University of Southern California Dissertations and Theses

The price and clinical impact of specialty drug management in the US healthcare system

(USC Thesis Other)

The price and clinical impact of specialty drug management in the US healthcare system

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
Page 1 of 73

The Price and Clinical Impact of Specialty Drug
Management in the US Healthcare System

By
Cho-Han Lee

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA

In Fulfillment of the Requirement for the Degree of
DOCTOR OF PHILOSOPHY
HEALTH ECONOMICS

August 2019

Page 2 of 73

Table of Contents
Acknowledgment ................................................................................................................................................................................... 3
Chapter 1 – Introduction ........................................................................................................................................................................ 4
Reference .......................................................................................................................................................................................... 6
Chapter 2 - Specialty Drug Price Trends in the Federal 340B Drug Discount Program ........................................................................ 8
Abstract ............................................................................................................................................................................................. 8
Introduction ....................................................................................................................................................................................... 9
Methods ........................................................................................................................................................................................... 10
Results ............................................................................................................................................................................................. 12
Discussion ....................................................................................................................................................................................... 13
Conclusion ...................................................................................................................................................................................... 16
References ....................................................................................................................................................................................... 17
Figure & Table ................................................................................................................................................................................ 20
Appendix ......................................................................................................................................................................................... 22
Chapter 3 - An Analysis of Granulocyte-Colony-Stimulating Factor Utilization for the Prevention of Chemotherapy-Induced Febrile
Neutropenia among Breast Cancer Patients in the United States ......................................................................................................... 24
Abstract ........................................................................................................................................................................................... 24
Introduction ..................................................................................................................................................................................... 25
Methods ........................................................................................................................................................................................... 26
Results ............................................................................................................................................................................................. 28
Discussion ....................................................................................................................................................................................... 29
Conclusion ...................................................................................................................................................................................... 31
Figure & Table ................................................................................................................................................................................ 32
References ....................................................................................................................................................................................... 38
Appendix ......................................................................................................................................................................................... 41
Chapter 4 - An Analysis of Granulocyte-Colony-Stimulating Factor Utilization for the Prevention of Chemotherapy-Induced Febrile
Neutropenia among Lung Cancer Patients in the United States........................................................................................................... 44
Abstract ........................................................................................................................................................................................... 44
Introduction ..................................................................................................................................................................................... 45
Methods ........................................................................................................................................................................................... 46
Results ............................................................................................................................................................................................. 48
Discussion ....................................................................................................................................................................................... 50
Conclusion ...................................................................................................................................................................................... 52
Figure and Table ............................................................................................................................................................................. 53
Reference ........................................................................................................................................................................................ 61
Appendix ......................................................................................................................................................................................... 65
Chapter 5 - Conclusion ........................................................................................................................................................................ 69
References ....................................................................................................................................................................................... 73

Page 3 of 73

Acknowledgment

This work would not have been possible without the financial support of the University of Southern
California Graduate Student Fellowship and Pharmedquest Research Fellowship. I am grateful to all of those with
whom I have had the pleasure to work during my dissertation and other related projects. Each of the members of
my Dissertation Committee has provided me extensive professional guidance in shaping my dissertation study.
I would especially want to thank Dr. Jeffrey McCombs, the chairman of my committee. As my teacher
and my mentor, he has taught me more than I could ever give him credits for here. His invaluable guidance
provided me with the tools and knowledge I needed to carry out my dissertation study. Also, without his
substantial assistance in the administrative work, I would not have been able to complete my dissertation and the
degree training successfully.
Nobody has been more important to me in the pursuit of academic achievement than the members of my
family. I would like to thank my parents, for their love and psychological and financial supports. Most
importantly, I want to thank my loving and supportive wife, Yi-Jen, for her patient listening and inexhaustible
inspiration. It will be impossible for me to complete my dissertation without her limitless support.

Page 4 of 73

Chapter 1 – Introduction

The past 20 years have seen a substantial increase in total drug spend in the US healthcare market,
especially for specialty drugs. (AARP Public Policy Institute, 2016; CVS Health; Diplomat Clinical Services,
2017; Penington andStubbings, 2016) To be designated as a specialty drug commonly requires that the drug is
either high cost, requires special handling during transportation and administration, requires patient monitoring
for safety and efficacy, requires a limited distribution network, or focuses on a rare disease. (Gleason et al., 2013)
Specialty drugs usually include self-injection products, agents infused in out-patient clinics or expensive oral
products. (Sullivan, 2008) Despite the various definition, the most common determination cited by insurance
payers is the high cost. (EMD Serono, 2013) The Medicare specialty tier threshold definition uses $600 per thirty-
days’ supply for payers to classify a drug into the specialty category. (Spatz, 2013) For some commercial plans,
the threshold was set ranged from $1000 to $1200. (Schondelmeyer, 2015)

Specialty drugs have attracted the attention of numerous global pharmaceutical companies and have thus
become the focus of increasing research investment. In 2006, only six specialty drugs gained approval by the
Food and Drug Administration (FDA) in the United States (US). The number increased quickly to 27 by 2014.
(PwC Health Research Institute, 2015) The sales of specialty drugs also have increased significantly in the US
market, accounting for $77.5 billion, or nearly 24% of drug spending in the US market in 2011. This sales
volume rose to $150.8 billion and 36% of the drug spending in 2015. (IMS Institute, 2016; IMS Institute for
Healthcare Informatics, 2015; UnitedHealth Center for Health Reform & Modernization, 2014) With the
expectation of steady growth in specialty drug market, predictions also showed the annual sales of specialty drugs
would increase to $402 billion and represent 47% of all medications by 2020. (IMS Institute, 2016; Truven Health
Analytics, 2016; UnitedHealth Center for Health Reform & Modernization, 2014) It is not surprising that the
utilization and pricing of specialty drugs has been subjected to efforts to manage price and utilization.

This dissertation study aimed to evaluate the price and clinical impact of selected examples of specialty
drug management in the US healthcare system. The first part of this dissertation is a published paper which
analyzed the specialty drug price in the 340B drug discount program. The 340B program allows eligible entities
purchasing drugs at a significant discounted, but commercially confidential price from manufactures. While the
number of 340B-eligible entity expends in the US vastly, the specialty drug price under the 340B program
becomes an issue concerned by policy makers, payers, and providers. To document the price trend of specialty
drug price in the 340B program, we used a purchasing record dataset from a regional 340B-contracted pharmacy
system to analyze the price trend in a 10-year timeframe.

Page 5 of 73

In the remaining parts of the dissertation considers the clinical impact of specialty drug management
efforts to control the utilization of granulocyte-colony stimulating factor (GCSF) which is used to prevent febrile
neutropenia (FN) cancer patients treated with chemotherapy. Specifically, breast cancer and lung cancer patients
frequently receive myelosuppressive chemotherapy which have significant risk of FN. GCSF is a costly specialty
drug used for the prevention of FN but which have frequent but serious side effect associated with
myelosuppressive chemotherapy administration. (Kuderer et al., 2006) Clinical guidelines have accepted the role
of GCSF prophylaxis to prevent chemotherapy-related FN. (Aapro et al., 2011; Freifeld et al., 2011; National
Comprehensive Cancer Network, 2017; Smith et al., 2015) However, less is to know about how the guideline
being implemented, and the effectiveness of GCSF prophylaxis on FN in real-world experience.

To address these questions, we analyzed the utilization and clinical/ economics outcomes of GCSF
prophylaxis among breast cancer and lung cancer patients who received myelosuppressive chemotherapy with a
10-year commercial health insurance claim dataset. A 10-year commercial health insurance claim dataset with
detailed patient-level social-economic status information will be used to identify clinical and economic outcomes
of using GCSF prophylaxis on FN episodes. We further investigate the association between patient’s social-
economic factors and the use of GCSF prophylaxis, to answer the policy question of how the prescribing of GCSF
prophylaxis being affected by non-clinical factors.

Page 6 of 73

REFERENCE
Aapro, M.S., Bohlius, J., Cameron, D.A., Lago, L.D., Donnelly, J.P., Kearney, N., Lyman, G.H., Pettengell, R.,
Tjan-Heijnen, V.C., Walewski, J., et al. (2011). 2010 update of EORTC guidelines for the use of granulocyte-
colony stimulating factor to reduce the incidence of chemotherapy-induced febrile neutropenia in adult patients
with lymphoproliferative disorders and solid tumours. Eur. J. Cancer 47, 8–32.
AARP Public Policy Institute (2016). Trends in Retail Prices of Brand Name Prescription Drugs Widely Used by
Older Americans, 2006 to 2015. Rx Price Watch Rep.
CVS Health What’s Special About Specialty?
Diplomat Clinical Services (2017). Specialty Drug Approvals: 2016 Highlights and 2017 Projections.
EMD Serono (2013). EMD Serono Specialty Digest
TM
, 9th Edition: Managed Care Strategies for Specialty
Pharmaceuticals.
Freifeld, A.G., Bow, E.J., Sepkowitz, K.A., Boeckh, M.J., Ito, J.I., Mullen, C.A., Raad, I.I., Rolston, K.V.,
Young, J.A.H., andWingard, J.R. (2011). Clinical practice guideline for the use of antimicrobial agents in
neutropenic patients with cancer: 2010 Update by the Infectious Diseases Society of America. Clin. Infect. Dis.
52.
Gleason, P.P., Alexander, G.C., Starner, C.I., Ritter, S.T., VanHouten, H.K., Gunderson, B.W., andShah, N.D.
(2013). Health plan utilization and costs of specialty drugs within 4 chronic conditions. J. Manag. Care Pharm. 19,
542–548.
IMS Institute (2016). Medicines Use and Spending in the U.S. A Review of 2015 and Outlook to 2020. IMS Inst.
Healthc. Informatics.
IMS Institute for Healthcare Informatics (2015). Medicines Use and Spending Shifts.
Kuderer, N.M., Dale, D.C., Crawford, J., Cosler, L.E., andLyman, G.H. (2006). Mortality, morbidity, and cost
associated with febrile neutropenia in adult cancer patients. Cancer 106, 2258–2266.
National Comprehensive Cancer Network (2017). Myeloid Growth Factors. NCCN Clin. Pract. Guidel. Oncol.
Penington, R., andStubbings, J.A. (2016). Evaluation of Specialty Drug Price Trends Using Data from
Retrospective Pharmacy Sales Transactions. J. Manag. Care Spec. Pharm. 22, 1010–1017.
PwC Health Research Institute (2015). Medical Cost Trend: Behind the Numbers 2016 (Chart Pack). Leadership.

Page 7 of 73

Schondelmeyer, S.W. (2015). Trends in Retail Prices of Specialty Prescription Drugs Widely Used by Older
Americans, 2006 to 2013. AARP Public Policy Institute, Rx Price Watch Rep.
Smith, T.J., Bohlke, K., Lyman, G.H., Carson, K.R., Crawford, J., Cross, S.J., Goldberg, J.M., Khatcheressian,
J.L., Leighl, N.B., Perkins, C.L., et al. (2015). Recommendations for the use of WBC growth factors: American
society of clinical oncology clinical practice guideline update. J. Clin. Oncol. 33, 3199–3212.
Spatz, I. (2013). Health Policy Brief: Specialty Pharmaceuticals.
Sullivan, S.D. (2008). The promise of specialty pharmaceuticals: are they worth the price? J. Manag. Care Pharm.
14, S3–S6.
Truven Health Analytics (2016). Specialty Pharmacy Trends. Heal. Leader’s Fact File.
UnitedHealth Center for Health Reform & Modernization (2014). THE GROWTH OF SPECIALTY
PHARMACY - Current trends and future opportunities.

Page 8 of 73

Chapter 2 - Specialty Drug Price Trends in the Federal 340B
Drug Discount Program

ABSTRACT
Background: The federal 340B Drug Discount Program provides access to significant drug price discounts for
healthcare organizations in the United States, serving a disproportional share of disadvantaged patients.

Objective: The objective of this paper is to analyze trends over a ten-year period (2006~2016) in the price of
specialty drugs contrasting the market price with the price paid under the 340B Program.

Methods: Pharmacy purchase records, including the 340B drug price and the wholesale acquisition cost (WAC),
were collected from a 340B-contract pharmacy group in Southern California between 2006 and 2016. Records
were used to calculate price changes in the annual average price paid. The average price was calculated as the
weighted price using purchasing volume for each year as weights. Separate time series of year-to-year price
changes were created by therapeutic class using the American Hospital Formulary Service (AHFS) Therapeutic
Classification system.

Results: The 340B price growth rate patterns were similar to the profile of the WAC prices over time across all
drug classes. The overall drug price growth rate per year over ten years for WAC prices was 15% and 10% for the
340B prices. For specialty drug classes the average growth rates per year were 14% for the WAC price and 6%
for the 340B price. For certain specialty drug classes, such as antineoplastic and antiretroviral drugs, the 340B
price inflation rates were significantly lower than the WAC price inflation rates after 2013.

Conclusion: The price inflation of specialty drugs exceeds the rate of inflation in the consumer price index in
prescription drugs. The 340B price shows a similar inflation pattern as the WAC price over time in the specialty
drug categories.

Page 9 of 73

INTRODUCTION
The past 20 years have seen a substantial increase in drug prices in the US healthcare market, especially
for specialty drugs. (AARP Public Policy Institute, 2016; CVS Health; Diplomat Clinical Services, 2017;
Penington andStubbings, 2016) To be designated as a specialty drug commonly requires that the drug is either
high cost, requires special handling during transportation and administration, requires patient monitoring for
safety and efficacy, requires a limited distribution network, or focuses on a rare disease. (Gleason et al., 2013)
Specialty drugs usually include self-injection products, agents infused in out-patient clinics or expensive oral
products. (Sullivan, 2008) Despite the various definition, the most common determination cited by insurance
payers is the high cost. (EMD Serono, 2013) The Medicare specialty tier threshold definition uses $600 per thirty-
days’ supply for payers to classify a drug into the specialty category. (Spatz, 2013) For some commercial plans,
the threshold was set ranged from $1000 to $1200. (Schondelmeyer, 2015)
Specialty drugs have attracted the attention of numerous global pharmaceutical companies and have thus
become the focus of increasing research investment. In 2006, only six specialty drugs gained approval by the
Food and Drug Administration (FDA) in the United States (US). The number increased quickly to 27 by 2014.
(PwC Health Research Institute, 2015) Researchers estimating that at least ten specialty drugs will obtain FDA
approvals and launch in the first quarter of 2018. (CVS Health, 2018) Meanwhile, the sales of specialty drugs
have increased significantly in the US market, accounting for $77.5 billion, or nearly 24% of drug spending in
the US market in 2011. This sales volume rose to $150.8 billion and 36% of the drug spending in 2015. (IMS
Institute, 2016; IMS Institute for Healthcare Informatics, 2015; UnitedHealth Center for Health Reform &
Modernization, 2014) With the expectation of steady growth in specialty drug market, predictions also showed the
annual sales of specialty drugs would increase to $402 billion and represent 47% of all medications by 2020.
(IMS Institute, 2016; Truven Health Analytics, 2016; UnitedHealth Center for Health Reform & Modernization,
2014)
Many pharmaceutical companies have shifted their research and development strategy from traditional
small-molecule “blockbuster” drugs to “niche buster” specialty drug, (Dolgin, 2010; Kumar Kakkar andDahiya,
2014; Song andHan, 2016) to offset the loss from the competition for drugs losing patent protection. Information
regarding the price of specialty drugs is becoming a major issue for healthcare stakeholders. Previous research has
estimated an inelastic demand for specialty drugs among patients compared to traditional drugs, (Goldman et al.,
2006) with studies also showing trends in drug prices among the specialty drugs domain upward. (Penington
andStubbings, 2016) Research documenting the price trends for specialty drugs relative to other drug prices in the
US is currently absent, however.

Page 10 of 73

There is no single reference drug price which is universally accepted in the US healthcare market. Four
published drug prices have been commonly used as the US reference price: the average wholesale price (AWP),
(Curtiss et al., 2010; Gencarelli, 2002; Salter, 2015) the average sales price (ASP), (Levinson, 2006, 2010; Salter,
2015) the average manufacturer price (AMP), (Levinson, 2005; Salter, 2015) and the wholesale acquisition cost
(WAC). (Curtiss et al., 2010; Gencarelli, 2002) Among these reference prices, the WAC price is the most
commonly accepted drug price benchmark in the US pharmaceutical market. (Curtiss et al., 2010) Another drug
price, the 340B price, has been less discussed.
The 340B drug price originates from the 340B Drug Discount Program, created by the US federal
government in 1992(Health Resources and Services Administration, 2017a) to promote the delivery of healthcare
services to the disadvantaged population.(Mulcahy et al., 2014) The 340B Program allows covered healthcare
entities, such as specialty care facilities and safety-net clinics, to purchase out-patient medications at a
significantly discounted price from WAC. This 340B drug price ceiling is set by law at least fifteen percent below
the AMP price for medications.(Health Resources and Services Administration, 2017b; Ruggeri andNolte, 2013;
Salter, 2015) Drug manufacturers who participate in the Medicaid program are required by the government to
provide drugs to covered entities at a cost no higher than the ceiling price in the 340B Program.
To become a 340B program covered entity, the healthcare system must serve a disproportionally large
number of Medicare/Medicaid beneficiaries or other low-income patients, and applied for the 340B program
eligibility. (Health Resources and Services Administration, 2017a) Covered entities can purchase medications
directly from manufacturers and wholesalers at the 340B price, and dispense these medications via entity-owned
pharmacies, contracted pharmacies, clinic dispensary or through outpatient clinics for physician-administrated
medicines. Alternatively, the covered entities can purchase the drugs from the manufacturers at the 340B price,
but dispense the medications through a network of 340B-contracted community pharmacies. (Mulcahy et al.,
2014) Covered entities are required to adhere to guidelines set forth by Health Resource and Services
Administration (HRSA) and are only allowed to dispense 340B drugs to eligible patients.
Several studies have discussed price trends for specialty drugs in the US healthcare market.
(Schondelmeyer, 2015; Stern andReissman, 2007; Trish et al., 2014) However, few studies address specialty drug
price inflation under the 340B Program. Our analysis documents the price inflation among specialty drugs in the
340B Program and compares these results to the corresponding price inflation data using the WAC price. We also
compare the 340B price inflation for specialty drug to the consumer price index in drug reported by the U.S
government, to visualize how the 340B specialty drug price changed over the regular drug price over time.

METHODS

Page 11 of 73

Medication purchasing records were collected from a 340B-contracted pharmacy network in the Southern
California area for the 2006–2016 period. Data included the National Drug Code (NDC), drug name, purchasing
quantity, AWP price value, 340B invoice price value and transaction dates. Because the WAC price was not
included in the purchasing records, we derived the WAC price based on the in-file AWP price with following
formula: (Curtiss et al., 2010)
WAC price value = AWP price value / 1.2
We used the American Hospital Formulary Service (AHFS)(American Hospital Formulary Service, 2017)
Therapeutic Classification code to classify the drugs for the analysis. The following formula for drug price per
therapeutic class was applied:
[Sum of (Drug Purchasing Quantity * Drug price) / Sum of Drug Quantity]
We calculated this volume-weighted drug price monthly and then calculated the average annual price by
therapeutic class. We then compared the current year drug price change relative to the previous year and reported
the annual price change as a percentage change by therapeutic class. For a sensitivity analysis, we utilized the
unweighted drug price values in both the WAC and 340B groups to examine whether the purchasing quantity
affected the drug price inflation calculations.
The First Databank Drug Database
©
was used to obtain the brand name status for all drugs in the
purchasing records by matching the NDC codes. We compared the average drug price inflation between the brand
name products and generic products in each reference price by the therapeutic class. While there are no consistent
criteria in defining specialty drugs, we select the drugs under the therapeutic classes that were commonly being
cited by specialty drug studies, and based on the purchase record availability. (Gleason et al., 2013; Goldman et
al., 2006; Hirsch et al., 2014; Kirchhoff, 2015; Penington andStubbings, 2016; Schondelmeyer, 2015; Spatz,
2013; Sullivan, 2008; Trish et al., 2014) Following therapeutic classes were selected: antiretroviral drugs, disease-
modifying antirheumatic drugs (DMARD), antineoplastic drugs, hematopoietic agents, interferons and
immunosuppressive agents. When comparing the overall 340B drug price and the WAC price, we excluded
certain drug classes due to the less therapeutic purpose, including dental agents, medical devices, diagnostic
agents, pharmaceutical aids, serums/toxoids/vaccines, and vitamins. The full AHFS codes used in the analysis are
listed in the Appendix.
To illustrate the trend of specialty drug price change relative to the general drug price, we implement the
consumer price index in the prescription drug from the U.S. Bureau of Labor Statistics (BLS) (U.S. Bureau of
Labor Statistics, 2017) to depict the general drug price inflation rate in the US between 2006 and 2016. The t-test
was applied to test the null hypodissertation at .1 and .05 level of whether the price inflation rates were the same

Page 12 of 73

between the WAC and 340B prices. Microsoft Excel
©
and SAS
©
9.3 were used for data preparation and analysis
process.

RESULTS
All drug price comparison:
We found no significant difference in the average price inflation between the WAC and 340B price values
for all drug classes across the 10-year study period (14.6% vs. 10.3%, p=0.39) (Table 1). For brand-name drugs,
both the WAC and 340B price values had a similar inflation trend, and both inflation rates were higher than the
general inflation, except for the years 2009, 2012 and 2014. For generic drugs, the price inflation trends were
similar between the WAC and 340B price values after 2007. Both the WAC and 340B price inflation rates were
lower than the general inflation rates between 2008 and 2010 for generic drugs.
Overall specialty drug price comparison:
The WAC price had a higher average inflation rate than the 340B price (14.1% vs. 6.4%, p > 0.1) (Table
1) over time for the overall specialty drugs. For the brand name drugs, the WAC price value showed a higher
average inflation rate than the 340B price value, although the difference was not significant (14.5% vs. 7.8%, p >
0.1) (Table 1). The inflation change patterns were similar between WAC and the 340B price values for brand
name specialty drugs (Graph 1). We found that for generic specialty drugs, the average 340B price inflation rate
was higher than the average WAC price inflation rate (22.9% vs. 19.6%, p > 0.1) (Table 1). The generic specialty
drugs have a trend of having higher inflation rates than the brand name specialty drugs in both WAC and 340B
price values. Moreover, we observed significant price inflation peaks in generic specialty drugs across time,
especially in 2008 and 2011 (Graph 1).
Antineoplastic drugs:
No statistically significant difference was found between the price inflation rates of the WAC and 340B
values across time (21.6% vs. 21.0%, p=0.99) (Table 1). For brand name antineoplastic drugs, the average price
inflation rates were the same between WAC and 340B price value. The generic antineoplastic drugs, meanwhile,
showed different average inflation rates between WAC and 340B price values, though the difference is
insignificant (Table 1). The inflation patterns between WAC and 340B price value were similar for both brand
name and generic antineoplastic drugs. We also observed a time-lagged association between the brand name and
generic drug price inflation rate in WAC and 340B price values (Graph 2).
Antiretroviral drugs:
The average WAC price growth rate was significantly higher than the average 340B price growth rate
across the target years (4.5% vs. -1.4%, p=0.05), regardless of the brand name status (Table 1). For brand name

Page 13 of 73

antiretroviral drugs, both the WAC and 340B price inflation rates were more stable than the generic drugs (Graph
3). The maximum price inflation rates were approximately 20% in brand name antiretroviral drugs compared to
350% in the generic drugs. The WAC price value had a slightly higher average inflation rate than the 340B
regarding the brand name products. The average inflation rate of the 340B price value was higher than that of the
WAC price value in the generic products (Table 1). Both the average 340B price and the average WAC price in
generic products experienced a high price peak in 2012 (Graph 3). Within both price values, we noted that the
generic antiretroviral drugs had a higher average inflation rate than the brand name products over the 10-year time
frame (Table 1, Graph 3), while the brand name products experienced relatively stable price inflation rates
overall. Such higher average inflation rate could result from the drug price peaks in 2010, 2012 and 2014.

