In 2005 and 2007, the Banff classification was modified to reduce the threshold for the borderline changes suspicious for acute T-cell-mediated rejection category (BL) so that tubulitis (t) scores of 1, 2 or 3 with interstitial inflammation (i) scores of 0 or 1, or interstitial inflammation (i2, i3) with mild tubulitis (t1) became the diagnostic criteria.1,2 Since then, the threshold for BL has remained a subject of controversy. A recent survey from Becker et al3 revealed that 67% of nephropathologists use the 1997 definition requiring i1, whereas 37% admitted exaggerating the i score in the presence of t score above 0. Notably, 48% of the nephropathologists who responded to the survey declared that they were unsatisfied with the current definition, with the 2 main criticisms being the low perceived clinical utility of the BL category and the difference between “no rejection” and BL being too low.3 Thus, a general disagreement remains. A new Banff working group has been created in 2013 to address issues related to T cell–mediated rejection (TCMR), including the borderline category.4
One important limitation of the BL category is that there is no clear evidence of a clinical-pathological correlation between the diagnosis and the patient outcomes, regardless of the treatment.5-10 In a study using the Banff 2009 classification, graft loss was observed in 2.4%, 6.7%, 20.4%, and 26.4% of patients with no rejection, BL, TCMR grade I, and TCMR grade II/III, respectively, at 8 years posttransplant.8 Another single-center retrospective study of 322 patients with indication biopsy between 1990 and 2014 showed a rate of 6.2% BL and 9.2% TCMR. In the latter study, biopsies were reviewed according to the Banff 2013 classification.9 Graft loss was not different in BL and TCMR (hazard ratio of 1.2 and 1.0, respectively) compared to normal biopsies at a median follow-up of 21 months postbiopsy. A recent meta-analysis of 5 studies that used various Banff classification versions to examine the outcome of indication biopsies found no functional recovery in 4% of BL, 0% to 25% of grade I TCMR, 11% to 20% of grade IIA, and 38% of grade IIB biopsies.10 To date, no specific data have examined the improvement in clinical utility of the new over the previous criteria for BL.
The aim of this study was to compare the outcomes of patients meeting the BLt1i0 versus the BL≥t1i1 threshold for BL. We found that this outcome was similar between patients with BLt1i0 and patients with no rejection, whereas patients with BL≥t1i1 had an outcome intermediary between those with no rejection and those with TCMR.
MATERIALS AND METHODS
Study Design and Population
This was a single-center, observational retrospective cohort study of kidney graft biopsies performed between January 2006 and January 2016 (Figure 1). We included all of the indication biopsies with a t score of 1 or greater and used the normal biopsies as controls. A total of 370 biopsies met these criteria. To study a “pure” cohort of patients with BL and compare them with patients with TCMR and nonrejectors, we first excluded the following (n = 83): cases with a concurrent secondary diagnosis of either glomerulonephritis, infectious pyelonephritis, or BK nephropathy. We then excluded cases with a component of antibody-mediated rejection (n = 85), which was defined as the presence of donor-specific antibody (DSA) with any transplant glomerulopathy, glomerulitis, peritubular capillaritis, or C4d-positive score except isolated g1 (9 cases), ptc1 (2 cases), or C4d1 by immunofluorescence cases that were negative by immunohistochemistry (3 cases). Among these 14 cases, 3 were normal biopsies, 7 were BLt1i0 and 4 were BL≥t1i1. Finally, we excluded repeat biopsies (n = 30). Importantly, all of the 42 cases classified as normal had a t0i0 score. The follow-up was conducted until January 15, 2017. No patient was lost to follow-up. Antirejection treatment was made of methylprednisolone for 3 days followed by prednisone tapering for all patients. Thymoglobulin was given to some patients with TCMR according to their transplant nephrologist assessment.
We used an additional independent validation cohort of 56 kidney graft biopsies from a different country. The study was approved by the institutional ethics committee of both centers. The reported clinical and research activities are consistent with the Principles of the Declaration of Istanbul.
