Institute of Health and Wellbeing, University of Glasgow, 1 Lilybank Gardens, Glasgow G12 8RZ, UK.
Institute of Health and Wellbeing, University of Glasgow, 1 Lilybank Gardens, Glasgow G12 8RZ, UK.
J Clin Epidemiol. 2022 Aug;148:74-80. doi: 10.1016/j.jclinepi.2022.04.020. Epub 2022 Apr 18.
To review comparative effectiveness research (CER) using physician's prescribing preference as an instrumental variable (PPP IV) in pharmacoepidemiology and to review methodological studies that use simulation to evaluate the performance of PPP IV in CER.
We conducted a review of CER using PPP IV and studies evaluating the use of PPP IV by using simulation methods. We searched Ovid, PubMed, and Google Scholar databases from 2005 to 2020.
We identified six simulation studies and 18 CER studies. The simulation studies explored the most suitable ways for using PPP IV in different settings (outcome types, sample size, and the prevalence of outcomes) which can be useful guidance for using PPP IV in CER. The CER studies identified show heterogeneity in terms of validation assumptions, estimation methods, and sample size. Not all applied studies used the methodological insights from the simulation studies. However, they all concluded that PPP is a valid IV.
Future CER should consider a range of methodological issues to improve the validity of findings when using PPP IV. Specifically, studies should consider the impact of a different choice of statistical methods, forms of proxy for measuring preference, time-varying exposures, and the type of outcome.
综述以医生处方偏好为工具变量（PPPIV）的药物流行病学中的比较效果研究（CER），并综述使用模拟方法评估 PPPIV 在 CER 中性能的方法学研究。
我们对使用 PPPIV 的 CER 进行了综述，并对使用模拟方法评估 PPPIV 使用的研究进行了综述。我们检索了 2005 年至 2020 年 Ovid、PubMed 和 Google Scholar 数据库。
我们确定了 6 项模拟研究和 18 项 CER 研究。模拟研究探讨了在不同环境（结局类型、样本量和结局的发生率）中使用 PPPIV 的最合适方法，这对 CER 中使用 PPPIV 具有有用的指导意义。所识别的 CER 研究在验证假设、估计方法和样本量方面存在异质性。并非所有应用研究都利用了模拟研究的方法学见解。然而，它们都得出结论认为，PPPI 是一种有效的 IV。
未来的 CER 应考虑一系列方法学问题，以提高使用 PPPIV 时发现结果的有效性。具体而言，研究应考虑不同统计方法选择、偏好测量的代理形式、时变暴露和结局类型的影响。
