.
Vis=Visual, SomMot=somatosensory/motor, SalVentAttn=Salience Ventral Attention, Cont=Executive Control

98
Figure 4.10 Unscaled version of Figure 6a for f~pain and Figure 8b for f~pain*multisite score. Networks are colorcoded and all spheres are the same size, regardless of the number of significant node pairs they are a part of.
Nodes with a significant relationship between both functional connectivity and pain
(Model 4) and functional connectivity and pain*multisite score (Model 5) were mostly present in
the somatosensory/motor network (Table 4.10, Figure 4.11), with an area of particularly dense
significant connections in the right hemisphere. To a lesser extent, the salience/ventral attention
network was also highlighted (Table 4.10, Figure 4.11). The top three network pairs of
overlapping significant connections were within somatosensory/motor (3.18%, significant 93
connections), within Salience/Ventral Attention (2.13%, 23 significant connections) and between
somatosensory/motor and salience/ventral attention (0.86%, 31 significant connections). The
areas unique to Models 4 and 5 were also examined but no single network was clearly
highlighted for either model; regions were distributed roughly in accordance with Models 4 and
5 run independently (Figures 4.6 and 4.8 versus Figure 4.11).

99
Figure 4.11 Areas of connectivity with significant relationships to both pain (Model 4) and pain*multisite score (Model
2). Vis=Visual, SomMot=somatosensory/motor, SalVentAttn=Salience Ventral Attention, Cont=Executive Control. a)
In the brain maps, spheres at each region were scaled by the number of connections. Bilateral somatosensory/motor
areas are the standout regions of overlap, with an area of particularly dense significant connections in the right
hemisphere. Spheres are color coded based on the network they belong to in the Schaefer atlas, displayed on the
axes of b), a connectivity matrix illustrating the nodes that had significant relationships with both pain and
pain*multisite score (Models 1 and 2, respectively). Color coding is now binary, with white indicating no significant
overlapping areas, and red indicating significant overlapping areas.

100
Figure 4.12 Significant areas unique to the pain only model (a and b) versus the those unique to the pain*multisite
score model (c and d)

101
Network Block
Significant
(%) Significant/Total (#)
'SomMot' 'SomMot' 3.18 93 2926
'SalVentAttn' 'SalVentAttn' 2.13 23 1081
'SomMot' 'SalVentAttn' 0.86 31 3619
'Cont' 'Cont' 0.60 8 1326
'Default' 'Default' 0.42 17 4095
'SalVentAttn' 'Cont' 0.25 6 2444
'DorsAttn' 'SalVentAttn' 0.23 5 2162
'SalVentAttn' 'Default' 0.14 6 4277
'Cont' 'Default' 0.13 6 4732
'DorsAttn' 'Cont' 0.13 3 2392
Table 4.10 Ten blocks with the highest percentage of connections significant for f~pain in Model 4. node pairs that
were significant for f~pain in Model 4 and f~ multisite score in Model 5. Column one lists the names of the two
networks that make up a given block, column two contains the percentage of connections in that block that are
significant for f~pain* multisite score and f~pain after FDR correction, and column three lists the number of
connections that were significant out of the total number of location pairs in a network-network block.Vis=Visual,
SomMot=somatosensory/motor, SalVentAttn=Salience Ventral Attention, Cont=Executive Control.
We found that pain significantly impacted mental and physical functioning as assessed
by the SF12: PCS-12 and MCS-12 (p<<0.01). Fixed effects for pain showed an overall negative
relationship with both mental and physical functioning: mental functioning was approximately
25% lower and physical functioning was approximately 30% lower for patients with a pain rating
of 10 versus patients with a pain rating of 0. Most of the random effects (individual function~pain
lines for each patient) followed the same trend (Figure 4.13). Similarly, pain ratings were well
predicted by whether the patient reported being in a flare (yes or no) (p<<0.01). GUPI pain
ratings were approximately 30% higher when patients reported being in a flare versus when
they reported no flare (Figure 4.14). This trend held true for both the fixed and most random
effects.

102
Figure 4.13 Impact of pain on physical and mental functioning as assessed by the SF12. As expected, we found that
pain significantly impacted a) mental and b) physical functioning (p<<0.01). Fixed effects for pain showed an overall
negative relationship with both mental and physical functioning: mental functioning was approximately 25% lower and
physical functioning was approximately 30% lower for patients with a pain rating of 10 versus patients with a pain
rating of 0. Most random effects (individual function~pain lines for each patient) followed the same trend. Black line
indicates the line of fixed effect, with individual patient lines in multicolor.
Figure 4.14 Pain ratings were well predicted by whether the patient reported being in a flare (yes or no) (p<<0.01).
Patient GUPI pain ratings were approximately 30% higher when reporting they were in a flare versus when they
reported no flare. This trend held true for both the fixed and most random effects. Black line indicates the line of fixed
effect, with individual patient lines in multicolor.

103
4.4. Discussion
Neural circuits underlying chronic pain fluctuations are not well understood, including the
effect of hypothesized clinical markers of centralized pain. In the present study, we used withinsubject, longitudinal analyses of pain fluctuations and clinical characteristics to address both.
We confirmed Symptom Patterns Study (SPS) participants experienced fluctuating pain without
overall improvement or worsening (Figure 4.2). This is in line with past research suggesting
long-term symptom profiles in UCPPS are generally stable for differing symptom burdens (in
this case high and low pain), but still fluctuate within individuals over time (Stephens-Shields et
al., 2016, Bradley et al., 2022).
Our findings suggest that, in UCPPS, ecologically valid pain rated in the moment versus
recalled, average pain from the past week activate the same areas in the pain network that
respond to experimentally induced acute pain in both healthy, pain free individuals and
individuals with chronic pain (Xu et al. 2021; Xu et al. 2020). However, recalled pain
(PainRecent) is more strongly associated with functional connectivity during empty bladder
scans, while current pain (PainNow) is more strongly associated with functional connectivity
during full bladder scans (Figure 4.3). Interestingly, these relationships appear to be modulated
by different clinical features, with the relationship between PainRecent and empty bladder being
primarily modulated by anxiety and the relationship between PainNow and full bladder being
primarily modulated by multisite pain (Figure 4.4).
We identified two major networks underlying pain fluctuations outside of the pain
network: somatosensory/motor and salience. Connectivity within and between these two
networks was the most likely to change as participants reported changes in pain both in the
empty bladder scans alone, and when the empty and full bladder scans were combined into a
single analysis (Figure 4.5, Figure 4.6, Table 4.5, Table 4.6). In the combined scans analysis
Model 5, only multisite pain significantly modified the relationship between pain and functional
connectivity. Connectivity within salience and somatosensory/motor networks was again

104
emphasized, but all networks saw many node pairs where the relationship between pain and
functional connectivity was significantly modified by multisite pain (Figure 4.8, Table 4.8, Table
4.9).
For connections with a significant relationship between functional connectivity and pain
(Model 4), our analysis primarily highlighted areas within the somatosensory/motor and salience
networks. Both networks play an important role in perception of pain: the insula and anterior
cingulate, which are nodes of the salience network (Hegarty et al., 2020), and somatosensory
cortex (S1) all process ascending pain signals. Salience network activity was previously shown
to be altered between patients and controls (De Ridder et al., 2022) and is associated with other
potential neural markers of centralized pain (McLain et al., 2022). Alterations in S1 activity have
been observed in chronic pain conditions and heightened pain sensitivity (Apkarian et al., 2005).
Highlighted nodes in salience included insula and anterior cingulate. The insula is
involved in the perception of acute pain and shows significant activity differences between
healthy controls and chronic pain patients (Borsook et al., 2013; Yam et al., 2018). Anterior
cingulate is proposed as mediating the affective components of pain (Vogt, 2005; Fuchs et al.,
2014; Xiao et al., 2021; Jee et al., 2023) and is implicated in pathological pain modulation in
animal models (Xiao et al. 2021). In somatosensory/motor, we identified widespread clusters
across sensorimotor regions, including large clusters both medially and laterally. While medial
sensorimotor regions are consistent with pain in the trunk/pelvis region of the body (painful body
region in UCPPS), the lateral clusters of the sensorimotor region were denser. There are
multiple possible interpretations of these results. The first is poor spatial specificity in functional
connectivity processing: the lateral clusters identified are near the insula. A second possibility is
the representation of the primary painful body region may be less significant to pain fluctuations
than lateral regions closer to the face and hand areas. The lateral regions could be interpreted
in the context of new research suggesting effector regions of the classic homunculus are
interspersed with integrative hubs representing multiple body regions (Gordon et al., 2023).

105
Understanding this complex spatial distribution of sensorimotor regions in pain is critical as
there are several unanswered therapeutic questions about neuromodulation of motor cortex for
pain (Lefaucheur et al., 2020): our results might suggest that, when stimulating M1, specifically
stimulating the primary painful body site may provide little additional benefit.
The default mode network (DMN) has been previously reported as having altered activity
in chronic pain patients (Baliki et al., 2014; Čeko et al., 2020; De Ridder et al., 2022) and
dysregulation of the interaction between DMN and salience is prevalent in chronic pain
conditions (Kucyi and Davis, 2015; Kim et al., 2018; van Ettinger-Veenstra et al., 2019; De
Ridder et al., 2022). Interestingly, our findings show relatively few significant connections
between the DMN and salience when compared to within salience, within somatosensory/motor,
and between salience and somatosensory/motor. While still significant, the connectivity of the
DMN with the salience network seems to have a less robust relationship with pain fluctuations
within individuals than previously seen between individuals.
Pain centralization involves heightened contribution of the central nervous system to
how much pain a patient experiences (Li et al., 1999, 2021; Wu et al., 2005; Harte et al., 2018;
Sellmeijer et al., 2018; Ellingsen et al., 2021) suggesting factors modifying the relationship
between brain activity and pain intensity can be understood as centrally amplifying or
suppressing. Common clinical features/questionnaires previously associated with centralized
pain are anxiety, depression, neuropathic symptoms assessed by the painDETECT
questionnaire, pain catastrophizing, and multisite pain. It is well established that chronic pain
patients have diverse clinical characteristics and that these characteristics are more likely to cooccur with centralized pain conditions (Borchers and Gershwin, 2015; Kutch et al., 2017; Conti
et al., 2020; van Ettinger-Veenstra et al., 2020; Ellingsen et al., 2021; Kratz et al., 2021).
However, our work suggests that multisite pain in particular modifies the within-person
relationship between brain activity and pain. This was not the case for the other clinical features
we examined.

106
The interaction between pain and multisite pain in Model 5 produced significant
connections across all networks, more than for pain alone in Model 4. As in Model 4, the areas
of densest connection were within the somatosensory/motor regions and salience networks, but
with a relatively higher medial emphasis and new emphasis on connectivity within subcortex
and executive control networks. Slightly lower in overall percent of significant connections, but
still more robust than in Model 4, was the connectivity between somatosensory/motor and
salience/ventral attention. A previous study identified increased salience-sensorimotor
connectivity in fibromyalgia and UCPPS patients with multisite pain (Kutch et al., 2017), while
stronger connectivity between salience, somatosensory, and default mode network is shown to
have a potentially causal relationship with the development of multisite pain in data from the
Adolescent Brain and Cognitive Development Study (Kaplan et al., 2022). Regions common to
Models 4 and 5 (Figure 4.8) show the networks of largest overlap are somatosensory/motor
and to a lesser extent, salience network. This suggests that multisite pain augments the
relationship between pain and functional connectivity, specifically in the somatosensory/motor
network.
Multisite pain's role as an indicator of centralized pain aligns with previous research
(Brummett et al., 2013, 2015; Kutch et al., 2017). However, the lack of significant results for
other clinical factors like depression, anxiety, and pain catastrophizing in Model 5, despite their
known associations with centralized pain (Borchers and Gershwin, 2015; Edwards et al., 2016;
van Ettinger-Veenstra et al., 2020; Ghasemi et al., 2022), is surprising. A possible explanation is
nicely summarized by the triple network model of pain (De Ridder et al., 2022) which comprises
three dimensions produced by independent pathways: pain, suffering, and inhibition. Changes
in pain intensity and multisite pain may be products of the pain pathway while depression,
anxiety, and pain catastrophizing are features specifically proposed as part of the suffering
pathway. Because of the potential similarity in neural substrate, the modifying effect of multisite
pain on the relationship between pain and functional connectivity may be stronger than that of

107
depression, anxiety, or pain catastrophizing when looking at the combined empty and full scan
data. This is indeed supported by the findings in Model 5, where multisite pain seems to
specifically modulate the relationship between brain activity and pain felt in the current moment,
while anxiety specifically modifies the relationship between functional connectivity and recalled
pain, a metric that might activate the suffering pathway as a patient is forced to reflect on their
symptoms over the past week. While a previous systematic review of the literature on selective
modulation of pain and suffering found no compelling evidence that the dimensions could be
independently modulated, it only focused on cognitive manipulations (Talbot et al., 2019). The
findings in this work may tap into the pathways in a more naturalistic way, revealing the
potentially different roles anxiety and multisite pain may play in the day-to day experience of
chronic pain. It is notable, however, that when integrating across states, only multisite pain
remains significant, despite the outcome variable being PainRecent. This perhaps suggests
multisite pain may have a larger impact on the overall levels of pain experienced, while anxiety
may have a larger impact on how intense/disruptive the chronic pain condition feels, even in the
absence of a triggering stimulus (empty bladder versus full). Further investigation into the
specific roles that depression, anxiety, and pain catastrophizing play in centralized pain
conditions is needed.
A potential limitation of this work is the lack of clarity on which parsing of the fMRI data
“best” captures the most relevant brain state for chronic pain. Past studies of the relationship
between brain activity and reported pain in chronic pain populations has largely focused on
acute, applied pain such as heat and electrical stimuli (Xu et al., 2021). This work aimed to
capture brain states more directly associated with ecologically valid pain: that is, pain that an
individual would experience naturally as a result of their chronic pain condition. The inclusion of
a full and then empty bladder scan was an attempt to capture the range of states an individual
with UCPPS might experience over the course of a given day, however the full bladder
paradigm might be considered acute application of a painful stimulus regardless of its ecological

108
validity. Previous work on the SPS dataset has examined the pain response to the FB paradigm
and found that it does cause discomfort in patients with UCPPS (Schrepf et al., 2023). The
paradigm should largely mirror the naturalistic experience of bladder filling throughout the day,
but it is still unclear whether this may be comparable to a non-ecologically valid acute pain
stimulus such thermal or electrical pain (Xu et al., 2021). For this reason, we included three
groupings of the brain data: empty only, full only, and empty and full together in a single model.
Future work should attempt longer resting state scans with multiple pain ratings throughout to
determine if the associations identified in this paper persist in the absence of a laboratory
applied stimulus.
We are aware of no other analysis examining the neural correlates of pain fluctuations
within individuals longitudinally on this time scale. Using individuals as their own controls is a
powerful form of analysis, more likely to capture brain regions important for the perception of
pain despite individual differences. We found that the primary networks of importance for
fluctuations in pain was the network previously identified to change in response to acute pain,
as well as, at the whole brain level, sensorimotor and salience networks. Multisite pain in
particular augments this relationship, although anxiety also plays a role when focusing
specifically on recalled pain during empty bladder states. Given past evidence that multisite pain
is linked to centralized pain conditions, we propose that multisite pain, more so than other
clinical factors considered, indicates central amplification of pain signals. This information will be
significant when considering treatments for patients with chronic pain, potentially reducing the
trial-and-error approach in identifying centralized pain conditions.

109
Chapter 5. Peak alpha frequency is associated with chronic pain
diagnosis but not pain intensity
Abstract
Background: Low peak alpha frequency (PAF) is an electroencephalography (EEG) outcome
reliably associated with increased acute pain sensitivity. However, existing research suggests
that the relationship between PAF and chronic pain is more variable. Indeed, while emerging
work has compared individuals with chronic pain to healthy controls, no previous studies have
examined differences in PAF between diagnoses or across chronic pain subtypes. This
inconsistency could be attributable to chronic pain groups typically being examined as
homogenous populations, without consideration being given to potential diagnosis-specific
differences. Methods: To address this gap, we reanalyzed a dataset of resting state EEG
previously used to demonstrate a lack of difference in PAF between individuals with chronic
pain and healthy controls. In this new analysis, we separated patients by diagnosis before
comparing PAF across three subgroups: chronic widespread pain (n=30), chronic back pain
(n=38), and healthy controls (n=87). We repeated the same analysis within separate, single
diagnostic group, and subtyped instead with an identified clinical marker of centralized pain:
pain distribution (localized n=14, widespread n=28). Results: We found that individuals with
chronic widespread pain had significantly higher global average PAF values than those with
chronic back pain [p=0.028, β=0.25 Hz]. We also found that, within a single diagnostic group,
patients with widespread pain had significantly lower PAF values than patients with localized
pain [p=0.012, β=-0.69 Hz]. These results controlled for age, sex, and depression. PAF was not
associated with pain intensity in either analysis. Conclusions: These novel findings reveal
diagnosis-specific differences in EEG recordings across individuals with chronic pain. Our work
suggests that PAF shifts are unlikely to be a general marker for chronic pain, and holds
important implications for future work exploring this measure in the context of pain physiology.