DISCUSSION
The medication purchases processed through the 340B Program increased exponentially over the past
twenty years, from $3.4 billion in 2006 to $16.2 billion in 2016. (Fein, 2016, 2017; Vandervelde andBlalock,
2016) Such spending growth can be contributed by multiple factors, including the rising number of covered
entities, contracted pharmacies, increasing use of expensive specialty drugs, and the inflation of 340B drug price.
The number of covered entities increased from 15,765 to 50,741 from 2006 to 2016, while the number of
contracted pharmacies rose from 1,104 to 35,181, respectively. (Office of Pharmacy Affairs, 2017) According to
the IMS report in 2016,(IMS Institute for Healthcare Informatics, 2016) 58% of the outpatient drug spending is
driven by specialty drugs. Studies also have predicted a 40% annual increase in specialty drug purchases among
outpatient drugs. (Express Scripts, 2017) Since most 340B price-eligible drugs are outpatient drugs, the high and
fast growth spending on specialty drugs in the 340B program is inevitable.
In this article, we analyzed the specialty drug price inflation in WAC and 340B price value for following
therapeutic classes: antiretroviral drugs, antineoplastic drugs, disease modifying anti-rheumatic drugs,
hematopoietic agents, interferons and immunosuppressive agents, and then categorized the price inflation rates by
their brand name status. Our selected drug classes have included the majority of the commonly prescribed
specialty drugs in the US drug market. (Penington andStubbings, 2016) Therefore we believe our analysis result
can represent the overall specialty drug price in the US healthcare system.
To document the drug price inflation over time, we utilized purchasing quantity-weighed drug prices in
both the WAC and 340B price values. Using the quantity-weighted price can reduce the potential bias from the
rarely obtained drugs and make the drug price trends more comparable with the experience from real-world
pharmacy management operation. We carried out a sensitivity analysis by implementing unweighted WAC and

Page 14 of 73

340B price values and found no significant difference between the weighted and unweighted results among most
of the drug classes. (Table 1)
In general, the 340B price has a lower inflation rate than the WAC price across the ten years included in
this study. We did not observe significantly different price inflation patterns between the WAC and the 340B
prices. Contrary to common belief, our analysis revealed that the brand name products did not always link to
higher price inflation rates. For instance, generic antiretroviral drugs and generic immunosuppressive agents
possessed higher price inflation rates in both the WAC and 340B prices when compared to their brand name
counterparts. For the analysis of the overall specialty drugs, we also observed higher inflation rates among the
generic products than brand name products in both the WAC and 340B price values. Such an effect could be
attributed to the marketing strategy of the brand name product manufacturers, who would use the 340B price
values to compete with the generic products once the latter became available on the market. For instance, we
observed a decrease in the 340B price inflation rates among the brand antiretroviral drugs after the generics, such
as abacavir (QuintilesIMS, 2014) and nevirapine (QuintilesIMS, 2013), began to emerge on the market in 2013.
Alternatively, rather than merely representing a healthcare policy tool, the 340B Program could also be applied as
a tool by manufacturers to increase market share amongst therapeutic areas with multiple comparative products.
Among the sub-class of the specialty drugs we analyzed, some of the drug classes demonstrated a time-
lagged relationship for the inflation rates among the brand name and generic products, such as the antineoplastic
agents. The pattern usually started with an inflation peak in the brand name products, followed by the generic
products. This observation implies that the brand name manufacturers maintained their role as the drug pricing
leader in the antineoplastic drug market over the 10-year study period.
Finally, we found both the WAC and 340B price inflation rates were higher than the prescription price
index reported by the U.S. Bureau of Labor Statistics (BLS). The BLS consumer price index for prescription
drugs may not adequately represent the drug price faced by the healthcare entity either. Based on the methodology
document, (U.S. Bureau of Labor Statistics, 2015) BLS sampled the drug prices from all healthcare service
providing sites, including hospitals, pharmacies and physician clinics, but only estimated the price at a consumer
out-of-pocket cost basis. The drug price analyzed in our study, nevertheless, is at the entity prospect instead. Since
patient perceived different drug cost compared to the provider, (Reinhardt, 2006) the substantial gap between the
drug price inflation rates reported in our analysis and the rates reported by BLS, is expected. Yet this finding also
suggests the current price index system is unable to reflect the purchase cost and the market price inflation for
specialty drugs. A separate price index reporting system for specialty drugs in the US is thus recommended for
health policy stakeholders and researchers in the drug policy analysis field.

Page 15 of 73

This analysis has several limitations. First, our analysis did not encompass all specialty drugs on the
market. We implemented a convenient method to locate specialty drugs, which is to include drugs from the
therapeutic classes commonly cited in the specialty drug studies, because of the lack of consensus on specialty
drug definition. These therapeutic classes, in general, have most of the high-cost, monitoring-required, and
injection/ infusion drugs (Goldman et al., 2006; Penington andStubbings, 2016; Sullivan, 2008), and are
distributed by specialty pharmacies. (EMD Serono, 2018; Express Scripts, 2018; IMS Institute for Healthcare
Informatics, 2016; Magellan Rx Management, 2018; Schondelmeyer, 2015) Some of the specialty drug products
might be missed from our analysis due to the purchasing record availability as well. (Table 1) Second, our
analysis results may not be applied to other therapeutic class systems, such as the Medi-Span Product Line
Generic Product Identifier (GPI) (Agency for Healthcare Research and Quality, 2014) or the Anatomical
Therapeutic Chemical Classification System (ATC). (WHO Collaborating Center for Drug Statistics
Methodology, 2018) To select the specialty drugs and to broaden the view of the drug price composition under
different drug pricing systems, we employed the AHFS therapeutic class data in the purchase records to
categorize the drug price. Since different drug classification systems may have different definitions in the
therapeutic classes, our analysis result might not be generalized to the therapeutic classification system other than
AHFS.
In addition, while some drugs might have different drug class codes in different years, our analysis did
not trace the code modification across time. Therefore, some of the average drug price inflation estimations may
be less accurate. The therapeutic class mixed with non-specialty drugs could bias the result as well. We used the
therapeutic class to group specialty drugs rather than applying a hand-picked drug list for the analysis. Some
specialty classes, such as the DMARDs, are a mixture of biological agents and traditional small molecule agents.
The price inflation estimation in such drug class could be skewed as a result. Finally, the pricing database we
utilized was based on the drug purchasing records of the 340B covered entities. We limited our analysis by the
real-world drug purchasing decisions as opposed to the wholesaler’s total drug catalog file to reflect the real-
world market situation of specialty drugs. Additionally, covered entities are bound to the scope of their HRSA
grants, patient population mix, and 340B patient eligibility. Some drugs might not be included in the analysis and
thus biased the result.
Further analysis is required to evaluate whether the 340B Program affects the patients’ and providers’
behavior in specialty drugs utilization, such as medication adherence and treatment selection. Additionally, it is
pertinent to investigate whether the 340B Program affects the overall healthcare utilization and access for the
disadvantaged population and fulfills the intent of the 340B Program. We also suggest a nationwide 340B
Program database that features the entities’ dataset, including medical claims and pharmacy claims, to allow
future researchers to investigate the health outcomes associated with the 340B Program.

Page 16 of 73

CONCLUSION
Our analysis presents the specialty drugs’ price trend in WAC and 340B from 2006 to 2016. The result
suggests the above-normal price inflation among specialty drugs across the time frame of focus. The traditional
consumer price index of the drug cannot satisfy the need of describing the price change in specialty drugs. A
specified drug price index system is required to reflect specialty drug price trend in the US market. Our analysis
also showed the 340B price value had a similar trending pattern to the WAC price across time in specialty drug
categories. The brand name status, therapeutic class, and marketing strategy potentially affect the price trend of
specialty drugs in the US. Consequently, a national 340B entity dataset, including medical claims and pharmacy
claims, is recommended for facilitating more comprehensive analyses regarding the health outcomes of the 340B
Drug Discount Program. Our study provides specialty drug management entities an insight into drug price trend
for specialty drugs. With the high pharmacy spending in specialty drug purchase, understanding the price trend in
specialty drug will help healthcare system allocate the resource more efficiently.

Page 17 of 73

REFERENCES
AARP Public Policy Institute (2016). Trends in Retail Prices of Brand Name Prescription Drugs Widely Used by
Older Americans, 2006 to 2015. Rx Price Watch Rep.
Agency for Healthcare Research and Quality (2014). Product Code Qualifier.
American Hospital Formulary Service (2017). AHFS Pharmacologic-Therapeutic Classification.
Curtiss, F.R., Lettrich, P., andFairman, K. a (2010). What is the price benchmark to replace average wholesale
price (AWP)? J. Manag. Care Pharm. 16, 492–501.
CVS Health What’s Special About Specialty?
CVS Health (2018). Specialty Pharmacy Pipeline: Anticipated Launches – 1th Quarter 2018 to 2nd Quarter 2018.
Drugs to Watch.
Diplomat Clinical Services (2017). Specialty Drug Approvals: 2016 Highlights and 2017 Projections.
Dolgin, E. (2010). Big pharma moves from “blockbusters” to “niche busters”. Nat. Med. 16, 837.
EMD Serono (2013). EMD Serono Specialty Digest
TM
, 9th Edition: Managed Care Strategies for Specialty
Pharmaceuticals.
EMD Serono (2018). EMD Serono Specialty Digest, 14th Edition.
Express Scripts (2017). The 2016 Drug Trend Report. Drug Trend Rep.
Express Scripts (2018). The 2017 Drug Trend Report. Drug Trend Rep.
Fein, A.J. (2016). 340B Purchases Hit $12 Billion in 2015—and Almost Half of the Hospital Market.
Fein, A.J. (2017). EXCLUSIVE: The 340B Program Hits $16.2 Billion in 2016; Now 5% of U.S. Drug Market.
Gencarelli, D.M. (2002). Average wholesale price for prescription drugs: is there a more appropriate pricing
mechanism? NHPF Issue Brief 1–19.
Gleason, P.P., Alexander, G.C., Starner, C.I., Ritter, S.T., VanHouten, H.K., Gunderson, B.W., andShah, N.D.
(2013). Health plan utilization and costs of specialty drugs within 4 chronic conditions. J. Manag. Care Pharm. 19,
542–548.
Goldman, D.P., Joyce, G.F., Lawless, G., Crown, W.H., andWilley, V. (2006). Benefit design and specialty drug
use. Health Aff. 25, 1319–1331.
Health Resources and Services Administration (2017a). 340B Drug Pricing Program.
Health Resources and Services Administration (2017b). 340B Drug Pricing Program Ceiling Price and
Manufacturer Civil Monetary Penalties Regulation (Federal).
Hirsch, B.R., Balu, S., andSchulman, K. a (2014). The impact of specialty pharmaceuticals as drivers of health
care costs. Health Aff. (Millwood). 33, 1714–1720.
IMS Institute (2016). Medicines Use and Spending in the U.S. A Review of 2015 and Outlook to 2020. IMS Inst.
Healthc. Informatics.

Page 18 of 73

IMS Institute for Healthcare Informatics (2015). Medicines Use and Spending Shifts.
IMS Institute for Healthcare Informatics (2016). Medicines Use and Spending in the U.S.: A Review of 2016 and
Outlook to 2021.
Kirchhoff, S.M. (2015). Specialty drugs: Background and policy concerns. Congr. Res. Serv.
Kumar Kakkar, A., andDahiya, N. (2014). The evolving drug development landscape: from blockbusters to niche
busters in the orphan drug space. Drug Dev. Res. 75, 231–234.
Levinson, D.R. (2005). Medicaid Drug Price Comparisons : Average Manufacturer Price to Published Prices.
Levinson, D.R. (2006). Calculation of Volume-Weighted Average Sales Price for Medicare Part B Prescription
Drugs.
Levinson, D.R. (2010). Average Sales Prices : Manufacturer Reporting and CMS Oversight.
Magellan Rx Management (2018). 2017 Medical Pharmacy Trend Report.
Mulcahy, A.W., Armstrong, C., Lewis, J., andMattke, S. (2014). The 340B Prescription Drug Discount Program:
Origins, Implementation, and Post-Reform Future.
Office of Pharmacy Affairs (2017). 340B Covered Entity and Contract Pharmacy Daily Report.
Penington, R., andStubbings, J.A. (2016). Evaluation of Specialty Drug Price Trends Using Data from
Retrospective Pharmacy Sales Transactions. J. Manag. Care Spec. Pharm. 22, 1010–1017.
PwC Health Research Institute (2015). Medical Cost Trend: Behind the Numbers 2016 (Chart Pack). Leadership.
QuintilesIMS (2013). Drug In Focus: Nevirapine.
QuintilesIMS (2014). Drug In Focus: Abacavir.
Reinhardt, U.E. (2006). The pricing of U.S. hospital services: chaos behind a veil of secrecy. Health Aff.
(Millwood). 25, 57–69.
Ruggeri, K., andNolte, E. (2013). Pharmaceutical pricing: The use of external reference pricing. RAND Corp.
Salter, M. (2015). Reference Pricing: An Effective Model for the U.S. Pharmaceutical Industry? Northwest. J. Int.
Law Bus. 35, 413.
Schondelmeyer, S.W. (2015). Trends in Retail Prices of Specialty Prescription Drugs Widely Used by Older
Americans, 2006 to 2013. AARP Public Policy Institute, Rx Price Watch Rep.
Song, C.H., andHan, J.-W. (2016). Patent cliff and strategic switch: exploring strategic design possibilities in the
pharmaceutical industry. Springerplus 5, 692.
Spatz, I. (2013). Health Policy Brief: Specialty Pharmaceuticals.
Stern, D., andReissman, D. (2007). Specialty pharmacy cost management strategies of private health care payers.
J. Manag. Care Pharm. 12, 736–744.
Sullivan, S.D. (2008). The promise of specialty pharmaceuticals: are they worth the price? J. Manag. Care Pharm.
14, S3–S6.

Page 19 of 73

Trish, E., Joyce, G., andGoldman, D.P. (2014). Specialty drug spending trends among medicare and medicare
advantage enrollees, 2007-11. Health Aff. 33, 2018–2024.
Truven Health Analytics (2016). Specialty Pharmacy Trends. Heal. Leader’s Fact File.
U.S. Bureau of Labor Statistics (2015). BLS Handbook of Methods, Chapter 17. The Consumer Price Index
(Updated 6/2015).
U.S. Bureau of Labor Statistics (2017). Consumer Price Index.
UnitedHealth Center for Health Reform & Modernization (2014). THE GROWTH OF SPECIALTY
PHARMACY - Current trends and future opportunities.
Vandervelde, A., andBlalock, E. (2016). 340B PROGRAM SALES FORECAST: 2016 - 2021. Berkeley Res. Gr.
WHO Collaborating Center for Drug Statistics Methodology (2018). Anatomical Therapeutic Chemical
Classification System.

Page 20 of 73

FIGURE & TABLE
▪ Table 1, Overall Price Inflation Rate Comparison:
Drug Class
Average
WAC
(Weighted)
Average 340B
(Weighted)
Difference
Difference,
Weighted
and
Unweighted
WAC
Difference,
Weighted
and
Unweighted
340B
Brand Generic
Average
WAC
Average
340B
Difference
Average
WAC
Average
340B
Difference
Overall Drug 14.7% 10.3% 4.5% -1.3% -0.1% 23.2% 15.0% 8.2% 10.4% 5.1% 5.3%
Overall - Specialty drugs only 14.1% 6.4% 7.8% 1.5% -0.5% 14.5% 7.8% 6.7% 19.6% 22.9% -3.3%
Antiretroviral 4.5% -1.4% 5.9%* -1.1% -0.1% 7.1% 2.4% 4.7% 23.2% 38.3% -15.0%
Antineoplastic 21.1% 21.0% 0.1% 1.2% -0.3% 32.7% 33.0% -0.2% 18.2% 33.1% -14.9%
DMARD 19.1% -3.2% 22.3% 7.8% 5.4% 21.7% -1.2% 22.9% --† --† --†
Hematopoietic Agents 12.3% 3.3% 9.0% 4.5% -1.3% 12.3% 3.3% 9.0% --† --† --†
Interferon -11.0% -11.5% 0.5% 0.6% 0.0% -11.0% -11.4% 0.5% --† --† --†
Immunosuppressive Agents 95.0% 111.0% -16.0% -55.4% -10.9% 28.7% -19.1% 47.8% 99.0% 169.4% -70.4%
*: p < 0.10; **: p < 0.05; †: No purchasing records found in dataset

Page 21 of 73

▪ Graph 1: Specialty Drug Price Inflation Rate Comparison

▪ Graph 2: Antineoplastic Drug Price Inflation Rate Comparison

▪ Graph 3: Antiretroviral Drug Price Inflation Rate Comparison

-50.00%
0.00%
50.00%
100.00%
150.00%
200.00%
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Brand name, WAC Brand name, 340B Generic, WAC
Generic, 340B Inflation (prescription drugs)
-100.00%
0.00%
100.00%
200.00%
300.00%
2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Brand name, WAC Brand name, 340B Generic, WAC
Generic, 340B Inflation (Prescription drugs)
-100.00%
0.00%
100.00%
200.00%
300.00%
400.00%
2009 2010 2011 2012 2013 2014 2015 2016
Brand name, WAC Brand name, 340B Generic, WAC
Generic, 340B Inflation (Prescription drugs)

Page 22 of 73

APPENDIX
Table A: AHFS codes for drug class inclusion / exclusion
AHFS Drug Class, Inclusion AHFS Drug Class, Exclusion
ANTINEOPLASTIC AGENTS DENTAL AGENTS
10000000 34000000
ANTIRETROVIRALS DEVICES
81808000 94000000
81808080 DIAGNOSTIC AGENTS
81808160 36040000
81808200 36260000
DISEASE-MODIFYING ANTIRHEUMATIC 36580000
92360000 36600000
HEMATOPOIETIC AGENTS 36840000
20160000 36882400
INTERFERONS PHARMACEUTICAL AIDS
81820000 96000000
IMMUNOSUPPRESSIVE AGENT SERUMS, TOXOIDS, AND VACCINES
92440000 80040000

80080000

80120000

VITAMINS

88040000

88080000

88120000

88160000

88200000

88240000

88280000

Page 23 of 73

Table B: Specialty Drug List
Antiretroviral Antiretroviral (Cont.) Antineoplastic (Cont.) Antineoplastic (Cont.)
Brand Generic Brand Generic
APTIVUS ABACAVIR TORISEL VINBLASTINE
ATRIPLA DIDANOSINE DR TREXALL VINCRISTINE
COMBIVIR LAMIVU/ZIDOVU TRISENOX VINORELBINE
COMPLERA LAMIVUDINE VELCADE DS Immunosuppressive Agents
CRIXIVAN NEVIRAPINE VINCASAR PFS Brand
DESCOVY ZIDOVUDINE XELODA CELLCEPT
EDURANT Antineoplastic ZOLADEX SAFE GENGRAF
EMTRIVA Brand ZYTIGA IMURAN
EPIVIR ABRAXANE Antineoplastic MYFORTIC
EPZICOM ADCETRIS DS Generic NEORAL
EVOTAZ TAB ADRIAMYCIN ANASTROZOLE Generic
FUZEON ADRUCIL SDV BICALUTAMIDE AZATHIOPRINE
GENVOYA AFINITOR ASD BLEOMYCIN CYCLOSPOR (MODIFIED)
INTELENCE ALIMTA CARBOPLATIN CYCLOSPORINE
INVIRASE ALKERAN CISPLATIN MYCOPHENOLATE MOFETIL
ISENTRESS ARIMIDEX CLADRIBINE MYCOPHENOLIC ACID
KALETRA AROMASIN CYCLOPHOSPHAM SIROLIMUS
LEXIVA AVASTIN CYTARABINE TACROLIMUS
NORVIR CAMPTOSAR DACARBAZINE Hematopoietic Agents
ODEFSEY CASODEX DOCETAXEL Brand
PREZCOBIX COSMEGEN DOXORUBICIN ARANESP
PREZISTA CYRAMZA ASD DOXORUB-LIPO EPOGEN
RETROVIR ELOXATIN AQU EPIRUBICIN GRANIX
REYATAZ ERBITUX ETOPOSIDE LEUKINE
SELZENTRY FASLODEX EXEMESTANE NEULASTA
STRIBILD FEMARA FLUOROURACIL NEUPOGEN
SUSTIVA GLEEVEC GEMCITABINE PROCRIT
TIVICAY HALAVEN ASD HYDROXYUREA Interferons
TRIUMEQ HERCEPTIN IDARUBICIN Brand
TRIZIVIR IDAMYCIN P/F IFOSFAMIDE INFERGEN
TRUVADA LUPRON DEPOT IRINOTECAN INTRON A
VIDEX EC MEGACE ES LETROZOLE INTRON A 18
VIRACEPT NEXAVAR ASD MEGESTROL INTRON A HSAF
VIRAMUNE OPDIVO ASD MERCAPTOPURIN PEGASYS
VIREAD PERJETA ASD METHOTREXATE PEGASYS 180
ZERIT RITUXAN MITOMYCIN PEG-INTRON
ZIAGEN TASIGNA ASD MITOXANTRONE PEG-INTRON RP
Antiretroviral TAXOTERE OXALIPLATIN

Generic TEMODAR PACLITAXEL

ABAC/LAMI/ZID TOPOSAR TAMOXIFEN

24

Chapter 3 - An Analysis of Granulocyte-Colony-Stimulating
Factor Utilization for the Prevention of Chemotherapy-Induced
Febrile Neutropenia among Breast Cancer Patients in the United
States

ABSTRACT
Background: Febrile neutropenia (FN) is a common immune suppressive condition and a severe side-effect
associated with myelosuppressive chemotherapy treatment. Granulocyte colony-stimulating factors (GCSF) are
used to prevent or ameliorate chemotherapy-induced FN. Clinical trial studies have shown that GCSF prophylaxis
significantly reduced the risk of FN in oncology patients receiving myelosuppressive chemotherapies. The
objective of this paper is to evaluate the real-world effectiveness of GCSF for FN prophylaxis in patients with
breast cancer.

Methods: A commercial health insurance database covering January 2007 to September 2016 is used for the
analysis. The analysis includes only newly diagnosed chemotherapy-naive breast cancer patients. A logistic
regression model was used to estimate the association between patient/treatment characteristics and the use of
GCSF prophylaxis. A multivariate Cox-regression model was estimated to test the association between the
emergence of FN and GCSF prophylaxis. Sensitivity analysis was carried out to test alternative model
specifications across FN-related outcomes.

Results: GCSF prophylaxis is significantly associated with lower risks in FN (HR=0.24, P <0.01) and mortality
(60-day mortality; HR=0.22, P<0.01) events in the first chemotherapy duration compared to the non-GCSF
prophylaxis cohort. The prevention effect is especially significant among the high-risk chemotherapy regimens
(HR=0.19, P <0.01). No significant association was found between the timing of prophylaxis initiation and the FN
event. GCSF prophylaxis had no significant prevention effects on infection-related hospitalization events.

Conclusion: GCSF prophylaxis may effectively prevent FN event among breast cancer patients who received
chemotherapy. There were no significant benefits of GCSF prophylaxis on infection and hospitalization events.

