Establishment of Predictive Models for Second Hepatoma Recurrence

Given that hepatoma frequently recurs after initial resection, it is often managed by liver transplantation. The persistently high recurrence rate even after transplantation underscores the need for better prognostic tools. Enhancing prediction accuracy before a second intervention is key to optimizing postoperative care and improving outcomes. We assessed which clinicopathological factor or molecular biomarkers could be optimally constructing prediction models to assist the treatment after resections, and prepare to second surgery.

To develop predictive models, we first evaluated CD46 and CD47 expression by immunohistochemistry using tissue microarrays from 196 patients who underwent resection for recurrent hepatoma. We then employed an artificial neural network and a classification and regression tree to optimize models that combined these biomarkers with pertinent clinical factors.

Survival analysis revealed that CD47 expression was significantly associated with both disease-free survival (DFS) and disease-specific survival (DSS) in patients experiencing a second recurrence. Significant differences in pathologic type, vein tumor thrombosis, Milan criteria, and CD47 expression were observed between patients with and without second recurrence. Likewise, these same factors-pathologic type, satellite lesions, Milan criteria, and CD47-also distinguished patients who died from a second recurrence from those with other outcomes (P < 0.05). By integrating these predictors, we developed classification models that achieved an accuracy of 85.0% for predicting second postoperative recurrence and 80.0% for predicting postoperative disease-specific prognosis.

We present multi-factor models that predict second recurrence and prognosis in recurrent HCC. This prognostic tool can inform personalized clinical management after resection and improve decision-making prior to transplantation.