110
5.1. Introduction
Chronic pain has been identified as a global research priority, with an estimated
worldwide prevalence of 10-25% (Goldberg and McGee, 2011). Given this impact, there is
growing interest in identifying objective markers that can shed light on pain etiology and guide
the development of intervention and prevention strategies. One marker of increasing interest is
peak alpha frequency (PAF). This electroencephalography (EEG) measure refers to the most
prominent frequency within the alpha band of brain wave activity, which typically ranges
between 7 and 13 Hz (McLain et al., 2022). Slowed PAF has been associated consistently with
increased acute pain sensitivity in healthy adults (Furman et al., 2018, 2019, 2020; Raghuraman
et al., 2019). However, literature examining the relationship between PAF and chronic pain is
variable, with reductions (Lim et al., 2016; Sarnthein et al., 2006; Sato et al., 2017; Sufianov et
al., 2014; de Vries et al., 2013), increases (Fauchon et al., 2022), and no differences (van den
Broeke et al., 2013; Schmidt et al., 2012; Ta Dinh et al., 2019; Witjes et al., 2021) in PAF being
observed between people with chronic pain and healthy controls. As previous work has
demonstrated that PAF is robust against common differences in EEG processing pipelines
(Chowdhury et al., 2023; McLain et al., 2022), such discrepancies are unlikely related to EEG
collection or processing methods.
The variability observed in chronic pain literature could be attributable to examining
chronic pain groups as relatively homogenous populations, without giving consideration to
potential diagnosis and etiology specific differences. Indeed, previous work has often averaged
findings across several diagnoses to form a single “chronic pain” group that is then compared to
controls (Zebhauser et al., 2023), obscuring possible differences across chronic pain diagnoses.
While PAF comparisons between healthy controls and individual chronic pain subtypes such as
back pain (Schmidt et al., 2012), neuropathic pain (Fauchon et al., 2022; Krupina et al., 2019)
and cancer-related pain (van den Broeke et al., 2013) do exist, no previous studies have
examined differences in PAF between distinct diagnoses. Past functional magnetic resonance

111
imaging (fMRI) studies suggest that this approach may in fact confound findings drawn from
neural recordings, as diagnostic categories and certain features of chronic pain, such as pain
distribution, differentially affect brain network function (Kaplan et al., 2022; Kutch et al., 2017).
Indeed, pain distribution was identified in Chapter 4 as a strong candidate marker of centralized
pain. In addition to there being no analysis of PAF differences between distinct diagnostic
groups, there has also been no examination of PAF differences between patients with different
pain distributions within a single diagnostic category. Some attempts have been made to identify
PAF shifts as being specific to neuropathic pain (Fauchon et al. 2022; Boord et al. 2007; Di
Pietro et al. 2018; Kim et al. 2019), but sample sizes are often small, and the definition of
“neuropathic” is not always consistent. As discussed in Chapter 1, some forms of neuropathic
pain are categorized as “central neuropathic,” meaning there could be some overlap between
the investigations into neuropathic pain populations and individuals with centralized pain
conditions.
In this study, we reanalyzed a previously published dataset of healthy controls and
individuals with chronic pain of mixed diagnoses as well as an independent sample of a single
diagnostic group, urologic chronic pelvic pain (UCPPS). We aimed to examine the effect of
diagnosis on PAF by subdividing the first dataset into healthy controls and the two largest
diagnostic categories within the chronic pain group: chronic back pain and chronic widespread
pain. Subsequently, we aimed to examine the effect of pain distribution by subtyping the group
of UCPPS patients based on their pain distribution types (localized versus widespread). While
the first dataset previously showed a lack of patient-control differences in PAF, we hypothesized
that the relationship between PAF and the presence of chronic pain is variable across
diagnoses, which may have been a confounding factor in previous analyses. We additionally
hypothesize that in the independent group of UCPPS patients, PAF will vary based on pain
distribution, a potential clinical marker for centralized pain.

112
5.2. Materials and Methods
5.2.1. Dataset one
5.2.1.1. Data acquisition
The raw EEG (https://osf.io/hbkms/) and associated participant data
(https://osf.io/srpbg/) used for this analysis were downloaded from Open Science Framework
under Creative Commons (Attribution-NonCommercial-ShareAlike 4.0) International Public
License without any restrictions from the collecting lab (May et al., 2021).
5.2.1.2. Participants
Details regarding data collection and eligibility criteria can be found in the original
publication of the dataset upon which this study is based (Ta Dinh et al., 2019). Briefly, to be
eligible for inclusion, participants in the chronic pain group required a clinical diagnosis of
chronic pain, a minimum pain duration of six months, and a minimum average pain intensity of
4/10 (numerical pain rating scale) over the preceding four weeks. Participants were excluded if
they presented with any acute changes in pain intensity due to events like surgery or injury, use
of benzodiazepines, or a diagnosis of a major neurological (epilepsy, stroke, or dementia) or
psychiatric (excluding depression) disease.
The files downloaded from OSF contained data from 101 people with chronic pain and
89 healthy control participants. In the chronic pain dataset, there were a total of 30 patients with
a diagnosis of chronic widespread pain (CWP) and 47 with a diagnosis of chronic back pain
(CBP). The remaining people with chronic pain (n=24) had one of the following diagnoses: joint
pain (n=6), unspecific neuropathic pain (n=5), postherpetic neuralgia (n=7), and polyneuropathic
pain (n=6). Two patients could not be processed due to incomplete or missing files: one patient
with back pain and one patient with polyneuropathic pain. In the healthy control dataset, data
from one participant could not be processed due to incomplete or missing files, and one
participant was excluded due to an absence of available demographic information. This resulted
in a total of 99 people with chronic pain and 87 healthy control participants. This differs slightly

113
from the numbers reported in Tah Dinh et al., 2019 (101 patients and 84 controls) and May et al.
2021, (101 patients and 88 controls). As we were interested in the relationship between
diagnosis and PAF, we focused on the two pain groups with the largest samples: chronic back
pain (n=46) and chronic widespread pain (n=30). These two pain groups were compared
against each other and the healthy control group (n=87).
5.2.1.3. EEG and survey collection
As described by Ta Dinh, et al., 2019, resting state data were acquired with participants in a
relaxed but awake state. Collections were performed in two, five-minute blocks (one eyes-open
and one eyes-closed) using a 64-channel Brain Vision EasyCap electrode system at a sampling
frequency of 1000 Hz and with an online bandpass filter of 0.016 and 250 Hz. Impedance was
kept below 20 kΩ. All of the standard, 10-20 electrodes were included, as well as electrodes
Fpz, CPz, POz, Oz, Iz, AF3/4, F5/6, FC1/2/3/4/5/6, FT7/8/9/10, C1/2/5/6, CP1/2/3/4/5/6,
TP7/8/9/10, P5/6, and PO1/2/9/10. Data were online referenced to FCz and grounded at AFz.
While both recordings with eyes-open and eyes-closed were included in the original dataset, the
final analyses of Ta Dinh, et al., 2019 and May et al. 2021 used only the eyes-closed data due
to better data quality and more stable results. Most papers in the PAF-pain literature also rely on
eyes-closed data (McLain et al., 2022). In keeping with these processing decisions, we used
only eyes-closed data for this analysis.
Clinical questionnaires were collected before the EEG recordings. Survey data that were
collected for both back pain and widespread pain in the original study were from the short-form
McGill Pain Questionnaire (Melzack, 1975), Beck Depression Inventory II (BDI-II) (Beck, 1996),
and State-Trait-Anxiety Inventory (Spielberger, 1970).
5.2.1.4. EEG Preprocessing
Parameters for preprocessing were primarily selected based on our previously published review
of the PAF-pain literature (McLain, et al., 2022). All offline data processing was performed with
custom EEGLAB v13.6.5.b and MATLAB (R2018b) scripts. Each participant had 5 minutes of

114
continuous eyes-closed signal. Preprocessing was conducted across all 64 channels of data for
each participant. The signal was first bandpass filtered from 1 to 100 Hz and a notch filter
applied from 58 to 62 Hz to attenuate electrical noise in the bandpassed signal. Data were
further pre-processed by ICA analysis to prepare the signal for automated artifact removal via
the Multiple Artifact Rejection Algorithm (MARA). MARA is an ICA-based, optimized linear
artifact classifier trained on data that has been visually inspected and scored manually for
artifact rejection. It reliably detects a range of biological and non-biological artifacts, including
eye movement, muscle, and electrical artifacts. We performed artifact rejection in an automated
manner using the default setting in MARA to reject any component with artifact probabilities
greater than 0.5 (Winkler et al. 2011; Winkler et al. 2014). The 50% threshold is implemented by
the original authors of the algorithm for automated removal (unsupervised implementation) and
is the level set in the Harvard Automated Processing Pipeline for Electroencephalography
(HAPPE) for optimal, automated removal of artifacts that balances data cleaning with
maintaining the integrity of the original data (Gabard-Durnam et al. 2018; Winkler et al. 2011;
Winkler et al. 2014). After artifact removal, each channel’s signal was re-referenced to the
common average of all channels for each participant. Preprocessed data were divided into oneminute epochs and epochs with peak-to-peak amplitude exceeding 80 mV were excluded from
further analysis. Within each remaining epoch, the power spectral density (PSD) with
normalized units of the EEG frequencies for each eyes-closed epoch were computed in
MATLAB using a moving average with 8-second windowing and no overlap. Eight-second
windowing was selected to minimize the impact of spectral leakage (Zalewska, 2020). For all
epochs surviving the 80 mV correction, PAF was calculated as the weighted sum of the alpha
spectrum divided by the total power, resulting in the ‘center’ of the spectral power, or the ‘center
of gravity (COG)’. This approach is shown to be the most stable measure of PAF determination
(Brötzner et al., 2014; Klimesch, 1997, 1999; Klimesch et al., 1993). The COG was calculated
and then averaged across all epochs at a given electrode for each participant, resulting in 64

115
COG values (one for each channel) for every participant. These values were further averaged
for a single, global average PAF value for each participant.
5.2.1.5. Data analysis and visualization
To visualize any initial differences between diagnostic groups, the power spectra for
each participant at each channel were normalized to the average power across all channels for
that participant. Then, this normalized power spectrum was averaged separately for each
diagnostic category (back pain, widespread pain, and healthy controls) and plotted across the
64 electrodes. To further visualize group differences, the average PAF value at each electrode
was calculated for each of the three groups (chronic back pain, chronic widespread pain, and
controls). The differences between the topographic, group average PAF values were then
visualized on a scalp map.
We previously demonstrated that the global average values capture over 95% of the
variance in topographic PAF values (McLain et al., 2022), so we chose to focus on the global
average values for the primary analysis. Our previous analysis, however, was also in a healthy
male population. Therefore, to replicate our previous analysis and confirm that global average
PAF still captures the majority of the variance in a population of healthy controls and patients
with chronic pain, we additionally ran a principal components analysis (PCA) across the PAF
values at every electrode for all participants in this data set, treating each participant as an
observation and each electrode as a variable.
To address the hypothesis that PAF values would be different between the unique
chronic pain diagnoses, we performed a linear regression analyses with global average PAF as
the outcome variable and diagnosis category as the predictor. Age has a well-documented
relationship with PAF values (Chiang et al., 2011; Osaka et al., 1999; Clark et al., 2004) and
while the literature on sex (Chiang et al., 2011; Cragg et al., 2011; Freschl et al., 2022) and
depression (Arns et al., 2012; Jiang et al., 2016; Schulman et al., 2011; Zhou et al., 2023) is
less consistent, we controlled for all three in the model. The reference level of the diagnosis

116
variable was set both to chronic widespread pain and chronic back pain so all three pairwise
comparisons could be examined. To address any potential spatial differences not captured in
the global average, we additionally ran a simplified model of PAF predicted by diagnostic
category controlled for age and sex at each electrode and corrected for multiple comparisons
using a Benjamini-Hochberg false discovery rate (FDR) correction (Benjamini and Hochberg
1995).
Another measure with a potential relationship to PAF is pain intensity (Furman et al.
2018; Raghuraman et al. 2019): this could not be entered into the primary analysis as healthy
controls did not have current pain scores. We instead ran an additional analysis in the patient
subgroups to determine if pain intensity had a significant relationship with PAF, and whether this
relationship survived the same correction for age, sex, and depression scores.
5.2.2. Dataset two
5.2.2.1. Participants
Participants were selected from two study populations: a study of chronic prostatitis
(CP) in males and an ongoing study of interstitial cystitis (IC)/bladder pain syndrome (BPS) in
females. IC/BPS and CP both fall under the umbrella diagnosis of urologic chronic pelvic pain
(UCPPS), with IC/BPS typically being diagnoses in females and CP typically being diagnoses in
males. Participants in both studies had resting state EEG data as well as clinical surveys for
depression, anxiety, pain distribution (number of painful body sites), and pain intensity.
Participants were eligible if they had a diagnosis of chronic prostatitis or interstitial
cystitis/bladder pain syndrome, were in pain the majority of the time over the past three months,
and had a minimum, worst pain rating of 4/10 in the past month. Additionally, patients were
older than 18 years of age, able to participate in the informed consent process, safe to be
scanned by magnetic resonance imaging, had no active urinary, anal, or genital infection, and
no severe, urgent, or debilitating medical condition. All aspects of the study conformed to the

117
principles described in the Declaration of Helsinki and were approved by our Institutional
Review Board. All participants provided informed consent.
The original pool of patients for the IC/BPS study comprised 29 females and the original
pool of patients for the CP study comprised 48 males. Patients were matched across the studies
for age and pain distribution to create a final, combined data set of 38 patients,19 from each
group.
5.2.2.2. EEG and survey collection
At the beginning of their visit, participants responded to a series of clinical questionaries and
demographic surveys. Anxiety and depression were both assessed with the Hospital Anxiety
and Depression scale (Zigmond and Snaith, 1983): possible scores for both anxiety and
depression scales 0-21. Participants also gave their current pain using the VAS, and marked all
locations on a body map where they experienced pain.
Continuous EEG data were collected using a 64-channel, ANT Neuro gel-based
electrode cap with sintered Ag/AgCl electrodes. The online reference was placed at the right
mastoid. Signal was acquired with eego sports acquisition software (v1.2.1) from an Ant Neuro
eego Sports amplifier (product number ee-202) at a sampling rate of 2048 Hz. Impedances for
all electrodes were kept below 15 kΩ.
Participants were lying supine and told to keep their head as still as possible, relax, and
not go to sleep. Participants in the CP study were then told to follow automated, alternating
voice commands to open or close their eyes. The continuous recording was annotated at the
beginning of each eyes-open/eyes-closed epoch: in total, there were ten minutes of continuous
EEG data with non-overlapping, one-minute epochs marked for five eyes-open and five eyesclosed periods. In the IC/BPS study, participants did not alternate between eyes open and eyes
closed, and instead did one continuous, ten-minute collection of eyes-closed EEG.

118
5.2.2.3. EEG Preprocessing
All offline data processing was performed with EEGLAB v2024.0 and MATLAB (2024a)
scripts. Participants in the IC/BPS study had 10 minutes of continuous eyes-closed data, and
participants in the CP group had 5, 1-minute segments of continuous data. To mimic the
structure of the IC/BPS collection, the CP data segments were interpolated into a single, 5-
minute continuous recording. Then data were processed using the same steps described in
5.2.1.2
The COG was calculated and then averaged across all epochs at a given electrode for
each participant, resulting in 64 PAF values (one for each channel) for every participant. These
values were further averaged for a single, global average PAF value for each participant.
5.2.2.4. Data analysis and visualization
Participants were dichotomized into widespread and localized pain subgroups based on their
body map data. Any patient with three or more painful body sites was classified as a patient with
widespread pain as in several past publications (Cashin et al. 2019; Wu et al. 2018; Kutch et al.
2017; Pan et al. 2019). We assessed the hypothesis that pain distribution would impact PAF
value in two ways. For our primary analysis, we ran a linear regression with global average PAF
as the outcome variable and widespread vs localized pain as the predictor of interest while also
controlling for age, sex, and depression as in the analysis of dataset one in 5.2.1.

119
Chronic pain groups Healthy controls
Total n=155 n=87
Widespread pain
n=30
Back pain
n=46
Age (years) 54.7±13.56 58.78 ± 13.66 57.71±14.24
Sex (f/m) 25/5 26/20 59/28
Pain at time of
collection (1-10)
5.81±1.87 5.00±1.85 NC
Anxiety
STAIX 1 40.13±8.43 41.11±11.26 30.44±5.95
STAIX 2 47.93±11.61 44.33±10.23 30.82±7.17
Depression 20.35±9.59 15.96±8.19 3.70±4.54
McGill Questionnaire
Sensory 18.33±4.84 12.72±5.62 NC
Affective 5.63±2.41 3.76±2.89 NC
Total 32.85±6.85 24.48±9.24 NC
Table 5.1 Demographic information and clinical characteristics for three groups in dataset one: chronic widespread
pain, chronic back pain, and healthy controls. NC=not collected
We did not have sufficient subjects to enter pain intensity into the model, so as in section 5.2.1,
we separately assessed the relationship between PAF and pain intensity with a simple
correlation. As in section 5.2.1, to address any potential spatial differences not captured in the
global average, we additionally ran the same model at each electrode and corrected for multiple
comparisons using a Benjamini-Hochberg FDR correction (Benjamini and Hochberg 1995).

