Division of General Surgery, McMaster University, Hamilton, ON, Canada.
Harvard T.H. Chan School of Public Health, Harvard University, Boston, MA, USA.
Int J Obes (Lond). 2023 Jul;47(7):546-553. doi: 10.1038/s41366-023-01298-1. Epub 2023 Apr 1.
Randomized controlled trials (RCTs) are regarded as high-level evidence, but the strength of their P values can be difficult to ascertain. The Fragility Index (FI) is a novel metric that evaluates the frailty of trial findings. It is defined as the minimum number of patients required to change from a non-event to event for the findings to lose statistical significance. This study aims to characterize the robustness of bariatric surgery RCTs by examining their FIs.
A search was conducted in MEDLINE, EMBASE, and CENTRAL from January 2000 to February 2022 for RCTs comparing two bariatric surgeries with statistically significant dichotomous outcomes. Bivariate correlation was conducted to identify associations between FI and trial characteristics.
A total of 35 RCTs were included with a median sample size of 80 patients (Interquartile range [IQR] 58-109). The median FI was 2 (IQR 0-5), indicating that altering the status of two patients in one treatment arm would overturn the statistical significance of results. Subgroup analyses of RCTs evaluating diabetes-related outcomes produced a FI of 4 (IQR 2-6.5), while RCTs comparing Roux-en-Y gastric bypass and sleeve gastrectomy had an FI of 2 (IQR 0.5-5). Increasing FI was found to be correlated with decreasing P value, increasing sample size, increasing number of events, and increasing journal impact factor.
Bariatric surgery RCTs are fragile, with only a few patients required to change from non-events to events to reverse the statistical significance of most trials. Future research should examine the use of FI in trial design.
随机对照试验（RCTs）被认为是高级别的证据，但 P 值的强度难以确定。脆弱指数（FI）是一种评估试验结果脆弱性的新指标。它定义为需要改变多少名患者的状态，才能使研究结果从无事件变为有事件，从而使结果失去统计学意义。本研究旨在通过检查减肥手术 RCTs 的 FI 来描述其稳健性。
在 2000 年 1 月至 2022 年 2 月期间，在 MEDLINE、EMBASE 和 CENTRAL 中进行了一项 RCTs 的搜索，比较了两种具有统计学显著二分类结局的减肥手术。进行了双变量相关性分析，以确定 FI 与试验特征之间的关联。
共纳入 35 项 RCT，中位数样本量为 80 例（四分位距[IQR]为 58-109）。FI 的中位数为 2（IQR 0-5），表明改变一个治疗组中两名患者的状态会推翻结果的统计学意义。评估糖尿病相关结局的 RCT 亚组分析得出 FI 为 4（IQR 2-6.5），而比较 Roux-en-Y 胃旁路术和袖状胃切除术的 RCT 则为 2（IQR 0.5-5）。FI 的增加与 P 值的降低、样本量的增加、事件数量的增加和期刊影响因子的增加有关。
减肥手术 RCTs 较为脆弱，仅需改变少数患者的状态，即可使大多数试验的统计学意义发生逆转。未来的研究应检验 FI 在试验设计中的应用。
