University of Southern California Dissertations and Theses

Predicting cognitive decline and dementia in elderly twins from indicators of early life oral health

(USC Thesis Other)

Predicting cognitive decline and dementia in elderly twins from indicators of early life oral health

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content PREDICTING COGNITIVE DECLINE AND DEMENTIA IN ELDERLY TWINS FROM INDICATORS OF EARLY LIFE ORAL HEALTH by Amber Watts Hall ________________________________________________________ A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (GERONTOLOGY) May 2009 Copyright 2009 Amber Watts Hall ii DEDICATION To my husband for his patience and support, the many friends who have shared this experience with me, my family, and to the beloved older adults in my life who have given me perspective. iii ACKNOWLEDGMENTS I would like to thank Margy Gatz for her tireless commitment to my professional development and to this project. She has provided me with an excellent example of how to balance mentorship and research. I offer sincere thanks to my committee members for their contributions to making this work its best. In particular, I thank Eileen Crimmins and Carol Prescott for helping to get the pilot grant that funded this project, for providing valuable feedback, and for demonstrating great interdisciplinary collaborations in their work. I am greatly appreciative to Tuck Finch for challenging me to deepen my understanding of this topic via many drafts of the theoretical manuscript he read and critiqued. Thanks to Brenda Plassman at Duke University for sharing her data and her expertise on many occasions along the way. Thanks to Bill Page at the Institute of Medicine for making these data available. I would also like to thank Jorgen Slots for comments on an earlier draft of the theoretical model as well as for letting me sit in on his oral microbiology class. I would like to thank Bernard Steinman for his help in rendering Figure 2.1 as well as for being a good friend. I want to acknowledge my lab mates and fellow graduate students for helping me to develop ideas, offering support, and positive feedback throughout each stage of this process. A paper closely resembling parts of chapter I and chapter II has been published in Neuropsychiatric Disease and Treatment, 4, 865-876. It is presented here with permission from Dove Press. Finally, I acknowledge grant support from NIA grant T32-AG00037 & NIA pilot grant P30 AG-017265. iv TABLE OF CONTENTS Dedication ii Acknowledgements iii List of Tables v List of Figures ix Abstract x Chapter I: Introduction 1 Chapter II: Proposed Theoretical Model 9 Chapter III: Research Design and Method 37 Chapter IV: Results for Hypothesis 1 58 Chapter V: Results for Hypothesis 2 77 Chapter VI: Discussion & Conclusion 94 Bibliography 104 Appendices Appendix A: Standards of Dental Treatment During the World War II Era Appendix B: Resolution of Twin Disagreement in the Report of Father’s Occupation 125 126 v LIST OF TABLES Table 1.1: Comparison of Characteristics of Previous Studies to Present Study 6 Table 2.1: Inflammatory Markers Associated with Cognitive Decline and Dementia in Epidemiological Studies 10 Table 3.1: Waves of Telephone Cognitive Screening in Duke Twins Study 39 Table 3.2: Selection Criteria for the Present Sample 42 Table 3.3: Descriptive Statistics for TICS-m and Dementia Diagnosis 47 Table 3.4: Description of Measures 48 Table 3.5: Summary Statistics for Primary Predictor Variables 54 Table 3.6: Correlations between Continuous Predictor Variables 55 Table 4.1: Group Means & Frequencies for Risk Factors in Case-Control Sample 59 Table 4.2: Numbers of Cases and Controls Exposed or Unexposed to Risk Factors 59 Table 4.3: Case Control Analyses Predicting Risk for Total Dementia (CONTINUOUS) 61 Table 4.4: Case Control Analyses Predicting Risk for Total Dementia (CATEGORICAL) 62 Table 4.5: Case Control Analyses Predicting Risk for AD (CONTINUOUS) 63 Table 4.6: Case Control Analyses Predicting Risk for AD (CATEGORICAL) 64 vi Table 4.7: Group Means for Risk Factors for Co-twin Sample 65 Table 4.8: Numbers of Pairs Exposed or Unexposed to Risk Factors within Dementia-concordant and Dementia-discordant Twin Pairs 67 Table 4.9: Comparative Risk Variables Occurring in Proband versus Control Twin 67 Table 4.10: Co-twin Control Analyses Predicting Total Dementia (CONTINUOUS) 70 Table 4.11: Co-twin Control Analyses Predicting Total Dementia (CATEGORICAL) 70 Table 4.12: Co-twin Control Analyses Predicting Total Dementia (COMPARATIVE RISK) 70 Table 4.13: Co-twin Control Analyses Predicting AD (CONTINUOUS) 71 Table 4.14: Co-twin Control Analyses Predicting AD (CATEGORICAL) 71 Table 4.15: Co-twin Control Analyses Predicting AD (COMPARATIVE RISK) 71 Table 4.16: Group Means & Frequencies for Risk Factors in Case-Control Sample in MZ Twins Only 73 Table 4.17: Numbers of Cases and Controls Exposed or Unexposed to Risk Factors in MZ Twins Only 73 Table 4.18: Group Means for Risk Factors for Co-Twin Control Sample in MZ Twins Only 74 Table 4.19: Numbers of Pairs Exposed or Unexposed to Risk Factors within Dementia-concordant and Dementia-discordant Twin Pairs in MZ Twins Only 74 Table 4.20: Comparative Risk Variables Occurring in Proband versus Control Twin in MZ Twins Only 75 vii Table 4.21: Co-twin Control Analyses Predicting Total Dementia in Discordant Pairs Only (CONTINUOUS) 75 Table 4.22: Co-twin Control Analyses Predicting Total Dementia in Discordant Pairs Only (CATEGORICAL) 75 Table 4.23: Co-twin Control Analyses Predicting Total Dementia in Discordant Pairs Only (COMPARATIVE RISK) 75 Table 4.24: Co-twin Control Analyses Predicting AD in Discordant Pairs Only (CONTINUOUS) 76 Table 4.25: Co-twin Control Analyses Predicting AD in Discordant Pairs Only (CATEGORICAL) 76 Table 4.26: Co-twin Control Analyses Predicting AD in Discordant Pairs Only (COMPARATIVE RISK) 76 Table 5.1: Mean TICS-m Scores and Ages at Each Wave of Measurement 77 Table 5.2: Paired-Samples T-Tests for Characteristics of Twin Pairs Discordant for Tooth Loss 79 Table 5.3: Correlations between Predictor Variables and TICS-m Scores 80 Table 5.4: Mixed Effects Model Results Predicting Baseline TICS-m Score in Non-demented Twins 82 Table 5.5: Mixed Effects Model Results Predicting Baseline TICS-m Score in Entire Sample 82 Table 5.6: Waves of TICS-m Data Selected within Twin Pairs among Non-demented 83 Table 5.7: Mixed Effect Model Results Predicting Change in TICS-m 84 Table 5.8: Mixed Effect Model Results Predicting Change in TICS-m from wave 1 to wave 2 in Non-demented Pairs 85 Table 5.9: Mixed Effect Model Results Predicting Change in TICS-m from wave 2 to wave 3 in Non-demented Pairs 85 viii Table 5.10: Mixed Effect Model Results Predicting Change in TICS-m from wave 3 to wave 4 in Non-demented Pairs 86 Table 5.11: Mixed Effect Model Results Predicting Change in TICS-m from wave 1 to wave 2 in Entire Sample 86 Table 5.12: Mixed Effect Model Results Predicting Change in TICS-m from wave 2 to wave 3 in Entire sample 87 Table 5.13: Mixed Effect Model Results Predicting Change in TICS-m from wave 3 to wave 4 in Entire sample 87 Table 5.14: Comparison of Growth Curve Model Fits over Age Bins 89 Table 5.15: Predictors of the Quadratic Growth Curve Model over Age Bins 92 Table 5.16 Comparison of Growth Curve Model Fits over Waves 93 Table 6.1 Rates of Tooth Loss, Attributable Risk, and Excess Fraction in Studies of Tooth Loss and Dementia 98 ix LIST OF FIGURES Figure 2.1: Oral Bacteria Enter the Bloodstream 17 Figure 2.2: Proposed Pathways between Periodontal Infection and Alzheimer’s Disease Pathology 25 Figure 3.1: Attrition from the Initial NAS-NRC Sample to Completion of First Wave of Duke Dementia Screening 38 Figure 3.2: Flowchart of Waves of Duke Dementia Screening 40 Figure 3.3: Frequency Distribution of the Continuous Number of Missing Teeth at Time of Entry into the Military 50 Figure 3.4 Percent Demented by Years of Education 55 Figure 3.5 Percent Demented by Father’s Occupational Status 56 Figure 3.6 Percent Demented by Height in Inches 56 Figure 3.7 Percent Demented by Number of Missing Teeth 57 Figure 5.1: Mean TICS-m Scores over Occasions of Measurement in the Entire Sample 78 Figure 5.2 Mean TICS-m Scores over Age in the Entire Sample 78 Figure 5.3 Linear and Quadratic Change in TICS-m Scores over Age 90 Figure 5.4 Piecewise Spline Model of TICS-m Scores over Occasion of Measurement 93 x ABSTRACT The purpose of this dissertation was to explore the relationship between tooth loss in young adulthood and dementia and cognitive decline in older adulthood in a sample of elderly twins. The study had two specific aims. The first aim was to determine whether tooth loss is associated with risk of developing dementia after accounting for low SES and other possible covariates. The second aim was to determine whether tooth loss is associated with baseline cognitive performance and cognitive decline on the modified Telephone Interview for Cognitive Status (TICS-m) after accounting for low SES and other possible covariates. A theoretical model is advanced proposing several possible mechanisms by which oral health may relate to dementia and cognitive decline. The literature review discusses how infectious pathogens and systemic inflammation may play a role in AD with regard to known pathologies including senile plaques, neuron death, neurofibrillary tangles, and cerebrovascular changes. A model of proposed pathways between periodontal infection and AD is presented including three possible mechanisms: a) direct effects of infectious pathogens, b) inflammatory response to pathogens, and c) the effects on vascular integrity. The role of gene polymorphisms is discussed. The study was conducted using military induction data from members of the NAS-NRC Registry of World War II Veterans who are members of the Duke Twins Study of Memory in Aging (N = 6,352). In case-control analyses, tooth loss in young adulthood did not significantly predict dementia. Tooth loss did predict xi poor baseline cognitive scores, although the effect was explained by the relationship between tooth loss and low educational attainment. Due to a low rate of tooth loss, the power to detect a relationship between tooth loss and dementia was .20 to .40 in our sample. Thus, results do not rule out a relationship between tooth loss and dementia in the population. Co-twin control analysis techniques were planned, although the low power effectively precluded meaningful co-twin control analyses. Consistent with previous research, older age and lower levels of educational attainment were associated with increased risk of dementia and poorer baseline cognitive scores. Additionally, low education predicted earlier age of dementia onset. 1 CHAPTER I: INTRODUCTION More than 4.5 million people in the U.S. currently suffer from Alzheimer’s disease (AD) with dramatic increases in prevalence expected in the upcoming years (Hebert, Scherr, Bienias, Bennett, & Evans, 2003). Some of the primary risk factors for dementia are not modifiable. These include advancing age and possession of particular genetic alleles. While knowledge of non-modifiable risk factors may help identify individuals at risk for dementia, knowledge of modifiable risk factors may help prevent or delay its development. Modifiable risk factors may include diet and exercise behaviors (Colcombe & Kramer, 2003; Kang, Ascherio, & Grodstein, 2005; Larson, Wang, Bowen, McCormick, Teri, Crane et al., 2006), educational attainment (Gatz, Svedberg, Pedersen, Mortimer, Berg, & Johansson, 2001), and avoidance of vascular disease (Honig, Kukull, & Mayeux, 2005; Luchsinger, Reitz, Tang, Shea, & Mayeaux, 2005). A few recent studies have contributed the intriguing finding that adult tooth loss increases the risk of dementia. Though the reasons for this relationship have not been fully explained, oral health may be a potentially modifiable risk factor for dementia, in particular for AD. Tooth Loss as a Risk Factor for Dementia In Swedish twins, Gatz, Mortimer, Fratiglioni, Johansson, Berg, Reynolds et al. (2006) found that those who had lost half or more of their teeth before age 35 were at 1.7 fold greater risk for AD after controlling for socioeconomic status (SES) and other factors. Within a twin pair, the demented twin was four times 2 more likely to have had worse oral health before age 35. Similarly, a case control study of AD in Japan found that loss of more than half of adult teeth by age 50 to 60 increased AD risk by 2.6 fold (Kondo, Niino, & Shido, 1994). In this study, periodontal disease (PDD) occurring 20-30 years prior to dementia onset was the most frequent cause of tooth loss. In a study of 144 nuns, having few or no teeth predicted a 4.3 fold higher risk of dementia, but only among those who did not carry the apolipoprotein (APOE) ε4 allele (Stein, Desrosiers, Donegan, Yepes, & Kryscio, 2007). The study included dental records that indicated whether teeth were lost due to PDD or other causes. In a Korean community population, persons with fewer teeth were at 1.6 fold greater risk of dementia in subsequent years if they did not have their missing teeth replaced by dentures (Kim, Stewart, Prince, Kim, Yang, Shin et al., 2007). In each of these studies, the measurement of tooth loss occurred prior to the onset of dementia, not subsequently, implicating oral health as a potential risk factor for dementia rather than a consequence of it. Possible Explanations for the Relationship between Tooth Loss and Dementia Despite evidence that oral disease may be a risk factor for dementia, little attention has been devoted to explaining the potential mechanisms for this association (see Kramer, Craig, Dasanayake, Brys, Glodzik-Sobanska, & deLeon, 2008; Stein, Scheff, & Dawson, 2006). It is unknown whether the two have a common cause or whether tooth loss may relate to processes that contribute to the development of dementia. Some of the causes of tooth loss are also potential risk factors for dementia. The most common causes of tooth loss, 3 periodontal disease and dental caries, are the result of chronic oral bacterial infection. Head injury could result in tooth loss as well as damage to the brain, though evidence that head injury is a risk factor for dementia is mixed (Jellinger, 2004). Infection and immune response to infection, especially inflammation, have been suspected to play a role in Alzheimer’s disease (e.g. Marx, Blasko, Pavelka, & Grubeck-Loebenstein, 1998; Perry, 2004). Infection and inflammation are also suspected risk factors for vascular disease (Mehta, Saldeen, & Rand, 1998; Blake & Ridker, 2002), and vascular disease may play a role in dementia pathology (Luchsinger et al., 2005; Honig, Kukull, & Mayeaux, 2005). Though we were not able to measure inflammation directly in the current study, it may be an underlying process. Inflammation is essential to immune response to infection and serves to protect and repair the body. In excess or chronic levels, however, inflammatory responses can become harmful. The possible role of inflammation in mediating the relationship between periodontal disease and dementia is discussed in detail in Chapter 2. The APOE ε4 allele, a known risk factor for dementia and vascular disease (Strittmatter, Saunders, Schmechel, Pericak-Vance, Enghild, Salvesen et al., 1993; Greenow, Pearce, & Ramji, 2005), was recently found to be more prevalent in edentulous older adults compared with matched controls who still possessed their teeth (Bergdahl, Bergdahl, Nyberg, & Nilsson, 2008). This suggests genetic factors could play a role in increasing risk for both tooth loss and dementia. 4 Early life socioeconomic conditions predict cognitive function and risk of Alzheimer’s disease and correlate with early life health conditions such as low birth weight, poor postnatal nutrition, and short adult height (Barker, Osmond, & Golding, 1990; Borenstein, Copenhaver, & Mortimer, 2006; Floud, Wachter, & Gregory, 1990). Periodontal disease is both more frequent and more severe among individuals with low SES (Borrell, Beck, & Heiss, 2006; Cunha-Cruz, Hujoel, & Nadanovsky, 2007). Therefore, it is necessary to account for SES to determine whether any relationship between tooth loss and later dementia is independent of the other disadvantages bestowed by poor SES. Purpose of the Present Research The purpose of this dissertation is to explore the relationship between tooth loss in young adulthood and dementia and cognitive decline in older adulthood in a sample of elderly twins. The findings may build on literature suggesting that exposure to risk factors in early life, particularly those related to infection, may be important predictors of mortality and dementia (Finch & Crimmins, 2004; Borenstein et al., 2006). It also proposes a detailed theoretical model of the roles of infection and inflammation in dementia and their possible relationship to tooth loss. The National Academy of Sciences-National Research Council (NAS- NRC) Twin Registry of World War II veterans was used to investigate the two aims of the study. The first aim was to determine the relationship between early adult tooth loss and later diagnosis of dementia accounting for low early life SES and other potential confounding variables. We hypothesized that greater tooth 5 loss would statistically increase the risk of dementia after accounting for these variables. The second aim was to determine whether tooth loss was associated with performance on tests of cognitive status and cognitive decline in demented and non-demented twins. We hypothesized that greater tooth loss would predict lower baseline scores and greater declines in cognitive performance after accounting for confounding variables. Contribution to the Literature Successful strategies for the prevention and treatment of dementia have been elusive. Exploration of the relationship between tooth loss and dementia may provide insight into modifiable risk factors for dementia as well as providing hints about the mechanisms leading to dementia pathology. Four known studies have previously investigated and reported a relationship between numbers of teeth lost and diagnosis of dementia and one found this relationship with poor mental status (See Table 1.1). The present study offers several advantages compared to previous studies. Only one of the previous studies used co-twin control allowing control for genetic and shared environmental influences. Four of the studies used a selected sample of individuals, for example Stein et al. (2007) studied nuns who were highly educated and all resided in Milwaukee. Whereas in the present study we use a population based sample offering greater generalizability including both AD and non-AD dementia cases. Another improvement of the present study over previous studies is the assessment of tooth loss. Two of the previous studies assessed tooth loss by 6 Table 1.1 Comparison of Characteristics of Previous Studies to Present Study Co-twin Population Objective Design Based Teeth Measure Present Study Yes Yes teeth observed Swedish Twins Yes Yes self report (Gatz et al., 2006) Milwaukee Nuns No No teeth observed (Stein et al., 2007) Korean Community No No 1 teeth observed (Kim et al., 2007) Japanese Study No No self report (Kondo et al., 1994) Japanese No No teeth observed Institutionalized (Shimazaki et al. 2001) Age when Dementia diagnosis & Teeth data collected mental status (age of tooth loss if retrospective) Present Study 16 - 34; dementia diagnosis & decline in mental status Swedish Twins 37 - 63 dementia diagnosis (Gatz et al., 2006) (before age 35) Milwaukee Nuns unknown 2 dementia diagnosis (Stein et al., 2007) Korean Community 65 - 80+ dementia diagnosis (Kim et al., 2007) Japanese Study 43 - 89 dementia diagnosis (Kondo et al., 1994) (50 - 69 years old) Japanese Institutionalized 59 - 107 mental status categorical 3 (Shimazaki et al. 2001) 1 Sample was not population based, but was likely to be representative 2 Study used dental records from life as a nun beginning in 1964, but ages were not standardized or specified in the report 3 Measures of mental status not specified in the report 7 self report, whereas in the veteran twins tooth loss was assessed during a medical exam at time of entry into the military. This assessment eliminates potential misreporting or recall bias in other studies in which many of the respondents recalled how many teeth were lost at a time prior to when the information was being collected. There is very little range of participants’ ages when tooth loss was measured in the veteran twins. In contrast, the Swedish twin study asked individuals aged 37 to 63 to recall their tooth loss at age 35. In the Japanese study, cases and controls ranging in age from 43 to 89 were asked about their dental history in the fifth and sixth decades of life. If tooth loss is measured at too old an age, the loss may be the consequence rather than the cause of any observed cognitive impairment. The contemporaneous direct assessment of tooth loss at an age prior to the possibility of dementia allows for better assessment of the hypothesis that poor oral health in early life is related to later risk of dementia. Poor oral health later in life is likely to increase risk of dementia through different mechanisms such as age-related declines in immune function or reductions in nutrition. The present study will also use tooth loss to predict baseline performance and changes in cognitive function among demented and non-demented individuals. This may allow detection of an effect of oral health on cognitive change prior to the onset of detectable dementia symptoms or in persons without dementia. While one study (Shimazaki et al., 2001) used dental health to predict poor mental status, no known studies have looked at both dementia diagnosis and cognitive performance over several occasions. 8 The present research used a twin design, which has the advantage of controlling for familial effects, i.e., genetic and early life experiences shared between the twins. To the extent that dementia is the product of multiple influences, including genetic (Gatz, Reynolds, Fratiglioni, Johansson, Mortimer, Berg et al., 2006), the twin design helps to isolate effects that may be specific to non-genetic exposures on which one member of a twin pair differs from the co- twin. Another advantage of the study is that information about tooth loss was collected prospectively 60 years ago which eliminates recall bias and problems of reverse causation. Organization of the Dissertation This dissertation will be presented in six chapters. The first is an introduction to the topic of oral health as a risk factor for dementia. The second chapter proposes a theoretical model describing the potential mechanisms by which oral health might influence dementia. A detailed description of the study design, method, and data used in the dissertation are given in chapter three. Though the data available do not allow testing all the pathways proposed in the theoretical model, chapters four and five contain results of data analyses for the first and second study aims respectively. Finally, chapter six provides a comprehensive discussion and conclusion for the entire project. 9 CHAPTER II: PROPOSED THEORETICAL MODEL Periodontal disease (PDD) is associated with increased risk of vascular disease and mortality in more than 200 reports and reviews (e.g., Arbes, Slade, & Beck, 1999; Beck, Garcia, Heiss, Vokonas, & Offenbacher, 1996; DeStefano, Anda, Kahn, Williamson, & Russell, 1993; Garcia, Krall, & Vokonas, 1998; Grau, Becher, Ziegler, Lichy, Buggle, Kaiser, et al., 2004; Loesche, Schork, Terpenning, Chen, Dominguez, & Grossman, 1998; Loesche, Schork, Terpenning, Chen, Kerr, & Dominguez, 1998; Wu, Trevisan, Genco, Dorn, Falkner, & Sempos, 2000). Far fewer studies have examined the relationship between oral health earlier in life and AD in late life (Gatz, Mortimer, Fratiglioni, Johansson, Berg, Reynolds, et al., 2006; Kim, Stewart, Prince, Kim, Yang, Shin, et al., 2007; Kondo, Niino, & Shido, 1994; Stein, Desrosiers, Donegan, Yepes, & Kryscio, 2007). Inflammation is recognized as a core process in atherosclerosis and cardiovascular disease (CVD), and may also have a major role in AD. Chronic inflammation, as measured by blood inflammatory markers (Table 2.1), is associated with increased risk for cognitive decline (Alley, Crimmins, Karlamangla, Hu, & Seeman, 2008; Weaver, Huang, Albert, Harris, Rowe, & Seeman, 2002; Yaffe, Lindquist, Penninx, Simonsick, Pahor, Kritchevsky, et al., 2003) and dementia (Engelhart, Geerlings, Meijer, Kiliaan, Ruitenberg, vanSwieten, et al., 2004; Schmidt, Schmidt, Curb, Masaki, White, & Launer, 2002; Tan, Beiser, Vasan, Roubenoff, Dinarello, Harris, et al., 2007). PDD is a 10 common source of chronic systemic infection. This review suggests mechanisms by which periodontal infection may promote AD. The proposed model gives a rationale for further experimental and clinical studies. Table 2.1 Inflammatory Markers Associated with Cognitive Decline and Dementia in Epidemiological Studies Cytokines Associated With Interleukin 1 (IL-1) Dementia (Framingham Study 1 ) Interleukin 6 (IL-6) Cognitive Decline (MacArthur Studies 2 ; Health ABC Study 3 ) Dementia (Rotterdam Study 4 ) Interleukin 10 (IL-10) Cognitive Decline (Leiden 85+ Study 5 ) Tumor necrosis factor Cognitive Decline (Leiden 85+ Study 5 ) alpha (TNF-α) Dementia (Framingham Study 1 ) Acute Phase Proteins C-reactive protein (CRP) Cognitive Decline (Health ABC Study 3 ; Greek Community 6 ) Dementia (Rotterdam Study 4 ; Honolulu-Asia Aging Study 7 ) α- 1 antichymotrypsin (ACT) Cognitive Decline (Longitudinal Aging Study Amsterdam 8 ) Dementia (Rotterdam Study 4 ) Cell Adhesion Molecules Intercellular adhesion Cognitive Decline (Greek Community 6 ) molecule (ICAM-1) Vascular cell adhesion Cognitive Decline (Greek Community 6 ) molecule (VCAM-1) 1 Tan et al., 2007; 2 Weaver et al., 2002; 3 Yaffe et al., 2003; 4 Engelhart et al., 2004; 5 van Exel et al., 2003; 6 Dimopoulos et al., 2006; 7 Schmidt et al., 2002; 8 Dik et al., 2005 11 Research has documented declines in self-care and oral hygiene that occur in people with dementia (e.g., Chalmers & Pearson, 2005; Meurman & Hamalainen, 2006). Our focus, however, is periodontal disease in early and midlife when it is clear that oral disease is an antecedent to dementia and when there would be no reason to suspect that dementia could have led to the oral disease. Periodontal Disease PDD is a group of conditions that cause inflammation and destruction of the gums, alveolar bone, and other structures that support the teeth. The etiology is complex involving the presence of pathogenic bacteria found in dental plaque and individual variation in host immune response. PDD is a common source of chronic systemic infection in humans (Garcia, Henshaw, & Krall, 2000; Li, Kollveit, Tronstad, & Olsen, 2000; Taylor, Tofler, Carey, Morel-Kopp, Philcox, Carter, et al., 2006). The bacterial pathogens most strongly implicated in chronic periodontal disease are porphyromonas gingivalis, tannerella forsythensis, treponema denticola, and actinobacillus actinomycetemcomitans. Elevated levels of interleukin 1 (IL-1) have been found in gingival crevicular fluid (GCF) of patients with experimentally induced gingivitis and active periodontal disease (Kinane, Winstanley, Adonogianaki, & Moughal, 1992; Masada, Persson, Kenney, Lee, Page, & Allison, 1990). In patients with advanced periodontitis, substantial reduction of IL-1 levels occurred after treatment (Masada et al., 1990). The simultaneous presence of p. gingivalis and t. forsythensis has been associated with increased GCF levels of inflammatory 12 mediators and associated with more severe disease (Airila-Mansson, Soder, Kari, & Meurman, 2006). In gingivitis, inflammatory mediators in the GCF do not penetrate deeply into tissues. Gingivitis advances to periodontitis when bacteria evade clearance by neutrophils and penetrate the deeper tissues (Offenbacher, 1996). Even low-grade infections in the oral cavity may be associated with moderate, sub-clinical systemic inflammatory response indicated by blood elevations of C-reactive protein (CRP) and interleukin 6 (IL-6) (D’Aiuto & Tonetti 2004; Taylor et al., 2006). Severe PDD can induce chronic inflammation and immune reactions that result in loss of bone and soft tissue that supports teeth in the jaws. Systemic inflammatory markers are commonly elevated in individuals with PDD. In a recent review (Loos, 2005), 8 studies showed that blood leukocytes and plasma levels of CRP were consistently higher in patients with periodontitis compared to healthy controls. In one study, carriers of common oral anaerobic bacteria had higher plasma levels of CRP (Bretz, Corby, Schork, Robinson, Coelho, Costa, et al., 2005). Furthermore, the severity of periodontal infection has been correlated with serum levels of inflammatory markers. For example, in the Atherosclerosis Risk in Communities study (ARIC), older adults with more extensive periodontal pockets had one-third higher plasma CRP than those with mild PDD (Slade, Ghezzi, Heiss, Beck, Riche, & Offenbacher, 2003). Other inflammatory proteins such as IL-6 and tumor necrosis factor alpha (TNF- α) have been found to be elevated in advanced PDD. For example, individuals 13 with extensive PDD had 2- to 4-fold higher mean plasma levels of IL-6 and TNF- α than those with mild or no disease (Bretz et al., 2005). The success of reducing CRP and other inflammatory markers with periodontal treatment has been mixed. Intervention studies of moderate to severe PDD use standard therapeutic scraping of dental calculus (scaling and root planing) alone or