Modeling APOE, morbidity, and mortality: a reproducibility challenge for genetic epidemiology

Erling Häggström Gunfridsson^*

• Centre for Demographic and Ageing Research (CEDAR), Umeå University, Umeå, Sweden

APOE is among the most extensively studied genetic loci in research on aging, morbidity, and mortality. Despite its well-established biological roles, empirical findings on the association between APOE and mortality remain inconsistent across studies. This heterogeneity is often attributed to biological complexity. In this Perspective, we argue that much of the apparent inconsistency instead reflects differences in modeling choices, variable definitions, and reporting practices, resulting in limited reproducibility and comparability. We highlight how pleiotropy, age-dependent effects, and selective survival make APOE particularly sensitive to analytical decisions. We focus on three underappreciated sources of irreproducibility: selective exclusion of rare APOE genotypes, lack of standardized baseline models, and routine adjustment for variables that are not confounders under Mendelian inheritance. We argue that all observed APOE genotypes should be included in primary analyses, that parsimonious baseline models adjusted only for variables independent of genotype should always be reported, and that overadjustment can obscure rather than clarify genetic effects. We propose a set of conceptual principles to improve reproducibility in studies of APOE, morbidity, and mortality, with implications for genetic epidemiology more broadly.