Background: Grading of Recommendations Assessment, Development and Evaluation (GRADE) has proposed a system for rating the certainty of a body of evidence (CoE) to determine confidence in the estimate. The GRADE approach has not only refreshed the way in which the evidence is presented, and even though it has become a fact of high-quality systematic reviews, its principles might be applicable to primary designs.
Objectives: Describe our experience applying the GRADE approach to the conclusions of a non-inferiority, pragmatic, multicenter, randomized clinical trial of Cardiac Rehabilitation (HYCARET Study).
Methods: This report is based on HYCARET study, which evaluated the non-inferiority of a hybrid cardiac rehabilitation (CR) program for patients with coronary artery disease. Following the guidelines for non-inferiority trials, we conducted intention-to-treat (ITT) and per-protocol (PP) analyses for the main outcome. We established the non-inferiority margin at an absolute risk difference of 5% for the proportions of cardiovascular events. For the PP analysis, we run four analyses for four different percentages of adherence to the intervention, 80%, 60%, 40%, and 20%. Then we plotted the effect estimates with their confidence intervals (Figure 1) and we judged the inconsistency, imprecision, and risk of bias following the GRADE guidelines.
Results: We decrease CoE in two levels because of imprecision. The loss of sample size according to adherence to the intervention in the PP analysis, and the wide confidence intervals crossing the non-inferiority threshold introduced imprecision to our result. However, the effect on recurrent cardiovascular events showed a consistent non-inferiority effect across all analyses. The conclusion of the study is therefore a hybrid CR program may be not inferior to a standard CR program in terms of recurrent cardiovascular events.
Conclusions: Domains of GRADE approach, such as inconsistency, imprecision, and risk of bias, might be transferable to draft the conclusions in non-inferiority trials. We hope that sharing our experience may help future authors of noninferiority trials not to focus only on thresholds in writing the conclusions of their papers and may consider including an assessment of the CoE. Patient involvement: Patients participated in the design of part of the intervention under study (counseling).