Predictors of one-year adverse outcomes after laparoscopic resection for hepatocellular carcinoma: Development and validation of an early-warning model

BACKGROUND The global burden of primary liver cancer (PLC) continues to rise. Although minimally invasive, especially laparoscopic, resection is increasingly performed for early-stage disease, 1-year adverse outcomes (recurrence, metastasis, or mortality) remain common. Widely used scores, such as the albumin-bilirubin grade, primarily assess hepatic reserve and may not fully reflect tumor biology or systemic inflammation for individualized early prognostic warning. This study aimed to develop and validate a least absolute shrinkage and selection operator (LASSO)-based model to predict 1-year adverse outcomes after minimally invasive PLC resection. AIM To identify predictors of short-term (1-year) adverse outcomes following minimally invasive PLC resection and construct an individualized postoperative prognostic model using LASSO regression. METHODS This retrospective study included patients with PLC who underwent minimally invasive resection at The Affiliated Suqian Hospital of Xuzhou Medical University between January 2019 and January 2023. Prognostic predictors were identified using LASSO regression and incorporated into a logistic regression model. Model performance and clinical utility were evaluated using receiver operating characteristic curves, calibration plots, and decision curve analysis. The dataset was randomly divided into training (n = 277) and internal validation (n = 144) cohorts. An external validation cohort of 138 patients with PLC (February 2023 to June 2024) was used to assess generalizability. RESULTS Receiver operating characteristic analysis indicated good performance of the logistic regression model based on six predictors, white blood cell count, tumor diameter, vascular invasion, portal vein infiltration, cirrhosis, and alpha-fetoprotein, with area under the curve (AUC) values of 0.756 [95% confidence interval (CI): 0.687-0.824] and 0.750 (95%CI: 0.659-0.841) in the training and internal validation cohorts, respectively. The model exhibited strong calibration (training, P = 0.6951; external validation, P = 0.5223) and clear net clinical benefit across risk thresholds. External validation further supported its generalizability (n = 138; AUC = 0.735, 95%CI: 0.640-0.830). Compared with albumin-bilirubin, the LASSO-based risk score showed higher though non-significant AUCs in the training (0.756 vs 0.691; DeLong P = 0.206) and external (0.735 vs 0.717; P = 0.803) cohorts and comparable performance in the internal validation cohort (0.750 vs 0.753; P = 0.968). CONCLUSION LASSO regression was used to identify six independent predictors of adverse 1-year outcomes after minimally invasive PLC resection. The resulting risk score model demonstrates reliable discrimination, calibration, and clinical utility for individualized prognostic assessment.