120
Figure 5.1 Group differences in global average PAF:back pain versus controls versus widespread pain. Patients with
back pain trend lower than widespread and control groups in PAF, but the only significant relationship after controlling
for age and sex was between the back pain and widespread pain groups.
5.3. Results
5.3.1. Dataset one
5.3.1.1. Participant characteristics
The demographic information for patients and controls is found in Table 5.1. The mean
age across all three groups was consistent. The proportion of females was higher in the
widespread pain (25f/5m) and control groups (59f/28m) than in the back pain group (26f/20m),
with all groups having more females than males. These data are consistent with epidemiological
data, where chronic pain conditions have an overall higher prevalence among women than men
(Greenspan et al., 2007), with this sex bias being more pronounced in chronic widespread pain
when compared to chronic back pain (Andrews et al., 2018; Maixner et al., 2016; Overstreet et

121
al., 2023; Viniol et al., 2013; Vos et al., 2020). Additionally, both pain groups scored higher than
controls in terms of depression and anxiety.
5.3.1.2. Differences in PAF between diagnoses
The mean PAF for each group (mean ± standard deviation) was 10.09 ± 0.37 for chronic
widespread pain patients, 9.78 ± 0.47 for back pain patients, and 9.92 ± 0.52 for healthy
controls (Figure 5.1). There was a significant difference between back pain and widespread
pain as a predictor of global average PAF after controlling for sex, age, and depression
(p=0.028). With chronic widespread pain as the reference category, back pain had a β value of -
0.25 Hz indicating that, holding all other variables constant, patients with back pain had
approximately 0.25 Hz lower PAF values than patients with chronic widespread pain (Table
5.2). There was no significant difference between chronic widespread pain and controls, nor

122
Figure 5.2 Averaged power spectrum for the three groups (chronic back pain, chronic widespread pain, and controls)
across four ROIs. Differences appear to increase moving posteriorly, with the least amount of overlap in the occipital
and parietal regions.
was there a difference between chronic back pain and controls. Age had a significant
(p=0.0004) but weak (β=-0.01) relationship with global average PAF. There was no significant
relationship between PAF and depression (Table 5.2).
Inspection of the normalized, averaged power spectra for each group (Figure 5.2)
revealed different general patterns of alpha activity across the three groups, particularly
between the widespread and back pain groups. As shown in Figure 5.1, the widespread group
spectra suggest an upshifted alpha compared to the back pain group spectra, with the controls
falling between the two. When comparing group means across all 64 electrodes (Figure 5.3), a
similar pattern was observed across the scalp, with the back pain group having lower PAF
values than the chronic widespread pain group and healthy controls falling in the middle. The
range of differences between back pain and widespread pain electrode means was 0.15-0.51
for PAF values. Areas of maximal difference between the two groups were generally across
occipital, parietal occipital, and temporal electrodes (Figure 5.3).
Figure 5.3 Distribution of the group differences in PAF at each channel: a. Back pain - controls, b. Widespread pain -
controls, c. widespread pain - back pain. The comparison of greatest difference is in the widespread pain - back pain
comparison (b.), with a particular emphasis on the posterior electrodes. The range of differences between back pain
and widespread pain electrode means was 0.15-0.51 Hz.

123
We ran a PCA across the PAF values at every electrode for all participants, treating
each participant as an observation and each electrode as a variable, and found that a single
principal component explained 88.84% of the variance across all three groups of participants.
This component was highly correlated with global average PAF values across all participants (r=
0.923). In running the regression at each electrode, no electrode survived the correction for
multiple comparisons. In the analysis of pain intensity in the patient subgroups, we found no
significant relationship between pain intensity at time of EEG collection and PAF values.
Table 5.2 Regression table for the models run on the global PAF values. diagnosis_ indicates which two conditions
were contrasted (e.g. diagnosis_widespread-back is the comparison of back pain PAF values against chronic
widespread pain as the reference level). Sex_female represents the binary variable ‘sex’ when female is compared
against male as the reference level. BDI=Beck depression index.
5.3.2. Dataset two
5.3.2.1. Participant characteristics
The demographic information for patients is found in Table 5.3. The mean age,
depression scores, anxiety scores, and pain at time of collection were consistent across both
groups. The proportion of females to males was equal across both groups. The widespread pain
group was larger than the localized pain group (n=24 versus n=14, respectively).

124
Total n=38
Widespread pain
n=24
Localized pain
n=14
Age (years) 43.17±12.65 45.86±15.56
Sex (f/m) 12/12 7/7
Depression 11.54±6.04 11.86±5.87
Anxiety 13.79±4.80 13.43±5.85
Pain at time of
collection (0-10)
3.50±2.43 3±2.72
Table 5.3 Demographic information and clinical characteristics for the two groups in dataset two: patients with
widespread pain (3 or more painful body sites) and patients with localized pain (2 or fewer painful body sites).
5.3.2.2. Differences in PAF based on pain distribution
In contrast to the findings in dataset one, dataset two showed that patients with widespread pain
had lower average PAF than patients with more localized pain: mean, global average PAF for
each group (mean ± standard deviation) was 9.58 ± 0.71 for chronic widespread pain patients
and 10.25 ± 0.83 for patients with localized pain. There was a significant difference between
localized and widespread pain as a predictor of global average PAF after controlling for sex,
age, and depression (p=0.012). With localized pain as the reference category, widespread pain
had a β value of -0.69 Hz indicating that, holding all other variables constant, patients with
widespread pain had approximately 0.69 Hz lower global average PAF values than patients with
localized pain within the UCPPS diagnosis. As in dataset one, depression and sex did not have
a significant relationship with global average PAF, however, unlike in dataset one, age also did
not have a significant effect (Table 5.4).

125
Variable Estimate SE tStat pValue
Sex_Female -0.24 0.42 0.58 0.567
Age -0.01 0.01 -0.67 0.506
PainDistribution_Widespread -0.69 0.26 -2.66 0.012*
HADSD 0.01 0.04 0.27 0.785
Number of observations R^2 Adjusted R^2
38.00 0.22 0.124
Table 5.4. Regression table for the models run on global PAF values. PainDistribution_Widespread that the localized
pain group was the reference level for pain dsitrubtion. Sex_female represents the binary variable ‘sex’ when female
is compared against male as the reference level. HADSD= Hospital Anxiety and Depression Scale- Depression
Unlike in dataset one, we found a significant relationship between pain category
(localized versus widespread) at every electrode (Figure 5.4): patients with widespread pain
had lower PAF values than patients with localized pain across the scalp. The range of B values
across the electrodes with localized pain as the reference category was -
0.80 - -
0.60 for PAF
values. Areas of maximal difference between the two groups were generally across the
sensorimotor and occipital regions. In the analysis of pain intensity, as in dataset one, we found
no significant relationship between pain intensity at time of EEG collection and PAF values.

126
Figure 5.4 β values widespread pain versus localized pain as a predictor of PAF values in the UCPPS dataset after
controlling for age, sex, and depression. The relationship was significant (p<0.05) at every electrode.
5.4. Discussion
This study is the first to demonstrate differences in peak alpha frequency (PAF) across
distinct diagnoses in chronic pain. Additionally, it is the first to demonstrate differences in PAF
between those with distributed versus localized pain. Comparing people with chronic back pain,
chronic widespread pain, and healthy controls, we found that the largest group difference is
between the two etiologically distinct patient groups, with chronic back pain patients downshifted
from healthy controls, and chronic widespread pain upshifted from healthy controls (Figure 5.1).
Separately, we found that in a single diagnostic group, patients with multisite pain, a feature
identified as a potential marker of centralized pain in Chapter 4, had lower PAF values than
those with chronic back pain (Figure 5.4). PAF holds strong promise as a neural marker due to
its stability, noninvasive capture, and well-documented association with acute pain sensitivity.
While our understanding of the connection between PAF and chronic pain remains incomplete,
the findings of this study represent an important step toward disentangling this relationship.
The findings across previous studies comparing PAF between people with chronic pain
and healthy controls are variable. A recent review of the literature comparing EEG/MEG

127
biomarkers for chronic pain reported that findings were a mix of decreased PAF in chronic pain,
increased PAF in chronic pain, and no significant differences compared to healthy controls
(Zebhauser et al., 2023). An earlier review found that, in the 12 papers comparing patients with
chronic pain to pain-free controls, eight found lower PAF in patients compared to controls, and
one found a positive relationship between PAF and pain intensity in people with major
depressive disorder and comorbid chronic pain. The remaining 3 papers reported no significant
relationship (McLain et al., 2022).
In past analyses and comparisons across papers, it is plausible to suggest that
diagnosis-specific shifts in PAF could have ‘washed out’ group-level differences. Indeed, our
findings may partially explain the inconsistencies summarized in the 2023 review of PAF and
chronic pain (Zebhauser et al., 2023), as the majority of papers finding no relationship were
conducted in populations of mixed pain types and diagnoses. The findings of this study highlight
the need to consider individual diagnoses in analyses of chronic pain. Further, the disparate
changes in PAF observed between diagnoses in the present paper align with the variable
clinical presentations across chronic pain conditions. For example, people with widespread pain
experience higher rates of chronic overlapping pain conditions, thought to be maintained by
central sensitization (Maixner et al. 2016; Aaron et al. 2000), have worse psychosomatic
symptoms (Viniol et al., 2013), and higher pain sensitivity (Mingorance et al., 2021) compared to
individuals with chronic back pain. Chronic widespread pain is associated with a broader
distribution and more steady presentation of pain than chronic back pain, which is often
characterized by intermittent exacerbations of relatively (though not exclusively) localized pain
(Viniol et al., 2013).
Chronic back pain and chronic widespread pain likely involve different levels of pain
centralization. Chronic back pain generally has more readily identifiable peripheral contributors
(Allegri et al., 2016; Mosabbir, 2022; Vora et al., 2010) than chronic widespread pain which,
apart from a degree of peripheral inflammation (Gerdle et al., 2024; Siracusa et al., 2021; Staud

128
and Smitherman, 2002), is thought to be centrally mediated (Sluka and Clauw 2016; Goubert et
al. 2017). While both conditions have demonstrated differences in brain function that distinguish
them from controls (Li et al. 2021; Bäckryd et al. 2017; Mosch et al. 2023; Napadow et al. 2010),
there is strong evidence that, in individuals with chronic widespread pain, differences in central
nervous system function underlie key clinical features such as pain distribution (Kaplan et al.
2022; Kutch et al. 2017). In a direct comparison of resting-state brain activity in chronic back
pain and chronic widespread pain, the two patient groups showed distinct patterns of
connectivity between the nucleus accumbens and the left ventral pallidum, left putamen, and
right caudate (Park et al. 2022). The differences in PAF observed in the present study could be
attributable to these differences in central nervous system involvement. Our previous work also
shows that PAF is inversely associated with activity in nodes of the salience network (McLain et
al., 2022) and our current findings show that patients with chronic back pain have significantly
lower PAF values than patients with chronic widespread pain. Our results may therefore indicate
that differences in salience network function could be driving opposite shifts in PAF and
resulting in inconsistent findings across analyses that do not account for differences in
diagnosis.
Previous literature on PAF shifts in chronic back pain and chronic widespread pain is
relatively limited. One other study of PAF in healthy controls compared to people with chronic
back pain found no significant differences between groups (Schmidt et al., 2012). Our findings in
chronic widespread pain, however, contrast with one earlier paper that reported lower PAF
values in patients with chronic widespread pain [fibromyalgia] than in healthy controls (Lim et
al., 2016). There are some important differences between the study populations and collection
methods in the previous study and the current examination of chronic widespread pain. For
example, the controls in the prior study had higher reported PAF values (10.3 ± 0.8 Hz) than
those in the present study (9.92 ± 0.52). Individuals with chronic widespread pain in the
previous study also demonstrated lower PAF (9.6 ± 0.6 Hz) than the individuals with chronic

129
widespread pain in our study (10.09 ± 0.37). The population of the prior study was entirely
female, while the present study is a mix of males and females (5 and 25, respectively). Further,
the collection in the prior study was performed with magnetoencephalography as opposed to the
EEG data examined in the present study. Given this variability, additional investigation is
warranted in a population with a more balanced sex distribution.
Looking at a single diagnostic group from a separate study, we found that differences in
PAF were also associated with the presence or absence of multisite pain, a hypothesized
marker of centralized pain conditions. We ran this analysis in a population of UCPPS subjects
subtyped into localized versus multisite pain, where scores for sex, age, depression, and
anxiety were balanced across the two groups (Table 5.3). While this aligns with the broader
findings that PAF shifts can be linked to heterogeneity in chronic pain populations and may be
linked to the level of pain centralization, in this analysis, PAF was lower for individuals with
widespread pain. Therefore, the shifts the analysis of a chronic pain condition characterized by
widespread pain versus chronic back pain were opposite the shifts found in the analysis of a
single diagnostic category subtyped by presence or absence of chronic pain. The findings from
the second analysis in UCPPS are more in line with what has previously been suggested by the
literature: low PAF may be specific to pain conditions involving sensitization or neuropathic pain
(Fauchon et al. 2022; Boord et al. 2007; Di Pietro et al. 2018; Kim et al. 2019). Without more
clinical and procedural information about the patients and EEG data examined in the dataset
one, it is difficult to specify what these opposite findings might indicate.
The primary limitation of this paper is our lack of information on chronic pain duration
and clinical presentation. Pain duration has been negatively associated with PAF (de Vries et
al., 2013) and while all patients included in this study had pain for at least 6 months, we were
unable to assess whether the total duration was balanced across the groups. The limited
information available regarding individual patient characteristics also means that we were
unable to determine how phenotypically distinct the chronic back pain and chronic widespread

130
samples in this analysis truly were. Diagnostic criteria were not provided for chronic widespread
pain nor chronic back pain, so it is plausible that there may have been overlap in clinical
presentations across groups. Previous research indicates that 12-49% of patients with chronic
back pain have a comorbid diagnosis of chronic widespread pain (Hüppe et al. 2004; Aaron et
al. 2000; Maixner et al. 2016; Aoyagi et al. 2019). Without more information about individual
patient characteristics, we cannot say for certain the amount of overlap in clinical characteristics
associated with centralized pain in this specific sample. Additionally, as this is a secondary
analysis, it is possible that there were inconsistencies in the collection across conditions of
which we are unaware. All EEG data were collected at the Technical University of Munich, but it
is unclear if the site and collection equipment were consistent across patient/control groups
(May et al., 2021; Ta Dinh et al., 2019). All of these factors may have contributed to the
inconsistency in the results from datasets one and two. However, the results still suggest that a
focus on PAF association with patient heterogeneity, rather than patient control differences
alone, is warranted.
Variability in chronic pain presentation and etiology often contributes to a trial-and-error
approach when assigning treatments. If, as suggested by this work and has been previously
hypothesized (Fauchon et al., 2022), shifts in PAF provide information about the type of pain a
person is experiencing, this measure could aid in delivering individualized treatments. This
would particularly be the case if PAF shifts with varying levels of pain centralization. PAF is
stable within individuals over weeks and months (Furman et al., 2019, 2020; Grandy et al.,
2013), and highly heritable (van Beijsterveldt and van Baal, 2002), making it an ideal candidate
for a neural marker in chronic pain. Further research is needed to understand if PAF will shift in
response to treatments for chronic pain, but there is some early evidence that it might change in
response to successful interventions (Ngernyam et al., 2015; Parker et al., 2021; Sato et al.,
2017; Sufianov et al., 2014).

131
This study provides novel insight regarding the need to consider diagnosis and markers
for centralized pain when examining shifts in PAF. Our findings suggest that shifts in PAF are
not likely a generalized marker for chronic pain (Fauchon et al., 2022), but instead exhibit
diagnosis and, perhaps, pain centralization-specific changes. Here we find a significant
difference in PAF values between patients with chronic back pain and patients with chronic
widespread pain, as well as between UCPPS patients with widespread versus localized pain
after controlling for age, sex, and depression. Future research should seek to compare PAF
values across multiple diagnoses while also incorporating measures of widespread pain to
better understand what aspect of chronic pain most influences shifts in PAF and the potential
role that centralization plays.
Chapter 6. Discussion
6.1. Findings from our work
Chronic pain is impactful and growing in prevalence, but it is not a homogenous
experience. Just as every patient has a unique life and set of circumstances, so too are their
experiences with chronic pain composed of variable features that create distinct clinical profiles
and patterns of neural activity. This work aimed to identify those clinical features and patterns of
brain activity that could disambiguate chronic pain states primarily maintained by peripheral
versus central drivers. Specifically, we sought to identify an electroencephalography (EEG)
measure that differed between peripheral versus centralized pain types and clinical features that
significantly modified the relationship between brain activity and self-reported pain levels. Are
there changes in specific brain networks or scores on clinical questionnaires (depression,
anxiety, pain catastrophizing, neuropathic pain likelihood, and pain distribution) that underlie the
individual variability in the experience of chronic pain? We propose that being able to subtype

132
pain patients based on pain source will present a way forward in personalizing treatment plans
and improving patient outcomes.
In Chapter 3, we investigated the stability and neural correlates of a potential EEG
marker for chronic pain states, peak alpha frequency (PAF). A recent review of the literature
comparing EEG/MEG biomarkers for chronic pain reported that findings were a mix of
decreased PAF in chronic pain, increased PAF in chronic pain, and no significant differences
compared to healthy controls (Zebhauser et al., 2023). However, these papers examined a
range of chronic pain conditions, with some studies combining multiple diagnoses into a single
“chronic pain” group to compare against controls. It is plausible to suggest that diagnosisspecific shifts in PAF could have ‘washed out’ group-level differences, in which case the
disparate findings could be more of a “feature than a bug,” as PAF could be used to subtype
pain patients. However, before addressing this possibility, we first needed to ensure that the
differences in findings weren’t solely due to the widely variable processing pipelines used
across the PAF-pain literature. By reviewing the literature and compiling the most common
differences in processing, we found PAF to be extremely robust against a variety of processing
decisions. Additionally, we found that global average PAF values capture the majority of the
variance between individuals (>95%), and that global average PAF values are associated with
activity in the salience network. PAF is stable within individuals over weeks and months
(Furman et al., 2019, 2020; Grandy et al., 2013), and highly heritable (van Beijsterveldt and van
Baal, 2002), which, paired with the analytic stability and association with pain-relevant cortical
areas from Chapter 3, make it an ideal candidate for a neural marker in chronic pain.
In Chapter 4, we examined the neural correlates of variations in ecologically valid
persistent pain both within and outside the traditional pain matrix using fMRI data in a population
of urologic chronic pelvic pain patients (n=492). We also investigated whether this relationship
was modified by putative clinical markers of centralized pain. We found that, within the pain
network, the relationship between pain and brain activity varied based on the timescale of pain

