The Fragility of Statistically Significant Randomized Controlled Trials in Plastic Surgery.

Abstract

The fragility index has been proposed as a metric to evaluate the robustness of statistically significant findings in randomized controlled trials. It measures the number of events that a trial result relies on to maintain statistical significance. This study examines the robustness of statistically significant results from randomized controlled trials in the plastic surgery literature. A systematic literature search of the 15 highest impact plastic surgery journals was conducted to identify randomized controlled trials published between 2000 and 2017 that reported a statistically significant dichotomous outcome (p < 0.05). The fragility index of each study was calculated using Fisher's exact test. Multiple linear regression was used to determine trial characteristics associated with the fragility index. The 90 eligible randomized controlled trials had a median sample size of 73.5 patients (25th to 75th percentile, 50 to 115) and a median of 20 events (25th to 75th percentile, 11 to 33.5) for the chosen outcome. The median fragility index was 1 (25th to 75th percentile, 0 to 4), indicating that statistical significance would be lost in half of the randomized controlled trials if a single patient had a change in event status. The fragility index was 0 in 24 of 90 (27 percent) randomized controlled trials, meaning the outcome immediately lost statistical significance on recalculation of the p value using Fisher's exact test. The results of randomized controlled trials in plastic surgery demonstrate substantial fragility, as statistically significant results were found to hinge on a small number of events. The fragility index offers an intuitive and simple metric to complement the p value and determine the confidence in the results of randomized controlled trials.