in combination with anti-inflammatory or anti-infective drugs. Scaling and root planing alone reduced CRP levels in one study (D’Aiuto & Tonetti 2004) but did not reduce CRP, interleukins, or TNF-α in two other studies (Fokkema, Loos, deSlegte, Burger, Piscaer, Ijzerman, et al., 2003; Ide, McPartlin, Coward, Crook, Lumb, & Wilson, 2003). Use of anti-inflammatory drugs in combination with scaling and root planing reduced CRP and haptoglobin in one study (Ebersole, Machen, Steffen, & Willmann, 1997) and in another study anti-infective drugs in combination with scaling and root planing reduced CRP, especially among those with elevated baseline CRP (Mattila, Vesanen, Valtonen, Nieminen, Palosuo, Rasi, et al., 2002). Extraction of all teeth in advanced PDD patients caused a reduction in blood CRP, fibrinogen, white cells, platelets, and plasminogen activator inhibitor-1 in another study (Taylor et al., 2006). Though the most effective treatment for reduction of inflammation in PDD remains unclear, treatment of periodontal infection may help to reduce systemic inflammation. PDD has been associated with increased risk of mortality, cardiovascular disease, and stroke in a large number of studies. For example, in the National Health and Nutrition Examination Study (NHANES), a large U.S. national sample, 14 adults with periodontitis had a 46% increased risk of independent all-cause mortality and a 25% increased risk of coronary heart disease (DeStefano et al., 1993). Other studies have shown that individuals with markers of poor oral health are at 2 to 4 times greater risk for stroke (Beck et al., 1996; Grau et al., 2004; Loesche, Schork, Terpenning, Chen, Kerr, et al., 1998; Wu et al., 2000). Severity of PDD has shown a dose-response relationship with disease outcomes and mortality (Arbes et al., 1999; Loesche, Schork, Terpenning, Chen, Dominguez, et al., 1998). A longitudinal dental study of veterans found that individuals with the deepest probing pocket depths and the greatest degree of alveolar bone loss had the highest independent mortality risk (Garcia et al., 1998). Though not all studies have supported PDD as a risk for CVD and cerebrovascular disease, most reports support a modest relationship (see Scannapieco, Bush, & Paju, 2003 for a systematic review). PDD may also be associated with increased risk for AD. In Swedish twins, Gatz et al. (2006) found that those who had lost half or more of their teeth before age 35 had a 1.7 fold greater risk for AD, after controlling for other factors. Within a twin pair, the demented twin was 4-times more likely to have had worse oral health before age 35. Similarly, a case control study of AD in Japan found that loss of more than half of adult teeth by age 50-60 increased AD risk by 2.6 fold (Kondo et al., 1994). In this study, PDD occurring 20-30 years prior to dementia onset was the most frequent cause of tooth loss. In a study of 144 nuns, having few or no teeth predicted a 4.3 fold higher risk of dementia, but only among those who did not carry the apolipoprotein (APOE) ε4 allele (Stein et al., 2007). The 15 study included dental records that indicated whether teeth were lost due to PDD or other causes. In a Korean community population, persons with fewer teeth were at 1.6 fold greater risk of dementia in subsequent years if they did not have their missing teeth replaced by dentures (Kim et al., 2007). Teeth are commonly lost as a result of periodontal infection and dental caries, both of which are caused by exposure to bacteria. Less commonly, teeth are lost due to trauma. Studies that measure markers of periodontal infection, not just tooth loss, will be important to establishing whether oral health may be related to AD. Systemic Infection: From Mouth to Bloodstream The mouth is a primary channel by which external organisms enter the body. Transient bacteremia occurs after tooth brushing and flossing as well as after normal dental procedures (Forner, Larsen, Kilian, & Holmstrup, 2006; Loos, 2005). In individuals with good oral and immune health, the transient bacteremia has few consequences. It is suspected that individuals with periodontal infection have higher levels of pathogenic accumulation and may experience transient bacteremia multiple times per day (Forner et al., 2006; Li et al., 2000; Loos, 2005). Gram positive and Gram negative oral bacteria contain several components within their membranes that can induce pro-inflammatory cytokines including IL-1, IL-6, and TNF-α. Lipopolysaccharide (LPS) endotoxin is the most often studied of these bacterial components (Beck et al., 1996; Wilson, Reddi, & Henderson, 1996). The oral cavity has several barriers, physical, electrical, and chemical, that inhibit penetration by pathogens. First, the surface epithelium tissue layer 16 provides a physical barrier composed of tight junctions and a chemical barrier containing peptide antibiotics called defensins. An electrical barrier affects the flow of electrons between the oral cavity and microbes that are introduced. The high reduction potential in the oral cavity increases oxidation of local bacteria. Another layer of protection is an immunological layer of antibody-forming cells. Finally, the reticuloendothelial system protects through phagocytosis to engulf and destroy bacteria upon entry into the blood (Loesche & Lopatin 2000; Weinberg, Krisanaprakornkit, & Dale, 1998). If these barriers are compromised by PDD, trauma, or immune suppression, microbes can disseminate to cause acute or chronic infection (Li et al., 2000). Once in the bloodstream, these bacteria can induce acute phase response characterized by increased white blood cell counts and the release of inflammatory cytokines. Supragingival plaque, above the gums, contains dense layers of Gram positive bacteria (see Figure 2.1). If these bacteria penetrate the epithelium, they may survive the oral cavity barriers and enter the bloodstream (Loesche & Lopatin, 2000). Though not specifically associated with PDD, the number of these bacteria increase in conditions of poor oral hygiene and may leak into the blood and result in the inflammation of distant systems. Subgingival plaque, below the gums, contains primarily Gram negative organisms in individuals with disease (Loesche & Lopatin, 2000). These get trapped in the space between the tooth and the gum tissue and provoke local inflammatory responses giving rise to PDD. PDD is characterized by chronic Gram-negative infection in which bacteria may enter the bloodstream via the 17 epithelium. Ulcerations in the gingival epithelium allow bacteria to spread from the pocket between the tooth and gum into the capillaries of the epithelium and thus into systemic circulation. Bacterial endotoxin (LPS contained in bacterial cell membranes) may enter the bloodstream. In PDD, the large surface area of the pockets allows a larger degree of LPS endotoxin to enter compared to the intact Figure 2.1. In periodontal disease, the gingival recedes from the tooth forming pockets through which bacteria may more easily enter the bloodstream. 18 endothelial lining and more limited pockets in individuals without disease. LPS induces cytokine production and its entry may alter blood coagulation and promote atherosclerosis and thrombogenesis (Stoll, Denning, & Weintraub, 2004; Valtonen, 1991). Cytokines produced locally, for example in the periodontium, are generally degraded locally. However, under conditions of repeated challenge, cytokine receptors may become saturated and less able to eliminate cytokines, thus allowing them to “spill over” into systemic circulation affecting serum levels of cytokines and acute phase reactants (Offenbacher, 1996). Chronic periodontal infection continuously attracts circulating leukocytes to inflamed periodontal tissues keeping immune cells activated (Fokkema et al., 2003). Bacteria are not the only pathogens that are likely to enter the system when defenses are vulnerable. Viruses are frequently detected in periodontal pockets (Kubar, Saygun, Ozdemir, Yapar, & Slots, 2005). In one study of 30 patients with advanced periodontitis, 78% had at least one of five viruses and 40% were co-infected with two to five viruses (Parra & Slots, 1996), particularly herpes viruses including cytomegalovirus (CMV), Epstein-Barr virus (EBV), and herpes simplex virus (HSV) found in 60%, 30%, and 20% of patients respectively. Herpes viruses also infect inflammatory cells in periodontitis (Contreras, Zadeh, Nowzari, & Slots, 1999). The potential role of bacterial and viral pathogens in AD pathogenesis is discussed in more detail below. Transmigration from Bloodstream to Brain Though the blood-brain barrier (BBB) generally prevents entry of substances into the brain, there is evidence that inflammatory cytokines can 19 enter or influence the brain under certain circumstances. For example, IL-6 and certain other blood cytokines elevated during the acute phase response enter the brain to initiate sickness behaviors including fever, malaise, reduced appetite, and decreased social contact (Perry, 2004). These behaviors are a normal response that is adaptive to conserve energy stores when fighting infection and possibly protect others from the spread of disease. IL-6 and other cytokines may be transported across the intact BBB through specific transport processes (Banks, Farr, & Morley, 2002; Pan & Kastin, 1999). IL-6 and other cytokines may also enter through fenestrated capillaries in circumventricular organs (including the pineal gland, vascular organ of the lamina terminalis, area postrem, and subfornical organ) near the base of the brain that are outside the BBB and are more easily penetrated (Banks et al., 2002; Perry, 2004). Infectious Pathogens and AD Exposure to infectious pathogens of various types is a possible risk factor for AD, although mixed results allow only tentative conclusions (Balin & Appelt, 2001; Holmes, El-Okl, Williams, Cunningham, Wilcockson, & Perry, 2003; Itzhaki, Wozniak, Appelt, & Balin, 2004; Kinoshita, 2004; Little, Hammond, MacIntyre, Balin, & Appelt, 2004; Mattson, 2004; Ringheim & Conant, 2004; Robinson, Dobson, & Lyons, 2004; Wozniak, Shipley, Combrinck, Wilcock, & Itzhaki, 2005). Research has focused primarily on Chlamydia pneumoniae, a bacterium, and herpes viruses including HSV and CMV. Very few studies have reported investigating the presence oral disease-related bacteria in AD (e.g., Riviere, Riviere, & Smith, 2002). 20 C. pneumoniae infection is characterized by chronic inflammation and is generally acquired orally or nasally. Although some studies found C. pneumoniae in post-mortem AD brains (Balin & Appelt, 2001), other labs failed to replicate these results (Gieffers, Reusche, Solbach, & Maass, 2000; Nochlin, Shaw, Campbell, & Kuo, 1999; Ring & Lyons, 2000; Wozniak, Cookson, Wilcock, & Itzhaki, 2003). C. pneumoniae antigens have been located in microglia and astroglia cells in brain regions with AD neuropathology (Itzhaki et al., 2004). An in vitro cell model of the BBB showed the possibility that C. pneumoniae could enter the brain (MacIntyre, Abramov, Hammond, Hudson, Arking, Little, et al., 2003). In other studies that have not been replicated, infective C. pneumoniae was obtained postmortem from an AD brain and induced amyloid plaques in a mouse model (Little et al., 2004). Treponema bacteria, a family of Gram negative spirochetes commonly associated with PDD, was found in the brains of AD patients with greater frequency than in non-AD controls (Riviere et al., 2002). The AD patients also had a greater variety of treponema species. The levels of bacteria measured in the blood did not differ between the AD patients and controls suggesting that AD subjects may be more susceptible to infection in the central nervous system (CNS) than controls. Antigens for two types of treponema were found in the trigeminal ganglia, pons, and hippocampus possibly indicating that the bacteria reached the brain via the trigeminal nerve. Other spirochetes have been found in the brain tissue, blood, and cerebrospinal fluid of AD patients (Miklossy, 1993), however the finding was not confirmed in another study (McLaughlin, Kin, Chen, 21 Nair, & Chan, 1999) and the link of spirochete presence to AD has been treated with skepticism (Hammond, Gage, & Terry, 1993). Viral infection may also be a risk factor for AD, particularly herpes viruses. Non-demented older adults with high serum levels of antibodies for CMV had faster rates of cognitive decline over a four year period than those with low levels of CMV antibodies (Aiello, Haan, Blythe, Moore, Gonzales, & Jagust, 2006). HSV-1 is found in a high proportion of non-AD elderly brains (Jamieson, Maitland, Wilcock, Yates, & Itzhaki, 1992) and is most often not found to be associated with dementia. This does not preclude the role of HSV in AD, because presence of a virus is not sufficient to result in disease. Individual factors such as immune function, pathogen virulence, and genetic factors determine the outcome of exposure to a virus. Latent viruses may be reactivated by immunosuppression, stress, or inflammation in the brain (Itzhaki & Wozniak, 2007). In AD brains, HSV-1 is present in areas of AD neuropathology. Remarkably, APOE ε4 allele carriers showed more than 15-fold risk of AD in the presence of HSV-1 in this sample. The APOE alleles may affect the degree or recurrence of viral damage, rather than susceptibility to infection with the virus (Itzhaki, Lin, Shang, Wilcock, Faragher, & Jamieson, 1997; Itzhaki & Wozniak, 2006). HSV-1 may contribute to the formation of amyloid plaques and abnormally phosphorylated tau protein, possibly by attenuating the processing of amyloid precursor protein (APP) into the toxic Aβ-peptide (Itzhaki & Wozniak 2007; Shipley, Parkin, Itzhaki, & Dobson, 2005). 22 Infection with viruses early in life may put individuals at risk of becoming re-infected throughout life, even chronically (Robinson et al., 2004). Since AD develops later in life, it is hypothesized that pathogens acquired early in life are not expressed until decades later. Age-related decline in immunity against pathogens or a long period of pathogen latency may offer explanations for the effects of pathogens found in older adulthood and rarely found in younger people (Itzhaki et al., 2004; Robinson et al., 2004). Systemic infection may result in the entry of cytokines into the brain and inflammation that reactivates latent HSV, further enhancing damage (Itzhaki & Wozniak, 2006). Viruses may infect multiple sites throughout the body including the periodontium and the brain without requiring a direct relationship between the different sites. It remains possible that PDD and AD are both partially, but independently, influenced by viruses. Inflammation and AD Inflammatory markers appear to be higher in persons with AD than normal age-matched controls. Blood IL-6 (Singh & Guthikonda, 1997) and α1- antichymotrypsin (ACT) levels (Licastro, Candore, Lio, Porcellini, Colonna- Romano, Franceschi, et al., 1995) are elevated in some AD patients compared to controls. In the brain, acute phase proteins IL-1, IL-6, S-100, CRP, and α2- macroglobulin are elevated in the temporal cortex of AD patients compared to controls (Griffin, Stanley, Ling, White, MacLeod, Perrot, et al., 1989; McGeer, Rogers, & McGeer, 1994; Wood, Wood, Ryan, Graff-Radford, Pilapil, Robitaille, et al., 1993). CRP and serum amyloid P are localized to the characteristic AD 23 extracellular amyloid deposits and neuronal tau protein aggregates post-mortem (Duong, Nikolaeva, & Acton, 1997; Kalaria,1992). Microvessels from AD brains release 60 to 88 percent more IL-1β, IL-6, and TNF-α than non-AD controls (Grammas & Ovase, 2001). Though some AD patients have elevated systemic inflammatory markers, it is not clear whether systemic inflammation precedes dementia or if neuroinflammation itself might result in systemic inflammation. Longitudinal studies, discussed below, suggest that elevated systemic inflammation predicts dementia months to years later; however, neurological changes can precede clinical signs of dementia by more than a decade. In community based samples, elevated blood inflammatory markers predict risk for dementia and incidence of cognitive impairment (Alley et al., 2008; Engelhart et al., 2004; Schmidt et al., 2002; Tan et al., 2007). In the Rotterdam Study, elevated blood IL-6, CRP, and ACT predicted increased risk of dementia onset over a year later (Engelhart et al., 2004). Over a follow up period of 25 years, men in the Honolulu-Asia Aging Study had a 3-fold increased risk for dementia in the top quartiles of inflammatory markers versus the lowest quartile at midlife (Schmidt et al., 2002). In individuals with AD, elevated IL-1β predicted rates of cognitive decline (Holmes et al., 2003). Patients with elevated markers preceding a baseline exam showed a greater rate of cognitive decline over a two month follow up period than those who did not have elevated levels prior to baseline. Though high levels of inflammatory markers predict dementia risk and cognitive decline in the demented, they may not be associated with decline in those with higher levels of normal cognitive function (Alley et al., 2008; Dik, 24 Jonker, Hack, Smit, Comijs, & Eikelenboom, 2005). For example, Dik and colleagues (2005) found that ACT was associated with declines in mental status, but not with measures of memory, fluid intelligence, or information processing speed. Model of Proposed Mechanisms for an Association between Periodontal Infection and AD This model proposes possible links between oral infection and the pathology of AD. It does not claim that oral infection or inflammation are the causes of AD, rather we propose that they may contribute to, exacerbate, and share risk factors with AD. Pathogenic bacteria in the oral cavity can lead to periodontal infection. Individuals vary in susceptibility to infection, partly due to the state of their oral hygiene and possibly due to particular genotypes that are more vulnerable to infection and have elevated inflammatory responses (discussed in more detail below). Once bacteria and/or viruses enter the blood stream, the infection may become systemic. From systemic circulation, pathogens and their products may cross the BBB and enter the brain. This may contribute to the development of AD pathology through three inter-related processes (see Figure 2.2). They are the direct effects of pathogenic products, the inflammatory response to these pathogens, and the effect on vascular integrity. These processes have been demonstrated to impact microglial activation, the production and formation of amyloid beta (Aβ) and tau protein, and cerebrovascular pathology. Microglial activation is associated with neuron death in AD though the causal direction remains undetermined. Aβ and tau proteins are 25 central contributors to inter-related AD pathologies, plaques and neurofibrillary tangles. Figure 2.2. Proposed Pathways between Periodontal Infection and Alzheimer’s Disease Pathology. LPS- lipopolysaccharide; APP- amyloid precursor protein; AB- Amyloid beta; NFTs neurofibrillary tangles Pathogen Products. The cell walls of Gram negative bacteria contain LPS that induces a number of host defenses. LPS stimulates certain inflammatory cytokines that are associated with microglial activation and altered processing of APP (Brugg, Dubreuil, Huber, Wollman, Delhaye-Bouchaud, & Mariani, 1995; Mattson, 2004). Animal studies show that chronic infusion of LPS into rat brains may result in long lasting inflammatory reaction with pathological changes. These changes include an increased number of activated astrocytes, increased number and density of reactive microglia, increase in IL-1β, TNF-α, and beta amyloid Oral Infection Systemic Infection Susceptibility to infection Interrelated Pathology Interrelated Mechanisms Interrelated Processes Vascular Changes Hyper- coagulation platelet aggregation Inflammation cytokines, acute phase reactants, complement, etc. Pathogen Products LPS endotoxin Viral effects Microglial activation APP / AB Tau protein phosphorylation atherogenesis / thrombogenesis Neuron death Plaques NFTs Cerebral microvessel pathology 26 precursor protein (βAPP), the degeneration of hippocampal pyramidal neurons, impairment in spatial working memory, and decreased size of hippocampus and temporal lobe associated with increased lateral ventricles (Hauss-Wegrzyniak, Vraniak, & Wenk, 2000). Viruses could also contribute directly to AD pathology. HSV has glycoproteins that are very similar in amino acid sequences to Aβ and tau protein and may aggregate like Aβ (Cribbs, Azizeh, Cotman, & LaFerla, 2000; Takakuwa, Goshima, Koshizuka, Murata, Daikoku, & Nishiyama, 2001). HSV may also impact the processing of APP (Benboudjema et al., 2003). For example, Shipley et al. (2005) reported that HSV-1 and HSV-2 infection of neuronal cells in vitro caused rapid decreases in cell levels of full length APP. The production of Aβ may be increased as a result of altered APP processing or may also result directly from HSV infection (Itzhaki & Wozniak, 2006; Satpute- Krishnan, DeGiorgia, & Bearer, 2003). Inflammation. It is clear that inflammation is involved in AD. However, it is not yet clear if inflammatory processes initiate pathological processes or merely contribute to disease progression. There are several mechanisms through which systemic inflammation might contribute to pathogenesis in AD. These include the priming of microglia, dysregulation of APP and Aβ processing and metabolism, the activation of microglia in response to Aβ, and a neurotoxic loop or vicious cycle in which immune response intended to be neuroprotective leads to exacerbation of the process. 27 The presence of primed microglia may influence the response of the brain to systemic infection. In aging individuals, changes in glia can cause exaggerated inflammatory responses (Licastro et al., 2005). It is hypothesized that in AD and other neurodegenerative diseases, microglia become activated, leading to higher production of inflammatory mediators and chronic overreaction to subsequent stimuli (Marx, Blasko, Pavelka, & Grubeck-Loebenstein, 1998; Perry, 2004). Microglia release many inflammatory mediators in the brain including acute phase proteins, complement factors, prostaglandins, free radicals, and cytokines (Perry, 2004). Aβ peptides in the brain may also potentiate monocyte transmigration from blood to brain (Brod, 2000). The processing and metabolism of APP and Aβ is critical in AD pathogenesis (Blasko & Grubeck-Loebenstein, 2003; vonBernhardi, Ramirez, Toro, & Eugenin, 2007). In AD, this balance is dysregulated. Aβ is aggregated as oligomers and fibrils which have varying neurotoxicity (Klein, Krafft, & Finch, 2001). Systemic inflammation and cytokine production can augment the regulation of APP and Aβ, though the effects are complex and may depend on which combination of mediators is present (Blasko & Grubeck-Loebenstein, 2003; Dziedzic, 2006; Heneka & O’Banion, 2007; Marx et al., 1998). Griffin and colleagues (1998) proposed that IL-1 is critical to the processing of βAPP, favoring continued Aβ deposition and the cyclical continuation of inflammatory response and cytokine overexpression. TNF-α and interferon gamma in combination also alter the metabolism of βAPP, trigger Aβ peptide production, and inhibit soluble APP secretion (Blasko, Marx, Steiner, 28 Hartmann, & Grubeck-Loebenstein, 1999). IL-1 stimulates APP synthesis and factors that lead to its amyloidogenic properties (Schmitt, Steiner, Klinger, Sztankay, & Grubeck-Loebenstein, 1996), contributes to the phosphorylation of tau protein favoring tangle formation, increases production of nitric oxide synthase fatal to cells, and increases the production of acetylcholinesterase responsible for the breakdown of acetylcholine which is important in learning and memory function (Mrak & Griffin, 2005). Beta secretase, a protease that cleaves APP and results in toxic Aβ peptides, may also be upregulated by inflammatory mediators (Heneka & O’Banion, 2007). In mice with APP gene mutations, systemic administration of LPS resulted in altered expression and processing of APP and increased production of Aβ (Sheng, Bora, Xu, Borchelt, Price, & Koliatsos, 2003). Mice with this gene mutation generate high levels of amyloidogenic Aβ. Aβ and its aggregation trigger the activation of microglial cells which respond by producing acute-phase proteins, complement components, prostaglandins, and cytokines some of which are neurotoxic or contribute to aggregation by stimulating Aβ production. This response may be more injurious than the plaques and tangles to which inflammatory processes are responding, resulting in neural damage and death (Bate & Williams, 2004; Blakso & Grubeck- Loebenstein, 2003; D’Andrea, Cole, & Ard, 2004; Dziedzic, 2006; Marx et al., 1998; McGeer & McGeer, 1995; Nagele, Wegiel, Venkataraman, Imaki, Wang, & Wegiel, 2004; Neuroinflammation Working Group, 2000; vonBernhardi et al., 2007; Wilson, Finch & Cohen, 2002). 29 Nonspecific responses of phagocytic cells can remove Aβ deposits, as well as recruit more microglia which produce more immune mediators. Cell attempts to dissolve plaques and tangles may cause toxicity via the release of neurotoxic substances including nitric oxide and other reactive oxygen species (Heneka & O’Banion, 2007; Marx et al., 1998). Pro-inflammatory cytokines including IL-6 and TNF-α can be directly toxic in high concentrations or can stimulate further Aβ production, aggregation, and toxicity (Blakso & Grubeck- Loebenstein 2003; Brod, 2000; Heneka & O’Banion, 2007; Marx et al., 1998; Perry, 2004). Chronic inflammation may result in chronic acute-phase protein secretion which favors the formation of Aβ fibrils. This fibrillar conformation of Aβ may be important in its ability to induce inflammation (Blasko & Grubeck- Loebenstein, 2003). Over sustained periods of time, these products may contribute to neurodegeneration via injury of surrounding non-infected cells resulting in neuron loss (Rogers, Strohmeyer, Kovelowski, & Li, 2002). Vascular Changes. Oral bacteria including streptococcus sanguis and p. gingivalis have been shown to result in the expression of platelet aggregation proteins that may play a role in the formation of atheromas and thrombi possibly contributing to vascular disease (Herzberg & Meyer, 1996; Herzberg & Meyer, 1998; Pham, Feik, Hammond, Rams, & Whitaker, 2002; Sharma, Novak, Sjoar, & Swank, 2000). High levels of atherosclerosis have been found to increase the risk of cognitive decline independent of other factors (Haan, Shemanski, Jagust, Manolio, & Kuller, 1999). Furthermore, the presence of inflammatory cytokines in the blood increases platelet aggregation in cerebral blood vessels which can lead 30 to atherogenesis and thrombus formation associated with strokes and hypoperfusion. Vascular changes in the brain may also contribute to the formation of AD pathology. Platelets are a primary source of APP (Bush, Martins, Rumble, Moir, Fuller, Milward, et al., 1990; Chen, Inestrosa, Ross, & Fernandez, 1995; Zandi & Breitner, 2001) and platelet aggregation associated with cerebrovascular pathology may increase the production of Aβ in the brain. In endothelial cells, Aβ causes the secretion of inflammatory proteins which upregulate the production of APP (Grammas & Ovase, 2001). Possible Role of Genes in Infection and Inflammation Acquisition of infectious disease requires the presence of both a pathogen and a susceptible host. The expression of disease depends on virulence of the pathogen and immune response of the host to the pathogen (Beck, Slade, & Offenbacher, 2000). Periodontal researchers (Beck et al., 1996) hypothesize a hyper-inflammatory phenotype that causes some individuals to have exaggerated inflammatory reactions in response to pathogens. Particular alleles in the HLA- DR3/4 or –DQ system are suspected to produce exaggerated inflammatory responses to infection. Carriers have peripheral blood monocytes that secrete 3 to 10 fold greater amounts of inflammatory mediators in response to LPS. For example, patients with refractory periodontitis, a condition in which oral health does not improve despite treatment and proper hygiene, release higher levels of IL-1β and prostaglandin-E2 than patients with non-refractory periodontitis (Hernichel-Gorbach, Kornman, Holt, Nichols, Meador, Kung, et al., 1994). 