133
being examined and state of the fMRI collection (empty versus full bladder). Pain at the time of
scan is more associated with functional connectivity during provoked, naturalistic pain (full
bladder), and pain in the past week is more associated with functional connectivity during nonprovoked/resting-state levels of chronic pain (empty bladder). We additionally found that anxiety
was a strong modifier of the relationship between empty bladder functional connectivity and pain
over the past week, while multisite pain was a strong modifier of the relationship between full
bladder functional connectivity and pain at the time of scan. When the analysis was expanded to
the whole brain, we only found a significant relationship between full bladder scans and pain (at
time of scan and, to a lesser extent, pain in the past week). The significant networks extended
beyond that of the traditional pain matrix, highlighting, in particular, the salience network, more
expansive regions of the somatosensory/motor network, and multiple nodes in the prefrontal
cortex/default mode network. We additionally found that when combining empty and full bladder
scans to capture the variation in brain activity that a patient might experience on any given day,
there were significant relationships between average pain in the last week and the functional
connectivity between and within salience, somatosensory/motor, and default mode network.
Further, this relationship was only significantly modified by multisite pain. Taken together, these
results suggest nodes both within and outside of the pain network contribute to self-reported
pain levels in UCPPS, and that this relationship is modified by widespread pain and, to a lesser
extent, anxiety. As in Chapter 3, we saw significant contributions from the salience network to
the experience of chronic pain.
In Chapter 5, we investigated whether PAF, which we associated with salience network
activity in Chapter 3, was different across diagnostic categories and clinical profiles. Specifically,
we compared PAF values in chronic back pain and chronic widespread pain, as well as in a
single population of UCPPS patients categorized by the primary clinical measure of interest
from Chapter 4: pain distribution/multisite pain. We found that PAF was indeed different
between pain groups (chronic widespread pain and chronic back pain), and had a stronger

134
relationship with heterogeneity in chronic pain patients than with patient, control differences. We
additionally found that within a single diagnostic category for chronic pain, UCPPS, PAF was
different amongst patients with widespread versus localized pain. Combined with the findings in
Chapters 3 and 4, this analysis indicates that PAF is not likely a generalized marker for chronic
pain (Fauchon et al., 2022), but instead exhibits diagnosis and potentially etiology-specific
changes that may be linked to the level of centralization.
6.2. Impact of Dissertation
Chronic pain has been identified as a global research priority, with an estimated
worldwide prevalence of 10-25% (Goldberg and McGee, 2011). Given this impact, there is
growing interest in identifying objective markers that can shed light on pain etiology and guide
the development of intervention and prevention strategies. However, there are very few
biomarkers developed/validated to the point of use in clinical practice for pain (C.-W. Woo et al.,
2017). Neuroimaging studies in the last 20 years have made significant progress in
understanding the brain circuitry that encodes pain (Martucci and Mackey 2018) but many
studies have focused on the blood-oxygen-level-dependent (BOLD) response to acutely applied
painful stimuli (Xu et al. 2021; Xu et al. 2020) as it is difficult to study naturalistic fluctuations in
chronic pain states. A large body of research has considered how resting-state functional
connectivity between BOLD signals differs between healthy individuals and those with chronic
pain (van der Miesen et al., 2019), but this presents the previously articulated issue of treating
chronic pain as a single, homogenous group. This is particularly problematic when variability in
chronic pain presentation and etiology often contributes to a trial-and-error approach when
assigning treatments.
This work is the first study to integrate fMRI, EEG, and clinical questionnaire data to find
a classifying marker of centralized pain. While there is a large body of literature investigating
mechanisms of pain in EEG, MRI, and clinical questionnaires individually, there has yet to be a

135
comprehensive study integrating the three in order to identify and understand the underlying
brain changes behind centralized pain markers. PAF is a marker of great interest in the pain
field, but the relationship between PAF and pain type has never been directly assessed. By
associating PAF with clinical markers for centralized pain identified with fMRI analysis in a large,
longitudinal data set, we have provided evidence for its potential in identifying patients with
centralized pain. While more research is needed to understand if PAF will shift in response to
treatments for chronic pain, there is some early evidence that it might change in response to
successful interventions (Ngernyam et al., 2015; Parker et al., 2021; Sato et al., 2017; Sufianov
et al., 2014), making it of particular clinical relevance.
6.3. Limitations and future work
This work highlights the need to dedicate more attention to heterogeneity in populations
of patients with chronic pain. While identifying differences between patients and healthy controls
is important, when it comes to directing treatments, identifying the underlying causes of
heterogeneity in patients with chronic pain may be more useful. An individual can typically
identify and communicate that they have been in pain for longer than three months. They,
however, likely cannot identify whether their pain stems more from peripheral sensitivity/damage
or central nervous system changes, an important factor to consider when designing a treatment
plan. While our work addresses this central issue, some limitations must be considered.
Across the three chapters, we lacked a dataset that had all three metrics of interest: fMRI, EEG,
and clinical questionnaires. While the work performed in each chapter can be synthesized to
push forward our understanding of centralized versus peripheral pain types, ideally all three
measures could be examined in a single data set. Direct association of differences in PAF with
changes in fMRI data within a chronic pain population would be particularly useful. Our work
suggests that activity in the salience network may be important in determining an individual’s

136
PAF value, but a direct analysis would be helpful for better targeting future neural interventions
such as repetitive transcranial magnetic stimulation (rTMS).
While we identified PAF as being robust against most processing decisions in Chapter 3,
using data sets from multiple institutions in a single analysis still runs the risk of unknown
confounds associated with differences in collection procedures and equipment. This is perhaps
best evidenced by Chapter 5, in which both datasets suggest PAF varies across chronic pain
groups but appears to be shifted in opposite directions. Without a replicated analysis in a larger
group of patients where the collection parameters are known to be consistent and the outcome
of interest is the difference between pain subtypes, it is impossible to know whether these
disparate findings are the result of another, unconsidered feature of the chronic pain groups, or
something site-specific.
In the same vein, this work had only a small data set (n=38) to examine the relationship
between PAF and pain distribution, a clinical marker identified in Chapter 4 as potentially
modifying the relationship between brain activity and perception of naturalistic fluctuations in
chronic pain. Future research should examine larger groups of chronic pain patients with EEG
data and measures of pain distribution so both variables can be entered into the same model.
This would allow a better understanding of whether PAF is truly an indicator of centralized pain,
or if something more specific to diagnostic boundaries is resulting in the shift. Based on
previous findings, it seems unlikely that PAF shifts are specifically associated with any one
diagnosis (Fauchon et al., 2022), and that it is more likely linked to pain source (Schmidt et al.
2012; Kim et al. 2019; Vuckovic et al. 2014), but this is speculative until a direct analysis of both
predictors is performed.

137
References
Aaron, L. A., Burke, M. M., & Buchwald, D. (2000). Overlapping conditions among patients with
chronic fatigue syndrome, fibromyalgia, and temporomandibular disorder. Archives of
Internal Medicine, 160(2), 221–227.
Ab Aziz, C. B., & Ahmad, A. H. (2006). The role of the thalamus in modulating pain. The
Malaysian Journal of Medical Sciences: MJMS, 13(2), 11–18.
Aiello, M., Salvatore, E., Cachia, A., Pappatà, S., Cavaliere, C., Prinster, A., Nicolai, E.,
Salvatore, M., Baron, J.-C., & Quarantelli, M. (2015). Relationship between
simultaneously acquired resting-state regional cerebral glucose metabolism and
functional MRI: a PET/MR hybrid scanner study. NeuroImage, 113, 111–121.
Allegri, M., Montella, S., Salici, F., Valente, A., Marchesini, M., Compagnone, C., Baciarello, M.,
Manferdini, M. E., & Fanelli, G. (2016). Mechanisms of low back pain: a guide for
diagnosis and therapy. F1000Research, 5.
https://doi.org/10.12688/f1000research.8105.2
Alshelh, Z., Brusaferri, L., Saha, A., Morrissey, E., Knight, P., Kim, M., Zhang, Y., Hooker, J. M.,
Albrecht, D., Torrado-Carvajal, A., Placzek, M. S., Akeju, O., Price, J., Edwards, R. R.,
Lee, J., Sclocco, R., Catana, C., Napadow, V., & Loggia, M. L. (2022). Neuroimmune
signatures in chronic low back pain subtypes. Brain: A Journal of Neurology, 145(3),
1098–1110.
Alter, B. J., Anderson, N. P., Gillman, A. G., Yin, Q., Jeong, J.-H., & Wasan, A. D. (2021).
Hierarchical clustering by patient-reported pain distribution alone identifies distinct
chronic pain subgroups differing by pain intensity, quality, and clinical outcomes. PloS
One, 16(8), e0254862.
Andrews, P., Steultjens, M., & Riskowski, J. (2018). Chronic widespread pain prevalence in the
general population: A systematic review. European Journal of Pain , 22(1), 5–18.
Angelakis, E., Lubar, J. F., Stathopoulou, S., & Kounios, J. (2004). Peak alpha frequency: an
electroencephalographic measure of cognitive preparedness. Clinical Neurophysiology:
Official Journal of the International Federation of Clinical Neurophysiology, 115(4), 887–
897.
Aoyagi, K., He, J., Nicol, A. L., Clauw, D. J., Kluding, P. M., Jernigan, S., & Sharma, N. K.
(2019). A Subgroup of Chronic Low Back Pain Patients With Central Sensitization. The
Clinical Journal of Pain, 35(11), 869–879.
Apkarian, A. V., Bushnell, M. C., Treede, R.-D., & Zubieta, J.-K. (2005). Human brain
mechanisms of pain perception and regulation in health and disease. European Journal
of Pain , 9(4), 463–484.
Arns, M., Drinkenburg, W. H., Fitzgerald, P. B., & Kenemans, J. L. (2012). Neurophysiological
predictors of non-response to rTMS in depression. Brain Stimulation, 5(4), 569–576.
Bäckryd, E., Tanum, L., Lind, A.-L., Larsson, A., & Gordh, T. (2017). Evidence of both systemic
inflammation and neuroinflammation in fibromyalgia patients, as assessed by a multiplex

138
protein panel applied to the cerebrospinal fluid and to plasma. Journal of Pain Research,
10, 515–525.
Baliki, M. N., Chialvo, D. R., Geha, P. Y., Levy, R. M., Harden, R. N., Parrish, T. B., & Apkarian,
A. V. (2006). Chronic pain and the emotional brain: specific brain activity associated with
spontaneous fluctuations of intensity of chronic back pain. The Journal of Neuroscience:
The Official Journal of the Society for Neuroscience, 26(47), 12165–12173.
Baliki, M. N., Mansour, A. R., Baria, A. T., & Apkarian, A. V. (2014). Functional reorganization of
the default mode network across chronic pain conditions. PloS One, 9(9), e106133.
Basbaum, A. I., & Fields, H. L. (1978). Endogenous pain control mechanisms: review and
hypothesis. Annals of Neurology, 4(5), 451–462.
Beck, A. T. (1996). Manual for the beck depression inventory-II. (No Title).
https://cir.nii.ac.jp/crid/1370564063990947215
Behzadi, Y., Restom, K., Liau, J., & Liu, T. T. (2007). A component based noise correction
method (CompCor) for BOLD and perfusion based fMRI. NeuroImage, 37(1), 90–101.
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: A practical and
powerful approach to multiple testing. Journal of the Royal Statistical Society, 57(1),
289–300.
Bergbom, S., Boersma, K., Overmeer, T., & Linton, S. J. (2011). Relationship among pain
catastrophizing, depressed mood, and outcomes across physical therapy treatments.
Physical Therapy, 91(5), 754–764.
Berger, H. (1929). Über das Elektrenkephalogramm des Menschen. European Archives of
Psychiatry and Neurological Sciences, 87, 527–570.
Berthelot, J.-M., Biha, N., Darrieutort-Laffite, C., Le Goff, B., & Maugars, Y. (2019). Are
painDETECT scores in musculoskeletal disorders associated with duration of daily pain
and time elapsed since current pain onset? Pain Reports, 4(3), e739.
Bigdely-Shamlo, N., Mullen, T., Kothe, C., Su, K.-M., & Robbins, K. A. (2015). The PREP
pipeline: standardized preprocessing for large-scale EEG analysis. Frontiers in
Neuroinformatics, 9, 16.
Birn, R. M., Molloy, E. K., Patriat, R., Parker, T., Meier, T. B., Kirk, G. R., Nair, V. A., Meyerand,
M. E., & Prabhakaran, V. (2013). The effect of scan length on the reliability of restingstate fMRI connectivity estimates. NeuroImage, 83, 550–558.
Bjørk, M. H., Stovner, L. J., Nilsen, B. M., Stjern, M., Hagen, K., & Sand, T. (2009). The occipital
alpha rhythm related to the “migraine cycle” and headache burden: A blinded, controlled
longitudinal study. Clinical Neurophysiology: Official Journal of the International
Federation of Clinical Neurophysiology, 120(3), 464–471.
Boord, P., Siddall, P. J., Tran, Y., Herbert, D., Middleton, J., & Craig, A. (2007).
Electroencephalographic slowing and reduced reactivity in neuropathic pain following
spinal cord injury. Spinal Cord, 46(2), 118–123.

139
Borchers, A. T., & Gershwin, M. E. (2015). Fibromyalgia: A Critical and Comprehensive Review.
Clinical Reviews in Allergy & Immunology, 49(2), 100–151.
Borsook, D., Edwards, R., Elman, I., Becerra, L., & Levine, J. (2013). Pain and Analgesia: The
Value of Salience Circuits. Progress in Neurobiology, 104, 93.
Bosma, R. L., Kim, J. A., Cheng, J. C., Rogachov, A., Hemington, K. S., Osborne, N. R., Oh, J.,
& Davis, K. D. (2018). Dynamic pain connectome functional connectivity and oscillations
reflect multiple sclerosis pain. Pain, 159(11), 2267–2276.
Bradley, C. S., Gallop, R., Sutcliffe, S., Kreder, K. J., Lai, H. H., Clemens, J. Q., Naliboff, B. D.,
& Multidisciplinary Approach to the Study of Chronic Pelvic Pain (MAPP) Research
Network. (2022). Long-Term Symptom Trajectories in Urologic Chronic Pelvic Pain
Syndrome: A MAPP Research Network Study. Urology, 169, 58–64.
Brötzner, C. P., Klimesch, W., Doppelmayr, M., Zauner, A., & Kerschbaum, H. H. (2014).
Resting state alpha frequency is associated with menstrual cycle phase, estradiol and
use of oral contraceptives. Brain Research, 1577, 36–44.
Brummett, C. M., Janda, A. M., Schueller, C. M., Tsodikov, A., Morris, M., Williams, D. A., &
Clauw, D. J. (2013). Survey criteria for fibromyalgia independently predict increased
postoperative opioid consumption after lower-extremity joint arthroplasty: a prospective,
observational cohort study. Anesthesiology, 119(6), 1434–1443.
Brummett, C. M., Urquhart, A. G., Hassett, A. L., Tsodikov, A., Hallstrom, B. R., Wood, N. I.,
Williams, D. A., & Clauw, D. J. (2015). Characteristics of fibromyalgia independently
predict poorer long-term analgesic outcomes following total knee and hip arthroplasty.
Arthritis & Rheumatology (Hoboken, N.J.), 67(5), 1386–1394.
Cagnie, B., Coppieters, I., Denecker, S., Six, J., Danneels, L., & Meeus, M. (2014). Central
sensitization in fibromyalgia? A systematic review on structural and functional brain MRI.
Seminars in Arthritis and Rheumatism, 44(1), 68–75.
Cashin, A. G., Traeger, A. C., Hübscher, M., Moseley, G. L., di Pietro, F., Parkitny, L., &
McAuley, J. H. (2019). Persistent Pain After Wrist or Hand Fracture: Development and
Validation of a Prognostic Model. The Journal of Orthopaedic and Sports Physical
Therapy, 49(1), 28–35.
Castro-Lopes, J. M., Tavares, I., & Coimbra, A. (1993). GABA decreases in the spinal cord
dorsal horn after peripheral neurectomy. Brain Research, 620(2), 287–291.
Čeko, M., Frangos, E., Gracely, J., Richards, E., Wang, B., Schweinhardt, P., & Catherine
Bushnell, M. (2020). Default mode network changes in fibromyalgia patients are largely
dependent on current clinical pain. NeuroImage, 216, 116877.
Chiang, A. K. I., Rennie, C. J., Robinson, P. A., van Albada, S. J., & Kerr, C. C. (2011). Age
trends and sex differences of alpha rhythms including split alpha peaks. Clinical
Neurophysiology: Official Journal of the International Federation of Clinical
Neurophysiology, 122(8), 1505–1517.

