Tamim H, Monfared A A Tahami, LeLorier J
Pharmacoepidemiology and Pharmacoeconomics Unit, University of Montreal Hospital Center, Hôtel-Dieu Hospital, Montreal, Quebec, Canada.
Pharmacoepidemiol Drug Saf. 2007 Mar;16(3):250-8. doi: 10.1002/pds.1360.
To control for protopathic bias, some studies have incorporated the concept of lag-time into their exposure definition (time period before the index date that was not considered in assessing exposure). The objective of this study was to introduce a procedure to identify the best lag-time to be applied in studies where control for protopathic bias is required.
We used data from a case-control study carried out to assess the association between exposure to proton pump inhibitors (PPIs) and risk of gastric cancer, using RAMQ databases. Exposure was defined as the number of defined daily doses of PPIs dispensed during the 5-year period prior to the index date (divided into four quartiles). Thirty-one different lag-times were applied (0-30 months) based on 1-month intervals. Logistic regression was used to estimate the matched odds ratio (OR) for each lag-time. The change point in the ln(ORs) was identified by applying a two-compartmental model and a segmented regression model.
A trend of decreasing ORs was found with the application of an increasing lag-time. As an illustration, the ORs for the 1st quartile of defined daily doses, when applying the 31 different lag-times, ranged between 3.52 when applying a 0 lag-time and 0.97 when applying a 30 months lag-time. Applying the two methods for the different lag-times showed that the ORs stabilized at around 6 months.
For the purpose of controlling for protopathic bias in pharmacoepidemiological studies, we have provided a method to assess the most appropriate lag-time that should be applied for the assessment of drug exposure.
为控制原发病性偏倚，一些研究已将滞后时间的概念纳入其暴露定义（在索引日期之前未被视为评估暴露的时间段）。本研究的目的是引入一种程序，以确定在需要控制原发病性偏倚的研究中应应用的最佳滞后时间。
我们使用了一项病例对照研究的数据，该研究利用魁北克医疗保险局（RAMQ）数据库评估质子泵抑制剂（PPI）暴露与胃癌风险之间的关联。暴露定义为索引日期前5年期间发放的PPI限定日剂量数（分为四个四分位数）。基于1个月的间隔应用了31种不同的滞后时间（0至30个月）。使用逻辑回归估计每个滞后时间的匹配比值比（OR）。通过应用双室模型和分段回归模型确定ln(OR)的变化点。
随着滞后时间的增加，发现OR呈下降趋势。例如，当应用31种不同的滞后时间时，限定日剂量第一四分位数的OR在应用0滞后时间时为3.52，在应用30个月滞后时间时为0.97之间。对不同滞后时间应用这两种方法表明，OR在大约6个月时稳定下来。
为了在药物流行病学研究中控制原发病性偏倚，我们提供了一种方法来评估评估药物暴露时应应用的最合适滞后时间。
