[NEJM综述]:主要结果失败-下一步应该做些什么?(4/4) | 中国病理生理学会危重病医学专业委员会
现在的位置: 首页时讯速递, 进展交流>正文
[NEJM综述]:主要结果失败-下一步应该做些什么?(4/4)
2017年01月17日 时讯速递, 进展交流 暂无评论

The Primary Outcome Fails — What Next?

主要结果失败-下一步应该做些什么?

Stuart J. Pocock, Ph.D., and Gregg W. Stone, M.D.

N Engl J Med 2016;375:861-70.

DOI: 10.1056/NEJMra1510064

佑安医院重症肝病科 宁琪琪 

Discussion

讨论

The 12 points explained above can be used to provide assistance in deciding what to do next when a trial fails to produce a positive finding for its primary outcome. Certainly one needs to be circumspect. Researchers may opt to move in one of three directions.

当试验的主要结局未能得到阳性结果时,以上解释的12条观点可以帮助我们决定下一步应该怎么办。当然我们需要谨慎。研究者应该在以下三个方向中选择其一。

Declare That the Trial Is Positive

宣布试验结果为阳性

Remarkable circumstances are usually required to report that a trial is positive even though the results of the primary outcome were not statistically significant at the prespecified level. The descriptions of the findings for the five trials listed in Table 2 provide a framework for debate as to whether each contained positive findings of clinical importance despite the negative primary outcome. However, although such considerations may inform guidelines committees, regulators are rarely swayed by such secondary analyses.

即使一项试验的主要结局未能达到预先确定的统计学显著差异水平,如果报告试验结果为阳性,通常需要极为严格的前提。表2描述的5项试验的结果为这一争论提供了框架,尽管试验的主要结果为阴性,是否每项试验都包含非常重要的阳性发现。 然而,尽管这些发现可能影响指南组委会的决定,但监管机构很少为之所动。

One notable exception was the CAPRICORN trial, which assessed the effects of carvedilol versus placebo after myocardial infarction in patients with left ventricular dysfunction.[37] The composite primary outcome — death or hospitalization from any cause –– failed to reach significance (hazard ratio, 0.92; 95% CI, 0.80 to 1.07; P = 0.30). But all-cause mortality alone did provide some evidence of benefit (hazard ratio, 0.77; 95% CI, 0.60 to 0.98; P=0.03) and, after much debate, led to FDA approval, perhaps because all-cause mortality had been the original primary outcome (an unfortunate switch was made by the investigators midtrial), and external evidence existed as to the effectiveness of beta-blockers in this population of patients.

一个明显的例外是CAPRICORN 试验,这项试验评估了左室功能障碍患者心肌梗塞后卡维地洛与安慰剂的疗效。复合终点(即全因死亡或住院)并无显著差异(风险比,0.92; 95% CI, 0.80 - 1.07; P = 0.30),但全因病死率提供了治疗获益的一些证据(风险比,0.77;95% CI,0.60-0.98;P = 0.03)。经过激烈辩论,FDA批准了卡维地洛,这可能是由于全因病死率才是最初的主要结局(研究者在试验过程中遗憾地作出了改变),同时外部证据表明β受体阻滞剂对这一人群的有效性。

Improve the Design of Future Trials

改善未来的试验设计

Trialists and sponsors usually have strong mechanistic support and background evidence that justifies the conduct of a major randomized trial. Hence, after a disappointing result, explanations are sought to guide the effort of designing a potential new trial. Aspects to consider include adjusting the treatment regimen, altering the study population, modifying the primary outcome, increasing the sample size, and improving other aspects of the trial that affected its quality. Such difficult and costly decisions should be based on realistic expectations rather than naive optimism.

