The Banff 2009 Working Proposal for Polyomavirus Nephropathy: A Critical Evaluation of its Utility as a Determinant of Clinical Outcome

Abstract

Clinical outcome in BK virus nephropathy(BKVN) was examined in relation to clinical and histologic parameters with reference to the Banff Working Proposal 2009, which emphasizes tubular injury and viral load. 71 patients were evaluated in three eras: (a) Era-I: No BKV PCR performed(n=36), (b) Era-II: PCR performed for rising creatinine(n=24), and (c) Era III: PCR performed for routine screening(n=11). 6/71(8.4%) patients were classified as Class A, 46/71(64.8%) as Class B and 19/71(26.8%) as Class C. Banff class A never occurred in era-I. It is a heterogeneous class that includes biopsies with inflammation that have hitherto been included in Class B. Higher inflammation, but not tubular injury, nor histologic viral load correlated with worse creatinine at 3 months. On long-term follow-up, class C associated with graft loss (hazard ratio 2.45, p=0.03). Clearance of viremia was associated with better graft survival at 5 years (46.0% versus 25.0%). Viruria clearance was infrequent (15.6%). In conclusion, the clinical utility of the Banff Working Proposal 2009 derives from scoring of fibrosis and not extent of tubular injury or viral cytopathic effect. The proposal is not superior to existing schemas that include assessment of inflammation, which is a well-known prognostic marker in other renal allograft diseases.

Introduction

Polyomavirus BK (BKV) infections are a common in kidney transplant patients. Serology shows past exposure in 60–90% of patients. Viral shedding in the urine can be demonstrable by cytology, polymerase chain reaction (PCR) or electron microscopy in 10–60% of subjects. Estimates of the prevalence of BK viremia range from 10–30% (1–3).

BKV nephropathy (BKVN) develops in 1–10%, with the exact incidence dependent on the intensity of immunosuppression as well as screening techniques implemented. The first step in the treatment of persistent BK viremia or BKVN is reduction of immunosuppression. With early diagnosis using regular screening the majority of patients respond favorably, but approximately 20% of still progress to graft loss (4). The actual toll of the disease is larger since 38% of patients continue to show decline in graft function, while 24% show a major decline characterized by sustained increase in serum creatinine of at least 50% compared to that observed at the time of diagnosis (5).

The ability to predict the clinical outcome in individual patients is an important goal of transplant physicians. Clinical factors reported to be associated with worse prognosis include deceased donor, female recipient (5), high serum creatinine (6), late diagnosis (7), and plasma viral load (5). As BKVN is ultimately a biopsy diagnosis, there has also been much interest in exploring the effect of histologic variables on the course of disease. The percentage of tubular cross sections showing infection and degree of interstitial fibrosis and tubular atrophy were identified as important in an early study from the Mayo Clinic (7). A composite system to stage the disease based on viral cytopathic effect, extent of inflammation and severity of fibrosis was first proposed by investigators at the University of Maryland (UMD) (8). Variations of this schema have been published by a multidisciplinary panel of polyomavirus investigators (9) and by the American Society of Transplantation (AST) (4).

Recently, the Banff Working Group has proposed a grading system which places considerable emphasis on the extent of virus induced tubular epithelial injury as measured by necrosis, cell lysis, intra-luminal shedding, and denudation of tubular basement membranes (10, 11). This premise needs further evaluation as it is recognized that clinical acute renal failure may be associated with only subtle signs of morphologic change at biopsy. Pathology findings are also subject to sampling error, since acute injury affects only short segments of the tubule and may only be seen in the medulla. Accordingly, we applied the Banff Working Proposal 2009 to 71 biopsies with BKVN to assess its predictive value in the clinical setting.

Methods

Clinical material

71 kidney transplant patients included in this study are from three different time periods were analyzed:

1. Era I: 1993–2000 (n=36): PCR testing for BKV DNA was not available. Most patients (22/36, 61.0%) received triple maintenance immunosuppression with methylprednisolone, calcineurin inhibitor and Mycophenolate mofetil or azathioprine.

