反向脆弱指数揭示Cochrane荟萃分析中P值0.05-0.20结果的高脆弱性
该meta-流行病学研究分析2025年发表的81篇Cochrane系统评价中280个P值0.05-0.20的非显著合并效应量,计算反向脆弱指数中位数为3,即改变3个事件即可使非显著结果变为显著,提示此类结果脆弱,不应被视为稳健的无效结论。
要点
- P值0.05-0.20的荟萃分析结果脆弱,中位反向脆弱指数仅3个事件。
- P值0.05-0.10的荟萃分析比0.10-0.20者更脆弱。
- 事件改变未显著增加异质性,非显著结果不应解读为无效。
核心结论
- P值0.05-0.20的荟萃分析结果具有统计脆弱性,反向脆弱指标可为非显著结果提供量化证据,挑战将其误读为稳健无效结论。
研究设计与样本
- meta-流行病学研究,分析2025年发表的81篇Cochrane系统评价中280个非显著合并效应估计。
- 计算反向脆弱指数(使非显著结果转为显著所需的最小结局事件改变数)及反向脆弱商。
- 结果以中位数(Q1-Q3)及基于10000次bootstrap迭代的95%CI表示。
主要发现
- 荟萃分析中位纳入5项研究、1002名参与者,干预组和对照组事件风险分别为11.7%和12.6%。
- 反向脆弱指数中位数为3(95%CI 2-3;Q1-Q3 2-5),即改变3个事件即可改变非显著结果状态。
- 反向脆弱商中位数为每1000名参与者2.7(95%CI 2.2-3.3;Q1-Q3 1.0-5.8)。
- P值0.05-0.10的荟萃分析比P值0.10-0.20者更脆弱。
- 事件改变未显著增加异质性。
临床怎么用
- P值0.05-0.20的荟萃分析不应被当作干预无效或无害的强证据。
- 研究者应报告非显著结果的反向脆弱指标,帮助临床医生和指南制定者判断证据可信度。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
OBJECTIVES: To quantify fragility of statistically nonsignificant meta-analysis results with P values between 0.05 and 0.20 using the reverse fragility index. STUDY DESIGN AND SETTING: This meta-epidemiological study examined Cochrane systematic reviews published in 2025. We calculated the reverse fragility index (minimum number of outcome event changes needed to convert a nonsignificant result to statistically significant) and reverse fragility quotient. Results represent medians (Q1-Q3) with 95% confidence interval (CI) from 10,000 bootstrap iterations. RESULTS: We analyzed 280 nonsignificant pooled effect estimates from 81 Cochrane systematic reviews. Meta-analyses included a median of five studies and 1002 participants with event risks of 11.7% and 12.6% in the intervention and control groups, respectively. The median reverse fragility index was 3 (95% CI, 2-3; Q1-Q3, 2-5), indicating that modifying three events among the primary studies included in a meta-analysis could change the status of nonsignificant results. The median reverse fragility quotient was 2.7 per 1000 participants (95% CI, 2.2-3.3; Q1-Q3, 1.0-5.8). Meta-analyses with P values 0.05-0.10 were more fragile than those with P values 0.10-0.20. Event changes did not substantially increase heterogeneity. CONCLUSION: Meta-analyses with P values between 0.05 and 0.20 showed statistical fragility. Reverse fragility metrics provide quantifiable evidence for nonsignificant results that may have a signal of clinical relevance and challenge their misinterpretation as robust null findings. PLAIN LANGUAGE SUMMARY: When researchers combine numerical results from multiple studies (called meta-analyses), they often find that results are not "statistically significant," meaning that they are more likely due to chance or error. These results are commonly misinterpreted using phrases such as "there is no difference" between the two groups, which can mislead readers about what the evidence shows. We examined 280 meta-analyses from Cochrane systematic reviews to test the stability of these results. We used a method called the "reverse fragility index," which counts how many events would need to change from not occurring to occurring or vice versa to shift a statistically nonsignificant result into statistically significant. We found that these statistically nonsignificant results are fragile. In more than half of the meta-analyses, changing three events or fewer was enough to make a nonsignificant result significant. This represents almost three event changes per 1000 people. The findings show that meta-analyses with P values between 0.05 and 0.20 should not be treated as strong evidence that an intervention does not work or does not cause harm. Instead, these results are fragile. Researchers should report how fragile their nonsignificant results are to help clinicians and guideline developers better understand how much they can trust the evidence when making clinical decisions.
引用
- Sharifan A, Harrod C, Dobrescu A, et al. Reverse fragility index shows high fragility in Cochrane meta-analyses with P values between 0.05 and 0.20: a meta-epidemiological study. J Clin Epidemiol. 2026. PMID: 42025973. doi:10.1016/j.jclinepi.2026.112285
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。