Optimism is not associated with two indicators of DNA methylation aging

Sample description

The WHI sample consisted of 3,298 women including 1,665 Whites, 961 Blacks, and 561 Hispanics, and 111 in the “Other” race/ethnicity category. Chronological age in the WHI sample ranged from 50-79 years (mean=64). Most women were married (or in marriage-like relationships: 56%), and most had education after high school (some college or an associate degree: 26%, or a college or graduate degree: 30%). The NAS sample consisted of 514 men (99% White) ranging in chronological age from 56-91 years (mean=73). Most men were married (76%), and had either a high school degree (20%) or college or graduate degrees (35%). Table 1, Table S1, and Table S2 [see Appendix A 1 for Table S1 and Table S2] provide additional details about participants.

Optimism and DNA methylation age

In WHI, when considering both dispositional optimism and DNAm age as continuous variables, we observed no association of optimism with IEAA (mean difference, β=0.07; 95% CI: -0.10,0.24; Table 2); however, we did observe an association with EEAA (mean difference, β=-0.35; 95% CI: -0.55,-0.13) in unadjusted models such that higher optimism was associated with lower DNAm age. In models that adjusted for basic confounders, we observed no relation of optimism to IEAA (β=0.02; 95% CI: -0.16,0.19; Table 2) and associations with EEAA were attenuated and no longer statistically significant (β=-0.06; 95% CI: -0.28,0.16). After further adjusting for the full set of covariates, associations were not meaningfully different (Table 2).

When evaluating the association between dispositional optimism and DNAm age in NAS, we observed a pattern of null findings in unadjusted as well in all the other models. For example, in our basic adjusted models, optimism was not associated with IEAA (β=-0.06; 95% CI: -0.55,0.42; Table 3) or EEAA (β=-0.23; 95% CI: -0.83,0.38).

We also evaluated quartiles of optimism in relation to both EEAA and IEAA, in the WHI and NAS. Findings were null in both cohorts for IEAA. In the WHI women, in the unadjusted model, individuals in the lowest (versus highest) quartile of optimism had a lower mean EEAA score (β=-0.73; 95% CI: -1.32,-0.15; Table 3). However, this association was no longer statistically significant after adjusting for additional covariates, and all findings with EEAA were null in the NAS.

Additional analyses

In both the WHI and NAS, after excluding individuals with depression, associations for optimism with either IEAA and EEAA remained consistently null (Appendix A 1, Table S3 and Table S4). Further, no statistically significant associations with IEAA and EEAA were evident for either the pessimism or the optimism subscales. Associations between the explanatory style optimism and both IEAA and EEAA were also uniformly null in all models (Table 4). In analyses stratified by race in WHI (Black and White women), associations remained consistently null across strata (data not shown). Findings from analyses that included the 736 WHI women who did not have the full set of covariates, were similarly null (data not shown). Finally, in WHI, after additionally adjusting for case-control status, WHI observational study membership, as well as clinical trial membership findings were uniformly null across all models (data not shown).

Study Population

Women’s Health Initiative

The Women’s Health Initiative (WHI) is a long-term study focused on identifying strategies for preventing major chronic diseases in postmenopausal women. Starting in 1993, racially and ethnically diverse women aged 50-79 years were recruited throughout the U.S. and entered either clinical trial(s) (WHI-CT; N=68,132) or the observational study (WHI-OS; N=93,676). Data were collected at 40 clinical centers throughout the country. At baseline, women completed self-administered questionnaires including information about sociodemographic factors, psychosocial characteristics, health behaviors, and chronic conditions. Further, after fasting overnight, they visited study clinics where certified clinical center staff collected blood specimens and performed anthropometric measurements.

The present study draws on data from two WHI substudies. The first substudy (Epigenetic Mechanisms of PM-Mediated Cardiovascular Disease Risk (EMPC; AS315; n=2,200) included a stratified random sample representative of the larger WHI CT population [27]. The second substudy (Integrative Genomics and Risk of Coronary Heart Disease and Related Phenotypes in the Women’s Health Initiative; BAA23; n=2,107) examined genomic determinants of coronary heart disease (CHD), using a nested case-control design (with oversampling of African Americans and Hispanics) in the WHI observational study and CT study populations. When data from both substudies were combined, there were 118 women who were in both substudies; thus, only data from EMPC was used when overlap existed (results were very similar when data from BAA23 were used instead). Further, 148 women were excluded because they had missing data on optimism and a further 7 were excluded due to missing data on the DNA methylation measures. Finally, some covariates were assessed only at baseline, and because some women had their blood drawn after baseline (n=736) they were excluded from primary analyses, resulting in a final analytic sample of 3,298 WHI women. In secondary analyses, we evaluated the main association of interest after including the 736 women, without fully control for covariates.

VA Normative Aging Study

The VA Normative Aging Study (NAS) is a longitudinal investigation of normal aging processes in community-dwelling men, initiated in 1963. The study enrolled healthy men aged 21 to 81 years who were free of known chronic medical conditions. Men provided information on demographic factors, medical history, psychosocial factors, and lifestyle factors on a regular basis. They were interviewed at the VA Boston hospital every 3-5 years, and also participated in a physical exam and laboratory tests. Blood was drawn at each physical exam. Eligibility for this study required continued participation as of the time when DNA samples were first collected. Dropout has been < 1% per year in this cohort. Among the active 1,749 NAS participants at the time, DNA samples were collected from 1999-2009 for 774 participants. Among these, we excluded from analysis 260 men who did not have an optimism measurement before a DNA sample was taken. The final analytic sample included 514 NAS men with both DNA and measures of optimism.

Because restrictions apply to the public availability of these data, they are available from the WHI and NAS study coordinators upon reasonable request.

Measures

Optimism assessment

WHI. In WHI, dispositional optimism was assessed at baseline using the Life Orientation Test-Revised (LOT-R). The measure has demonstrated good discriminant and convergent validity, as well as good reliability [33]. Negatively worded items were reverse coded, and then all items were summed to create a composite score that ranged from 6 to 30, with higher scores indicating higher optimism. Following standard WHI protocol, the score was set to missing if study respondents were missing any of the LOT-R items. To facilitate comparisons of effect sizes across studies, we standardized optimism scores (Mean(M)=0, Standard Deviation(SD)=1). Internal consistency reliability was high in the analytic sample at baseline (Cronbach α=0.75). To assess the possibility of discontinuous or threshold effects, we also created quartiles of optimism based on the score distribution in the sample. Mean optimism scores by quartile were: 19, 23, 24, and 27. Following prior work on dispositional optimism [34], in secondary analyses, we also evaluated the optimistic (Cronbach α=0.77) and pessimistic (Cronbach α=0.74) subscales of LOT-R, each composed of three items from the overall scale.

NAS. In NAS, dispositional optimism was assessed using both the Revised Optimism-Pessimism Scale (PSM-R) and the original Life Orientation Test (LOT). The LOT is the parent scale from which the LOT-R was largely derived and includes 8 items that contribute to the scale score. Items were reverse coded as necessary and summed to create a composite score that ranged from 7 to 32 in this sample, with higher scores indicating higher optimism. Again, we standardized optimism scores (M=0, SD=1). Internal consistency reliability was high in the analytic sample at baseline (Cronbach α=0.78). We created quartiles of optimism based on the score distribution in the sample and mean optimism scores by quartile were: 16, 20, 22, and 26. In secondary analyses, we evaluated the optimistic (Cronbach α=0.75) and pessimistic (Cronbach α=0.80) subscales of LOT, each composed of four items from the overall scale.

Another measure of optimism based on explanatory style was assessed using the PSM–R, developed and validated by Malinchoc, Offord, and Colligan [35,36]. This bipolar scale measures the way individuals explain the causes of both good and bad events, and characterized explanatory style on a continuum from optimistic to pessimistic, by using 263 items selected from the revised Minnesota Multiphasic Personality Inventory (MMPI–2). Following the scoring algorithm, items were combined into a composite bipolar score reflecting a more optimistic explanatory style at the low end of the continuum and a more pessimistic one at the high end. This measure has predicted reduced risk of heart disease and slower lung function decline in the NAS cohort [3,37]. Internal consistency reliability was high in the analytic sample at baseline (Cronbach α=0.78), and prior research demonstrates this scale has high test-retest reliability of 0.90 [36]. To assess the possibility of discontinuous or threshold effects, we also created quartiles of optimism based on the score distribution in the sample. Mean scores by quartile were: 58, 47, 41, and 32.

Assessment of DNA methylation and DNA methylation age acceleration

WHI. Methylation analysis for both substudies was performed using the Illumina Infinium Human-Methylation450 Beadchip, which measures single-CpG resolution DNA methylation levels at 485,577 unique CpG sites in the human genome. The BAA23 WHI substudy samples were processed at the HudsonAlpha Institute of Biotechnology. The EMPC WHI substudy samples were processed at the Northwestern University Genomics Core. All measurements were quality-controlled and batch adjusted as described elsewhere [27].

NAS. Methylation data were generated at the Northwestern University Genomics Core Facility. All measurements underwent quality-control. The Bioconductor minfi package Illumina-type background correction without normalization was used to preprocess the samples and generate methylation beta values to compute DNAm-age [38].

Epigenetic age acceleration scores. In all of the analyses, we considered both the intrinsic epigenetic age acceleration score (IEAA) [16] and the extrinsic epigenetic age acceleration score (EEAA) [15]. The two scores were calculated identically in WHI and NAS based on scripts developed by Horvath (the IEAA variable is denoted as AAHOAdjCellCounts and the EEAA variable is denoted as BioAge4HAStaticAdjAge). Both the EEAA and IEAA measures represent accelerated biological aging that exceeds chronological age, with positive values indicating that epigenetic age is higher than chronological age. Both measures are calculated by obtaining the residual when regressing DNAm age on chronological age. Intrinsic epigenetic age acceleration captures properties of aging that are independent of cell type and organ whereas extrinsic epigenetic acceleration likely reflects both epigenetic variation and age-related changes in cell distributions in blood [27]. In WHI, the correlation between IEAA and EEAA was 0.36 and in NAS the correlation was 0.32.

Assessment of potential confounders and other covariates

WHI. Potential confounders included sociodemographic factors and depression. Sociodemographic variables were obtained from the baseline questionnaire and included age (continuous), race (White, Black/African-American, Hispanic/Latino, Other), marital status (married/marriage like-relationship, divorced/single, widowed), education (less than high school, high school graduate, some college or associate degree, college or more), income (<$20,000, $20,000-$49,999, $50,000-$74,999, $75,000+, don’t know). Depression status (yes/no) was defined according to Burnam Screening Algorithm questionnaire that includes 6 items from the Center for Epidemiologic Studies Depression Scale (CES-D) and two from the Diagnostic Interview Scale (DIS) [39,40], with a cutoff of ≥0.06 indicating depression [39]. Health conditions were self-reported (yes/no) and included: hypertension, high cholesterol, cardiovascular disease, diabetes, stroke, cancer. Height and weight were measured by trained staff and used to calculate Body Mass Index (BMI) in kg/m². and then, three BMI categories were created (<24.9, 25.0 to 29.9, ≥30.0). Potential confounding or intermediate variables included the following health behaviors, all self-reported on the baseline questionnaire: cigarette smoking (never, former, current smoker), physical activity (weekly expenditure of metabolic equivalent tasks (METs; <3 METs/week, 3-<9 METs/week, 9-<19 METs/week, 18-<27 METs/week, 27+ METs/week), alcohol intake (non-drinker, current drinker), diet (122-item food frequency questionnaire (FFQ) [41]; overall diet quality was quantified using the Alternative Healthy Eating Index (scale 0-110; the AHEI includes 11 different diet components).

NAS. Potential confounders included sociodemographic factors, depression, and chronic conditions. Sociodemographic variables were obtained from the baseline questionnaire and included age (continuous), educational attainment (years), race/ethnicity (white/non-white), marital status (married/not married). Depressive symptoms were assessed with a subscale from the Brief Symptom Inventory [42], with a cutoff of ≥0.638 indicating depression. Information about health conditions was self-reported (yes/no), updated every 3-5 years, and included: coronary heart disease, stroke, diabetes, and cancer. BMI in kg/m² was calculated from weight and height measured by study staff (<24.9, 25.0-29.9, ≥30.0). Further variables included the following health behaviors (updated every 3-5 years), which could be confounders or intermediates: cigarette smoking status (current, former, never), physical activity (created based on questions asking about energy expenditure (e.g., frequency of sports activities, flights of stairs climbed/day, distance walked, etc., to calculate the following metabolic equivalent of tasks (METs) categories; <3 METs/week, 3-<9 METs/week, 9-<19 METs/week, 18-<27 METs/week, 27+ METs/week), alcohol intake, (<2 drinks versus 2+ drinks), and diet (frequency of fruit and vegetable consumption).

Statistical analysis

In primary analyses, we considered optimism as a continuous standardized variable, where associations represent the change in DNAm age as a function of a 1 standard deviation increase in optimism, and also evaluated optimism categorized into quartiles (based on the sample-specific distribution of scores). All models were run separately within each cohort. Several sets of models were tested using ordinary least square regression. The first model did not adjust for any covariates. The second model added basic potential confounders including race, education, income, marital status, chronic conditions, and depression. A third model further added variables that could be potential confounders or intermediates including physical activity, smoking, BMI, diet, and alcohol consumption.

We conducted several secondary analyses. First, to test potential residual confounding due to depression, we excluded those with high levels of depressive symptoms. Second, to evaluate whether associations differed by the optimism or pessimism subscale, we separately evaluated these two subscales. Third, to evaluate the consistency of results across different measures of optimism, we conducted analyses in NAS using the PSM-R instead of the LOT. Fourth, In WHI we conducted stratified analyses in Black and White women. Fifth, we evaluated findings after adding to our sample 736 WHI women who had blood draws after the baseline assessment, and who did not have full covariate data. Finally, to evaluate potential confounding caused by case-control status (CHD-no CHD over follow-up), WHI observational study membership, or clinical trial membership (hormone therapy (HT), dietary modification (DM), or calcium and Vitamin D supplementation (CaD)), we adjusted for all of these factors in the WHI cohort.

All analyses were completed using Stata (StataCorp. 2017. Stata Statistical Software: Release 15.0. College Station, TX: StataCorp LP) or R 3.4.1 (R Core Team (2017). R Foundation for Statistical Computing, Vienna, Austria).

Ethics Approval

For WHI, institutional review board approval was obtained at each clinical center and all participants provided written informed consent. For NAS, participants provided written informed consent to the VA Institutional Review Board.

Availability of data and materials

The data that support the findings of this study are available upon reasonable request to Ron Spiro (aspiro3@bu.edu) for NAS data. Details regarding access to all WHI data are available at: https://www.whi.org/researchers/data/Pages/Home.aspx."