2. Era II: 2001–2005 (n=24): BKV PCR in the urine and plasma was performed at the time of a biopsy necessitated by graft dysfunction. Patients received antibody induction with thymoglobulin or Campath-1H followed by steroid-free tacrolimus monotherapy.

3. Era III: 2006–2010 (n=11): Patients were screened for urine and plasma BKV DNA monthly for 6 months and 3–6 monthly for 24 months. All patients had Alemtuzumab induction therapy, and 7/11 received tacrolimus monotherapy as maintenance immunosuppression.

Clinical parameters were retrospectively reviewed in accordance with IRB protocol # 0602155.

Short-term outcome

For each biopsy, we obtained a baseline serum creatinine (4 months prior to diagnosis), creatinine at the time of allograft biopsy, and a post-therapy creatinine 3 months thereafter. If changes in immunosuppression led to >70% reversal of the rise in serum creatinine, the case was classified as “complete response”. Graft dysfunction episodes with 30–70% reversal qualified for “partial response”, less than 30% of reversal and less than 30% increase as “non-response”, and rise in creatinine >30% as “progression”. Clearance of viremia or viruria was defined by the availability of 3 or more consecutive negative PCR tests for BKV DNA spanning an interval of at least 3 weeks.

Assessment of long-term outcome of BKVN

The median graft follow-up period of BKVN patients was 4.2 (range 0.5–17.9) years. Death censored graft loss was defined as return to dialysis or re-transplantation. Analysis of the cause of death was based on a perusal of medical records, review of biopsy findings, and examination of allograft nephrectomy specimens, when available.

Pathologic examination

Hematoxylin & Eosin, Periodic acid Schiff, and Masson’ silver trichrome stains were performed on all biopsies. Diagnosis of BKVN was based on viral cytopathic effect and confirmed by immunohistochemical staining for polyomavirus large-T antigen and/or in situ hybridization for BK virus DNA. Histologic lesions were scored using the Banff 1997 schema of renal allograft pathology. However, inflammation and tubulitis were scored over the entire biopsy including areas with fibrosis, as it is now recognized that the total inflammatory score in renal allograft biopsies has more prognostic relevance (12–15). All biopsies with BKVN were classified using (a) The Working Proposal discussed at the 10th Banff conference, (b) The UMD System, and (c) The AST schema, using criteria summarized in Table 1 and Figure 1. Histologic viral load was assessed semi-quantitatively as the percentage of tubules that stained positive for BKV using a 3 tier system (<1%, 1–10%, and >10%, as per The Banff Working Proposal 2009). Although the working proposal classifies BKVN into Stage A, B and C, we have used the terms class A, B and C in lieu thereof, since other Banff schemas use the word stage to refer to the extent of fibrosis.

Table 1.

Salient features of histologic grading systems applied to study patients.

	Banff working Proposal*	University of Maryland***	American Society of Transplantation***
Class A	Variable number of virus infected cells with NO or MINIMAL injury to tubular epithelial cells.	Variable number of virus infected cells with ANY degree of tubular injury but NO or NEGLIGIBLE INFLAMMATION	Virus infection and cytopathic effect in <25% of biopsy, with NO/NEGLIGIBLE INFLAMMATION specified to be <10% of tissue
Class B	Tubular epithelial cell necrosis or lysis with denudation of basement membrane across a length of more than 2 cells**	Variable number of virus infected cells with ANY degree of tubular injury AND SIGNIFICANT INFLAMMATION affecting less than 25% (pattern B1), 25–50% (pattern B2) or >50% (pattern B3) of the core biopsy.	Essentially similar to University of Maryland, except that B1, B2 B3 are assigned progressively increasing degrees of cytopathic effect, atrophy & fibrosis
Class C	Any degree of tubular injury with INTERSTITIAL FIBROSIS affecting >50% of cortex	Variable number of infected cells with ANY degree of tubular injury AND tubular ATROPHY/FIBROSIS affecting >50% of core biopsy.	Same as University of Maryland System

^*The Banff Working Proposal was conceived during the 10^th Banff Conference on Allograft Pathology in 2009 (10, 11). This proposal describes Stages A, B, and C to different evolutionary stages of the disease. To avoid confusion with other Banff Schemas where the term stage refers specifically to degrees of fibrosis, we suggest that the terms class A, B and C would be more appropriate. A detailed description of the Banff working proposal is available at: http://www.uncnephropathology.org/documents/BanffDraftforPolyomavirusNephropathyStaging.pdf

^**A rather conspicuous feature of the Working Proposal is that the degree of inflammation is not taken into account in the staging of disease.

