Suissa Samy, Dell'Aniello Sophie, Webster-Clark Michael Andrew, Ernst Pierre
Centre for Clinical Epidemiology, Lady Davis Institute-Jewish General Hospital, Montreal, QC, Canada.
Department of Medicine, McGill University, Montreal, QC, Canada.
Int J Chron Obstruct Pulmon Dis. 2025 Jul 5;20:2273-2285. doi: 10.2147/COPD.S527497. eCollection 2025.
Recent observational studies have reported that prompt initiation of single-inhaler triple therapy after a COPD exacerbation is more effective than delayed initiation. We show that their study design, by "peeking into the future" to define the timing of treatment initiation, introduces time-related biases, particularly protopathic bias. These biases can be avoided using the "cloning" approach to emulate a randomized trial approach.
We formed a cohort of patients with COPD who had an exacerbation (index) after September 2017, using the United Kingdom's Clinical Practice Research Datalink (CPRD). Using the "cloning" trial emulation technique, each subject was assigned to both the prompt and the delayed initiator arms as of the index date and censored according to their treatment over time. The Cox model was used to compare the incidence of the first exacerbation after the index exacerbation, over one year, after weighing by inverse probability of censoring. We also replicated the biased approach of the recent studies, based on peeking into the future.
The cohort included 91,958 eligible subjects who had an exacerbation, generating 91,958 prompt initiator clones and 91,958 delayed initiator clones. The hazard ratio (HR) of a moderate or severe exacerbation, comparing prompt versus delayed initiators, was 0.98 (95% CI: 0.80-1.19), while it was 1.26 (95% CI: 0.81-1.96) for severe exacerbation. The replication of the time-related biased approach comparing prompt with delayed initiation resulted, correspondingly, in HRs of 0.73 (95% CI: 0.65-0.81) and 0.58 (95% CI: 0.46-0.74).
Using a trial emulation approach, prompt treatment with single-inhaler triple therapy after a COPD exacerbation was not more effective than delayed treatment at reducing the incidence of subsequent exacerbations. The method used by previous studies, suggesting significant effectiveness with prompt treatment initiation, was affected by time-related biases induced by peeking into the future. A randomized controlled trial can confirm these findings.
近期的观察性研究报告称，慢性阻塞性肺疾病（COPD）急性加重后迅速开始单吸入器三联疗法比延迟开始更有效。我们发现，他们的研究设计通过“窥探未来”来确定治疗开始的时间，引入了与时间相关的偏差，尤其是原发病偏差。使用“克隆”方法来模拟随机试验方法可以避免这些偏差。
我们使用英国临床实践研究数据链（CPRD），组建了一组在2017年9月后发生急性加重（索引）的COPD患者队列。使用“克隆”试验模拟技术，从索引日期起，将每个受试者同时分配到迅速启动组和延迟启动组，并根据他们随时间的治疗情况进行删失。采用Cox模型，在按删失逆概率加权后，比较索引急性加重后一年内首次急性加重的发生率。我们还基于窥探未来，复制了近期研究中有偏差的方法。
该队列包括91958名发生急性加重的合格受试者，生成了91958个迅速启动克隆和91958个延迟启动克隆。比较迅速启动者与延迟启动者，中度或重度急性加重的风险比（HR）为0.98（95%CI：0.80 - 1.19），而重度急性加重的HR为1.26（95%CI：0.81 - 1.96）。相应地，比较迅速启动与延迟启动的与时间相关的有偏差方法的复制结果，中度或重度急性加重的HR为0.73（95%CI：0.65 - 0.81），重度急性加重的HR为0.58（95%CI：0.46 - 0.74）。
采用试验模拟方法，COPD急性加重后迅速使用单吸入器三联疗法治疗在降低后续急性加重发生率方面并不比延迟治疗更有效。先前研究使用的方法表明迅速开始治疗有显著效果，但受到了窥探未来所导致的与时间相关偏差的影响。一项随机对照试验可以证实这些发现。
