Temporal evaluation of postoperative day 1 biomarkers for diagnosed portal vein thrombosis after splenectomy in Wilson’s disease

BACKGROUND Postoperative day 1 (POD1) laboratory changes may be associated with portal vein thrombosis (PVT) diagnosed after splenectomy in Wilson's disease, but retrospective sampling and surveillance schedules can leave the biological sequence uncertain.AIM To reassess discrimination, calibration and uncertainty of archived POD1 model probabilities in a same-centre temporal cohort and examine the limits of their clinical interpretation.METHODS The retrospective cohort included 771 patients operated on during 2001-2019: 612 in development and 159 in temporal validation, with 160 and 42 PVT diagnoses, respectively.The previously reported ExtraTrees probabilities were retained without refitting or recalibration.We recomputed performance, five-bin calibration, risk-stratum outcomes and decision-curve uncertainty using 2000 patient bootstrap resamples, and performed recorded-diagnosis landmark and management-stratified sensitivity analyses.RESULTS Temporal-validation area under the curve was 0.835 (95% confidence interval: 0.758-0.905),average precision was 0.760 (0.641-0.858), and Brier score was 0.136 (0.110-0.165).Calibration-in-the-large was -0.091 (-0.461 to 0.265), and calibration slope was 1.483 (1.066 to 2.132).PVT was diagnosed in 6/53, 10/56 and 26/50 temporal-validation patients in the < 15%, 15%-35% and > 35% strata.No diagnoses were recorded on POD0-POD1, but sampling timestamps and contemporaneous negative imaging were unavailable.A 4 / 36 POD7 recorded-diagnosis landmark retained 157 validation patients and 40 subsequent diagnoses (area under the curve 0.827; 95% confidence interval: 0.737-0.901).CONCLUSION Archived probabilities discriminate recorded 30-day PVT in this same-centre cohort.Unresolved predictor-outcome ordering, management and follow-up coding, calibration uncertainty and unavailable original model serialization preclude claims of established incident-risk prediction or clinical readiness.Source-record adjudication and external validation remain necessary.