^***The University of Maryland (8) and AST (4) schema are essentially identical, except that the AST schema specifically defines ‘no or negligible inflammation’ in the Maryland system as inflammation affecting <10% of the sampled core. The Maryland system proposes a straight forward sub-categorization of class B into categories B1 thru B3 based on % area affected by inflammation and atrophy. Sub-classification of class B in the AST schema assigns progressively increasing amounts of viral cytopathic effect to subclass B1, B2 and B3. However, this stipulation makes it difficult to apply the classification to many biopsies where inflammation and fibrosis are out of proportion to the degree of viral cytopathic effect.

Figure 1.

Representative findings used to classify BKVN. Class A (upper panel) depicts mild inflammation on Hematoxylin-Eosin, mild fibrosis on trichrome, and minimal tubular injury. Immunohistochemistry for LTA shows that virus infected tubular epithelial cells remain attached to the underlying basement membrane, without cell lysis (A). In class B (middle panel), mild inflammation and fibrosis are again present, but immunohistochemistry shows that more than two virus infected cells have shed into the lumen resulting in denudation of the tubular basement membrane. Finally, in class C (lowest panel) the defining feature is the presence of interstitial fibrosis affecting >50% of cortex, irrespective of the number of virus infected cells shed into the tubular lumen. A color version of this figure is available online.

Statistical analysis

JMP version 8 (SAS Institute, Cary, NC, USA) was used for all analyses, except for 3×3 and 4×2 Fisher’s exact probability tests, which were performed STATA-v11 (StataCorp LP, College-Station, TX, USA). Analysis of variance (ANOVA) and Kruskal-Wallis test were used to compare continuous data. Categorical data was analyzed by chi-square or Fisher’s exact probability tests. Adjusted standardized residuals in the range ±1.96 and ±2.56 were considered to be discriminant at confidence levels of 95% and 99% respectively (16, 17). Kaplan-Meier estimates and log-rank tests were employed to evaluate graft and patient survival. Logistic regression analysis was used to determine which risk factors independently affected clinical outcome. Age, sex, and specific variables associated with p-values <0.1 by univariate analysis were entered into multivariate analysis. Results were expressed as odds ratios with respective 95% confidence intervals. Cox’s proportional-hazard regression analysis was used to assign hazard ratios (HR) to factors predictive of death censored graft loss.

RESULTS

Demographic characteristics of the patients studied are presented in Table 2. Induction therapy with Alemtuzumab was given to all cases diagnosed during the regular screening era, and never used in era I. Tacrolimus monotherapy was commonly used in era II and III, while triple therapy and other more conventional immunosuppression regimens were typically administered in era I. Male/Female ratio, age at transplantation, proportions of cadaveric versus live donors, degree of HLA mismatch count and etiology of end-stage renal disease were not statistically different amongst the study groups.

Table 2.

Demographic and clinical characteristics of 71 patients with BKVN classified by three eras

