类风湿关节炎中JAK抑制剂与TNF抑制剂的心血管和死亡率结局：一项目标试验模拟（CARDINAL-RA）。
完整摘要 · 中文译文
目的：本研究旨在通过靶向试验模拟，比较在具有较高心血管风险的类风湿关节炎（RA）患者中，Janus激酶抑制剂（JAKi）与肿瘤坏死因子抑制剂（TNFi）的心血管结局和死亡结局。 方法：在TriNetX数据库中，对年龄≥50岁且至少有1项心血管危险因素、开始使用JAKi或TNFi的成人RA患者，采用新用药者、活性对照设计进行比较。改良的1:1倾向性评分匹配纳入了人口统计学特征、心血管危险因素、合并症、个体常规合成改善病情抗风湿药（csDMARDs）、口服糖皮质激素、既往非TNF生物制剂暴露史以及基线炎症标志物。主要结局为主要不良心血管事件（MACE）（非致死性心肌梗死或非致死性缺血性卒中）；次要结局为全因死亡、心力衰竭住院和静脉血栓栓塞（VTE），并预设白内障作为阴性对照。随访从用药开始后90天起计。采用Cox模型估计风险比，并计算5年Kaplan-Meier累积发生率和每1000人绝对风险差异。队列分别按药物、性别、心血管风险分层和用药起始时代进行匹配；E值用于量化未测量混杂的敏感性。 结果：在18,757对匹配患者中，JAKi未与MACE风险增加相关（HR 1.002，0.923-1.088；每1000人+2.11）。次要结局全因死亡在JAKi组较高（HR 1.158，1.058-1.268；p = 0.0014；每1000人+10.4；E值1.586），而心力衰竭住院（HR 1.019，0.955-1.088）和VTE（HR 1.054，0.961-1.156）无差异；阴性对照结 对照组未显示超额风险（HR 0.934, 0.867-1.007）。没有任何药物在MACE或死亡率方面显示显著超额（死亡率HR：托法替布1.052，巴瑞替尼1.213，乌帕替尼1.090）；唯一名义上显著的药物层面发现是巴瑞替尼的VTE（HR 1.364, 1.028-1.809；198例事件）。包含死亡的三组分复合终点无显著性（HR 1.058, 0.992-1.127）。未观察到性别、心血管风险负担或索引年份的一致效应修饰。 结论：JAKi启动与MACE（主要结局）、心力衰竭住院或VTE的超额风险无关。作为次要结局观察到全因死亡率的适度超额，且未发现任何单一药物的信号。该关联的稳健性适中（E值1.586），且未在任何药物特异性队列中重现，因此不能排除残余通道偏倚。MACE阴性结果不应被解读为类效应的心血管安全性保证，死亡率信号也不应被视为已确立的因果效应。
完整摘要 · 英文原文
OBJECTIVES: This study aims to compare cardiovascular and mortality outcomes of Janus kinase inhibitors (JAKi) versus tumor necrosis factor inhibitors (TNFi) in rheumatoid arthritis (RA) at elevated cardiovascular risk using target trial emulation. METHODS: Within TriNetX, adults with RA aged ≥50 years and ≥1 cardiovascular risk factor initiating JAKi or TNFi were compared in a new-user, active-comparator design. Improved 1:1 propensity score matching incorporated demographics, cardiovascular risk factors, comorbidities, individual conventional synthetic DMARDs, oral corticosteroids, prior non-TNF biologic exposure, and baseline inflammatory markers. The primary outcome was MACE (nonfatal myocardial infarction or nonfatal ischemic stroke); the secondary outcomes were all-cause mortality, heart failure hospitalization, and venous thromboembolism (VTE), with cataract as a prespecified negative control. Follow-up began 90 days after initiation. Cox models estimated hazard ratios, with 5-year Kaplan-Meier cumulative incidences and absolute risk differences per 1,000. Cohorts were separately matched by agent, sex, cardiovascular-risk stratum, and index era; E-values quantified sensitivity to unmeasured confounding. RESULTS: In 18,757 matched pairs, JAKi was not associated with excess MACE (HR 1.002, 0.923-1.088; +2.11 per 1,000). All-cause mortality, a secondary outcome, was higher with JAKi (HR 1.158, 1.058-1.268; p = 0.0014; +10.4 per 1,000; E-value 1.586), whereas heart failure hospitalization (HR 1.019, 0.955-1.088) and VTE (HR 1.054, 0.961-1.156) did not differ; the negative control showed no excess (HR 0.934, 0.867-1.007). No agent showed a significant excess in MACE or mortality (mortality HR 1.052 tofacitinib, 1.213 baricitinib, and 1.090 upadacitinib); the only nominally significant agent-level finding was VTE with baricitinib (HR 1.364, 1.028-1.809; 198 events). A three-component composite including death was non-significant (HR 1.058, 0.992-1.127). No consistent effect modification was seen by sex, cardiovascular risk burden, or index year. CONCLUSION: JAKi initiation was not associated with excess MACE (primary outcome), heart failure hospitalization, or VTE. A modest excess in all-cause mortality was observed as a secondary outcome, with no signal for any individual agent. This association was of modest robustness (E-value 1.586) and was not reproduced in any agent-specific cohort, so residual channeling bias cannot be excluded. The null MACE result should not be read as class-wide cardiovascular safety reassurance nor the mortality signal as an established causal effect.