Neighborhood factors and incident stroke: reproducibility of associations across two nationwide cohorts

Research linking adverse neighborhood context with disparities in incident stroke may reflect publication bias for chance associations. We compared results from the REasons for Geographic and Racial Differences in Stroke (REGARDS, n = 25,126, aged ≥45 years, 41 % Non-Hispanic Black, 12 % < high school degree, 2003-2022) study to the Health and Retirement Study (HRS, n = 12,969, aged >50 years, 15 % Non-Hispanic Black, 20 % < high school degree, 2004-2022). We estimated Cox models predicting stroke for 44 census tract variables representing demographic, socioeconomic, labor force, and housing conditions, evaluating inter-cohort consistency of main and race-stratified estimates. Follow-up in REGARDS (median = 12.1 years; IQR: 6.5, 14.9) was similar to HRS (median = 12.6; IQR: 7.0, 17.7). Cumulative stroke incidence was lower in REGARDS (6.6 %, adjudicated) than HRS (15.4 %, reported). Census tract-level household income, median rent, and home values were higher in HRS. Thirty-two of the 44 census tract variables evaluated had associations with incident stroke that differed between cohorts by less than log(0.05). For example, the proportion of housing units built 1980-1999 was associated with stroke incidence in both REGARDS (HR per SD = 0.94 [95 % CI: 0.88, 0.99)] and HRS (HR per SD = 0.92 [95 % CI: 0.87, 0.99) after adjustment for individual-level confounders and state of residence. Five predictors had significant (p < 0.05) interactions with race in HRS, but none of these interactions replicated in REGARDS. Strengthening the evidence base linking neighborhood disadvantage with stroke disparities is essential. Systematic exploration of how heterogeneity in sample composition and outcome ascertainment contributes to diverging findings across studies is needed.