Jackson Bradford E, Greenup Rachel A, Strassle Paula D, Deal Allison M, Baggett Chris D, Lund Jennifer L, Reeder-Hayes Katie E
Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, NC, USA.
Department of Surgery, Duke University, Durham, NC, USA; Duke Cancer Institute, Duke University, Durham, NC, USA.
Surg Oncol. 2021 Jun;37:101539. doi: 10.1016/j.suronc.2021.101539. Epub 2021 Mar 5.
Surgical health services researchers are increasingly utilizing observational data to assess associations between treatments and outcomes, especially since some procedures are unable to be evaluated through randomized controlled trials. However, the results of many of these studies may be affected by the presence of immortal-time bias, which exists when treatment does not occur on Day 0 of the study. This bias can result in researchers overestimating a treatment benefit, or even observe a treatment benefit when none exists. In this paper, we describe what immortal-time bias is, the challenges it presents, and how to recognize and address it using the real-world example of surgical resection of the primary tumor for stage IV breast cancer throughout. In our example, we guide researchers and illustrate how the early studies, which did not account for immortal-time bias, suggested a protective benefit of surgery, and how these results were supplanted by more recent studies through identifying and addressing immortal-time bias in their design and analyses.
外科健康服务研究人员越来越多地利用观察性数据来评估治疗与结果之间的关联，特别是因为一些手术无法通过随机对照试验进行评估。然而，许多此类研究的结果可能会受到永生时间偏倚的影响，当治疗不是在研究的第0天进行时，就会出现这种偏倚。这种偏倚可能导致研究人员高估治疗益处，甚至在不存在治疗益处时观察到治疗益处。在本文中，我们将描述什么是永生时间偏倚、它带来的挑战，以及如何通过贯穿始终的IV期乳腺癌原发肿瘤手术切除的真实案例来识别和解决它。在我们的案例中，我们指导研究人员，并说明早期未考虑永生时间偏倚的研究如何表明手术具有保护作用，以及最近的研究如何通过在设计和分析中识别和解决永生时间偏倚而取代了这些结果。