31 Polymorphisms for genes involved in inflammatory process are a logical target for exploration in understanding a possible common cause or link between periodontal disease and Alzheimer’s disease. Several candidate genes have been proposed that may potentially link PDD and CVD (Goteiner, Ashmen, Lehrman, Janal, & Eskin, 2008; Kornman, Pankow, Offenbacher, Beck, diGiovine, & Duff, 1999; Kornman & Duff, 2001). While no publications have directly considered gene polymorphisms common to PDD and AD, IL-1 and TNF- α polymorphisms have been associated with both diseases (Diehl, Wang, Brooks, Burmeister, Califano, Wang, et al., 1999; Donati, Berglundh, Hytonen, Hahn-Zoric, Hanson, & Padyukov, 2005; Galbraith, Hendley, Sanders, Palesch, & Pandey, 1999; Grimaldi, Casadei, Ferri, Veglia, Licastro, Annoni, et al., 2000; McCusker, Curran, Dynan, McCullagh, Urquhart, Middleton, et al., 2001; Nicoll, Mrak, Graham, Stewart, Wilcock, MacGowan, et al., 2000; Rainero, Bo, Ferrero, Valfre, Vaula, & Pinessi, 2004). Not all studies have confirmed these associations (Craandijk, van Krugten, Verweij, van der Velden, & Loos, 2002; Folwaczny, Glas, Torok, Mende, & Folwaczny, 2004) and more research on genetic polymorphisms is needed to explain discrepant findings (Loos, 2005). The APOE ε4 allele is implicated in susceptibility to infection or the degree of resulting damage from viruses (Itzhaki et al., 1997; Lin, Graham, MacGowan, Wilcock, & Itzhaki, 1998; Lin, Wozniak, Esiri, Klenerman, & Itzhaki, 2001; Wozniak et al., 2005). AD brains positive for HSV were 17 times more likely to carry the APOE ε4 allele compared to HSV negative non-AD brains suggesting 32 that the combination of an APOE ε4 allele and HSV may confer higher risk for AD than either alone (Itzhaki et al.,1997). APOE ε4 alleles may also play a role in inflammatory response to infection. For example, mice with APOE ε4 genotypes have greater elevations of inflammatory cytokines in systemic circulation and in the brain in response to LPS than those with APOE ε3 (Lynch, Tang, Wang, Vitek, Bennett, Sullivan, et al., 2003). The ability of astroglial cells to phagocytize Aβ may also depend on APOE type (Heneka & O’Banion, 2007). ApoE and other genetic markers of hyper-inflammatory response could be used to identify individuals at risk for targeted prevention and treatment. Establishing a Relationship Many more studies are needed to validate an association between periodontal disease and dementia, including prospective studies that directly measure oral microbiology. Several requirements must be met before a causal relationship between PDD and AD can be established. First, PDD must precede dementia. Among the few longitudinal or archival studies that have reported tooth loss as a risk factor for dementia, two have measured periodontal disease as a possible cause for the tooth loss (Shimazaki, Soh, Saito, Yamashita, Koga, Miyazaki, et al., 2001; Stein et al., 2007), although none has established periodontal disease per se as the risk factor for dementia. Next, PDD must be correlated with AD. A few studies have suggested this is to be the case (e.g., Gatz, Mortimer, et al., 2006; Kondo et al., 1994; Stein et al., 2007) although tooth loss was used as the index of PDD in these studies. Finally, all other possible 33 contributors must be controlled for. While this is not entirely possible, steps have been and could be taken to account for some potential contributing factors. This includes external factors that might contribute to both tooth loss and dementia separately such as viral infection in different body systems, head injury, low socioeconomic status, malnutrition, or an exaggerated inflammatory profile. Eventually, randomized-controlled intervention trials with a long follow-up period would be needed to establish whether preventive oral health measures could reduce the risk of AD. For inflammation to be clearly established as a mediator for the relationship between PDD and dementia, it should be present in both oral disease and dementia as has been demonstrated through several studies (Licastro et al., 1995; Loos, 2005; Singh & Guthikonda, 1997). Treating PDD should reduce inflammation (D’Aiuto & Tonetti, 2004; Ebersole et al., 1997; Mattila et al., 2002; Taylor et al., 2006) and reducing inflammation should lead to reduced incidence of dementia. Epidemiological studies suggest that non- steroidal anti-inflammatory drugs (NSAIDs) may protect against the development of AD if taken in midlife, many years prior to the diagnosis of dementia (e.g., Hayden, Zandi, Khachaturian, Szekely, Fotuhi, Norton, et al., 2007). However, evidence from epidemiological studies, fundamental pathology, and clinical trials suggest that COX-2 inhibitors fail to reduce AD pathology and have not been shown to be effective in individuals who have already developed dementia (McGeer & McGeer, 2007; Townsend & Pratico, 2005). 34 No published studies have measured inflammation concurrently with PDD and dementia to allow a determination of its role in the relationship. It is necessary to determine whether inflammatory processes in the brain are initiated or exacerbated by systemic infection and inflammation resulting from PDD, or whether they reflect completely independent sources (e.g., immune response to AD pathology). Conclusion This review and synthesis of the association between PDD and AD integrates research from disparate fields. Despite evidence that oral disease may be a risk factor for dementia, little attention has been devoted to explaining the potential mechanisms for this association (see Kramer et al., 2008; Stein, Scheff, & Dawson, 2006). It is proposed that bacterial and viral infections commonly found in PDD may impact the brain, either directly or via systemic signals to the brain, and contribute to the development of AD. Periodontal infections may result in harmful pathogenic products leading to systemic inflammatory responses. Elevated systemic inflammatory response may contribute to the exacerbation of existing brain pathologies. Infections may also contribute to vascular pathology with the potential to impact brain function. PDD and AD may also share common risk factors such as genetic polymorphisms related to production of inflammatory mediators. Though there are theoretical reasons for suspecting that pathogens may play a role in AD, controversy remains given the mixed results of post-mortem brain studies (Kinoshita, 2004). Further evidence is needed to determine whether 35 pathogens contribute uniquely to AD or are involved in a common cause. For example, perhaps those carrying the APOE ε4 allele may have a hyper- inflammatory response to challenges from pathogen products or be highly vulnerable to infection. Few studies have attempted to link oral health with AD diagnosis or disease pathology and none have investigated the role of inflammation as a potential mediator. Information about history of chronic infection among AD patient populations would help to investigate this hypothesis. Longitudinal data measuring periodontal status, levels of inflammatory markers, and cognitive status would be ideal. If systemic infection and inflammation prove to be contributors to AD, several preventive measures and treatment strategies would be implied. Such developments may be particularly significant given the paucity of promising preventative strategies for AD at the present time. Focus on the prevention of oral and other sources of systemic infection would be warranted. Timely treatment of periodontal infection might be indicated to reduce risk of systemic infection and inflammation. Though successful treatment of PDD has shown reductions in inflammation (e.g., Taylor et al., 2006), it remains to be determined whether this would reduce risk of CVD or dementia. Identification of individuals susceptible to infection and hyper-inflammation would allow for targeted prevention to reduce contact with pathogens and treatment strategies to reduce harm from hyper-inflammatory incidents. 36 The model presented here should enable researchers to test specific hypotheses regarding the multiple inter-related mechanisms that may be responsible for an association between oral infection, inflammation, and the development of Alzheimer pathology. In particular, research should examine whether pathogen products and inflammation resulting from periodontal infection are related to these same processes suspected to contribute to pathology in AD or whether the two diseases merely share common risk factors. 37 CHAPTER III: RESEARCH DESIGN AND METHOD Research Design The present research uses data from twins, which allow for the use of both case-control and co-twin control designs (Lichtenstein, DeFaire, Floderus, Svartengren, Svedberg, & Pedersen, 2002). Case control allows us to compare individuals with dementia to genetically unrelated individuals who do not have dementia as would be done in a non-twin study. Co-twin control is a matched pair design in which one twin serves as a control for the other. The comparison of a monozygotic (MZ) twin with dementia to his non-demented twin provides a virtually exact match on age and genetic factors and allows some control over environmental history and family background (Page, 1995). Using dizygotic (DZ) twins also allows some control over environmental history, age, and family background with the genetic similarity of siblings. Sample The sample for the present study comprised members of the National Academy of Sciences-National Research Council (NAS-NRC) Registry of World War II veteran male twins who are part of the Duke Twins Study of Memory in Aging (See Figure 3.1). In 1955, NAS-NRC began to identify twins where both members of the pair had served in the Armed Forces during World War II. The birth certificates of approximately 93% of the primarily white male twins born between 1917 and 1927 were matched with veteran status files from the Department of Veterans Affairs to create the NAS-NRC Twin Registry. The 38 registry is made up of 15,924 complete pairs of veteran twins (31,848 individuals; see Breitner, Welsh, Gau, McDonald, Steffens, Saunders, et al., 1995; Jablon, Neel, Gershowitz, & Atkinson, 1967; Page, 2002). Pairs were excluded if one or more of the twins was deceased or out of the country. Birth Certificates twins born 1917-1927 VA Military Service Records 1940s NAS-NRC Veteran Twin Registry N = 31,848 (15,924 complete pairs) Screened at 1 st wave N = 12,709 (5,699 pairs) One or both twins in a pair deceased, no longer in U.S., refused, not located Duke Twin Study Dementia Screening Potentially Eligible N = 18,426 (9,213 pairs) One or both twins out of country N = 78 Eligible N = 15,934 One or both twins deceased N = 2,414 (1,207 pairs) Refused N = 1,650 Not Located N = 1,575 Surviving pairs not out of country, per NAS data 1988 Figure 3.1. Attrition from the Initial NAS-NRC Sample to Completion of First Wave of Duke Dementia Screening (adapted from Breitner et al., 1995). In 1989, the Duke Twins Study (Breitner et al., 1995) attempted to contact all veteran twin pairs still residing in the U.S. where both were still living. A set of 18,426 individuals was determined to be potentially eligible. After sending an introductory letter to potentially eligible participants, 1,235 previously unrecorded 39 deaths were discovered in 1,207 pairs that were not studied further. An additional 78 individuals were out of the country and therefore those pairs were ineligible. Of the remaining eligible participants, 1,650 refused participation and another 1,575 were unable to be located despite tracing efforts. From 1990-1992, the remaining 12,709 men (including 5,699 complete pairs) aged 63-73 years old were screened for dementia by telephone using the Modified Telephone Interview for Cognitive Status (TICS-m) described in the Procedures section. This screening constituted the first wave of data collection. Participants considered to be not demented were subsequently re-screened in the following waves (See Table 3.1). The mean age of participants at the time they were screened in wave 1 was 66.1 (SD = 2.8). Table 3.1. Waves of Telephone Cognitive Screening in Duke Twins Study Wave 1 complete pairs only 1990-1992 Wave 2 living individuals, regardless of twin mortality, cognitively normal 1993-1995 at last screening Wave 3 complete pairs only, cognitively normal at last screening 1996-1998 Wave 4 complete pairs only, cognitively normal at last screening 2001-2002 Notes: 1. Co-twins of demented or cognitively impaired but not demented (CIND) individuals who had completed at least one in-person assessment were re-assessed at regular intervals (typically every 12-18 months) using the telephone protocol. 2. Beginning at Wave 3, the TICS-m threshold was raised to education adjusted TICS-m of >28. Criteria for being deemed not demented included: 1) an education adjusted TICS-m score > 28 at previous screening wave, 2) an education adjusted TICS-m < 28 and a negative dementia questionnaire (DQ) at previous 40 screening wave, 3) answers on a proxy cognitive function measure indicating ‘not demented’ at previous screening wave, 4) answers on a proxy cognitive function measure indicating ‘suspected dementia’ and a negative DQ at previous screening wave, or 5) a diagnosis of cognitively normal based on an in-person assessment. There were four waves of screening (See Figure 3.2). Starting Sample Wave 1: 18,424 Wave 2: Not Available Wave 3: 10,410 Wave 4: 8,216 Either Twin Deceased Wave 1: 2,414 Wave 2: Not Available Wave 3: 1,476 Wave 4: 1,115 Completed Screen Wave 1: 12,709 Wave 2: 11,160 Wave 3: 7,026 Wave 4: 5,022 Refused Wave 1: 1,650 Wave 2: Not Available Wave 3: 1,536 Wave 4: 1,492 Not Located/ No Proxy Wave 1: 1,651 Wave 2: Not Available Wave 3: 382 Wave 4: 587 Not Targeted for DQ Wave 1: 11,122 Wave 2: 9,020 Wave 3: 5,699 Wave 4: 3,756 Targeted for DQ Wave 1: 1,587 Wave 2: 2,140 Wave 3: 1,329 Wave 4: 1,266 Either Twin Deceased Wave 1: 20 Wave 2: 90 Wave 3: 22 Wave 4: 16 Not Located/ No Proxy/ Refused Wave 1: 276 Wave 2: 403 Wave 3: 170 Wave 4: 133 DQ Completed Wave 1: 1,291 Wave 2: 1,647 Wave 3: 1,137 Wave 4: 1,117 Targeted for Clinical Assessment Wave 1: 87 Wave 2: 119 Wave 3: 158 Wave 4: 288 Not Targeted for Clinical Assessment Wave 1: 1,204 Wave 2: 1,521 Wave 3: 979 Wave 4: 829 Deceased Wave 1: 5 Wave 2: 10 Wave 3: 7 Wave 4: 18 Refused/ Not Located Wave 1: 5 Wave 2: 16 Wave 3: 12 Wave 4: 29 Proband Clinical Assessment Completed Wave 1: 77 Wave 2: 93 Wave 3: 139 Wave 4: 241 Figure 3.2. Flowchart of Waves of Duke Dementia Screening (adapted from Plassman et al., 2006). 41 Participants suspected to be demented at any wave (education adjusted TICS-m score < 28 or answers on the proxy measure indicated suspected dementia) were targeted for follow up with the DQ as described in Procedures below. At each wave, some participants were found to be deceased, unable to be located, or refused to participate in the DQ. Participants were also excluded if their co-twin was deceased (except in wave 2). After completing the DQ, participants were classified as “suspected dementia” and targeted for clinical assessment or classified as “no dementia” in which case they were subsequently re-screened. Of those targeted for clinical assessment, some were unable to be fully assessed and diagnosed if they were deceased, unable to be located, or refused. When the Registry was constructed, selected information was abstracted for each Registry member from his military record and archived on microfiche film. Due to a subsequent fire at the St. Louis military record repository, for most of these men, these archived microfiche data were all that remained of their military record. As part of the present study, archived records were scanned into .PDF files and the information subsequently entered into a database. The present sample includes all dementia cases and all co-twins of dementia cases, excluding dementia concordant pairs whose onset was less than three years apart (and pairs in which the non-demented co-twin died less than 3 years after the onset of his demented twin or had less than 3 years of follow-up after onset of the proband). Data were selected for complete pairs, not for single individuals. See Table 3.2 for a description of the selection criteria. 42 Pairs were also excluded if the cognitive status of one twin was unknown, the zygosity was unknown, or the military records were unavailable. The present sample also includes non-demented control pairs. The controls were complete pairs with known zygosity, data on at least two waves of measurement where neither twin was demented, and where the last screening for each indicated they were not demented. Due to budget constraints, extraction of military records was completed for 91% of the potential control pairs. Selected control pairs were prioritized for inclusion if they had available APOE data, had available data for 3 or more screening waves, had available data for a mailed questionnaire, or were among the oldest pairs living beyond age 70. Table 3.2. Selection Criteria for the Present Sample Demented Inclusion of twin pairs 1. Concordant for dementia with > 3 years difference in age of onset (N = 55) 2. Discordant for dementia (N = 261) Exclusion of twin pairs 1. Concordant for dementia with < 3 years difference in age of onset (N = 64) 2. Cognitive status of co-twin is unknown (N = 36) 3. Zygosity unknown (N = 22) 4. Co-twin died before onset of proband or <3 years after onset of proband, or had < 3 years follow up after onset of proband (N = 68) 5. Excluded for unknown reason, one twin had non-AD dementia (N = 6) Control Inclusion of twin pairs (N = 2,860) 1. Neither twin is demented 2. Two of more waves of TICS-m data are available Exclusion of twin pairs 1. Zygosity unknown (N = 116) 2. Fewer than two waves of TICS-m in both twins (N = 5,692) 3. No data available for co-twin (N = 783) 4. Excluded due to budget constraints (N = 283) Total number of complete pairs included in the study (N = 3,176) 43 Case Control Sample The case control sample consists of 371 cases diagnosed with dementia (all dementia types combined) and 5,981 controls who screened negative for dementia at the last wave of measurement. That is, all individuals in the concordant pairs (55 pairs, 110 individuals) and half of the individuals in the discordant pairs (261 pairs, 261 individuals) are demented. All of the individuals in the control pairs (2,860 pairs, 5,720 individuals) and the other half of the individuals in the discordant pairs (261 pairs, 261 individuals) are not demented. Non-demented controls include individuals categorized as cognitively normal and “cognitively impaired but not demented” (CIND). Individuals who screened positive initially, but were found cognitively normal upon further evaluation, were categorized as cognitively normal. The categorization of CIND was determined by expert consensus after review of collected clinical information and medical records. The sample includes twin pairs and therefore contains data dependencies. Analytic approaches for handling data dependencies are described in the results chapters. Co-twin Control Sample The co-twin control sample consists of 316 twin pairs. In 261 of those pairs, one twin has dementia and the other does not have dementia. The remaining 55 pairs include twins who both have a current diagnosis of dementia but who are discordant for onset. Discordant onset is defined as a difference in age of dementia onset of 3 years or greater. This definition of discordance is based on previous research by Breitner, Gau, Welsh, Plassman, McDonald, 44 Helms et al. (1994). Prior to the establishment of this definition, studies did not consider the length of follow up needed to accurately define a pair as discordant or possible inaccuracies involved in establishing age of dementia onset. Although an improvement from studies with no such definition, debate exists regarding the best definition for discordance. Gatz, Pedersen, Crowe, and Fiske (2000) suggest that a 5 year interval allows a better margin of error for the estimated age of onset. In the present study analyses are conducted both including and excluding pairs discordant for dementia. For the purposes of the co-twin control analysis, the first twin to develop dementia is defined as the proband while the second to develop dementia is defined as the control. Among the 316 discordant pairs, 52% are MZ. This is the same percentage of MZ pairs in the entire sample. Procedures Identification of dementia in the Duke Twins Study was based on a telephone screening procedure followed by an in-person diagnostic assessment. Telephone Screening Procedure Study participants were contacted via telephone and were administered the Modified Telephone Interview for Cognitive Status (TICS-m; Welsh, Breitner, & Magruder-Habib, 1993). The TICS-m is a brief assessment (5-10 minutes) modeled after the Mini Mental State Exam. It assesses a range of cognitive functions including orientation, immediate and delayed verbal memory, calculation, language, repetition, general knowledge, and abstraction. It contains 21 items and has a total score of 50 points. For cases in which the twin in question was unable to participate directly in the TICS-m, a nine-item proxy 45 questionnaire or the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE; Jorm & Jacomb, 1989) was administered to an informant to assess signs and symptoms of dementia. For participants who were suspected to have cognitive dysfunction based on the TICS-m or the proxy screening instrument, a second phase of telephone screening occurred in which the Dementia Questionnaire (DQ) was administered to an informant, usually a spouse or other close relative (Breitner et al., 1995). The DQ assesses chronological history of cognitive and functional symptoms as well as relevant medical history. The telephone screening procedure has been shown to have good sensitivity and specificity for identifying individuals with dementia (Gallo & Breitner, 1995). Descriptive information regarding TICS-m scores at each wave are found in Table 3.3. In-Person Diagnostic Assessment For individuals suspected to have dementia based on the telephone screening procedure, an in-person diagnostic assessment was conducted (Breitner et al., 1995). The in-person assessment was conducted in the participants’ homes by a research nurse and psychometrician. Nurses and psychometricians were trained to proficiency in each test procedure. The evaluations were 3-4 hour structured assessments including a standard battery of neuropsychological tests, a neurological examination, and measure of blood pressure. The information collected included a chronological history of cognitive and functional changes, detailed medical history and current medications, current neuropsychiatric symptoms, measures of severity of cognitive and functional 46 impairments, and family history of memory problems (Plassman, Steffens, Burke, Welsh-Bohmer, Newman, Drosdick, et al., 2006). Relevant medical records, including neuroimaging results, were sought from the participants’ personal physicians. Onset of dementia was estimated as the age at which the individual clearly met DSM-III-R criteria based on a review of the history of cognitive symptoms. The average age at onset in our sample was 73.0 (SD = 6.2). Final dementia diagnoses were determined by expert consensus and fell into three general categories: demented, CIND, or cognitively normal. Those who were found to be demented were assigned to specific diagnostic categories including AD, vascular dementia, and other types of dementia. Dementia diagnoses were based on DSM-III-R criteria for dementia (American Psychiatric Association, 1987). The criteria used to diagnose AD were from the National Institute of Neurological and Communicative Disease and Stroke and Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA; McKhann, Drachman, Folstein, Katzman, Price, & Stadlan, 1984). For vascular dementia, an operational version of the National Institute of Neurological Disorders and Stroke- Association Internationale pour la Recherche et l’Enseignement en Neurosciences (NINDS-AIREN) criteria were used (Roman, Tetemichi, Erkinjuntti, Cummings, Masdeu, Garcia, et al., 1993). Similarly widely accepted criteria were used for other types of dementia. To reflect that dementia is often the consequence of more than one pathological process, primary and secondary diagnoses were assigned denoting multiple etiologies. The non-demented co- twins of individuals with dementia or participants classified as CIND who had 47 completed at least one in-person assessment were re-tested with the telephone screening procedure at regular intervals. If they were thought to have crossed the threshold for dementia, they would proceed again to the in-person diagnostic assessment. In the present sample, non-demented controls include individuals categorized as CIND and cognitively normal. Descriptive information regarding dementia status and diagnosis types is found in Table 3.3. Table 3.3. Descriptive Statistics for TICS-m and Dementia Diagnosis TICS-m Scores range M (SD) N Wave 1 7-49 33.8 (4.2) 6,126* Wave 2 1-49 33.5 (4.9) 6,047 Wave 3 10-50 34.0 (4.8) 4,980 Wave 4 3-50 32.7 (4.9) 3,586 Current Dementia Status N % of sample Non-demented 5,981 94.2 Total demented all types 371 5.8 Dementia by Clinical Type (primary diagnosis) N % of demented AD 232 62.5 Vascular Dementia 54 14.6 Fronto-temporal Dementia 8 2.2 Parkinson’s Disease Dementia 13 3.5 Lewy Body Dementia 8 2.2 Other Dementia or Undetermined etiology 56 15.0 Cases with secondary diagnosis 165 44.5 *TICS-m screening data not available at wave 1 for the following reasons: N Did not complete, reason unknown 161 Proxy interview completed 46 Did not complete, proxy refused 7 Did not complete, proxy unable to be located 2 Pilot subject and screening not completed according to protocol 10 Total missing 226 48 Measures Descriptions of the measures, their sources, and dates of collection are summarized in Table 3.4. Summary statistics for the primary variables of interest can be found in Table 3.5 and correlations between predictor variables in Table 3.6 (not adjusted for non-independence). Table 3.4. Description of Measures Source Contains Variables Date Measured Military Record Physical Exam Number of teeth lost At military entry Height 1936-1955 Questionnaire 2 (Q2) & Father’s occupational prestige 1967-1972 NORC prestige ratings (NAS-NRC) Duke Twins Study TICS-m 1990-2002 Dementia diagnosis Waves 1-4 Education Zygosity Physical Examination from Military Record The military records of the Registry members contain demographic and health data, including a physical examination conducted at the time of each participant’s entry into the armed forces. The average age of individuals at time of entry into the military was 19.9 (SD = 2.2). The physical examination records include an open-ended text field with information on medical defects, health history, number of teeth missing and other dental observations including the number of cavities, capped or crowned teeth, and the replacement of teeth with a bridge or dentures. The number of teeth missing was coded both as a continuous number and as a categorical proportion of teeth missing (i.e. greater than or less 49 than half of all teeth lost). Teeth that have been replaced by dentures or bridges are considered to be missing. The number of missing third molars (wisdom teeth) was not included in the total number of missing teeth when they were identified as such. Third molars generally emerge in young adulthood, around the time when these men were entering the military. Therefore, they are likely to be missing because they have not yet erupted rather than as the result of disease or accident. It is also possible for third molars to be congenitally absent. For more detail on dental practice during the WWII era see Appendix A. The mean number of missing teeth was 3.8 (SD = 2.9). See Figure 3.3 for the frequency distribution. Over 25% of the sample were missing more than 4 teeth (N = 1,618). Less than 1% of the sample were missing half or more of their teeth (N = 40). As the number of teeth missing was reported in an open-ended text field, it is likely that the modal number of missing teeth = 4 is due to reporting of wisdom teeth as missing without identifying them as wisdom teeth to be subtracted from the total number of teeth missing. 50 0 200 400 600 800 1000 1200 1400 1600 1800 0 2 4 6 8 10 12 14 16 18 20 22 26 number of missing teeth frequency Figure 3.3. Frequency Distribution of the Continuous Number of Missing Teeth at Time of Entry into the Military. Twelve percent of the sample was missing data for the number of teeth. Military records are present for all individuals in the dataset, so missing tooth information does not indicate missing military record. The group missing data for the number of teeth has a higher mean education (t = -1.99, p < .05) and father’s occupational prestige (t = -2.76, p < .01) than the group who are not missing data for this variable. It is therefore possible that a lack of data indicates a lack of dental problems. The number of cases diagnosed with dementia does not differ between individuals for whom data on the number of teeth is missing or not missing (X 2 = .45, Fisher’s Exact Test for significance = .56). As the Fisher’s Exact Test cannot correct for non-independence of twins, we also used a logistic regression using a SAS macro to adjust the confidence intervals. Individuals with 51 missing data for number of teeth did not have a higher likelihood of dementia diagnosis (OR = 1.12, CIs 0.80, 1.57). Previous studies have used adult height as an indicator of childhood economic conditions and nutritional deprivation and have linked adult height to mortality, vascular disease, reproductive cancers, and other diseases (e.g. Arnesen & Forsdahl, 1985, Barker, Osmond, & Golding, 1990; Blackwell, Hayward, & Crimmins, 2001; Floud, Wachter, & Gregory, 1990). In the present study young adult height was taken from the military physical exam record and is used as a marker for early life health status. In the present sample, height is weakly negatively correlated with the number of missing teeth (r = -.08, p < .001). Father’s Occupational Prestige from Mailed Questionnaire All Registry members were mailed a questionnaire during the time period of 1967-1972 (called Q2). Q2 included items regarding parental occupation using U.S. Census occupational coding. Over 90% of the mother’s occupation responses were missing or were uninformative regarding social status (i.e. housewife). Therefore, father’s occupation was chosen. Occupational prestige was ascertained from these codes using the National Opinion Research Center (NORC) prestige scale. The NORC prestige scale is based on studies of public attitudes regarding the prestige of various occupations that were then mapped onto the list of occupations from the U.S. Bureau of the Census (Reiss, 1961; Hodge, Siegel, & Rossi, 1964). For example, a bootblack has a prestige score of 9, and a farmer has a score of 41, while a physician has a score of 82. In the present sample, prestige ranged from 13.5 to 82 points. Prestige scores are 52 weakly negatively correlated with the number of missing teeth (r = -.08, p < .001). A full table of prestige scores can be found in a book by Lin (1976). We used these prestige scores as a measure of early life socioeconomic status. Duncan’s socioeconomic index (Reiss, 1961) is a more often cited instrument for this purpose, but its calculation requires knowledge of income, which is not available from these data. Two raters independently rated twins’ responses. All disagreements between twins’ reports of father’s occupation (20% of all twin pairs) were resolved according to the criteria listed in Appendix B. Education Years of education were included as a covariate as determined by self report in the Duke Twins dementia screening. Level of education is positively correlated with scores on the TICS-m at all four waves (range r = .38 to .41, p < .001) and with father’s occupational prestige (r = .29, p < .001). Education is negatively correlated with number of missing teeth at entry into the military (r = - .13, p < .001) and with age at time of first screening (r = -.04, p <.01). The mean number of years of education differed for demented (M = 13.2, SD = 3.3) and non-demented subjects (M = 13.8, SD = 3.1), t = 3.43, p = .001. As age at time of first screening also differs for demented (M = 67.5, SD = 2.98) and non- demented subjects (M = 66.0, SD = 2.81), effects of education were analyzed after accounting for age differences among demented and non-demented subjects. 