140
Chowdhury, N. S., Skippen, P., Si, E., Chiang, A. K. I., Millard, S. K., Furman, A. J., Chen, S.,
Schabrun, S. M., & Seminowicz, D. A. (2023). The reliability of two prospective cortical
biomarkers for pain: EEG peak alpha frequency and TMS corticomotor excitability.
Journal of Neuroscience Methods, 385, 109766.
Cioffi, C. L. (2018). Modulation of Glycine-Mediated Spinal Neurotransmission for the Treatment
of Chronic Pain. Journal of Medicinal Chemistry, 61(7), 2652–2679.
Clauw, D. J. (2014). Fibromyalgia: a clinical review. JAMA: The Journal of the American Medical
Association, 311(15), 1547–1555.
Clemens, J. Q., Kutch, J. J., Mayer, E. A., Naliboff, B. D., Rodriguez, L. V., Klumpp, D. J.,
Schaeffer, A. J., Kreder, K. J., Clauw, D. J., Harte, S. E., Schrepf, A. D., Williams, D. A.,
Andriole, G. L., Lai, H. H., Buchwald, D., Lucia, M. S., van Bokhoven, A., Mackey, S.,
Moldwin, R. M., … Landis, J. R. (2020). The Multidisciplinary Approach to The Study of
Chronic Pelvic Pain (MAPP) Research Network*: Design and Implementation of the
Symptom Patterns Study (SPS). Neurourology and Urodynamics, 39(6), 1803.
Cohen, D., & Cuffin, B. N. (1983). Demonstration of useful differences between
magnetoencephalogram and electroencephalogram. Electroencephalography and
Clinical Neurophysiology, 56(1), 38–51.
Cohen, S. P., Vase, L., & Hooten, W. M. (2021). Chronic pain: an update on burden, best
practices, and new advances. The Lancet, 397(10289), 2082–2097.
Conti, Y., Vatine, J.-J., Levy, S., Levin Meltz, Y., Hamdan, S., & Elkana, O. (2020). Pain
Catastrophizing Mediates the Association Between Mindfulness and Psychological
Distress in Chronic Pain Syndrome. Pain Practice: The Official Journal of World Institute
of Pain, 20(7), 714–723.
Corcoran, A. W., Alday, P. M., Schlesewsky, M., & Bornkessel-Schlesewsky, I. (2018). Toward
a reliable, automated method of individual alpha frequency (IAF) quantification.
Psychophysiology, 55(7), e13064.
Corlier, J., Carpenter, L. L., Wilson, A. C., Tirrell, E., Gobin, A. P., Kavanaugh, B., & Leuchter,
A. F. (2019). The relationship between individual alpha peak frequency and clinical
outcome with repetitive Transcranial Magnetic Stimulation (rTMS) treatment of Major
Depressive Disorder (MDD). Brain Stimulation, 12(6), 1572–1578.
Corlier, J., Tadayonnejad, R., Wilson, A. C., Lee, J. C., Marder, K. G., Ginder, N. D., Wilke, S.
A., Levitt, J., Krantz, D., & Leuchter, A. F. (2021). Repetitive transcranial magnetic
stimulation treatment of major depressive disorder and comorbid chronic pain: response
rates and neurophysiologic biomarkers. Psychological Medicine, 1–10.
Cragg, L., Kovacevic, N., McIntosh, A. R., Poulsen, C., Martinu, K., Leonard, G., & Paus, T.
(2011). Maturation of EEG power spectra in early adolescence: a longitudinal study.
Developmental Science, 14(5), 935–943.
Cuffin, B. N., & Cohen, D. (1979). Comparison of the magnetoencephalogram and
electroencephalogram. Electroencephalography and Clinical Neurophysiology, 47(2),
132–146.

141
Dahlhamer, J., Lucas, J., Zelaya, C., Nahin, R., Mackey, S., DeBar, L., Kerns, R., Von Korff, M.,
Porter, L., & Helmick, C. (2018). Prevalence of Chronic Pain and High-Impact Chronic
Pain Among Adults - United States, 2016. MMWR. Morbidity and Mortality Weekly
Report, 67(36), 1001–1006.
da Silva, J. T., & Seminowicz, D. A. (2019). Neuroimaging of pain in animal models: a review of
recent literature. Pain Reports, 4(4), e732.
da Silva, F. H. L., Vos, J. E., Mooibroek, J., & van Rotterdam, A. (1980). Relative contributions
of intracortical and thalamo-cortical processes in the generation of alpha rhythms,
revealed by partial coherence analysis. In Electroencephalography and Clinical
Neurophysiology (Vol. 50, Issues 5-6, pp. 449–456). https://doi.org/10.1016/0013-
4694(80)90011-5
Davis, K. D., Flor, H., Greely, H. T., Iannetti, G. D., Mackey, S., Ploner, M., Pustilnik, A., Tracey,
I., Treede, R.-D., & Wager, T. D. (2017). Brain imaging tests for chronic pain: medical,
legal and ethical issues and recommendations. Nature Reviews. Neurology, 13(10),
624–638.
De Ridder, D., Vanneste, S., Smith, M., & Adhia, D. (2022). Pain and the Triple Network Model.
Frontiers in Neurology, 13. https://doi.org/10.3389/fneur.2022.757241
de Vries, M., Wilder-Smith, O. H., Jongsma, M. L. A., van den Broeke, E. N., Arns, M., van
Goor, H., & van Rijn, C. M. (2013). Altered resting state EEG in chronic pancreatitis
patients: toward a marker for chronic pain. Journal of Pain Research, 6, 815–824.
Di Pietro, F., Macey, P. M., Rae, C. D., Alshelh, Z., Macefield, V. G., Vickers, E. R., &
Henderson, L. A. (2018). The relationship between thalamic GABA content and resting
cortical rhythm in neuropathic pain. Human Brain Mapping, 39(5), 1945–1956.
Dydyk, A. M., & Givler, A. (2022). Central Pain Syndrome. In StatPearls. StatPearls Publishing.
Edwards, R. R., Cahalan, C., Mensing, G., Smith, M., & Haythornthwaite, J. A. (2011). Pain,
catastrophizing, and depression in the rheumatic diseases. Nature Reviews.
Rheumatology, 7(4), 216–224.
Edwards, R. R., Dworkin, R. H., Turk, D. C., Angst, M. S., Dionne, R., Freeman, R., Hansson,
P., Haroutounian, S., Arendt-Nielsen, L., Attal, N., Baron, R., Brell, J., Bujanover, S.,
Burke, L. B., Carr, D., Chappell, A. S., Cowan, P., Etropolski, M., Fillingim, R. B., …
Yarnitsky, D. (2016). Patient phenotyping in clinical trials of chronic pain treatments:
IMMPACT recommendations. Pain, 157(9), 1851–1871.
Eklund, A., Nichols, T. E., & Knutsson, H. (2016). Cluster failure: Why fMRI inferences for
spatial extent have inflated false-positive rates. Proceedings of the National Academy of
Sciences of the United States of America, 113(28), 7900–7905.
Eller-Smith, O. C., Nicol, A. L., & Christianson, J. A. (2018). Potential Mechanisms Underlying
Centralized Pain and Emerging Therapeutic Interventions. Frontiers in Cellular
Neuroscience, 12, 35.

142
Ellingsen, D.-M., Beissner, F., Moher Alsady, T., Lazaridou, A., Paschali, M., Berry, M., Isaro,
L., Grahl, A., Lee, J., Wasan, A. D., Edwards, R. R., & Napadow, V. (2021). A picture is
worth a thousand words: linking fibromyalgia pain widespreadness from digital pain
drawings with pain catastrophizing and brain cross-network connectivity. Pain, 162(5),
1352–1363.
Esteban, O., Markiewicz, C. J., Blair, R. W., Moodie, C. A., Isik, A. I., Erramuzpe, A., Kent, J. D.,
Goncalves, M., DuPre, E., Snyder, M., Oya, H., Ghosh, S. S., Wright, J., Durnez, J.,
Poldrack, R. A., & Gorgolewski, K. J. (2019). fMRIPrep: a robust preprocessing pipeline
for functional MRI. Nature Methods, 16(1), 111–116.
Esteban, O., Markiewicz, C. J., Burns, C., Goncalves, M., Jarecka, D., Ziegler, E., Berleant, S.,
Ellis, D. G., Pinsard, B., Madison, C., Waskom, M., Notter, M. P., Clark, D., ManhãesSavio, A., Clark, D., Jordan, K., Dayan, M., Halchenko, Y. O., Loney, F., … Ghosh, S.
(2022). nipy/nipype: 1.8.3. Zenodo. https://doi.org/10.5281/ZENODO.596855
Fauchon, C., Kim, J. A., El-Sayed, R., Osborne, N. R., Rogachov, A., Cheng, J. C., Hemington,
K. S., Bosma, R. L., Dunkley, B. T., Oh, J., Bhatia, A., Inman, R. D., & Davis, K. D.
(2022). Exploring sex differences in alpha brain activity as a potential neuromarker
associated with neuropathic pain. Pain, 163(7), 1291–1302.
Fan, L., Li, H., Zhuo, J., Zhang, Y., Wang, J., Chen, L., Yang, Z., Chu, C., Xie, S., Laird, A. R.,
Fox, P. T., Eickhoff, S. B., Yu, C., & Jiang, T. (2016). The Human Brainnetome Atlas: A
New Brain Atlas Based on Connectional Architecture. Cerebral Cortex , 26(8), 3508–
3526.
Fauchon, C., Kim, J. A., El-Sayed, R., Osborne, N. R., Rogachov, A., Cheng, J. C., Hemington,
K. S., Bosma, R. L., Dunkley, B. T., Oh, J., Bhatia, A., Inman, R. D., & Davis, K. D.
(2022). Exploring sex differences in alpha brain activity as a potential neuromarker
associated with neuropathic pain. Pain, 163(7), 1291–1302.
Ferguson, A. R., Huie, J. R., Crown, E. D., Baumbauer, K. M., Hook, M. A., Garraway, S. M.,
Lee, K. H., Hoy, K. C., & Grau, J. W. (2012). Maladaptive spinal plasticity opposes spinal
learning and recovery in spinal cord injury. Frontiers in Physiology, 3, 399.
Finnern, M. M., Kleinböhl, D., Flor, H., Benrath, J., & Hölzl, R. (2021). Differential sensory and
clinical phenotypes of patients with chronic widespread and regional musculoskeletal
pain. Pain, 162(1), 56–70.
Fló, A., Gennari, G., Benjamin, L., & Dehaene-Lambertz, G. (2022). Automated Pipeline for
Infants Continuous EEG (APICE): A flexible pipeline for developmental cognitive studies.
Developmental Cognitive Neuroscience, 54, 101077.
Freschl, J., Azizi, L. A., Balboa, L., Kaldy, Z., & Blaser, E. (2022). The development of peak
alpha frequency from infancy to adolescence and its role in visual temporal processing:
A meta-analysis. Developmental Cognitive Neuroscience, 57, 101146.
Freynhagen, R., Baron, R., Gockel, U., & Tölle, T. R. (2006). painDETECT: a new screening
questionnaire to identify neuropathic components in patients with back pain. Current
Medical Research and Opinion, 22(10), 1911–1920.

143
Fuchs, P. N., Peng, Y. B., Boyette-Davis, J. A., & Uhelski, M. L. (2014). The anterior cingulate
cortex and pain processing. Frontiers in Integrative Neuroscience, 8, 35.
Furman, A. J., Meeker, T. J., Rietschel, J. C., Yoo, S., Muthulingam, J., Prokhorenko, M.,
Keaser, M. L., Goodman, R. N., Mazaheri, A., & Seminowicz, D. A. (2018). Cerebral
peak alpha frequency predicts individual differences in pain sensitivity. NeuroImage,
167, 203–210.
Furman, A. J., Prokhorenko, M., Keaser, M. L., Zhang, J., Chen, S., Mazaheri, A., &
Seminowicz, D. A. (2020). Sensorimotor Peak Alpha Frequency Is a Reliable Biomarker
of Prolonged Pain Sensitivity. In Cerebral Cortex (Vol. 30, Issue 12, pp. 6069–6082).
https://doi.org/10.1093/cercor/bhaa124
Furman, A. J., Thapa, T., Summers, S. J., Cavaleri, R., Fogarty, J. S., Steiner, G. Z., Schabrun,
S. M., & Seminowicz, D. A. (2019). Cerebral peak alpha frequency reflects average pain
severity in a human model of sustained, musculoskeletal pain. In Journal of
Neurophysiology (Vol. 122, Issue 4, pp. 1784–1793).
https://doi.org/10.1152/jn.00279.2019
Gabard-Durnam, L. J., Mendez Leal, A. S., Wilkinson, C. L., & Levin, A. R. (2018). The Harvard
Automated Processing Pipeline for Electroencephalography (HAPPE): Standardized
Processing Software for Developmental and High-Artifact Data. Frontiers in
Neuroscience, 12, 97.
Georgopoulos, V., Akin-Akinyosoye, K., Zhang, W., McWilliams, D. F., Hendrick, P., & Walsh,
D. A. (2019). Quantitative sensory testing and predicting outcomes for musculoskeletal
pain, disability, and negative affect: a systematic review and meta-analysis. Pain, 160(9),
1920–1932.
Gerdle, B., Dahlqvist Leinhard, O., Lund, E., Lundberg, P., Forsgren, M. F., & Ghafouri, B.
(2024). Pain and the biochemistry of fibromyalgia: patterns of peripheral cytokines and
chemokines contribute to the differentiation between fibromyalgia and controls and are
associated with pain, fat infiltration and content. Frontiers in Pain Research (Lausanne,
Switzerland), 5, 1288024.
Ghadimi, M., & Sapra, A. (2023). Magnetic Resonance Imaging Contraindications. In
StatPearls. StatPearls Publishing.
Ghasemi, M., Behnaz, F., Hassanzad, N., & Taheri, F. (2022). Evaluating the Relationship
Between Depression and Pain Anxiety with Pain Catastrophizing in Patients with
COVID-19. Anesthesiology and Pain Medicine, 12(1).
https://doi.org/10.5812/aapm.119354
Goldberg, D. S., & McGee, S. J. (2011). Pain as a global public health priority. BMC Public
Health, 11, 770.
Goldenberg, D. L. (1992). Fibromyalgia, chronic fatigue, and myofascial pain syndromes. In
Current Opinion in Rheumatology (Vol. 4, Issue 2, pp. 247–258).
https://doi.org/10.1097/00002281-199204000-00019

144
Goldman, R. I., Stern, J. M., Engel, J., Jr, & Cohen, M. S. (2002). Simultaneous EEG and fMRI
of the alpha rhythm. In NeuroReport (pp. 2487–2492).
https://doi.org/10.1097/01.wnr.0000047685.08940.d0
Gordon, E. M., Chauvin, R. J., Van, A. N., Rajesh, A., Nielsen, A., Newbold, D. J., Lynch, C. J.,
Seider, N. A., Krimmel, S. R., Scheidter, K. M., Monk, J., Miller, R. L., Metoki, A.,
Montez, D. F., Zheng, A., Elbau, I., Madison, T., Nishino, T., Myers, M. J., …
Dosenbach, N. U. F. (2023). A somato-cognitive action network alternates with effector
regions in motor cortex. Nature, 617(7960), 351–359.
Gorgolewski, K., Burns, C. D., Madison, C., Clark, D., Halchenko, Y. O., Waskom, M. L., &
Ghosh, S. S. (2011). Nipype: a flexible, lightweight and extensible neuroimaging data
processing framework in python. Frontiers in Neuroinformatics, 5, 13.
Goubert, D., Danneels, L., Graven-Nielsen, T., Descheemaeker, F., & Meeus, M. (2017).
Differences in Pain Processing Between Patients with Chronic Low Back Pain,
Recurrent Low Back Pain, and Fibromyalgia. Pain Physician, 20(4), 307–318.
Grandy, T. H., Werkle-Bergner, M., Chicherio, C., Schmiedek, F., Lövdén, M., & Lindenberger,
U. (2013). Peak individual alpha frequency qualifies as a stable neurophysiological trait
marker in healthy younger and older adults. Psychophysiology, 50(6), 570–582.
Greenspan, J. D., Craft, R. M., LeResche, L., Arendt-Nielsen, L., Berkley, K. J., Fillingim, R. B.,
Gold, M. S., Holdcroft, A., Lautenbacher, S., Mayer, E. A., Mogil, J. S., Murphy, A. Z.,
Traub, R. J., & Consensus Working Group of the Sex, Gender, and Pain SIG of the
IASP. (2007). Studying sex and gender differences in pain and analgesia: a consensus
report. Pain, 132 Suppl 1(Suppl 1), S26–S45.
Greve, D. N., & Fischl, B. (2009). Accurate and robust brain image alignment using boundarybased registration. NeuroImage, 48(1), 63–72.
Griffith, J. W., Stephens-Shields, A. J., Hou, X., Naliboff, B. D., Pontari, M., Edwards, T. C.,
Williams, D. A., Clemens, J. Q., Afari, N., Tu, F., Lloyd, R. B., Patrick, D. L., Mullins, C.,
Kusek, J. W., Sutcliffe, S., Hong, B. A., Lai, H. H., Krieger, J. N., Bradley, C. S., …
Landis, J. R. (2016). Pain and Urinary Symptoms Should Not be Combined into a Single
Score: Psychometric Findings from the MAPP Research Network. The Journal of
Urology, 195(4 Pt 1), 949–954.
Grosveld, F. M., Jansen, B. H., Hasman, A., & Visser, S. L. (1976). La reconnaissance des
individus à l’intérieur d'un groupe de 16 sujets normaux. Revue
D’electroencephalographie et de Neurophysiologie Clinique, 6(2), 295–297.
Gudmundsson, S., Runarsson, T. P., Sigurdsson, S., Eiriksdottir, G., & Johnsen, K. (2007).
Reliability of quantitative EEG features. Clinical Neurophysiology: Official Journal of the
International Federation of Clinical Neurophysiology, 118(10), 2162–2171.
Haegens, S., Cousijn, H., Wallis, G., Harrison, P. J., & Nobre, A. C. (2014). Inter- and intraindividual variability in alpha peak frequency. NeuroImage, 92, 46–55.
Hah, J. M., Aivaliotis, V. I., Hettie, G., Pirrotta, L. X., Mackey, S. C., & Nguyen, L. A. (2022).
Whole Body Pain Distribution and Risk Factors for Widespread Pain Among Patients

