Initial biliary drainage strategy and in-hospital outcomes in malignant biliary obstruction: A national analysis.

Abstract

583 Background: Optimal first-line biliary drainage for malignant biliary obstruction (MBO) remains debated. Endoscopic retrograde cholangiopancreatography (ERCP) and percutaneous biliary drainage (PBD) are both used, sometimes sequentially within the same admission. Comparative in-hospital effectiveness across real-world strategies is unclear. Methods: We analyzed 2020–2022 NIS adults hospitalized with malignant biliary obstruction who underwent ERCP and/or percutaneous biliary drainage, classifying the index strategy as ERCP-only, PBD-only, ERCP-first, or PBD-first. To address confounding by indication, we estimated multinomial propensity scores. Effects were estimated with survey-weighted marginal structural models. Results: Among 11,133 discharges (weighted ≈55,645), initial strategies were ERCP-only 91.2%, PBD-only 7.5%, ERCP-first 1.05%, and PBD-first 0.33%. After inverse-probability weighting, baseline covariates were balanced across strategies. Compared with ERCP-first, ERCP-only had lower in-hospital mortality (adjusted risk 3.10% vs 10.8%; aOR 0.26, 95% CI 0.14–0.49). PBD-only (aOR 0.83, 95% CI 0.43–1.61) and PBD-first (aOR 0.66, 95% CI 0.17–2.61) did not differ significantly from ERCP-first. ERCP-only was also associated with shorter length of stay (IRR 0.46, 95% CI 0.41–0.52) and lower total charges (cost ratio 0.54, 95% CI 0.47–0.61); PBD-only showed intermediate reductions (LOS IRR 0.77, 95% CI 0.68–0.88; cost ratio 0.75, 95% CI 0.65–0.86). Conclusions: ERCP-only initial strategy for MBO was associated with lower in-hospital mortality and reduced resource use compared with an ERCP-first combined approach. These findings support prioritizing definitive endoscopic drainage when feasible and may inform practice and trial design. In-hospital outcomes by initial biliary drainage strategy for malignant biliary obstruction (NIS 2020–2022). Strategy (index admission) Weighted share, % Mortality — aOR vs ERCP-first (95% CI) LOS — IRR vs ERCP-first (95% CI) Total charges — Cost ratio vs ERCP-first (95% CI) ERCP-only 91.2 0.26 (0.14–0.49) 0.46 (0.41–0.52) 0.54 (0.47–0.61) PBD-only 7.5 0.83 (0.43–1.61) 0.77 (0.68–0.88) 0.75 (0.65–0.86) ERCP-first (reference) 1.05 1.00 (Ref) 1.00 (Ref) 1.00 (Ref) Models are survey-weighted marginal structural models with stabilized IPTW multiplied by NIS discharge weights; hospitals as PSUs and NIS strata applied. Mortality modeled with logistic regression (aORs); LOS with Poisson log link (IRRs); charges with gamma log link (cost ratios). Adjusted risks (mortality): ERCP-only 3.10% vs ERCP-first 10.8% (risk difference −7.7 percentage points; 95% CI −13.5 to −1.9).PBD-first represented 0.33% of discharges; estimates were imprecise and are omitted here for clarity (mortality aOR vs ERCP-first 0.66, 95% CI 0.17–2.61); sensitivity analyses excluding PBD-first yielded similar results.