25

INTRODUCTION
Febrile neutropenia (FN) is defined as absolute neutrophil count (ANC) <1000/mm
3
with concomitant
systemic temperature >38.3°C, or a sustained body temperature of ≥38°C for more than 1 hr. (National Cancer
Institute, 2009) Due to the life-threatening nature of FN, patients often require an extended hospital stay, leading
to increased mortality and higher healthcare cost. (Kuderer et al., 2006) Oncologists can reduce the risk of
myelosuppression by either modifying the chemotherapy interval or reduce the chemotherapy dose, both of which
may reduce chemotherapy treatment response and decreased survival in some malignancies. (Wildiers and Reiser,
2011) Prophylactic use of granulocyte colony-stimulating factors (GCSF) has been shown to prevent or
ameliorate chemotherapy-induced FN, (Bow, 2011; Timmer-Bonte et al., 2005; Wingard et al., 2012) yet the
evidence concerning the impact of GCSF on FN-associated hospitalization has been inconsistent. (Choi et al.,
2014; Potosky et al., 2011; Ramsey et al., 2010; Waters et al., 2013; Wright et al., 2013)
The substantial cost of GCSF imposes a significant burden on healthcare payers and patients and thus is a
significant contributor to the total cost of cancer-related healthcare service in the US. (Magellan Rx Management,
2018; 2017) GCSF accounted for 10% of the total oncology drug expenditures among both commercial and
Medicare plans in 2017. (Magellan Rx Management, 2018) Evidence-based guidelines have been developed for
GCSF prophylaxis to prevent chemotherapy-related FN (Aapro et al., 2011; Freifeld et al., 2011; National
Comprehensive Cancer Network, 2017; Smith et al., 2015). These guidelines commonly stratified FN risk
according to patient risk factors and the myelosuppressive nature of the chemotherapy regimen. These regimens
are classified as high (≥20% of FN incidence rate), intermediate (10–20%), or low (<10%) potential for the
development of FN. Primary GCSF prophylaxis is recommended for patients with high-FN risk chemotherapy
regimens and is not recommended for low-FN risk regimens users. The impact of GCSF on economic outcomes
remained inconclusive. (Barnes et al., 2014a; Dulisse et al., 2013; Kawatkar et al., 2017)
The evidence is incomplete concerning GCSF’s effectiveness in real-world clinical practice. Clinical
guidelines are based on well-controlled clinical trials, which may differ significantly from real-world practice.
(Truong et al., 2016) Retrospective studies of GCSF’s effectiveness focused on patients aged >65 years, despite
the increasing incidence of cancers in patients aged <50 years.(Campos, 2017; Johnson et al., 2013) Although age
represents an essential factor impacting bone marrow recovery, GCSF’s effectiveness among younger patients is
not delineated in the literature. Some studies also suggested the timing of implementing GCSF may be critical to
its effectiveness as FN prophylaxis on high-FN risk regimen patients. (Weycker et al., 2016, 2018) The evidence
for the impact of time to GCSF prophylaxis on other FN risk regimen patients is required.
This study’s objective is to evaluate the real-world effectiveness of GCSF in preventing FN in
chemotherapy-naï ve breast cancer patients using health insurance claims collected over a 10-year period. We

26

restricted our analysis to the first treatment duration and evaluated the factors associated with GCSF prophylaxis
use. We further analyzed the effectiveness of GCSF prophylaxis in FN prevention controlling for
patient/treatment characteristics. Alternative model specifications and FN-related outcomes, including infection,
hospitalization, and mortality events, were analyzed as well.

METHODS
Patient Selection and Chemotherapy Index Day:
Our analytic data were sourced from De-identified Clinformatics® Data Mart (OptumInsight, Eden
Prairie, MN). The dataset included 51,988,191 enrollees with their full medical claims covering Jan-2007 to Sept-
2016. We located all female breast cancer patients through ICD-9CM code 174 or ICD-10-CM codes C5001,
C5011, C5021, C5031, C5041, C5051, C5061, C5081, and C5091 (For specific codes see the Appendix). Patients
with <180-days of continuous enrollment prior to the first day of breast cancer diagnosis were removed. We also
excluded patients with the history of bone marrow or organ transplantation from the analysis. To prevent
including breast cancer patients who may use off-label chemotherapy (e.g., bevacizumab for ophthalmology
diseases), patients with macular edema, proliferative diabetic retinopathy, retinal vein occlusion, or age-related
macular degeneration were also excluded.
We located chemotherapy and GCSF claims through HCPCS Level 2 codes J9000–J9999 with additional
HCPCS “C” codes. (Centers for Medicare & Medicaid Services, 2006; National Institutes of Health, 2017) The
first day of chemotherapy with a breast cancer diagnosis was then set as the index day. Patients used GCSF before
the index day were excluded. We defined the patient’s treatment regimen as the combination of all chemotherapy
drugs used during the treatment period defined as continuous treatment up to the emergence of a 60-day gap in
chemotherapy drug use. Patients with chemotherapy claims prior to the date of first breast cancer diagnosis were
removed as potential chemotherapy off-label users.
FN-risk Regimen:
The risk classification for the patient’s initial chemotherapy regimen was defined using data from NCCN
Clinical Practice Guidelines in Oncology for Myeloid Growth Factors. (National Comprehensive Cancer
Network, 2017) Since only high- and intermediate-risk regimens were defined in the NCCN guidelines, we
manually searched for low-risk chemotherapy regimens for breast cancer, which were reviewed by a group of
oncologists and pharmacists specialized in breast cancer chemotherapy care. For the chemotherapy regimens with
dose-dense variants, such as doxorubicin and cyclophosphamide, we used the highest FN risk among the variants.
FN Prophylaxis:

27

We defined GCSF prophylaxis as GCSF use prior to an FN event incidence. Patients initiating GCSF on
the same day as the emergence of FN or later, and those never using GCSF, were categorized as the non-
prophylaxis GCSF group. We included pre-chemotherapy intravenous (IV)-antibiotic prophylaxis in the analysis
if the patient had used IV antibiotics within 14 days before chemotherapy initiation and without concomitant FN
and infection events.
Outcomes:
An FN event was identified if the concomitant diagnosis codes for neutropenia (288.x in ICD-9; D70 in
ICD-10) and fever (780.6x in ICD-9; R502x, R508x, R509x in ICD-10) are reported. We broadened our outcome
definition to include microbiology infections using diagnosis and DRG codes. The date of the first post-index
hospital admission was recorded separately for infection-related hospitalizations and all-cause hospitalizations
based on the DRG codes (Appendix). Mortality following chemotherapy initiation was measured using the
monthly death records provided by the insurance dataset. Detailed codes utilized to identify FN and other
outcomes are listed in the Appendix.
Analytic Methods:
A logistic regression model was estimated the factors associated with the use of GCSF prophylaxis,
including patient/treatment characteristics. A Cox-regression model was estimated to document the temporal
association between GCSF prophylaxis and FN events and other health outcomes controlling for patient/treatment
factors and time trends. To capture the treatment year effect on the outcomes, we transformed the treatment year
into a time trend variable, where the year 2007 was coded t=0, increasing to t=10 for a patient treated in 2016.
The odds ratio/hazard ratio for this time trend variable measures the average yearly trend in the outcome variable.
All outcomes were censored on the final day of receiving chemotherapy treatment in the initial course of
treatment, except for all-cause mortality (Appendix - Figure 1). For all-cause mortality, we considered the
censored death event at 60-/90-day time intervals post-chemotherapy initiation due to the HIPPA requirements
that restrict claim dataset’s death dates to the year and month. Therefore, we hypothetically set the death date to
the last day of the death month to construct the time-to-death estimates.
Several sensitivity analyses were conducted. We re-estimated our core models using interaction terms of
between GCSF prophylaxis and the FN risk regimens. We also established the effect of GCSF prophylaxis by the
days of delay since the initiation of chemotherapy. This specification estimates the effects of using GCSF
prophylaxis at the day of chemotherapy initiation (index day), 24-hr post-chemotherapy initiation (day+1), 48-hr
later (day+2), 72-hr later (day+3) and 96-hr or more post-chemotherapy initiation (day+4 and above). Alternative
models were estimated for all outcome measures. All the statistical models were estimated using SAS
©
9.4.

28

RESULTS
Patients
Figure 1 provides data on the patient identification process. A total of 452,157 female breast patient
cancer patients were identified, which translates into a 1.7% prevalence in the female population in the database.
This prevalence reported is consistent with the rate of 2.0% reported in the SEER Program. (National Cancer
Institute) Most excluded patients (86%) did not use chemotherapy, suggesting they were in early stages (I/IIa).
This rate is higher than the national statistics of 66.2%, which likely reflects a younger patient population
typically found in the commercial health plans. (National Cancer Institute)
Patient/ Treatment Characteristics
Table 1 shows patient characteristics of the analytic samples. A total of 17,265 female patients met the
inclusion and exclusion criteria for this analysis. For patients using chemotherapy, 39.5% used GCSF
prophylaxis. Most patients were aged 50–64 years in both GCSF (40.7%) and non-prophylaxis groups (40.7%).
Chronic pulmonary disease was the most common comorbidity among both groups (15.9% vs. 16.1%,
respectively; not significant), with no significant comorbidity distribution differences observed between the
prophylaxis and non-prophylaxis groups, except for congestive heart failure (2.3% vs. 3.1%, respectively;
P<0.01).
Table 2 presents the treatment characteristics identified in the analytic dataset. The high-FN-risk
chemotherapy regimen is the most commonly employed in both the prophylaxis (70.3%) and non-prophylaxis
groups (54.1%). For the prophylaxis group, most GCSF prophylaxis was used on day+1 (65.2%), while 49.7% of
the non-prophylaxis group patients eventually used GCSF, most likely to treat emerging FN. Surgical
interventions are more frequently used than radiological interventions prior to the initiation of chemotherapy,
suggesting most of the patients were with localized diseases and using adjuvant chemotherapies.
Factors Correlated with the Use of GCSF Prophylaxis
Figure 2 summarizes the patient characteristics associated with the use of GCSF as prophylaxis. Patients
using low FN-risk regimens (OR=0.13, P<0.05), radiotherapy (OR=0.79, P<0.05), having surgery (OR=0.38,
P<0.05) or with a pre-chemotherapy FN experience (OR=0.79, P<0.05) were less likely to use GCSF prophylaxis.
Patient age between 50 and 65 (OR=0.92, P<0.05) or above age ≥ 65 (OR=0.88, P>0.05) are significantly less
likely to use GCSF prophylaxis. Patients with pre-chemotherapy IV antibiotic prophylaxis (OR=1.44, P<0.05) and
insurance coverage with a preferred provider organization (PPO) insurance plan (OR=1.38, P<0.05) are
significantly correlated with a higher probability of using GCSF prophylaxis. GCSF prophylaxis usage increased
by about 7% per year over the study period.

29

The Core Model of FN Risk:
Figure 3 presents the results of the core Cox-regression model. GCSF prophylaxis is associated with a
significant reduction in the FN event (HR=0.24, P<0.01). The intermediate-, low- and unknown-risk regimens all
have significantly lower risks of an emerging FN event than patients using the high-risk regimen (HR=0.68, 0.26,
0.65; P<0.01 for all variables, respectively). These results are consistent with the expected risk tendency of
chemotherapy regimens reported in the clinical guidelines.
Pre-chemotherapy surgical intervention is related to a significantly increased risk of FN (HR=1.45,
P<0.01). The pre-chemotherapy radiotherapy experience is significantly associated with decreased risks of FN.
The risk reduction effects related to radiotherapy could reflect more careful neutropenia management among
radiologic therapy patients. Pre-chemotherapy FN experience and pre-chemotherapy IV-antibiotic prophylaxis are
associated with significant FN risk increases (HR=1.38, P<0.01; HR=1.21, P<0.01, respectively), suggesting that
treating physician have responded to a higher baseline FN risks among these patients.
Sensitivity Analysis, GCSF Prophylaxis:
Table 3 summarizes the sensitivity analyses results with the GCSF prophylaxis effects on all outcomes
with different model specifications (see Appendix for details). The core model showed using GCSF prophylaxis
reduced the overall risks of FN, FN-related hospitalization and both 60- and 90-day all-cause mortality (HR=0.24,
P<0.05; HR=0.80, P<0.05; HR=0.22, P<0.05; HR=0.23, P<0.05, respectively). The interaction term model
reveals using GCSF on patients with high-risk regimens effectively reduced the risks of FN, FN-related
hospitalization, infection-related hospitalization, and 60-/90-day all-cause mortality (HR=0.19, P<0.05; HR=0.65,
P<0.05; HR=0.79, P<0.05; HR=0.11, P<0.05; HR=0.27, P<0.05, respectively). No significant risk-reductions
were found of using GCSF prophylaxis with the low-risk regimens across outcomes, nor were significant effect
differences found between the different GCSF prophylaxis timing.

DISCUSSION
In this study, we estimate the effect of GCSF prophylaxis in preventing FN in breast cancer patients
receiving chemotherapy using a health insurance claim dataset. Our analysis confirmed the FN risk gradient
among the chemotherapy regimens and showed that GCSF prophylaxis reduced the risks in FN, FN-related
hospitalization, infection-related hospitalization, and overall mortality in real-world clinical practice. While GCSF
prophylaxis benefited the high-risk regimen users the most, no significant prophylaxis benefits were found in the
low-risk regimen users.

30

Recent studies (Goyal et al., 2018; Weycker et al., 2016, 2017, 2018) suggested the time of initiation of
GCSF prophylaxis may be critical for FN prevention. Our analysis did not find that the timing of GCSF
prophylaxis significantly impacted patient outcomes. Early GCSF prophylaxis (index day use) was estimated to
be less effective than day+1 use in FN prevention, but the estimated effect was insignificant. We also observed
that late GCSF prophylaxis (day+4 or above) was slightly more effective than day+1 prophylaxis, although again
the effect was not statistically significant.
Our analysis highlighted a regional effect of using GCSF prophylaxis and the risk of FN, as reported in
other studies.(Caggiano et al., 2005; Du et al., 2005; Hershman et al., 2007) However, the lower prophylaxis rates
and higher FN risks in these regions could be associated with the insurance company’s regional distribution of
membership, policy design, physician behavior, and patient ethnicity.(Barnes et al., 2014b) Patients with the PPO
plan had a significantly higher probability of using GCSF prophylaxis than those with the POS plan. Potosky et
al. (Potosky et al., 2011) observed a similar result, alongside patients being less likely to be given GCSF
prophylaxis in the managed care plans than other plans.
Our study advances the literature on GCSF prophylaxis using real-world data in several way. First,
different chemotherapy regimens and patient factors result in variations in chemotherapy course duration. Cox-
regression model address these variations and other issues related to censored clinical outcomes than other studies
using logistic regression methods (Kawatkar et al., 2017; Weycker et al., 2016, 2017, 2018). Second, previous
studies only discuss the high- and intermediate-risk regimens while our analysis conducted a comprehensive
investigation by including myelosuppressive chemotherapy regimens with all risk classes, especially the low-risk
regimens. Finally, we implemented multiple sensitivity analyses to test the different outcomes and model
specifications, which enhances the study’s internal validity while providing more information for healthcare
decision-makers.
Our analysis has limitations. First, The decision to use of GCSF prophylaxis to prevent chemotherapy-
induced febrile neutropenia is based primarily on the patient’s risk factors of neutropenic fever. (Skoetz N et al.,
2015) Therefore, the decision to using GCSF prophylaxis and FN-related health outcomes will be simultaneously
affected by these factors which must be accounted for in the analysis. These correlations make it more challenging
to measure the beneficial impact of GCSF prophylaxis, biasing downward any estimated effect. Econometrics
methods could be implemented to alleviating the confounder bias, such as the interterminal variable (IV) and
difference-in-difference methods. Our database lacked the availability for the zip-code level geographic
information and prevented the implementation of the known IVs from most of the studies. (Earle et al., 2001; Pan,
2010) Given that this outcome variable is a patient-level event and not a clinical outcome measured before and
after treatment, the difference-in-difference approach is not feasible in this analysis either. Changes in clinical

31

guidelines over time could be used as a time-dependent factors to create the needed for difference-in-difference
design. No significant changes were found in clinical guidelines during the data period, except for the addition of
new chemotherapy treatment regimens. As a result, both the IV and the difference-in-difference methods might
not be feasible for our analysis.
Our analysis is limited to a medical paid claims dataset, which provided limited drug dose information.
We used the worst-case scenario for FN risk for the dose-intensive regimens. This approach inflates the number
of high- and intermediate-risk regimens and may underestimate the effects associated with these two risk
regimens. Our dataset did not provide lab values or pathology reports. With the cancer stage and neutrophil count
may be influential factors contributing to chemotherapy regimen and GCSF prophylaxis use, (National
Comprehensive Cancer Network, 2017) these missing factors would introduce the threat of patient selection
biases in our analysis. To address this, we controlled for the FN-risk regimen and GCSF use prior to the FN event
in the regression analysis. Finally, our analysis did not include oral-antibiotic prophylaxis use due to the
inaccessibility to pharmacy claims. This limitation may restrict our interpretations in antibiotics prophylaxis in
regard to FN episodes.

CONCLUSION
In this retrospective claim analysis, we evaluated those factors affecting the use of GCSF prophylaxis for
FN and the effect of GCSF prophylaxis on FN risk with different myelosuppressive chemotherapy regimens. We
also found GCSF prophylaxis is effective in preventing FN events in the high-FN risk chemotherapy regimen,
while no significant benefits were found when using the prophylaxis in the low-risk regimens. The timing of
GCSF prophylaxis did not significantly affect the outcomes either. Further studies in the effectiveness of oral
antibiotics prophylaxis and the association between GCSF prophylaxis and chemotherapy discontinuation are
warranted.

32

FIGURE & TABLE
Figure 1: Patient selection flow chart

All breast cancer patients, 2007 Q1~2016 Q3, no organ / bone
marrow transplant history
N=452,157
N=62,896
N=50,298
N= 26,953
N= 17,470
Final analytic cohort:
N= 17,265
Remove patients never received any cytotoxic chemotherapy
N=389,261 (86.1%)
Remove patient with diagnosis in macular edema, proliferative diabetic
retinopathy, retinal vein occlusion, or macular degeneration (to prevent off-lave
chemotherapy drug use)
N=12,598 (20.0%)
Remove patient without more than 180-days enrollment prior to the first cancer
diagnosis date
N=23,345 (46.4%)
Remove patient with other cancers prior to BC)
N=9483 (35.2%)
Remove patient received chemotherapy prior to the first day of BC diagnosis (to
prevent off-label chemotherapy drug use)
N=205 (1.2%)

33

Table 1: Patient Characteristics

Patient Characteristics
GCSF Prophylaxis No GCSF Prophylaxis
P-value (χ
2
test)
N (% in column) N (% in column)
Overall, the total 6813

10452

Gender [Female] 6813 100.0% 10452 100.0%

Age

<50 2136 31.4% 2945 28.2%
<0.01
50~64 2775 40.7% 4259 40.7%
65+ 1902 27.9% 3248 31.1%
Insurance Type

Exclusive Provider Organization 588 8.6% 798 7.6%
<0.01
Health Maintenance Organization 1329 19.5% 2394 22.9%
Indemnity 44 0.6% 122 1.2%
Other* 1331 19.5% 2025 19.4%
Point of Service 3175 46.6% 4669 44.7%
Preferred Provider Organization 346 5.1% 444 4.2%
Region

Midwest 1426 20.9% 3064 29.3%
<0.01
Other** 112 1.6% 172 1.6%
Northeast 361 5.3% 850 8.1%
South 3520 51.7% 4232 40.5%
West 1394 20.5% 2134 20.4%
Comorbidity (6 months prior to BC diagnosis) ***

Myocardial Infarction 76 1.1% 148 1.4% 0.09
Congestive Heart Failure 160 2.3% 321 3.1% <0.01
Peripheral Vascular Disease 206 3.0% 368 3.5% 0.07
Cerebrovascular Disease 332 4.9% 532 5.1% 0.52
Chronic Pulmonary Disease 1083 15.9% 1682 16.1% 0.29
Connective Tissue Disease-Rheumatic Disease 200 2.9% 340 3.3% 0.24
Mild Liver Disease 243 3.6% 362 3.5% 0.72
Diabetes without complications 975 14.3% 1468 14.0% 0.62
Diabetes with complications 151 2.2% 237 2.3% 0.82
Renal Disease 195 2.9% 323 3.1% 0.39
*: Has an insurance plan switch or without insurance information
**: Has more than 1 service state in the enrollment record or without the state information
***: Comorbidities with less than 1% of the total patient were not reported in the table

34

Table 2: Chemotherapy Course Characteristics

Course Characteristics
GCSF Prophylaxis No GCSF Prophylaxis
P-value (χ
2
test)
N (% in column) N (% in column)
Overall, the total 6813

10452

Treatment Year

2007 259 3.8% 463 4.4%
<0.01
2008 784 11.5% 1410 13.5%
2009 792 11.6% 1361 13.0%
2010 716 10.5% 1144 10.9%
2011 724 10.6% 1154 11.0%
2012 780 11.4% 1123 10.7%
2013 758 11.1% 1021 9.8%
2014 714 10.5% 1059 10.1%
2015 742 10.9% 984 9.4%
2016 544 8.0% 733 7.0%
FN Risk Regimen

High 4787 70.3% 5654 54.1%
<0.01
Intermediate 517 7.6% 1286 12.3%
Low 60 0.9% 604 5.8%
Unknown 1449 21.3% 2908 27.8%
GCSF Use

Use GCSF in-course 6813 100.0% 5193 49.7% <0.01
Use GCSF at day 0 734 10.8% 335 3.2% <0.01
Use GCSF at day 1 4443 65.2% 2903 27.8% <0.01
Use GCSF at day 2 285 4.2% 167 1.6% <0.01
Use GCSF at day 3 363 5.3% 233 2.2% <0.01
Use GCSF at day 4 or later 988 14.5% 1555 14.9% 0.50
Peg-filgrastim 6558 96.3% 4603 44.0% <0.01
Filgrastim 830 12.2% 1072 10.3% <0.01
Sargramostim 54 0.8% 121 1.2% 0.02
Antibiotics Use

Pre-chemotherapy IV Antibiotics Prophylaxis** 285 4.2% 312 3.0% <0.01
Radiotherapy & Surgery Experience

Pre-chemotherapy radiotherapy 606 8.9% 1399 13.4% <0.01
Pre-chemotherapy surgery 6578 96.6% 9260 88.6% <0.01
Pre-chemotherapy FN 507 7.4% 974 9.3% <0.01
Outcomes

FN 1710 25.1% 5903 56.5% <0.01
FN-related hospitalization 168 2.5% 283 2.7% 0.33
Infection Event 1612 23.7% 2202 21.1% <0.01
Infection-related hospitalization 289 4.2% 459 4.4% 0.64
Hospitalization (all-cause) 1870 27.5% 2537 24.3% <0.01
60-day all-cause mortality* 56 0.54% 5 0.07% <0.01
90-day all-cause mortality* 100 0.96% 9 0.13% <0.01
*: Since the index day of chemotherapy
**: Using IV antibiotics 14 days prior to the start of chemotherapy, without concomitant FN/Infection events

35

Figure 2, Logistic Regression (Outcome = use GCSF prophylaxis)

Chemotherapy FN Risk Regimen [vs. High]
Intermediate
Low
Unknown
Surgery / Radiotherapy Experience [vs. No]
Pre-course surgery
Pre-course radiotherapy
Pre-course FN [vs. No]
Pre-course IV antibiotics prophylaxis [vs. No]
Age [vs. <50]
50 ≤ Age < 65
65 ≤ Age
Treatment Year Effect [vs. Previous year]
Insurance Plan [vs. Point-of-Service Plan]
Exclusive Provider Organization Plan
HMO Plan
Indemenity Plan
Other Plan
PPO Plan
Service Region [vs. South]
Midwest
Northeast
Other
West
Comorbidities
Myocardial Infarction
Congestive Heart Failure
Peripheral Vascular Disease
Cerebrovascular Disease
Dementia
Chronic Pulmonary Disease
Connective Tissue Disease-Rheumatic Disease
Peptic Ulcer Disease
Mild Liver Disease
Diabetes without complications
Diabetes with complications
Paraplegia and Hemiplegia
Renal Disease
Moderate or Severe Liver Disease
AIDS/HIV
Parameter
0.51 [0.45, 0.57]
0.13 [0.10, 0.17]
0.64 [0.59, 0.69]
3.83 [3.27, 4.47]
0.66 [0.60, 0.74]
0.79 [0.70, 0.89]
1.44 [1.21, 1.72]
0.92 [0.86, 1.00]
0.88 [0.79, 0.97]
1.07 [1.06, 1.08]
0.98 [0.87, 1.11]
1.16 [1.05, 1.28]
0.72 [0.50, 1.03]
1.11 [1.00, 1.23]
1.38 [1.18, 1.63]
0.53 [0.49, 0.57]
0.51 [0.44, 0.58]
0.92 [0.71, 1.19]
0.96 [0.88, 1.04]
0.87 [0.64, 1.17]
0.87 [0.71, 1.08]
0.88 [0.73, 1.06]
0.98 [0.84, 1.14]
0.79 [0.41, 1.53]
1.00 [0.91, 1.09]
0.90 [0.75, 1.09]
1.14 [0.82, 1.56]
1.03 [0.86, 1.22]
1.04 [0.94, 1.15]
0.90 [0.71, 1.13]
1.15 [0.72, 1.83]
1.04 [0.85, 1.27]
1.53 [0.45, 5.26]
0.61 [0.21, 1.79]
OR [95% CI]
0.00 0.50 1.00 1.50 2.00 2.50 3.00 3.50 4.00
Odss Ratio (95% CI)