145
Presenting with Abdominal Pain: A Retrospective Cohort Study. Pain and Therapy,
11(2), 683–699.
Halgren, M., Ulbert, I., Bastuji, H., Fabó, D., Erőss, L., Rey, M., Devinsky, O., Doyle, W. K.,
Mak-McCully, R., Halgren, E., Wittner, L., Chauvel, P., Heit, G., Eskandar, E., Mandell,
A., & Cash, S. S. (2019a). The Generation and Propagation of the Human Alpha
Rhythm. https://doi.org/10.1101/202564
Halgren, M., Ulbert, I., Bastuji, H., Fabó, D., Erőss, L., Rey, M., Devinsky, O., Doyle, W. K.,
Mak-McCully, R., Halgren, E., Wittner, L., Chauvel, P., Heit, G., Eskandar, E., Mandell,
A., & Cash, S. S. (2019b). The generation and propagation of the human alpha rhythm.
Proceedings of the National Academy of Sciences of the United States of America,
116(47), 23772–23782.
Harris, R. E., Williams, D. A., McLean, S. A., Sen, A., Hufford, M., Gendreau, R. M., Gracely, R.
H., & Clauw, D. J. (2005). Characterization and consequences of pain variability in
individuals with fibromyalgia. Arthritis and Rheumatism, 52(11).
https://doi.org/10.1002/art.21407
Harte, S. E., Harris, R. E., & Clauw, D. J. (2018). The neurobiology of central sensitization.
Journal of Applied Biobehavioral Research, 23(2), e12137.
Hegarty, A. K., Yani, M. S., Albishi, A., Michener, L. A., & Kutch, J. J. (2020). Salience network
functional connectivity is spatially heterogeneous across sensorimotor cortex in healthy
humans. NeuroImage, 221, 117177.
Henderson, L. A., Peck, C. C., Petersen, E. T., Rae, C. D., Youssef, A. M., Reeves, J. M.,
Wilcox, S. L., Akhter, R., Murray, G. M., & Gustin, S. M. (2013). Chronic pain: lost
inhibition? The Journal of Neuroscience: The Official Journal of the Society for
Neuroscience, 33(17), 7574–7582.
Hill, C. E., Reynolds, E. L., Burke, J. F., Banerjee, M., Kerber, K. A., Magliocco, B., Esper, G. J.,
Skolarus, L. E., & Callaghan, B. C. (2021). Increasing Out-of-Pocket Costs for
Neurologic Care for Privately Insured Patients. Neurology, 96(3), e322–e332.
Hosier, G. W., Doiron, R. C., Tolls, V., & Nickel, J. C. (2018). The X-Y factor: Females and
males with urological chronic pelvic pain syndrome present distinct clinical phenotypes.
Canadian Urological Association Journal = Journal de l’Association Des Urologues Du
Canada, 12(6), E270–E275.
Hughes, S. W., & Crunelli, V. (2005). Thalamic mechanisms of EEG alpha rhythms and their
pathological implications. The Neuroscientist: A Review Journal Bringing Neurobiology,
Neurology and Psychiatry, 11(4), 357–372.
Hughes, S. W., Lörincz, M., Cope, D. W., Blethyn, K. L., Kékesi, K. A., Rheinallt Parri, H.,
Juhász, G., & Crunelli, V. (2004). Synchronized Oscillations at α and θ Frequencies in
the Lateral Geniculate Nucleus. In Neuron (Vol. 42, Issue 2, pp. 253–268).
https://doi.org/10.1016/s0896-6273(04)00191-6

146
Hüppe, A., Brockow, T., & Raspe, H. (2004). Chronische ausgebreitete Schmerzenund Tender
Points bei Rückenschmerzen in derBevölkerung. Zeitschrift Für Rheumatologie, 63(1),
76–83.
Hyun, J. W., Jang, H., Yu, J., Park, N. Y., Kim, S. H., Huh, S. Y., Kim, W., Park, M. S., Oh, J.,
Park, K. D., & Kim, H. J. (2020). Comparison of Neuropathic Pain in Neuromyelitis
Optica Spectrum Disorder and Multiple Sclerosis. Journal of Clinical Neurology , 16(1),
124–130.
Jann, K., Koenig, T., Dierks, T., Boesch, C., & Federspiel, A. (2010). Association of individual
resting state EEG alpha frequency and cerebral blood flow. NeuroImage, 51(1), 365–
372.
Jee, H. J., Zhu, E., Sun, M., Liu, W., Zhang, Q., & Wang, J. (2023). Anterior cingulate cortex
regulates pain catastrophizing-like behaviors in rats. Molecular Brain, 16(1), 71.
Jenkinson, M., Bannister, P., Brady, M., & Smith, S. (2002). Improved optimization for the robust
and accurate linear registration and motion correction of brain images. NeuroImage,
17(2), 825–841.
Jiang, H., Popov, T., Jylänki, P., Bi, K., Yao, Z., Lu, Q., Jensen, O., & van Gerven, M. A. J.
(2016). Predictability of depression severity based on posterior alpha oscillations.
Clinical Neurophysiology: Official Journal of the International Federation of Clinical
Neurophysiology, 127(4), 2108–2114.
Ji, R.-R., Nackley, A., Huh, Y., Terrando, N., & Maixner, W. (2018). Neuroinflammation and
Central Sensitization in Chronic and Widespread Pain. Anesthesiology, 129(2), 343–
366.
Kandel, E., & Professor of Biochemistry and Molecular Biophysics Thomas M Jessell, M D.
(2000). Principles of Neural Science, Fourth Edition. McGraw-Hill
Companies,Incorporated.
Kaplan, C. M., Schrepf, A., Mawla, I., Ichesco, E., Boehnke, K. F., Beltz, A., Foxen-Craft, E.,
Puglia, M. P., 2nd, Tsodikov, A., Williams, D. A., Hassett, A. L., Clauw, D. J., Harte, S.
E., & Harris, R. E. (2022). Neurobiological antecedents of multisite pain in children. Pain,
163(4), e596–e603.
Kilpatrick, L. A., Kutch, J. J., Tillisch, K., Naliboff, B. D., Labus, J. S., Jiang, Z., Farmer, M. A.,
Apkarian, A. V., Mackey, S., Martucci, K. T., Clauw, D. J., Harris, R. E., Deutsch, G.,
Ness, T. J., Yang, C. C., Maravilla, K., Mullins, C., & Mayer, E. A. (2014). Alterations in
resting state oscillations and connectivity in sensory and motor networks in women with
interstitial cystitis/painful bladder syndrome. The Journal of Urology, 192(3), 947–955.
Kim, J. A., Bosma, R. L., Hemington, K. S., Rogachov, A., Osborne, N. R., Cheng, J. C., Oh, J.,
Crawley, A. P., Dunkley, B. T., & Davis, K. D. (2019). Neuropathic pain and pain
interference are linked to alpha-band slowing and reduced beta-band
magnetoencephalography activity within the dynamic pain connectome in patients with
multiple sclerosis. Pain, 160(1), 187–197.

147
Kim, J., Kang, I., Chung, Y.-A., Kim, T.-S., Namgung, E., Lee, S., Oh, J. K., Jeong, H. S., Cho,
H., Kim, M. J., Kim, T. D., Choi, S. H., Lim, S. M., Lyoo, I. K., & Yoon, S. (2018). Altered
attentional control over the salience network in complex regional pain syndrome.
Scientific Reports, 8. https://doi.org/10.1038/s41598-018-25757-2
Kim, S., Lee, J., & Boone, D. (2022). Protective and Risk Factors at the Intersection of Chronic
Pain, Depression, Anxiety, and Somatic Amplification: A Latent Profile Approach.
Journal of Pain Research, 15, 1107–1121.
Kisler, L. B., Kim, J. A., Hemington, K. S., Rogachov, A., Cheng, J. C., Bosma, R. L., Osborne,
N. R., Dunkley, B. T., Inman, R. D., & Davis, K. D. (2020). Abnormal alpha band power
in the dynamic pain connectome is a marker of chronic pain with a neuropathic
component. NeuroImage. Clinical, 26, 102241.
Klimesch, W. (1997). EEG-alpha rhythms and memory processes. In International Journal of
Psychophysiology (Vol. 26, Issues 1-3, pp. 319–340). https://doi.org/10.1016/s0167-
8760(97)00773-3
Klimesch, W. (1999). EEG alpha and theta oscillations reflect cognitive and memory
performance: a review and analysis. Brain Research. Brain Research Reviews, 29(2-3),
169–195.
Klimesch, W., Schimke, H., & Pfurtscheller, G. (1993). Alpha frequency, cognitive load and
memory performance. Brain Topography, 5(3), 241–251.
Kratz, A. L., Whibley, D., Alschuler, K. N., Ehde, D. M., Williams, D. A., Clauw, D. J., & Braley,
T. J. (2021). Characterizing chronic pain phenotypes in multiple sclerosis: a nationwide
survey study. Pain, 162(5), 1426–1433.
Krupina, N. A., Churyukanov, M. V., Kukushkin, M. L., & Yakhno, N. N. (2019). Central
Neuropathic Pain and Profiles of Quantitative Electroencephalography in Multiple
Sclerosis Patients. Frontiers in Neurology, 10, 1380.
Kucyi, A., & Davis, K. D. (2015). The dynamic pain connectome. Trends in Neurosciences,
38(2). https://doi.org/10.1016/j.tins.2014.11.006
Kucyi, A., Salomons, T. V., & Davis, K. D. (2013). Mind wandering away from pain dynamically
engages antinociceptive and default mode brain networks. Proceedings of the National
Academy of Sciences of the United States of America, 110(46), 18692–18697.
Kutch, J. J., Ichesco, E., Hampson, J. P., Labus, J. S., Farmer, M. A., Martucci, K. T., Ness, T.
J., Deutsch, G., Apkarian, A. V., Mackey, S. C., Klumpp, D. J., Schaeffer, A. J.,
Rodriguez, L. V., Kreder, K. J., Buchwald, D., Andriole, G. L., Lai, H. H., Mullins, C.,
Kusek, J. W., … MAPP Research Network. (2017). Brain signature and functional impact
of centralized pain: a multidisciplinary approach to the study of chronic pelvic pain
(MAPP) network study. Pain, 158(10), 1979–1991.
Laumann, T. O., Gordon, E. M., Adeyemo, B., Snyder, A. Z., Joo, S. J., Chen, M.-Y., Gilmore,
A. W., McDermott, K. B., Nelson, S. M., Dosenbach, N. U. F., Schlaggar, B. L., Mumford,
J. A., Poldrack, R. A., & Petersen, S. E. (2015). Functional System and Areal
Organization of a Highly Sampled Individual Human Brain. Neuron, 87(3), 657–670.

148
Lee, D. M., Pendleton, N., Tajar, A., O’Neill, T. W., O’Connor, D. B., Bartfai, G., Boonen, S.,
Casanueva, F. F., Finn, J. D., Forti, G., Giwercman, A., Han, T. S., Huhtaniemi, I. T.,
Kula, K., Lean, M. E. J., Punab, M., Silman, A. J., Vanderschueren, D., Moseley, C. M.,
… EMAS study group. (2010). Chronic widespread pain is associated with slower
cognitive processing speed in middle-aged and older European men. Pain, 151(1), 30–
36.
Lee, J.-J., Kim, H. J., Čeko, M., Park, B.-Y., Lee, S. A., Park, H., Roy, M., Kim, S.-G., Wager, T.
D., & Woo, C.-W. (2021). A neuroimaging biomarker for sustained experimental and
clinical pain. Nature Medicine, 27(1), 174–182.
Lefaucheur, J.-P., Aleman, A., Baeken, C., Benninger, D. H., Brunelin, J., Di Lazzaro, V.,
Filipović, S. R., Grefkes, C., Hasan, A., Hummel, F. C., Jääskeläinen, S. K., Langguth,
B., Leocani, L., Londero, A., Nardone, R., Nguyen, J.-P., Nyffeler, T., Oliveira-Maia, A.
J., Oliviero, A., … Ziemann, U. (2020). Evidence-based guidelines on the therapeutic
use of repetitive transcranial magnetic stimulation (rTMS): An update (2014-2018).
Clinical Neurophysiology: Official Journal of the International Federation of Clinical
Neurophysiology, 131(2), 474–528.
Lever, I., Cunningham, J., Grist, J., Yip, P. K., & Malcangio, M. (2003). Release of BDNF and
GABA in the dorsal horn of neuropathic rats. The European Journal of Neuroscience,
18(5), 1169–1174.
Li, J., Simone, D. A., & Larson, A. A. (1999). Windup leads to characteristics of central
sensitization. Pain, 79(1). https://doi.org/10.1016/S0304-3959(98)00154-7
Lim, M., Kim, J. S., Kim, D. J., & Chung, C. K. (2016). Increased Low- and High-Frequency
Oscillatory Activity in the Prefrontal Cortex of Fibromyalgia Patients. Frontiers in Human
Neuroscience, 10, 111.
Lindgren, K. A., Larson, C. L., Schaefer, S. M., Abercrombie, H. C., Ward, R. T., Oakes, T. R.,
Holden, J. E., Perlman, S. B., Benca, R. M., & Davidson, R. J. (1999). Thalamic
metabolic rate predicts EEG alpha power in healthy control subjects but not in
depressed patients. In Biological Psychiatry (Vol. 45, Issue 8, pp. 943–952).
https://doi.org/10.1016/s0006-3223(98)00350-3
Lindquist, M. A., Geuter, S., Wager, T. D., & Caffo, B. S. (2019). Modular preprocessing
pipelines can reintroduce artifacts into fMRI data. Human Brain Mapping, 40(8), 2358–
2376.
Linton, S. J., & Shaw, W. S. (2011). Impact of psychological factors in the experience of pain.
Physical Therapy, 91(5), 700–711.
Liu, R., Kurihara, C., Tsai, H.-T., Silvestri, P. J., Bennett, M. I., Pasquina, P. F., & Cohen, S. P.
(2017). Classification and Treatment of Chronic Neck Pain: A Longitudinal Cohort Study.
Regional Anesthesia and Pain Medicine, 42(1), 52–61.
Li, W., Gong, Y., Liu, J., Guo, Y., Tang, H., Qin, S., Zhao, Y., Wang, S., Xu, Z., & Chen, B.
(2021). Peripheral and Central Pathological Mechanisms of Chronic Low Back Pain: A
Narrative Review. Journal of Pain Research, 14, 1483–1494.

149
Li, X.-H., Matsuura, T., Xue, M., Chen, Q.-Y., Liu, R.-H., Lu, J.-S., Shi, W., Fan, K., Zhou, Z.,
Miao, Z., Yang, J., Wei, S., Wei, F., Chen, T., & Zhuo, M. (2021). Oxytocin in the anterior
cingulate cortex attenuates neuropathic pain and emotional anxiety by inhibiting
presynaptic long-term potentiation. Cell Reports, 36(3).
https://doi.org/10.1016/j.celrep.2021.109411
Llinás, R. R., Ribary, U., Jeanmonod, D., Kronberg, E., & Mitra, P. P. (1999). Thalamocortical
dysrhythmia: A neurological and neuropsychiatric syndrome characterized by
magnetoencephalography. Proceedings of the National Academy of Sciences of the
United States of America, 96(26), 15222–15227.
Lopez-Jornet, P., Molino-Pagan, D., Parra-Perez, P., & Valenzuela, S. (2017). Neuropathic Pain
in Patients with Burning Mouth Syndrome Evaluated Using painDETECT. Pain Medicine,
18(8), 1528–1533.
Louis ES, Frey L (2016) Electroencephalography (EEG): an introductory text and atlas of normal
and abnormal findings in adults, children, and infants. American Epilepsy Society,
Chicago, IL,. https://doi.org/10.5698/978-0-9979756-0-4
Lorenz, J., Minoshima, S., & Casey, K. L. (2003). Keeping pain out of mind: the role of the
dorsolateral prefrontal cortex in pain modulation. Brain: A Journal of Neurology, 126(Pt
5), 1079–1091.
Maixner, W., Fillingim, R. B., Williams, D. A., Smith, S. B., & Slade, G. D. (2016). Overlapping
Chronic Pain Conditions: Implications for Diagnosis and Classification. The Journal of
Pain: Official Journal of the American Pain Society, 17(9 Suppl), T93–T107.
Martucci, K. T., & Mackey, S. C. (2018). Neuroimaging of Pain: Human Evidence and Clinical
Relevance of Central Nervous System Processes and Modulation. Anesthesiology,
128(6), 1241–1254.
Mawla, I., Schrepf, A., Ichesco, E., Harte, S. E., Klumpp, D. J., Griffith, J. W., Strachan, E.,
Yang, C. C., Lai, H., Andriole, G., Magnotta, V. A., Kreder, K., Clauw, D. J., Harris, R. E.,
Quentin Clemens, J., Richard Landis, J., Mullins, C., Rodriguez, L. V., Mayer, E. A., &
Kutch, J. J. (2020). Natural bladder filling alters resting brain function at multiple spatial
scales: a proof-of-concept MAPP Network Neuroimaging Study. In Scientific Reports
(Vol. 10, Issue 1). https://doi.org/10.1038/s41598-020-76857-x
May, E. S., Gil Ávila, C., Ta Dinh, S., Heitmann, H., Hohn, V. D., Nickel, M. M., Tiemann, L.,
Tölle, T. R., & Ploner, M. (2021). Dynamics of brain function in patients with chronic pain
assessed by microstate analysis of resting-state electroencephalography. Pain, 162(12),
2894–2908.
McLain, N. J., Yani, M. S., & Kutch, J. J. (2022). Analytic consistency and neural correlates of
peak alpha frequency in the study of pain. Journal of Neuroscience Methods, 368,
109460.
Melzack, R. (1975). Pain Mechanisms: A New Theory.
Melzack, R. (1987). The short-form McGill Pain Questionnaire. Pain, 30(2), 191–197.

