Tanscatheter aortic valve replacement for asymptomatic aortic stenosis - A revisited and contrarian meta-analysis

ImportanceUsing aggregated data, two recent meta-analyses have concluded that early aortic valve replacement (AVR) was associated with reduced adverse events compared to clinical surveillance in severe but asymptomatic aortic stenosis. However, individual patient data was not used and the possibility and extent of bias due to the unblinding trial design were not considered. ObjectiveUsing reconstructed individual patient level data, the possiblity of early bias was investigated and a meta-analysis of longer term benefits was performed using one year landmark data. Evidence ReviewFour randomized trials, as identified from previous systematic reviews, showed important clinical and statistical heterogeneity in year one AVR crossovers to cardiovascular hospitalizations. To minimize any early bias from unblinding, one year landmark analyses were performed separately for each trial and combined in a Bayesian (hierarchical) meta-analysis. FindingsThe largest trial with a TAVR intervention arm was the only trial to show improved outcomes in the first year, driven almost completely by an approximate two fold increase in the crossover rate compared to previous SAVR intervention trials. A one year landmark meta-analysis showed no long term benefit for AVR compared to CS for the primary outcome of mortality and cardiovascular hospitalizations for any individual study or for the pooled result (RR 0.70, 95% CI 0.34 - 1.08). Conclusions and RelevanceThe early benefit with TAVR in asymptomatic patients with severe aortic stenosis appears more driven by bias than by efficacy. Landmark analysis accounting for this potential bias show no longer term advantage for early AVR compared to clinical surveillance in this population. Key PointsO_ST_ABSQuestionC_ST_ABSDoes early intervention for severe asymptomatic aortic stenosis improve patient outcomes compared to clinical surveillance.? FindingsA systematic review suggested early benefits were likely attributable not to interventional efficacy but rather bias due to an unblinded design for a subjective outcome. A one-year landmark meta-analysis showed no long term benefit for early intervention compared to clinical surveillance for the primary composite outcome of mortality and cardiovascular hospitalizations (RR 0.70, 95% CI 0.34 - 1.08). MeaningAfter accounting for possible early bias, landmark meta-analysis shows no longer term advantage for early intervention compared to clinical surveillance in this population.