In the Banff classification of kidney transplant pathology the "borderline changes" category falls short of a diagnosis of mild acute rejection, with the recommendation that no treatment is a possible clinical approach. We reviewed the clinical course of patients whose renal allograft biopsies showed "borderline changes" to determine how often these histologic findings actually represented acute rejection. Between January 1992 and June 1994, 351 biopsy specimens were obtained from 170 renal allografts and graded according to the Banff criteria. Eighty-one biopsy specimens were classified as "borderline changes" (23%). Of these, 59 had Banff scores of i1, t1, and vO; the remaining 22 had scores of i2, t1, and vO (i = interstitial infiltrate, t = tubulitis, and v = vasculitis). Medical record review showed that nearly all the "borderline" biopsies had been performed because of an elevated creatinine (Cr; 78 of 81 [96%]), with a mean increase of 1.1 +/- 0.1 mg/dL (+/- SE) over baseline. Most of the patients with "borderline changes" and elevated Cr were treated for acute rejection (61 of 78 [78%]); some with pulse steroids alone (29 of 61 [48%]), the rest with antilymphocyte antibody (32 of 61 [52%]). Among all 61 patients with "borderline" biopsies treated for rejection, 26 had a complete response (43%), 17 had a partial response (28%), and 18 had no response (30%). Interpretation of these changes in Cr, however, was confounded by intercurrent conditions in 28 of the patients. A group of 33 patients was therefore identified in whom a "borderline changes" biopsy was obtained, who were treated for rejection, and in whom all other identifiable causes of elevated Cr other than possible acute rejection had been systematically eliminated from consideration. In this group the mean Cr was 2.0 +/- 0.1 mg/dL at baseline, 3.3 +/- 0.2 mg/dL at the time of biopsy, and 2.2 +/- 0.1 mg/dL 1 month after treatment (P < 0.001 Cr at biopsy v Cr 1 month later). Among these 33 patients, 19 had a complete response (58%), 10 had a partial response (30%), and four had no response (12%). Therefore, the Cr in 88% of the patients in this group was lower 1 month after treatment for rejection than it was at the time of the biopsy. Follow-up biopsies were performed within 1 month of the "borderline" biopsy in 24 cases; these showed "borderline changes" (five of 24 [21%]), mild acute rejection (eight of 24 [33%]), or moderate to severe acute rejection (11 of 24 [46%]). We conclude that in the clinical setting of deteriorating renal graft function with mild elevation of serum Cr, the "borderline changes" biopsy frequently represents acute rejection. Antirejection treatment is therefore appropriate in the majority of cases. The reader should bear in mind that the current study is retrospective, with no control group. The risk of loosely interpreting these data is that some patients will be treated without due cause. Banff "borderline changes" should be used as part of an algorithm, but not the sole criterion, for therapeutic decision making.