Validation of an alpha-fetoprotein model to predict recurrence after liver resection for hepatocellular carcinoma

To validate a previously reported alpha-fetoprotein (AFP) model (including three variables: preoperative image-diagnosed tumor number and size and AFP level) for the prediction of recurrence in hepatocellular carcinoma (HCC) patients who have undergone liver resection (LR). This retrospective study enrolled patients who underwent curative LR for newly diagnosed HCC in our institution between 2011 and 2018. The probabilities of overall survival (OS) and recurrence were compared according to the aforementioned AFP model. A total of 838 patients were included. AFP score ≥ 3 versus ≤ 2 independently predicted recurrence and OS. However, net reclassification improvements (NRI) indicated that the AFP model was not superior to the Barcelona Clinic Liver Cancer (BCLC) system for predicting 1-year recurrence (p = 0.746). Relatedly, we developed a modified AFP model based on our cohort. The modified AFP score ≥ 3 versus ≤ 2 independently predicted recurrence and OS. However, NRI again indicated that the modified AFP model was not superior to the BCLC system for predicting 1-year recurrence (p = 0.69). Patients with a modified AFP score ≤ 2 had a risk of recurrence similar to that of patients with a modified AFP score ≥ 3 in BCLC stage 0-A (p = 0.57). However, patients with a modified AFP score ≤ 2 had a lower risk of recurrence than patients with a modified AFP score ≥ 3 in BCLC stage B-C (p = 0.02). The original AFP model was not feasible in our cohort. However, the modified AFP model may be useful for predicting recurrence in BCLC B-C patients who underwent LR in our cohort.