Addressing "immortal time bias" in the evaluation of treatment outcomes for post-infarction ventricular septal defect: implications for clinical practice and research

Abstract

Abstract Introduction Post-infarction ventricular septal defect (VSD) is a rare but life-threatening complication of acute myocardial infarction (AMI). Surgical closure (SC) has traditionally been the standard treatment, but in recent years, percutaneous transcatheter (PC) closure has emerged as an alternative. This study aims to investigate the potential for "immortal time bias" in treatment selection for post-infarction VSD. Methods 8,022 patients who presented with STEMI, of whom 80 (1%) had a VSD. Among these: 26 SC, 20 underwent PC, and 34 received medical management (MM). The primary endpoint was in-hospital mortality. We used the Kaplan–Meier curves (Fig 1C) to plot survival, and to address immortal time bias, we used the time to closure as a time-dependent covariate and conducted a landmark analysis using time-dependent Cox proportional hazard regression. Significance was set at P≤0.05. Results The baseline and admission characteristics in the 3 groups are summarized in Table 1. The mean age was 64.4 years, and most were men (78.8%) with comorbidities such as DM(48.8%) and HTN(57.5%). At presentation, patients managed with MM were more likely to have high-risk features, including lower systolic blood pressure, LVEF; and higher right ventricular dysfunction. 21 patients (26.3%) presented with cardiogenic shock, which was more common in the MM group (SC 11.5%, PC 10%, vs MM 47.1%; P=0.001). The delay between the onset of AMI-arrival was similar in the SC and PC groups (P=0.95) but longer compared to MM (6 vs 5 vs 0.8 days, respectively, P<0.001). Patients managed with MM had a shorter time from onset of STEMI-VSD diagnosis (4 [2-9] SC, 4 [2-20] PC, and 2 [1-4] days for MM; P=0.03). The time from onset of STEMI-closure was 22 (12-45) days, with no significant difference in closure achievers (P=0.78) (Fig 1A). Only 19 patients (23.8%) received reperfusion therapy, most receiving thrombolytic treatment, and only 8 receiving pPCI. Inotropics, vasopressors, and invasive ventilation were used more frequently in the MM group. To evaluate the potential for survival bias, logistic regression was performed with the closure of VSD versus the time of AMI-event (closure or death) as the outcome variable for the entire cohort. The analysis showed a time-dependent gradient of reaching closure (P<0.001) with OR=1.15 (1.08-1.23) per day since AMI. This finding was corroborated by a spline that showed linearity df=1 X2=13.2 P<0.001. In the time-dependent analysis of the cohort, there were no significant differences in mortality (P=0.331) with a HR=1.69 (0.75-3.81; P=0.204) for SC and 0.78 (0.33-1.87; P=0.578) for PC vs MM (Fig1C). Conclusion Results showed a time-dependent gradient of reaching VSD closure, and a time-varying analysis showed no differences in mortality between SC or PC compared to MM. The findings of this study suggest that future studies evaluating outcomes for post-infarction VSD should take into account the potential for immortal time bias.Table 1Figure 1