36

Figure 3, Cox-Regression, Core model (Outcome = FN event)

GCSF Prophylaxis [vs. No]
Chemotherapy FN Risk Regimen [vs. High]
Intermediate
Low
Unknown
Surgery / Radiotherapy Experience [vs. No]
Pre-course surgery
Pre-course radiotherapy
Pre-course FN [vs. No]
Pre-course IV Antibiotics Prophylaxis [vs. No]
Age [vs. <50]
50 ≤ Age < 65
65 ≤ Age
Treatment Year Effect [vs. Previous year]
Insurance Plan [vs. Point-of-Service Plan]
Exclusive Provider Organization Plan
HMO Plan
Indemenity Plan
Other Plan
PPO Plan
Service Region [vs. South]
Midwest
Northeast
Other
West
Comorbidities
Myocardial Infarction
Congestive Heart Failure
Peripheral Vascular Disease
Cerebrovascular Disease
Dementia
Chronic Pulmonary Disease
Connective Tissue Disease-Rheumatic Disease
Peptic Ulcer Disease
Mild Liver Disease
Diabetes without complications
Diabetes with complications
Paraplegia and Hemiplegia
Renal Disease
Moderate or Severe Liver Disease
AIDS/HIV
Parameter
0.24 [0.23, 0.25]
0.68 [0.63, 0.74]
0.26 [0.22, 0.30]
0.65 [0.62, 0.69]
1.45 [1.31, 1.60]
0.71 [0.66, 0.77]
1.38 [1.28, 1.49]
1.21 [1.07, 1.37]
0.96 [0.91, 1.01]
0.84 [0.78, 0.91]
1.00 [0.99, 1.01]
1.01 [0.93, 1.10]
0.90 [0.84, 0.96]
0.83 [0.65, 1.06]
1.06 [0.98, 1.14]
0.95 [0.84, 1.07]
1.13 [1.07, 1.20]
1.31 [1.20, 1.43]
1.11 [0.92, 1.33]
1.06 [1.00, 1.13]
1.00 [0.81, 1.24]
1.04 [0.89, 1.21]
0.90 [0.78, 1.04]
1.00 [0.89, 1.12]
0.56 [0.31, 1.00]
1.02 [0.96, 1.08]
0.95 [0.83, 1.09]
0.95 [0.75, 1.21]
0.96 [0.85, 1.09]
0.95 [0.88, 1.02]
0.94 [0.79, 1.12]
0.91 [0.64, 1.29]
1.13 [0.98, 1.30]
1.20 [0.50, 2.89]
0.90 [0.45, 1.79]
HR [95% CI]
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80
Hazard Ratio (95% CI)

37

Table 3, Cox-Regression, Sensitivity Analysis

Model / Intervention parameters / Hazard Ratio
Outcome:
FN
Outcome: FN-
related
Hospitalization
Outcome:
Infection
Outcome:
Infection-related
Hospitalization
Outcome:
Hospitalization
(All-cause)
Outcome:
60-day All-
cause
Mortality
Outcome:
90-day All-
cause
Mortality
Core Model
GCSF Prophylaxis [vs. no prophylaxis] 0.24** 0.80* 1.07* 0.88 1.17** 0.22** 0.23**
Risk Regimen [vs. High]

Intermediate 0.68** 0.73 0.96 0.78 1.03 3.94** 4.45**
Low 0.26** 0.32** 1.02 0.65 1.10 4.92** 4.52**
Unknown 0.65** 0.96 1.11** 1.00 1.20** 4.01** 3.37**
Surgery / Radiotherapy Experience [vs. no]

Pre-chemotherapy surgery 1.45** 1.82* 1.26** 1.41 1.54** 1.06 1.27
Pre-chemotherapy radiotherapy 0.71** 0.88 1.16** 1.02 1.20** 2.95** 3.49**
Pre-chemotherapy IV Antibiotics Prophylaxis [vs. no] 1.21** 1.35 1.46** 1.41** 1.21** 1.30 1.35
GCSF Prophylaxis Effect by Risk Class of Chemotherapy
GCSF Prophylaxis*Risk Regimen [vs. no prophylaxis]
High 0.19** 0.65** 1.07 0.79* 1.13** 0.11* 0.27*
Intermediate 0.31** 0.70 1.33* 0.71 1.40** 1.08 0.57
Low 0.90 † 1.19 † 1.06 † †
Unknown 0.42** 1.41 1.03 1.21 1.19** 0.10* 0.07**
Risk Regimen [vs. High]

Intermediate 0.62** 0.70 0.89 0.78 0.95 2.89* 4.13**
Low 0.23** 0.32** 1.01 0.68 1.09 4.64** 4.78**
Unknown 0.55** 0.72* 1.13* 0.85 1.17** 3.87** 3.67**
Surgery / Radiotherapy Experience [vs. no]

Pre-chemotherapy surgery 1.42** 1.76* 1.26** 1.37 1.54** 1.07 1.28
Pre-chemotherapy radiotherapy 0.73** 0.91 1.16** 1.04 1.20** 2.92** 3.43**
Pre-chemotherapy IV Antibiotics Prophylaxis [vs. no] 1.21** 0.80 0.91 0.88 1.44** † 0.62
GCSF Prophylaxis Effect by Prophylaxis Timing
GCSF Prophylaxis * Prophylaxis day [vs. no prophylaxis]
Day 0 0.27** 0.53* 0.97 0.68 1.23** 0.64 0.34
Day 1 0.24** 0.86 1.11* 0.96 1.16** 0.08* 0.17**
Day 2 0.26** 1.06 1.30* 1.00 1.54** † †
Day 3 0.24** 0.72 1.09 0.75 1.13 0.59 0.72
Day 4 or later 0.22** 0.64 0.95 0.68* 1.05 0.39 0.24*
Risk Regimen [vs. High]

Intermediate 0.68** 0.74 0.96 0.78 1.03 3.85** 4.43**
Low 0.26** 0.32** 1.03 0.66 1.11 4.80** 4.50**
Unknown 0.65** 0.97 1.11** 1.01 1.20** 3.92** 3.34**
Surgery / Radiotherapy Experience [vs. no]

Pre-chemotherapy surgery 1.45** 1.81* 1.25** 1.39 1.54** 1.06 1.27
Pre-chemotherapy radiotherapy 0.71** 0.89 1.16** 1.03 1.20** 2.93** 3.48**
Pre-chemotherapy IV Antibiotics Prophylaxis [vs. no] 1.21** 0.81 0.91 0.88 1.44** † 0.64

*: P <0.05; **: P<0.01; †: No/ extremely small sample size

38

REFERENCES
Aapro, M.S., Bohlius, J., Cameron, D.A., Lago, L.D., Donnelly, J.P., Kearney, N., Lyman, G.H., Pettengell, R.,
Tjan-Heijnen, V.C., Walewski, J., et al. (2011). 2010 update of EORTC guidelines for the use of granulocyte-
colony stimulating factor to reduce the incidence of chemotherapy-induced febrile neutropenia in adult patients
with lymphoproliferative disorders and solid tumours. Eur. J. Cancer 47, 8–32.
Barnes, G., Pathak, A., andSchwartzberg, L. (2014a). Pharmacoeconomics of granulocyte colony-stimulating
factor: A critical review. Adv. Ther. 31, 683–695.
Barnes, G., Pathak, A., andSchwartzberg, L. (2014b). G-CSF utilization rate and prescribing patterns in United
States: associations between physician and patient factors and GCSF use. Cancer Med. 3, 1477–1484.
Bow, E.J. (2011). Fluoroquinolones, antimicrobial resistance and neutropenic cancer patients. Curr. Opin. Infect.
Dis. 24, 545–553.
Caggiano, V., Weiss, R.V., Rickert, T.S., andLinde-Zwirble, W.T. (2005). Incidence, cost, and mortality of
neutropenia hospitalization associated with chemotherapy. Cancer 103, 1916–1924.
Campos, F.G. (2017). Colorectal cancer in young adults: A difficult challenge. World J. Gastroenterol. 23, 5041–
5044.
Centers for Medicare & Medicaid Services (2006). Billing of Temporary “C” HCPCS Codes by Non-Outpatient
Prospective Payment System (Non-OPPS) Providers.
Choi, M.R., Solid, C. a, Chia, V.M., Blaes, A.H., Page, J.H., Barron, R., andArneson, T.J. (2014). Granulocyte
colony-stimulating factor (G-CSF) patterns of use in cancer patients receiving myelosuppressive chemotherapy.
Support. Care Cancer 22, 1619–1628.
Du, X.L., Lairson, D.R., Begley, C.E., andFang, S. (2005). Temporal and geographic variation in the use of
hematopoietic growth factors in older women receiving breast cancer chemotherapy: Findings from a large
population-based cohort. J. Clin. Oncol. 23, 8620–8628.
Dulisse, B., Li, X., Gayle, J.A., Barron, R.L., Ernst, F.R., Rothman, K.J., Legg, J.C., andKaye, J.A. (2013). A
retrospective study of the clinical and economic burden during hospitalizations among cancer patients with febrile
neutropenia. J Med Econ 16, 720–735.
Earle, C.C., Tsai, J.S., Gelber, R.D., Weinstein, M.C., Neumann, P.J., andWeeks, J.C. (2001). Effectiveness of
chemotherapy for advanced lung cancer in the elderly: instrumental variable and propensity analysis. J. Clin.
Oncol. 19, 1064–1070.
Freifeld, A.G., Bow, E.J., Sepkowitz, K.A., Boeckh, M.J., Ito, J.I., Mullen, C.A., Raad, I.I., Rolston, K.V.,
Young, J.A.H., andWingard, J.R. (2011). Clinical practice guideline for the use of antimicrobial agents in
neutropenic patients with cancer: 2010 Update by the Infectious Diseases Society of America. Clin. Infect. Dis.
52.
Goyal, R.K., Tzivelekis, S., Rothman, K.J., Candrilli, S.D., andKaye, J.A. (2018). Time trends in utilization of G-
CSF prophylaxis and risk of febrile neutropenia in a Medicare population receiving adjuvant chemotherapy for
early-stage breast cancer. Support. Care Cancer 26, 539–548.
Hershman, D., Neugut, A.I., Jacobson, J.S., Wang, J., Tsai, W.Y., McBride, R., Bennett, C.L., andGrann, V.R.
(2007). Acute myeloid leukemia or myelodysplastic syndrome following use of granulocyte colony-stimulating
factors during breast cancer adjuvant chemotherapy. J. Natl. Cancer Inst. 99, 196–205.

39

Johnson, R.H., Chien, F.L., andBleyer, A. (2013). Incidence of breast cancer with distant involvement among
women in the United States, 1976 to 2009. JAMA 309, 800–805.
Kawatkar, A.A., Farias, A.J., Chao, C., Chen, W., Barron, R., Vogl, F.D., andChandler, D.B. (2017).
Hospitalizations, outcomes, and management costs of febrile neutropenia in patients from a managed care
population. Support. Care Cancer 1–9.
Kuderer, N.M., Dale, D.C., Crawford, J., Cosler, L.E., andLyman, G.H. (2006). Mortality, morbidity, and cost
associated with febrile neutropenia in adult cancer patients. Cancer 106, 2258–2266.
Magellan Rx Management (2018). 2017 Medical Pharmacy Trend Report.
National Cancer Institute Fast Stats: An interactive tool for access to SEER cancer statistics. Surveillance
Research Program.
National Cancer Institute (2009). Common Terminology Criteria for Adverse Events v4.0 (CTCAE).
National Comprehensive Cancer Network (2017). Myeloid Growth Factors. NCCN Clin. Pract. Guidel. Oncol.
National Institutes of Health (2017). Procedure Codes for SEER-Medicare Analyses.
Pan, X. (2010). Estimation Of G-CSF effectiveness in reducing neutropenia hospitalization among non-Hodgkin’s
Lymphoma (NHL)patients treated with anthracycline-based chemotherapy. University of Iowa.
Potosky, A.L., Malin, J.L., Kim, B., Chrischilles, E.A., Makgoeng, S.B., Howlader, N., andWeeks, J.C. (2011).
Use of colony-stimulating factors with chemotherapy: Opportunities for cost savings and improved outcomes. J.
Natl. Cancer Inst. 103, 979–982.
Ramsey, S.D., McCune, J.S., Blough, D.K., McDermott, C.L., Clarke, L., Malin, J.L., andSullivan, S.D. (2010).
Colony-stimulating factor prescribing patterns in patients receiving chemotherapy for cancer. Am. J. Manag. Care
16, 678–686.
Skoetz N, Bohlius J, Engert A, Monsef I, Blank O, andVehreschild JJ (2015). Prophylactic antibiotics or G ( M ) -
CSF for the prevention of infections and improvement of survival in cancer patients receiving myelotoxic
chemotherapy ( Review ). Cochrane Database Syst. Rev.
Smith, T.J., Bohlke, K., Lyman, G.H., Carson, K.R., Crawford, J., Cross, S.J., Goldberg, J.M., Khatcheressian,
J.L., Leighl, N.B., Perkins, C.L., et al. (2015). Recommendations for the use of WBC growth factors: American
society of clinical oncology clinical practice guideline update. J. Clin. Oncol. 33, 3199–3212.
Timmer-Bonte, J.N., deBoo, T.M., Smit, H.J., Biesma, B., Wilschut, F. a, Cheragwandi, S. a, Termeer, A.,
Hensing, C. a, Akkermans, J., Adang, E.M., et al. (2005). Prevention of chemotherapy-induced febrile
neutropenia by prophylactic antibiotics plus or minus granulocyte colony-stimulating factor in small-cell lung
cancer: a Dutch Randomized Phase III Study. J. Clin. Oncol. 23, 7974–7984.
Truong, J., Lee, E.K., Trudeau, M.E., andChan, K.K.W. (2016). Interpreting febrile neutropenia rates from
randomized, controlled trials for consideration of primary prophylaxis in the real world: a systematic review and
meta-analysis. Ann. Oncol. Off. J. Eur. Soc. Med. Oncol. 27, 608–618.
Waters, G.E., Corrigan, P., Gatesman, M., andSmith, T.J. (2013). Comparison of pegfilgrastim prescribing
practice to national guidelines at a university hospital outpatient oncology clinic. J. Oncol. Pract. 9, 203–206.
Weycker, D., Li, X., Figueredo, J., Barron, R., Tzivelekis, S., andHagiwara, M. (2016). Risk of chemotherapy-
induced febrile neutropenia in cancer patients receiving pegfilgrastim prophylaxis: does timing of administration

40

matter? Support. Care Cancer 24, 2309–2316.
Weycker, D., Bensink, M., Lonshteyn, A., Doroff, R., andChandler, D. (2017). Risk of chemotherapy-induced
febrile neutropenia by day of pegfilgrastim prophylaxis in US clinical practice from 2010 to 2015. Curr. Med.
Res. Opin. 33, 2107–2113.
Weycker, D., Hanau, A., Lonshteyn, A., Bowers, C., Bensink, M., Garawin, T., andChandler, D. (2018). Risk of
chemotherapy-induced febrile neutropenia with same-day versus next-day pegfilgrastim prophylaxis among
patients aged ≥65 years: a retrospective evaluation using Medicare claims. Curr. Med. Res. Opin. 0, 1–11.
Wildiers, H., andReiser, M. (2011). Relative dose intensity of chemotherapy and its impact on outcomes in
patients with early breast cancer or aggressive lymphoma. Crit. Rev. Oncol. Hematol. 77, 221–240.
Wingard, J.R., Eldjerou, L., andLeather, H. (2012). Use of antibacterial prophylaxis in patients with
chemotherapy-induced neutropenia. Curr. Opin. Hematol. 19, 21–26.
Wright, J.D., Neugut, A.I., Ananth, C.V., Lewin, S.N., Wilde, E.T., Lu, Y.-S., Herzog, T.J., andHershman, D.L.
(2013). Deviations From Guideline-Based Therapy for Febrile Neutropenia in Cancer Patients and Their Effect on
Outcomes. JAMA Intern. Med. 173, 559.
(2017). Clinical advances in the management of febrile neutropenia. Am. J. Manag. Care - Suppl. October.

41

APPENDIX
Figure A1: Algorithm of chemotherapy cycle & course

Table A1: Sensitivity Analysis (Full coefficients)
Model / Intervention parameters / Hazard Ratio
Outcome:
FN
Outcome: FN-
related
Hospitalization
Outcome:
Infection
Outcome:
Infection-related
Hospitalization
Outcome:
Hospitalization
(All-cause)
Outcome:
60-day All-
cause
Mortality
Outcome:
90-day All-
cause
Mortality
Core Model
GCSF Prophylaxis [vs. no prophylaxis] 0.24** 0.80* 1.07* 0.88 1.17** 0.22** 0.23**
Risk Regimen [vs. High]

Intermediate 0.68** 0.73 0.96 0.78 1.03 3.94** 4.45**
Low 0.26** 0.32** 1.02 0.65 1.10 4.92** 4.52**
Unknown 0.65** 0.96 1.11** 1.00 1.20** 4.01** 3.37**
Surgery / Radiotherapy Experience [vs. no]

Pre-chemotherapy surgery 1.45** 1.82* 1.26** 1.41 1.54** 1.06 1.27
Pre-chemotherapy radiotherapy 0.71** 0.88 1.16** 1.02 1.20** 2.95** 3.49**
Pre-chemotherapy FN [vs. no] 1.38** 0.80 0.91 0.88 1.44** 0.01 0.64
Pre-chemotherapy IV Antibiotics Prophylaxis [vs. no] 1.21** 1.35 1.46** 1.41** 1.21** 1.30 1.35
Year Trend 1.00 0.97 0.99 0.98* 1.18** 0.74** 0.76**
Age [vs. <50]

50~65 0.96 1.30* 1.00 1.28* 1.05 1.51 1.58
65+ 0.84** 1.19 1.09 1.29* 1.50** 7.83** 5.42**
Insurance Plan [vs. Point-of-Service Plan]

Exclusive Provider Organization Plan 1.01 1.00 0.93 0.92 1.22* 0.56 0.51
HMO Plan 0.90** 0.83 1.09 0.84 2.22** 0.75 1.05
Indemnity Plan 0.83 2.09* 0.63* 1.40 0.98 0.01 0.46
Other Plan 1.06 1.39* 1.24** 1.34* 4.17** 0.60 0.67
PPO Plan 0.95 1.35 1.14 1.16 2.99** 0.81 1.13
Service Region [vs. South]

Midwest 1.13** 1.28* 1.14** 1.20* 1.64** 0.52 0.85
Northeast 1.31** 0.92 1.08 1.00 1.69** 0.56 0.68
Other 1.11 0.62 0.87 0.52 0.49** 0.01 0.01
West 1.06 0.75 0.89* 0.73** 0.59** 0.66 0.86
Comorbidity [vs. no]

Myocardial Infarction 1.00 1.02 1.20 1.11 1.25* 0.74 0.89
Congestive Heart Failure 1.04 1.08 1.27* 1.27 1.09 2.54* 2.61**
Peripheral Vascular Disease 0.90 0.90 1.14 1.08 1.06 0.64 0.58
Cerebrovascular Disease 1.00 0.84 1.02 0.80 1.11 1.02 1.11
Dementia 0.56 0.01 0.84 0.94 1.11 0.01 0.01
Chronic Pulmonary Disease 1.02 1.14 1.40** 1.22* 1.14** 1.51 1.47
Connective Tissue Disease-Rheumatic Disease 0.95 1.12 1.14 1.34 1.00 1.26 1.18
Peptic Ulcer Disease 0.95 0.87 1.51** 1.41 0.91 2.95 1.39
Mild Liver Disease 0.96 1.18 1.08 1.05 1.13 0.95 0.88
Diabetes without complications 0.95 0.96 1.09 1.02 1.08 1.09 1.48
Diabetes with complications 0.94 1.13 1.13 1.31 1.04 1.12 0.57
Paraplegia and Hemiplegia 0.91 0.47 1.22 1.05 1.27 0.01 0.01
Renal Disease 1.13 1.27 1.06 1.14 1.20** 2.10 1.64
Moderate or Severe Liver Disease 1.20 0.01 0.82 0.01 0.82 0.01 0.01
AIDS/HIV 0.90 0.01 1.02 0.01 1.44 33.82** 19.37**
GCSF Prophylaxis Effect by Risk Class of Chemotherapy
GCSF Prophylaxis*Risk Regimen [vs. no prophylaxis]

High 0.19** 0.65** 1.07 0.79* 1.13** 0.11* 0.27*
Intermediate 0.31** 0.70 1.33* 0.71 1.40** 1.08 0.57
Low 0.90 0.01 1.19 0.01 1.06 0.01 0.01

42

Unknown 0.42** 1.41 1.03 1.21 1.19** 0.10* 0.07**
Risk Regimen [vs. High]

Intermediate 0.62** 0.70 0.89 0.78 0.95 2.89* 4.13**
Low 0.23** 0.32** 1.01 0.68 1.09 4.64** 4.78**
Unknown 0.55** 0.72* 1.13* 0.85 1.17** 3.87** 3.67**
Surgery / Radiotherapy Experience [vs. no]

Pre-chemotherapy surgery 1.42** 1.76* 1.26** 1.37 1.54** 1.07 1.28
Pre-chemotherapy radiotherapy 0.73** 0.91 1.16** 1.04 1.20** 2.92** 3.43**
Pre-chemotherapy FN [vs. no] 1.21** 0.80 0.91 0.88 1.44** 0.01 0.62
Pre-chemotherapy IV Antibiotics Prophylaxis [vs. no] 1.21** 0.80 0.91 0.88 1.44** 0.01 0.62
Year Effect 1.00 0.97 0.99 0.97* 1.18** 0.74** 0.76**
Age [vs. <50]

50~65 0.96 1.30* 1.00 1.27* 1.05 1.54 1.59
65+ 0.85** 1.21 1.09 1.30* 1.51** 8.01** 5.44**
Insurance Plan [vs. Point-of-Service Plan]

Exclusive Provider Organization Plan 1.01 1.00 0.93 0.92 1.22* 0.57 0.52
HMO Plan 0.90** 0.83 1.09 0.84 2.22** 0.76 1.05
Indemnity Plan 0.82 2.12* 0.63* 1.42 0.98 0.01 0.45
Other Plan 1.06 1.40* 1.24** 1.34* 4.17** 0.60 0.67
PPO Plan 0.94 1.34 1.14 1.15 2.98** 0.79 1.11
Service Region [vs. South]

Midwest 1.13** 1.28* 1.14** 1.20* 1.64** 0.52 0.85
Northeast 1.32** 0.92 1.08 1.00 1.69** 0.57 0.69
Other 1.11 0.63 0.87 0.52 0.49** 0.01 0.01
West 1.06 0.75 0.89** 0.73** 0.59** 0.65 0.86
Comorbidity [vs. no]

Myocardial Infarction 1.01 1.03 1.19 1.11 1.25* 0.73 0.86
Congestive Heart Failure 1.04 1.09 1.27* 1.27 1.09 2.54* 2.60**
Peripheral Vascular Disease 0.90 0.89 1.14 1.08 1.06 0.68 0.61
Cerebrovascular Disease 1.01 0.85 1.02 0.80 1.11 1.01 1.10
Dementia 0.56 0.01 0.83 0.94 1.11 0.01 0.01
Chronic Pulmonary Disease 1.01 1.14 1.40** 1.22* 1.14** 1.48 1.46
Connective Tissue Disease-Rheumatic Disease 0.95 1.12 1.14 1.34 1.00 1.25 1.17
Peptic Ulcer Disease 0.95 0.87 1.51** 1.41 0.91 3.06 1.39
Mild Liver Disease 0.96 1.19 1.08 1.06 1.13 0.96 0.88
Diabetes without complications 0.95 0.96 1.09 1.02 1.08 1.10 1.48
Diabetes with complications 0.95 1.14 1.13 1.32 1.05 1.13 0.58
Paraplegia and Hemiplegia 0.91 0.47 1.22 1.04 1.27 0.01 0.01
Renal Disease 1.13 1.27 1.06 1.14 1.20** 2.06 1.61
Moderate or Severe Liver Disease 1.28 0.01 0.83 0.01 0.83 0.01 0.01
AIDS/HIV 0.89 0.01 1.03 0.01 1.45 34.62** 19.30**
GCSF Prophylaxis Effect by Prophylaxis Timing
GCSF Prophylaxis * Prophylaxis day [vs. no prophylaxis]

