Simultaneous mediation effect estimation of APOE through multiple neuropathological pathways on cognitive outcomes

Abstract

AbstractBackgroundIndividual associations of APOE genotype with Alzheimer’s disease neuropathologic changes (ADNC) have been studied extensively, but the effects through comorbid neuropathological pathways are rarely evaluated simultaneously. In addition, conventional statistical methods for mediation analysis rely heavily on assumptions on causal directions among different pathologies.MethodUsing the National Alzheimer’s Coordinating Center database, we use state‐of‐the‐art mediation analysis methods for multiple mediators effect decomposition to simultaneously estimate the indirect effects of the APOE ε4 allele through different pathologies, including ADNC, TDP‐43, and Lewy body disease, on cognitive outcomes. Unlike the conventional Baron and Kenny approach, we do not assume a causal structure between different pathologies, making causal interpretations under fewer assumptions. We control for cerebrovascular diseases, cardiovascular diseases, and age at death that are potentially affected by the APOE ε4 allele and could confound the pathologies.ResultOur results show that the APOE ε4 allele has a negative effect on cognitive outcomes (MoCA: ‐4.97 (‐6.92, ‐3.03); CDR Sum of Boxes: 1.34 (0.70, 1.98)). Most of the effect goes through ADNC (MoCA: ‐3.37 (‐4.68, ‐2.05); CDR Sum of Boxes: 1.20 (0.92, 1.49)), and there is evidence that the indirect effect through ADNC is significant for both outcomes. There is a small indirect effect through TDP‐43 pathology of the same direction as the total causal effects. It is significant for the standard CDR sum of boxes (0.26 (0.13, 0.39)) but not for the MoCA score (‐0.37 (‐0.81, 0.07)). There is little evidence of an indirect effect through Lewy body pathology for both outcomes. In addition, the effect that goes through the dependence or interaction between the pathologies is small.ConclusionWe utilize novel causal mediation analysis methods on neuropathology data to quantify the effects of APOE genotype on dementia risk through different ADRD pathologies without requiring knowledge of their causal relationship. Most of the effect of the APOE ε4 allele goes through ADNC, as expected, but there is also evidence of a smaller indirect effect through TDP‐43. This work provides insights into the understanding of the dementia disease process and introduces a new statistical tool for future AD and related dementias research.