Cytel Canada Inc., Vancouver, BC, Canada.
School of Public Health, Faculty of Medicine, Imperial College London, London, United Kingdom.
J Med Internet Res. 2021 Mar 12;23(3):e26718. doi: 10.2196/26718.
This paper aims to provide a perspective on data sharing practices in the context of the COVID-19 pandemic. The scientific community has made several important inroads in the fight against COVID-19, and there are over 2500 clinical trials registered globally. Within the context of the rapidly changing pandemic, we are seeing a large number of trials conducted without results being made available. It is likely that a plethora of trials have stopped early, not for statistical reasons but due to lack of feasibility. Trials stopped early for feasibility are, by definition, statistically underpowered and thereby prone to inconclusive findings. Statistical power is not necessarily linear with the total sample size, and even small reductions in patient numbers or events can have a substantial impact on the research outcomes. Given the profusion of clinical trials investigating identical or similar treatments across different geographical and clinical contexts, one must also consider that the likelihood of a substantial number of false-positive and false-negative trials, emerging with the increasing overall number of trials, adds to public perceptions of uncertainty. This issue is complicated further by the evolving nature of the pandemic, wherein baseline assumptions on control group risk factors used to develop sample size calculations are far more challenging than those in the case of well-documented diseases. The standard answer to these challenges during nonpandemic settings is to assess each trial for statistical power and risk-of-bias and then pool the reported aggregated results using meta-analytic approaches. This solution simply will not suffice for COVID-19. Even with random-effects meta-analysis models, it will be difficult to adjust for the heterogeneity of different trials with aggregated reported data alone, especially given the absence of common data standards and outcome measures. To date, several groups have proposed structures and partnerships for data sharing. As COVID-19 has forced reconsideration of policies, processes, and interests, this is the time to advance scientific cooperation and shift the clinical research enterprise toward a data-sharing culture to maximize our response in the service of public health.
本文旨在探讨 COVID-19 大流行背景下的数据共享实践。科学界在抗击 COVID-19 方面取得了多项重要进展，目前在全球范围内有超过 2500 项临床试验注册。在疫情迅速变化的情况下，我们看到大量试验在没有结果的情况下进行。很可能有大量试验因缺乏可行性而提前停止，而不是因为统计学原因。提前因可行性而停止的试验，根据定义，统计上的效力不足，因此容易得出不确定的结论。统计效力不一定与总样本量呈线性关系，即使患者人数或事件略有减少，也会对研究结果产生重大影响。鉴于大量临床试验在不同地理和临床环境下研究相同或相似的治疗方法，人们还必须考虑到，随着试验总数的增加，大量假阳性和假阴性试验的出现会增加公众对不确定性的看法。由于大流行的不断变化性质，用于计算样本量的对照组风险因素的基本假设比在记录良好的疾病情况下更为复杂，这使得这个问题变得更加复杂。在非大流行环境下，应对这些挑战的标准答案是评估每个试验的统计效力和偏倚风险，然后使用荟萃分析方法汇总报告的汇总结果。对于 COVID-19 来说，这个解决方案是不够的。即使使用随机效应荟萃分析模型，仅使用汇总报告数据也很难调整不同试验的异质性，尤其是在缺乏共同数据标准和结果衡量标准的情况下。迄今为止，已经有几个小组提出了数据共享的结构和合作伙伴关系。由于 COVID-19 迫使人们重新考虑政策、流程和利益，现在是推进科学合作并将临床研究企业转向数据共享文化的时候了，以最大限度地提高我们为公共卫生服务的反应。
