The Fragility of Statistically Significant Findings in Pediatric Critical Care Randomized Controlled Trials.

Abstract

OBJECTIVES\nThe Fragility Index measures the number of events on which the statistical significance of a result depends and has been suggested as an adjunct statistical assessment for interpretation of trial results. This study aimed to assess the robustness of statistically significant results from pediatric critical care randomized controlled trials with dichotomous outcomes.\n\n\nDATA SOURCES\nA previously published scoping review of pediatric critical care randomized controlled trials (www.PICUtrials.net).\n\n\nSTUDY SELECTION\nA total of 342 trials were screened for inclusion. After applying inclusion/exclusion criteria, 43 fulfilled eligibility criteria and were included in the analysis.\n\n\nDATA EXTRACTION\nCalculation of Fragility Index for trials reporting a statistically significant dichotomous outcome, and analysis of the relationship between trial characteristics and Fragility Index.\n\n\nDATA SYNTHESIS\nThe median Fragility Index was 2 (interquartile range, 1-6). The median sample size was 98 (interquartile range, 50-148) and sample size demonstrated a strong correlation with the Fragility Index (r = 0.729; n = 43; p < 0.001). The median number of outcome events was 8 (interquartile range, 4-15) and the total number of outcome events also showed a strong correlation with the Fragility Index (r = 0.728; n = 43; p < 0.001).\n\n\nCONCLUSIONS\nResults from pediatric critical care randomized controlled trials with dichotomous outcomes reporting statistically significant findings often hinge on a small number of outcome events. Clinicians should exercise caution when interpreting results of trials with a low Fragility Index.