Introduction: Observational studies are increasingly being used to provide evidence on the real-world effectiveness of medications for preventing vascular diseases, such as stroke. We investigated whether the real-world effectiveness of treatment with lipid-lowering medications after ischemic stroke is affected by prevalent user bias. Methods: An observational cohort study of 90-day survivors of ischemic stroke, using person-level data from the Australian Stroke Clinical Registry (2012-2016; 45 hospitals) linked to administrative (pharmaceutical, hospital, death) records. Use of, and adherence to (proportion of days covered <80% [poor adherence] vs. ≥80% [good adherence]), lipid-lowering medications within 90 days post-discharge was determined from pharmaceutical records. Users were further classified as prevalent (continuing) or new-users, based dispensing within 90 days prior to stroke. Propensity score-adjusted Cox regression was used to evaluate the effectiveness of lipid-lowering medications on outcomes (all-cause mortality, all-cause and cardiovascular disease readmission) within the subsequent year. Analyses were undertaken using prevalent-user (all users vs. non-users) and new-user designs (new-users vs. non-users). Results: Of 11,217 eligible patients (median age 72 years, 42% female), 9,294 (83%) used lipid-lowering medications within 90 days post-discharge, including 5,479 new-users. In both prevalent-user and new-user designs, non-users (vs. users) had significantly greater rates of mortality (hazard ratio [HR] 2.35, 95% CI 1.89-2.92) or all-cause readmissions (HR 1.22, CI 1.05-1.40), but not cardiovascular disease readmission. In contrast, associations between having poor (vs. good) adherence on outcomes were stronger among new-users than all-users. Among new-users, having poor adherence was associated with greater rates of mortality (HR 1.48, CI 1.12-1.96), all-cause readmission (HR 1.14, CI 1.02-1.27), and cardiovascular disease readmission (HR 1.20, CI 1.01-1.42). Conclusions: The real-world effectiveness of treatment with lipid-lowering medications after stroke is attenuated when evaluated based on prevalent-user rather than new-user design. These findings may have implications for designing studies on the real-world effectiveness of secondary prevention medications.


中文翻译：

---

降脂药物对中风后结局的真实效果：新用户设计的潜在影响

简介：越来越多的观察性研究被用于提供证据，证明药物在预防血管疾病（如中风）方面的真实有效性。我们调查了缺血性中风后降脂药物治疗的真实效果是否受到普遍的用户偏见的影响。方法：一项针对缺血性中风 90 天幸存者的观察性队列研究，使用来自澳大利亚中风临床登记处（2012-2016 年；45 家医院）的个人数据与行政（药物、医院、死亡）记录相关联。从药物记录确定出院后 90 天内降脂药物的使用和依从性（覆盖天数的比例 <80% [依从性差] 与 ≥80% [依从性好]）。用户被进一步分类为普遍（持续）或新用户，中风前 90 天内基于配药。倾向评分调整后的 Cox 回归用于评估降脂药物在随后一年内对结果（全因死亡率、全因和心血管疾病再入院）的有效性。使用普遍用户（所有用户与非用户）和新用户设计（新用户与非用户）进行分析。结果：在 11,217 名符合条件的患者（中位年龄 72 岁，42% 为女性）中，9,294 名 (83%) 在出院后 90 天内使用了降脂药物，其中包括 5,479 名新用户。在流行用户和新用户设计中，非用户（与用户相比）的死亡率（风险比 [HR] 2.35，95% CI 1.89-2.92）或全因再入院率（HR 1.22，CI 1.05-1.40)，但不是心血管疾病再入院。相比之下，与所有用户相比，新用户对结果的依从性差（与好）之间的关联更强。在新使用者中，依从性差与更高的死亡率（HR 1.48，CI 1.12-1.96）、全因再入院率（HR 1.14，CI 1.02-1.27）和心血管疾病再入院率（HR 1.20，CI 1.01- 1.42). 结论：当基于普遍用户而不是新用户设计进行评估时，中风后降脂药物治疗的真实效果会减弱。这些发现可能对设计二级预防药物在现实世界中有效性的研究有影响。全因再入院（HR 1.14，CI 1.02-1.27）和心血管疾病再入院（HR 1.20，CI 1.01-1.42）。结论：当基于普遍用户而不是新用户设计进行评估时，中风后降脂药物治疗的真实效果会减弱。这些发现可能对设计二级预防药物在现实世界中有效性的研究有影响。全因再入院（HR 1.14，CI 1.02-1.27）和心血管疾病再入院（HR 1.20，CI 1.01-1.42）。结论：当基于普遍用户而不是新用户设计进行评估时，中风后降脂药物治疗的真实效果会减弱。这些发现可能对设计二级预防药物在现实世界中有效性的研究有影响。