To assess the effect of sampling error on renal allograft biopsies, we determined the concordance of diagnoses between 2 biopsy samples from the same renal allograft and the frequency with which 1 biopsy sample would underdiagnose or lead to the undertreatment of acute rejection. Two core samples from the same allograft biopsy procedure were labeled as core A and core B and presented to both unblinded and blinded pathologists, and each pathologist independently assigned an acute and a chronic rejection grade. A set of clinical data with pertinent prebiopsy information was combined with either the core A or core B histopathological diagnosis and presented to 3 transplant nephrologists who made treatment recommendations for each combination. Two cores were obtained in 79 allograft biopsies. Core pairs differed by > or = 1 grade of acute rejection in 30% and 50% of cases for unblinded and blinded pathologist readings, respectively. Moderate or severe acute rejection would have been missed with a 1 core in 9.5% of cases, increasing to 25.6% if only biopsy pairs containing at least 1 reading of moderate or severe acute rejection are included. Therapy would have failed to be increased with a single core in 7.5% of cases, increasing to 10.5% if only pairs containing at least one recommendation of an increase in therapy are included. The use of 2 cores of renal allograft tissue provides better diagnostic information and thereby leads to appropriate increases in antirejection therapy without increasing the complication rate of the procedure.