The prediction of abstinence from ethanol may be crucial to the optimal selection of liver transplantation candidates with alcoholism. Of 84 consecutive end-stage alcoholic patients who underwent transplantation (1986-1994) at our institution, we analyzed 63 long-surviving recipients for pretransplantation variables to predict posttransplantation abstinence (follow-up: 49.3 +/- 21 mo). Thirty-three pretransplantation variables were reviewed from our transplantation data base and supplemented and confirmed with interviews with recipients. The psycho-social inclusion criteria included the following: patient recognition of alcoholism, a domicile, an occupation, and at least one close personal relationship. The incidence of abstinence from ethanol was (50/63) 79%. A logistic regression of the 33 variables in conjunction with our above inclusion criteria accurately predicted abstinence (90% accuracy, chi2 model, P < .00001) based on the absence of previous history of any illicit drug use (Drug Use: yes = 1/no = 0), the presence of an active, personal life insurance policy (Life Ins: yes = 1/no = 0), number of alcoholic sisters (ETOH-SIS), and the length of pretransplantation abstinence (PRE-TRANS-ABS, mos): Prob. of abstinence = 1/1 + e(-F), F = -0.33 +/- 0.89 (DRUG USE) -1.02 (LIFE INS) -1.68 (ETOH-SIS) +0.24 (PRE-TRANS-ABS). In contrast, receiver-operating characteristic curve analysis found that 7 and 9 months of pretransplantation abstinence were the best cut-off points in predicting subsequent abstinence, but poor utility was noted at these points with this specific value alone (sensitivity 61-84%, specificity 64-68%). A separate analysis of high-risk patients with poly-drug use (n = 15, alcohol recidivism 8/15, 53%) and the remaining low-risk group of purely alcohol dependent patients (n = 48, alcohol recidivism 5/48, 10%) found no combination of variables was predictive of abstinence in either group. The length of pretransplantation abstinence is a relatively poor predictor of posttransplantation abstinence. Variables of comorbid substance use, social function, and possibly family history are more predictive in conjunction with our standard criteria and might be useful as tools in evaluating liver transplantation candidates whose primary diagnosis is alcohol-induced cirrhosis.