Aging
Causal estimands for cognitive change in the presence of truncation by death L. Paloma Rojas-Saunero* L. Paloma Rojas-Saunero Yixuan Zhou Joan A. Casey Lan Wen Elizabeth Rose Mayeda
Alzheimer’s disease and Alzheimer’s disease-related dementias present a growing public health burden. Research efforts that aim to identify causal mechanisms and potential targets of intervention to reduce dementia risk often estimate the effect of the exposure of interest on the rate of change in cognitive function. A major challenge in these studies is that death acts as a truncation event, and cognitive function becomes unobservable. There is limited work defining causal estimands in this setting, and current literature relies on linear mixed models, generalized estimating equations with inverse probability of censoring weights, or joint longitudinal-survival models, with little attention to interpretation of “rate of change” in the context of truncation by death. This work aims to discuss potential estimands for cognitive change as the outcome of interest in the presence of truncation by death, including causal effects under the hypothetical scenario where death is prevented, the survivor average causal effect, causal separable effects, and the estimand conditional on survival (i.e., “while alive”). We outline identifiability assumptions and identification strategies, and clarify the interpretation of estimates. To illustrate and compare estimated effects, we conducted a case study using data from the U.S. Health and Retirement Study and U.S. life tables to simulate alternative data generation mechanisms based on directed causal graphs, using effects of early-life and late-life environmental exposures on cognitive change over 20 years