Preserving Trial Endpoint Specificity and Cause of Death Attribution in Cardiovascular Trials

Background

The MARINER (Medically Ill Patient Assessment of Rivaroxaban vs Placebo in Reducing Post-Discharge Venous Thrombo-Embolism Risk) trial examined the efficacy of rivaroxaban on venous thromboembolism (VTE) following discharge in high-risk medical inpatients. The trial did not meet its primary endpoint, in part due to a lesser effect of rivaroxaban on “VTE-related death” than on nonfatal VTE.

Objectives

The objective of this exploratory research was to examine the impact of more specific fatal VTE definitions on trial outcome HRs through readjudication of death endpoints.

Methods

Primary source documents for the 241 deaths in the MARINER trial were reviewed by blinded adjudicators not involved with the original trial. Prespecified definitions for VTE-related death were used, and “Death of Unknown Etiology” was allowed instead of the original endpoint “Cannot rule out pulmonary embolism.” Original event determinations for nonfatal events were used in this analysis. HRs and 95% CIs for rivaroxaban vs placebo were calculated for prespecified cardiovascular outcome composites.

Results

Rereviewed death cases showed strong concordance with original results, except deaths originally categorized as “Cannot rule out pulmonary embolism” were redistributed, largely to undetermined death (60%). The readjudicated MARINER primary endpoint using only confirmed fatal VTE events revealed a HR of 0.46 (95% CI: 0.23-0.91) vs the original HR of 0.76 (95% CI: 0.52-1.1).

Conclusions

This post-hoc, exploratory analysis of endpoint design demonstrates that designing specific trial endpoints can minimize the risk of type II error. In the trial design stage, it is important to preserve endpoint specificity to allow accurate hypothesis testing. Using standardized endpoint definitions, such as for VTE-related death, across trials can help achieve this goal.