	Era-I (‘93-‘00)	Era-II (‘01-‘05)	Era-III (‘06-‘10)	p value
Number of cases	36	24	11
Male/Female	23/13	16/8	11/0	0.06
Age at transplant	48±16	48±13	48±21	1.0
Deceased donor	32 (88.9%)	16 (66.7%)	8 (72.7%)	0.1
First graft	28 (77.8%)	22 (91.7%)	9 (81.8%)	0.4
HLA mismatch	3.8±1.4	4.4±1.7	4.3±1.1	0.3
Etiology of ESRD
Glomerulonephritis	10 (27.8%)	3 (12.6%)	2 (18.2%)	0.7
Hypertension	6 (16.6%)	5 (20.8%)	1 (9.1%)
Diabetes	10 (27.8%)	11 (45.8%)	5 (45.4%)
Others	10 (27.8%)	5 (20.8%)	3 (27.3%)
Antibody induction*
Thymoglobulin	0 (0%)	1 (4.3%)	0 (0%)	<0.0001
Alemtuzumab	0 (0%)	12 (52.2%)	10 (100%)
No antibody	36 (100%)	10 (43.5%)	0 (0%)
Maintenance immunosuppression
Triple immunosuppression**	22 (61.0%)	9 (37.5%)	1 (9.1%)	<0.0001
Tac + PSL	10 (27.8%)	2 (8.3%)	0 (0.0%)
Tac + MMF	1 (2.8%)	3 (12.5%)	3 (27.3%)
Tac monotherapy	1 (2.8%)	9 (37.5%)	7 (63.6%)
Time post transplantation (d)	420±265	676±752	354±351	0.09
Best serum creatinine (mg/dl)	1.4±0.5	1.4±0.7	1.3±0.3	0.8
Baseline serum creatinine (mg/dl)	1.7±0.6	1.7±0.7	1.3±0.4	0.2
Serum creatinine at diagnosis (mg/dl)	2.6±1.0	2.8±1.0	2.4±1.0	0.5
History of acute rejection before BKVN diagnosis	23 (63.9%)	13 (54.2%)	2 (22.2%)	0.03
BK viral load in urine (copies/ml)***	---	1.4E+09 (3.4E+08 – 1.2E+10)	8.4E+08 (8.4E+07 – 1.7E+09)	0.3
BK viral load in plasma (copies/ml)***	---	1.2E+05 (3.4E+03 – 5.1E+05)	1.9E+04 (1.0E+04 – 2.8E+05)	0.8

ESRD; end-stage renal disease, Tac; tacrolimus, MMF; mycophenolate mofetil, AZA; azathiopurine. PSL; prednisolone

^*Induction therapies are not available in 1 case in era II and 1 case in era III who received transplantation in other institutes

^**Tac + MMF + PSL or Tac + AZA + PSL or Tac + sirolimus + PSL

^***Data was expressed as median and inter-quartile range.

The diagnosis was made by allograft biopsy performed for a rise in the serum creatinine. Tubular epithelium showed viral cytopathic effect and BKV DNA was documented by immunohistochemical staining for SV40 large-T antigen and/or in-situ hybridization. Plasma and urinary PCR for BKV were not available in era I. On the day of diagnosis of BKVN, patients with Era II were characterized by high median urinary load (1.4E+09 copies/ml, interquartile range (IQR) 3.4E+08–1.2E+10, percentile) and plasma load (1.2E+05 copies/ml, IQR 3.4E+08–1.2E+10, n= 24). However, these values were not statistically different compared to era III, wherein the median urinary load was 8.4E+08 copies/ml (IQR 8.4E+07–1.7E+09) and plasma load 1.9E+04 copies/ml (IQR 1.0E+04–2.85+05, n=11).

The baseline serum creatinine, defined as the lowest value in a 4 month interval prior to diagnosis (1.7±0.6mg/dl) and the serum creatinine at the time of diagnosis (2.6±1.0mg/dl) in era I was comparable to that seen in era II (baseline 1.7±0.7mg/dl, time of diagnosis 2.8±1.0mg/dl), and era III (baseline 1.3±0.4mg/dl; time of diagnosis 2.4±1.0mg/dl). These relatively high serum creatinine values reflect the late time of onset of BKVN, and the occurrence of multiple rejection episodes prior to diagnosis (Table 2). Assignment of Banff histopathologic class to the biopsies was greatly facilitated by an algorithm in which initial examination of immunohistochemistry or in-situ hybridization stained slides was performed at 10× magnification. Class B could be provisionally diagnosed by demonstrating three or more detached cells in the tubular lumen, with denudation of the underlying tubular basement membranes. Distinction from class C required a review of the trichrome stain to confirm that the degree of fibrosis did not affect >50% of tissue. In cases with one or two detached cells in the tubules, all slides had to be carefully evaluated for epithelial shedding and denudation of tubular basement membranes for staging purposes. Amongst class B biopsies, 6 specimens showed less than 2 detached epithelial cells, but denudation of the tubular basement membranes affected a greater part of the tubular circumference.

