Challenges and new insights on estimating the effect of reintervention after EVAR

Background The effect of reintervention after elective endovascular abdominal aortic aneurysm repair (EVAR) on survival remains uncertain. Existing studies report inconsistent findings and may be vulnerable to immortal time bias because reintervention occurs at variable times during follow-up, and to confounding bias from time-evolving postoperative factors such as endoleak. Methods We retrospectively analyzed Vascular Quality Initiative registry data for patients undergoing elective EVAR. Reintervention within 2 years after EVAR was modeled as a time-varying exposure, and the outcome was all-cause mortality. The primary analysis used sequential target trial emulation using pooled sequential Cox proportional-hazards models to account for delayed reintervention timing and adjust for baseline and time-varying confounders. A sequential instrumental variable Cox analysis was also performed to address potential unmeasured confounding. Results In the primary sequential Cox analysis and the IV analysis extension, reintervention was associated with higher all-cause mortality (HR: 1.24 [1.09–1.41], p < 0.01 and HR: 1.27 [1.07–1.52], p = 0.01, respectively). In contrast, the naïve Cox model analysis, which is vulnerable to immortality bias, showed no significant association between reintervention and mortality, and the unadjusted time-dependent Cox model, which is vulnerable to unmeasured confounding bias, yielded a stronger adverse association. Conclusion After accounting for reintervention timing and time-varying confounding, post-EVAR reintervention was associated with modestly higher all-cause mortality. These findings do not establish that reintervention is overused or intrinsically harmful; rather, they suggest that reintervention should not be assumed to universally confer a survival benefit and may be a marker of complicated post-EVAR disease course, device failure, endoleak, higher-risk anatomy, or disease progression. These findings highlight the importance of addressing immortal time bias, time-dependent confounding and unmeasured confounding when evaluating vascular procedures performed after follow-up begins.