M. Maria Glymour’s research while affiliated with University of Massachusetts Boston and other places

Mixed evidence on how statins affect dementia risk may reflect variability in model specifications. Alternate specifications are rarely systematically compared. Using an emulated trial design framework, we investigated variation in the estimated effect of statin initiation on dementia across alternative (1) eligibility criteria, (2) confounding variable sets, and (3) outcome definitions. Kaiser Permanente Northern California members’ linked electronic health records from 1996 to 2020 were used to identify statin initiation and dementia diagnoses. Statin initiators were matched on age and low-density lipoprotein cholesterol with up to 5 non-initiators. Possible covariates included clinical (n = 1.4 million); socioeconomic and behavioral (n = 265,224); and genetic (n = 69,573) variables. Using Cox proportional-hazards models, we estimated variation across 1.27 million intent-to-treat estimates for statin initiation varying specification of eligibility, outcome definition, and covariates. Estimated hazard ratios (HRs) for statin initiation on dementia across all specifications ranged from 0.93 to 1.47. The variance of estimates due to model specification differences was 7.6 times larger than the average variance of specific estimates due to finite sample size. Three modeling decisions notably attenuated coefficients [ln(HR)]: requiring a run-in period prior to the emulated trial start date (0.034); adjustment for diabetes (0.030) and cardiovascular disease (0.039); and excluding the first year of follow-up (0.041). HRs from models with all three specifications ranged from 0.99 to 1.15. No specification we evaluated consistently generated protective effects. Estimates of the association between statin initiation and dementia leveraging real world data are sensitive to model specification, especially decisions related to clinical covariates and time-at-risk.

INTRODUCTION Intimate partner violence (IPV) victimization is highly common among women and associated with adverse health consequences that may be linked to dementia risk. METHODS Nurses’ Health Study II participants (N = 14,771) reported adult (age ≥ 18) emotional, physical, and sexual IPV in 2001/2008 and completed the Cogstate Brief Battery 2014–2019 (4/6 maximum assessments). Any versus no IPV and IPV subtypes were used to predict cognition in confounder‐adjusted generalized estimating equation models weighted to account for attrition. RESULTS Mean age at baseline was 61.0 years (standard deviation = 4.6); 46.5% reported any IPV (42.3% emotional, 22.6% physical, and 11.3% sexual). IPV victimization was associated with 0.029 SD unit (95% confidence interval [CI]: −0.068, 0.009) lower global cognitive score but not rate of cognitive change. Among IPV types, emotional IPV had the strongest association (β = −0.048; 95% CI: −0.075, −0.020) with cognitive scores. DISCUSSION Gendered social experiences such as IPV may influence dementia risk. Further assessment of IPV in aging cohorts is needed. Highlights IPV predicted lower average cognitive score over follow‐up. Emotional abuse had the largest associations with cognitive score among subtypes. We found no differences in rate of cognitive score change by violence exposure. Even modest impacts of violence would translate to large population effects. Gendered experiences warrant additional research in understanding dementia risk.

INTRODUCTION Decreases in body mass index (BMI) may be early consequences of Alzheimer's disease (AD) pathophysiological changes. Previous research in the UK Biobank estimated that AD‐related genes began affecting BMI around age 47. We assessed whether this result could be replicated using longitudinal data in an independent cohort. METHODS Using All of Us (AOU) (N = 197,619, aged 30+) data, we estimated linear mixed models for associations of Z‐scored AD‐Genetic Risk Score (AD‐GRS) with BMI, stratified by decade of age. We calculated the earliest age at which AD‐GRS was associated with differences in BMI using cross‐validated models adjusted for demographics. RESULTS Higher AD‐GRS was statistically associated with lower BMI in participants aged 60–70 (b = −0.060 [−0.113, −0.007]). Best fitting models suggested the inverse association of AD‐GRS and BMI emerged beginning at ages 47–54. DISCUSSION AD genes accelerate age‐related weight loss starting in middle age. Highlights Understanding when physiological changes from amyloid pathology begin is key for AD prevention. Our findings indicate that AD‐associated genes accelerate midlife weight loss, starting between 47 and 54 years. AD prevention research should consider that disease pathology likely begins by middle age.