Using the Banff Working Proposal 2009, the relative proportion of BKVN class was different in eras I-III (p=0.04, Figure 2). Class A disease was never diagnosed in era I (0/36 cases, p<0.01, adjusted standardized residual −2.59), and was equally prevalent in era II (4/24, 16.7%), and era III (2/11, 18.1%). The occurrence of class A lowered the relative frequency of class B in eras II and III, but the differences were not statistically significant. The class of BKVN at initial diagnosis did not show any obvious relationship with the time post-transplant (596±454, 422±437, and 645±648 days for class A, B and C, respectively), serum creatinine at diagnosis (2.5±0.9, 2.5±0.9, and 3.0±1.1mg/dl respectively), or the magnitude of increase in serum creatinine from baseline (0.7±0.1, 1.0±0.6, and 1.2±0.7mg/dl respectively). Application of the UMD/AST Classification to the same cases resulted in only one biopsy with class A disease in era I. This is because biopsies without interstitial inflammation are uncommon in our practice. All Banff class A biopsies were reassigned to UMD/AST class B1, B2 or B3. Assignment of biopsies to class C did not differ between the three schemas.

Figure 2.

Frequency distribution of the classes of BKVN in three different eras.

Banff scores for individual pathologic lesions in the biopsies examined are shown in Table 3. The distribution of interstitial inflammation scores was significantly different across class A, B, and C (p=0.03). In particular, grade 3 inflammation was higher in Banff class C disease (13/19, 68.4%, p<0.01, adjusted standardized residual 3.36) than Banff class A (1/6, 16.7%) and Banff class B (12/46, 26.1%). Grade 1 inflammation was marginally higher in Banff class A disease (2/6, 33.3%) compared to Banff class B (10/46, 21.7%) and Banff class C (2/19, 10.5%) diseases. It is notable that more than one half of biopsies with putatively early class A disease in the Banff Working Proposal 2009 showed moderate (grade t2) or severe (grade t3) inflammation. In contrast, the UMD/AST Classification requires the inflammation score in class A to be zero. Therefore, all Banff class A biopsies in this study were designated as class B1, B2, B3 in the UMD/AST Classification (Table 4). Scores for tubulitis generally paralleled the degree of inflammation. Severe interstitial fibrosis (grade ci3) was by definition confined to class C in both the Banff Working Proposal 2009 and UMD/AST Classification (Tables 3, 4). Tubular atrophy was generally proportionate to the degree of interstitial fibrosis, and all ct3 cases fell into the category of class C disease. Arteriosclerosis did not differ by class of BKVN.

Table 3.

Distribution of Banff i, t, ci, ct and cv scores in BKVN biopsies subclassified by the Banff Working Proposal 2009

Class	i0	i1	i2	i3	t0	t1	t2	t3
A	0 (0.0%)	2 (33.3%)	3 (50.0%)	1 (16.7%)	0 (0.0%)	2 (33.3%)	2 (33.3%)	2 (33.3%)
B	1 (2.2%)	10 (21.7%)	23 (50.0%)	12 (26.1%)	2 (4.4%)	8 (17.4%)	8 (17.4%)	28 (60.8%)
C	0 (0.0%)	2 (10.5%)	4 (21.1%)	13 (68.4%)	0 (0.0%)	2 (10.5%)	1 (5.3%)	16 (84.2%)
Class	ci0	ci1	ci2	ci3	ct0	ct1	ct2	ct3
A	0 (0.0%)	5 (83.3%)	1 (16.7%)	0 (0.0%)	0 (0.0%)	5 (83.3%)	1 (16.7%)	0 (0.0%)
B	7 (15.2%)	25 (54.4%)	14 (30.4%)	0 (0.0%)	6 (13.0%)	25 (54.4%)	15 (32.6%)	0 (0.0%)
C	0 (0.0%)	0 (0.0%)	0 (0.0%)	19 (100%)	0 (0.0%)	0 (0.0%)	0 (0%)	19 (100.0%)
Class	cv0*	cv1	cv2	cv3
A	3 (50.0%)	2 (33.3%)	1 (16.7%)	0 (0.0%)
B	14 (32.6%)	19 (44.2%)	10 (23.2%)	0 (0.0%)
C	4 (28.6%)	6 (42.9%)	3 (21.4%)	1 (7.1%)

^*In 3 class B and 5 class C biopsies “cv“ scores could not be determined as no interlobular artery was available for evaluation.

