Long-Term Survival and Beneficiaries of Adjuvant Anti-PD-1 Therapy in Resected Hepatocellular Carcinoma.

[BACKGROUND] Early and late hepatocellular carcinoma (HCC) recurrences, which are driven by residual and de novo tumors, respectively, differ in biology and potential treatment susceptibility. The effectiveness of adjuvant PD-1 inhibitors (aPD-1) across these distinct recurrence patterns remains unclear.

[PATIENTS AND METHODS] A total of 3436 cases were included from 9 centers in China. Among these, 1662 patients (2019-2023) were propensity score-matched to compare the effects of aPD-1 versus active surveillance on survival outcomes. Machine learning (ML) models were trained on 1774 patients (2014-2018) to predict early recurrence (≤ 2 years). The efficacy of aPD-1 was further assessed in ML-predicted subgroups, using a 2-year landmark analysis to evaluate its effect on recurrence etiology.

[RESULTS] The logistic regression model achieved the best performance, with an AUC of 0.818 in the test set. Patients classified as high risk showed significantly poorer disease-free survival (DFS) and overall survival. aPD-1 significantly improved DFS in high-risk patients (HR 0.69, 95% CI 0.57-0.84, p < 0.001), but no benefit was observed in low-risk patients (HR 0.99, 95% CI 0.71-1.39, p = 0.950). Landmark analysis further demonstrated that aPD-1 preferentially reduced early recurrence in high-risk HCC (HR 0.67, 95% CI 0.54-0.83, p < 0.001), while showing limited efficacy against late recurrence (HR 0.81, 95% CI 0.49-1.31, p = 0.380).

[CONCLUSIONS] aPD-1 demonstrates time-dependent efficacy, primarily reducing early recurrence through controlling residual micro-metastases rather than preventing de novo tumors. A ML model accurately identified patients most likely to benefit, highlighting the importance of risk-guided adjuvant therapy.