A Bayesian method for adverse effects estimation in observational studies with truncation by death.

Area of Science:

• Biostatistics
• Observational Studies
• Geriatric Medicine

Background:

• High mortality rates in observational studies of geriatric or severely ill patients complicate intervention effect evaluation.
• Outcome "truncation" by death hinders accurate comparison of adverse event prevalence between interventions.
• Existing methods like survivor average causal effect do not fully account for the interplay between adverse events and mortality.

Purpose of the Study:

• To propose a novel Bayesian method for handling outcome truncation by death in observational studies.
• To enable simultaneous comparison of interventions' effects on both mortality and adverse events across the entire study sample.

Main Methods:

• A Bayesian approach is used to impute unobserved mortality and adverse event outcomes for participants under the alternative intervention.
• A composite ordinal outcome is defined, integrating death and adverse events into a severity scale.
• The method is applied to compare heart failure incidence in geriatric Type II diabetes patients treated with sulfonylureas versus dipeptidyl peptidase-4 inhibitors.

Main Results:

• The proposed Bayesian method allows for a comprehensive analysis of intervention effects, including those who die during the study.
• The composite outcome facilitates a unified assessment of treatment impact on both mortality and adverse events.
• The application demonstrates the method's utility in a real-world clinical scenario involving diabetes medications.

Conclusions:

• The developed Bayesian imputation technique effectively addresses outcome truncation by death.
• This method provides a more complete understanding of intervention effects by considering the entire patient sample.
• The approach is valuable for comparative effectiveness research in high-mortality populations.