150
Mendell, L. M. (1966). Physiological properties of unmyelinated fiber projection to the spinal
cord. Experimental Neurology, 16(3), 316–332.
Mendell, L. M. (2014). Constructing and deconstructing the gate theory of pain. Pain, 155(2),
210–216.
Mendell, L. M., & Wall, P. D. (1965). RESPONSES OF SINGLE DORSAL CORD CELLS TO
PERIPHERAL CUTANEOUS UNMYELINATED FIBRES. Nature, 206, 97–99.
Menon, V., & Uddin, L. Q. (2010). Saliency, switching, attention and control: a network model of
insula function. Brain Structure & Function, 214(5-6), 655–667.
Mierau, A., Klimesch, W., & Lefebvre, J. (2017). State-dependent alpha peak frequency shifts:
Experimental evidence, potential mechanisms and functional implications.
Neuroscience, 360, 146–154.
Mingorance, J. A., Montoya, P., Miranda, J. G. V., & Riquelme, I. (2021). An Observational
Study Comparing Fibromyalgia and Chronic Low Back Pain in Somatosensory
Sensitivity, Motor Function and Balance. Healthcare (Basel, Switzerland), 9(11).
https://doi.org/10.3390/healthcare9111533
Molins, A., Stufflebeam, S. M., Brown, E. N., & Hämäläinen, M. S. (2008). Quantification of the
benefit from integrating MEG and EEG data in minimum l2-norm estimation.
NeuroImage, 42(3), 1069–1077.
Moore, K. A., Kohno, T., Karchewski, L. A., Scholz, J., Baba, H., & Woolf, C. J. (2002). Partial
peripheral nerve injury promotes a selective loss of GABAergic inhibition in the
superficial dorsal horn of the spinal cord. The Journal of Neuroscience: The Official
Journal of the Society for Neuroscience, 22(15), 6724–6731.
Mosabbir, A. (2022). Mechanisms behind the Development of Chronic Low Back Pain and Its
Neurodegenerative Features. Life , 13(1). https://doi.org/10.3390/life13010084
Morsø, L., Olsen Rose, K., Schiøttz-Christensen, B., Sowden, G., Søndergaard, J., &
Christiansen, D. H. (2021). Effectiveness of stratified treatment for back pain in Danish
primary care: A randomized controlled trial. European Journal of Pain , 25(9), 2020–
2038.
Mosch, B., Hagena, V., Herpertz, S., & Diers, M. (2023). Brain morphometric changes in
fibromyalgia and the impact of psychometric and clinical factors: a volumetric and
diffusion-tensor imaging study. Arthritis Research & Therapy, 25(1), 81.
Murphy, K. J., & Brunberg, J. A. (1997). Adult claustrophobia, anxiety and sedation in MRI.
Magnetic Resonance Imaging, 15(1), 51–54.
Muthuraman, M., Moliadze, V., Mideksa, K. G., Anwar, A. R., Stephani, U., Deuschl, G., Freitag,
C. M., & Siniatchkin, M. (2015). EEG-MEG Integration Enhances the Characterization of
Functional and Effective Connectivity in the Resting State Network. PloS One, 10(10),
e0140832.

151
Mutschler, I., Ball, T., Wankerl, J., & Strigo, I. A. (2012). Pain and emotion in the insular cortex:
evidence for functional reorganization in major depression. Neuroscience Letters,
520(2), 204–209.
Nadar, M. S., Jasem, Z., & Manee, F. S. (2016). The Cognitive Functions in Adults with Chronic
Pain: A Comparative Study. Pain Research & Management: The Journal of the
Canadian Pain Society = Journal de La Societe Canadienne Pour Le Traitement de La
Douleur, 2016, 5719380.
Napadow, V., LaCount, L., Park, K., As-Sanie, S., Clauw, D. J., & Harris, R. E. (2010). Intrinsic
brain connectivity in fibromyalgia is associated with chronic pain intensity. Arthritis and
Rheumatism, 62(8), 2545–2555.
News, Announcements, and IASP. (2017). Crisis, 38(5), 360–361.
Newson, J. J., & Thiagarajan, T. C. (2018). EEG Frequency Bands in Psychiatric Disorders: A
Review of Resting State Studies. Frontiers in Human Neuroscience, 12, 521.
Ngernyam, N., Jensen, M. P., Arayawichanon, P., Auvichayapat, N., Tiamkao, S., Janjarasjitt,
S., Punjaruk, W., Amatachaya, A., Aree-uea, B., & Auvichayapat, P. (2015). The effects
of transcranial direct current stimulation in patients with neuropathic pain from spinal
cord injury. Clinical Neurophysiology: Official Journal of the International Federation of
Clinical Neurophysiology, 126(2), 382–390.
Nir, R.-R., Sinai, A., Raz, E., Sprecher, E., & Yarnitsky, D. (2010). Pain assessment by
continuous EEG: association between subjective perception of tonic pain and peak
frequency of alpha oscillations during stimulation and at rest. Brain Research, 1344, 77–
86.
Nolte, J., Vanderah, T., & Gould, D. (2016). Nolte’s the Human Brain: An Introduction to Its
Functional Anatomy. Elsevier Philadelphia, PA.
Ohrbach, R., Sharma, S., Fillingim, R. B., Greenspan, J. D., Rosen, J. D., & Slade, G. D. (2020).
Clinical Characteristics of Pain Among Five Chronic Overlapping Pain Conditions.
Journal of Oral & Facial Pain and Headache, 34(Suppl), s29–s42.
Osaka, M., Osaka, N., Koyama, S., Okusa, T., & Kakigi, R. (1999). Individual differences in
working memory and the peak alpha frequency shift on magnetoencephalography1We
reported our preliminary findings in the proceedings of the Third Pan-Pacific Conference
on Brain Topography [9].1. Cognitive Brain Research, 8(3), 365–368.
Overstreet, D. S., Strath, L. J., Jordan, M., Jordan, I. A., Hobson, J. M., Owens, M. A., Williams,
A. C., Edwards, R. R., & Meints, S. M. (2023). A Brief Overview: Sex Differences in
Prevalent Chronic Musculoskeletal Conditions. International Journal of Environmental
Research and Public Health, 20(5). https://doi.org/10.3390/ijerph20054521
Pan, F., Tian, J., Aitken, D., Cicuttini, F., & Jones, G. (2019). Pain at Multiple Sites Is
Associated With Prevalent and Incident Fractures in Older Adults. Journal of Bone and
Mineral Research: The Official Journal of the American Society for Bone and Mineral
Research, 34(11), 2012–2018.

152
Parker, T., Huang, Y., Raghu, A. L. B., FitzGerald, J., Aziz, T. Z., & Green, A. L. (2021).
Supraspinal Effects of Dorsal Root Ganglion Stimulation in Chronic Pain Patients.
Neuromodulation: Journal of the International Neuromodulation Society, 24(4), 646–654.
Parkes, L., Fulcher, B., Yücel, M., & Fornito, A. (2018). An evaluation of the efficacy, reliability,
and sensitivity of motion correction strategies for resting-state functional MRI.
NeuroImage, 171, 415–436.
Park, S. H., Baker, A. K., Krishna, V., Mackey, S. C., & Martucci, K. T. (2022). Altered restingstate functional connectivity within corticostriatal and subcortical-striatal circuits in
chronic pain. Scientific Reports, 12(1), 12683.
Pascual-Marqui, R. D., Esslen, M., Kochi, K., & Lehmann, D. (2002). Functional imaging with
low-resolution brain electromagnetic tomography (LORETA): a review. Methods and
Findings in Experimental and Clinical Pharmacology, 24 Suppl C, 91–95.
Phillips, K., & Clauw, D. J. (2011). Central pain mechanisms in chronic pain states--maybe it is
all in their head. Best Practice & Research. Clinical Rheumatology, 25(2), 141–154.
Posthuma, D., Neale, M. C., Boomsma, D. I., & de Geus, E. J. (2001). Are smarter brains
running faster? Heritability of alpha peak frequency, IQ, and their interrelation. Behavior
Genetics, 31(6), 567–579.
Power, J. D., Mitra, A., Laumann, T. O., Snyder, A. Z., Schlaggar, B. L., & Petersen, S. E.
(2014). Methods to detect, characterize, and remove motion artifact in resting state fMRI.
NeuroImage, 84, 320–341.
Clemens, Q.J., Calhoun, E. A., Litwin, M. S., McNaughton-Collins, M., Kusek, J. W., Crowley, E.
M., & Richard Landis, J. (2009). Validation of a Modified National Institutes of Health
Chronic Prostatitis Symptom Index to Assess Genitourinary Pain in Both Men and
Women. Urology, 74(5), 983.
Raghuraman, N., Wang, Y., Schenk, L. A., Furman, A. J., Tricou, C., Seminowicz, D. A., &
Colloca, L. (2019). Neural and behavioral changes driven by observationally-induced
hypoalgesia. Scientific Reports, 9(1), 19760.
Reimer, M., Hüllemann, P., Hukauf, M., Keller, T., Binder, A., Gierthmühlen, J., & Baron, R.
(2017). Prediction of response to tapentadol in chronic low back pain. European Journal
of Pain , 21(2), 322–333.
Rifbjerg-Madsen, S., Christensen, A. W., Boesen, M., Christensen, R., Danneskiold-Samsøe,
B., Bliddal, H., Dreyer, L., Locht, H., & Amris, K. (2018). The course of pain
hypersensitivity according to painDETECT in patients with rheumatoid arthritis initiating
treatment: results from the prospective FRAME-cohort study. Arthritis Research &
Therapy, 20(1), 105.
Richard Clark, C., Veltmeyer, M. D., Hamilton, R. J., Simms, E., Paul, R., Hermens, D., &
Gordon, E. (2004). Spontaneous alpha peak frequency predicts working memory
performance across the age span. International Journal of Psychophysiology: Official
Journal of the International Organization of Psychophysiology, 53(1), 1–9.

153
Rosenstiel, A. K., & Keefe, F. J. (1983). The use of coping strategies in chronic low back pain
patients: relationship to patient characteristics and current adjustment. Pain, 17(1).
https://doi.org/10.1016/0304-3959(83)90125-2
Saalmann, Y. B., Pinsk, M. A., Wang, L., Li, X., & Kastner, S. (2012). The pulvinar regulates
information transmission between cortical areas based on attention demands. Science,
337(6095), 753–756.
Salinsky, M. C., Oken, B. S., & Morehead, L. (1991). Test-retest reliability in EEG frequency
analysis. Electroencephalography and Clinical Neurophysiology, 79(5), 382–392.
Salovey, P., Smith, A. F., Turk, D. C., Jobe, J. B., & Willis, G. B. (1993). The accuracy of
memory for pain. APS Journal, 2(3), 184–191.
Sarnthein, J., Stern, J., Aufenberg, C., Rousson, V., & Jeanmonod, D. (2006). Increased EEG
power and slowed dominant frequency in patients with neurogenic pain. Brain: A Journal
of Neurology, 129(Pt 1), 55–64.
Sato, G., Osumi, M., & Morioka, S. (2017). Effects of wheelchair propulsion on neuropathic pain
and resting electroencephalography after spinal cord injury. Journal of Rehabilitation
Medicine: Official Journal of the UEMS European Board of Physical and Rehabilitation
Medicine, 49(2), 136–143.
Schaefer, A., Kong, R., Gordon, E. M., Laumann, T. O., Zuo, X.-N., Holmes, A. J., Eickhoff, S.
B., & Yeo, B. T. T. (2018). Local-Global Parcellation of the Human Cerebral Cortex from
Intrinsic Functional Connectivity MRI. Cerebral Cortex , 28(9), 3095–3114.
Scherrer, K. H., Ziadni, M. S., Kong, J.-T., Sturgeon, J. A., Salmasi, V., Hong, J., Cramer, E.,
Chen, A. L., Pacht, T., Olson, G., Darnall, B. D., Kao, M.-C., & Mackey, S. (2021).
Development and validation of the Collaborative Health Outcomes Information Registry
body map. Pain Reports, 6(1), e880.
Schmidt, S., Naranjo, J. R., Brenneisen, C., Gundlach, J., Schultz, C., Kaube, H., Hinterberger,
T., & Jeanmonod, D. (2012). Pain ratings, psychological functioning and quantitative
EEG in a controlled study of chronic back pain patients. PloS One, 7(3), e31138.
Schrepf, A., Kaplan, C., Harris, R. E., Williams, D. A., Clauw, D. J., As-Sanie, S., Till, S.,
Clemens, J. Q., Rodriguez, L. V., Van Bokhoven, A., Landis, R., Gallop, R., Bradley, C.,
Naliboff, B., Pontari, M., O’Donnell, M., Luo, Y., Kreder, K., Lutgendorf, S. K., & Harte, S.
E. (2023). Stimulated whole-blood cytokine/chemokine responses are associated with
interstitial cystitis/bladder pain syndrome phenotypes and features of nociplastic pain: a
multidisciplinary approach to the study of chronic pelvic pain research network study.
Pain, 164(5), 1148–1157.
Schulman, J. J., Cancro, R., Lowe, S., Lu, F., Walton, K. D., & Llinás, R. R. (2011). Imaging of
thalamocortical dysrhythmia in neuropsychiatry. Frontiers in Human Neuroscience, 5,
69.
Schwab, S., Koenig, T., Morishima, Y., Dierks, T., Federspiel, A., & Jann, K. (2015). Discovering
frequency sensitive thalamic nuclei from EEG microstate informed resting state fMRI.
NeuroImage, 118, 368–375.

154
Sellmeijer, J., Mathis, V., Hugel, S., Li, X.-H., Song, Q., Chen, Q.-Y., Barthas, F., Lutz, P.-E.,
Karatas, M., Luthi, A., Veinante, P., Aertsen, A., Barrot, M., Zhuo, M., & Yalcin, I. (2018).
Hyperactivity of Anterior Cingulate Cortex Areas 24a/24b Drives Chronic Pain-Induced
Anxiodepressive-like Consequences. The Journal of Neuroscience: The Official Journal
of the Society for Neuroscience, 38(12), 3102.
Shi, Q., Wang, X. S., Mendoza, T. R., Pandya, K. J., & Cleeland, C. S. (2009). Assessing
persistent cancer pain: a comparison of current pain ratings and pain recalled from the
past week. Journal of Pain and Symptom Management, 37(2), 168–174.
Shirvalkar, P., Prosky, J., Chin, G., Ahmadipour, P., Sani, O. G., Desai, M., Schmitgen, A.,
Dawes, H., Shanechi, M. M., Starr, P. A., & Chang, E. F. (2023). First-in-human
prediction of chronic pain state using intracranial neural biomarkers. Nature
Neuroscience, 26(6), 1090–1099.
Simis, M., Pacheco-Barrios, K., Uygur-Kucukseymen, E., Castelo-Branco, L., Battistella, L. R., &
Fregni, F. (2021). Specific Electroencephalographic Signatures for Pain and Descending
Pain Inhibitory System in Spinal Cord Injury. Pain Medicine .
https://doi.org/10.1093/pm/pnab124
Siracusa, R., Paola, R. D., Cuzzocrea, S., & Impellizzeri, D. (2021). Fibromyalgia:
Pathogenesis, Mechanisms, Diagnosis and Treatment Options Update. International
Journal of Molecular Sciences, 22(8). https://doi.org/10.3390/ijms22083891
Sluka, K. A., & Clauw, D. J. (2016). Neurobiology of fibromyalgia and chronic widespread pain.
Neuroscience, 338, 114–129.
Soni, A., Wanigasekera, V., Mezue, M., Cooper, C., Javaid, M. K., Price, A. J., & Tracey, I.
(2019). Central Sensitization in Knee Osteoarthritis: Relating Presurgical Brainstem
Neuroimaging and PainDETECT-Based Patient Stratification to Arthroplasty Outcome.
Arthritis & Rheumatology (Hoboken, N.J.), 71(4), 550–560.
Spielberger, C. D. (1970). Manual for the State-Trait Anxiety Inventory (self-evaluation
questionnaire). (No Title). https://cir.nii.ac.jp/crid/1370285712575158016
Staud, R. (2006). Biology and therapy of fibromyalgia: pain in fibromyalgia syndrome. Arthritis
Research & Therapy, 8(3), 208.
Staud, R., Boissoneault, J., Lai, S., Mejia, M. S., Ramanlal, R., Godfrey, M. M., & Stroman, P.
W. (2021). Spinal cord neural activity of patients with fibromyalgia and healthy controls
during temporal summation of pain: an fMRI study. Journal of Neurophysiology, 126(3),
946–956.
Staud, R., Craggs, J. G., Robinson, M. E., Perlstein, W. M., & Price, D. D. (2007). Brain activity
related to temporal summation of C-fiber evoked pain. Pain, 129(1-2), 130–142.
Staud, R., & Smitherman, M. L. (2002). Peripheral and central sensitization in fibromyalgia:
pathogenetic role. Current Pain and Headache Reports, 6(4), 259–266.
Stein, A. von, von Stein, A., & Sarnthein, J. (2000). Different frequencies for different scales of
cortical integration: from local gamma to long range alpha/theta synchronization. In

155
International Journal of Psychophysiology (Vol. 38, Issue 3, pp. 301–313).
https://doi.org/10.1016/s0167-8760(00)00172-0
Stephens-Shields, A. J., Clemens, J. Q., Jemielita, T., Farrar, J., Sutcliffe, S., Hou, X., Landis, J.
R., & MAPP Research Network. (2016). Symptom Variability and Early Symptom
Regression in the MAPP Study: A Prospective Study of Urological Chronic Pelvic Pain
Syndrome. The Journal of Urology, 196(5), 1450–1455.
Strigo, I. A., Simmons, A. N., Matthews, S. C., Craig, A. D. B., & Paulus, M. P. (2008).
Association of major depressive disorder with altered functional brain response during
anticipation and processing of heat pain. Archives of General Psychiatry, 65(11), 1275–
1284.
Sufianov, A. A., Shapkin, A. G., Sufianova, G. Z., Elishev, V. G., Barashin, D. A., Berdichevskii,
V. B., & Churkin, S. V. (2014). Functional and metabolic changes in the brain in
neuropathic pain syndrome against the background of chronic epidural electrostimulation
of the spinal cord. Bulletin of Experimental Biology and Medicine, 157(4), 462–465.
Sun, L., Liu, R., Guo, F., Wen, M.-Q., Ma, X.-L., Li, K.-Y., Sun, H., Xu, C.-L., Li, Y.-Y., Wu, M.-
Y., Zhu, Z.-G., Li, X.-J., Yu, Y.-Q., Chen, Z., Li, X.-Y., & Duan, S. (2020). Parabrachial
nucleus circuit governs neuropathic pain-like behavior. Nature Communications, 11(1),
5974.
Suvas, S. (2017a). Role of Substance P Neuropeptide in Inflammation, Wound Healing, and
Tissue Homeostasis. Journal of Immunology , 199(5), 1543–1552.
Suzuki, K., Haruyama, Y., Kobashi, G., Sairenchi, T., Uchiyama, K., Yamaguchi, S., & Hirata, K.
(2021). Central Sensitization in Neurological, Psychiatric, and Pain Disorders: A
Multicenter Case-Controlled Study. Pain Research & Management: The Journal of the
Canadian Pain Society = Journal de La Societe Canadienne Pour Le Traitement de La
Douleur, 2021, 6656917.
Ta Dinh, S., Nickel, M. M., Tiemann, L., May, E. S., Heitmann, H., Hohn, V. D., Edenharter, G.,
Utpadel-Fischler, D., Tölle, T. R., Sauseng, P., Gross, J., & Ploner, M. (2019). Brain
dysfunction in chronic pain patients assessed by resting-state electroencephalography.
Pain, 160(12), 2751–2765.
Takazawa, T., & MacDermott, A. B. (2010). Synaptic pathways and inhibitory gates in the spinal
cord dorsal horn. Annals of the New York Academy of Sciences, 1198, 153–158.
Talbot, K., Madden, V. J., Jones, S. L., & Moseley, G. L. (2019). The sensory and affective
components of pain: are they differentially modifiable dimensions or inseparable aspects
of a unitary experience? A systematic review. British Journal of Anaesthesia, 123(2),
e263–e272.
Tatum, W. O., Mani, J., Jin, K., Halford, J. J., Gloss, D., Fahoum, F., Maillard, L., Mothersill, I., &
Beniczky, S. (2022). Minimum standards for inpatient long-term video-EEG monitoring: A
clinical practice guideline of the international league against epilepsy and international
federation of clinical neurophysiology. Clinical Neurophysiology: Official Journal of the
International Federation of Clinical Neurophysiology, 134, 111–128.

