Critical appraisal of the clinical and pathologic predictors of survival after resection of large hepatocellular carcinoma

Abstract

Hypothesis: A subset of patients with hepatocellular carcinoma (HCC) with a diameter of 10 cm or larger may benefit from hepatic resection.

Design: Retrospective study of a multi-institutional database.

Setting: Five major hepatobiliary centers.

Patients: We identified 300 patients who underwent hepatic resection for HCC 10 cm or larger.

Main outcome measures: Clinical and pathologic data were collected, and prognostic factors were evaluated by univariate and multivariate analyses. Patient survival was stratified according to a clinical scoring system and pathologic T classification.

Results: The perioperative mortality rate was 5%. At a median follow-up of 32 months, the median survival was 20.3 months, and the 5-year actuarial survival rate was 27%. Four clinical factors-alpha-fetoprotein of 1000 ng/mL or higher, multiple tumor nodules, the presence of major vascular invasion, and the presence of severe fibrosis-were significant predictors of poor survival (all P<.05). Patients were assigned a clinical score according to the following risk factors: 1, no factor; 2, one or two factors; or 3, three or four factors. On the basis of the clinical score, patients could be stratified into only 2 distinct prognostic groups: no factor (score of 1) vs 1 or more factors (score of 2 or 3) (P<.001). In contrast, when patients were stratified according to pathologic T classification, 3 distinct groups were identified: T1 vs T2 vs T3 and T4 combined (P<.001). Fifty-six percent of the patients with a clinical score of 2 and 20% of patients with a clinical score of 3 actually had T1 or T2 disease on pathologic examination.

Conclusions: Patients with large HCCs should be considered for liver resection as this treatment is associated with a 5-year survival rate exceeding 25%. Clinical predictors should not be used to exclude patients from surgical resection because these factors do not reliably predict outcome.