Fragility Analysis of Statistically Significant Outcomes of Randomized Control Trials in Spine Surgery A Systematic Review

摘要

Study Design. Systematic review. Objectives. The aim of this study was to assess the robustness of statistically significant outcomes from randomized control trials (RCTs) in spine surgery using Fragility Index (FI) which is a novel metric measuring the number of events upon which statistical significance of the outcome depends. Summary of Background Data. Many trials in Spine surgery were characterized by fewer outcome events along with small sample size. FI helps us identify the robustness of the results from such studies with statistically significant dichotomous outcomes. Methods. We conducted independent and in duplicate, a systematic review of published RCTs in spine surgery from PubMed Central, Embase, and Cochrane Database. RCTs with 1:1 prospective study design and reporting statistically significant dichotomous primary or secondary outcomes were included. FI was calculated for each RCT and its correlation with various factors was analyzed. Results. Seventy trials met inclusion criteria with a median sample size of 133 (interquartile range [IQR]: 80-218) and median reported events per trial was 38 (IQR: 13-94). The median FI score was 2 (IQR: 0-5), which means if we switch two patients from nonevent to event, the statistical significance of the outcome is lost. The FI score was less than the number of patients lost to follow-up in 28 of 70 trials. The FI score was found to positively correlated with sample size (r = 0.431, P = 0.001), total number of outcome events (r = 0.305, P = 0.01) while negatively correlated with P value (r = -0.392, P = 0.001). Funding, journal impact-factor, risk of bias domains, and year of publication did not have a significant correlation. Conclusion. Statistically significant dichotomous outcomes reported in spine surgery RCTs are more often fragile and outcomes of the patients lost to follow-up could have changed the significance of results and hence it needs caution before transcending their results into clinical application. The addition of FI in routine reporting of RCTs would guide readers on the robustness of the statistical significance of outcomes. RCTs with FI >= 5 without any patient lost to follow-up can be considered to have clinically robust results.