The use of broad-specificity cross-reactive groups (CREGs) at the A and B HLA loci has been proposed as a means to improve both access and outcome for renal transplantation but has not yet been studied for liver transplantation. We retrospectively analyzed our results for all adult liver transplantations performed at our center between 1989 and 1996 for which HLA typing and complete 3-year follow-up data were available. Two hundred eight transplantations were studied, with a mean follow-up of 66 +/- 2 months (range, 36 to 110 months); A and B loci were converted to CREGs by the method of Rodey. Thirteen percent of the patients with 0 CREG mismatches had hepatic artery thrombosis versus 2% for those with 1 or more mismatches (odds ratio, 6.7; 95% confidence interval, 2.6 to 17.3 by univariate analysis; odds ratio, 3.5; 95% confidence interval, 1.1 to 11.7 by logistic regression analysis). These events occurred significantly later in the 0-CREG mismatch group (21 +/- 7 v 4 +/- 2 months posttransplantation; P =.04 by Student's t-test). Graft survival rates were significantly lower for patients with 0 CREG mismatches (56% v 68% at end of study; P =.05 by Cox-Mantel test). The number of CREG mismatches had no effect on the frequency of rejection, steroid-resistant rejection, or infectious complications, including cytomegalovirus. Neither total nor individual A, B, or DR HLA matching had an effect on outcome. There appears to be little evidence that intentional CREG matching would improve outcomes for liver transplantation and, under some circumstances, could be deleterious.