Importance Detection bias occurs when an exposure is associated with a systematic difference in outcome ascertainment or diagnosis. For dementia research, diagnosed health conditions that bring patients into frequent interaction with health care may increase the chance that an individual receives a dementia diagnosis. Objective To evaluate potential detection bias or misdiagnosis bias in evaluation of clinical factors associated with dementia using electronic health record (EHR) data. Design, Setting, and Participants This prospective cohort study used data from 2 population-based volunteer cohorts: UK Biobank (UKB) and All of Us (AOU). Participants were aged 55 years or older, were dementia-free at baseline, and had linked EHRs. Participants in UKB were followed up from baseline (2006-2010) until December 2022, and in AOU, from baseline (2017-2022) until July 2022. Data were analyzed from November 2023 through February 2025. Exposures Diagnoses of type 2 diabetes, depression, hypertension, urinary tract infection, kidney stones, forearm fracture, and gastrointestinal (GI) bleeding. Main Outcomes and Measures Rate of incident all-cause dementia diagnosis from EHRs and associations between clinical exposures and incident dementia diagnosis, assessed using Cox proportional hazards regression models. Results Among 228 392 participants from UKB (n = 137 374; mean [SD] age at baseline, 62.5 [4.1] years; 53.8% female) and AOU (n = 91 018; mean [SD] age at baseline, 66.9 [7.8] years; 57.1% female), those with a history of a clinical exposure at baseline had higher dementia incidence rates compared with those without such history. For example, among participants with a history of GI bleeding, the dementia incidence rates were 3.0 (UKB) and 7.7 (AOU) per 1000 person-years compared with 2.2 (UKB) and 2.4 (AOU) per 1000 person-years among those without a history of GI bleeding. All exposures were significantly associated with incident dementia, with hazard ratios (HRs) ranging from 1.18 (95% CI, 1.00-1.40) to 3.51 (95% CI, 3.08-4.01). Risk of incident dementia was typically highest in the first year following exposure diagnosis and attenuated thereafter. For example, in the first year after GI bleeding, there were larger elevations in risk of incident dementia (HR, 2.17 [95% CI, 1.46-3.22] in UKB; HR, 2.56 [95% CI, 1.62-4.04] in AOU) compared with 1 to 5 years after bleeding (HR, 1.46 [95% CI, 1.15-1.86] in UKB; HR, 2.14 [95% CI, 1.63-2.81] in AOU). Conclusions and Relevance In this cohort study of 2 large datasets, diagnoses of several conditions associated with varying risks of dementia were associated with a higher short-term likelihood of dementia diagnosis. This finding suggests that diagnostic bias or misdiagnoses may lead to spurious associations between conditions requiring clinical care and subsequent dementia diagnoses.

Positive childhood experiences (PCEs) have unknown effects on late life cognition and Alzheimer’s Disease biomarkers. We examined 406 Asian, 1179 Black, 349 Latinx, and 498 White KHANDLE and STAR study participants with data on PCEs, longitudinal cognitive measures, MRI (n = 560), and amyloid PET (n = 281). We conducted mediation and multigroup models within the structural equation modeling framework allowing us to examine the direct association of PCEs with episodic memory level and change as well as the indirect effects of PCEs through education. We additionally conducted linear regressions examining the association of PCEs with MRI and amyloid PET outcomes. Average participant age was 74 (53–90) and 62% were female. Overall, PCEs were positively associated with memory intercept and change. Education significantly mediated the association between PCEs and memory intercept. PCEs were not associated with hippocampal volume or amyloid burden in the combined sample or across individual ethnocultural groups. PCEs are positively related to episodic memory through the promotion of educational attainment.

INTRODUCTION Mixed evidence on how statin use affects risk of Alzheimer's disease and related dementias (ADRD) may reflect heterogeneity across sociodemographic factors. Few studies have sufficient power to evaluate effect modifiers. METHODS Kaiser Permanente Northern California (KPNC) members (n = 705,061; n = 202,937 with sociodemographic surveys) who initiated statins from 2001 to 2010 were matched on age and low‐density lipoprotein cholesterol (LDL‐C) with non‐initiators and followed through 2020 for incident ADRD. Inverse probability‐weighted Cox proportional hazards models were used to evaluate effect modification by age, gender, race/ethnicity, education, marital status, income, and immigrant generation. RESULTS Statin initiation (vs non‐initiation) was not associated with ADRD incidence in any of the 32 subgroups (p > .05). Hazard ratios ranged from 0.964 (95% CI: 0.923 to 1.006) among Asian‐identified participants to 1.122 (95% CI: 0.995 to 1.265) in the highest income category. DISCUSSION Sociodemographic heterogeneity appears to have little to no influence on the relationship between statin initiation and dementia. Highlights The study includes a large and diverse cohort from Kaiser Permanente (N = 705,061). An emulated trial design of statin initiation on dementia incidence was used. Effect modification by sociodemographic factors was assessed. There were no significant Alzheimer's disease and related dementias (ADRD) risk differences in 32 sociodemographic subgroups (p > 0.05).

