The utility of the MELD score in predicting mortality following liver resection for metastasis

Background The MELD score has been demonstrated to be predictive of hepatectomy outcomes in mixed patient samples of primary and secondary liver cancers. Because MELD is a measure of hepatic dysfunction, prior conclusions may rely on the high prevalence of cirrhosis observed with primary lesions. This study aims to evaluate MELD score as a predictor of mortality and develop a risk prediction model for patients specifically undergoing hepatic metastasectomy. Methods ACS-NSQIP 2005–2013 was analyzed to select patients who had undergone liver resections for metastases. A receiver operating characteristic (ROC) analysis determined the MELD score most associated with 30-day mortality. A literature review identified variables that impact hepatectomy outcomes. Significant factors were included in a multivariable analysis (MVA). A risk calculator was derived from the final multivariable model. Results Among the 14,919 patients assessed, the mortality rate was 2.7%, and the median MELD was 7.3 (range = 34.4). A MELD of 7.24 was identified by ROC (sensitivity = 81%, specificity = 51%, c-statistic = 0.71). Of all patients above this threshold, 4.4% died at 30 days vs. 1.1% in the group ≤7.24. This faction represented 50.1% of the population but accounted for 80.3% of all deaths (p < 0.001). The MVA revealed mortality to be increased 2.6-times (OR = 2.55, 95%CI 1.69–3.84, p < 0.001). A risk calculator was successfully developed and validated. Conclusions MELD > 7.24 is an important predictor of death following hepatectomy for metastasis and may prompt a detailed assessment with the provided risk calculator. Attention to MELD in the preoperative setting will improve treatment planning and patient education prior to oncologic liver resection.