The biopsies were assessed and graded according to the Banff '97 criteria and those subsequently updated.1,2,4,11 For the first cohort, all the biopsies were routinely presented at the weekly transplant clinical-pathological case reviews, during which challenging cases were reviewed by the 2 kidney pathologists (E.L. and J.R.) to ensure consistency in the evaluation. For the validation cohort, 56 samples with the diagnosis of BL were retrieved from the archives of Columbia University Medical Center. They were confirmed and reclassified jointly at multihead scope by 2 renal pathologists (I.B. and D.S.).
Definition of Outcomes
The primary outcome was a combination of death-censored graft failure or doubling of the serum creatinine level, defined as a persistent doubling from the level recorded on the day of the biopsy. In sensitivity analyses, graft loss alone was used as the outcome.
Clinical and pathological characteristics between the groups were compared using an analysis of variance, χ2 test, or Kruskal-Wallis test, as appropriate. The composite outcome was assessed using the Kaplan-Meier method with the log-rank test and followed by Cox models. Violations of the proportional hazards assumption were examined by plots of the logarithm of the negative logarithm of the estimated survivor function versus log time and by Schoenfled residuals. Statistical analyses were performed using Stata 11.0 (StataCorp, College Station, TX). All of the tests were 2-tailed, and a P value less than 0.05 was considered significant.
Proportion and Characteristics of Patients Meeting the BLt1i0 Versus BL≥t1i1 Classification in the Initial Cohort
One hundred seventy-two biopsies met the study criteria, including the nonrejectors and patients with TCMR. Among these cases, 42 (24%) were normal biopsies (t0i0), 111 (65%) were BL, and 19 (11%) were TCMR (Table 1). Specifically, among the BL cases, 75 were BLt1i0 and 36 were BL≥t1i1. Within this last group, most patients had a score of t1i1 (22 patients), followed by t2i0 or t2i1 (6 patients), t1i2 (4 patients), and t3i0 or t3i1 (4 patients). Patients had an average age of 42 years and were more likely to be male. In these patients with indication biopsies, serum creatinine levels were nonsignificantly lower in normal biopsies and the BLt1i0 group than in the 2 other groups (P = 0.11). Notably, the time posttransplant was higher in BLt1i0 cases than in the 3 other groups. The proportion of patients who received acute antirejection treatment was 0%, 16%, 56%, and 100% in normal, BLt1i0, BL≥t1i1, and TCMR biopsies, respectively (P<0.01). Acute TCMR scores (i, t, and intimal arteritis [v]) differed by study design, and the interstitial fibrosis (ci) and arteriolar hyalinosis scores were significantly different between groups.
Composite Endpoint in BL Versus Nonrejection and TCMR
We first assessed whether the outcome of patients with BL was similar to that of patients with no rejection or those with TCMR. At 5 years, the occurrence of the composite endpoint was 5%, 8%, and 21% for nonrejectors, BL, and TCMR patients, respectively. The survival curves are shown in Figure 2.
Composite Endpoint in BLt1i0 Versus BL≥t1i1, Nonrejection and TCMR
We then stratified the BL group into BLt1i0 and BL≥t1i1 and examined whether these subgroups shared a similar prognosis. The Kaplan-Meier plot shown in Figure 3 depicts the relationship between each of these histological diagnoses and the composite endpoint. At 5 years, the proportion of events was 5% in each of the BLt1i0 and nonrejector groups compared with 14% and 21% for the BL≥t1i1 and TCMR groups, respectively.
We further assessed the difference between BLt1i0 and BL≥t1i1 patients in an external cohort of 56 BL patients from a different center. Characteristics of this cohort are detailed in (Supplemental Table S1,http://links.lww.com/TP/B589). In this cohort, 13 (23%) patients reached the endpoint, at a median time of 12 (4-21) months postbiopsy. The proportion of patients who reached the composite endpoint was significantly lower in the BLt1i0 than the BL≥t1i1 patients (8 vs 36%, 2/25 vs 11/31, P = 0.02). As displayed by the Kaplan-Meier plot in (Supplemental Figure S1,http://links.lww.com/TP/B589), the survival was significantly lower in the BL≥t1i1 group (P = 0.02).
