Assessment of non-inferiority with meta-analysis: example of hypofractionated radiation therapy in breast and prostate cancer

Jane-Chloé Trone,Silvy Laporte,Patrick Mismetti,Paul Jacques Zuffrey,Edouard Ollier,Michel Cucherat,Nicolas Magné,Céline Chapelle

doi:10.1038/s41598-020-72088-2

Abstract

The aim of this study was to propose a methodology for the assessment of non-inferiority with meta-analysis. Assessment of hypofractionated RT in prostate and breast cancers is used as an illustrative example. Non-inferiority assessment of an experimental treatment versus an active comparator should rely on two elements: (1) an estimation of experimental treatment’s effect versus the active comparator based on a meta-analysis of randomized controlled trials and (2) the value of an objective non-inferiority margin. This margin can be defined using the reported effect of active comparator and the percentage of the active comparator’s effect that is desired to be preserved. Non-inferiority can then be assessed by comparing the upper bound of the 95% confidence interval of experimental treatment’s effect to the value of the objective non-inferiority margin. Application to hypofractionated RT in breast cancer showed that hypofractionated whole breast irradiation (HWBI) appeared to be non-inferior to conventionally fractionated RT for local recurrence. This was not the case for accelerated partial breast irradiation (APBI). Concerning overall survival, non-inferiority could not be claimed for either HWBI or APBI. For prostate cancer, the lack of demonstrated significant superiority of conventional RT versus no RT precluded any conclusion regarding non-inferiority of hypofractionated RT.

Highlights

The aim of this study was to propose a methodology for the assessment of non-inferiority with meta-analysis
The risk ratio (RR) for local recurrence was 0.50 and the RR for OS was 0.92 in favor of RT
To calculate the non-inferiority margins for local recurrence corresponding to 50% of the effect of conventionally fractionated RT, we considered both the value of HRCRTVsCtrl (0.50) and the value of the upper bound of its 95% confidence interval (95% CI) UBCRTVsCtrl (0.55)

Summary

Introduction

The aim of this study was to propose a methodology for the assessment of non-inferiority with meta-analysis. Non-inferiority assessment of an experimental treatment versus an active comparator should rely on two elements: (1) an estimation of experimental treatment’s effect versus the active comparator based on a meta-analysis of randomized controlled trials and (2) the value of an objective non-inferiority margin. This margin can be defined using the reported effect of active comparator and the percentage of the active comparator’s effect that is desired to be preserved. Traditional superiority trials are not well suited to evaluate efficacy of such treatments and non-inferiority trials have been proposed The latter aim to demonstrate that a new treatment is not “inferior”, not “unacceptably worse” to the standard by a too large amount. The assessment of non-inferiority from the literature faces two issues: (1) the choice of the non-inferiority margin and (2) the non-adapted design of most clinical studies

Objectives

Methods

Results

Conclusion