In sensitivity analyses for cohort studies, which of the following is a typical approach?

• A. Changing exposure definitions, lag times, and subgroups
• B. Relying only on the primary exposure definition
• C. Randomizing exposure post hoc
• D. Ignoring time-to-event data

Answer: (A)

Sensitivity analyses in cohort studies test how robust the observed association is to plausible changes in the analytic approach. A typical approach is to vary exposure definitions, lag times, and subgroup analyses. By changing how exposure is defined (for example, using different exposure windows or intensity thresholds), we check whether misclassification or measurement error could be driving results. Adjusting lag times tests whether the presumed delay between exposure and outcome affects the estimate, which matters for diseases with incubation or latency. Exploring results within subgroups helps assess consistency and potential effect modification, indicating whether findings are generalizable or context-specific. If estimates remain similar across these variations, the findings are shown to be more robust to reasonable assumptions.

Relying only on the primary exposure definition would miss these robustness checks. Randomizing exposure post hoc is not an ethical or feasible option in observational cohorts and is not a typical sensitivity analysis. Ignoring time-to-event data discards valuable information and is not representative of how these analyses are conducted.