Prognostic factors and staging systems in childhood hepatoblastoma.

The objectives of this study were to investigate clinical and patho-histological characteristics of childhood hepatoblastoma on their value as prognostic factors, and to evaluate the predictive impact of different staging systems for liver tumors on 72 patients treated in the German Pediatric Liver-Tumor Study HB89. Statistical analysis was performed by comparing patients' disease-free survival with characteristics and stages. Multivariate analysis was done by the Cox proportional-hazards model, the recursive partitioning and amalgamation model (RECPAM) and the model of clustering by response (CBR). The following characteristics were significantly related with prognosis: tumor involvement of one vs. both liver lobes, multifocal disseminated vs. unifocal growth pattern in the liver, distant metastases, vascular invasion, fetal vs. embryonal differentiation, and serum alpha-fetoprotein; patients with values of < or = 100 ng/ml or > or = 1,000,000 ng/ml had a worse outcome than those with 100 to 1,000,000 ng/ml. Multivariate analysis with the 3 models revealed that tumor-growth pattern, serum alpha-fetoprotein and, in the Cox and CBR models, vascular invasion also are independent prognostic factors, permitting the allocation of hepatoblastoma patients to 1 of 2 prognostic groups for differential therapy. Post-surgical staging and the conventional TNM system for liver carcinoma had a high predictive value, in contrast to a TNM system proposed by the UICC for testing on childhood liver tumors. We therefore propose that the TNM system for liver carcinoma be applied for comparison of treatment results in hepatoblastoma.