156
Taylor, B. K. (2009). Spinal inhibitory neurotransmission in neuropathic pain. Current Pain and
Headache Reports, 13(3), 208–214.
Tement, S., Pahor, A., & Jaušovec, N. (2016). EEG alpha frequency correlates of burnout and
depression: The role of gender. Biological Psychology, 114, 1–12.
Todd, A. J. (2015). Plasticity of inhibition in the spinal cord. Handbook of Experimental
Pharmacology, 227, 171–190.
Trainor, A., & Miner, J. (2008). Pain treatment and relief among patients with primary headache
subtypes in the ED. The American Journal of Emergency Medicine, 26(9), 1029–1034.
Tupper, S. M., Rosenberg, A. M., Pahwa, P., & Stinson, J. N. (2013). Pain intensity variability
and its relationship with quality of life in youths with juvenile idiopathic arthritis. Arthritis
Care & Research, 65(4). https://doi.org/10.1002/acr.21850
Valet, M., Sprenger, T., Boecker, H., Willoch, F., Rummeny, E., Conrad, B., Erhard, P., & Tolle,
T. R. (2004). Distraction modulates connectivity of the cingulo-frontal cortex and the
midbrain during pain—an fMRI analysis. Pain, 109(3), 399.
van Beijsterveldt, C. E. M., & van Baal, G. C. M. (2002). Twin and family studies of the human
electroencephalogram: a review and a meta-analysis. In Biological Psychology (Vol. 61,
Issues 1-2, pp. 111–138). https://doi.org/10.1016/s0301-0511(02)00055-8
van den Broeke, E. N., Wilder-Smith, O. H. G., van Goor, H., Vissers, K. C. P., & van Rijn, C. M.
(2013). Patients with persistent pain after breast cancer treatment show enhanced alpha
activity in spontaneous EEG. Pain Medicine , 14(12), 1893–1899.
van der Miesen, M. M., Lindquist, M. A., & Wager, T. D. (2019). Neuroimaging-based
biomarkers for pain: state of the field and current directions. Pain Reports, 4(4), e751.
Van Dijk, K. R. A., Sabuncu, M. R., & Buckner, R. L. (2012). The influence of head motion on
intrinsic functional connectivity MRI. NeuroImage, 59(1), 431–438.
van Ettinger-Veenstra, H., Boehme, R., Ghafouri, B., Olausson, H., Wicksell, R. K., & Gerdle, B.
(2020). Exploration of Functional Connectivity Changes Previously Reported in
Fibromyalgia and Their Relation to Psychological Distress and Pain Measures. Journal
of Clinical Medicine Research, 9(11). https://doi.org/10.3390/jcm9113560
van Ettinger-Veenstra, H., Lundberg, P., Alföldi, P., Södermark, M., Graven-Nielsen, T., Sjörs,
A., Engström, M., & Gerdle, B. (2019). Chronic widespread pain patients show disrupted
cortical connectivity in default mode and salience networks, modulated by pain
sensitivity. Journal of Pain Research, 12, 1743.
Vera-Portocarrero, L. P., Zhang, E.-T., Ossipov, M. H., Xie, J. Y., King, T., Lai, J., & Porreca, F.
(2006). Descending facilitation from the rostral ventromedial medulla maintains nerve
injury-induced central sensitization. Neuroscience, 140(4), 1311–1320.
Viniol, A., Jegan, N., Leonhardt, C., Brugger, M., Strauch, K., Barth, J., Baum, E., & Becker, A.
(2013). Differences between patients with chronic widespread pain and local chronic low

157
back pain in primary care--a comparative cross-sectional analysis. BMC Musculoskeletal
Disorders, 14, 351.
Vogt, B. A. (2005). Pain and emotion interactions in subregions of the cingulate gyrus. Nature
Reviews. Neuroscience, 6(7), 533–544.
Vora, A. J., Doerr, K. D., & Wolfer, L. R. (2010). Functional anatomy and pathophysiology of
axial low back pain: disc, posterior elements, sacroiliac joint, and associated pain
generators. Physical Medicine and Rehabilitation Clinics of North America, 21(4), 679–
709.
Vos, T., Lim, S. S., Abbafati, C., Abbas, K. M., Abbasi, M., Abbasifard, M., Abbasi-Kangevari,
M., Abbastabar, H., Abd-Allah, F., Abdelalim, A., Abdollahi, M., Abdollahpour, I.,
Abolhassani, H., Aboyans, V., Abrams, E. M., Abreu, L. G., Abrigo, M. R. M., AbuRaddad, L. J., Abushouk, A. I., … Murray, C. J. L. (2020). Global burden of 369 diseases
and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the
Global Burden of Disease Study 2019. The Lancet, 396(10258), 1204–1222.
Vuckovic, A., Hasan, M. A., Fraser, M., Conway, B. A., Nasseroleslami, B., & Allan, D. B.
(2014). Dynamic oscillatory signatures of central neuropathic pain in spinal cord injury.
The Journal of Pain: Official Journal of the American Pain Society, 15(6), 645–655.
Vuckovic, A., Jajrees, M., Purcell, M., Berry, H., & Fraser, M. (2018). Electroencephalographic
Predictors of Neuropathic Pain in Subacute Spinal Cord Injury. The Journal of Pain:
Official Journal of the American Pain Society, 19(11), 1256.e1–e1256.e17.
Wang, J., Sun, H., Cui, B., Yang, H., Shan, Y., Dong, C., Zang, Y., & Lu, J. (2021). The
Relationship Among Glucose Metabolism, Cerebral Blood Flow, and Functional Activity:
a Hybrid PET/fMRI Study. Molecular Neurobiology. https://doi.org/10.1007/s12035-021-
02305-0
Wang, Z., & Xu, Z.-Z. (2021). The Parabrachial Nucleus as a Key Regulator of Neuropathic
Pain. Neuroscience Bulletin, 37(7), 1079–1081.
Ware, J., Kosinski, M., & Keller, S. D. (1996). A 12-Item Short-Form Health Survey: construction
of scales and preliminary tests of reliability and validity. Medical Care, 34(3).
https://doi.org/10.1097/00005650-199603000-00003
Winkler, I., Brandl, S., Horn, F., Waldburger, E., Allefeld, C., & Tangermann, M. (2014). Robust
artifactual independent component classification for BCI practitioners. Journal of Neural
Engineering, 11(3), 035013.
Winkler, I., Haufe, S., & Tangermann, M. (2011). Automatic classification of artifactual ICAcomponents for artifact removal in EEG signals. Behavioral and Brain Functions: BBF, 7,
30.
Wirsich, J., Jorge, J., Iannotti, G. R., Shamshiri, E. A., Grouiller, F., Abreu, R., Lazeyras, F.,
Giraud, A.-L., Gruetter, R., Sadaghiani, S., & Vulliémoz, S. (2021). The relationship
between EEG and fMRI connectomes is reproducible across simultaneous EEG-fMRI
studies from 1.5T to 7T. NeuroImage, 231, 117864.

158
Witjes, B., Baillet, S., Roy, M., Oostenveld, R., J P M Huygen, F., & C de Vos, C. (2021).
Magnetoencephalography reveals increased slow-to-fast alpha power ratios in patients
with chronic pain. Pain Reports, 6(2), e928.
Woo, C.-W., Chang, L. J., Lindquist, M. A., & Wager, T. D. (2017). Building better biomarkers:
brain models in translational neuroimaging. In Nature Neuroscience (Vol. 20, Issue 3,
pp. 365–377). https://doi.org/10.1038/nn.4478
Worsley, K. J. (2001). Statistical analysis of activation images. In P. Jezzard, P. M. Matthews, &
S. M. Smith (Eds.), Functional MRI: An Introduction to Methods. Oxford Scholarship
Online.
Wu, L.-J., Toyoda, H., Zhao, M.-G., Lee, Y.-S., Tang, J., Ko, S. W., Jia, Y. H., Shum, F. W. F.,
Zerbinatti, C. V., Bu, G., Wei, F., Xu, T.-L., Muglia, L. J., Chen, Z.-F., Auberson, Y. P.,
Kaang, B.-K., & Zhuo, M. (2005). Upregulation of Forebrain NMDA NR2B Receptors
Contributes to Behavioral Sensitization after Inflammation. The Journal of Neuroscience:
The Official Journal of the Society for Neuroscience, 25(48), 11107–11116.
Wu, F.-L., Shih, Y.-F., Lee, S.-H., Luo, H.-J., & Wang, W. T.-J. (2018). Development of a clinical
prediction rule to identify patients with plantar heel pain likely to benefit from
biomechanical anti-pronation taping: A prospective cohort study. Physical Therapy in
Sport: Official Journal of the Association of Chartered Physiotherapists in Sports
Medicine, 31, 58–67.
Xiao, X., Ding, M., & Zhang, Y.-Q. (2021). Role of the Anterior Cingulate Cortex in Translational
Pain Research. Neuroscience Bulletin, 37(3), 405–422.
Xu, A., Larsen, B., Baller, E. B., Scott, J. C., Sharma, V., Adebimpe, A., Basbaum, A. I.,
Dworkin, R. H., Edwards, R. R., Woolf, C. J., Eickhoff, S. B., Eickhoff, C. R., &
Satterthwaite, T. D. (2020). Convergent neural representations of experimentallyinduced acute pain in healthy volunteers: A large-scale fMRI meta-analysis.
Neuroscience and Biobehavioral Reviews, 112, 300–323.
Xu, A., Larsen, B., Henn, A., Baller, E. B., Scott, J. C., Sharma, V., Adebimpe, A., Basbaum, A.
I., Corder, G., Dworkin, R. H., Edwards, R. R., Woolf, C. J., Eickhoff, S. B., Eickhoff, C.
R., & Satterthwaite, T. D. (2021). Brain Responses to Noxious Stimuli in Patients With
Chronic Pain: A Systematic Review and Meta-analysis. JAMA Network Open, 4(1),
e2032236.
Yam, M. F., Loh, Y. C., Tan, C. S., Adam, S. K., Manan, N. A., & Basir, R. (2018). General
Pathways of Pain Sensation and the Major Neurotransmitters Involved in Pain
Regulation. International Journal of Molecular Sciences, 19(8).
https://doi.org/10.3390/ijms19082164
Yani, M. S., Fenske, S. J., Rodriguez, L. V., & Kutch, J. J. (2019). Motor cortical
neuromodulation of pelvic floor muscle tone: Potential implications for the treatment of
urologic conditions. Neurourology and Urodynamics, 38(6), 1517–1523.
Zakoscielna, K. M., & Parmelee, P. A. (2013). Pain variability and its predictors in older adults:
depression, cognition, functional status, health, and pain. Journal of Aging and Health,
25(8). https://doi.org/10.1177/0898264313504457

159
Zalewska, E. (2020). Is So Called “Split Alpha” in EEG Spectral Analysis a Result of
Methodological and Interpretation Errors? Frontiers in Neuroscience, 14, 608453.
Zebhauser, P. T., Hohn, V. D., & Ploner, M. (2023). Resting-state electroencephalography and
magnetoencephalography as biomarkers of chronic pain: a systematic review. Pain,
164(6), 1200–1221.
Zeilhofer, H. U. (2005). The glycinergic control of spinal pain processing. Cellular and Molecular
Life Sciences: CMLS, 62(18), 2027–2035.
Zhou, P., Wu, Q., Zhan, L., Guo, Z., Wang, C., Wang, S., Yang, Q., Lin, J., Zhang, F., Liu, L.,
Lin, D., Fu, W., & Wu, X. (2023). Alpha peak activity in resting-state EEG is associated
with depressive score. Frontiers in Neuroscience, 17, 1057908.
Zigmond, A. S., & Snaith, P. (1983). The hospital anxiety and depression scale.

Abstract (if available)

Abstract Chronic pain is an impactful condition characterized by no widely effective treatments and a subset of patients who never fully recover. Understanding the origins of pain is critical to informing effective treatments. One relevant metric by which to divide pain origin is peripheral versus centralized: peripheral pain can be defined as direct activation of the nociceptors (damage or threat in the periphery) while centralized pain is compounded by dysregulation of the central nervous system. Past work suggests peripheral and centralized pain are mediated by different neural circuitry. Additionally, symptom pattern studies indicate that centralized pain conditions have distinct symptom profiles from conditions that are driven primarily by peripheral nociceptive input. This work is focused on the combination of clinical phenotypes with two complementary modalities for measuring brain function: functional magnetic resonance imaging (fMRI) and electroencephalography (EEG). The overall goal of our proposed work is to determine how resting state peak alpha frequency (rs-PAF), a promising EEG marker, is associated with phenotypic features of centralized pain and what it reflects about changes in brain function using fMRI. This project will answer important questions about pain patients with heightened involvement of their central nervous system in pain amplification and chronification, allow for targeted assignment of existing therapies, and provide the basis for future development of targeted treatments based on the neural circuits identified.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Brain-based prediction of chronic pain progression: a longitudinal study of urologic chronic pelvic pain syndrome using baseline resting state connectivity from the periaqueductal gray

PDF

Decoding the neurological and genetic underpinnings of chronic pain

PDF

Heart, brain, and breath: studies on the neuromodulation of interoceptive systems

PDF

Brain and behavior correlates of intrinsic motivation and skill learning

PDF

Functional magnetic resonance imaging characterization of peripheral form vision

PDF

Affective neuropsychiatric symptoms and neural connectivity in the early stages of Alzheimer’s disease

PDF

Functional models of fMRI BOLD signal in the visual cortex

PDF

The acute impact of glucose and sucralose on food decisions and brain responses to visual food cues

PDF

Selective inhibition of inflammatory and neuropathic cold pain

PDF

Elucidating the elusive role of TRPA1 in neurogenic inflammatory cold pain

PDF

Behavioral and neural influences of interoception and alexithymia on emotional empathy in autism spectrum disorder

PDF

An iPSC-based biomarker strategy to identify neuroregenerative responders to allopregnanolone

PDF

Axon guidance cues in development of the mammalian auditory circuit

Asset Metadata

Creator McLain, Natalie Jo (author)

Core Title Identifying neural markers of centralized pain

School School of Dentistry

Degree Doctor of Philosophy

Degree Program Biokinesiology

Degree Conferral Date 2024-08

Publication Date 01/21/2025

Defense Date 07/21/2024

Publisher Los Angeles, California (original), University of Southern California (original), University of Southern California. Libraries (digital)

Tag biomarkers,central nervous system,centralized pain,chronic pain,clinical phenotyping,clinical profiles,EEG,fMRI,MRI,neural markers,OAI-PMH Harvest,PAF,pain processing,peak alpha frequency,resting state,subtyping,urologic chronic pelvic pain

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kutch, Jason (committee chair), Jann, Kay (committee member), Schweighofer, Nicolas (committee member), Smith, Beth (committee member)

Creator Email natalieemclain@gmail.com,nmclain08@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113998FYN

Unique identifier UC113998FYN

Identifier etd-McLainNata-13272.pdf (filename)

Legacy Identifier etd-McLainNata-13272

Document Type Dissertation

Format theses (aat)

Rights McLain, Natalie Jo

Internet Media Type application/pdf

Type texts

Source 20240730-usctheses-batch-1186 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu