Let’s talk methodology…
In this randomized, open-label, non-inferiority trial, the authors analyze a primary composite outcome that includes recurrent VTE and major bleeding in cancer patients taking edoxaban, a factor X inhibitor developed by the study sponsor, Daiichi-Sankyo. They conclude that when compared with LMWH, edoxaban is non-inferior for the aforementioned composite outcome. Because composites occur more often than its components, using composites increase the study efficiency by decreasing the study time and sample size with subsequent costs reduction.
Non-inferiority trials are a useful tool when it is unethical to compare a new drug with placebo because a standard of care is available or because the trial involves serious outcomes such as mortality or in this case recurrence of VTE. Non-inferiority trials are usually confused with equivalence trials but these are different. Equivalence is applied mainly for comparison of generic versus marketed drug preparations to assess for pharmacokinetic differences. Otherwise, equivalence is impossible to be achieved between 2 drugs with different mechanisms of action such as edoxaban and LMWH. A non-inferiority approach may be used when the new treatment – edoxaban – seems to offer advantages such as better compliance, cost or availability over the standard of care – LMWH. Importantly, several considerations need to be accounted for when non-inferiority is part of the study design:
- As equivalence cannot be determined, a margin of non-inferiority should be established a-priori. This is defined as an acceptable margin of non-inferiority based on clinical judgment and statistical reasoning and it could be guided by previous trials that compared the control drug (LMWH) with placebo. In this case, the margin was a hazard ratio of 1.5 based on the unmet need for an alternative to parenteral LMWH and the advantages of oral therapy, allowing non-inferiority if the edoxaban population presented the composite outcome at 1.5 times the rate per unit time as the LMWH population or 50% the hazard.
- Importantly, the hazard ratio must be interpreted conjointly with time-to-event analysis or Kaplan-Meier curves. In this study, the curves cross for the composite primary outcome indicating non-inferiority, however, when interpretation is performed for each component, edoxaban appears to have less recurrent VTE but more major bleeding events when compared to LMWH. Once these 2 opposing components are combined into a composite, there is a sense of compensation.
- Intention-to-treat (ITT) analysis is indicated in superiority trials to preserve randomization – similar characteristics in both groups – avoiding attrition bias, however, including dropouts in non-inferiority trials can lead to bias towards equivalence. For this reason, in non-inferiority trials, it is recommended to perform both ITT and per-protocol analysis and both should demonstrate non-inferiority for it to be concluded. It this study a modified ITT analysis was performed including all patients who underwent randomization and received at least one dose of the assigned drug. These results were confirmed by sensitivity analysis of the primary outcome with a per-protocol approach.
In conclusion, non-inferiority trials should be cautiously executed and interpreted. Further studies with diverse factor X inhibitors are needed to validate these results and physicians must discuss risks and benefits of oral versus subcutaneous anticoagulation in cancer patients for whom quality of life might be the decisive variable.
Daniela Guerrero Vinsard MD.
Internal Medicine Resident at the University of Connecticut Health Center.