Figure 3.

Modeling cancer as a time-varying predictor eliminates the inverse relationship between cancer and Alzheimer’s disease (AD). Hazard rate ratios for the effect of cancer diagnosis on AD estimated using Cox regression. Panels A and B are run separately by sex and stratified by birth year. No additional covariates are included in the models. Panel (A) treats cancer as time-independent predictor and Panel (B) treats cancer diagnosis as time-varying predictor. TVEM = time-varying effect model (Models 1 and 2 in Figure 4). Source: Utah Population Database, Utah Cancer Registry, Centers for Medicare and Medicaid Services.