Aidan Tan and Philippe Ravaud, Mailman School of Public Health, Columbia University, New York, NY; Aidan Tan, National University Hospital, Singapore; Raphael Porcher, Perrine Crequit, Philippe Ravaud, and Agnes Dechartres, Hôpital Hôtel-Dieu, Assistance Publique-Hôpitaux de Paris; Raphael Porcher, Philippe Ravaud, and Agnes Dechartres, Université Paris Descartes, Sorbonne Paris Cité; Raphael Porcher, Perrine Crequit, Philippe Ravaud, and Agnes Dechartres, Centre de Recherche Epidémiologie et Statistique, INSERM U1153; and Perrine Crequit, Philippe Ravaud, and Agnes Dechartres, Cochrane France, Paris, France.
J Clin Oncol. 2017 May 20;35(15):1686-1694. doi: 10.1200/JCO.2016.71.2109. Epub 2017 Apr 4.
Purpose We aimed to compare treatment effect sizes between overall survival (OS) and progression-free survival (PFS) in trials of US Food and Drug Administration-approved oncology immunotherapy drugs with results posted at ClinicalTrials.gov . Methods We searched ClinicalTrials.gov for phase II to IV cancer trials of Food and Drug Administration-approved immunotherapy drugs and selected those reporting results for both OS and PFS. For each trial, we extracted the hazard ratios (HRs) with 95% CIs for both outcomes and evaluated the differences by a ratio of HRs (rHRs): the HR for PFS to that for OS. We performed a random effects meta-analysis across trials to obtain a summary rHR. We also evaluated surrogacy of PFS for OS by the coefficient of determination and the surrogacy threshold effect, the minimal value of HR for PFS to predict a non-null effect on OS. Results We identified 51 trials assessing 14 drugs across 15 conditions. Treatment effect sizes were 17% greater, on average, with PFS than with OS (rHR, 0.83; 95% CI, 0.79 to 0.88; I = 34.4%; P = .01; τ = 0.0129). Nearly one half of the trials (n = 23, 45%) showed statistically significant benefits for PFS but not for OS. Differences were great for trials of obinutuzumab (rHR, 0.21; 95% CI, 0.08 to 0.54), bevacizumab (rHR, 0.75; 95% CI, 0.67 to 0.84), and rituximab (rHR, 0.79; 95% CI, 0.64 to 0.98). The coefficient of determination was 38% and the surrogacy threshold effect was 0.50. Conclusion Treatment effect sizes in trials of immunotherapy drugs were greater for PFS than for OS, with important differences for some drugs, which is consistent with surrogacy metrics. Caution must be taken when interpreting PFS in the absence of OS data.
目的 我们旨在比较美国食品和药物管理局批准的肿瘤免疫治疗药物临床试验中总生存期（OS）和无进展生存期（PFS）的治疗效果大小，并将结果发布在 ClinicalTrials.gov 上。
方法 我们在 ClinicalTrials.gov 上搜索了 II 期至 IV 期癌症试验，这些试验针对的是美国食品和药物管理局批准的免疫治疗药物，并选择了同时报告 OS 和 PFS 结果的试验。对于每个试验，我们提取了这两个结局的风险比（HR）及其 95%置信区间（CI），并通过 HR 比值（rHR）评估差异：PFS 的 HR 与 OS 的 HR 之比。我们对试验进行了随机效应荟萃分析，以获得汇总 rHR。我们还通过决定系数和替代阈值效应评估了 PFS 对 OS 的替代程度，替代阈值效应是预测 OS 非零效应的最小 PFS HR 值。
结果 我们确定了 51 项试验，涉及 15 种条件下的 14 种药物。平均而言，PFS 的治疗效果比 OS 大 17%（rHR，0.83；95%CI，0.79 至 0.88；I = 34.4%；P =.01；τ = 0.0129）。近一半的试验（n = 23，45%）显示 PFS 有统计学意义的益处，但 OS 没有。奥滨尤妥珠单抗（rHR，0.21；95%CI，0.08 至 0.54）、贝伐珠单抗（rHR，0.75；95%CI，0.67 至 0.84）和利妥昔单抗（rHR，0.79；95%CI，0.64 至 0.98）的试验差异较大。决定系数为 38%，替代阈值效应为 0.50。
结论 免疫治疗药物临床试验中 PFS 的治疗效果比 OS 大，对于某些药物存在显著差异，这与替代指标一致。在没有 OS 数据的情况下，解读 PFS 时必须谨慎。
