External validation of hepatocellular carcinoma recurrence model after surgery using the Royston-Altman approach

Liver cancer has a high recurrence rate (up to 70 %) after hepatectomy, which significantly compromises long-term survival. No study has comprehensively compiled and externally validated Cox models for postoperative relapse.

We systematically searched the relevant articles from PubMed and Web of Science databases. Patients with hepatocellular carcinoma who underwent curative resection in the Department of Liver Surgery, West China Hospital of Sichuan University were retrospectively recruited and followed up regularly. Level 1 and Level 2 validation were conducted for the Cox models according to Royston-Altman approach. Kaplan-Meier curves were generated for each Level-2 validation.

Our cohort was used to externally validate ten of the sixteen studies. The calibration slope (γ) of the five Cox models based on the Prognostic Index were 0.240, 0.383, 0.484, 0.634 and 1.129 in Level 1 validation and -0.030, 0.252, 0.264, 0.289 and 0.627 in Level 2 validation. Only one model effectively distinguished the recurrence risk between high and low-risk groups in Level 2 validation.

Without refitting the model, an optimal model was identified in both Level 1 and Level 2 validation. Our research methodology may be applicable to external validation studies across different regions and institutions.