试验者及资助方通常都有较强的理论支持和背景证据来实施一项大型的随机试验。因此,在得到令人失望的结果后,应该寻求解释以指导新试验的设计。需要考虑的方面包括调整治疗方案,改变研究人群,修订主要结局指标,增加样本量,以及改进影响试验质量的其他方面。应当根据实际的期望而非天真的乐观情绪作出上述困难且代价不菲的决定。

For example, after numerous open-label studies were conducted with highly positive results, renal denervation failed to substantially reduce blood pressure in patients with refractory hypertension in the sham-controlled SYMPLICITY HTN-3 trial.[38] Proposed explanations for this finding (which few expected) include an unfavorable mix of patients (some of whom had hypertension with an underlying cause that made a response to renal denervation unlikely), inadequate delivery of radiofrequency energy, changes in the drug treatment, and the failure to control for regression to the mean. Blinded mechanistic trials are ongoing in patients with hypertension who are not taking any antihypertensive medication to determine whether renal denervation does indeed “work,” before additional, large-scale trials are conducted.

例如,在众多开放标签研究得到阳性结果后,采用假手术为对照的SYMPLICITY HTN-3 对照试验发现,对于难治性高血压患者,肾脏去神经治疗不能显著降低患者的血压。对于这一发现(很少人预见到)的解释包括不恰当的患者混杂(有些患者高血压的病因不可能对肾脏去神经治疗有反应),射频能量不足,药物治疗的改变,以及未能控制趋均数回归。在进行其他大样本试验前,正在进行设盲的机制试验验证肾脏去神经治疗对于那些未服用任何降压药物的高血压患者是否真的有用。

Abandon the Treatment as Ineffective

放弃无效的治疗

The purpose of randomized trials is to distinguish between treatments that are effective and those that are not. Unfortunately, many innovations land in the second category. Hence, if the overall results of a trial show little or no evidence of treatment efficacy, and especially if there are safety issues, it may be wise to desist from further investigation. Such a conclusion may finally have been reached for thrombus aspiration in patients with acute myocardial infarction. After many years of mixed results from smaller trials, two large randomized trials [39,40] have now convincingly shown that routine thrombus aspiration has no benefit.

随机试验的目的是鉴别有效治疗与无效治疗。遗憾的是,许多创新在第二阶段即已夭折。因此,如果试验的总体结果很少或未能显示疗效证据,尤其当存在安全问题时,停止进一步研究是明智的。在急性心肌梗塞患者血栓抽吸试验中即得到了这样的结论。历经多年的小样本试验彼此矛盾的结果,两项大样本随机试验已经令人信服地表明,常规血栓抽吸没有益处。

Conclusions

结论

When the primary outcome of a trial fails to achieve statistical significance, we propose that researchers ask a series of searching questions that will help them clarify whether the new treatment may still have value. The options are to claim “success” anyway on the basis of the total evidence (an option that is rarely used), to plan a future trial with design improvements (a costly option), or to accept that the new treatment is likely to be ineffective (a frustrating option). However, the best option is to avoid this scenario altogether through rigorous upfront planning. By making sure that there is evidence of strong pathophysiological and mechanistic underpinnings that are common to both the new therapy and the disease, by selecting appropriate patients and end points, by calculating an adequate sample size, by paying meticulous attention to dosing, definitions of disease and outcomes, and all procedural processes, and by anticipating the ways in which the trial might fail and give rise to criticisms, one can enhance the likelihood of reaching a decisive conclusion.

当试验的主要结局未能达到统计学意义时,我们建议研究者提出一系列的探索性问题,以帮助他们阐明新的治疗方案是否仍有价值。此时的选择包括根据总体证据宣布试验“成功”(很少采用的选择),在改进设计的基础上计划未来的试验(代价不菲的选择),或接受新的治疗可能无效的结论(令人沮丧的选择)。然而,最好的选择是通过严格的前期规划避免这种情况的发生。通过确保新治疗方法及疾病的生理学及机制方面的充分证据,选择合适的患者及终点,计算足够的样本量,仔细确定剂量,疾病以及结局的定义,以及关注所有的操作流程,并预期可能导致试验失败的原因,我们能够提高获得确切结果的可能性。

给我留言

您必须 [ 登录 ] 才能发表留言!

×
腾讯微博