Day 0 0.27** 0.53* 0.97 0.68 1.23** 0.64 0.34
Day 1 0.24** 0.86 1.11* 0.96 1.16** 0.08* 0.17**
Day 2 0.26** 1.06 1.30* 1.00 1.54** 0.01 0.01
Day 3 0.24** 0.72 1.09 0.75 1.13 0.59 0.72
Day 4 or later 0.22** 0.64 0.95 0.68* 1.05 0.39 0.24*
Risk Regimen [vs. High]

Intermediate 0.68** 0.74 0.96 0.78 1.03 3.85** 4.43**
Low 0.26** 0.32** 1.03 0.66 1.11 4.80** 4.50**
Unknown 0.65** 0.97 1.11** 1.01 1.20** 3.92** 3.34**
Surgery / Radiotherapy Experience [vs. no]

Pre-chemotherapy surgery 1.45** 1.81* 1.25** 1.39 1.54** 1.06 1.27
Pre-chemotherapy radiotherapy 0.71** 0.89 1.16** 1.03 1.20** 2.93** 3.48**
Pre-chemotherapy FN [vs. no] 1.38** 1.34 1.46** 1.41** 1.21** 1.29 1.33
Pre-chemotherapy IV Antibiotics Prophylaxis [vs. no] 1.21** 0.81 0.91 0.88 1.44** 0.01 0.64
Year Trend 1.00 0.97 0.99 0.98 1.18** 0.74** 0.76**
Age [vs. <50]

50~65 0.95 1.30* 1.00 1.28* 1.05 1.51 1.58
65+ 0.84** 1.20 1.09 1.29* 1.50** 7.81** 5.41**
Insurance Plan [vs. Point-of-Service Plan]

Exclusive Provider Organization Plan 1.01 1.00 0.93 0.92 1.22* 0.55 0.51
HMO Plan 0.90** 0.83 1.08 0.84 2.22** 0.74 1.04
Indemnity Plan 0.83 2.09* 0.63* 1.40 0.98 0.01 0.46
Other Plan 1.06 1.38* 1.23** 1.33* 4.17** 0.60 0.67
PPO Plan 0.95 1.34 1.14 1.15 2.97** 0.80 1.13
Service Region [vs. South]

Midwest 1.13** 1.29* 1.14** 1.20* 1.64** 0.53 0.86
Northeast 1.31** 0.92 1.08 0.99 1.69** 0.57 0.68
Other 1.11 0.62 0.88 0.52 0.49** 0.01 0.01
West 1.06 0.76 0.89* 0.73** 0.59** 0.65 0.87
Comorbidity [vs. no]

Myocardial Infarction 1.00 1.02 1.20 1.10 1.25* 0.73 0.90
Congestive Heart Failure 1.04 1.09 1.27* 1.28 1.09 2.52* 2.57**
Peripheral Vascular Disease 0.90 0.89 1.14 1.08 1.06 0.65 0.59

43

Cerebrovascular Disease 1.00 0.84 1.02 0.80 1.11 1.03 1.11
Dementia 0.56 0.01 0.83 0.95 1.11 0.01 0.01
Chronic Pulmonary Disease 1.02 1.14 1.41** 1.22* 1.14** 1.49 1.46
Connective Tissue Disease-Rheumatic Disease 0.95 1.12 1.14 1.34 1.00 1.27 1.18
Peptic Ulcer Disease 0.95 0.85 1.50** 1.39 0.91 2.99 1.42
Mild Liver Disease 0.96 1.18 1.08 1.05 1.13 0.95 0.88
Diabetes without complications 0.95 0.95 1.09 1.02 1.07 1.11 1.50
Diabetes with complications 0.94 1.15 1.14 1.33 1.05 1.12 0.56
Paraplegia and Hemiplegia 0.91 0.46 1.21 1.04 1.28 0.01 0.01
Renal Disease 1.13 1.27 1.06 1.14 1.20** 2.10 1.66
Moderate or Severe Liver Disease 1.20 0.01 0.81 0.01 0.81 0.01 0.01
AIDS/HIV 0.90 0.01 1.05 0.01 1.43 33.09** 19.22**

*: P<0.05; **:P<0.01; †: No Sample Existed

44

Chapter 4 - An Analysis of Granulocyte-Colony-Stimulating
Factor Utilization for the Prevention of Chemotherapy-Induced
Febrile Neutropenia among Lung Cancer Patients in the United
States

ABSTRACT
Background: Febrile neutropenia (FN) is a common but severe immune suppressive condition associated with
myelosuppressive chemotherapy. Granulocyte-colony stimulating factors (GCSF) are used to prevent or
ameliorate chemotherapy-induced FN. Clinical trial studies have shown GCSF prophylaxis significantly reduced
the risk of chemotherapy-induced FN in oncology patients receiving chemotherapies. The objective of this paper
is to evaluate the real-world effectiveness of GCSF for primary FN prophylaxis in patients with lung cancer.

Methods: A commercial health insurance databased covering January 2007 to December 2017 was used for the
analysis. The analysis included only newly diagnosed chemotherapy-naive lung cancer patients. Logistic
regression models were used to estimate the association between patient/treatment characteristics and the proper
and improper use of GCSF prophylaxis. A multivariate Cox-regression model was estimated to test the
association between the emergence of an FN event and GCSF prophylaxis. Sensitivity analysis was carried out to
test the robustness of core results across FN-related outcomes and alternative model specifications.

Results: GCSF prophylaxis is significantly associated with chemotherapy FN-risk regimen, radiotherapy,
immunotherapy, geographic area, and provider characteristics. GCSF prophylaxis significantly reduced FN risk
(HR=0.74, P <0.01) compared to the non-prophylaxis cohort. Concomitant GCSF prophylaxis and radiotherapy
increased the risk of infection event (HR=1.23, P<0.05). GCSF prophylaxis had no significant FN prevention
effects on infection-related hospitalization, all-cause hospitalization, and ER use.

Conclusion: GCSF prophylaxis may effectively prevents FN event among lung cancer patients who received
chemotherapy. There were no significant benefits of GCSF prophylaxis on the risk of other FN-related health
outcomes.

45

INTRODUCTION
Febrile neutropenia (FN) is defined as absolute neutrophil count (ANC) <1000/mm
3
with concomitant
systemic temperature >38.3°C, or a sustained body temperature of ≥38°C for more than 1 hour. (National Cancer
Institute, 2009) FN is a frequent and severe side-effect associated with myelosuppressive chemotherapy. Due to
the life-threatening nature of FN, patients often require hospital stays, leading to increased mortality and higher
healthcare costs. (Kuderer et al., 2006) Oncologists can reduce the risk of myelosuppression by either modifying
the chemotherapy interval or reducing the chemotherapy dose, both of which may reduce treatment response and
decrease survival in some malignancies. (Wildiers andReiser, 2011) Prophylactic use of granulocyte colony-
stimulating factors (GCSF) has been shown to prevent or ameliorate chemotherapy-induced FN, (Bow, 2011;
Timmer-Bonte et al., 2005; Wingard et al., 2012) yet the evidence concerning the impact of GCSF on FN-
associated hospitalization has been inconsistent. (Choi et al., 2014; Potosky et al., 2011; Ramsey et al., 2010;
Waters et al., 2013; Wright et al., 2013)
The substantial cost of GCSF is a significant contributor to the total cost of cancer-related healthcare
service in the US. (Magellan Rx Management, 2018; 2017) GCSF accounted for 10% of the total oncology drug
expenditures among both commercial and Medicare plans in 2017. (Magellan Rx Management, 2018) Evidence-
based guidelines have been developed for GCSF use as primary prophylaxis to prevent chemotherapy-related FN.
(Aapro et al., 2011; Freifeld et al., 2011; National Comprehensive Cancer Network, 2017; Smith et al., 2015a)
These guidelines commonly stratified FN risk according to patient risk factors and the myelosuppressive nature of
the chemotherapy regimen. These regimens are classified as high (≥20% of FN incidence rate), intermediate (10–
20%), or low (<10%) potential for development of FN. Primary GCSF prophylaxis is recommended for patients
with high-FN risk chemotherapy regimens and is not recommended for low-FN risk regimens users. The impact
of GCSF prophylaxis on economic outcomes remained inconclusive. (Barnes et al., 2014a; Dulisse et al., 2013;
Kawatkar et al., 2017)
The evidence is incomplete concerning GCSF’s effectiveness in the real-world clinical practice. Clinical
guidelines are based on well-controlled clinical trials which may differ significantly from real-world
practice.(Truong et al., 2016) Retrospective studies of GCSF effectiveness have focused on patients aged >65
years, despite the increasing incidence of cancers in patients aged <50 years. (Campos, 2017; Johnson et al.,
2013) Some studies also have found that worsening health outcomes of using GCSF prophylaxis among patients
received concomitant radiotherapy. (Cui et al., 2015; Murakami et al., 2018; Srivastava et al., 2018; Woolf et al.,
2016)
This study’s objective is to evaluate the real-world effectiveness of GCSF in preventing FN in
chemotherapy-naï ve lung cancer patients using health insurance claims collected over a 10-year period. Treatment

46

decisions in the first week of chemotherapy were then used to define GCSF prophylaxis and treatment
characterstics.

METHODS
Patient Selection and Chemotherapy Index Day:
Our analytic data were sourced from De-identified Clinformatics® Data Mart (OptumInsight, Eden
Prairie, MN). The dataset included 51,988,191 enrollees covering January 2007 to December 2017. We located all
lung cancer patients through ICD-9-CM code 162 or ICD-10-CM codes C34 (See Appendix). Patients with <180-
days of continuous insurance enrollment prior to their first lung cancer diagnosis were excluded. We also
excluded patients with the history of organ transplantation, bone marrow transfer or prior FN from the analysis.
Chemotherapy claims, including immonotherapy and target therapy agents, were classified using HCPCS Level 2
codes J9000–J9999 and HCPCS “C” codes. (Centers for Medicare & Medicaid Services, 2006; National Institutes
of Health, 2017) The first day of chemotherapy after the lung cancer diagnosis was then set as the index day.
Patients used GCSF before the index day were excluded from the analysis as well.
Provider specialty, chemotherapy service location, and patient characteristics were coded based on the
index day claim. The date of use of GCSF, antibiotics, radiation therapy, and surgical interventions were recorded
to document the mix of these services used during the first week of chemotherapy. We defined the patient’s
treatment regimen as the combination of all chemotherapy drugs used during the treatment period defined as
continuous treatment up to the emergence of a 60-day gap in chemotherapy drug use.

FN-risk Regimen:
The risk classification for the patient’s initial chemotherapy regimen was defined using data from NCCN
Clinical Practice Guidelines in Oncology for Myeloid Growth Factors. (National Comprehensive Cancer
Network, 2017) Since only high- and intermediate-riskregimens are defined in the NCCN guidelines, we added
low-risk regimens as defined in the ASCO/Cleveland Clinic consensus guideline (Martin Goodman et al., 2016)
and a manually search for other low-risk chemotherapy regimens for lung cancer. For chemotherapy regimens
having inconsistent FN risk levels across different guidelines, or risk dependent on dose, we used the highest FN
risk to define the patient’s FN risk.

Treatments Used as FN Prophylaxis:
We defined GCSF prophylaxis as GCSF use prior to an emerging diagnosis of FN or infection. Next, we
defined guideline-consistent GCSF prophylaxis as GCSF use within 72 hours (day+3) after the initiation of
chemotherapy. (National Comprehensive Cancer Network, 2017) Patients initiating GCSF prophylaxis after

47

day+4, and those never using GCSF prophylaxis, were categorized as the non-guideline/non-prophylaxis
comparison group. Antibiotics prophylaxis was defined using the same prophylaxis definition as GCSF. We also
defined pre-chemotherapy antibiotic use as prophylaxis if the patient had used IV/oral antibiotics within 14 days
before the chemotherapy initiation and without concomitant FN/ infection events. Post-index interventions,
including radiotherapy, antibiotics prophylaxis, immunotherapy, and target therapy, are recorded until day +7 to
capture the pattern of the initial treatment period. HCPCS codes J2820, J1440, J1441, J1442, J2505, and Q5101
were employed to identify the use of GCSF. Oral and IV antibiotics were identified from pharmacy claims by
using the American Hospital Formulary Service therapeutic drug class codes. (American Hospital Formulary
Service, 2017)

Outcomes:
An FN event was identified using concomitant diagnosis codes for neutropenia (288.x in ICD-9; D70 in
ICD-10) and fever (780.6x in ICD-9; R502x, R508x, R509x in ICD-10). We broadened our outcome definition to
include microbiology infections using diagnosis and DRG codes. The date of the first post-index hospital
admission was recorded separately for infection-related hospitalizations, all-cause hospitalizations and ER use
based on the DRG codes (Appendix) and billing location records.

Analytic Methods:
A logistic regression model was estimated to document the factors associated of the use of GCSF
prophylaxis, including patient/ treatment characteristics, SES factors and provider information. A Cox-regression
model was estimated to document the association between GCSF prophylaxis and FN events and other health
outcomes controlling for patient characteristics, time trends and other factors. All outcomes were censored on the
last paid claim in the patient’s data.
We implemented the propensity score methods which are frequently applied in retrospective studies to
reduce the impact of unobserved differences between treated and untreated patients (Rosenbaum andRubin, 1985;
ROSENBAUM and RUBIN, 1983) by matching treated and untreated patients. A 1:2 greedy matching method for
the robustness check was used based on the recommendation by Austin (Austin, 2010). Post-matching propensity
score distribution and the standardized differences among regressors are estimated to document the precision of
these matching efforts. We also include additional Inverse Probability Treatment Weighting (IPTW) method for
the robustness check based on the suggestion by Burden et al.. (Burden et al., 2017)
Several sensitivity analyses were conducted. We re-estimated the core models using interaction terms of
between GCSF prophylaxis and concomitant radiation therapy and the FN risk regimens. These analyses
quantified possible differential effects of GCSF prophylaxis across chemotherapy risk levels and concomitant

48

cancer-related interventions. We also documented the effect of GCSF prophylaxis by the days of delay since the
initiation of chemotherapy. This specification estimates the effects of using GCSF prophylaxis at 24-hr later
(day+1), 48-hr later (day+2), and 72-hr later (day+3) post-chemotherapy initiation. Sensitivity models were
estimated for all outcome measures. All the statistical models were estimated using SAS
©
9.4.

RESULTS
Patient Selection
Table 1 provides data on the patient identification process. A total of 246,067 lung patient cancer patients
were identified, representing an approximate 0.43% prevalence rate in the database. The prevalence reported here
is higher than the rate of 0.17% reported in the SEER Program. (National Cancer Institute) A study sample of
21,177 patients met the inclusion and exclusion criteria for the analysis. Most excluded patients (57%) did not use
chemotherapy, suggesting they were in early stages (I/II).

Patient/ Treatment Characteristics of the GCSF Prophylaxis and Non-prophylaxis Populations
Table 2 shows 8.0% of lung cancer chemotherapy patients received GCSF for FN prophylaxis within
day+3 of the chemotherapy initiation. This is consistent with guideline recommendations. Most patients were
aged older than 65 years in both prophylaxis and no prophylaxis groups (70.4% vs. 69.7%, P> 0.05). Chronic
pulmonary disease was the most common comorbidity for both groups and had a significant distribution
difference (43.2% vs. 39.0%; P < 0.05).
The low-FN-risk regimen was the most commonly employed chemotherapy regimen in both the GCSF
prophylaxis and non-prophylaxis treatment groups (37.5% vs. 51.8%, respectively). Most GCSF prophylaxis was
used on day+1 (53.3%). While 31.9% of the non-prophylaxis patients eventually used GCSF, only 4.7% of the
GCSF use was initiated before day+3 implying either the late emergence of FN events or the delayed prophylaxis.

Factors Correlated with the Use of GCSF Prophylaxis/ Improper Prophylaxis
Patients with high, intermediate and undocumented FN-risk chemotherapy regimens were more likely to
use GCSF prophylaxis than patients with the low-risk regimens (Graph 1, High: OR=1.47; Intermediate:
OR=2.33; Undocumented: OR=1.49; all P< 0.05). This is in accordance with the guideline of not using GCSF
prophylaxis in conjunction with low-risk regimens. The use of radiotherapy was associated with a less GCSF
prophylaxis use (Concomitant radiotherapy: OR=0.38; Pre-chemo radiotherapy: OR=0.85; all P<0.05), which is
consistent to the clinical findings of poorer clinical outcomes for patients using radiotherapy and GCSF
concomitantly. (Cui et al., 2015; Murakami et al., 2018; Srivastava et al., 2018) The yearly trend showed a 7%

49

utilization reduction per year (OR=0.93, P<0.05), suggesting the improved awareness of the over use of GCSF.
The higher likelihoods of use GCSF prophylaxis among pre-chemotherapy IV and concomitant oral antibiotics
prophylaxis patients (OR=1.76, P<0.05; OR=1.39, P <0.05; respectively) suggested the higher baseline FN risks
among patients treated with this FN prophylaxis treatments.
Only 10.2% of the study patients were categorized as improper prophylaxis (Table 2). Geographic area
was a strong predictor for the improper prophylaxis (Graph 2). Patients living in the Mountain area were more
likely to be prophylaxis improperly than the ones in the South Atlantic area (OR=1.42, P<0.05). No significant
associations were found between the insurance plan type and improper prophylaxis.

The Core Model of FN Risk:
The results of the core analysis of the factors impacting the patient’s risk of developing FN are reported in
Graph 3. A significant 25% FN risk reduction was found for GCSF prophylaxis (HR=0.75, P<0.05). The use of
immunotherapy was associated with a significant risk reduction in FN event (HR=0.71; P<0.05). Using pre-
chemotherapy IV antibiotic prophylaxis had a significant risk increasement in FN (HR=1.19, P<0.05). This result
likely reflects physicians’ concerns for patient’s baseline FN risk rather than IV antibiotics causing FN.
No significant associations were found between the risks of FN and insurance plans. The Mountain area,
though had a higher likelihood in improper prophylaxis, did not show a worsen FN risk compared to the South
Atlantic area (HR=0.95, P>0.05). The use of hospital-outpatient center was associated with a lower FN risk than
regular clinics (HR=0.91, P<0.05). Providers in non-hematology/oncology specialties were associated with a
lower risk in FN event comparing to hematology/oncology specialty (HR=0.83, P<0.05), suggesting patients with
higher risks of FN or worse well-being might be treated by oncologists rather other specialties. Finally, Patients
with chronic pulmonary diseases, rheumatic diseases, and renal diseases were significantly associated with higher
risks in FN events (HR=1.09; HR=1.14; HR=1.19; all P<0.05, respectively).

Selection Bias Robustness Check and Sensitivity Analysis:
Table 3 summarizes the selection bias robustness check and the sensitivity analysis results. The full model
results are reported in the Appendix. The standardized differences showed the patient characteristics between pre-
and post-1:2 matching results were balanced (Appendix). The effects of GCSF prophylaxis on FN were consistent
across the core and PS adjustment methods (Unmatched: HR=0.74; 1:2 Matching: HR=0.73; IPTW: HR=0.80; all
P<0.05). The effects of GCSF prophylaxis on other outcomes were similar in the magnitude and pattern of
statistical significances.

50

The core model showed using GCSF prophylaxis reduces the risk in FN event significantly; while no
significant benefits were found on the hospitalization-associated outcomes. The interaction terms on FN-risk
regimens showed using GCSF prophylaxis were effective among all the risk regimens. The benefits of GCSF
prophylaxis on FN risk were more significant for patients with high- and the intermediate-risk regimens (High:
HR=0.71; Intermediate: HR=0.57, all P <0.05, respectively) relative to patients treated by low risk regimens.
Using GCSF prophylaxis on high/ intermediate risk regimens also decreased the risk of infection event
significantly (HR=0.75; HR=0.87, all P<0.05, respectively), but did not benefit other health outcomes. Using
GCSF prophylaxis on day+1 or day+3 provided significant risk reductions on FN (HR=0.76; HR=0.69; all
P<0.05, respectively). No significant effect difference was found between day+1 and day+3 on FN prevention.
The concomitant use of GCSF prophylaxis and radiotherapy did not increase the risk of FN but did
significantly increase the risk of an infection event and an infection-related hospitalization (HR=1.23; HR=1.66,
all P<0.05, respectively). This finding confirmed the worsen health outcome of using GCSF and concomitant
radiotherapy.

DISCUSSION
This analysis established that FN risk remains significant in real world clinical practice and that GCSF
can lower the risk of FN across all risk categories of chemotherapy. The unobserved patient factors that contribute
to the FN risk is not accessible through claims data. However, we showed that prior antibiotic use and patient’s
surgical history are independent predictors of FN risk. In the propensity matched cohort, use of GCSF also
reduced the risk of overall-infection events.
As in all comparative effectiveness analyses using retrospective data, we took several steps to reduce the
possibility of endogeneity bias from the analysis. Previous research has documented several potential
confounding variables for development of chemotherapy-induced FN in patients with lung cancer. These risk
factors include age, chemotherapy regimens and line of therapy, concurrent radiation, prior surgery, organ
dysfunction, and general well-being (performance status). (National Comprehensive Cancer Network, 2017;
Skoetz N et al., 2015; Smith et al., 2015b) We deployed conservative case selection to focus our analysis on
patients who are receiving their initial chemotherapy ; however, it is possible that these patients before this
study’s 6-month wash out period, possibly when covered by other insurance and thus the chemotherapy episode in
this analysis is not their first course of treatment. Similarly, we have the gap of the time of chemotherapy which
may be different from the age at diagnosis. Excluding patients with pre-existing neutropenia and prior history of
FN strengthens the selection process for those whom neutropenia can be attributed to chemotherapy rather than
other comorbidities. Patients with liver and renal dysfunction are excluded from most clinical trials, and this

51

dataset provides real-world evidence for higher likelihood of neutropenia in these patients and also benefit from
GCSF.
Our analysis found the use of GCSF prophylaxis within the 72 hours of chemotherapy initiation was
effective in reducing the risk of FN but not the risk of other outcomes. This results is not consistent with other
retrospective studies. This could be due to the heterogeneity of patient selection in other studies which included
multiple types of cancer patients, (Mitchell et al., 2016; Tan et al., 2011) or using logistic regressions. (Kawatkar
et al., 2017; Weycker et al., 2016, 2017, 2018) We implemented Cox-regression models to better address the
inconsistent duration issue and the censored clinical outcomes. While other studies only discuss the high- and
intermediate-risk regimens, our analysis conducted a comprehensive investigation by including chemotherapy
regimens with all risk classes.
Our analytic results found regional and prescriber effects on the decision to use GCSF prophylaxis and
on the risk of FN as reported in other studies. (Barnes et al., 2014b; Du et al., 2005; Wright et al., 2013) Studies
have indicated the GCSF prophylaxis utilization varied widely across geographic regions. (Du et al., 2005; Wright
et al., 2013) where the regional healthcare resource disparity and the regional-specific provider behavior could be
the key roles. Prescriber specialty also contribute to the variance of prophylaxis decision, which could be resulted
from the different acknowledge of GCSF and FN. (Bennett et al., 1999) While the PPO plan was associated with
more GCSF use than the managed care plan as Potosky et. al found, (Potosky et al., 2011) we observed
significantly higher risks in the infection- and the all-cause hospitalization events in the PPO over the HMO plan.
Since patients with advanced or recurrence cancers may tend to use less-restrictive health plans, it is possible to
see worsen health outcomes in the managed care plans.
Results from the interaction models showed the concomitant use of GCSF and radiotherapy lead to
undesirable outcomes. This is consistent with clinical observations among lung cancer patients. (Cui et al., 2015;
Murakami et al., 2018; Srivastava et al., 2018) Considering the high presence of the combined use of radiation
therapy and chemotherapy among lung cancer patients, our finding highlighted the need for careful coordination
between radio-oncologists and hematologists treatment plan design. Though recent studies (Goyal et al., 2018;
Weycker et al., 2016, 2017, 2018) suggested the time of initiation of GCSF prophylaxis may be critical for FN
prevention, our analysis did not note the timing effect significantly. The day+2 use of GCSF did not show
effective FN risk reduction, which likely reflects the relatively low number of day+2 users compared to the
number of patient initiating GCSF prophylaxis on day+1 or day+3.
Our analysis has limitations. First, it is limited to a insurance claims dataset, which provided limited drug
dose information. We were unable to quantify FN risk associated with dose-intensive regimens. For regimens
where the FN risk is classed by the drug dose, we assigned the highest FN risk classification to the patient

52

regimen based on the worst-case scenario for the medications used. This approach inflates the number of high-
and intermediate-risk regimens and may underestimate the effects associated with these two risk regimens.
Second, our dataset did not provide detailed provider characteristics and detailed insurance plan information. We
are unable to throughtly investigate the interaction between provider, insurance plan design and health outcomes.
Third, the geographically estimated SES factors assigned to each patient in the database were never found to be
correlated with GCSF prophylaxis use or patient outcomes. The high percentages of the unknown categories
among the SES variables also limited their value as independent variables in the regression models. Finally, the
lack of zip-code level geographic variables prevents us from using known econometrics methods to adjust the
potential selection bias. Yet the robustness check with propensity score matching and IPTW has shown the bias is
minor with the acceptable credibility of the estimation.

