BACKGROUND: There is a shortage of randomized controlled trials (RCTs) on which to base guidelines in liver surgery. The feasibility of conducting an adequately powered RCT in liver surgery using the dichotomous endpoints surgery-related mortality or morbidity was examined. METHODS: Articles published between January 2002 and November 2007 with mortality or morbidity after liver surgery as primary endpoint were retrieved. Sample size calculations for a RCT aiming to show a relative reduction of these endpoints by 33, 50 or 66 per cent were performed. RESULTS: The mean operative mortality rate was 1.0 per cent and the total morbidity rate 28.9 per cent; mean rates of bile leakage and postresectional liver failure were 4.4 and 2.6 per cent respectively. The smallest numbers of patients needed in each arm of a RCT aiming to show a 33 per cent relative reduction were 15 614 for operative mortality, 412 for total morbidity, 3446 for bile leakage and 5924 for postresectional liver failure. CONCLUSION: The feasibility of conducting an adequately powered RCT in liver surgery using outcomes such as mortality or specific complications seems low. Conclusions of underpowered RCTs should be interpreted with caution. A liver surgery-specific composite endpoint may be a useful and clinically relevant solution to pursue.