Published online Nov 24, 2026. doi: 10.5306/wjco.126582
Revised: September 24, 2026
Accepted: November 2, 2026
Published online: November 24, 2026
Processing time: 101 Days and 18.1 Hours
Postoperative local-recurrence estimates could inform surveillance discussions after soft tissue sarcoma surgery, but machine-learning performance can be over
To develop and evaluate a model for binary 3-year local-recurrence status among patients with ascertainable status and compare it with prespecified logistic-regre
This retrospective two-institution study analyzed 1035 development patients with ascertainable binary recurrence status at 3 years (331 recurrences) and 212 patients in a second-institution cohort (67 recurrences). The estimand was recurrence-status probability conditional on 3-year status being ascertainable, not cumulative incidence in all postoperative patients. The locked analytical package did not re
CatBoost had a mean internal area under the receiver operating characteristic curve (AUC) of 0.8126 and Brier score of 0.1584. In the second institution, its AUC was 0.8050 [95% confidence interval (CI): 0.7433-0.8668], Brier score 0.1615 (95%CI: 0.1331-0.1912), log loss 0.4884 (95%CI: 0.4146-0.5678), calibration intercept 0.148 (95%CI: -0.222 to 0.490), and calibration slope 0.985 (95%CI: 0.703-1.357). The CatBoost-minus-six-variable-logistic differences were -0.0083 for AUC (95%CI: -0.0480 to 0.0299) and 0.0043 for Brier score (95%CI: -0.0112 to 0.0198). A secondary full 12-variable logistic model also showed no material difference from CatBoost.
In the protocol-frozen second-institution re-evaluation, CatBoost retained discrimination but did not demonstrate incremental performance over logistic comparators. The study was not designed or powered to establish equiva
Core Tip: We compared 243 configurations across 14 base-model families plus three prespecified ensembles by repeated nested cross-validation. The selected CatBoost model was frozen and applied without recalibration in a protocol-frozen second-institution re-evaluation. CatBoost retained an area under the receiver operating characteristic curve of 0.8050 but did not demonstrate incremental performance over either a six-variable or full 12-variable logistic model; this was not an equivalence or non-inferiority analysis. The browser interface is provided for reproducibility and independent evaluation, not for clinical decision-making.