Multivariable Analysis of the Composite Endpoint by Biopsy Result
We conducted multivariable survival analyses on the combined, multicenter cohort of patients, totaling 228 patients and 35 events. Adjusted and unadjusted models for the composite outcome are displayed in Tables 2A and 2B. First, the models demonstrate no signal for a worse outcome in BLt1i0 compared with nonrejectors (Table 2A). In contrast, the hazard ratio (HR) for the BL≥t1i1 and TCMR, albeit nonsignificant because of the low number of events, suggests that these 2 groups behave differently than nonrejectors. Next, we wanted to assess whether there was a difference between BLt1i0 and BL≥t1i1. The models showing this pairwise comparison consistently indicate that BLt1i0 is associated with a better outcome than BL≥t1i1 (HR, 3.8; 95% confidence interval [CI], 1.3-11.5; P = 0.02, adjusted model 2).
To assess the robustness of the findings, we first repeated the analysis using death-censored graft loss only as the endpoint. Twenty-four events occurred during the study period. The magnitude of the HRs displayed in Supplemental Tables S2A and S2B (http://links.lww.com/TP/B589) are consistent with those presented in Tables 2A and 2B. Next, we repeated the analysis of the composite endpoint using the time posttransplant instead of the time postbiopsy as the time of reference. As shown in Supplemental Figure S2 (http://links.lww.com/TP/B589) this analysis over 15 years posttransplant suggest that the behavior of BLt1i0 is similar to nonrejectors, but different than BL≥t1i1 and TCMR (P < 0.001).
In this study, we found that as a single group, patients meeting the t1i0 threshold for the BL category did not differ from nonrejectors with regard to the occurrence of the composite endpoint of death-censored graft loss or doubling of the serum creatinine level at 5 years postbiopsy. However, the outcome of BL≥t1i1 patients was significantly worse than that of patients with BLt1i0, which suggests that BL≥t1i1 may be the relevant score in terms of prognosis. These results were consistent in an external validation cohort and robust to adjustment for the antirejection treatment received.
From an immunological standpoint, it is interesting to question why the addition of interstitial inflammation seems to modify the prognosis of mild tubulitis. It is now clear that inflammation is a key factor related to graft outcome in the presence of interstitial fibrosis and tubular atrophy.12-14 As suggested previously by Mengel et al,14 tubulitis is a sign of deterioration of epithelial cells and shows that these cells have lost their ability to exclude invading immune cells. In contrast, interstitial infiltration of the renal parenchyma by leucocytes is indirect evidence of the recent chemoattraction of these cells after chemokine secretion by the graft tissue. Therefore, it is plausible that interstitial infiltration better reflects ongoing immune injury to the graft than tubulitis. de Freitas et al,7 after their analysis by microarray of a cohort of 40 BL patients, concluded that the histological assessment was more accurate when it was based primarily on the i score rather than the t score.
One of the strengths of the current study is that it specifically compares BLt1i0 with BL≥t1i1 in indication biopsies. Most of the data published thus far analyzed BL as a single category and were based on protocol only or both protocol and indication biopsies, so a direct comparison with our results was not possible. Here, we excluded protocol biopsies because of the little benefit previously shown in treating patients with subclinical rejection meeting even the TCMR level.15 Therefore, the current equipoise regarding the BL category is related to indication biopsies. Given the sample size, it was not possible to test the hypothesis that the impact of treating BL≥t1i1 cases would be different than treating BLt1i0; however, the observation that the prognosis of BL≥t1i1 was closer to that of TCMR and worse than that of BLt1i0 makes this a plausible speculation. The findings reported here add to the accumulating evidence that the heterogeneity of the current BL category precludes testing whether treating patients with this diagnosis provides benefit.3,7 Importantly, during the study period, the biopsy reports were based on the most recent Banff classification so that patients with BLt1i0 were “flagged” for the transplant nephrologists as potentially having BL. Therefore, the management of these patients and consequently their graft outcomes reflect the real-life setting of using the most recent Banff classification rigorously.
Of note, Table 1 shows that the median time posttransplantation at biopsy was significantly higher in BLt1i0 patients than in the other groups (83 months for BLt1i0 vs 38, 7, and 13 months for nonrejectors, BL ≥ t1i1, and TCMR, respectively). In all of the cases, the biopsy was triggered by a rise in serum creatinine level variation or a rise/de novo proteinuria. The observation that BLt1i0 was found predominantly on aging kidneys may indicate that variations in serum creatinine levels are used misleadingly in these grafts when a decision is made to proceed to a biopsy.