53 Zygosity Zygosity in the twin registry was determined from questionnaire, military records, and fingerprint records (Jablon et al., 1967). For 492 pairs, zygosity was determined by blood or buccal DNA samples (Reed, Plassman, Tanner, Dick, Rinehart, & Nichols, 2005). In categorical analyses, education was divided into two categories corresponding to less than 12 years of education, or 12 or more years of education. This is a traditional cutoff for educational attainment. As no information was available to establish a threshold for low occupational prestige level based on the NORC scale, father’s occupational prestige was divided into two categories separating roughly the bottom 10 percent (actually 8.7%) from the remaining 90 percent. Height was divided into two categories separating those fewer than 69 inches tall from those 69 inches or taller. The cutoff for height was based on previous research on height as a marker of socioeconomic status which suggested that the mean height of males born in the 1920s and 1930, albeit not in the U.S., belonging to the lowest social class was around 68 or 69 inches (Arnesen & Forsdahl, 1985; Barker, Osmond, & Golding, 1990). The mean height in our sample is 68.1 inches. Previous studies of tooth loss and dementia used categories corresponding to missing half or more of all teeth. This cutoff was used in the case control analyses. Although we attempted to use this cutoff in the co-twin sample, there were not enough individuals missing more than half of all teeth to allow for co-twin control analyses. We therefore, divided tooth loss into the two categories that were closest to missing half of one’s teeth 54 that still had a large enough number of individuals to allow the analyses to be run. The categories were those missing 12 or fewer teeth vs. those missing 12 or more. See Figures 3.4 to 3.7 for description of the number of dementia cases by education, father’s occupational prestige, height, and number of missing teeth. Table 3.5. Summary Statistics for Primary Predictor Variables M (SD) % of sample At time of entry into the military Number of missing teeth 3.8 (2.9) Missing half or greater of all teeth 0.7% Missing fewer than half of all teeth 99.3% Missing 12 or more teeth 1.4% Missing Fewer than 12 teeth 98.6% Height (inches) 68.1 (2.6) Fewer than 69 inches 28.8% 69 inches or taller 71.2% Mailed questionnaire Father’s occupational prestige* 42.0 (11.7) Prestige less than 30 8.7% Prestige higher than 30 91.3% At time of dementia screening Education (years) 13.7 (3.1) Fewer than 12 years 15.2% 12 years or more 84.8% Zygosity Monozygotic 51.9% Dizygotic 48.1% * Note that a prestige score of 41 is indicative of status of occupations including: farmer, plumber, tailor, and railroad conductor. 55 Table 3.6 Correlations between Continuous Predictor Variables Age Educ Occup Height Teeth 1. 2. 3. 4. 5. 1. Age -- 2. Education (years) -.04** -- 3. Father’s occupational prestige -.02 .29** -- 4. Height (inches) .02 .17** .10** -- 5. Missing Teeth -.07** -.13** -.08** -.08** -- ** p <.001 0 20 40 60 80 100 120 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 20 Years of Education % Demented Figure 3.4. Percent Demented by Years of Education. 56 0 2 4 6 8 10 12 14 16 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 24 father's occupational status % Demented Figure 3.5 Percent Demented by Father’s Occupational Status. 0 10 20 30 40 50 60 53 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 Height (inches) % Demented Figure 3.6. Percent Demented by Height in Inches. 57 0 10 20 30 40 50 60 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 25 26 32 # Missing Teeth % Demented Figure 3.7. Percent Demented by Number of Missing Teeth. 58 CHAPTER IV: RESULTS FOR HYPOTHESIS 1 Hypothesis 1: A greater number of teeth lost will increase risk for dementia Case Control Analysis Strategy We used logistic regression models to determine whether greater tooth loss increases the risk of dementia. We estimated the odds ratio (OR) and 95% confidence intervals (CIs) for each individual risk factor. Then all risk factors were included simultaneously in a multivariate model. Standard case control analyses were used to compare dementia cases with biologically unrelated controls to be able to compare the results from this sample with other case-control studies. Because members of twin pairs were included in the analyses in the same way as single participants, confidence intervals were corrected using robust standard errors. This was accomplished by using a macro in SAS that increases variance estimates in proportion to the degree of correlation between twin pairs making the confidence intervals more conservative (Lin, 1994; Moradi, Adami, Ekborm, Wedren, Terry, Floderus, et al., 2002). We first calculated ORs for variables measured continuously and then calculated the odds for each variable after division into discrete categories as described in chapter 3. We included age, zygosity, education, father’s occupational prestige, height, and number of missing teeth in the models. For age, the inverse of birth year was used. See Table 4.1 for group means and frequencies of the variables included in the continuous analysis. See Table 4.2 for a description of the number of individuals exposed or unexposed to the 59 categorical risk factors among cases and controls. The same tables for the MZ twins only can be found at the end of the chapter (Tables 4.16 to 4.17). Table 4.1. Group Means & Frequencies for Risk Factors in Case-Control Sample Dementia No Dementia (N=371) (N=5,981) Variable M (SD) M (SD) Birth year 1922.18 (2.96) 1923.58 (2.76) Education 13.20 (3.31) 13.77 (3.08) Prestige 42.54 (11.48) 42.01 (11.70) Height 68.04 (2.64) 68.11 (2.56) Teeth lost 3.64 (2.67) 3.82 (2.88) N (%) N (%) Monozygotic 201 (54.2) 3,097 (51.8) Table 4.2 Numbers of Cases and Controls Exposed or Unexposed to Risk Factors Variable Demented (N=371) Non- demented (N = 5,981) Exposed N Unexposed N Missing N Exposed N Unexposed N Missing N Low education 84 287 0 878 5,097 6 Low father’s occupational prestige 25 253 93 419 4,405 1,157 Short adult height (<69 inches) 276 95 0 1,729 4,236 16 Missing ½ or more of all teeth 3 328 40 37 5,230 714 Zygosity MZ (vs. DZ) 201 170 0 3,097 2,884 0 Among all cases of dementia, those with less than a high school education had a younger average age of onset (M = 71.8, SD = 6.7) than dementia cases who completed high school or more education (M = 73.5, SD = 6.0), t = 2.19, p 60 =.03. Among all individuals with dementia, completing less than a high school education predicted, on average, 1.7 years earlier onset (β = -1.68, p = .03). Case Control Results for Total Dementia Table 4.3 summarizes the results for the case-control analyses in total dementia using the variables measured continuously. In univariate models (model 1), older age and lower levels of education predicted increased risk of dementia. Zygosity, father’s occupational prestige, height, and number of missing teeth failed to predict the risk of dementia at the level of statistical significance in the univariate models. In the next set of models (models 2-4), the effects of father’s occupational prestige, height, and missing teeth were each tested after accounting for age, zygosity, and education. In model 2, the addition of father’s occupational prestige made education non-significant, suggesting that the two variables share variance. The effects of height and number of missing teeth were unchanged. When these were all put together in one model (model 5), only age remained a significant predictor. Table 4.4 summarizes the results for the case-control analyses in total dementia using the variables categorically. The pattern of results was similar to the continuous models. In univariate models (model 1), older age and completing fewer than 12 years of education predicted increased risk of dementia. Zygosity, father’s occupational prestige, height, and number of missing teeth failed to predict the risk of dementia at the level of statistical significance in the univariate models. 61 Table 4.3 Case Control Analyses Predicting Risk for Total Dementia (CONTINUOUS) Model 1 Model 2 Model 3 Univariate Multivariate Multivariate OR (95%CI) OR (95%CI) OR (95%CI) Age 1.18 (1.14, 1.23)* 1.18 (1.14, 1.23)* 1.18 (1.14, 1.22)* Zygosity (MZ=1, DZ=2) 0.91 (0.74, 1.12) 0.90 (0.71, 1.16) 0.90 (0.72, 1.11) Education 0.94 (0.91, 0.98)* 0.96 (0.91, 1.00) 0.95 (0.91, 0.99)* Father’s occup prestige 1.00 (0.99, 1.01) 1.01 (1.00, 1.02) Height 0.99 (0.95, 1.03) 1.00 (0.95, 1.04) Missing Teeth 0.98 (0.94, 1.02) Model 4 Model 5 Multivariate Multivariate OR (95%CI) OR (95%CI) Age 1.20 (1.16, 1.25)* 1.20 (1.15, 1.26)* Zygosity (MZ=1, DZ=2) 0.92 (0.73, 1.15) 0.92 (0.70, 1.19) Education 0.95 (0.92, 0.99)* 0.96 (0.91, 1.01) Father’s occup prestige 1.01 (1.00, 1.02) Height 0.99 (0.93, 1.04) Missing Teeth 0.99 (0.95, 1.02) 0.99 (0.95, 1.03) Notes: 1. Confidence intervals adjusted for dependent observations using SAS macro 2. Odds ratios are significant where 95% CIs do not include 1.00 3. For age, inverse of birth year was used for ease of interpretation In the next set of models (models 2-4), the effects of father’s occupational prestige, height, and missing teeth were each tested after accounting for age, zygosity, and education. Father’s occupational prestige, height, and missing teeth were not significant predictors in these models. When these were all put together in one model (model 5), only age and low education remained significant. 62 Table 4.4 Case Control Analyses Predicting Risk for Total Dementia (CATEGORICAL) Model 1 Model 2 Model 3 Univariate Multivariate Multivariate OR (95%CI) OR (95%CI) OR (95%CI) Age 1.18 (1.14, 1.23)* 1.18 (1.14, 1.24)* 1.18 (1.14, 1.23)* Zygosity (MZ=1, DZ=2) 0.91 (0.74, 1.12) 0.90 (0.70, 1.15) 0.89 (0.72, 1.10) Low Education (<HS) 1.70 (1.32, 2.19)* 1.59 (1.16, 2.18)* 1.72 (1.33, 2.23)* Low Prestige (low 10%) 1.04 (0.68, 1.59) 0.96 (0.62, 1.48) Height <69 inches 1.19 (0.93, 1.51) 1.13 (0.88, 1.44) Missing > ½ of Teeth 1.29 (0.39, 4.34) Model 4 Model 5 Multivariate Multivariate OR (95%CI) OR (95%CI) Age 1.21 (1.16, 1.25)* 1.21 (1.15, 1.26)* Zygosity (MZ=1, DZ=2) 0.91 (0.73, 1.14) 0.90 (0.69, 1.18) Low Education (<HS) 1.66 (1.26, 2.19)* 1.53 (1.09, 2.15)* Low Prestige (low 10%) 0.94 (0.59, 1.49) Height < 69 inches 1.09 (0.81, 1.46) Missing > ½ of Teeth 0.99 (0.28, 3.58) 1.28 (0.34, 4.75) Notes: 1. Confidence intervals adjusted for dependent observations using SAS macro 2. Odds ratios are significant where 95% CIs do not include 1.00 3. For age, inverse of birth year was used for ease of interpretation Case Control Results for Alzheimer’s Disease Table 4.5 summarizes the results for the case-control analyses excluding cases of non-AD dementia using the variables measured as continuous. In univariate models (model 1), older age and lower levels of education predicted increased risk of AD. Zygosity, father’s occupational prestige, height, and number of missing teeth failed to predict the risk of AD at the level of statistical significance in the univariate models. In the next set of models (models 2-4), the effects of father’s occupational prestige, height, and missing teeth were each tested after accounting for age, zygosity, and education. Father’s occupational prestige, height, and missing 63 teeth were not significant predictors in these models. In model 2, the addition of father’s occupational prestige made education non-significant, suggesting that the two variables share variance. When these were all put together in one model (model 5), only age remained significant. Table 4.5. Case Control Analyses Predicting Risk for AD (CONTINUOUS) Model 1 Model 2 Model 3 Univariate Multivariate Multivariate OR (95%CI) OR (95%CI) OR (95%CI) Age 1.18 (1.13, 1.24)* 1.18 (1.12, 1.24)* 1.18 (1.13, 1.23)* Zygosity (MZ=1,DZ=2) 0.84 (0.65, 1.10) 0.80 (0.59, 1.09) 0.83 (0.64, 1.09) Education 0.95 (0.91, 0.99)* 0.96 (0.91, 1.02) 0.96 (0.91, 1.00) Father’s occup prestige 1.00 (0.90, 1.11) 1.00 (0.99, 1.02) Height 0.99 (0.94, 1.04) 1.00 (0.94, 1.05) Missing Teeth 0.97 (0.91, 1.03) Model 4 Model 5 Multivariate Multivariate OR (95%CI) OR (95%CI) Age 1.21 (1.15, 1.26)* 1.21 (1.15, 1.28)* Zygosity (MZ=1,DZ=2) 0.84 (0.63, 1.11) 0.78 (0.56, 1.08) Education 0.96 (0.91, 1.01) 0.97 (0.91, 1.03) Father’s occup prestige 1.01 (0.99, 1.02) Height 1.01 (0.94, 1.08) Missing Teeth 0.98 (0.93, 1.03) 0.99 (0.94, 1.05) Notes: 1. Confidence intervals adjusted for dependent observations using SAS macro 2. Odds ratios are significant where 95% CIs do not include 1.00 3. For age, inverse of birth year was used for ease of interpretation Table 4.6 summarizes the results for the case-control analyses excluding cases of non-AD dementia using the variables categorically. The pattern of results was similar to the continuous models. In univariate models (model 1), older age predicted increased risk of AD. Completing fewer than 12 years of education did not significantly predict increased risk of AD, though it was near the 64 threshold. Zygosity, father’s occupational prestige, height, and number of missing teeth failed to predict the risk of AD at the level of statistical significance in the univariate models. In the next set of models (models 2-4), the effects of father’s occupational prestige, height, and missing teeth were each tested after accounting for age, zygosity, and education. Father’s occupational prestige, height, and missing teeth were not significant predictors in these models. In model 3, education became significant when short height was included in the model, but father’s occupational prestige was not included. When these were all put together in one model (model 5), only age remained significant. Table 4.6. Case Control Analyses Predicting Risk for AD (CATEGORICAL) Model 1 Model 2 Model 3 Univariate Multivariate Multivariate OR (95%CI) OR (95%CI) OR (95%CI) Age 1.18 (1.13, 1.24)* 1.19 (1.13, 1.25)* 1.18 (1.13, 1.24)* Zygosity (MZ=1,DZ =2) 0.84 (0.65, 1.10) 0.80 (0.59, 1.09) 0.84 (0.64, 1.09) Low Education (< HS) 1.40 (1.00, 1.95) 1.42 (0.95, 2.12) 1.41 (1.01, 1.98)* Low prestige (low10%) 1.05 (0.62, 1.77) 0.97 (0.57, 1.65) Height <69 inches 1.25 (0.92, 1.70) 1.20 (0.88, 1.64) Missing > ½ of Teeth 2.05 (0.61, 6.90) Model 4 Model 5 Multivariate Multivariate OR (95%CI) OR (95%CI) Age 1.21 (1.16, 1.27)* 1.21 (1.15, 1.28)* Zygosity (MZ=1,DZ=2) 0.84 (0.63, 1.11) 0.77 (0.56, 1.08) Low Education (< HS) 1.34 (0.93, 1.92) 1.36 (0.89, 2.09) Low prestige (low10%) 0.94 (0.53, 1.65) Height <69 inches 0.97 (0.68, 1.39) Missing > ½ of Teeth 1.66 (0.46, 6.00) 2.11 (0.56, 7.86) Notes: 1. Confidence intervals adjusted for dependent observations using SAS macro 2. Odds ratios are significant where 95% CIs do not include 1.00 3. For age, inverse of birth year was used for ease of interpretation 65 Co-Twin Control Analysis Strategy Co-twin control analyses were conducted in which one twin served as the matched control for his co-twin. Pairs in which neither twin had dementia were not included in these analyses. The co-twin sample consists of 316 twin pairs in which at least one of the twins has dementia. The diagnosis of demented versus non-demented was used to define the case and control in the pair. In pairs where both twins were demented but onset was asynchronous, the first diagnosed was used to define the case and the second diagnosed as the control. See Table 4.7 for group means and frequencies of the variables included in the continuous analysis. See Table 4.8 for a description of the number of individuals exposed or unexposed to the categorical risk factors among cases and controls. The proband is defined as the demented twin in discordant pairs and the first twin demented in concordant pairs. A description of the occurrence of the proband having a higher rate of the risk variable than his co-twin can be found in Table 4.9. The same tables for the MZ twins only can be found at the end of the chapter (Tables 4.18 to 4.20). Table 4.7. Group Means for Risk Factors for Co-twin Control Sample Discordant for dementia Concordant Dementia No Dementia 1 st demented 2 nd demented (N =261) (N = 261) (N=55) (N=55) Education 13.30 (3.39) 13.35 (3.27) 12.80 (2.93) 12.86 (3.36) Height 68.11 (2.39) 67.98 (2.63) 68.39 (2.77) 68.47 (2.80) Missing Teeth 3.93 (2.69) 4.06 (2.77) 2.81 (2.04) 3.59 (2.90) Note: Birth year, father’s occupational prestige, and zygosity not included in co-twin analyses because they are identical within a twin pair 66 The majority of twin pairs were concordant for educational attainment category (more or less than high school). In pairs in which both twins had dementia, 91% were in the same educational attainment category. In pairs where only one twin had dementia, 84% were in the same category of educational attainment. This difference was not statistically significant. In pairs in which both twins had dementia, but differed in educational status (N=5), the proband (first demented) had the lower level of education 80% of the time (N=4). In pairs in which only one twin had dementia, but differed in educational status (N = 35), 43% of those with low education were the proband (demented twin) (N = 15). This difference was not statistically significant. The mean number of years of education in dementia concordant pairs was 12.60 (SD = 3.1). The mean number of years of education in the demented twin in dementia discordant pairs was 13.46 (SD = 3.4). In dementia discordant pairs, the average age of dementia onset was earlier (M = 72.7, SD = 6.3) than the age of onset in dementia concordant pairs (M = 74.0, SD = 6.0) t = 1.95, p = .05. A regression using low educational status to predict age of onset in dementia concordant pairs showed that twins with less than a high school education had, on average, 2.5 years earlier onset than twins who completed high school (β = -2.45, p = .05). 67 Table 4.8. Numbers of Pairs Exposed or Unexposed to Risk Factors within Dementia-concordant and Dementia-discordant Twin Pairs Risk Variable Pairs Discordant for Dementia (N=261) Pairs Concordant for Dementia (N = 55) Discordant for Exposure concordant for exposure N missing N discordant for exposure Concor- dant for exposure N missing N proband exposed N total exposed N proband exposed N total exposed N Low Education 15 35 220 6 4 5 50 0 Father’s Occup Prestige 0 0 261 0 0 0 55 0 Short Height < 69 inches 38 24 199 0 3 4 48 0 > ½ of teeth missing 0 1 211 49 0 0 49 6 >12 teeth missing 2 3 207 49 0 0 49 6 Notes: 1. Proband is defined as the demented twin in discordant pairs or the first twin demented in concordant pairs 2. Father’s occupational prestige is identical within all twin pairs. It is therefore not included in the co-twin control analyses. Table 4.9. Comparative Risk Variables Occurring in Proband versus Control Twin Comparative Risk Variable Pairs Discordant for Dementia (N=261) Pairs Concordant for Dementia (N = 55) Discordant for Exposure Missing data or identical within a pair N Discordant for Exposure Missing data or identical within a pair N Proband Exposed N Control Exposed N Proband Exposed N Control Exposed N Lower education than co-twin 82 77 102 21 14 20 Shorter height than co-twin 88 97 76 18 15 22 Missing more teeth than co-twin 77 87 97 19 22 14 68 As in the case control analyses, number of teeth lost was used to predict dementia diagnosis after adjusting for several covariates. We used conditional logistic regression to obtain hazard ratios (HRs) with twin pair as the stratum (PROC PHREG in SAS) to compare each dementia case to his co-twin. We estimated the HR and 95% CIs for each individual risk factor. Then, all predictors were included simultaneously in a multivariate model. We first calculated HRs for variables measured continuously. We included education, height, and missing teeth in the models. As father’s occupational prestige, birth year, and zygosity are identical for twins, these variables were not included in the analyses. In categorical analyses, education was divided into two categories corresponding to less than 12 years of education, or 12 or more years of education. Height was divided into two categories separating those less than 69 inches tall from those 69 inches or taller. As the number of individuals missing greater than half of their teeth was not great enough to run co-twin models, tooth loss was divided into two categories corresponding to individuals who had lost 12 or more of their teeth versus fewer than 12. A comparative risk approach was also used. That is, the twin with more of each risk factor was compared to his co-twin with less of that risk factor. In this case, twins with identical amounts of the risk factors or with missing data were excluded from the analyses. Variables were constructed to indicate whether a twin had less education than his co-twin, shorter height than his co-twin, or was missing more or fewer teeth than his co-twin. Pairs with identical values for education, height, and missing teeth were not included in these analyses. 69 Co-Twin Control Results for Total Dementia Table 4.10 summarizes the results for the co-twin control analysis in total dementia using the variables measured as continuous. In univariate and multivariate models, education, height, and number of missing teeth failed to predict dementia. Table 4.11 summarizes the results for the co-twin control analysis in total dementia using the variables measured as categorical. In univariate and multivariate models, education, height, and number of missing teeth failed to predict dementia. Table 4.12 summarizes the results for the co-twin control analysis in total dementia using comparative risk. None of the predictors was significant in univariate or multivariate models. The same analyses were conducted after excluding pairs concordant for dementia with the same results (see Tables 4.21 through 4.23 at the end of the chapter). Co-Twin Control Results for AD Table 4.13 summarizes the results for the co-twin control analysis in AD using the variables measured as continuous. In univariate and multivariate models, education, height, and number of missing teeth failed to predict dementia. Table 4.14 summarizes the results for the co-twin control analysis in AD using the variables measured as categorical. In univariate and multivariate models, education, height, and number of missing teeth failed to predict AD. 70 Co-twin Control Analyses (using pair number as strata) Predicting Total Dementia (N=632/ 316 pairs) 4.10 CONTINUOUS Model 1 Model 2 (N=556) Univariate Multivariate HR (95%CI) HR (95%CI) Education 0.90 (0.79, 1.02) 0.94 (0.82, 1.07) Height 1.13 (0.91, 1.41) 1.12 (0.86, 1.46) Missing Teeth 1.05 (0.90, 1.23) 1.06 (0.90, 1.24) 4.11 CATEGORICAL Model 1 Model 2 (N=556) Univariate Multivariate HR (95%CI) HR (95%CI) Less than high school 1.80 (0.60, 5.37) 1.23 (0.37, 4.10) Height < 69 inches 0.50 (0.20, 1.24) 0.39 (0.12, 1.26) Missing >12 teeth 1.00 (0.14, 7.10) 1.33 (0.17, 10.18) 4.12 COMPARATIVE RISK Model 1 Model 2 (N=174) Univariate Multivariate HR (95%CI) HR (95%CI) Twin with more/less education 0.87 (0.55, 1.38) 0.94 (0.43, 2.03) Twin with shorter/taller height 1.03 (0.66, 1.59) 0.61 (0.26, 1.42) Twin with more/less teeth 1.25 (0.79, 1.99) 0.84 (0.36, 1.93) Notes: 1. Hazard ratios are significant where 95% CIs do not span above and below 1.00 Table 4.15 summarizes the results for the co-twin control analysis in AD using comparative risk. None of the predictors was significant in univariate or multivariate models. The same analyses were conducted after excluding pairs concordant for dementia with the same results (see Tables 4.24 through 4.26 at the end of the chapter). 71 Co-twin Control Analyses (using pair number as strata) Predicting AD (N=384/ 192 pairs) 4.13 CONTINUOUS Model 1 Model 2 (N=342) Univariate Multivariate HR (95%CI) HR (95%CI) Education 1.01 (0.83, 1.22) 1.03 (0.84, 1.26) Height 1.18 (0.86, 1.62) 1.14 (0.78, 1.68) Missing Teeth 0.97 (0.79, 1.19) 0.98 (0.80, 1.20) 4.14 CATEGORICAL Model 1 Model 2 (N=342) Univariate Multivariate HR (95%CI) HR (95%CI) Less than high school 1.00 (0.14, 7.10) 0.43 (0.03, 6.21) Height < 69 inches 0.22 (0.05, 1.03) 0.29 (0.06, 1.38) Missing >12 teeth 1.00 (0.06, 15.99) 0.66 (0.03, 15.05) 4.15 COMPARATIVE RISK Model 1 Model 2 (N=119) Univariate Multivariate HR (95%CI) HR (95%CI) Twin with more/less education 0.81 (0.43, 1.53) 0.79 (0.35, 3.23) Twin with shorter/taller height 1.60 (0.84, 3.05) 1.07 (0.35, 3.23) Twin with more/less teeth 1.11 (0.59, 2.10) 0.83 (0.29, 2.37) Notes: 1. Hazard ratios are significant where 95% CIs do not span above and below 1.00 Power As the sizes in the group missing half of more of their teeth (N=40) and missing less than half of their teeth (N=5,556) are dramatically different, the power to test our hypothesis was reduced. The concept of effective sample size states that a comparison between two groups with sample sizes x and y has equivalent power to a comparison of two groups with the sample size equal to the harmonic mean of x and y. The harmonic mean for our sample missing half or more of their teeth and the sample missing less than half of their teeth is 79.43 72 meaning that our sample has equivalent power to a study with 79 people in each group, or a total sample size of 158. Using the effective sample size figures in a power estimating calculator for logistic regression adjusting for covariates (Tosteson, Buzas, Demidenko, & Karagas, 2003) suggests the present sample has power of .39 if there is no measurement error and .17 if there is measurement error. We would need a sample of at least 391 persons missing half or more of their teeth and no additional persons missing less than half of their teeth to accomplish 90% power with measurement error. This estimate is limited for several reasons. First, the predictor variable is dichotomous and the power estimate formula is based on continuous predictors. Second, this estimate does not adjust for non-independence of our sample. Finally, this estimate is far greater than the sample sizes used in other studies of tooth loss because the proportion of the population in which tooth loss was observed was much higher. It is based on the expectation that in this sample tooth loss occurs at a rate of less than 1%. 73 Case Control Tables in MZ Pairs Only Table 4.16. Group Means & Frequencies for Risk Factors in Case-Control Sample in MZ Twins Only Dementia No Dementia (N=371) (N=5,981) Variable M (SD) M (SD) Birth year 1922.06 (2.83) 1923.58 (2.75) Education 13.27 (3.24) 13.87 (3.01) Prestige 41.41 (10.07) 42.22 (11.76) Height 67.89 (2.78) 67.96 (2.52) Missing teeth 3.86 (2.76) 3.81 (2.77) Table 4.17 Numbers of Cases and Controls Exposed or Unexposed to Risk Factors in MZ Twins Only Risk Variable Demented (N=201) Non- demented (N =3.097 ) Exposed N Unexposed N Missing N Exposed N Unexposed N Missing N Low education 43 158 0 418 2,675 4 Low Father’s occupational prestige 14 138 49 208 2,332 557 Short adult height (<69 inches) 147 54 0 2265 824 8 Half or more of teeth missing 2 175 24 14 2,714 369 74 Co-Twin Control Tables in MZ Pairs Only Table 4.18. Group Means for Risk Factors for Co-Twin Control Sample in MZ Twins Only Discordant for dementia Concordant Dementia No Dementia 1 st 2 nd (N =131) (N = 131) (N=36) (N=36) Education 13.46 (3.14) 13.69 (3.37) 12.73 (3.11) 12.72 (2.98) Height 67.79 (2.46) 67.73 (2.70) 68.11 (2.98) 68.31 (2.82) Missing Teeth 4.16 (2.79) 4.23 (2.68) 2.55 (2.14) 3.56 (3.21) Note: Birth year, prestige, and zygosity not included in co-twin analyses because they are identical within a twin pair Table 4.19. Numbers of Pairs Exposed or Unexposed to Risk Factors within Dementia- concordant and Dementia-discordant Twin Pairs in MZ Twins Only Risk Variable Pairs Discordant for Dementia (N=129) Pairs Concordant for Dementia (N = 36) Discordant for Exposure Concor- dant for Exposure N Missing N Discordant for Exposure Concor- dant for Exposure N Missing N Proband Exposed N Total Exposed N Proband Exposed N Total Exposed N Low Education 5 9 116 4 1 2 34 0 Short Height (<69 inches) 9 4 116 0 0 2 34 0 > ½ of teeth missing 0 0 104 25 0 0 30 6 > 12 missing teeth 1 0 103 25 0 0 30 6 Notes: 1. Proband is defined as the demented twin in discordant pairs or the first twin demented in concordant pairs 2. Father’s occupational prestige is identical within all twin pairs. It is therefore not included in the co-twin control analyses. 75 Table 4.20 Comparative Risk Variables Occurring in Proband versus Control Twin in MZ Twins Only Comparative Risk Variable Pairs Discordant for Dementia (N=129) Pairs Concordant for Dementia (N = 36) Discordant for Exposure Missing data or identical within a pair N Discordant for Exposure Missing data or identical within a pair N Proband Exposed N Control Exposed N Proband Exposed N Control Exposed N Lower Education than co-twin 34 30 65 14 6 16 Shorter height than co-twin 37 32 60 13 7 16 Missing more teeth than co-twin 40 38 51 10 14 12 Co-twin Control Analyses (using pair number as strata) Predicting Total Dementia in Discordant Pairs Only (N=522/ 261 pairs) 4.21 CONTINUOUS Model 1 Model 2 Univariate Multivariate HR (95%CI) HR (95%CI) Education 0.90 (0.79, 1.02) 0.94 (0.82, 1.07) Height 1.13 (0.91, 1.41) 1.12 (0.86, 1.46) Missing Teeth 1.05 (0.90, 1.23) 1.06 (0.90, 1.24) 4.22 CATEGORICAL Model 1 Model 2 Univariate Multivariate HR (95%CI) HR (95%CI) Less than high school 1.80 (0.60, 5.37) 1.23 (0.37, 4.10) Height < 69 inches 0.50 (0.20, 1.24) 0.39 (0.12, 1.26) Missing >12 teeth 1.00 (0.14, 7.10) 1.33 (0.17, 10.18) 4.23 COMPARATIVE RISK Model 1 Model 2 Univariate Multivariate HR (95%CI) HR (95%CI) Twin with more/less education 0.87 (0.55, 1.38) 0.94 (0.43, 2.03) Twin with shorter/taller height 1.03 (0.66, 1.59) 0.61 (0.26, 1.42) Twin missing more/less teeth 1.25 (0.79, 1.99) 0.84 (0.36, 1.93) Notes: 1. Hazard ratios are significant where 95% CIs do not span above and below 1.00 76 Co-twin Control Analyses (using pair number as strata) Predicting AD in Discordant Pairs Only (N=304/ 152 pairs) 4.24 CONTINUOUS Model 1 Model 2 Univariate Multivariate HR (95%CI) HR (95%CI) Education 1.01 (0.83, 1.22) 1.03 (0.84, 1.26) Height 1.18 (0.86, 1.62) 1.14 (0.78, 1.68) Missing Teeth 0.97 (0.79, 1.19) 0.98 (0.80, 1.20) 4.25 CATEGORICAL Model 1 Model 2 Univariate Multivariate HR (95%CI) HR (95%CI) Less than high school 1.00 (0.14, 7.10) 0.43 (0.03, 6.21) Height < 69 inches 0.22 (0.05, 1.03) 0.29 (0.06, 1.38) Missing >12 teeth 1.00 (0.06, 15.99) 0.66 (0.03, 15.05) 4.26 COMPARATIVE RISK Model 1 Model 2 Univariate Multivariate HR (95%CI) HR (95%CI) Twin with more/less education 0.81 (0.43, 1.53) 0.79 (0.26, 2.46) Twin with shorter/taller height 1.60 (0.84, 3.05) 1.07 (0.35, 3.23) Twin with more/less teeth 1.11 (0.59, 2.10) 0.83 (0.29, 2.37) Notes: 1. Hazard ratios are significant where 95% CIs do not span above and below 1.00 77 CHAPTER V: RESULTS FOR HYPOTHESIS 2 Hypothesis 2: A greater number of teeth lost in young adulthood will predict lower scores on the telephone interview for cognitive status (TICS-m) in older adulthood and greater decline in the TICS-m score. To test hypothesis 2, we used mixed effects modeling and growth curve modeling. Table 5.1 summarizes the mean TICS-m scores and ages at occasion of measurement. Figures 5.1 and 5.2 graphically illustrate the mean TICS-m scores over occasion and over age respectively. For mean scores over occasion, note that individuals participated in different numbers of waves. In the graph over age, individuals did not participate every year, but data are included for each age at which they participated (at up to four occasions). Table 5.1 Mean TICS-m Scores and Ages at Each Wave of Measurement Non-demented Only TICS-m score Age at Time of Measurement N M (SD) M (SD) Wave 1 5,581 34.0 (4.0) 66.0 (2.8) Wave 2 5,589 33.7 (4.7) 69.0 (2.8) Wave 3 4,629 34.3 (4.7) 72.6 (2.7) Wave 4 3,426 32.9 (4.7) 76.9 (2.7) Total Sample TICS-m score Age at Time of Measurement N M (SD) M (SD) Wave 1 6,126 33.8 (4.2) 66.1 (2.8) Wave 2 6,047 33.5 (4.9) 69.1 (2.8) Wave 3 4,980 34.0 (4.8) 72.6 (2.8) Wave 4 3,586 32.7 (4.9) 77.0 (2.7) TICS-m = Modified Telephone Interview for Cognitive Status 78 Mean TICS-m Scores over Occasion of Measurement 32 32.5 33 33.5 34 34.5 1 2 3 4 Occasion of Measurement Mean TICS-m Score Figure 5.1 Mean TICS-m Scores over Occasions of Measurement in the Entire Sample Mean TICS-m Scores over Age 28 29 30 31 32 33 34 35 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 Age in years Mean TICS-m Score Figure 5.2 Mean TICS-m Scores over Age in the Entire Sample 79 Paired samples t-tests were used to determine whether pairs discordant for tooth loss differed significantly on baseline TICS-m score, change in TICS-m score between first and last available measurements (regardless of availability of the same wave for twin), baseline age at TICS-m testing, years of education, and height (see Table 5.2). These analyses were conducted excluding 1,246 pairs in the non-demented sample (1,357 pairs in the total sample) in which twins had identical numbers of teeth lost or when data were not available for both twins. None of the variables differed significantly within a twin pair discordant for tooth loss. Table 5.2 Paired-Samples T-Tests for Characteristics of Twin Pairs Discordant for Tooth Loss Non-demented Only Twin missing more teeth Twin missing fewer teeth N M (SD) M (SD) Baseline TICS-m score 1,601 33.6 (4.1) 33.8 (4.1) TICS-m change score 1,601 -1.0 (4.5) -1.0 (4.3) Baseline age 1,598 66.2 (2.9) 66.2 (2.9) Education (years) 1,614 13.4 (3.0) 13.5 (3.0) Height (inches) 1,611 68.0 (2.5) 68.0 (2.6) Total Sample Twin missing more teeth Twin missing fewer teeth N M (SD) M (SD) Baseline TICS-m score 1,785 33.3 (4.5) 33.5 (4.3) TICS-m change score 1,646 -1.2 (4.7) -1.2 (4.5) Baseline age 1,782 66.3 (2.9) 66.3 (2.9) Education (years) 1,814 13.4 (3.0) 13.5 (3.1) Height (inches) 1,816 68.0 (2.5) 68.1 (2.6) TICS-m = Modified Telephone Interview for Cognitive Status Note: Father’s occupational prestige not included because it is identical within a twin pair 80 Correlations between the predictor variables and TICS-m scores can be found in Table 5.3. Higher baseline TICS-m score was correlated with greater declines in TICS-m score, younger baseline age, higher levels of education and occupational prestige, taller height, and fewer missing teeth. Declines in TICS-m score were significantly correlated with higher baseline age and higher levels of education. The number of missing teeth was significantly negatively correlated with all variables except the change in TICS-m score. In the total sample, education was negatively correlated with baseline age, though this was not the case in the non-demented only sample. These correlations are not adjusted for non-independence in the sample. Table 5.3 Correlations between Predictor Variables and TICS-m Scores Non-demented Only (N=5,720) 1. 2. 3. 4. 5. 6. 7. 1. Baseline TICS-m score -- 2. TICS-m change score -.33*** -- 3. Baseline age -.08*** -.09*** -- 4. Education (years) .41*** .06*** -.02 -- 5. Father’s occupational prestige .14*** .03 -.02 .29*** -- 6. Height (inches) .10*** .02 .04** .18*** .11*** -- 7. Missing Teeth -.07*** -.01 -.06*** -.14*** -.07*** -.08*** -- Total Sample (N=6,292) 1. Baseline TICS-m score -- 2. TICS-m change score -.31*** -- 3. Baseline age -.12*** -.10*** -- 4. Education (years) .40*** .06*** -.03** -- 5. Father’s occupational prestige .12*** .03* -.01 .29*** -- 6. Height (inches) .09*** .02 .04** .17*** .10*** -- 7. Missing Teeth -.06*** -.01 -.07*** -.13*** -.08*** -.08*** -- TICS-m = Modified Telephone Interview for Cognitive Status * p < .05, ** p < .01, *** p <.001 81 Predicting Baseline TICS-m Score We used mixed effects models (PROC MIXED in SAS) to determine whether tooth loss in young adulthood predicts baseline TICS-m score in older adulthood. Non-independence of twin pairs is handled in this analysis by evaluating twins at the level of the pair (small cluster of 2) rather than the individual level. First, we estimated the effect of each predictor variable individually. Then all predictors were included simultaneously in a multivariate model. Table 5.4 summarizes the results of the analysis. In the univariate analyses (Model 1), all variables were significant predictors of baseline TICS-m score. Increasing age and a higher number of missing teeth predicted lower TICS-m scores. Higher levels of education and father’s occupational prestige and taller height in young adulthood predicted higher scores. Twins from dizygotic pairs had lower baseline TICS-m scores compared to twins from monozygotic pairs. If education is not included in the model (Model 2), the number of missing teeth remains significant adjusting for all other covariates. In a full multivariate model (Model 3), baseline age, zygosity, education, and height remain significant predictors. It is possible that in models 2 and 3 we are over-controlling for education given that it is highly correlated with baseline TICS-m score (R = .41, p < .0001). However, it is also possible that missing teeth is an indicator of low education rather than an independent predictor of low cognitive performance. Missing teeth is negatively correlated with education (R = -.13, p < .001). 82 Table 5.4 Mixed Effects Model Results Predicting Baseline TICS-m Score in Non-demented Twins (N = 4,023) Model 1 Model 2 Model 3 Univariate Multivariate Multivariate Unstd B (SE) Unstd B (SE) Unstd B (SE) Baseline age -.11 (.02)* -.12 (.02)* -.11 (.02)* Zygosity (MZ=1, DZ=2) -.28 (.12)* -.42 (.14)* -.31 (.12)* Education .51 (.02)* .49 (.02)* Father’s occupational prestige .05 (.01)* .04 (.01)* .01 (.01) Height .16 (.02)* .14 (.02)* .06 (.02)* Missing teeth -.09 (.02)* -.07 (.02)* -.02 (.02) Note: Using first TICS score available regardless of twin availability at same wave We ran the same analyses including the entire sample (both demented and non-demented individuals) finding the same pattern of results. See Table 5.5 Table 5.5 Mixed Effects Model Results Predicting Baseline TICS-m Score in Entire Sample (N = 6,288) Model 1 Model 2 Model 3 Univariate Multivariate Multivariate Unstd B (SE) Unstd B (SE) Unstd B (SE) Baseline age -.19 (.02)* -.21 (.02)* -.18 (.02)* Zygosity (MZ=1, DZ=2) -.31 (.13)* -.40 (.14)* -.30 (.13)* Education .53 (.02)* .51 (.02)* Father’s occupational prestige .04 (.01)* .04 (.01)* .00 (.01) Height .15 (.02)* .13 (.03)* .05 (.02)* Missing teeth -.08 (.02)* -.07 (.02)* -.03 (.02) Note: Using first TICS score available regardless of twin availability at same wave Predicting Change in TICS-m Score We considered change in TICS-m scores in two different ways. We first used mixed effects modeling to predict a change score calculated across two waves of measurement. Then we used growth curve modeling considering the change in TICS-m over age instead of over waves. For mixed effects analysis predicting a TICS-m change score, waves of TICS-m data were selected for the 83 occasions on which both twins in a pair had available data. Data from the measurement occasion that maximized the number of years between the baseline score and most recent follow up score were used (see Table 5.6). We calculated the change in TICS-m score by subtracting the baseline score from the follow up score most recently available for both twins. To account for differences in the number of years between baseline and follow up occasions of measurement, we included a measure of this duration in the models. The mean number of years between baseline and follow up was 7.7 years (SD = 3.3). Pairs that did not have two waves of scores in common (N=30) were not included in the analysis. Table 5.6 Waves of TICS-m Data Selected within Twin Pairs among Non-demented Duration Baseline Follow up N Total N 3 waves (9-12 years) Wave 1 Wave 4 1,257 1,257 2 waves (4-8 years) Wave 1 Wave 3 863 (6-9 years) Wave 2 Wave 4 62 925 1 wave (1-5 years) Wave 1 Wave 2 599 (1-5 years) Wave 2 Wave 3 42 (3-6 years) Wave 3 Wave 4 7 648 No 2 waves in common within a twin pair 30 Total pairs 2,860 ________________________________________________________________ We used mixed effects models (PROC MIXED in SAS) to determine whether tooth loss predicts decline in the TICS-m score from baseline to follow up. First, we estimated the effect of each predictor variable individually. Then all predictors were included simultaneously in a multivariate model. Table 5.7 84 summarizes the results of the analysis. In the univariate (Model 1) and multivariate models (Model 3), baseline age, baseline TICS-m score, duration between baseline and follow up, and education significantly predicted change in TICS-m score. Increasing age and duration between waves predicted larger declines in TICS-m score. A higher baseline score also predicted greater declines. Higher levels of education predicted smaller declines. Number of missing teeth significantly predicted change in TICS-m when adjusting only for baseline age, baseline TICS-m and duration between waves (Model 2). Adding education and other variables to the model eliminated any significant contribution by the number of missing teeth (Model 3). Table 5.7 Mixed Effect Model Results Predicting Change in TICS-m (N = 4,023) Model 1 Model 2 Model 3 Univariate Multivariate Multivariate Unstd B (SE) Unstd B (SE) Unstd B (SE) Baseline age -.13 (.02)* -.20 (.02)* -.19 (.02)* Baseline TICS-m -.39 (.01)* -.39 (.01)* -.50 (.02)* Duration between baseline & follow up -.11 (.02)* -.08 (.02)* -.11 (.02)* Zygosity (DZ is ref group) .22 (.12) -.03 (.13) Education .07 (.02)* .35 (.02)* Father’s occupational prestige .01 (.01) .01 (.01) Height .03 (.02) .02 (.02) Missing teeth -.01 (.02) -.06 (.02)* -.01 (.02) Disregarding the availability of data for matched pairs, we conducted mixed models using all the available data. We predicted change in TICS-m score between waves 1 and 2, waves 2 and 3, and waves 3 and 4 (see Tables 5.8, 5.9, & 5.10). In the univariate (Model 1) and multivariate models (Model 3), baseline age, baseline TICS-m score, and education significantly predicted change in 85 TICS-m score from wave 1 to wave 2. Older age and higher baseline scores predicted larger declines in TICS-m score. Higher levels of education predicted smaller declines. Number of missing teeth significantly predicted change in TICS- m when adjusting only for baseline age and baseline TICS-m (Model 2). Adding education and other variables to the model eliminated any significant contribution by the number of missing teeth (Model 3). In models predicting the change in TICS-m between waves 2 and 3, only baseline age was a significant predictor (See Table 5.9). The same was true for change between waves 3 and 4 (See Table 5.10). Table 5.8. Mixed Effect Model Results Predicting Change in TICS-m from wave 1 to wave 2 in Non-demented Pairs Model 1 Model 2 Model 3 Univariate Multivariate Multivariate Unstd B (SE) Unstd B (SE) Unstd B (SE) Baseline age -.08 (.02)* -.11 (.02)* -.10 (.02)* Baseline TICS-m -.35 (.01)* -.36 (.01)* -.48 (.02)* Zygosity (DZ is ref group) -.17 (.11) -.06 (.12) Education .11 (.02)* .35 (.02)* Father’s occupational prestige .01 (.01) .00 (.01) Height .03 (.02) .03 (.02) Missing teeth -.02 (.02) -.06 (.02)* .00 (.02) Table 5.9. Mixed Effect Model Results Predicting Change in TICS-m from wave 2 to wave 3 in Non-demented Pairs. Model 1 Model 2 Model 3 Univariate Multivariate Multivariate Unstd B (SE) Unstd B (SE) Unstd B (SE) Baseline age -.06 (.02)* -.06 (.02)* -.05 (.03)* Baseline TICS-m -.02 (.02) -.02 (.02) -.02 (.02) Zygosity (DZ is ref group) .17 (.13) -.08 (.16) Education .03 (.02) .02 (.03) Father’s occupational prestige .01 (.01) .01 (.01) Height .02 (.02) .00 (.03) Missing teeth -.03 (.02) -.03 (.02) -.03 (.03) Note: Using TICS scores available regardless of twin availability at same wave 86 Table 5.10. Mixed Effect Model Results Predicting Change in TICS-m from wave 3 to wave 4 in Non-demented Pairs. Model 1 Model 2 Model 3 Univariate Multivariate Multivariate Unstd B (SE) Unstd B (SE) Unstd B (SE) Baseline age -.09 (.03)* -.11 (.03)* -.13 (.03)* Baseline TICS-m -.03 (.02) -.03 (.02) .01 (.03) Zygosity (DZ is ref group) -.23 (.15) .13 (.18) Education -.01 (.02) .01 (.03) Father’s occupational prestige -.00 (.01) -.02 (.01) Height -.04 (.03) -.01 (.04) Missing teeth .05 (.03) .04 (.03) .03 (.03) Note: Using TICS scores available regardless of twin availability at same wave We repeated the analyses including the entire sample (both demented and non-demented individuals) finding the same pattern of results. See Tables 5.11, 5.12, and 5.13.The one exception is that father’s occupational prestige is a significant univariate predictor of TICS-m scores when including individuals who became demented (in Table 5.11), whereas it is not significant when those with dementia were excluded (in Table 5.8). Table 5.11 Mixed Effect Model Results Predicting Change in TICS-m from wave 1 to wave 2 in Entire Sample Model 1 Model 2 Model 3 Univariate Multivariate Multivariate Unstd B (SE) Unstd B (SE) Unstd B (SE) Baseline age -.09 (.02)* -.14 (.02)* -.12 (.02)* Baseline TICS-m -.34 (.01)* -.35 (.01)* -.47 (.02)* Zygosity (DZ is ref group) -.17 (.11) -.00 (.12) Education .11 (.02)* .34 (.02)* Father’s occupational prestige .01 (.01)* .01 (.01) Height .04 (.02) .03 (.02) Missing teeth -.01 (.02) -.05 (.02)* .01 (.02) Note: Using TICS scores available regardless of twin availability at same wave 87 Table 5.12 Mixed Effect Model Results Predicting Change in TICS-m from wave 2 to wave 3 in Entire Sample Model 1 Model 2 Model 3 Univariate Multivariate Multivariate Unstd B (SE) Unstd B (SE) Unstd B (SE) Baseline age -.08 (.02)* -.08 (.02)* -.07 (.03)* Baseline TICS-m -.02 (.02) -.03 (.02) -.03 (.02) Zygosity (DZ is ref group) .15 (.13) -.10 (.16) Education .03 (.02) .01 (.03) Father’s occupational prestige .00 (.01) .00 (.01) Height .04 (.02) .03 (.03) Missing teeth -.03 (.02) -.04 (.02) -.04 (.03) Note: Using TICS scores available regardless of twin availability at same wave Table 5.13 Mixed Effect Model Results Predicting Change in TICS-m from wave 3 to wave 4 in Entire Sample Model 1 Model 2 Model 3 Univariate Multivariate Multivariate Unstd B (SE) Unstd B (SE) Unstd B (SE) Baseline age -.12 (.03)* -.13 (.03)* -.15 (.03)* Baseline TICS-m -.04 (.02) -.03 (.02) .02 (.02) Zygosity (DZ is ref group) -.20 (.15) .10 (.18) Education -.01 (.02) .01 (.03) Father’s occupational prestige -.00 (.01) -.00 (.01) Height -.05 (.03) -.02 (.04) Missing teeth .04 (.03) .03 (.03) .03 (.03) Note: Using TICS scores available regardless of twin availability at same wave Growth Curves Growth curves for the change in TICS-m scores were estimated (using Mplus, Muthen & Muthen, 2007) across 2-year age bins ranging from 62 years to 85 years. This approach allowed us to describe the slope of change in TICS-m scores across age, rather than estimating change between waves of measurement. We used full information maximum likelihood (FIML) estimation to make use of all the data. To eliminate the problem of non-independence in the 88 estimation of these models, we randomly selected one individual from each twin pair (N=3,176) using a random number generator. To describe the change in TICS-m scores, we compared three models (see Table 5.14). In this type of modeling, model fit is assessed by comparison to a baseline model. Reductions in the X 2 values and log likelihoods indicate improvements in fit per the number of degrees of freedom and free parameters respectively. We also use root mean squared error of approximation (RMSEA) for which a smaller number represents a better fit. For the comparative fit index (CFI) numbers greater than .90 are generally considered acceptable in fit (Byrne, 1998). First, a model predicting no change was estimated to establish a basis for comparison (X 2 (df) = 667 (75), RMSEA = .05). A linear model offered improvement from the no change model (X 2 (df) = 311 (72), RMSEA = .03). This model assumes that the amount of change is equivalent from year to year. Next, we estimated a quadratic model that allows for non-linear change. This model provided an improved fit over the linear model (X 2 (df) = 150 (68), RMSEA = .02) (see Figure 5.3). It does not appear to have much curvature due to the small, but significant quadratic term estimate (-0.06). The models were centered to minimize the correlation between level and slope. This does not affect model fit. 89 Table 5.14 Comparison of Growth Curve Model Fits over Age Bins No Change Linear Quadratic Estimate (SE) Estimate (SE) Estimate (SE) Mean level 33.47 (0.07) 34.49 (0.09) 33.44 (0.08) slope1 -- -3.18 (0.20) -0.35 (0.02) slope 2 -- -- -0.06 (0.01) Variance level 12.30 (0.40) 9.00 (0.60) 15.08 (0.55) slope1 -- 9.79 (3.14) 0.11 (0.03) slope2 -- -- 0.00 (0.00) error 10.13 (0.17) 9.40 (0.19) 8.92 (0.19) Covariance level, slope1 -- 4.01 (1.15) 0.42 (0.09) level, slope2 -- -- -0.12 (0.03) slope1, slope2 -- -- 0.01 (0.00) -2LL/ 29145 / 3 28967 / 6 28886 / 10 Parameters Д 2LL/ -- 178 / 3 259 / 7 Parameters X 2 / df 667 / 75 311 / 72 150 / 68 Д X 2 / df -- 356 / 3 517 / 7 CFI .82 .93 .98 RMSEA .05 .03 .02 Note: Basis weights for each model are centered to minimize covariance between level and slope 90 Linear & Quadratic Change in TICS-m Scores over Age Buckets -450 -400 -350 -300 -250 -200 -150 -100 -50 0 62 64 66 68 70 72 74 76 78 80 82 84 Age in Years Change in TICS-m Scores Linear Quadratic Figure 5.3 Linear and Quadratic Change in TICS-m Scores over Age After establishing the quadratic model as the best fitting description of change in TICS-m scores, we added our predictor variables to the model to determine whether they influenced the intercepts and slopes of the change in TICS-m scores over age (see Table 5.15). The predictor variables were centered on the mean, except for education which was centered on twelve years. First, we assessed the effect of tooth loss on the model. A greater number of teeth lost predicted a lower baseline TICS-m score (level). The number of teeth lost did not predict either the linear or quadratic slopes. In the next model we added level of education. The effect of tooth loss on the mean TICS-m score was diminished to below the level of significance. Education significantly predicted the mean TICS- m level, but did not predict either the linear or quadratic slopes. We then ran a model with all the predictors simultaneously (zygosity, education, father’s 91 occupational prestige, height, and missing teeth). The number of missing teeth was not a significant predictor of the baseline TICS-m score (level) after adjusting for all other predictors. Education continued to predict the level. Height also predicted the level, but not the slope. None of the other variables was a significant predictor of level or slope. Finally, in the mixed effects models, education predicted change between waves 1 and 2, but not between any other occasions. This may imply a practice effect is occurring in which test takers improve their scores after having had first exposure to a test. To test this concept, we ran growth curve models over the four occasions of measurement. From the shape of change illustrated in Figure 5.1, there does not appear to be an increase in scores between time 1 and time 2 in the overall sample. Several growth curves were fitted to change in TICS-m scores over waves. A piecewise spline model with a knot point at time 2 was estimated, but was unreliable because it had a negative slope variance. The best fitting was a piecewise spline model in which the change between times 3 and 4 were estimated as a piece separate from the changes between times 1, 2, and 3 (See Figure 5.4). See Table 5.16 for a summary of model fits. We added education as a predictor to the best fitting model. Education significantly predicted both level (Estimate = 0.69, standard error = 0.03) and slope (Estimate = 0.07, standard error (0.01). This is in contrast to the models over age in which education only predicted level but not slope. 92 Table 5.15 Predictors of the Quadratic Growth Curve Model over Age Bins Model1 Model 2 Model 3 Estimate (SE) Estimate (SE) Estimate (SE) Missing Teeth (level) -1.20 (0.31)* -0.39 (0.28) -0.33 (0.28) Missing Teeth (linear) 0.03 (0.06) 0.02 (0.06) 0.02 (0.06) Missing Teeth (quadratic) 0.01 (0.02) 0.01 (0.02) 0.01 (0.02) Education (level) -- 6.39 (0.24)* 6.17 (0.26)* Education (linear) -- 0.08 (0.06) 0.07 (0.07) Education (quadratic) -- -0.03 (0.02) -0.02 (0.02) Zygosity (level) -- -- 0.01 (0.15) Zygosity (linear) -- -- -0.01 (0.04) Zygosity (quadratic) -- -- -0.00 (0.01) Father’s occup prestige (level) -- -- 0.11 (0.07) Father’s occup prestige (linear) -- -- 0.01 (0.02) Father’s occup prestige (quad) -- -- -0.01 (0.01) Height (level) -- -- 0.73 (0.29)* Height (linear) -- -- -0.03 (0.07) Height (quadratic) -- -- -0.03 (0.02) Mean level 33.44 (0.08)* 32.32 (0.09)* 32.35 (0.11)* Mean slope (linear) -0.35 (0.02)* -0.36 (0.02)* -0.35 (0.03)* Mean slope (quadratic) -0.06 (0.01)* -0.05 (0.01)* -0.05 (0.01)* Variance level 14.97 (0.55)* 11.03 (0.44)* 10.97 (0.44)* Variance slope (linear) 0.11 (0.03)* 0.11 (0.03)* 0.11 (0.03)* Variance slope (quadratic) 0.00 (0.00) 0.00 (0.00) 0.00 (0.00) Covariance level, slope1 0.43 (0.09)* 0.38 (0.08)* 0.38 (0.08)* Covariance level, slope 2 -0.12 (0.03)* -0.10 (0.03)* -0.10 (0.03)* Covariance slope 1, slope 2 0.01 (0.00)* 0.01 (0.00)* 0.01 (0.00)* Error variance 8.92 (0.19)* 8.93 (0.19)* 8.93 (0.19)* -2LL/ parameters 29309 / 13 29676 / 16 35975 / 25 X 2 / df 157 / 77 168 / 86 201 / 113 CFI .98 .98 .98 RMSEA .02 .02 .02 93 Table 5.16 Comparison of Growth Curve Model Fits over Waves No Change Linear Quadratic Estimate (SE) Estimate (SE) Estimate (SE) -2LL/ parameters 29246 / 3 29107 / 6 29028 / 10 Change in 2LL/parameters -- 139 / 3 218 / 7 X 2 / df 566 / 11 289 / 8 132 / 4 Change in X 2 / df -- 277 / 3 434 / 7 CFI .84 .92 .96 RMSEA .13 .11 .10 Piecewise Latent Piecewise knot point time 2 knot point time 3 Estimate (SE) Estimate (SE) Estimate (SE) -2LL/ parameters 29056 / 10 29029 / 8 28996 / 10 Change in 2LL/parameters 190 / 7 217 / 5 250 / 7 X 2 / df 187 / 4 134 / 6 67 / 4 Change in X 2 / df 379 / 7 432 / 5 499 / 7 CFI .95 .96 .98 RMSEA .12 .08 .07 Piecewise Spline Model of TICS-m Change over Occasions of Measurement 31 31.5 32 32.5 33 33.5 34 Time 1 Time 2 Time 3 Time 4 TICS-m Score Figure 5.4. Piecewise Spline Model of TICS-m Scores over Occasions of Measurement 94 CHAPTER VI: DISCUSSION & CONCLUSION The purpose of the present research was to determine whether tooth loss in early adulthood predicts increased risk for dementia and cognitive decline in later life after accounting for possible covariates. Three previous studies reported that tooth loss in middle- and old-age predicted an increased risk for subsequent dementia. One study found that retrospective report of tooth loss prior to age 35 was a significant risk factor for dementia. As yet, few explanations have been posited for why tooth loss would increase dementia risk. In Chapter II, a theoretical model was proposed for how oral infection, as a major cause of tooth loss, may contribute to the development of dementia. The data analyses reported in Chapters IV and V were intended to test a small part of that model to determine whether a marker of oral infection, tooth loss, could predict dementia or declines in cognitive performance indicative of neural changes. Early Life Exposures Infection activates the immune system, including inflammatory processes, which places demands on the human body’s limited resources. Exposure to infection in early life, along with poor nutrition (Barker, 2004), has been hypothesized to divert resources away from growth and development resulting in shorter adult height (Crimmins & Finch, 2006; McDade, 2003). Furthermore, exposure to infection and inflammation over the lifetime may have a cumulative impact on morbidity and mortality (Bengtsson & Lindstrom, 2000; Costa, 2000; Finch & Crimmins, 2004; Georges, Rupprecht, Blankenberg, Poirier, Bickel, 95 Hafner, et al., 2003). As tooth loss often results from bacterial infection, it may indicate chronic exposure to infection and inflammation. For this reason, we suspected early life tooth loss might influence later risk of dementia. Adult height was included as a covariate to provide a possible indication of whether early life exposure to infection might be responsible for any relationship found between tooth loss and dementia. Exposure to infection and poor nutrition are more likely among people of low socioeconomic status, as is tooth loss (Borrell et al., 2006; Cunha-Cruz et al., 2007). For this reason, father’s occupational prestige and level of educational attainment were included as covariates. Main Results The first study aim was to evaluate the effect of tooth loss on the risk of dementia after accounting for the covariates. Tooth loss failed to predict total dementia and AD in both case-control and co-twin control analyses. Losing half or more of all teeth predicted odds ratios of greater than 1, but the confidence intervals were too large to obtain significance. Height did not lend predictive power in any of these models. The second study aim was to determine whether tooth loss predicts cognitive performance on the TICS-m at baseline and changes in performance after accounting for education and other covariates. We used both mixed models and growth curve models to test this hypothesis. Growth curve models estimate the change in TICS-m scores over ages of the participants, in contrast to mixed models which include change across occasions of measurement. The results of the two approaches were in agreement. Tooth loss did not predict baseline TICS- 96 m scores after accounting for other variables. Though