Prediction of hospital mortality in ruptured hepatocellular carcinoma

Abstract

Aim: Prognosis of ruptured hepatocellular carcinoma (HCC) was poor. There is no prognostic score to guide treatment strategy in these patients. Methods: A retrospective analysis of all ruptured HCC patients in a university center from 2003 to 2016 was performed. A predictive score for hospital mortality was calculated using clinical and tumor parameters. Significant predictors for hospital mortality was analyzed using Cox regression and predictive performance of the score was assessed using receiver operating characteristics (ROC) curve. The scoring system was subsequently validated in a prospective observational study from 2017 to 2018. Results: The training set was based on 315 ruptured HCC patients with an overall hospital mortality of 43.5%. Transarterial embolization was the mainstay of treatment (46%) and 16.2% received surgical treatment. On multivariate analysis, MELD>14 [HR7.322 (4.128–12.988), P < 0.001], bilobar HCC [HR2.555 (1.383–4.719), P = 0.003], known history of HCC [HR3.129 (1.771–5.529), P < 0.001] and complication of variceal bleeding [HR53.414 (5.919–481.999), P < 0.001] were significant predictors for hospital mortality. The ROC curve of training set had an area under curve of 0.834, P < 0.001. In the prospective validation set, 41 patients were included and overall hospital mortality was similar (46.3%). The median MELD was 13.33(6–29), median size of HCC was 11.5(2–23) and most had multifocal disease. In the ROC curve of validated cohort, area under curve was 0.744(0.59–0.898), P = 0.008. When the score was >5, patients had 100% hospital mortality. Conclusion: We have developed a reliable scoring system to predict hospital mortality after HCC rupture and treatment could be futile if the score was >5.