CONCLUSION
In this retrospective claim analysis of patients with lung cancer, we evaluated those factors affecting the
use of GCSF prophylaxis on FN and the effect of GCSF prophylaxis on FN with different chemotherapy
regimens. We found chemotherapy regimen, immunotherapy, radiotherapy, insurance plan, service region,
chemotherapy site, and provider specialty affects GCSF prophylaxis use. GCSF prophylaxis is effective in
preventing FN events, while no significant benefits were found when using the prophylaxis on other health
outcomes. Using GCSF prophylaxis is effective among all risk regimens. The day+1 and day+3 GCSF
prophylaxis reduced the FN risk significantly but did not affect other health outcomes significantly. Further
studies in the association between provider characteristics, insurance products, and GCSF prophylaxis
effectiveness will be carried out in the future.

53

FIGURE AND TABLE
Table 1, Patient Selection Criteria:
Selection Criteria Inclusion Exclusion
Patients with primary lung cancer diagnosis 246067

Never use chemotherapy

139484
Has other primary cancer diagnosis history before LC

47298
Has any secondary cancer diagnosis code before LC

8198
Patients without 180-day enrollment before 1st lung cancer diagnosis

26034
Used any chemotherapy agents prior to LC diagnosis

384
Had bone marrow transfer history prior to LC diagnosis

0
Has organ transplantation history prior to LC diagnosis

3
Used GCSF before the initiation of chemotherapy

402
Had FN before chemotherapy

2908
Without complete patient characteristics (gender, birth year, etc.)

4
Without using target therapy/ chemotherapy agents for non-lung cancer
specific
175
Total 21177

54

Table 2, Patient & Treatment Characteristics:
Patient Characteristics
GCSF Prophylaxis
(+) (N=1701)
GCSF Prophylaxis
(-) (n=19476)
P-value
(Chi-sq/
T)
N (%) N (%)
Overall, total 1701 8.0% 19476 92.0%

Gender

Female 850 50.0% 9457 48.6%
0.26
Male 851 50.0% 10019 51.4%
Age

< 50 44 2.6% 629 3.2%
0.33 50~64 459 27.0% 5286 27.1%
65+ 1198 70.4% 13561 69.6%
Insurance Type

Exclusive Provider Organization 66 3.9% 594 3.0%
< 0.05
Health Maintenance Organization 530 31.2% 6297 32.3%
Indemnity 8 0.5% 198 1.0%
Other 598 35.2% 7186 36.9%
Point of Service 271 15.9% 3396 17.4%
Preferred Provider Organization 228 13.4% 1805 9.3%
Region

EAST NORTH CENTRAL 194 11.4% 3080 15.8%
< 0.05
EAST SOUTH CENTRAL 100 5.9% 1008 5.2%
MIDDLE ATLANTIC 106 6.2% 1519 7.8%
MOUNTAIN 184 10.8% 1208 6.2%
NEW ENGLAND 70 4.1% 945 4.9%
PACIFIC 82 4.8% 2126 10.9%
SOUTH ATLANTIC 606 35.6% 4870 25.0%
Unknown 4 0.2% 81 0.4%
WEST NORTH CENTRAL 183 10.8% 2375 12.2%
WEST SOUTH CENTRAL 172 10.1% 2264 11.6%
Comorbidity

Myocardial Infarction 93 5.5% 1003 5.1% 0.57
Congestive Heart Failure 136 8.0% 1570 8.1% 0.92
Peripheral Vascular Disease 207 12.2% 2549 13.1% 0.28
Cerebrovascular Disease 207 12.2% 2150 11.0% 0.15
Dementia 11 0.6% 106 0.5% 0.58
Chronic Pulmonary Disease 739 43.4% 7622 39.1% < 0.05
Connective Tissue Disease-Rheumatic Disease 62 3.6% 629 3.2% 0.36
Peptic Ulcer Disease 34 2.0% 313 1.6% 0.22
Mild Liver Disease 138 8.1% 1169 6.0% < 0.05
Diabetes without complications 346 20.3% 3720 19.1% 0.21
Diabetes with complications 101 5.9% 1152 5.9% 0.97
Paraplegia and Hemiplegia 10 0.6% 121 0.6% 0.87
Renal Disease 105 6.2% 1277 6.6% 0.54
Cancer 0 0.0% 0 0.0% --
Moderate or Severe Liver Disease 5 0.3% 41 0.2% 0.48
Metastatic Carcinoma 0 0.0% 0 0.0% --
AIDS/HIV 3 0.2% 19 0.1% 0.33
Social-economic factors

Education level

High School Diploma / Less than 12th Grade 648 38.1% 7110 36.5%
0.12
Less than Bachelor Degree 821 48.3% 9345 48.0%
Bachelor Degree Plus 142 8.3% 1735 8.9%
Unknown 90 5.3% 1286 6.6%
Home Ownership

Unknown 471 27.7% 5668 29.1%
0.12 Probable Homeowner 1174 69.0% 13027 66.9%
Probable Renter 56 3.3% 781 4.0%
Household Income Range

Unknown 340 20.0% 4289 22.0%
0.56
<$40K 524 30.8% 5775 29.7%

55

$40K-$49K 141 8.3% 1574 8.1%
$50K-$59K 136 8.0% 1450 7.4%
$60K-$74K 162 9.5% 1767 9.1%
$75K-$99K 174 10.2% 2051 10.5%
$100K+ 224 13.2% 2570 13.2%
Networth Range

$500K+ 266 15.6% 3347 17.2%
< 0.05
$150K-$249K 270 15.9% 3462 17.8%
$25K-$149K 360 21.2% 3766 19.3%
$250K-$499K 218 12.8% 2312 11.9%
<$25K 331 19.5% 3485 17.9%
Unknown 256 15.0% 3104 15.9%
Race

Asian 27 1.6% 436 2.2%
< 0.05
Black 198 11.6% 2318 11.9%
Hispanic 72 4.2% 1113 5.7%
Unknown 216 12.7% 2708 13.9%
White 1188 69.8% 12901 66.2%
Treatment Year

2007 39 2.3% 556 2.9%
< 0.05
2008 156 9.2% 1660 8.5%
2009 176 10.3% 1751 9.0%
2010 189 11.1% 1759 9.0%
2011 196 11.5% 1773 9.1%
2012 201 11.8% 1826 9.4%
2013 196 11.5% 1649 8.5%
2014 176 10.3% 1647 8.5%
2015 161 9.5% 1794 9.2%
2016 113 6.6% 2350 12.1%
2017 98 5.8% 2711 13.9%
Treatment Season

Spring (March ~ May) 436 25.6% 4648 23.9%
0.24
Summer (June~ August) 410 24.1% 4597 23.6%
Fall (September~ November) 368 21.6% 4497 23.1%
Winter (December~ February) 469 27.6% 5734 29.4%
FN Risk Regimen

High 152 8.9% 1523 7.8%
< 0.05
Intermediate 415 24.4% 2394 12.3%
Low 638 37.5% 10093 51.8%
Unknown 496 29.2% 5466 28.1%
Regimen Combination

Single 41 2.4% 1842 9.5%
< 0.05
Multiple 1660 97.6% 17634 90.5%
Immunotherapy 53 3.1% 1718 8.8% < 0.05
PD-L1 53 3.1% 1677 8.6%
0.26
CTLA4 0 0.0% 41 0.2%
Target Therapy 30 1.8% 314 1.6% 0.12
EGFR 9 0.5% 162 0.8%
0.97 PDGFR 0 0.0% 1 0.0%
VEGF 4 0.2% 67 0.3%
Chemotherapy Provider Specialty

0.0%

Hematology/ Oncology specialty 1053 61.9% 11780 60.5%
0.25
Non-Hematology/ Oncology specialty 648 38.1% 7696 39.5%
Chemotherapy Location

Hospital outpatient facility 548 32.2% 5496 28.2%
< 0.05 Physician office / clinics 1097 64.5% 12475 64.1%
Other location 56 3.3% 1505 7.7%
GCSF Prophylaxis Use

Peg-filgrastim 1619 95.2% 4233 21.7% < 0.05
Filgrastim 178 10.5% 1797 9.2% 0.09
Sargramostim 21 1.2% 186 1.0% 0.26
Use GCSF at day +1 907 53.3% 499 2.6% < 0.05

56

Use GCSF at day +2 117 6.9% 67 0.3% < 0.05
Use GCSF at day +3 677 39.8% 364 1.9% < 0.05
Antibiotics Use

IV Antibiotic Prophylaxis within 7 days of chemotherapy initiation 13 0.8% 135 0.7% 0.73
IV Antibiotics prophylaxis, 1~7 days before chemotherapy initiation 66 3.9% 441 2.3% < 0.05
Oral Antibiotic Prophylaxis within 7 days of chemotherapy initiation 55 3.2% 469 2.4% < 0.05
Non-Quinolones 13 0.8% 172 0.9%
0.06
Quinolones 42 2.5% 297 1.5%
Oral Antibiotics prophylaxis, 1~7 days before chemotherapy initiation 24 1.4% 308 1.6% 0.58
Non-Quinolones 16 0.9% 190 1.0%
0.49
Quinolones 8 0.5% 118 0.6%
Radiotherapy & Surgery Experience

Use radiotherapy within 7 days of chemotherapy initiation 129 7.6% 3398 17.4% < 0.05
Pre-course radiotherapy 253 14.9% 3989 20.5% < 0.05
Pre-course surgery 199 11.7% 2221 11.4% 0.71
Pre-chemo ER/ Hospitalization history

Pre-chemo ER use, within 14 days before chemo 132 7.8% 1238 6.4% < 0.05
Hospitalization, admitted within 14 days before chemo initiation 15 0.9% 239 1.2%

Outcomes

FN 723 42.5% 8716 44.8% 0.07
Infection Event 1172 68.9% 12833 65.9% < 0.05
Use ER 1069 62.8% 10787 55.4% < 0.05
Infection-related hospitalization 399 23.5% 3873 19.9% < 0.05
All-cause Hospitalization 1327 78.0% 14634 75.1% < 0.05

57

Table 3, Selection Bias Robustness Check and Sensitive Analysis
Model Specifications
Outcome: FN
Outcome:
Infection
Event
Outcome:
Infection-related
Hospitalization
Outcome: All-
cause
Hospitalization
Outcome: ER
HR P-value HR P-value HR P-value HR P-value HR P-value
Core Setting (Unmatched)
Use GCSF prophylaxis 0.75 <.0001 0.94 0.03 1.02 0.69 1.02 0.56 0.99 0.84
1:2 Propensity Score Matching
Use GCSF prophylaxis 0.75 <.0001 0.94 0.03 1.02 0.69 1.00 0.91 0.99 0.84
IPTW
Use GCSF prophylaxis 0.80 <.0001 0.98 0.44 1.11 0.07 1.04 0.17 1.02 0.57
Use GCSF prophylaxis at Day 1/ Day2/ Day3 (Unmatched)
At Day1 0.76 <0.05 1.00 0.91 1.01 0.91 1.02 0.61 1.02 0.69
At Day2 1.04 0.75 0.86 0.18 0.78 0.24 1.04 0.68 0.95 0.66
At Day3 0.69 <0.05 0.87 <0.05 1.08 0.32 0.97 0.57 0.97 0.56
G-prophylaxis * Radiotherapy (Unmatched)
Interaction 1.06 0.62 1.23 <0.05 1.66 <0.05 1.19 0.08 1.04 0.71
G-prophylaxis * Risk Regimen (Unmatched)
High-risk Regimen 0.71 <0.05 0.75 <0.05 0.81 0.25 1.01 0.91 0.89 0.29
Intermediate-risk Regimen 0.57 <0.05 0.87 <0.05 0.94 0.58 0.97 0.65 0.91 0.16
Undocumented-risk Regimen 0.86 <0.05 1.03 0.63 1.07 0.52 1.01 0.87 1.12 0.06
Low-risk Regimen 0.83 <0.05 0.96 0.42 1.11 0.23 1.02 0.72 0.99 0.87

58

Graph 1, Logistic Regression Model for GCSF prophylaxis Use:
*SES factors & Comorbidities were not reported due to no significant findings

Chemotherapy FN Risk Regimen [vs. High]
High
Intermediate
Undocumented
Surgery / Radiotherapy Experience [vs. No]
Pre-chemo surgery
Pre-chemo radiotherapy
Radiotherapy
Immunotherapy [vs. No]
Target Therapy [vs. No]
Antibiotics prophylaxis [vs. No]
Pre-chemo IV antibiotics prophylaxis
IV antibiotics prophylaxis
Pre-chemo oral antibiotics prophylaxis
Oral antibiotics prophylaxis
Female [vs. Male]
Age [vs.>= 65]
Age < 50
50 ≤ Age < 65
Treatment Year Effect [vs. Previous year]
Treatment Season [vs. Summer]
Fall
Winter
Spring
Insurance Plan [vs. HMO]
Exclusive Provider Organization Plan
Indemenity Plan
Other Plan
Point-of-service Plan
PPO Plan
Service Region [vs. South Atlantic]
East North Central
East South Central
Middle Atlantic
Mountain
New England
Pacific
West North Central
West South Central
Unknown
Chemotherapy Service Location [vs. Physician office/ Clinics]
Hospital Out-patient Center
Other Location
Provider Sepcialty [vs. Hematology/Oncology]
Race [vs. White]
Asian
Black
Hispanic
Unknown
Pre-chemo ER/ Hospitalization
Use ER within 14 days before chemo initiation
Hospitalization within 14 days before chemo initiation
Parameter
1.47 [1.21, 1.77]
2.33 [2.03, 2.67]
1.49 [1.31, 1.69]
0.96 [0.81, 1.12]
0.86 [0.74, 0.99]
0.39 [0.32, 0.47]
0.39 [0.29, 0.52]
0.61 [0.35, 1.08]
1.76 [1.34, 2.32]
1.16 [0.64, 2.08]
0.82 [0.53, 1.25]
1.39 [1.04, 1.86]
1.04 [0.94, 1.15]
0.91 [0.79, 1.05]
0.73 [0.52, 1.02]
0.93 [0.91, 0.95]
0.96 [0.83, 1.11]
0.92 [0.80, 1.06]
1.03 [0.89, 1.19]
1.06 [0.79, 1.42]
0.37 [0.18, 0.76]
0.96 [0.84, 1.11]
0.84 [0.70, 1.01]
1.22 [1.01, 1.47]
0.51 [0.43, 0.61]
0.82 [0.65, 1.04]
0.55 [0.44, 0.69]
1.33 [1.09, 1.61]
0.47 [0.36, 0.61]
0.38 [0.29, 0.50]
0.52 [0.19, 1.44]
0.57 [0.48, 0.69]
0.71 [0.59, 0.86]
1.88 [1.46, 2.41]
0.71 [0.51, 1.01]
0.74 [0.59, 0.94]
0.93 [0.62, 1.38]
0.87 [0.73, 1.03]
0.79 [0.61, 1.02]
1.12 [0.92, 1.36]
1.00 [0.81, 1.23]
1.14 [0.99, 1.31]
OR [95% CI]
0.25 0.75 1.25 1.75 2.25 2.75
Odss Ratio (95% CI)
Logistic Regression, Outcome = use GCSF prophylaxis

59

Graph 2, Improper Prophylaxis:
*SES factors & Comorbidities were not reported due to no significant findings

Surgery / Radiotherapy Experience [vs. No]
Pre-chemo surgery
Pre-chemo radiotherapy
Radiotherapy
Immunotherapy [vs. No]
Target Therapy [vs. No]
Antibiotics prophylaxis [vs. No]
Pre-chemo IV antibiotics prophylaxis
IV antibiotics prophylaxis
Pre-chemo oral antibiotics prophylaxis
Oral antibiotics prophylaxis
Female [vs. Male]
Age [vs.>= 65]
Age < 50
50 ≤ Age < 65
Treatment Year Effect [vs. Previous year]
Treatment Season [vs. Summer]
Fall
Winter
Spring
Insurance Plan [vs. HMO]
Exclusive Provider Organization Plan
Indemenity Plan
Other Plan
Point-of-service Plan
PPO Plan
Service Region [vs. South Atlantic]
East North Central
East South Central
Middle Atlantic
Mountain
New England
Pacific
West North Central
West South Central
Unknown
Chemotherapy Service Location [vs. Physician office/ Clinics]
Hospital Out-patient Center
Other Location
Provider Sepcialty [vs. Hematology/Oncology]
Pre-chemo ER/ Hospitalization
Use ER within 14 days before chemo initiation
Hospitalization within 14 days before chemo initiation
Race [vs. White]
Asian
Black
Hispanic
Unknown
Parameter
1.18 [0.96, 1.46]
0.98 [0.81, 1.18]
0.42 [0.34, 0.53]
0.75 [0.51, 1.11]
1.00 [0.00, 2.00]
1.92 [1.30, 2.85]
0.92 [0.35, 2.40]
1.04 [0.59, 1.85]
1.37 [0.88, 2.12]
1.03 [0.89, 1.19]
0.95 [0.77, 1.17]
0.80 [0.50, 1.30]
0.96 [0.94, 0.99]
1.09 [0.88, 1.35]
1.02 [0.83, 1.25]
1.14 [0.92, 1.41]
0.71 [0.44, 1.17]
0.46 [0.19, 1.11]
1.04 [0.85, 1.28]
1.05 [0.80, 1.37]
1.31 [0.99, 1.74]
0.70 [0.54, 0.89]
0.68 [0.47, 0.99]
0.69 [0.50, 0.95]
1.42 [1.06, 1.91]
0.70 [0.48, 1.01]
0.58 [0.42, 0.81]
0.61 [0.19, 2.01]
0.69 [0.52, 0.90]
0.90 [0.68, 1.18]
1.20 [0.86, 1.67]
0.81 [0.54, 1.23]
0.97 [0.72, 1.32]
0.55 [0.39, 0.79]
1.09 [0.87, 1.36]
0.65 [0.37, 1.13]
0.86 [0.67, 1.10]
1.05 [0.76, 1.46]
1.10 [0.83, 1.45]
OR [95% CI]
0.25 0.75 1.25 1.75 2.25
Odss Ratio (95% CI)
Logistic Regression, Outcome = Improper Prophylaxis

60

Graph 3, Core Model:
*SES factors were not reported due to no significant findings

GCSF Prophylaxis [vs. No]
Chemotherapy FN Risk Regimen [vs. Low]
High
Intermediate
Undocumented
Surgery / Radiotherapy Experience [vs. No]
Pre-chemo surgery
Pre-chemo radiotherapy
Radiotherapy
Immunotherapy [vs. No]
Target Therapy [vs. No]
Antibiotics Prophylaxis [vs. No]
IV Antibiotics Prophylaxis
Oral Antibiotics Prophylaxis
Pre-chemo IV Antibiotics Prophylaxis
Pre-chemo Oral Antibiotics Prophylaxis
Female [vs. Male]
Age [vs. ≥65]
Age < 50
50 ≤ Age < 65
Treatment Year Effect [vs. Previous year]
Treatment Season [vs. Summer]
Fall
Winter
Spring
Insurance Plan [vs. Point-of-Service Plan]
Exclusive Provider Organization Plan
Indemenity Plan
Other Plan
Point-of-service Plan
PPO Plan
Service Region [vs. South Atlantic]
East North Central
East South Central
Middle Atlantic
Mountain
New England
Pacific
West North Central
West South Central
Unknown
Race [vs. White]
Asian
Black
Hispanic
Unknown
Chemotherapy Service Location [vs. Physician office/ Clinics]
Hospital Out-patient Center
Other Location
Provider Sepcialty [vs. Hematology/Oncology]
Pre-chemo ER/ Hospitalization
Use ER within 14 days before chemo initiation
Hospitalization within 14 days before chemo initiation
Comorbidities
Myocardial Infarction
Congestive Heart Failure
Peripheral Vascular Disease
Cerebrovascular Disease
Dementia
Chronic Pulmonary Disease
Connective Tissue Disease-Rheumatic Disease
Peptic Ulcer Disease
Mild Liver Disease
Diabetes without complications
Diabetes with complications
Paraplegia and Hemiplegia
Renal Disease
Moderate or Severe Liver Disease
AIDS/HIV
Parameter
0.75 [0.70, 0.81]
1.56 [1.46, 1.68]
1.63 [1.54, 1.73]
1.21 [1.16, 1.28]
0.95 [0.89, 1.01]
1.05 [0.99, 1.11]
0.95 [0.90, 1.01]
0.71 [0.65, 0.78]
1.16 [0.97, 1.38]
1.17 [0.93, 1.46]
0.93 [0.81, 1.05]
1.19 [1.05, 1.35]
1.03 [0.88, 1.20]
1.00 [0.96, 1.04]
0.97 [0.86, 1.10]
0.99 [0.93, 1.04]
1.00 [0.99, 1.00]
1.03 [0.97, 1.09]
1.03 [0.98, 1.09]
1.01 [0.95, 1.07]
0.97 [0.85, 1.10]
0.97 [0.80, 1.18]
1.03 [0.98, 1.09]
1.03 [0.96, 1.10]
0.93 [0.86, 1.02]
1.15 [1.08, 1.23]
1.04 [0.94, 1.14]
1.41 [1.30, 1.53]
0.95 [0.87, 1.05]
1.01 [0.91, 1.13]
0.92 [0.84, 1.00]
0.88 [0.62, 1.24]
1.18 [1.10, 1.27]
0.92 [0.85, 0.99]
0.97 [0.84, 1.12]
1.04 [0.98, 1.11]
1.14 [1.04, 1.24]
1.05 [0.98, 1.11]
0.91 [0.83, 1.00]
1.03 [0.92, 1.15]
0.83 [0.77, 0.90]
1.13 [1.03, 1.23]
1.19 [1.12, 1.26]
1.09 [0.99, 1.20]
1.05 [0.97, 1.13]
1.01 [0.95, 1.08]
1.02 [0.95, 1.09]
1.23 [0.94, 1.60]
1.10 [1.05, 1.14]
1.14 [1.03, 1.28]
1.15 [0.99, 1.34]
0.97 [0.89, 1.06]
1.03 [0.97, 1.09]
1.04 [0.94, 1.14]
0.98 [0.75, 1.27]
1.20 [1.10, 1.31]
0.65 [0.40, 1.06]
1.38 [0.74, 2.55]
HR [95% CI]
0.30 0.50 0.70 0.90 1.10 1.30 1.50 1.70
Hazard Ratio (95% CI)
Cox Regression, Outcome = FN

61

REFERENCE
Aapro, M.S., Bohlius, J., Cameron, D.A., Lago, L.D., Donnelly, J.P., Kearney, N., Lyman, G.H., Pettengell, R.,
Tjan-Heijnen, V.C., Walewski, J., et al. (2011). 2010 update of EORTC guidelines for the use of granulocyte-
colony stimulating factor to reduce the incidence of chemotherapy-induced febrile neutropenia in adult patients
with lymphoproliferative disorders and solid tumours. Eur. J. Cancer 47, 8–32.
American Hospital Formulary Service (2017). AHFS Pharmacologic-Therapeutic Classification.
Austin, P.C. (2010). Statistical Criteria for Selecting the Optimal Number of Untreated Subjects Matched to Each
Treated Subject When Using Many-to-One Matching on the Propensity Score. Am. J. Epidemiol. 172, 1092–
1097.
Barnes, G., Pathak, A., andSchwartzberg, L. (2014a). Pharmacoeconomics of granulocyte colony-stimulating
factor: A critical review. Adv. Ther. 31, 683–695.
Barnes, G., Pathak, A., andSchwartzberg, L. (2014b). G-CSF utilization rate and prescribing patterns in United
States: associations between physician and patient factors and GCSF use. Cancer Med. 3, 1477–1484.
Bennett, C.L., Bishop, M.R., Tallman, M.S., Somerfield, M.R., Feinglass, J., andSmith, T.J. (1999). The
association between physician reimbursement in the US and use of hematopoietic colony stimulating factors as
adjunct therapy for older patients with acute myeloid leukemia: Results from the 1997 American Society of
Clinical Oncology survey. Ann. Oncol. 10, 1355–1359.
Bow, E.J. (2011). Fluoroquinolones, antimicrobial resistance and neutropenic cancer patients. Curr. Opin. Infect.
Dis. 24, 545–553.
Burden, A., Roche, N., Miglio, C., Hillyer, E.V, Postma, D.S., Herings, R.M., Overbeek, J.A., Khalid, J.M.,
vanEickels, D., andPrice, D.B. (2017). An evaluation of exact matching and propensity score methods as applied
in a comparative effectiveness study of inhaled corticosteroids in asthma. Pragmatic Obs. Res. 8, 15–30.
Campos, F.G. (2017). Colorectal cancer in young adults: A difficult challenge. World J. Gastroenterol. 23, 5041–
5044.
Centers for Medicare & Medicaid Services (2006). Billing of Temporary “C” HCPCS Codes by Non-Outpatient
Prospective Payment System (Non-OPPS) Providers.
Choi, M.R., Solid, C. a, Chia, V.M., Blaes, A.H., Page, J.H., Barron, R., andArneson, T.J. (2014). Granulocyte
colony-stimulating factor (G-CSF) patterns of use in cancer patients receiving myelosuppressive chemotherapy.
Support. Care Cancer 22, 1619–1628.
Cui, Y.H., Suh, Y., Lee, H.J., Yoo, K.C., Uddin, N., Jeong, Y.J., Lee, J.S., Hwang, S.G., Nam, S.Y., Kim, M.J., et
al. (2015). Radiation promotes invasiveness of non-small-cell lung cancer cells through granulocyte-colony-
stimulating factor. Oncogene 34, 5372–5382.
Du, X.L., Lairson, D.R., Begley, C.E., andFang, S. (2005). Temporal and geographic variation in the use of
hematopoietic growth factors in older women receiving breast cancer chemotherapy: Findings from a large
population-based cohort. J. Clin. Oncol. 23, 8620–8628.
Dulisse, B., Li, X., Gayle, J.A., Barron, R.L., Ernst, F.R., Rothman, K.J., Legg, J.C., andKaye, J.A. (2013). A
retrospective study of the clinical and economic burden during hospitalizations among cancer patients with febrile
neutropenia. J Med Econ 16, 720–735.