tooth loss was predictive of baseline cognitive score in the univariate model, the addition of education to the model eliminated the effect, which suggests that tooth loss may be an indicator of low education in our sample. Greater numbers of teeth lost were correlated with lower levels of education, and education is highly correlated with TICS-m scores making it difficult to disentangle the independent contributions of tooth loss and education to cognitive performance and decline. The implications of education as a predictor of TICS-m performance will be discussed later. The change in TICS-m scores over age was best described by a quadratic growth model. This suggests that decline in mental status accelerates at later ages and is consistent with previous research on change in mental status (e.g., Lyketsos, Chen & Anthony, 1999; Wilson, Li, Aggarwal, Barnes, McCann, Gilley et al., 2004). Tooth loss did not predict the change in TICS-m scores. Shorter young adult height predicted lower TICS-m scores at baseline after accounting for all covariates including education and father’s occupational prestige, an indicator of childhood SES. This finding is consistent with other research suggesting that short adult height is associated with low educational attainment and poor adult cognitive performance (e.g. Abbott, White, Ross, Petrovich, Masaki, Snowden et al., 1998; Magnusson, Rasmussen, & Gyllensten, 2006). Height did not predict the change in TICS-m scores over age. Though previous studies demonstrated that tooth loss is a risk factor for dementia, the present study did not replicate this finding. There are several potential explanations for this. First, our sample had considerably fewer 97 individuals who had lost half or more of their teeth by the age of measurement (see Table 6.1), compared to previous studies (Gatz et al., 2006; Kim et al., 2007; Kondo et al., 1994; Stein et al., 2007). The low rate of tooth loss in our sample resulted in a failure to test the proposed hypothesis rather than a true failure to replicate results of previous studies. Health examinations required for enlistment in the military may have excluded those in poorest oral health from being included in the sample as men with the most severe, untreated dental disease may have been excluded (Asbell, 1988). For more information about the history of dental treatment in the military during the World War II era, refer to Appendix A. To have adequate statistical power to detect a significant effect (p <.05) in our case control analysis given our observed effect size (OR = 1.28) we would need approximately 391 cases who lost half or more of their teeth in the case control sample (Tosteson et al., 2003). Calculations of attributable risk (AR) allow an estimation of the proportion of cases among those exposed to the risk factor (i.e. tooth loss) that is due to that exposure (Greenland & Robins, 1988). The AR of the current and previous studies, found in Table 6.1, range from 2 to 20 percent. Calculation of an excess fraction (EF) describes the proportion of exposed cases that are in excess of the cases unexposed to the risk factor in question. The EF of the current and previous studies, found in Table 6.1, suggest that in two studies, the number of cases experiencing tooth loss did not occur in excess of the number of cases who did not experience tooth loss. When 98 dementia cases exposed to tooth loss occurred in excess of cases not exposed to tooth loss the amount of excess ranged from 16 to 31 percent. Table 6.1 Rates of Tooth Loss, Attributable Risk, and Excess Fraction in Studies of Tooth Loss and Dementia Missing >1/2 of Teeth Attributable Excess Risk Fraction Cases Controls Present Study 3 / 331 37 / 5,267 2% not in excess Swedish Twins 147 / 310 987 / 3,063 6% not in excess (Gatz et al., 2006) Milwaukee Nuns 21 / 32 57 / 112 10% 31% (Stein et al., 2007) Korean Community 33 / 57 241 / 629 6% 16% (Kim et al., 2007) Japanese Study 38 / 60 49 / 120 20% 27% (Kondo et al., 1994) Notes Kim et al. study divided teeth at 14 teeth or more rather than half (16). Attributable risk: number of cases exposed/total number exposed minus number of cases unexposed/total number unexposed. Excess Fraction: number of cases exposed/total number of cases minus number of cases unexposed/total number of cases. Second, previous studies of tooth loss as a risk factor for dementia measured tooth loss at older ages (see Table 1.1) compared to our study in which tooth loss was measured in young adulthood (mean age 20 years). Several studies have shown that increasing age is a risk factor for loss of periodontal attachment (Ismael, Morrison, Burt, Caffesse, & Kavanagh, 1990). The greater number of years a person has lived, the greater the opportunity to acquire oral disease. It is also possible that the risks of oral infection in young adulthood do not exert much influence over dementia risk, while oral infection in 99 older age may. The timing of exposure to oral infection may be of importance in its contribution to dementia risk. Third, it is possible that the causes of tooth loss in young adults differ substantially from the causes of tooth loss in older adulthood. Young men are particularly prone to higher risk taking that may result in accidents and injuries. As we do not have clear records on the causes of tooth loss in the present study, we are not able to ensure whether tooth loss was due to infection, malnutrition, head injury, or other causes. To truly test our hypotheses would require a larger number of individuals with tooth loss, varying ages at the time of tooth loss, and measures of the causes of tooth loss (i.e., infection, injury, etc.). To test the theoretical model proposed in Chapter II, it would also be advantageous to have a measure of chronic inflammation. Secondary Results Older age and low levels of education predicted increased risk of dementia in our study. Both have been established as risk factors for dementia in previous research (e.g., Cummings, Vinters, Cole, & Khachaturian, 1998; Hebert et al., 2003; Launer, Andersen, Dewey, Letenneur, Ott, Amaducci, et al., 1999). Low levels of education were also associated with an earlier onset of dementia. In the case-control sample, having less than a high school education was a significant risk factor for total dementia. As education measured continuously was not a significant predictor, this suggests educational attainment below a threshold is a risk factor. 100 Attaining less than a high school education did not predict increased risk of AD after accounting for other variables. It might be the case that education matters more in overall dementia than in AD specifically. For example, in a study of nuns, low education predicted risk for dementia, but was not associated with neuropathological criteria for AD in a sub-sample of brains autopsied (Mortimer, Snowdon, & Markesbery, 2003). However, this idea contradicts the conclusions of a meta-analysis of 19 studies (Caamano-Isorna, Corral, Montes-Martinez, & Takkouche, 2006) demonstrating that low education predicts AD (OR = 1.80, CIs 1.43-2.27) and total dementia (OR = 1.59, 1.26-2.01), but not non-AD dementia (OR = 1.32 CIs 0.92 - 1.88). Another study by Gatz et al. (2001) reported that low education predicted AD, but not total dementia in Swedish twins. Overall, research findings on the effects of education on dementia and AD have been quite mixed (Caamano-Isorna et al., 2006) and warrant further research. Education did not predict diagnosis of dementia in the co-twin sample when comparing individuals to their twins. That is, controlling for shared genetic and environmental influences, less education did not make a twin at greater risk for dementia than his co-twin. However, in pairs concordant for dementia, twins with less than a high school education had earlier onset of dementia by an average of 2.5 years. This suggests that education does not prevent the occurrence of dementia, but may delay onset or diagnosis in individuals with dementia. This supports the cognitive reserve hypothesis which suggests that higher education is related to delays in the onset of decline, though it is usually followed by a more rapid acceleration of decline after the onset or diagnosis of 101 the disease (Hall, Derby, LeValley, Katz, Verghese, & Lipton, 2007; Stern, Albert, Tang, & Tsai, 1999; Wilson, Li, Aggarwal, Barnes, McCann, Gilley, et al., 2004). In multivariate mixed effect models, baseline TICS-m scores were predicted by age, zygosity, and education. Age and education are well established predictors of cognitive performance (Schaie, 1993; Salthouse, 1999). DZ twins were more likely to have lower baseline TICS-m scores compared to MZ twins. Although studies have demonstrated that cognitive ability is highly genetically heritable (e.g., Pedersen, Plomin, Nesselroade, & McClearn, 1992), it is unclear why dizygotic twins would have lower scores than MZ twins despite their tendency to have greater variability from their co-twins. In multivariate mixed effect models, age, baseline TICS-m score, and education predicted changes in TICS-m score between the first and second waves of cognitive testing. Only age predicted change in TICS-m score between waves 2, 3, and 4. In the growth curve models, which estimate change over age rather than measurement occasions, education predicted the baseline TICS-m score, but not change. This may reflect a practice effect over waves of measurement whereby more highly educated individuals benefited more from previous exposure to the test, but did not show differences in change over age. Again, older age is a well established predictor of cognitive decline. Higher levels of education predicted smaller declines in TICS-m scores, while higher baseline TICS-m scores predicted steeper declines. This is likely due, at least in part, to floor effects. That is, those who started with low scores on the measure did not have room left to decline. 102 Confounded Contributions of Education, Tooth Loss, and TICS-m The effects of tooth loss on TICS-m performance were accounted for by education, suggesting that tooth loss may indicate low education in our sample. However, it is difficult to determine exactly what education indicates. Education is frequently used as a measure of SES. We included father’s occupational prestige as another indicator of SES. In some models, father’s occupational status seemed to account for the same variance as education causing the each to nullify the contribution of variance by the other. This suggests that education does, in part, represent SES. On the other hand, the finding that education is highly correlated with TICS-m scores suggests that higher levels of education also indicate intelligence. This is a difficult relationship to disentangle. The Army general classification test (AGCT) was designed to assess intellectual aptitude without heavy reliance on education beyond early elementary school. In a sub- sample of the veteran twins (Potter, Helms, & Plassman, 2008), the AGCT was highly correlated with education (r = 0.52). Future studies may be able to determine whether education and tooth loss are unique predictors of dementia by including other measures of SES (e.g., income), and intelligence (e.g., IQ). Conclusion & Future Directions Tooth loss in young adulthood did not predict dementia or cognitive decline in older adulthood. Although tooth loss did predict poor baseline cognitive scores, the effect was explained by its relationship to low educational attainment. Consistent with previous research, older age and lower levels of educational attainment were associated with increased risk of dementia and poorer baseline 103 cognitive scores. Attaining less than a high school education also predicted an earlier dementia onset. Although tooth loss did not predict dementia or cognitive decline in the present study, we cannot conclude that there is no relationship between them in the general population. A limitation of the present study was the low rate of tooth loss among the presumably healthy young veterans compared to other study populations. Tooth loss in the present sample was measured at approximately age 20, which may have little or no relationship to late life cognitive decline and dementia, while tooth loss at later ages may have a relationship to cognitive decline and dementia. Consideration should be given to the importance of age at the time of tooth loss and the impact this may have on the risk for dementia. Future research on oral health as a risk factor for dementia will be benefited by large samples of individuals with tooth loss caused by oral infection and measures of inflammation. Longitudinal studies measuring inflammation concurrently with PDD and dementia would be ideal. More broadly, research on the role of the immune system and sources of chronic inflammation would provide insight on potential contributing factors to dementia. For example asthma, rhinitis, and eczema were shown to increase risk of dementia (Eriksson, Gatz, Dickman, Fratiglioni, & Pedersen, 2008). To test the model proposed in Chapter II, it is of particular importance to determine whether systemic inflammation is related to neural inflammation found in dementia. Such research will contribute to the ongoing search for preventable causes of dementia and to our understanding of possible mechanisms for the disease. 104 BIBLIOGRAPHY Abbott, R. D., White, L. R., Ross, G. W., Petrovitch, H., Masaki, K. H., Snowdon, D. A. et al., (1998). Height as a marker of childhood development and late- life cognitive function: The Honolulu-Asia Aging Study. Pediatrics, 102, 602-609. Aiello, A. E., Haan, M. N., Blythe, L., Moore, K., Gonzales, J. M., & Jagust, W. (2006). The influence of latent viral infection on rate of cognitive decline over 4 years. Journal of the American Geriatrics Society, 54, 1046-1054. Airila-Mansson, S., Soder, B., Kari, K., & Meurman, J. H. (2006). Influence of combinations of bacteria on the levels of prostaglandin E2, interleukin-1β, and granulocyte elastase in gingival crevicular fluid and on the severity of periodontal disease. Journal of Periodontology, 77, 1025-1031. Alley, D. E., Crimmins, E. M., Karlamangla, A., Hu, P. & Seeman, T.E. (2008). Inflammation and rate of cognitive change in high functioning older adults. Journal of Gerontology: Medical Sciences, 63A, 50-55. American Psychiatric Association. (1987). Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition. Washington, DC: American Psychological Association. Arbes S. J., Slade, G. D., & Beck, J. D. (1999). Association between extent of periodontal attachment loss and self-reported history of heart-attack: An analysis of NHANES III data. Journal of Dental Research, 78, 1777-1782. Arnesen, E., & Forsdahl, A. (1985). The TromsØ heart study: Coronary risk factors and their association with living conditions during childhood. Journal of Epidemiology and Community Healthy, 39, 210-214. Asbell, M. B. (1988). Dentistry: A historical perspective. Dorrance and Company: Bryn Mawr, Pennsylvania. Balin, B. J., & Appelt, D. M. (2001). Role of infection in Alzheimer's disease. Journal of the American Osteopathic Association, 101, S1-6. Banks, W. A., Farr, S. A., & Morley, J. E. (2002). Entry of blood-borne cytokines into the central nervous system: Effects on cognitive processes. Neuroimmunomodulation, 10, 319-327. Barker, D. J. P. (2004). The developmental origins of well-being. Philosophical Transactions of the Royal Society of London: Biological Sciences, 359, 1359-1366. 105 Barker, D. J. P., Osmond, C., & Golding, J. (1990). Height and mortality in the counties of England and Wales. Annals of Human Biology, 17, 1-6. Bate, C., & Williams, A. (2004). Detoxified lipopolysaccharide reduces microglial cell killing of prion-infected neurons. NeuroReport, 15, 2765-2768. Beck J., Garcia, R., Heiss, G., Vokonas, P.S., & Offenbacher, S. (1996). Periodontal disease and cardiovascular disease. Journal of Periodontology, 67, 1123-1137. Beck, J. D., Slade, G., & Offenbacher, S. (2000). Oral disease, cardiovascular disease, and systemic inflammation. Periodontology, 23, 110-120. Benboudjema, L., Mulvey, M., Gao, Y., Pimplikar, S. W., & Mohr, I. (2003). Association of the herpes simplex virus type 1 USl1 gene product with the cellular kinesin ligh-chain-related protein PAT1 results in the redistribution of both polypeptides. Journal of Virology, 77, 9192-9203. Bengtsson, T. & Lindstrom, M. (2000). Childhood misery and disease in later life: The effects on mortality in old age of hazards experience in early life, southern Sweden, 1760-1894. Population Studies, 54, 263-277. Blackwell, D. L., Hayward, M. D., & Crimmins, E. M. (2001). Does childhood health affect chronic morbidity in later life? Social Science and Medicine, 52, 1269-1284. Blake G. J., & Ridker, P.M. (2002). Inflammatory bio-markers and cardiovascular risk prediction. Journal of Internal Medicine, 252, 283-294. Blasko, I., & Grubeck-Loebenstein, B. (2003). Role of the immune system in the pathogenesis, prevention and treatment of Alzheimer’s disease. Drugs and Aging, 20, 101-113. Blasko, I., Marx, F., Steiner, E., Hartmann, T, & Grubeck-Loebenstein, B. (1999). TNFα plus IFNγ induce the production of Alzheimer β-amyloid peptides and decrease the secretion of APPs. The FASEB Journal, 13, 63-68. Borenstein, A. R., Copenhaver, C. I., & Mortimer, J. A. (2006). Early-life risk factors for Alzheimer disease. Alzheimer Disease & Associated Disorders, 20, 63-72. Borrell, L. N., Beck, J. D., & Heiss, G. H. (2006). Socioeconomic disadvantage and periodontal disease: The dental atherosclerosis risk in communities study. American Journal of Public Health, 96, 332-339. 106 Breitner, J. C. S., Gau, B. A., Welsh, K. A., Plassman, B. L., McDonald, W. M., Helms, M. J., et al. (1994). Inverse association of anti-inflammatory treatments and Alzheimer’s disease: Initial result of a co-twin control study. Neurology, 44, 227-232. Breitner, J. C. S., Welsh, K. A., Gau, B. A., McDonald, W. M., Steffens, D. C., Saunders, A. M., et al. (1995). Alzheimer's disease in the NAS-NRC Registry of aging twin veterans. III. Detection of cases, longitudinal results, and observations on twin concordance. Archives of Neurology, 52,763- 771. Breteler, M. M. B. (2000). Vascular risk factors for Alzheimer’s disease: An epidemiologic perspective. Neurobiology of Aging, 21, 153-160. Bretz, W. A., Corby, P. M., Schork, N. J., Robinson, M. T., Coelho, M., Costa, S., et al. (2005). Longitudinal analysis of heritability for dental caries traits. Journal of Dental Research, 84, 1047-1051. Bretz, W. A., Weyant, R. J., Corby, P.M., Ren, D., Weissfeld, L. Kritchevsky, S. B., et al. (2003). Systemic inflammatory markers, periodontal disease, and periodontal infections in an elderly population. Journal of the American Geriatrics Society, 53, 1532-1537. Brod, S. A. (2000). Unregulated inflammation shortens human functional longevity. Inflammation Research, 49, 561-570. Brugg, B., Dubreuil, Y. L., Huber, G., Wollman, E. E., Delhaye-Bouchaud, N., & Mariani, J. (1995). Inflammatory processes induce β-amyloid precursor protein changes in mouse brain. Proceedings of the National Academy of Sciences, 92, 3032-3035. Bush, A. I., Martins, R. N., Rumble, B., Moir, R., Fuller, S., Milward, E., et al. (1990). The amyloid precursor protein of Alzheimer’s disease is released by human platelets. Journal of Biological Chemistry, 265, 15977-15983. Byrne, B. M. (1998). Structural Equation Modeling: Basic Concepts, Applications, and Programming. Mahwah, New Jersey: Lawrence Erlbaum Associates. Caamano-Isorna, F., Corral, M., Montes-Martinez, A., & Takkouche, B. (2006). Education and dementia: A meta-analytic study. Neuroepidemiology, 26, 226-232. Carranza, F., & Shklar, G. (2003). History of periodontology. Quintessence Publishing Company: Chicago, Illinois. 107 Chalmers, J., & Pearson, A. (2005). Oral hygiene care for residents with dementia: A literature review. Journal of Advanced Nursing, 52, 410-419. Chen, M., Inestrosa, N. C., Ross, G. S., & Fernandez, H. L. (1995). Platelets are the primary source of amyloid beta-peptide in human blood. Biochemical and Biophysical Research Communications, 213, 96-103. Cobb, J. L., Wolf, P. A., Au, R., White, R., & D’Agostino, R. B. (1995). The effect of education on the incidence of dementia and Alzheimer’s disease in the Framingham Study. Neurology, 45, 1707-1712. Colcombe, S. & Kramer, A. F. (2003). Fitness effects on the cognitive function of older adults: a meta-analytic study. Psychological Science, 14, 125-130. Conry, J. P., Messer, L. B., Boraas, J. C., Aeppli, D. P., & Bouchard, T. J. (1993). Dental caries and treatment characteristics in human twins reared apart. Archives of Oral Biology, 38, 937-943. Contreras, A., Zadeh, H. H., Nowzari, H., & Slots, J. (1999). Herpes virus infection of inflammatory cells in human periodontitis. Oral Microbiology and Immunology, 14, 206-212. Copeland, L.B., Krall, E. A., Brown, L. J., Garcia, R. I, & Streckfus, C. F. (2004). Predictors of tooth loss in two U.S. adult populations. Journal of Public Health Dentistry, 64, 31-37. Costa, D. L. (2000). Understanding the twentieth century decline in chronic conditions among older men. Demography, 37, 53-72. Cunha-Cruz, J., Hujoel, P. P., & Nadanovsky, P. (2007). Secular trends in socio- economic disparities in edentulism: USA 1972-2001. Journal of Dental Research, 86, 131-136. Craandijk, J., van Krugten, M. V., Verweij, C. L., van der Velden, U., & Loos, B. G. (2002). Tumor necrosis factor-α gene polymorphisms in relation to periodontitis. Journal of Clinical Periodontology, 29, 28-34. Cribbs, D. H., Azizeh, B. Y., Cotman, C. W., & LaFerla, F. M. (2000). Fibril formation and neurotoxicity by a herpes simplex virus glycoprotein B fragment with homology to the Alzheimer’s A beta peptide. Biochemistry, 39, 5988-5994. Crimmins, E. C., & Finch, C. E. (2006). Infection, inflammation, height, and longevity. Proceedings of the National Academy of Science, 103, 498-503. 108 Cummings, J. L., Vinters, H. V., Cole, G. M., & Khachaturian, Z. S., (1998). Alzheimer’s disease: Etiologies, pathophysiology, cognitive reserve, and treatment opportunities, Neurology, 51(Suppl1), S2-S17. Dannefer, D. (2003). Cumulative advantage/disadvantage and the life course: Cross-fertilizing age and social science theory. Journals of Gerontology: Social Sciences, 58, S327-S337. D’Aiuto, F., & Tonetti, D. (2004). Periodontal disease and C-reactive protein- associated cardiovascular risk. Journal of Periodontal Research, 39, 236- 241. D’Andrea, M. R, Cole, G. M., & Ard, M. D. (2004). The microglial phagocytic role with specific plaques types in the Alzheimer disease brain. Neurobiology of Aging, 25, 675-683. DeStefano, F., Anda, R.F., Kahn, H.S., Williamson, D. F., & Russell, C. M. (1993). Dental disease and risk of coronary heart disease and mortality. British Medical Journal, 306, 688-691. Diehl, S. R., Wang, Y., Brooks, C. N., Burmeister, J. A., Califano, J. V., Wang, S. et al. (1999). Linkage disequilibrium of interleukin-1 genetic polymorphisms with early-onset periodontitis. Journal of Periodontology, 70, 418-430. Dik, M. G., Jonker, C., Hack, C. E., Smit, J. H., Comijs, H. C., & Eikelenboom, P. (2005). Serum inflammatory proteins and cognitive decline in older persons. Neurology, 64, 1371-1377. Dimopoulos, N., Piperi, C., Salonicioti, A., Mitropoulos, P., Kallai, E. Liappas, I., et al. (2006). Indices of low-grade chronic inflammation correlate with early cognitive deterioration in an elderly Greek population. Neuroscience Letters, 398, 118-123. Donati, M., Berglundh, T., Hytonen, A. M., Hahn-Zoric, M., Hanson, L. A., & Padyukov, L. (2005). Association of the -159 CD14 gene polymorphism and lack of association of the -308 TNFA and Q551R IL-4A polymorphisms with severe chronic periodontitis in Swedish Caucasians, Journal of Clinical Periodontology, 32, 474-489. Duong T., Nikolaeva, M., & Acton, P. J. (1997). C-reactive protein-like immunoreactivity in the neurofibrillary tangles of Alzheimer’s disease. Brain Research, 749, 152-156. 109 Dziedzic, T. (2006). Systemic inflammatory markers and risk of dementia. American Journal of Alzheimer’s Disease and Other Dementias, 21, 258- 262. Ebersole, J. L., Machen, R. L., Steffen, M. J., & Willmann, D. E. (1997). Systemic acute-phase reactants, C-reactive protein and haptoglobin in adult periodontitis. Clinical and Experimental Immunology, 107, 347-352. Engelhart, M. J., Geerlings, M. I., Meijer, J., Kiliaan, A., Ruitenberg, A., van Swieten, J. C., et al. (2004). Inflammatory proteins in plasma and the risk of dementia. Archives of Neurology, 61, 668-672. Eriksson, U. K., Gatz, M., Dickman, P. W., Fratiglioni, L., & Pedersen, N. L. (2008). Asthma, eczema, rhinitis and the risk for dementia. Dementia and Geriatric Cognitive Disorders, 25, 148-156. Finch, C. E., & Crimmins, E. M. (2004). Inflammatory exposure and historical changes in human life-spans. Science, 17, 1736-1739. Floud, R., Wachter, K., & Gregory, A. (1990). Height, health, and history: Nutritional status in the United Kingdom 1750-1980. Cambridge: Cambridge University Press. Fokkema, S. J., Loos, B. G., deSlegte, C., Burger W., Piscaer, M., Ijzerman, Y., et al. (2003). Increased release of IL-12p70 by monocytes after periodontal therapy. Journal of Clinical Periodontology, 30, 1091-1096. Folwaczny, M., Glas, J., Torok, H. P., Mende, M., & Folwaczny, C. (2004). Lack of association between the TNF-α G-308 A promoter polymorphism and periodontal disease. Journal of Clinical Periodontology, 31, 449-453. Forner, L., Larsen, T., Kilian, M., & Holmstrup, P. (2006). Incidence of bacteremia after chewing, tooth brushing, and scaling in individuals with periodontal inflammation. Journal of Clincal Periodontology, 33, 401-407. Fratiglioni, L., Grut, M., Forsell, Y., Viitanen, M., Grafstrom, M., Holmen, K., et al. (1991). Prevalence of Alzheimer’s disease and other dementias in an elderly urban population: Relationship with age, sex, and education. Neurology, 41, 1886-1892. Galbraith, G. M. P., Hendley, T. M., Sanders, J. J., Palesch, Y. & Pandey, J. P. (1999). Polymorphic cytokine genotypes as markers of disease severity in adult periodontitis, Journal of Clinical Periodontology, 26, 705-709. 110 Gallo, J. J., & Breitner, J. C. S. (1995). Alzheimer’s disease in the NAS-NRC Registry of Ageing Twin Veteran, IV. Performance characteristics of a two- stage telephone screening procedure for Alzheimer’s dementia. Psychological Medicine, 25, 1211-1220. Garcia, R. I., Henshaw, M. M., & Krall, E. A. (2000). Relationship between periodontal disease and systemic health. Periodontology, 25, 21-36. Garcia, R. I., Krall E. A., & Vokonas, P. S. (1998). Periodontal disease and mortality from all causes in the VA dental longitudinal study. Annals of Periodontology, 3, 339-349. Gatz, M., Mortimer, J. A., Fratiglioni, L., Johansson, B., Berg, S., Reynolds, C. A., et al. (2006). Potentially modifiable risk factors for dementia in identical twins. Alzheimer’s & Dementia, 2, 110-117. Gatz, M., Pedersen, N. L, Crowe, M., & Fiske, A. (2000). Defining discordance in twin studies of risk and protective factors for late life disorders. Twin Research, 3, 159-164. Gatz, M., Reynolds, C. A., Fratiglioni, L., Johansson, B., Mortimer, J. A., Berg, S., et al. (2006). Role of genes and environments for explaining Alzheimer disease. Archives of General Psychiatry, 63, 168-174. Gatz, M., Svedberg, P., Pedersen, N.L., Mortimer, J.A., Berg, S., & Johansson, B. (2001). Education and the risk of Alzheimer’s disease: Findings from the Study of Dementia in Swedish Twins. Journals of Gerontology: Psychological Sciences, 56, 292-300. Georges, J. L., Rupprecht, H. J., Blankenberg, S., Poirier, O., Bickel, C., Hafner, G., et al., 2003. Impact of pathogen burden in patients with coronary artery disease in relationship systemic inflammation and variation in genes encoding cytokines. American Journal of Cardiology, 92, 515-521. Gieffers, J., Reusche, E., Solbach, W., & Maass, M. (2000). Failure to detect Chlamydia pneumoniae in brain sections of Alzheimer’s disease patients. Journal of Clinical Microbiology, 38, 881-881. Goteiner, D., Ashmen, R., Lehrman, N., Janal, M. N. & Eskin, B. (2008). Presence and significance of interleukin-1 polymorphism in patients who present with acute coronary syndrome, angina, and chronic periodontitis: An epidemiologic pilot study. Journal of Periodontology, 79, 138-143. Grammas, P. & Ovase, R. (2001). Inflammatory factors are elevated in brain microvessels in Alzheimer’s disease. Neurobiology of Aging, 22, 837-842. 111 Grau, A. J., Becher, H., Ziegler, C. M., Lichy, C., Buggle, F., Kaiser, C., et al. (2004). Periodontal disease as a risk factor for ischemic stroke. Stroke, 35, 496-501. Greenland, S., & Robins, J. M. (1988). Conceptual problems in the definition and interpretation of attributable fractions. American Journal of Epidemiology, 128, 1185-1197. Greenow, K., Pearce, N. J., & Ramji, D. P. (2005). The key role of apolipoprotein in atherosclerosis. Journal of Molecular Medicine, 83, 329-342. Griffin, W. S. T., Stanley, L. C., Ling, C., White, V., MacLeod, V., Perrot, L. J., et al. (1989). Brain interleukin 1 and S-100 immunoreactivity are elevated in Down syndrome and Alzheimer’s disease. Proceedings of the National Academy of Sciences, 86, 7611-7615. Griffin, W. S. T., Sheng, J. G., Royston, M. C., Gentleman, S. M., McKenzie, J. E., & Graham, D. I., et al. (1998). Glial-neuronal interactions in Alzheimer’s disease: The potential role of a ‘cytokine cycle’ in disease progression. Brain Pathology, 8, 65-72. Grimaldi, L. M. E., Casadei, V. M., Ferri, C., Veglia, F., Licastro, F. Annoni, G., et al. (2000). Association of early-onset Alzheimer’s disease with an interleukin-1α gene polymorphism. Annals of Neurology, 47, 361-365. Haan, M. N., Shemanski, L., Jagust, W. J., Manolio, T. A., & Kuller, L. (1999). The role of APOE ε4 in modulating effects of other risk factors for cognitive decline in elderly persons. Journal of the American Medical Association, 281, 40-46. Hall, C. B., Derby, C., LeValley, A., Katz, M. J., Verghese, J. & Lipton, R. B., (2007). Education delays accelerated decline on a memory test in persons who develop dementia. Neurology, 69, 1657-1664. Hammond, R. R., Gage, F. H., & Terry, R. D. (1993). Alzheimer’s disease and spirochetes; a questionable relationship. Neuroreport, 4, 840. Hauss-Wegrzyniak, B., Vraniak, P. D., & Wenk, G. L. (2000). LPS-induced neuroinflammatory effects do not recover with time. Neuroreport, 11, 1759-1763. Hayden, K. M., Zandi, P. P., Khachaturian, A. S., Szekely, C. A., Fotuhi, M., Norton, M. C., et al. (2007). Does NSAID use modify cognitive trajectories in the elderly?: The Cache County Study. Neurology, 69, 275-282. 