We have used a formal transplant protocol to select patients with alcoholic liver disease (ALD) for transplantation. We retrospectively analyzed all the patients with ALD who were referred specifically for transplantation to our Liver Unit between 1987 and 1994. Patients were selected for liver transplantation if they had end-stage liver disease and had remained abstinent from the time they were medically advised to stop alcohol intake. Of the 180 patients referred for transplantation, 43 (none of whom were transplanted) had case records insufficiently complete for full analysis; this may bias the analysis. Of the remaining 137 patients, 39 were transplanted and 4 were awaiting transplantation at the time of analysis. Of the patients who were not accepted for transplantation, 13 died during the assessment, 7 were considered to be unlikely to survive the procedure, 29 were found to be medically unsuitable, 16 psychologically unsuitable, 7 patients refused the offer of transplantation, and an additional 19 either showed clinical improvement or were considered too well for transplantation. Special investigations, such as brain computerized tomography (CT) scan and echocardiograph, changed the clinical decision to transplant in only a small number of cases (4% and 5%, respectively). Nine of the transplanted patients died and the remaining were followed up for a median of 25 (range, 7-63) months. One year actuarial survival for the transplanted patients was 79%, for those considered too sick was 0%, for medically unsuitable patients was 44%, for psychologically unsuitable patients was 65% and for those considered too well was 94%. Only 5 of the transplanted patients (13%) reverted to drinking. The observed actuarial survival of nontransplanted patients was compared with the expected survival calculated by 'the Beclere model.' The observed actuarial survival in the nontransplanted groups was much better than anticipated from the Beclere model, which therefore, is not applicable to our patients. The proportional hazards regression analysis of our nontransplanted patients identified serum bilirubin, serum albumin, blood urea, ascites, and spontaneous bacterial peritonitis as factors significantly predictive of their probability of survival. Using a model based on these parameters, the expected survival of our transplanted patients was calculated. Although we applied the model to a different population, the observed actuarial survival in the transplanted patients was found to be better than their expected survival (P < or = .001). Our protocol was useful in selecting suitable patients with ALD for liver transplantation, which resulted in significant survival advantage with low recidivism rate.