Drawing causal inference from observational studies is challenging because the exposure-outcome relationship is likely confounded. Additional hurdles present themselves when there is a random time element to the outcome. While many biomarkers can be recorded in vivo, others can only be measured by analyzing tissues collected during autopsy. This is the case, for example, of many neurological biomarkers that require sampling brain tissues; outcome data collection can only occur when a participant dies. Comparing observed biomarker values across exposure groups can be highly misleading when the exposure under consideration affects survival. This occurs because such comparison ignores that observation times then tend to be different across exposure groups, and that biomarker values typically vary temporally across time. In this work, we present a causal inference framework for studying the effect of a point-exposure on a time-varying biomarker process that can only be sampled at death.