For many important questions about influences on clinical and public health outcomes, no single study can provide a decisive answer. The perfect study—a large, diverse, well-conducted trial randomizing all relevant versions of a treatment and comprehensively tracking all relevant health outcomes—is never feasible. Instead, we must draw conclusions by piecing together evidence from multiple imperfect studies. A systematic framework for combining disparate, complementary sources of evidence is emerging. We introduce this framework, called evidence triangulation; summarize key approaches based on delineating likely biases due to confounding, measurement, and selection; and review some methods for combining evidence. We illustrate the issues using the example of estimating the effects of alcohol use on dementia. The central tenet of evidence triangulation is to identify the most important weaknesses for any given study approach (and for each specific study applying that approach) and, if necessary, to identify which new sources of evidence that do not share these weaknesses are required. Almost certainly, the new studies will have weaknesses, but when results are consistent across studies that rest on different assumptions, and for which biases should be unrelated, the conclusions are on much sturdier ground.

Background: Many studies link adverse neighborhood context with racial myocardial infarction (MI) disparities. Research may reflect strong publication bias for chance associations in main or race-specific effects. Hypothesis: We theorized that associations would vary in direction as well as magnitude across two national cohorts due to differences in sample composition, outcome ascertainment, and parent study sampling schema. Methods: We compared results from the REasons for Geographic and Racial Differences in Stroke (REGARDS, n=25,143, aged ≥ 45 years, 42% Non-Hispanic (NH) Black; 2003-2018) study to the Health and Retirement Study (HRS, n=14,1941, aged > 50 years, 13% Non-Hispanic Black; 2004-2018). We estimated Cox models predicting MI for 51 American Community Survey (ACS) census tract (CT) variables and evaluated consistency of main and racially-stratified estimates between cohorts. Results: Follow-up in REGARDS (median=11.5 years; IQR: 6.5, 13.6) was similar to HRS (median=13.1; IQR: 8.3, 14.1), as was cumulative MI incidence (6.2% and 7.1%). The proportions of NH White and NH Black adults were at least moderately collinear in both samples ( r among White REGARDS participants-0.95, among Black REGARDS participants=-0.94, among White HRS participants= -0.63, and among Black HRS participants-=0.82). Sixteen ACS variables had estimated effects with incident MI that differed by at least 0.05 between cohorts. The estimated effect of the percentage of adults with less than a high school diploma was stronger in REGARDS (HR per SD: 1.10; 95% CI: 1.04, 1.17) than HRS (HR per SD: 1.04; 95% CI: 0.94, 1.15); the estimated effect of the percentage of residents in poverty was attenuated in REGARDS (HR per SD: 1.05; 95% 1.00, 1.11) compared to HRS (HR per SD: 1.17; 95% 1.07, 1.27). Differences in the estimated effects for 12 ACS variables across racial strata (p<0.05) identified in REGARDS were not corroborated in HRS. Conclusions: Neighborhood socioeconomic associations with MI across two national studies broadly replicated in direction but differed in magnitude. Inadequate statistical power and sample differences at the level of participant as well as of neighborhood likely contributed to inconsistent estimated effects by race.

Introduction: Though poor oral health is linked to elevated risk of later-life cardiovascular disease (CVD) and Alzheimer’s Disease and Related Dementias (ADRD) via systemic inflammation, the impact of financial barriers to oral care on CVD and ADRD incidence remains unquantified. Hypothesis: We hypothesized that older adults reporting financial barriers to oral care -particularly women, people marginalized due to racial/ethnic identity, or with periodontitis - have elevated risk for CVD and ADRD. Methods: We used data from All of Us participants aged at least 55 years (N=65,770, 77.2% Non-Hispanic White, 57.9% women). Our primary exposure was self-reported difficulty accessing oral care due to cost within the past year, defined as yes or no. We identified congestive heart failure (CHF), cerebrovascular accident (CVA), myocardial infarction (MI) and ADRD outcomes in linked electronic health records. Using Cox proportional hazards models, we estimated associations of difficulty accessing dental care with each outcome. We assessed heterogeneity by gender, racial/ethnic identity, and prevalent periodontitis diagnosis. We estimated the population attributable fractions (PAF) of each outcome associated with difficulty accessing oral care using national prevalence estimates. Results: Participants who reported difficulty affording oral care had a higher incidence of CVA (adjusted HR=1.72; 95% CI: 0.72, 3.86) and CHF (HR=1.22; 95% CI: 1.01, 1.47). Estimated effects on MI (HR=1.11; 95% CI: 0.85, 1.44) and ADRD (HR=1.02; 95% CI: 0.61, 1.69) were not robust to adjustment for demographic, health behavioral, and clinical confounders. We did not observe differences in these associations by race/ethnicity, gender, or periodontitis. The estimated PAFs suggest that eliminating financial barriers to oral care could, over three years, prevent 9% of incident stroke, 4% of MI, 6% of CHF, and 4% of ADRD. Conclusion: Addressing financial barriers to needed oral care remains an unrealized policy target for CVD and ADRD prevention.