Methods to evaluate overall and mediated treatment effects in the presence of death
Embargo Date
2027-02-12
OA Version
Citation
Abstract
In clinical practice, treatment decisions are often based on both the probability of survival and other clinical outcomes, such as Quality of Life (QoL) scores or neurocognitive test scores. When interest in a research study lies in both survival and other clinical outcomes, death before the follow-up assessment results in undefined clinical outcomes. Many authors refer to this setting as “truncation by death” to distinguish it from settings where the outcome is simply missing. It is well known that analyses without careful consideration of undefined outcomes due to death can lead to flawed treatment decisions. In this thesis, we focus on methods to evaluate overall and mediated treatment effects in the presence of death in three different projects. First, we advocate not always treating death as a mechanism through which clinical outcomes are missing or censored, but rather as part of the outcome measure. We propose summarizing the clinical benefit of a treatment by combining death and the clinical outcome into a composite outcome. We introduce the survival-incorporated median, the median of the composite outcome, as a simple and useful summary measure to inform clinical practice. We illustrate the application of the survival-incorporated median by analyzing a study that compares the clinical benefit of two treatment regimens in prostate cancer patients. Second, we develop an estimation method for the survival-incorporated median using observational data for a causal interpretation. Combining Inverse Probability of Treatment Weighting and a quantile estimation procedure, we propose an estimator in settings with point treatment and time-varying treatment. We prove consistency for the proposed estimator. Simulations demonstrate the usefulness of this estimation procedure in various settings. We apply the method to invetigate the effect of statins on cognitive changes in elderly individuals. Last, we investigate the relationship between Apolipoprotein E (APOE), lipids, and cognitive function using data from the Long Life Family Study (LLFS) and the New England Centenarian Study (NECS). We use a linear mixed model to estimate the rate of change in the Telephone Interview for Cognitive Status (TICS) score for each participant. We perform mediation analysis to examine both the direct effect of APOE and its indirect effect through lipids on the rate of change in TICS score. We discover no significant evidence of an indirect effect of APOE through lipids, but a significant direct protective effect of APOE e2 on the rate of change in TICS score.
Description
2024
License
Attribution 4.0 International