62

Freifeld, A.G., Bow, E.J., Sepkowitz, K.A., Boeckh, M.J., Ito, J.I., Mullen, C.A., Raad, I.I., Rolston, K.V.,
Young, J.A.H., andWingard, J.R. (2011). Clinical practice guideline for the use of antimicrobial agents in
neutropenic patients with cancer: 2010 Update by the Infectious Diseases Society of America. Clin. Infect. Dis.
52.
Goyal, R.K., Tzivelekis, S., Rothman, K.J., Candrilli, S.D., andKaye, J.A. (2018). Time trends in utilization of G-
CSF prophylaxis and risk of febrile neutropenia in a Medicare population receiving adjuvant chemotherapy for
early-stage breast cancer. Support. Care Cancer 26, 539–548.
Johnson, R.H., Chien, F.L., andBleyer, A. (2013). Incidence of breast cancer with distant involvement among
women in the United States, 1976 to 2009. JAMA 309, 800–805.
Kawatkar, A.A., Farias, A.J., Chao, C., Chen, W., Barron, R., Vogl, F.D., andChandler, D.B. (2017).
Hospitalizations, outcomes, and management costs of febrile neutropenia in patients from a managed care
population. Support. Care Cancer 1–9.
Kuderer, N.M., Dale, D.C., Crawford, J., Cosler, L.E., andLyman, G.H. (2006). Mortality, morbidity, and cost
associated with febrile neutropenia in adult cancer patients. Cancer 106, 2258–2266.
Magellan Rx Management (2018). 2017 Medical Pharmacy Trend Report.
Martin Goodman, L., Moeller, M.B., Azzouqa, A.-G., Guthrie, A.E., Dalby, C.K., Earl, M.A., Cheng, C., Pennell,
N.A., Shapiro, M., Velcheti, V., et al. (2016). Reduction of Inappropriate Prophylactic Pegylated Granulocyte
Colony-Stimulating Factor Use for Patients With Non–Small-Cell Lung Cancer Who Receive Chemotherapy: An
ASCO Quality Training Program Project of the Cleveland Clinic Taussig Cancer Institute. J. Oncol. Pract. 12,
e101–e107.
Mitchell, S., Li, X., Woods, M., Garcia, J., Hebard-Massey, K., Barron, R., andSamuel, M. (2016). Comparative
effectiveness of granulocyte colony-stimulating factors to prevent febrile neutropenia and related complications in
cancer patients in clinical practice: A systematic review. J. Oncol. Pharm. Pract. 1078155215625459-.
Murakami, K., Kudo, Y., Kakihana, M., Matsubayashi, J., Okano, T., Shimada, Y., Nagao, T., Kajiwara, N.,
Ohira, T., andIkeda, N. (2018). Granulocyte colony stimulating factor-producing lung cancer with severe anemia
of inflammation. Gen. Thorac. Cardiovasc. Surg. 66, 415–418.
National Cancer Institute Fast Stats: An interactive tool for access to SEER cancer statistics. Surveillance
Research Program.
National Cancer Institute (2009). Common Terminology Criteria for Adverse Events v4.0 (CTCAE).
National Comprehensive Cancer Network (2017). Myeloid Growth Factors. NCCN Clin. Pract. Guidel. Oncol.
National Institutes of Health (2017). Procedure Codes for SEER-Medicare Analyses.
Potosky, A.L., Malin, J.L., Kim, B., Chrischilles, E.A., Makgoeng, S.B., Howlader, N., andWeeks, J.C. (2011).
Use of colony-stimulating factors with chemotherapy: Opportunities for cost savings and improved outcomes. J.
Natl. Cancer Inst. 103, 979–982.
Ramsey, S.D., McCune, J.S., Blough, D.K., McDermott, C.L., Clarke, L., Malin, J.L., andSullivan, S.D. (2010).
Colony-stimulating factor prescribing patterns in patients receiving chemotherapy for cancer. Am. J. Manag. Care
16, 678–686.
Rosenbaum, P.R., andRubin, D.B. (1985). Constructing a Control Group Using Multivariate Matched Sampling
Methods That Incorporate the Propensity Score. Am. Stat. 39, 33–38.

63

ROSENBAUM, P.R., andRUBIN, D.B. (1983). The central role of the propensity score in observational studies
for causal effects. Biometrika 70, 41–55.
Skoetz N, Bohlius J, Engert A, Monsef I, Blank O, andVehreschild JJ (2015). Prophylactic antibiotics or G ( M ) -
CSF for the prevention of infections and improvement of survival in cancer patients receiving myelotoxic
chemotherapy ( Review ). Cochrane Database Syst. Rev.
Smith, T.J., Bohlke, K., Lyman, G.H., Carson, K.R., Crawford, J., Cross, S.J., Goldberg, J.M., Khatcheressian,
J.L., Leighl, N.B., Perkins, C.L., et al. (2015a). Recommendations for the use of WBC growth factors: American
society of clinical oncology clinical practice guideline update. J. Clin. Oncol. 33, 3199–3212.
Smith, T.J., Bohlke, K., Lyman, G.H., Carson, K.R., Crawford, J., Cross, S.J., Goldberg, J.M., Khatcheressian,
J.L., Leighl, N.B., Perkins, C.L., et al. (2015b). Recommendations for the Use of WBC Growth Factors:
American Society of Clinical Oncology Clinical Practice Guideline Update. J. Clin. Oncol. 33, 3199–3212.
Srivastava, R., Ajmeri, A.N., Al-Qawasmi, L., Murphy, C., Lebowicz, Y.Z., Akpanudo, S.M., andJamil, M.O.
(2018). Safety of use of granulocyte colony stimulating factor during concurrent chemotherapy and thoracic
radiotherapy in limited-stage small-cell lung cancer. J. Clin. Oncol. 36, e20570–e20570.
Tan, H., Tomic, K., Hurley, D., Daniel, G., Barron, R., andMalin, J. (2011). Comparative effectiveness of colony-
stimulating factors for febrile neutropenia: a retrospective study. Curr. Med. Res. Opin. 27, 79–86.
Timmer-Bonte, J.N., deBoo, T.M., Smit, H.J., Biesma, B., Wilschut, F. a, Cheragwandi, S. a, Termeer, A.,
Hensing, C. a, Akkermans, J., Adang, E.M., et al. (2005). Prevention of chemotherapy-induced febrile
neutropenia by prophylactic antibiotics plus or minus granulocyte colony-stimulating factor in small-cell lung
cancer: a Dutch Randomized Phase III Study. J. Clin. Oncol. 23, 7974–7984.
Truong, J., Lee, E.K., Trudeau, M.E., andChan, K.K.W. (2016). Interpreting febrile neutropenia rates from
randomized, controlled trials for consideration of primary prophylaxis in the real world: a systematic review and
meta-analysis. Ann. Oncol. Off. J. Eur. Soc. Med. Oncol. 27, 608–618.
Waters, G.E., Corrigan, P., Gatesman, M., andSmith, T.J. (2013). Comparison of pegfilgrastim prescribing
practice to national guidelines at a university hospital outpatient oncology clinic. J. Oncol. Pract. 9, 203–206.
Weycker, D., Li, X., Figueredo, J., Barron, R., Tzivelekis, S., andHagiwara, M. (2016). Risk of chemotherapy-
induced febrile neutropenia in cancer patients receiving pegfilgrastim prophylaxis: does timing of administration
matter? Support. Care Cancer 24, 2309–2316.
Weycker, D., Bensink, M., Lonshteyn, A., Doroff, R., andChandler, D. (2017). Risk of chemotherapy-induced
febrile neutropenia by day of pegfilgrastim prophylaxis in US clinical practice from 2010 to 2015. Curr. Med.
Res. Opin. 33, 2107–2113.
Weycker, D., Hanau, A., Lonshteyn, A., Bowers, C., Bensink, M., Garawin, T., andChandler, D. (2018). Risk of
chemotherapy-induced febrile neutropenia with same-day versus next-day pegfilgrastim prophylaxis among
patients aged ≥65 years: a retrospective evaluation using Medicare claims. Curr. Med. Res. Opin. 0, 1–11.
Wildiers, H., andReiser, M. (2011). Relative dose intensity of chemotherapy and its impact on outcomes in
patients with early breast cancer or aggressive lymphoma. Crit. Rev. Oncol. Hematol. 77, 221–240.
Wingard, J.R., Eldjerou, L., andLeather, H. (2012). Use of antibacterial prophylaxis in patients with
chemotherapy-induced neutropenia. Curr. Opin. Hematol. 19, 21–26.
Woolf, D.K., Slotman, B.J., andFaivre-Finn, C. (2016). The Current Role of Radiotherapy in the Treatment of
Small Cell Lung Cancer. Clin. Oncol. 28, 712–719.

64

Wright, J.D., Neugut, A.I., Ananth, C.V., Lewin, S.N., Wilde, E.T., Lu, Y.-S., Herzog, T.J., andHershman, D.L.
(2013). Deviations From Guideline-Based Therapy for Febrile Neutropenia in Cancer Patients and Their Effect on
Outcomes. JAMA Intern. Med. 173, 559.

65

APPENDIX
Table 1: Pre- and Post- 1:2 Propensity Matching Standardized Difference
Variable
Description
Pre-
Match
Case
Mean
Pre-
Match
Control
Mean
Pre-
Match
Diff
Pre-
Match
Stand
Diff (%)
Post-
Match
Case
Mean
Post-
Match
Control
Mean
Post-
Match
Diff
Post-
Match
Stand
Diff (%)
Pre-
Match
T-test
p-value
Post-
Match
T-test
p-value
FN Risk Regimen – High 0.0946 0.0829 0.0117 4.12 0.0914 0.0879 0.0036 1.26 0.1318 0.6871
FN Risk Regimen - Intermediate 0.2345 0.1200 0.1144 30.30 0.2325 0.2139 0.0186 4.47 <.0001 0.1512
FN Risk Regimen - Undocumented 0.3051 0.3235 -0.0185 -3.98 0.3070 0.3076 -0.0007 -0.14 0.1383 0.9639
Immunotherapy 0.0291 0.0799 -0.0508 -22.51 0.0294 0.0216 0.0078 4.97 <.0001 0.1210
Target-therapy 0.0168 0.0167 0.0001 0.07 0.0150 0.0163 -0.0013 -1.05 0.9786 0.7387
Radiotherapy 0.0758 0.1700 -0.0942 -29.00 0.0764 0.0705 0.0059 2.25 <.0001 0.4690
Pre-chemo Radiotherapy 0.1477 0.1962 -0.0486 -12.90 0.1489 0.1437 0.0052 1.48 <.0001 0.6359
IV Antibiotics Prophylaxis 0.00842 0.00720 0.0012 1.38 0.00849 0.00751 0.0010 1.10 0.6159 0.7281
Oral Antibiotics Prophylaxis 0.0330 0.0247 0.0083 4.97 0.0307 0.0320 -0.0013 -0.75 0.0777 0.8114
Pre-chemo Oral Antibiotics Prophylaxis 0.0117 0.0143 -0.0027 -2.37 0.0118 0.00980 0.0020 1.90 0.3527 0.5504
Male 0.5065 0.5176 -0.0111 -2.23 0.5062 0.4997 0.0065 1.31 0.4038 0.6765
Age between 50 and 64 0.2843 0.2812 0.0031 0.69 0.2867 0.2828 0.0039 0.87 0.7966 0.7814
Age < 50 0.0278 0.0336 -0.0058 -3.34 0.0268 0.0268 0.0000 0.00 0.1925 1.0000
Year Trend 4.9372 5.5385 -0.6013 -20.74 4.9536 4.9226 0.0310 1.09 <.0001 0.7235
Season: Winter 0.2798 0.2932 -0.0134 -2.96 0.2802 0.2825 -0.0023 -0.51 0.2695 0.8711
Season: Spring 0.2487 0.2379 0.0108 2.52 0.2476 0.2541 -0.0065 -1.51 0.3411 0.6308
Season: Fall 0.2306 0.2300 0.0005 0.12 0.2319 0.2266 0.0052 1.24 0.9627 0.6910
Insurance Plan: EPO 0.0421 0.0336 0.0085 4.48 0.0418 0.0382 0.0036 1.83 0.1079 0.5611
Insurance Plan: IND 0.00518 0.00929 -0.0041 -4.85 0.00523 0.00588 -0.0007 -0.88 0.0381 0.7768
Insurance Plan: OTH 0.3277 0.3564 -0.0287 -6.04 0.3292 0.3383 -0.0091 -1.94 0.0246 0.5361
Insurance Plan: POS 0.1742 0.1876 -0.0134 -3.47 0.1757 0.1803 -0.0046 -1.20 0.1985 0.7031
Insurance Plan: PPO 0.1328 0.0964 0.0364 11.44 0.1313 0.1277 0.0036 1.07 <.0001 0.7320
Area: EAST NORTH CENTRAL 0.1140 0.1539 -0.0399 -11.73 0.1143 0.1185 -0.0042 -1.32 <.0001 0.6733
Area: EAST SOUTH CENTRAL 0.0544 0.0521 0.0023 1.01 0.0549 0.0575 -0.0026 -1.13 0.7029 0.7180
Area: MIDDLE ATLANTIC 0.0648 0.0789 -0.0141 -5.46 0.0653 0.0640 0.0013 0.53 0.0333 0.8651
Area: MOUNTAIN 0.1108 0.0674 0.0433 15.25 0.1110 0.1009 0.0101 3.29 <.0001 0.2900
Area: NEW ENGLAND 0.0363 0.0480 -0.0118 -5.86 0.0366 0.0369 -0.0003 -0.17 0.0201 0.9558
Area: PACIFIC 0.0518 0.0964 -0.0446 -17.07 0.0516 0.0434 0.0082 3.84 <.0001 0.2265
Area: UNKNOWN 0.00324 0.00398 -0.0007 -1.24 0.00327 0.00359 -0.0003 -0.56 0.6273 0.8573
EDU: Less than Bachelor Degree 0.4832 0.4831 0.0001 0.01 0.4820 0.4889 -0.0069 -1.37 0.9960 0.6612
EDU: Bachelor Degree Plus 0.0868 0.0890 -0.0022 -0.78 0.0875 0.0839 0.0036 1.28 0.7721 0.6810
EDU: Unknown 0.0505 0.0611 -0.0106 -4.62 0.0496 0.0480 0.0016 0.76 0.0723 0.8083

66

Variable
Description
Pre-
Match
Case
Mean
Pre-
Match
Control
Mean
Pre-
Match
Diff
Pre-
Match
Stand
Diff (%)
Post-
Match
Case
Mean
Post-
Match
Control
Mean
Post-
Match
Diff
Post-
Match
Stand
Diff (%)
Pre-
Match
T-test
p-value
Post-
Match
T-test
p-value
Home Ownership: Unknown 0.2785 0.2872 -0.0087 -1.94 0.2776 0.2727 0.0049 1.10 0.4683 0.7258
Home Ownership” Probable Renter 0.0330 0.0387 -0.0057 -3.07 0.0333 0.0274 0.0059 3.43 0.2345 0.2813
Income: Unknown 0.2008 0.2163 -0.0155 -3.81 0.2005 0.1999 0.0007 0.16 0.1572 0.9584
Income: $40K-$49K 0.0816 0.0801 0.0015 0.56 0.0816 0.0816 0.0000 0.00 0.8324 1.0000
Income: $50K-$59K 0.0790 0.0762 0.0029 1.07 0.0777 0.0738 0.0039 1.48 0.6863 0.6349
Income: $60K-$74K 0.0952 0.0913 0.0039 1.34 0.0954 0.0927 0.0026 0.89 0.6137 0.7745
Income: $75K-$99K 0.1023 0.1052 -0.0029 -0.95 0.1025 0.1091 -0.0065 -2.12 0.7232 0.4996
Income: $100K+ 0.1367 0.1360 0.0007 0.19 0.1365 0.1434 -0.0069 -1.98 0.9430 0.5292
Net Worth: Unknown 0.1541 0.1672 -0.0130 -3.54 0.1535 0.1541 -0.0007 -0.18 0.1895 0.9539
Net Worth: $25K-$149K 0.2189 0.1981 0.0208 5.13 0.2195 0.2113 0.0082 1.99 0.0581 0.5249
Net Worth: $150K-$249K 0.1256 0.1189 0.0068 2.07 0.1267 0.1205 0.0062 1.88 0.4337 0.5456
Net Worth: $250K-$499K 0.1937 0.1828 0.0109 2.78 0.1927 0.1933 -0.0007 -0.17 0.2920 0.9579
Net Worth: $500K+ 0.1522 0.1595 -0.0073 -2.02 0.1515 0.1633 -0.0118 -3.23 0.4531 0.3049
Race: Asian 0.0168 0.0219 -0.0051 -3.69 0.0170 0.0153 0.0016 1.29 0.1433 0.6817
Race: Black 0.1179 0.1177 0.0002 0.06 0.1189 0.1199 -0.0010 -0.30 0.9815 0.9232
Race: Hispanic 0.0453 0.0555 -0.0102 -4.64 0.0457 0.0474 -0.0016 -0.78 0.0700 0.8050
Race: Unknown 0.1250 0.1352 -0.0102 -3.03 0.1241 0.1205 0.0036 1.10 0.2627 0.7256
Location: ACUTE-CARE OUTPATIEN 0.2934 0.2805 0.0129 2.86 0.2933 0.3008 -0.0075 -1.64 0.2815 0.6000
Location: Other 0.0369 0.0683 -0.0314 -14.10 0.0372 0.0294 0.0078 4.37 <.0001 0.1709
Spec: GENERAL ACUTE-CARE HOSPIT 0.2876 0.2785 0.0090 2.01 0.2880 0.2949 -0.0069 -1.51 0.4495 0.6302
Spec: INTERNAL/FAMILY MEDICINE 0.0453 0.0635 -0.0181 -7.99 0.0457 0.0376 0.0082 4.09 0.0013 0.1987
Spec: OTHER 0.0233 0.0393 -0.0160 -9.19 0.0229 0.0245 -0.0016 -1.07 0.0001 0.7330
Myocardial Infarction 0.0551 0.0516 0.0034 1.52 0.0542 0.0581 -0.0039 -1.70 0.5635 0.5887
Congestive Heart Failure 0.0816 0.0781 0.0036 1.31 0.0823 0.0807 0.0016 0.60 0.6198 0.8486
Peripheral Vascular Disease 0.1179 0.1256 -0.0078 -2.37 0.1182 0.1094 0.0088 2.78 0.3785 0.3725
Cerebrovascular Disease 0.1172 0.1099 0.0073 2.31 0.1169 0.1202 -0.0033 -1.01 0.3810 0.7474
Dementia 0.00583 0.00562 0.0002 0.27 0.00588 0.00849 -0.0026 -3.09 0.9185 0.3081
Chronic Pulmonary Disease 0.4333 0.3961 0.0372 7.55 0.4331 0.4105 0.0225 4.56 0.0044 0.1445
Connective Tissue Disease-Rheumatic Disease 0.0376 0.0323 0.0053 2.87 0.0379 0.0372 0.0007 0.34 0.2959 0.9125
Peptic Ulcer Disease 0.0194 0.0163 0.0031 2.36 0.0196 0.0229 -0.0033 -2.27 0.3928 0.4636
Mild Liver Disease 0.0764 0.0609 0.0156 6.16 0.0764 0.0771 -0.0007 -0.25 0.0269 0.9376
Diabetes without complications 0.2073 0.1893 0.0180 4.51 0.2064 0.2182 -0.0118 -2.87 0.0863 0.3601
Diabetes with complications 0.0563 0.0560 0.0004 0.15 0.0568 0.0555 0.0013 0.57 0.9541 0.8559
Paraplegia and Hemiplegia 0.00583 0.00626 -0.0004 -0.55 0.00588 0.00621 -0.0003 -0.42 0.8383 0.8934
Renal Disease 0.0615 0.0631 -0.0015 -0.64 0.0614 0.0614 0.0000 0.00 0.8115 1.0000

