Abstract
BACKGROUND: Considering the heavy burden on healthcare resources owing to HBV infection and the broad feasibility of transient elastography techniques in China, we aimed to construct and corroborate a liver stiffness measure (LSM)-dictated prediction model concerning hepatocellular carcinoma (HCC) development among CHB patients. METHODS: A retrospective cohort study was conducted, involving 713 consecutive patients with CHB. These patients were randomly assigned to the derivation (n = 534) and internal validation (n = 179) cohorts, respectively. Variable selection was optimized using the least absolute shrinkage and selection operator (LASSO) regression and subsequent multivariate Cox regression analysis. A corresponding nomogram was built and compared regarding discrimination, calibration, and risk stratification across the whole population. To further verify the generalizability of the predictive model, we integrated data from multiple external centers to construct two external validation cohorts for evaluation (n = 1084 and n = 623). RESULTS: During a median follow-up duration of 57 months, 48 (8.99%) patients in the derivation cohort and 18 (10.06%) patients in the internal validation cohort developed HCC. Following the LASSO alongside Cox regression analyses, 5 variables were retained and constituted the LEAST model (LSM, age, albumin, sex, and platelet) and resulting nomogram. Our proposed model demonstrated sufficiently discriminative abilities to predict cumulative HCC development, as indicated by a time-dependent area under the curve (tdAUC) of 0.838 (95% CI 0.752-0.925), 0.898 (95% CI 0.851-0.944), and 0.907 (95% CI 0.856-0.959) over 3, 5, and 8 years, respectively. Nomogram-derived risk strata can appropriately identify patients at high risk of developing HCC. Our prediction model exhibited numerically the highest AUC compared to several previous scores. Moreover, the validity and generalizability of the LEAST model were verified in 2 independent external validation cohorts, confirmed in the calibration and stratification performance. CONCLUSIONS: The LEAST model could predict HCC development in CHB patients, facilitating the identification of high-risk patients who might benefit from enhanced surveillance or early therapy.