Answers are grounded in this paper and cite their sources.
To derive and internally evaluate an exploratory clinicoradiologic model for splanchnic vein thrombosis (SVT) detected on index contrast-enhanced CT (CECT) in acute pancreatitis (AP), and to assess its association with pancreatic necrosis volume (PNV). This single-center retrospective study included 360 adults with AP. Non-contrast CT (NCCT) preceded index CECT in all patients. SVT detected on index CECT was considered prevalent rather than incident. A ten-predictor elastic-net logistic model was evaluated using repeated nested cross-validation and bootstrap resampling. Its performance was compared with PNV-only models using ordinal and nominal coding. Among 360 patients (276 men; median age, 41 years [IQR, 34–50]), 47 (13.1%) had SVT. The ten-predictor model achieved an AUC of 0.765 (95% CI 0.679–0.838; bootstrap-corrected AUC, 0.762) and a Brier score of 0.099. At a training-derived threshold of 0.14, mean sensitivity, specificity, positive predictive value, and negative predictive value were 0.66, 0.77, 0.31, and 0.94, respectively, with 15.9 missed SVT events per 360-patient validation repeat. Across matched folds, mean AUCs were 0.772 for the ten-predictor model and 0.763 for ordinal PNV-only (paired ΔAUC, 0.009; 95% CI − 0.027 to 0.046; P = 0.608). With nominal PNV coding, the paired ΔAUC was 0.013 (95% CI − 0.027 to 0.053; P = 0.510). Brier scores and decision-curve net benefit were similar, whereas mean average precision was higher for the ten-predictor model under both codings (ΔAP, 0.074 and 0.089; unadjusted secondary P = 0.006 for both). The ten-predictor model showed moderate internal discrimination but no clear advantage over PNV-only models in AUC, Brier score, or decision-curve net benefit. Higher average precision is hypothesis-generating. Independent validation is required before clinical application. In this exploratory cohort, PNV was associated with SVT detected on index CECT. Its descriptive discrimination was higher than that of the recorded composite imaging scores, but this finding does not establish clinical superiority. The results are hypothesis-generating and require independent validation.
No takes yet. Share an insight, caveat, or question.
Xiang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: