Termination Based on Event Accrual in Per Protocol Versus Intention to Treat in the ROCKET AF Trial

Area of Science:

• Clinical Trials Methodology
• Cardiovascular Research
• Pharmacology

Background:

• Event-driven clinical trials rely on accumulating primary efficacy events for termination.
• Determining the optimal cohort for tracking events in noninferiority trials with conditional superiority examination is crucial.
• The ROCKET AF trial provides a relevant case study for evaluating different event-tracking strategies.

Purpose of the Study:

• To compare the impact of tracking primary efficacy events in the per-protocol cohort versus the intention-to-treat (ITT) cohort on clinical trial termination and results.
• To analyze how different termination points affect noninferiority and superiority conclusions in the ROCKET AF trial.

Main Methods:

• Utilized data from the ROCKET AF trial (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation).
• Identified the date when 429 primary efficacy events occurred in the ITT cohort during the ITT period, mirroring the per-protocol cohort termination.
• Performed hypothetical noninferiority and superiority analyses based on this earlier ITT-based termination date.

Main Results:

• Termination of the ROCKET AF trial could have occurred 226 days earlier if events were tracked in the ITT cohort.
• Rivaroxaban would have met noninferiority (HR, 0.77; 95% CI, 0.62-0.96) and, unlike the original analysis, superiority (HR, 0.82; 95% CI, 0.68-0.99) if terminated based on ITT events.
• Safety profiles, including lower intracranial hemorrhage risk with rivaroxaban, remained consistent across both termination scenarios.

Conclusions:

• Clinical trial termination based on ITT event accrual versus per-protocol can significantly influence trial outcomes.
• The choice of cohort for event tracking is critical, especially when considering drug discontinuation rates and off-treatment event rates.
• This finding has implications for the design and interpretation of future event-driven clinical trials, particularly those assessing noninferiority with conditional superiority.