After the completion of a clinical trial, its conclusion generally depends on the results of statistical analysis of the main outcome, that is, whether the P-value in the hypothesis test is less than the α level of the hypothesis test, usually α=0.05. The size of the P-value indicates the sufficient degree of reason for making the hypothesis judgment, and can be interpreted as to determine whether a conclusion is statistically significant but does not involve the difference in the degree of drug effects or other effects. Fragility index, which is, the minimum number of patients required to change the occurrence of a target outcome event to a non-target outcome event from a statistically significant outcome to a non-significant outcome, can be used to assist in understanding of clinical trial statistical inference results and assisting in clinical decision making This paper discusses the concept, calculation method and clinical application of the fragility index, and recommends that the fragility index be routinely reported in all future randomized controlled trials to help patient clinicians and policymakers make appropriate and optimal decisions.
ObjectiveTo evaluate the robustness of pediatrics Clinical evidence-based evidence using fragility index and to explore the factors influencing fragility index. MethodsWe searched the PubMed, Embase, and Scopus databases to collect relevant literature on systematic reviews and meta-analyses in the field of pediatrics, and calculated the fragility index. The rank sum test was used to compare differences between groups with different outcome types, different levels of statistical significance, and different sample sizes. Spearman correlation analysis was used to explore the association between the fragility index and sample size, as well as the year of publication. ResultsA total of 152 systematic reviews, including 573 meta-analyses, were included, with a median fragility index of 6 (3, 10). Most meta-analyses chose the risk ratio (RR) as the effect measure (387/573, 67.5%), the Mantel-Haenszel method (412/573, 71.9%) as the synthesis method, and the fixed-effect model (300/573, 57.4%) as the assumed model. The Mann-Whitney test showed no statistically significant difference in the fragility index between meta-analyses with safety outcomes and those with efficacy outcomes (P=0.397), and no statistically significant difference between meta-analyses with significant results and those with non-significant results (P=0.520). The Kruskal-Wallis test found a statistically significant difference in sample size among groups with different fragility indices (P<0.001). Spearman correlation analysis found a positive correlation between the fragility index and sample size (ρ=0.39, P<0.001), but no statistically significant correlation with the year of publication (P=0.235). ConclusionThe fragility index of clinical evidence-based evidence published in pediatrics journals is generally low, and the robustness of the results is not high, so it is necessary to be cautious when making evidence-based decisions. Furthermore, the larger the sample size included in the meta-analysis, the higher the fragility index, and incorporating more trials and populations can facilitate the increase in the robustness of the meta-analysis results.