Hierarchical endpoints in critical care: A post-hoc exploratory analysis of the standard versus accelerated initiation of renal-replacement therapy in acute kidney injury and the intensity of continuous renal-replacement therapy in critically ill patients trials.

Abstract

To perform a post-hoc reanalysis of the Standard versus Accelerated Initiation of Renal-Replacement Therapy in Acute Kidney Injury (STARRT-AKI) and the Intensity of Continuous Renal-Replacement Therapy in Critically Ill Patients (RENAL) trials through hierarchical composite endpoint analysis using win ratio (WR). All patients with complete information from the STARRT-AKI (which compared accelerated versus standard approaches for renal replacement therapy - RRT initiation) and RENAL (which compared two different RRT doses in critically ill patients) trials were selected. WR was defined as a hierarchical composite endpoint using 90-day mortality, RRT dependency at 90-days, intensive care unit (ICU) length-of-stay (LOS), and hospital LOS (primary analysis); values above the unit represent a benefit of the intervention for the hierarchical composite endpoint. A secondary analysis replacing LOS by days alive and free of RRT was performed. Stratified analyses were performed according to illness severity score, surgical status, and the presence of sepsis. The WR analysis produced 2,141,830 pairs for the STARRT-AKI trial and 536,446 pairs for the RENAL trial, respectively. The WR results for STARRT-AKI and RENAL were 1.04 (95% confidence interval [CI] 0.96-1.13; p = 0.33) and 1.02 (95% CI; 0.90-1.15; p = 0.75) for the primary analysis, and 0.88 (95% CI; 0.79-0.99; p = 0.03) and 1.02 (95% CI; 0.87-1.21; p = 0.77) for the secondary analysis, respectively. The stratified analysis of the primary suggested possible benefit of the accelerated-strategy in the STARRT-AKI trial for non-surgical patients with sepsis, while the secondary analysis suggested possible harm of the accelerated-strategy for surgical patients without sepsis. There was no evidence of heterogeneity in treatment effects in stratified analyses in the RENAL trial. WR approach using a hierarchical composite endpoint is feasible for trials in critical care nephrology. The primary re-analyses of the STARRT-AKI and RENAL trials both yielded neutral results; however, there was suggestion of heterogeneity in treatment effect in stratified analyses of the STARRT-AKI trial by surgical status and sepsis. Selection of the endpoints and hierarchical ordering before trial design using the WR approach can have important implications for trial interpretation. ClinicalTrials.gov number NCT02568722 (STARRT-AKI) and NCT00076219 (RENAL).