67

Variable
Description
Pre-
Match
Case
Mean
Pre-
Match
Control
Mean
Pre-
Match
Diff
Pre-
Match
Stand
Diff (%)
Post-
Match
Case
Mean
Post-
Match
Control
Mean
Post-
Match
Diff
Post-
Match
Stand
Diff (%)
Pre-
Match
T-test
p-value
Post-
Match
T-test
p-value
Moderate or Severe Liver Disease 0.00324 0.00196 0.0013 2.51 0.00327 0.00196 0.0013 2.56 0.3902 0.4323
AIDS/HIV 0.00130 0.000948 0.0003 1.04 0.00131 0.000327 0.0010 3.43 0.7140 0.3173
*All categorical variables are translated to dummies (0/1)
Table 2: Detailed Core Model Result
# of the sample used in the Cox regression
Outcome
Total
Freq.
Outcome
Total
Freq.
Outcome
Total
Freq.
Outcome
Total
Freq.
Outcome
Total
Freq.
Event (-),
censored
11738
Event (-),
censored
7172
Event (-),
censored
16905
Event (-),
censored
5216
Event (-),
censored
9321
Event (+) 9439 Event (+) 14005 Event (+) 4272 Event (+) 15961 Event (+) 11856
Parameters
Outcome: FN
Outcome: any Infection
event (including
hospitalization)
Outcome: Infection-
related Hosp
Outcome: anytime All-
cause Hosp
Outcome: anytime ER
Hazard Ratio P-value Hazard Ratio P-value Hazard Ratio P-value Hazard Ratio P-value Hazard Ratio P-value
Use GCSF Prophylaxis within 3 days of chemo
initiation
0.747 <.0001 0.936 0.0335 1.021 0.6949 1.017 0.562 0.993 0.8361
Concomitant radiotherapy (within 7-day of chemo
initiation)
0.945 0.0544 0.938 0.0078 0.945 0.1779 1.238 <.0001 0.932 0.0057
Risk regimen: Undocumented [ref= low] 1.209 <.0001 0.988 0.5704 1.063 0.1032 0.799 <.0001 0.97 0.1788
Risk regimen: High [ref= low] 1.557 <.0001 0.966 0.2854 1.186 0.002 1.2 <.0001 1.008 0.8281
Risk regimen: Intermediate [ref= low] 1.628 <.0001 0.994 0.8262 1.162 0.0016 1.016 0.5308 1.051 0.0828
Use immunotherapy drug [ref=no] 0.706 <.0001 0.893 0.0016 0.87 0.0312 0.732 <.0001 0.91 0.0096
Use target therapy drug [ref=no] 1.151 0.1196 0.944 0.4709 1.021 0.8763 0.874 0.0645 0.835 0.0426
Pre-chemo surgery 0.943 0.0664 0.763 <.0001 0.633 <.0001 0.996 0.8727 0.776 <.0001
Pre-chemo radiotherapy 1.045 0.1219 1.199 <.0001 1.309 <.0001 1.249 <.0001 1.239 <.0001
IV antibiotics prophylaxis (within 7-day of chemo
initiation)
1.16 0.1994 0.89 0.2491 0.88 0.472 1.607 <.0001 1.031 0.7469
Oral antibiotics prophylaxis (within 7-day of chemo
initiation)
0.921 0.2153 1.046 0.3976 0.971 0.7665 0.858 0.0032 1.028 0.6533
Pre-chemo IV antibiotics prophylaxis (1~7day
before chemo initiation)
1.184 0.0099 0.876 0.0215 1.095 0.3223 1.07 0.1551 0.965 0.5063
Pre-chemo oral antibiotics prophylaxis (1~7day
before chemo initiation)
1.022 0.7867 1.12 0.0937 0.941 0.6303 1.097 0.1539 1.064 0.415
Gender, Female [ref=Male] 0.992 0.6912 1.072 <.0001 0.836 <.0001 0.986 0.3857 0.941 0.0011
Age < 50 [ref= 65+] 0.968 0.5988 0.927 0.1526 0.945 0.5471 1.091 0.0845 0.711 <.0001
50 <= Age < 65 [ref= 65+] 0.981 0.5065 0.958 0.0677 0.944 0.174 0.975 0.2467 0.842 <.0001
Year trend 0.99 0.0141 1.009 0.0051 1.027 <.0001 1.129 <.0001 1.017 <.0001
Treatment season: Fall [ref= summer] 1.025 0.4262 1.001 0.976 1.077 0.1009 1.029 0.2324 1.021 0.4412
Treatment season: Spring [ref= summer] 1.001 0.9728 0.997 0.8957 0.98 0.6426 1.007 0.7568 0.999 0.9573
Treatment season: Winter [ref=summer] 1.027 0.3606 1.051 0.0346 0.989 0.8009 1.03 0.1853 0.997 0.9213
Insurance plan: Exclusive Provider Organization
Plan [ref= HMO]
0.963 0.5669 0.838 0.0016 1.134 0.2013 0.784 <.0001 0.106 <.0001
Insurance plan: Indemenity Plan [ref= HMO] 0.968 0.7405 0.869 0.1017 0.845 0.3074 0.897 0.198 0.052 <.0001
Insurance plan: Other Plan [ref= HMO] 1.028 0.3531 0.917 0.0004 1.134 0.0038 1.539 <.0001 0.987 0.5902
Insurance plan: POS plan [ref= HMO] 1.022 0.5575 0.887 0.0001 1.191 0.0023 0.842 <.0001 0.11 <.0001
Insurance plan: PPO Plan [ref= HMO] 0.93 0.0924 0.962 0.2589 1.194 0.0035 1.345 <.0001 0.958 0.2057
Division: EAST NORTH CENTRAL [ref= South
Atlantic]
1.15 <.0001 0.957 0.1208 1.054 0.2663 1.149 <.0001 1.046 0.1323
Division: EAST SOUTH CENTRAL [ref= South
Atlantic]
1.031 0.5523 1.065 0.1225 1.027 0.7102 1.03 0.4158 1.045 0.3137
Division: MIDDLE ATLANTIC [ref= South Atlantic] 1.407 <.0001 1.039 0.2829 1.07 0.2551 1.048 0.1269 1.116 0.0022
Division: MOUNTAIN [ref= South Atlantic] 0.947 0.2597 0.835 <.0001 0.664 <.0001 0.63 <.0001 1.101 0.0243
Division: NEW ENGLAND [ref= South Atlantic] 1.008 0.8789 0.998 0.968 1.083 0.2649 1.071 0.0651 1.126 0.0048
Division: Other [ref= South Atlantic] 0.911 0.0486 0.861 <.0001 0.643 <.0001 0.611 <.0001 0.869 0.0014
Division: PACIFIC [ref= South Atlantic] 0.871 0.4366 0.764 0.0622 0.587 0.1953 0.419 <.0001 0.995 0.9766
Division: WEST NORTH CENTRAL [ref= South
Atlantic]
1.179 <.0001 0.849 <.0001 0.986 0.7915 1.165 <.0001 1.078 0.0217
Division: WEST SOUTH CENTRAL [ref=South
Atlantic]
0.916 0.0244 0.957 0.1665 0.689 <.0001 0.669 <.0001 1.021 0.5817
Race: Asian [ref= White] 0.964 0.6241 0.85 0.008 0.9 0.3788 0.969 0.589 0.869 0.0459
Race: Black [ref= White] 1.038 0.2557 0.933 0.0101 1.012 0.7985 1.031 0.2138 1.098 0.001
Race: Hispanic [ref= White] 1.132 0.0061 0.973 0.4813 0.985 0.8389 1.018 0.6394 1.001 0.9869
Race: Unknown [ref= White] 1.041 0.1957 0.934 0.0082 0.895 0.0195 0.947 0.0247 1.028 0.3168
Location: ACUTE-CARE OUTPATIEN
[ref=OFFICE/CLINIC]
0.91 0.0452 1.062 0.1111 1.254 0.0028 1.818 <.0001 1.233 <.0001
Location: UNKNOWN [ref=OFFICE/CLINIC] 1.021 0.7249 0.978 0.6268 0.116 <.0001 0.248 <.0001 0.138 <.0001
Specialty: Non-Hematology/Oncology
[ref=HEMATOLOGY & ONCOLOGY]
0.828 <.0001 0.984 0.6487 0.829 0.0101 1.055 0.141 0.897 0.0088
Use ER within 14 days before chemotherapy
initiation [ref=No]
1.121 0.0111 1.24 <.0001 1.149 0.0326 1.119 0.0009 1.84 <.0001
Hospitalization within 14 days before
chemotherapy initiation [ref=No]
1.185 <.0001 1.369 <.0001 1.27 <.0001 1.207 <.0001 1.147 <.0001

68

Myocardial Infarction 1.081 0.1068 1.038 0.3493 1.099 0.1679 1.034 0.3833 1.012 0.7847
Congestive Heart Failure 1.036 0.3826 1.213 <.0001 1.193 0.0022 1.098 0.003 1.191 <.0001
Peripheral Vascular Disease 1.011 0.7407 1.052 0.0664 0.956 0.3654 1.016 0.529 1.038 0.2028
Cerebrovascular Disease 1.005 0.8784 1.048 0.1067 1.108 0.0431 1.074 0.0089 1.088 0.0062
Dementia 1.242 0.1107 1.263 0.0353 1.622 0.0059 1.215 0.0599 1.331 0.011
Chronic Pulmonary Disease 1.086 0.0003 1.17 <.0001 1.139 0.0001 1.023 0.1968 1.065 0.0019
Connective Tissue Disease-Rheumatic Disease 1.129 0.03 1.173 0.0006 1.092 0.2906 1.075 0.1067 1.213 <.0001
Peptic Ulcer Disease 1.146 0.0838 0.978 0.7445 0.953 0.6845 0.94 0.3161 1.134 0.0622
Mild Liver Disease 0.973 0.539 0.947 0.1325 1.016 0.8082 1.014 0.6753 0.922 0.0417
Diabetes without complications 1.014 0.6461 1.036 0.1487 0.998 0.9648 1.047 0.0494 1.025 0.3412
Diabetes with complications 1.035 0.4992 1.006 0.8841 1.071 0.3513 1.058 0.1518 1.068 0.1258
Paraplegia and Hemiplegia 0.973 0.8414 0.952 0.6662 1.124 0.5235 0.983 0.8686 1.148 0.2254
Renal Disease 1.185 0.0001 1.109 0.0043 1.252 0.0005 1.02 0.5836 1.122 0.0026
Moderate or Severe Liver Disease 0.677 0.1209 1.312 0.123 0.821 0.5801 1.043 0.8015 1.193 0.3568
AIDS/HIV 1.341 0.3558 0.931 0.7908 0.756 0.6288 1.332 0.2127 0.92 0.7637

Graph 1: Propensity Score Distribution – Before Matching

Graph 2: Propensity Score Distribution – After 1:2 Matching

69

Chapter 5 - Conclusion

In the past 20 years, the market of specialty drugs, which including most injectable and biological agents,
has increased significantly in the US. These agents are usually used for severe diseases with limited alternative
treatment options and have much higher costs compared to traditional medications. With the increasing demand
for specialty drugs and the fast growth of specialty drug expenditure, the management of specialty drugs has
become a critical issue for policy stakeholders, insurance payers, and health care providers. The objective of this
dissertation was to provide an evaluation of some of the tools used in the U.S. health care system to control the
price and utilization of specialty drugs.
Three main approaches for healthcare management are commonly implemented in specialty drug
management – price controls, insurance benefit design, and utilization control which often using treatment
guidelines and care management. (Tu andSamuel, 2012) Insurance design interventions often use formulary
constraints and tier-based copayments designed to force the use of substitutes or limit the access to certain
specialty drugs based on the characteristics of group plan purchasers. The formulary restrictions and tier-based
copayments decreased the financial risk for payers by reducing negotiated prices via discounts and by reducing
the unnecessary use of specialty drugs. However, these approaches are criticized for limiting treatment options
and transferring financial burden to patients. (Goldman et al., 2006)
The 340B program is a Federal program which mandates that eligible healthcare entities be allowed to
purchase drugs with a significant price discount, the so-call 340B price. Healthcare systems have to serve a
proportional disadvantage population to obtain the permission of enrolling the 340B program and using 340B
price. Since the accessibility of 340B price is transferable, there is also a trend among healthcare systems to
acquire 340B-eligible-ready entities, especially the 340B-eligible oncology clinics, in the past ten years. (Butcher,
2014; Conti andBach, 2014) Because the 340B program is designed to benefit uninsured and indigent patient
covered by government programs, other insurance payers do not benefit from the 340B price and patients do not
pay drug prices directly. The drug price approach in the case of the 340B program, therefore, only reduces the
financial risk among healthcare providers serving indigent patients. (Conti and Bach, 2013)
Utilization management is the most common healthcare management tool used by insurance providers.
(Tu andSamuel, 2012) Prior-authorization, for instance, is widely practiced in drug and medical intervention
management among insurance payers and requires providers to submit evidence of the proper use of expensive
interventions. Proper use is frequently measured as adherence to specific treatment guidelines which must be
documented before initiating treatment. Ideally, insurance payers aim to avoid unnecessary treatment and reduce

70

healthcare resource waste. Though the use of utilization management might result in provider resistance or
provider withdraw from the network, this approach has the potential to benefit patients by improving patient
outcomes as adherence to evidence-based guideline ensures the effectiveness of treatment, decrease the risk of
adverse events and improves economic benefits in the long run. (Cabana et al., 1999)
In this dissertation study, we evaluate selected drug pricing and utilization management tools used in
specialty drug management. We first investigate issues related to specialty drug pricing by documenting price
trends for specialty drugs in the 340B program. With the rapid expansion of 340B entities and the growth of
specialty drug use, information about specialty drug price trends under the 340B program is vital for policy
stakeholders and pharmacy managers. We compared the trend of specialty drug price in 340B program with
regular pharmacy purchasing price and found the price trends in these two price systems are similar. While minor
differences were observed between different therapeutic classes, the overall price trend of specialty drug is
proportionate to the average drug price inflation rate in the US. The fast-increasing drug prices among specialty
drugs, therefore, need extra attention in the development of management methods.
We subsequently evaluated clinical guideline adherence for the use of a vital specialty drug in oncology,
granulocyte colony-stimulating factors (GCSF), which is used to prevent febrile neutropenia (FN) and delays in
chemotherapy in cancer patients. Febrile neutropenia is an immune suppressive state characterized by an absolute
neutrophil count (ANC) <1000/mm
3
with concomitant systemic temperature >38.3°C, or a sustained body
temperature of ≥38°C for more than 1 hr. FN is a frequent but severe side-effect associated with
myelosuppressive chemotherapy administration that commonly requires hospitalization and the initiation of board
spectrum antibiotic treatment. Due to the life-threatening nature of FN, patients often require hospital stays,
leading to increased mortality and higher healthcare cost. To prevent or ameliorate chemotherapy-induced FN,
with randomized controlled trials showing GCSF effectively increases ANC and thus reduces the risk of FN in
oncology patients receiving myelosuppressive chemotherapies.
The substantial cost of GCSF imposes significant burdens on healthcare payers and patients and thus is a
significant contributor to the total cost of cancer-related healthcare service in the US. Therefore, several clinical
associations, such as the NCCN and ASCO, have enacted guidelines in GCSF use. These guidelines
recommended using GCSF prophylaxis on patients with high-risk factors, especially the high-FN-risk
chemotherapy regimens users. For the low-risk patients, such as the low-FN-risk regimens users, using GCSF
prophylaxis was not recommended. Both clinical guidelines also listed patient risk factors and the FN risk level
among chemotherapy regimens based on clinical studies. With the evidence-based design, these clinical
guidelines have been widely recognized by clinicians and been utilized by insurance payers as the reference of
proper GCSF use in the utilization management approach.

71

In our analysis, we first discussed the factors associated with the use of GCSF prophylaxis, from patient
characteristics, social-economic status, treatment characteristics to provider characteristics. Then we used the
guideline-recommended risk factors to model the effectiveness of GCSF prophylaxis in terms of health outcomes
in FN, infection and hospitalization events. While the current GCSF guidelines did not specialize in the type of
cancer, we separate the discussion of GCSF prophylaxis on health outcomes by the type of cancer. We choose
breast cancer and lung cancer patients as our analytic cohort, due to their high prevalence in the US population
and the high frequency of using chemotherapy as the treatment strategy.
Our analysis results showed the use of GCSF prophylaxis is strongly associated with the choice of FN-
risk regimen. Nevertheless, not all high-risk regimen users received GCSF prophylaxis. We found 46% high-risk
regimen users received GCSF prophylaxis among breast cancer patients. For lung cancer patients, only 9% of
high-risk regimen users took GCSF prophylaxis. Though the guidelines recommended no prophylaxis for the low-
risk regimen users, we still find about 9% low-risk regimen users being prescribed GCSF prophylaxis in breast
cancer, and 5% of the low-risk regimen users used GCSF prophylaxis in lung cancer, respectively.
We are not surprised by the finding of different prophylaxis rates within each risk regimens by the type of
cancer since other risk factors would affect the use of prophylaxis. The more important question is, whether the
use of GCSF prophylaxis among each risk regimens provided meaningful protection in FN and other health
outcomes. Our analytic results showed using the prophylaxis on the low-risk regimen users provided no
significant FN prevention effect in breast cancer (HR=0.9, P>0.05). Contradicted to the finding in breast cancer,
using GCSF prophylaxis with the low-risk regimen still provided significant FN protection in lung cancer patients
(HR=0.83, P<0.05). The inconsistent effectiveness of using GCSF prophylaxis with FN risk regimen by cancer
suggested us the need for discussion in cancer-specific GCSF prophylaxis.
Our analysis also showed the use of GCSF prophylaxis might not necessarily benefit health outcomes
other than FN. In breast cancer, using GCSF prophylaxis was associated with reduced risks in FN and mortality
event only (FN: HR=0.24, P<0.05; 60-day mortality: HR=0.22, P<0.05). In lung cancer, only the risk of FN event
was reduced significantly (HR=0.74, P<0.05). The finding suggests the use of GCSF prophylaxis did not alleviate
the hospitalization event at all. While further studies are required to document the association between the GCSF
prophylaxis and the hospitalization-related outcomes, our finding suggested the economic benefit brought by the
prophylaxis might be limited since the no-benefit was found on the most costly outcome.
There are several retrospective studies focused on the association between GCSF prophylaxis and health
outcomes. Most of the studies were done by using the Medicare-SEER database. The Medicare-SEER database
provided comprehensive information in patient characteristics, diagnosis history, medication use, and cancer
stage. However, due to the nature of the Medicare program, the Medicare-SEER database has a relatively

72

restrictive patient population with older than 65 or with end-stage renal diseases. By using commercial insurance
database with broader age groups, our analytic results are more generalized and represented for common breast/
lung cancer patients, especially for the younger cohorts.
Finally, our analysis provided additional discussion for using antibiotics prophylaxis to against FN event.
No significant prophylaxis effects were found for the use of IV/oral antibiotics in FN prevention. While we saw a
risk-increase of using IV prophylaxis before the initiation of chemotherapy, this could be linked to the higher
baseline infection risk among patients. Considering our analysis did not optimize in the use of antibiotics
prophylaxis, further studies are required to establish the link between the use of antibiotic prophylaxis and health
outcomes.
The rapidly emerging and the sky-rocketing expense of specialty drugs have raised concerns in the
healthcare system from payers to providers and need managemental interventions. Innovative management
approaches are emerging in the healthcare field to manage the use of specialty drugs, such as outcome-based
contract payment (Amgen, 2017) and value-based payment. (Bentley et al., 2017) With the increased complexity
in the specialty drug category and the reimbursement policy, it is essential to have a throughout understanding
from the effectiveness of the outcomes for the specialty drug management tools. In this dissertation study, we
have shown the power of using real-world data, from pharmacy purchasing records to health insurance claims,
and retrospective studies to investigate the trend, utilization and the outcomes associated to the management of
specialty drugs. To evaluate the new management tools for specialty drug, we believe the use of real-world data
support will be indispensable. We expect in the future, with the interdisciplinary support from different real-world
databases, such as insurance claims, medical records, and social-economic status survey results, the more-detailed
interaction between patients, providers, and payers in terms of specialty drugs will be depicted and finally
improve the effectiveness, and the outcomes of specialty drug management approaches.

73

REFERENCES
Amgen (2017). Amgen And Harvard Pilgrim Agree To First Cardiovascular Outcomes-Based Refund Contract
For Repatha® (Evolocumab).
Bentley, T.G.K., Cohen, J.T., Elkin, E.B., Huynh, J., Mukherjea, A., Neville, T.H., Mei, M., Copher, R., Knoth,
R., Popescu, I., et al. (2017). Measuring the Value of New Drugs: Validity and Reliability of 4 Value Assessment
Frameworks in the Oncology Setting. J. Manag. Care Spec. Pharm. 23, S34–S48.
Butcher, L. (2014). Unintended Consequences: How Government Policies Have Increased the Cost of Cancer
Care—Part II: 340B Drug Discounts Have Fueled the Migration of Cancer Care. Oncol. Times 36, 10–12.
Cabana, M.D., Rand, C.S., Powe, N.R., Wu, A.W., Wilson, M.H., Abboud, P.A., andRubin, H.R. (1999). Why
don’t physicians follow clinical practice guidelines? A framework for improvement. JAMA 282, 1458–1465.
Conti, R.M., and Bach, P.B. (2013). Cost consequences of the 340B drug discount program. JAMA 309, 1995–
1996.
Conti, R.M., and Bach, P.B. (2014). The 340B drug discount program: hospitals generate profits by expanding to
reach more affluent communities. Health Aff. (Millwood). 33, 1786–1792.
Goldman, D.P., Joyce, G.F., Lawless, G., Crown, W.H., andWilley, V. (2006). Benefit design and specialty drug
use. Health Aff. 25, 1319–1331.
Tu, H.T., andSamuel, D.R. (2012). Limited options to manage specialty drug spending. Res. Brief 1–13.

Abstract (if available)

Abstract The past 20 years have seen a substantial increase in total specialty drug spend in the US healthcare market. Specialty drugs also have attracted the attention of pharmaceutical companies and have thus become the focus of increasing research investment and sale. The utilization and pricing of specialty drugs have been subjected to efforts to manage price and utilization. This dissertation aimed to evaluate the price and clinical impact of selected examples of specialty drug management in the US healthcare system. We first illustrated the specialty drug price trend in the 340B drug discount program. A specialty drug, granulocyte-colony stimulating factor, was further analyzed for its effectiveness in chemotherapy-induced febrile neutropenia prevention among lung cancer and breast cancer patients by using a real-world health insurance database. Finally, we answered the policy question of how the prescribing of GCSF prophylaxis being affected by non-clinical factors.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The value of novel antihyperlipidemic treatments in the U.S. healthcare system: Reducing the burden of cardiovascular diseases and filling the gap of low adherence in statins

PDF

Pharmacotherapy among patients with type 2 diabetes

PDF

Value in oncology care and opportunities for improvement

PDF

Outcomes of antibiotic use among children with acute respiratory tract infections

PDF

Impact of pharmacy-based transitional care on healthcare utilization and costs

PDF

The effects of tobacco exposure on hormone levels and breast cancer risk among young women

PDF

Three essays on comparative economic and clinical outcomes of antipsychotics using retrospective claims data

PDF

Multilevel sociodemographic correlates of the health and healthcare utilization of childhood cancer survivors

PDF

Economic, clinical, and behavioral outcomes from medical and pharmaceutical treatments

PDF

Understanding primary nonadherence to medications and its associated healthcare outcomes: a retrospective analysis of electronic medical records in an integrated healthcare setting

PDF

Effects of a formulary expansion on the use of atypical antipsychotics and health care services by patients with schizophrenia in the California Medicaid Program

PDF

Evaluating cancer treatments with health economics methods

PDF

Investigating the complexity of the tumor microenvironment's role in drug response

PDF

Factors that may impact drug disposition and metabolism

PDF

Documenting the impact of under-treatment for chronic disorders

PDF

Total systems engineering evaluation of invasive pediatric medical therapies conducted in non-clinical environments

PDF

The disease burden among aging people living with HIV/AIDS

PDF

Comparison of Cox regression and machine learning methods for survival analysis of prostate cancer

PDF

Value in health in the era of vertical integration

PDF

The causal-effect of childhood obesity on asthma in young and adolescent children

Asset Metadata

Creator Lee, Cho-Han (author)

Core Title The price and clinical impact of specialty drug management in the US healthcare system

School School of Pharmacy

Degree Doctor of Philosophy

Degree Program Health Economics

Publication Date 07/30/2021

Defense Date 08/15/2019

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

Tag 340B,breast cancer,febrile neutropenia,GCSF,granulocyte-colony stimulating factor,lung cancer,OAI-PMH Harvest,specialty drug

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor McCombs, Jeffrey S. (committee chair)

Creator Email chohanle@usc.edu,hsnu101105@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-200616

Unique identifier UC11663339

Identifier etd-LeeChoHan-7697.pdf (filename),usctheses-c89-200616 (legacy record id)

Legacy Identifier etd-LeeChoHan-7697.pdf

Dmrecord 200616

Document Type Dissertation

Format application/pdf (imt)

Rights Lee, Cho-Han

Type texts

Source 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 a...

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