Fragility index and risk of bias in cardiovascular outcome trials for new antidiabetic drugs

Abstract

Background: The alarm set by rosiglitazone, a thiazolidindione that seemed to enhance the existing two to three-fold increased risk of cardiovascular disease comprised in T2DM, induced that regulatory agencies issued a guidance for proving cardiovascular safety for new antidiabetic therapies. Even though thiazolidindiones’ safety was proven afterwards, cardiovascular outcome trials (CVOTs) have been conducted since 2008, using mainly a non-inferiority trial design. These aim for the rejection of their null hypothesis; that the new drug is not unacceptably worse than the standard treatment. In order to achieve that, the upper bound of the 95% confidence interval (CI) cannot exceed a HR ≤1,3 for the primary end point of combined cardiovascular death, nonfatal stroke and nonfatal myocardial infarction. Surprisingly, some of these trials happen to find a cardiovascular outcome HR lower than 1, meaning that cardiovascular events are lower in the experimental group than in the control group. And therefore, are the mainstay of the pretended cardiovascular risk reduction of some new antidiabetic agents such as SGLT-2i or GLP-1RA, ending in the inclusion of these drugs in recent widely used clinical practice recommendations’ documents. Objectives: The aim of this study is to assess the robustness and methodological quality of CVOTs conducted for new antidiabetic drugs after the 2008 FDA guidance, that justify changing clinical practice recommendations. Methods: In order to accomplish that aim, this narrative review applied the Fragility Index (FI) and Fragility Quotient (FQ) to eligible trials, together with the assessment of the Cochrane “Risk of Bias” Tool RoB 2.0 for all CVOTs conducted since 2008 as a measure of internal validity. Duplicate revision is on the way, so that preliminary results are reported. Results: FI was applied to 7 eligible CVOTs claiming for statistical superiority, obtaining a median FI of 50 (IQR 19-62) and a median FQ of 1,1% (IQR 0.2-1.3), value that was surpassed in 100% of the

trials by the median number of patients lost to follow-up and rate of premature discontinuation. Sub- analyses conducted for 3 CVOTs claiming for statistical superiority that were not eligible for FI,

showed no statistical significance. RoB assessment conducted for 21 CVOTs resulted in some concerns as the overall appraisal for 75% of them. Conclusions: CVOTs conducted after the 2008 FDA guidance for new antidiabetic drugs for T2DM do not provide enough statistical robustness and do have some concerns regarding their methodological quality to justify changing clinical practice recommendations. Hence, creating some hesitation on the reliability of the results and whether is ethical to treat with these additional drug therapies that group of patients. Our findings demonstrate that there is a need for critical review for CVOTs, which would probably lead to the revision of treatment recommendations