112 Hebert, L. E., Scherr, P. A., Bienias, J. L., Bennett D. A., & Evans, D. A. (2003). Alzheimer disease in the US population: prevalence estimates using the 2000 census. Archives of Neurology, 60, 1119-1122. Heneka, M. T., & O’Banion, M. K. (2007). Inflammatory processes in Alzheimer’s disease. Journal of Neuroimmunology, 184, 69-91. Hernichel-Gorbach, E., Kornman, K. S., Holt, S. C., Nichols, F., Meador, H., Kung, J. T., et al. (1994). Host responses in patients with generalized refractory periodontitis. Journal of Periodontology, 65, 8-16. Herzberg, M. C., & Meyer, M. W. (1996). Effects of oral flora on platelets: Possible consequences in cardiovascular disease. Journal of Periodontology, 65, 8-16. Herzberg, M. C., & Meyer, M. W. (1998). Dental plaque, platelets, and cardiovascular diseases. Annals of Periodontology, 3, 151-160. Hodge, R. W., Siegel, P. M., & Rossi, P. H. (1964). Occupational prestige in the United States, 1925-1963. The American Journal of Sociology, 70, 286- 302. Holmes, C., El-Okl, M., Williams, A.L., Cunningham, C., Wilcockson, D., & Perry, V. H. (2003). Systemic infection, interleukin 1 beta, and cognitive decline in Alzheimer's disease. Journal of Neurology, Neurosurgery & Psychiatry. 74, 788-789. Honig, L. S., Kukull, W., & Mayeux, R. (2005). Atherosclerosis and AD: Analysis of data from the US National Alzheimer’s Coordinating Center. Neurology, 64, 494-500. Ide, M., McPartlin, D., Coward, P. Y., Crook, M., Lumb, P., & Wilson, R. F. (2003). Effect of treatment of chronic periodontitis on levels of serum markers of acute-phase inflammatory and vascular responses. Journal of Clinical Periodontology, 30, 334-340. Itzhaki, R. F., & Wozniak, M. A. (2006). Herpes simplex virus type 1, apolipoprotein E, and cholesterol: A dangerous liaison in Alzheimer’s disease and other disorders. Progress in Lipid Research, 45, 73-90. Itzhaki, R. F., & Wozniak, M. A. (2007). Viral infection and cognitive decline. Journal of the American Geriatrics Society, 55, 131. 113 Itzhaki, R. F., Wozniak, M. A., Appelt, D. M., & Balin, B. J. (2004). Infiltration of the brain by pathogens causes Alzheimer’s disease. Neurobiology of Aging, 25, 619-627. Itzhaki, R. F., Lin, W., Shang, D., Wilcock, G. K., Faragher, B., & Jamieson, G. A. (1997). Herpes simplex virus type 1 in the brain and risk of Alzheimer’s disease. Lancet, 349, 241-244. Ismael, A. I., Morrison, E. C., Burt, B. A., Caffesse, R. G., & Kavanagh, M.T. (1990). Natural history of periodontal disease in adults: Findings from the Tecumseh Periodontal Disease Study, 1959-1987. Journal of Dental Research, 69, 430-435. Jablon, S., Neel, J. V., Gershowitz, H., & Atkinson, G. F., (1967). The NAS-NRC twin panel: methods of construction of the panel, zygosity diagnosis and proposed use. American Journal of Human Genetics, 19, 133-161. Jamieson, G. A., Maitland, M. J., Wilcock, G. K., Yates, C. M., & Itzhaki, R. F. (1992). Herpes simplex virus type 1 DNA is present in specific regions of brain from aged people with and without senile dementia of the Alzheimer type. Journal of Pathology, 167. 365-368. Jellinger, K. A. (2004). Head injury and dementia. Current Opinion in Neurology, 17, 719-723. Jorm, A. F., & Jacomb, P. A. (1989.). The Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): Socio-demographic correlates, reliability, validity and some norms. Psychological Medicine, 19, 1015-1022. Kalaria, R. N. (1992). Serum amyloid P and related molecules associated with the acute-phase response in Alzheimer’s disease. Research in Immunology, 143, 637-641. Kang, J. H., Ascherio, A., & Grodstein, F. (2005). Fruit and vegetable consumption and cognitive decline in aging women. Annals of Neurology 57, 713-720. Kim, J., Stewart, R., Prince, M., Kim, S., Yang, S. Shin, I., et al. (2007). Dental health, nutritional status and recent-onset dementia in a Korean community population. International Journal of Geriatric Psychiatry, 22, 850-855. Kinane, D. F., Winstanley, F. P., Adonogianaki, E., Moughal, & N.A. (1992). Bioassay of interleukin-1 (IL-1) in human gingival crevicular fluid during experimental gingivitis. Archives of Oral Biology, 37, 153-156. 114 Kinoshita, J. (2004). Pathogens as a cause of Alzheimer’s disease. Neurobiology of Aging, 25, 639-640. Klein, W. L., Krafft, G. A., & Finch, C. E. (2001). Targeting small Aβ oligomers: The solution to an Alzheimer’s disease conundrum? Trends in Neurosciences, 24, 219-224. Kondo, K., Niino, M., & Shido, K. (1994). A case-control study of Alzheimer’s disease in Japan: Significance of lifestyles. Dementia, 5, 314-326. Kornman, K. S. & Duff, G. W. (2001). Candidate genes as potential links between periodontal and cardiovascular disease. Annals of Periodontology, 6, 48- 57. Kornman, K. S., Pankow, J., Offenbacher, S., Beck, J., diGiovine, F., & Duff, G. W. (1999). Interleukin-1 genotypes and the association between periodontitis and cardiovascular disease. Journal of Periodontal Research, 34, 353-357. Kramer, A. R., Craig, R. G., Dasanayake, A. P., Brys, M., Glodzik-Sobanska, L., & deLeon, M. J. (2008). Inflammation and Alzheimer’s disease: Possible role of periodontal disease. Alzheimer’s & Dementia, 4, 242-250. Kubar, A., Saygun, I., Ozdemir, A., Yapar, M., & Slots, J. (2005). Real-time polymerase chain reaction quantification of human cytomegalovirus and Epstein-Barr virus in periodontal pockets and the adjacent gingiva of periodontitis lesions. Journal of Periodontal Research, 40, 97-104. Larson, E. B., Wang, L., Bowen, J. D., McCormick, W. C., Teri, L., Crane, P., et al. (2006). Exercise is associated with reduced risk for incident dementia among persons 65 years of age and older. Annals of Internal Medicine, 144, 73-81. Launer, L. J., Andersen, K., Dewey, M. E., Letenneur, L., Ott, A., Amaducci, L. A., et al., (1999). Rates and risk factors for dementia and Alzheimer’s disease: Results from EURODEM pooled analyses, Neurology, 52, 78-84. Li, X., Kolltveit, K. M., Tronstad, L., & Olsen, I. (2000). Systemic diseases caused by oral infection. Clinical Microbiology Reviews, 13, 547-558. Licastro, F., Candore, G., Lio, D., Porcellini, E., Colonna-Romano, G., Franceschi, C., et al. (2005). Innate immunity and inflammation in ageing: A key for understanding age-related disease. Immunity and Ageing, 2, 8- 22. 115 Licastro, F., Morini, M. C., Polazzi, E., & Davis, L. J. (1995). Increased serum α1- antichymotrypsin in patients with probable Alzheimer’s disease: An acute phase reactant without the peripheral acute phase response. Journal of Neuroimmunology, 57, 71-75. Lin, D. Y. (1994). Cox regression analysis of multivariate failure time data: the marginal approach. Statistical Medicine, 13, 2233-2247. Lin, N. (1976). Foundations of social research. New York: McGraw Hill. Lin, W. R., Graham, J., MacGowan, S. M., Wilcock, G. K., & Itzhaki, R. F. (1998). Alzheimer's disease, herpes virus in brain, apolipoprotein E4 and herpes labialis. Alzheimer's Report 1, 173-178. Lin, W. R., Wozniak, M. A., Esiri, M. M., Klenerman, P., & Itzhaki, R. F. (2001). Herpes simplex encephalitis: involvement of apolipoprotein E genotype. Journal of Neurology, Neurosurgery, and Psychiatry, 70, 117-119. Lichtenstein, P., DeFaire, U., Floderus, B., Svartengren, M., Svedberg, P., & Pedersen, N. L. (2002). The Swedish Twin Registry: A unique resource for clinical, epidemiological and genetic studies. Journal of Internal Medicine, 252, 184-205. Little, C. S., Hammond, C. J., MacIntyre, A., Balin, B. J., & Appelt, D. M. (2004). Chlamydia pneumoniae induces Alzheimer-like amyloid plaques in brains of BALB/c mice. Neurobiology of Aging, 25, 419-429. Loesche, W.J., & Lopatin, D.E. (2000). Interactions between periodontal disease, medical diseases and immunity in the older individual. Periodontology, 16, 80-105. Loesche, W. J., Schork, A., Terpenning, M. S., Chen, Y. M., Dominguez, B. L., & Grossman, N. (1998). Assessing the relationship between dental disease and coronary heart disease in elderly U.S. veterans. Journal of the American Dental Association, 129, 301-311. Loesche, W. J., Schork, A., Terpenning, M. S., Chen, Y. M., Kerr, C., & Dominguez, B. L. (1998). The relationship between dental disease and cerebral vascular accident in elderly United States veterans. Annals of Periodontology, 3, 161-174. Loos B. G. (2005). Systemic markers of inflammation in periodontitis. Journal of Periodontology, 76, 2106-2115. 116 Luchsinger, J. A., Reitz, C., Tang, M. X., Shea, S., & Mayeux, R. (2005). Aggregation of vascular risk factors and risk of incident Alzheimer’s disease. Neurology, 65, 545-551. Lyketsos, C. G., Chen, L., Anthony, J. C. (1999). Cognitive decline in adulthood: An 11.5 year follow up of the Baltimore epidemiologic catchment area study. American Journal of Psychiatry, 156, 58-65. Lynch, J. R., Tang, W., Wang, H., Vitek, M. P., Bennett, E. R., Sullivan, P. M., et al. (2003). APOE genotype and an ApoE-mimetic peptide modify the systemic and central nervous system inflammatory response. Journal of Biological Chemistry, 278, 48529-48533. MacIntyre, A., Abramov, R., Hammond, C. J., Hudson, A. P., Arking, E. J., Little, C. S., et al. (2003). Chlamydia pneumoniae infection promotes the transmigration of monocytes through human brain endothelial cells. Journal of Neuroscience Research, 71, 740-750. Magnusson, P. K. E., Rasmussen, F., & Gyllensten, U. B. (2006). Height at age 18 years is a strong predictor of attained education later in life: Cohotr study of over 950000 Swedish men. International Journal of Epidemiology, 35, 658-663. Masada, M. P., Persson, R., Kenney, J. S., Lee, S. W., Page, R. C., & Allison, A. C. Measurement of interleukin-1α and -1β in gingival crevicular fluid: Implication for the pathogenesis of periodontal disease, Journal of Periodontal Research, 25, 156-163. Mattila, K. J., Vesanen, M., Valtonen, V., Nieminen, M., Palosuo, T., Rasi, V., et al. (2002). Effect of treating periodontitis on C-reactive protein levels: A pilot study. BMC Infectious Diseases, 2, 30-32. Mattson, M. P. (2004). Infectious agents and age-related neurodegenerative disorders. Ageing Research Reviews, 3, 105-120. Marx, F., Blasko, I., Pavelka, M., & Grubeck-Loebenstein (1998). The possible role of the immune system in Alzheimer’s disease. Experimental Gerontology, 33, 871-881. McCusker, S. M., Curran, M. D., Dynan, K. B., McCullagh, C. D., Urquhart, D. D., Middleton, D., et al. (2001). Association between polymorphism in regulation region of gene encoding tumour necrosis factor α and risk of Alzheimer’s disease and vascular dementia: A case-control study. Lancet, 357, 436-439. 117 McDade, T. W. (2003). Life history theory and the immune system: Steps toward a human ecological immunology, Yearbook of Physical Anthropology, 46, 100-125. McGeer, P. L., & McGeer, E. G. (1995). The inflammatory response system of brain: Implications for therapy of Alzheimer and other neurodegenerative diseases. Brain Research Reviews, 21, 195-218. McGeer, P. L., & McGeer, E. G. (2007). NSAIDs and Alzheimer disease: Epidemiological, animal model and clinical studies. Neurobiology of Aging, 28, 639-647. McGeer, P. L., Rogers, J., & McGeer, E. G. (1994). Neuroimmune mechanisms in Alzheimer disease pathogenesis. Alzheimer’s Disease and Associated Disorders, 8, 149-158. McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., & Stadlan, E. M. (1984). Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-ADRDA Work Group under the auspices of the Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology, 34, 939-944. McLaughlin, R., Kin, N. M., Chen, M. F., Nair, N. P., & Chan, E. C. (1999). Alzheimer’s disease may not be a spirochetosis. Neuroreport, 10, 1489- 1491. Metha, J. L., Saldeen, T. G. P., & Rand. K. (1998). Interactive role of infection, inflammation, and traditional risk factors in atherosclerosis and coronary artery disease. Journal of the American College of Cardiology, 31, 1217- 1225. Meurman, J. H., & Hamalainen, P. (2006). Oral health of people with dementia. Gerodontology, 23, 3-32. Miklossy, J. (1993). Alzheimer’s disease- a spirochetosis? Neuroreport, 4, 841- 848. Moradi, T., Adami, H., Ekbom, A., Wedren, S., Terry, P., Floderus, B., et al. (2002). Physical activity and risk for breast cancer a prospective cohort study among Swedish twins. International Journal of Cancer, 100, 76-81. Mortimer, J. A., Snowdon, D. A., & Markesbery, W. R. (2003). Head circumference, education, and risk of dementia: Findings from the Nun Study. Journal of Clinical and Experimental Neuropsychology, 25, 671- 679. 118 Mrak, R. E., & Griffin, W. S. T. (2005). Glia and their cytokines in progression of neurodegeneration. Neurobiology of Aging, 26, 349-354. Muthen, L. K. & Muthen, B. O. (2007). Mplus User’s Guide. Fourth Edition. Los Angeles, CA: Muthen & Muthen. Mucci, L. A., Bjorkman, L., Douglass, C. W., & Pedersen, N. L. (2005). Environmental and heritable factors in the etiology of oral diseases: A population-based study of Swedish twins. Journal of Dental Research, 84, 800-805. Nagele, R. G., Wegiel, J., Venkataraman, V., Imaki, H., Wang, J., & Wegiel J. (2004). Contribution of glial cells to the development of amyloid plaques in Alzheimer’s disease. Neurobiology of Aging, 25, 663-674. Neuroinflammation Working Group. (2000). Inflammation and Alzheimer’s disease. Neurobiology of Aging, 21, 383-421. Nicoll, J. A. R., Mrak, R. E., Graham, D. I., Stewart, J., Wilcock, G., MacGowan, S., et al., (2000). Association of interluekin-1 gene polymorphisms with Alzheimer’s disease. Annals of Neurology, 47, 365-368. Nochlin, D., Shaw, C. M., Campbell, L. A., & Kuo, C. C. (1999). Failure to detect Chlamydia pneumoniae in brain tissues of Alzheimer’s disease. Neurology, 53. 1888. Offenbacher, S. (1996). Periodontal disease: Pathogenesis. Annals of Periodontology, 1, 821-878. O’Rand, A. M. (2001). Stratification and the life course: The forms of life-course capital and their interrelationships. In R. H. Binstock & L. K. George (Eds.) Handbook of Aging and the Social Sciences 5 th edition. San Diego: Academic Press. 197-213. Page, W. F. (1995). Annotation: The National Academy of Sciences-National Research Council Twin Registry. American Journal of Public Health, 85, 617. Page, W. F. (2002). The NAS-NRC Twin Registry of WWII military veteran twins. Twin Research, 5, 493-496. Pan, W., & Kastin, A. J. (1999). Penetration of neurotrophins and cytokines across the blood-brain/ blood-spinal cord barrier. Advanced Drug Delivery Reviews, 36, 291-298. 119 Pedersen, N. L., Plomin, R., Nesselroade, J. R., & McClearn, G. E. (1992). A quantitative genetic analysis of cognitive abilities during the second half of the life span. Psychological Science, 3, 346-352. Parra, B., & Slots, J. (1996). Detection of human viruses in periodontal pockets using polymerase chain reaction. Oral Microbiology and Immunology, 11, 289-293. Perry, V.H. (2004). The influence of systemic inflammation on inflammation in the brain: implications for chronic neurodegenerative disease. Brain, Behavior, and Immunity, 18, 407-413. Pham, K., Feik, D., Hammond, B. F., Rams, T. E., & Whitaker, E. J. (2002). Aggregation of human platelets by gingipain-R from porphyromonas gingivalis cells and membrane vesicles. Platelets, 12, 21-30. Plassman, B. L., Steffens, D. C., Burke, J. R., Welsh-Bohmer, K. A., Newman, T. N., Drosdick, D., et al. (2006). Duke Twins Study of Memory in Aging in the NAS-NRC Twin Registry. Twin Research and Human Genetics, 9, 950-957. Potter, G. G., Helms, M. J., & Plassman, B L. (2008). Associations of job demands and intelligence with cognitive performance among men in late life. Neurology, 70, 1803-1808. Qiu, C., Winblad, B., & Fratiglioni, L. (2005). The age-dependent relation of blood pressure to cognitive function and dementia. Lancet Neurology, 4, 487- 499. Rainero, I., Bo, M., Ferrero, M. Valfre, W., Vaula, G., & Pinessi, L. (2004). Association between the interleukin-1α gene and Alzheimer’s disease: A meta-analysis. Neurobiology of Aging, 25, 1293-1298. Reed, T., Plassman, B. L., Tanner, C., Dick, D. M., Rinehart, S. A., & Nichols, W. C. (2005). Verification of self-report of zygosity determined via DNA testing in a subset of the NAS-NRC Twin Registry 40 years later. Twin Research and Human Genetics, 8, 362-367. Reiss, A. J. (1961). Occupations and social status. New York: Free Press. Ring, R. H., & Lyons, J. M. (2000). Failure to detect Chlamydia pneumoniae in the late-onset Alzheimer’s brain. Journal of Clinical Microbiology, 38, 2591-2594. 120 Ringheim, G. E., & Conant, K. (2004). Neurodegenerative disease and the neuroimmune axis (Alzheimer’s and Parkinson’s disease, and viral infections). Journal of Neuroimmunology, 147, 43-49. Riviere, G. R., Riviere, K. H., & Smith, K. S. (2002). Molecular and immunological evidence of oral Treponema in the human brain and their association with Alzheimer’s disease. Oral Microbiology and Immunology, 17, 113-118. Robinson, S. R., Dobson, C., & Lyons, J. (2004). Challenges and directions for the pathogen hypothesis of Alzheimer’s disease. Neurobiology of Aging, 25, 629-637. Rogers, J., Strohmeyer, R., Kovelowski, C. J., & Li, R. (2002). Microglia and inflammatory mechanisms in the clearance of amyloid beta peptide. Glia, 40, 260-269. Roman, G. C., Tetemichi, T. K., Erkinjuntti, T., Cummings, J. L., Masdeu, J. C., Garcia, J. H., et al. (1993). Vascular dementia: diagnostic criteria for research studies: report of the NINDS-AIREN international workshop. Neurology, 43, 250-260. Salthouse, T. A. (1999). Theories of Cognition. In V. L. Bengtson & K. W. Shaie (Eds.) Handbook of Theories of Aging. Springer: New York. Satpute-Krishnan, P., DeGiorgia, J. A., & Bearer, E. L. (2003). Fast anterograde transport of herpes simplex virus: role for the amyloid precursor protein of Alzheimer’s disease. Aging Cell, 2, 305-318. Scannapieco, F. A., Bush, R. B., & Paju, S. (2003). Associations between periodontal disease and risk for atherosclerosis, cardiovascular disease, and stroke. A systematic review. Annals of Periodontology, 8, 38-53. Schaie, K. W. (1998). The Seattle Longitudinal Studies of adult intelligence. In M. P. Lawton & T. A. Salthouse (Eds.) Essential Papers on the Psychology of Aging. New York University Press: New York. Schmidt, R., Schmidt, H., Curb, J. D., Masaki, K., White, L. R., & Launer, L. J. (2002). Early inflammation and dementia: A 25-year follow-up of the Honolulu-Asia aging study. Annals of Neurology, 52, 168-174. Schmitt, T. L., Steiner, E., Klinger, P., Sztankay, A. & Grubeck-Loebenstein, B. (1996). The production of an amyloidogenic metabolite of the Alzheimer amyloid β precursor protein (APP) in thyroid cells is stimulated by interleukin 1β, but inhibited by interferon gamma. Journal of Clinical Endocrinology and Metabolism, 81, 1666-1669. 121 Sharma, A., Novak, E. K., Sojar, H. T., & Swank, R. T. (2000). Porphyromonas gingivalis platelet aggregation activity: Outer membrane vesicles are potent activators of murine platelets. Oral Microbiology and Immunology, 15, 393-396. Sheng, J. G., Bora, S. H., Xu, G., Borchelt, D. R., Price, V. E., & Koliatsos, V. E. (2003). Lipopolysaccharide-induced-neuroinflammation increases intracellular accumulation of amyloid precursor protein and amyloid β peptide in APPswe transgenic mice. Neurobiology of Disease, 14, 133- 145. Shimazaki, Y., Soh, I., Saito, T., Yamashita, Y., Koga, T., Miyazaki, H., et al. (2001). Influence of dentition status on physical disability, mental impairment, and mortality in institutionalized elderly people. Journal of Dental Research, 80, 340-345. Shipley, S. J., Parkin, E. T., Itzhaki, R. F., & Dobson, S. S. (2005). Herpes simplex virus interferes with amyloid precursor protein processing. BMC Microbiology, 5, 48-55. Singh, V. K., & Guthikonda, P. (1997). Circulating cytokines in Alzheimer’s disease. Journal of Psychiatric Research, 31, 657-660. Skoog, I., Lernfelt, B., Landahl, S., Palmertz, B., Andreasson, L., Nilsson, L., et al., (1996). 15 year longitudinal study of blood pressure and dementia. Lancet, 347, 1141-1145. Slade, G. D., Ghezzi, E. M., Heiss, G., Beck, J. D., Riche, E., & Offenbacher, S. (2003). Relationship between periodontal disease and C-reactive protein among adults in the Atherosclerosis Risk in Communities Study. Archives of Internal Medicine, 163, 1172-1179. Stein, P. S., Desrosiers, M., Donegan, S., Yepes, R. J., & Kryscio, R. J. (2007). Tooth loss, dementia and neuropathology in the Nun Study. Journal of the American Dental Association, 138, 1314-1322. Stein, P. S., Scheff, S., & Dawson, D. R. (2006). Alzheimer’s disease and periodontal disease: Mechanisms underlying a potential bi-directional relationship. Grand Rounds in Oral Systemic Medicine, 3, 14-24. Stern, Y., Albert, S., Tang, M. X., & Tsai, W. Y. (1999). Rate of memory decline in AD is related to education and occupation. Cognitive Reserve? Neurology, 53, 1942-1947. 122 Stewart, R., & Hirani, V. (2007). Dental health and cognitive impairment in an English national survey population. Journal of the American Geriatrics Society, 55, 1410-1414. Stoll, L. L., Denning, G. M., & Weintraub, N. L. (2004). Potential role of endotoxin as a proinflammatory mediator of atherosclerosis. Arteriosclerosis, Thrombosis, and Vascular Biology, 24, 2227-2236. Strittmatter, W. J., Saunders, A. M., Schmechel, D., Pericak-Vance, M., Enghild, J., Salvesen, G. S., et al. (1993). Apolipoprotein E: High-avidity binding to B-amyloid and increased frequency of type 4 alleles in late onset familial Alzheimer’s disease. Proceedings of the National Academy of Sciences, 90, 1977-1981. Takakuwa, H., Goshima, F., Koshizuka, T., Murata, T., Daikoku, T., & Nishiyama, Y. (2001). Herpes simplex virus encodes a virion-associated protein which promotes long cellular processes in over-expressing cells. Genes to Cells, 6, 955-956. Tan, Z. S., Beiser, A. S., Vasan, R. S., Roubenoff, R., Dinarello, C. A., Harris, T. B., et al. (2007). Inflammatory markers and the risk of Alzheimer disease: The Framingham study. Neurology, 68, 1902-1908. Taylor, B. A., Tofler, G. H., Carey, H. M. R., Morel-Kopp, M. C., Philcox, S., Carter, T. R., et al. (2006). Full mouth tooth extraction lowers systemic inflammatory and thrombotic markers of cardiovascular risk. Journal of Dental Research, 85, 74-78. Tosteson, T. D., Buzas, J. S., Demidenko, E., & Karagas, M. (2003). Power and sample size calculations for generalized regression models with covariate measurement error. Statistics in Medicine, 22, 1069-1082. Townsend, K. P., & Pratico, D. (2005). Novel therapeutic opportunities for Alzheimer’s disease: Focus on nonsteroidal anti-inflammatory drugs. The FASEB Journal, 19, 1592-1601. Valtonen, V. V. (1991). Infection as a risk factor for infarction and atherosclerosis. Annals of Medicine, 23, 539-543. van Exel, E., Craen, A. J. M., Remarque, E. J., Gussekloo, J., Houx, P., Bootsma-van der Wiel, A., et al. (2003). Interaction of atherosclerosis and inflammation in elderly subjects with poor cognitive function. Neurology, 61, 1695-1701. 123 vonBernhardi, R., Ramirez, G., Toro, R., & Eugenin, J. (2007). Pro-inflammatory conditions promote neuronal damage mediated by amyloid precursor protein and decrease its phagocytosis and degradation by microglial cells in culture. Neurobiology of Disease, 26, 153-164. Weaver, J. D., Huang, M. H., Albert, M., Harris, T., Rowe, J. W., & Seeman, T. E. (2002). Interleukin-6 and risk of cognitive decline: MacArthur Studies of Successful Aging. Neurology, 59, 371-378. Weinberg, A., Krisanaprakornkit, S., & Dale, B. A. (1998). Epithelial antimicrobial peptides: Review and significance for oral applications. Critical Reviews in Oral Biology and Medicine, 9, 399-414. Welsh, K. A., Breitner, J. C. S., & Magruder-Habib, K. M. (1993). Detection of dementia in the elderly using telephone screening of cognitive status. Neuropsychiatry, Neuropsychology, and Behavioral Neurology, 6, 103- 110. Wilson, C. J., Finch, C. E., & Cohen, H. J. (2002). Cytokines and cognition: The case for a head-to-toe inflammatory paradigm. Journal of the American Geriatrics Society, 50, 2041-2056. Wilson, M., Reddi, K., & Henderson, B. (1996). Cytokine-inducing components of periodontopathogenic bacteria. Journal of Periodontal Research, 31, 393- 407. Wilson, R. S., Li, Y., Aggarwal, N. T., Barnes, L. L., McCann, J. J., Gilley, D. W., et al. (2004). Education and the course of cognitive decline in Alzheimer’s disease. Neurology, 63, 1198-1202. Wood, J. A., Wood, P. L., Ryan, R., Graff-Radford, N. R., Pilapil, C. Robitaille, Y., et al. (1993). Cytokine indices in Alzheimer’s temporal cortex: No changes in mature IL-1β or IL-1RA but increases in the associated acute phase proteins IL-6, α2 macroglobulin, and C-reactive protein. Brain Research, 629, 245-252. Wozniak, M. A., Cookson, A., Wilcock, G. K., & Itzhaki, R. F. (2003). Absence of chlamydia pneumoniae in brain of vascular dementia patients. Neurobiology of Aging, 24, 761-765. Wozniak, M. A, Shipley, S. J., Combrinck, M., Wilcock, G. K., & Itzhaki, R. F. (2005). Productive herpes simplex virus in brain of elderly normal subjects and Alzheimer’s disease patients. Journal of Medical Virology, 75, 300- 306. 124 Wu, T., Trevisan, M., Genco, R. J., Dorn, J. P., Falkner, K.L., & Sempos, C. T. (2000). Periodontal disease and risk of cerebrovascular disease: The first National Health and Nutrition Examination Survey and its follow-up study. Archives of Internal Medicine, 160, 2749-2755. Yaffe, K., Lindquist, K., Penninx, B. W., Simonsick, E. M. Pahor, M., Kritchevsky S., et al. (2003). Inflammatory markers and cognition in well-functioning African-American and white elders. Neurology, 61, 76-80. Zandi, P. P., & Breitner, J. C. S. (2001). Do NSAIDs prevent Alzheimer’s disease? And if so, why? The epidemiological evidence. Neurobiology of Aging, 22, 811-817. 125 APPENDIX A: STANDARDS OF DENTAL TREATMENT DURING THE WORLD WAR II ERA Ninety-seven percent of the members of our sample entered the military between 1936 and 1946. The remaining 3% entered between 1947 and 1955. At the time the military dental examinations were conducted, periodontal infection was referred to as pyorrhea (or pyorrhea alveolaris). Pocket probing was becoming a routine way to measure the depth of periodontal pockets. In the 1940s, surgical treatment for periodontal infection was introduced. However, many dentists rejected surgical treatments despite the availability of anesthesia and X-rays (Carranza & Shklar, 2003). The standard method of periodontal treatment involved the removal of calculus and other deposits on the teeth, called scaling. Some dentists used acids to remove calculus. It was also believed at the time that improper contact of the upper and lower teeth contributed to periodontal disease and attempts were made to correct this. It was common dental practice to replace missing teeth with artificial substitutes as is reported for many of the members of the NAS-NRC sample. Health campaigns during the twentieth century advocated oral hygiene for the prevention of gingivitis. During military recruitment for World War I, dental treatment experienced rapid development to ensure the availability of soldiers eligible for duty, as those with untreated dental disease would have been excluded. By the time members of the NAS-NRC sample entered the military during the World War II era, dental treatment was well established in the armed forces (Asbell, 1988). 126 APPENDIX B: RESOLUTION OF TWIN DISAGREEMENT IN THE REPORT OF FATHER’S OCCUPATION For disagreements in which occupational status differed by 10 points or less, the scores were averaged (N=313). For disagreements in which occupational status differed by greater than 10 points (N=326): 1. If both responses were in the same occupational category, a more specific occupation term was accepted over a less specific term (N=31) 2. If both responses were in the same occupational category but neither was more specific, the score for the general category was accepted (identified as “not elsewhere classified” for each category) (N=2) 3. If the score for the general category was substantially higher than either of the scores reported by the twins and the difference was 12 or less, the two scores were averaged (N=4) 4. Typographical errors were corrected if reversal of numbers would easily match twin and industry code (N=1) 5. Occupations that more closely matched reported industry codes were accepted over occupations that did not match reported industry codes (N=56) 6. If industry data provided by one twin was missing or generic, occupation data was accepted from the twin with non-missing or more specific data (N=69) Unable to be resolved (N=124) More than one rule applied (N=39)