The main limitation of this study was that the number of events recorded was small. However, despite the fact that it limited the number of potential confounders included in the models, we believe that this does not invalidate the reported findings. Vittinghoff and McCulloch16 found only mild bias in binary regression models when the number of events per predictor variable was relaxed, even in the range of 2 to 4. The authors suggested that the required events per predictor variable can be particularly relaxed when the aim of modeling is to demonstrate appropriate control of confounding factors, which is the technique that we used here with progressive adjustment. Moreover, the sample size of the patients with reported BL here is larger than that in recently published cohorts.7-9 Second, we focused on the contrast between BLt1i0 and BL ≥ t1i1 because this was the most recent update made in Banff for the BL category. Ideally, a more comprehensive comparison of the BL distribution above t1i1, for instance, t1i1 versus t2-3i1 or t1i2-3, would be ideal, but was not possible using the current study population.
In summary, there appear to be notable differences between the prognosis of patients with BLt1i0 and those with BL ≥ t1i1, which suggests that these cases should not be considered in the same diagnostic category. These data support the notion that the current Banff criteria should be revisited for patients with low-grade tubulitis in the absence of interstitial inflammation.
The authors would like to thank all the patients who participated in the study.
1. Solez K, Colvin RB, Racusen LC, et al. Banff '05 Meeting Report: differential diagnosis of chronic allograft injury and elimination of chronic allograft nephropathy ('CAN'). Am J Transplant
2. Solez K, Colvin RB, Racusen LC, et al. Banff 07 classification of renal allograft pathology: updates and future directions. Am J Transplant
3. Becker JU, Chang A, Nickeleit V, et al. Banff borderline changes suspicious for acute T-cell mediated rejection: where do we stand? Am J Transplant
4. Haas M, Sis B, Racusen LC, et al. Banff 2013 meeting report: inclusion of c4d-negative antibody-mediated rejection and antibody-associated arterial lesions. Am J Transplant
5. Gloor JM, Cohen AJ, Lager DJ, et al. Subclinical rejection in tacrolimus-treated renal transplant recipients. Transplantation
6. Matoza JR, Danguilan RA, Chicano S. Impact of Banff borderline acute rejection among renal allograft recipients. Transplant Proc
7. de Freitas DG, Sellares J, Mengel M, et al. The nature of biopsies with “borderline rejection” and prospects for eliminating this category. Am J Transplant
8. Wu K, Budde K, Lu H, et al. The severity of acute cellular rejection defined by Banff classification is associated with kidney allograft outcomes. Transplantation
9. Arias-Cabrales C, Redondo-Pachon D, Perez-Saez MJ, et al. Renal graft survival according to Banff 2013 classification in indication biopsies. Nefrologia
10. Lamarche C, Cote JM, Senecal L, et al. Efficacy of acute cellular rejection treatment according to Banff score in kidney transplant recipients: a systematic review. Transplant Direct
11. Loupy A, Haas M, Solez K, et al. The Banff 2015 Kidney Meeting Report: current challenges in rejection classification and prospects for adopting molecular pathology. Am J Transplant
12. Park WD, Griffin MD, Cornell LD, et al. Fibrosis with inflammation at one year predicts transplant functional decline. J Am Soc Nephrol
13. Mannon RB, Matas AJ, Grande J, et al. Inflammation in areas of tubular atrophy in kidney allograft biopsies: a potent predictor of allograft failure. Am J Transplant
14. Mengel M, Reeve J, Bunnag S, et al. Scoring total inflammation is superior to the current Banff inflammation score in predicting outcome and the degree of molecular disturbance in renal allografts. Am J Transplant
15. Rush D, Arlen D, Boucher A, et al. Lack of benefit of early protocol biopsies in renal transplant patients receiving TAC and MMF: a randomized study. Am J Transplant
16. Vittinghoff E, McCulloch CE. Relaxing the rule of ten events per variable in logistic and Cox regression. Am J Epidemiol