Tomlinson Laurie A, Fogarty Damian G, Douglas Ian, Nitsch Dorothea
Faculty of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London, UK.
Department of Nephrology, Belfast Health and Social Care Trust, Belfast, Northern Ireland, UK.
Nephrol Dial Transplant. 2017 Apr 1;32(suppl_2):ii40-ii46. doi: 10.1093/ndt/gfw349.
Pharmacoepidemiology studies are increasingly used for research into safe prescribing in chronic kidney disease (CKD). Typically, patients prescribed a drug are compared with patients who are not on the drug and outcomes are compared to draw conclusions about the drug effects. This review article aims to provide the reader with a framework to critically appraise such research. A key concern in pharmacoepidemiology studies is confounding, in that patients who have worse health status are prescribed more drugs or different agents and their worse outcomes are attributed to the drugs not the health status. It may be challenging to adjust for this using statistical methods unless a comparison group with a similar health status but who are prescribed a different (comparison) drug(s) is identified. Another challenge in pharmacoepidemiology is outcome misclassification, as people who are more ill engage more often with the health service, leading to earlier diagnosis in people who are frequent attenders. Finally, using replication cohorts with the same methodology in the same type of health system does not ensure that findings are more robust. We use two recent papers that investigated the association of proton pump inhibitor drugs with CKD as a device to review the main pitfalls of pharmacoepidemiology studies and how to attempt to mitigate against potential biases that can occur.
药物流行病学研究越来越多地用于慢性肾脏病（CKD）安全处方的研究。通常，将使用某种药物的患者与未使用该药物的患者进行比较，并比较结果以得出关于药物效果的结论。这篇综述文章旨在为读者提供一个框架，以便对这类研究进行批判性评估。药物流行病学研究中的一个关键问题是混杂因素，即健康状况较差的患者会被开更多的药或不同的药物，而他们较差的预后被归因于药物而非健康状况。除非确定一个健康状况相似但使用不同（对照）药物的对照组，否则使用统计方法对此进行调整可能具有挑战性。药物流行病学中的另一个挑战是结局错误分类，因为病情较重的人更频繁地使用医疗服务，导致经常就诊的人更早被诊断出来。最后，在同一类型的卫生系统中使用相同方法的重复队列并不能确保研究结果更可靠。我们使用两篇最近研究质子泵抑制剂药物与CKD关联的论文，作为一种手段来审视药物流行病学研究的主要缺陷，以及如何试图减轻可能出现的潜在偏倚。
