METHODS: Registered in PROSPERO (CRD420251233077). We searched six databases from inception to November 2025 for adult observational cohorts comparing endoscopic closure versus surgery. Outcomes included treatment success, mortality, major morbidity, reoperation, length of stay, and fasting duration. Two reviewers independently screened studies, extracted data, and assessed bias using ROBINS-I. Random-effects models pooled risk ratios (RR) and mean differences (MD); GRADE rated certainty. Sensitivity and geographic subgroup analyses assessed robustness and effect modification.
RESULTS: Four retrospective cohorts (n = 123; 52 endoscopic, 71 surgical) from Portugal, Korea, and Malaysia were included across care settings. Treatment success showed no clear difference (RR 1.00, 95% CI 0.94-1.06; I2 = 0%; low certainty). Mortality was rare and imprecise (8 events; RR 0.26, 95% CI 0.06-1.16; I2 = 0%; very low certainty). Hospital stay was shorter with endoscopic management (MD - 9.23 days, 95% CI - 13.74 to - 4.73; I2 = 43%; low certainty). Fasting duration did not differ significantly and was heterogeneous. No geographic subgroup effect was detected (P = 0.95). Sensitivity analysis supported robustness, except for hospital-stay heterogeneity driven by referred cases in one study.
CONCLUSIONS: In observational cohorts, endoscopic closure was typically used for immediately recognized, smaller perforations in favorable clinical conditions, whereas surgery was preferentially used for delayed diagnosis, larger defects, or suspected contamination, introducing substantial confounding by indication. Accordingly, the pooled estimates should not be interpreted as evidence of equivalence. In carefully selected patients (immediate recognition,