Using a single non-inferiority margin or a single preserved fraction for an entire pharmacological class may be invalid for the analysis of non-inferiority: A case study of statin non-inferiority trials

Abstract

Background: The use of a single non-inferiority margin or a single preserved fraction (PF) in non-inferiority trials for an entire pharmacological class will spare investigators from the extensive efforts required to define a new margin each time a member of this group is chosen as an active comparator (the recommended approach by regulators). However, the validity of this approach has not been assessed. Objectives: To assess the validity of using a single margin or a single PF for all non-inferiority trials within a pharmacological class (statins). Methods: A search in PubMed, EMBASE, and CENTRAL resulted in 7 active-controlled non-inferiority trials for treating hyperlipidemia. The impact of using a single margin (6% reduction of low-density lipoprotein cholesterol from the baseline) was assessed by evaluating how this margin corresponds to the PF for each comparator statin in the 7 trials. PF is the fraction of the effect of the comparator statin that was preserved by the test statin (the higher the PF the stricter the margin). The use of a single PF was assessed by re-analyzing non-inferiority in the included trials with new margins (based on the single PF) for each comparator statin and compare the new results with those of the original trials (in which a PF was assumed to be chosen for each comparator in each trial). Results: The use of a single margin resulted in PFs that range between 81% and 89% for the different comparators. This means that the stringency of demonstrating non-inferiority, in terms of the PF, varies among the comparator statins. For example, non-inferiority of a test statin to 10 mg atorvastatin will be demonstrated if it at least preserved 84% of the effect of atorvastatin that was pooled from the historical placebo-controlled trials (both the test and comparator statins are equipotent). However, this PF may become higher or lower if another equipotent statin is chosen as a comparator instead of 10 mg atorvastatin. The use of single PF resulted in 4 of 9 (44%) different non-inferiority conclusions compared with the original analyses. This means that the new margins were either wider or narrower compared with the original ones. Conclusions: The threshold of demonstrating non-inferiority with a single margin or single preserved fraction of the effect per pharmacological class may not be consistent with using a margin/PF for each comparator separately, which may also be invalid for the analysis of non-inferiority.