Abstract (if available)

Abstract The purpose of this dissertation was to explore the relationship between tooth loss in young adulthood and dementia and cognitive decline in older adulthood in a sample of elderly twins. The study had two specific aims. The first aim was to determine whether tooth loss is associated with risk of developing dementia after accounting for low SES and other possible covariates. The second aim was to determine whether tooth loss is associated with baseline cognitive performance and cognitive decline on the modified Telephone Interview for Cognitive Status (TICS-m) after accounting for low SES and other possible covariates.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

More knowledge, better plans: a study of heterogeneity in dementia prevalence and mid-life cognitive changes

PDF

Accuray of subjective cognitive complaints in a longitudinal context: the effect of depression and dementia status

PDF

Life course implications of adverse childhood experiences: impacts on elder mistreatment, subjective cognitive decline, and caregivers' health

PDF

Contributions of dynamic cerebrovascular function to cognitive decline and dementia: development and validation of a novel neuroimaging approach

PDF

Cognitive health and self-rated memory in later life: linkages to race/ethnicity, multimorbidity, and survival

PDF

Self-perceptions of Aging in the Context of Neighborhood and Their Interplay in Late-life Cognitive Health

PDF

Cross-sectional association of blood pressure, antihypertensive medications, MRI volumetric measures and cognitive function scores in an aging population

Asset Metadata

Creator Watts Hall, Amber (author)

Core Title Predicting cognitive decline and dementia in elderly twins from indicators of early life oral health

School Leonard Davis School of Gerontology

Degree Doctor of Philosophy

Degree Program Gerontology

Publication Date 03/02/2009

Defense Date 12/04/2008

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

Tag cognitive decline,dementia,Inflammation,OAI-PMH Harvest,oral health,periodontal disease,tooth loss,Twins

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Gatz, Margaret (committee chair), Crimmins, Eileen M. (committee member), Finch, Caleb E. (committee member), Prescott, Carol A. (committee member)

Creator Email amberwat@usc.edu,amberwattshall@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1992

Unique identifier UC1447296

Identifier etd-Watts-2401 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-148420 (legacy record id),usctheses-m1992 (legacy record id)

Legacy Identifier etd-Watts-2401.pdf

Dmrecord 148420

Document Type Dissertation

Rights Watts Hall, Amber

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu