Early Antiplatelet Therapy in Upper Gastrointestinal Bleeding with Coronary Artery Disease: A Landmark Analysis of MIMIC-IV and eICU-CRD Databases.

INTRODUCTION

The safety of early antiplatelet administration versus withholding in patients with concurrent upper gastrointestinal bleeding (UGIB) and coronary artery disease (CAD) or myocardial infarction (MI) remains unresolved. This study aimed to evaluate the association between early antiplatelet therapy timing and mortality in this population.

METHODS

This retrospective cohort study included 3,422 adults with UGIB and concurrent CAD or MI from the Medical Information Mart for Intensive Care IV (MIMIC-IV) dataset (2008-2019). Patients were classified as having antiplatelet therapy ordered within 24 hours (AP-24h; n=1,530), between 24 and 72 hours (AP-24-72h; n=368), or no antiplatelet therapy within 72 hours (No-AP; n=1,524; reference). This study quantified immortal time bias using 72-hour mortality distributions, applied landmark analysis to eliminate this bias, and externally validated findings using the eICU Collaborative Research Database (eICU-CRD). After excluding 87 patients dying within 72 hours, the MIMIC-IV landmark cohort comprised 3,335 patients. The primary outcome was 1-year all-cause mortality from the 72-hour landmark point. eICU-CRD external validation applied the same phenotype and exposure timing; after excluding 137 early deaths, the validation cohort included 2,657 patients. The eICU-CRD outcome was post-landmark hospital mortality.

RESULTS

Landmark analysis demonstrated a 72-hour mortality rate of 3.7% in No-AP versus 1.8% in AP-24h, confirming misclassification of early critically ill patients into the reference group. Adjusted landmark Cox regression showed no significant difference in 1-year all-cause mortality for AP-24h (hazard ratio (HR): 0.910, 95% confidence interval (CI): 0.806-1.027, p=0.125) or AP-24-72h (HR: 0.884, 95% CI: 0.733-1.066, p=0.197) compared with No-AP. In-hospital mortality was similar across groups (8.7-9.8%); adjusted logistic regression showed no significant differences for AP-24h (OR: 1.056, 95% CI: 0.786-1.419, p=0.718) or AP-24-72h (OR: 1.007, 95% CI: 0.646-1.568, p=0.976). Thirty-day mortality was also similar for AP-24h (HR: 0.934, 95% CI: 0.780-1.118, p=0.455) and AP-24-72h (HR: 1.027, 95% CI: 0.785-1.343, p=0.847). In the eICU-CRD validation cohort, post-landmark hospital mortality was 8.69%, 5.07%, and 8.00% for AP-24h, AP-24-72h, and No-AP, respectively. Fully adjusted validation models showed no increased in-hospital mortality for AP-24h (OR: 0.97, 95% CI: 0.62-1.53, p=0.895) or AP-24-72h (OR: 0.74, 95% CI: 0.32-1.69, p=0.469).

CONCLUSION

After correcting for immortal time bias using landmark analysis, early antiplatelet administration in UGIB patients with CAD or MI was not associated with increased 1-year all-cause mortality or in-hospital mortality in the MIMIC-IV dataset, and this finding was externally supported in the eICU-CRD dataset.