Table 4.

Distribution of Banff i, t, ci, ct and cv scores in class A and class B biopsies subclassified using the UMD/AST Classification^*

Class	i0	i1	i2	i3	t0	t1	t2	t3
A	1 (100%)	0 (0%)	0 (0%)	0 (0%)	1 (100%)	0 (0%)	0 (0%)	0 (0%)
B1	0 (0%)	12 (100%)	0 (0%)	0 (0%)	1 (8.3%)	7 (58.4%)	4 (33.3%)	0 (0%)
B2	0 (0%)	0 (0%)	26 (100%)	0 (0%)	0 (0%)	3 (11.5%)	6 (23.1%)	17 (65.4%)
B3	0 (0%)	0 (0%)	0 (0%)	13 (100%)	0 (0%)	0 (0%)	0 (0%)	13 (100%)
Class	ci0	ci1	ci2	ci3	ct0	ct1	ct2	ct3
A	1 (100%)	0 (0%)	0 (0%)	0 (0%)	1 (100%)	0 (0%)	0 (0%)	0 (0%)
B1	1 (8.3%)	10 (83.4%)	1 (8.3%)	0 (0%)	1 (8.3%)	10 (83.4%)	1 (8.3%)	0 (0%)
B2	4 (15.4%)	12 (46.2%)	10 (38.4%)	0 (0%)	3 (11.5%)	12 (46.2%)	11 (42.3%)	0 (0%)
B3	1 (7.7%)	8 (61.5%)	4 (30.8%)	0 (0%)	1 (7.7%)	8 (61.5%)	4 (30.8%)	0 (0%)
Stage	cv0**	cv1	cv2	cv3
A	1 (100%)	0 (0%)	0 (0%)	0 (0%)
B1	4 (36.4%)	4 (36.4%)	3 (27.2%)	0 (0%)
B2	6 (24.0%)	14 (56.0%)	5 (20.0%)	0 (0%)
B3	6 (50.0%)	3 (25.0%)	3 (25.0%)	0 (0%)

^*Class C assignments were identical for the Banff Working Proposal, UMD and AST classifications. Therefore, histologic scores for class C that are provided in Table 3 are not repeated here. Class A and B assignments were identical for the UMD and AST Classifications. However, with respect to subcategories of class B, discrepancies between extent of viral cytopathic effect and degree of inflammation made the AST classification difficulty to use. Therefore, subclass assignment into B1, B2 and B3 categories was done using the UMD system.

^**In one class B1, one class B2, one class B3, and five class C biopsies “cv“ scores could not be determined as no interlobular artery was available for evaluation.

Short-term outcome based on serum creatinine measured 3 months post-diagnosis of BKVN differed in the three eras (Figure 3A, p=0.01). Specifically, the percentage of patients showing progressive rise of serum creatinine was higher in era I compared to eras II and III (23/36, 63.9%, versus 8/24, 33.3% versus and 3/11, 27.3%). The sum of cases showing either partial or complete responses tended to be higher in Banff class A (33.3%+16.7%=50.0%) compared to Banff class B (23.9%) and C (21.1%), but the difference did not reach statistical significance (Figure 3B). Severe inflammation (grade i3) was commoner in patients who showed progression of serum creatinine (18/34, 52.9% versus 8/37, 21.6%, p=0.01, Figure 4A). In contrast severe tubulitis (grade t3) had no discriminating power in this setting (Figure 4B). Many biopsies with Banff class A disease showed moderate (grade t2) or severe (grade t3) inflammation. This may account for the increased tendency to graft loss even in this group of patients (OR=1.47 vide infra).

Figure 3.

Therapeutic response to changes in immunosuppression stratified by era of diagnosis (A) or class of BKVN (B and C).

Figure 4.

Distribution of inflammation (A) and tubulitis (B) scores in biopsies stratified by progression of disease or lack thereof.

Logistic regression analyses of different clinical parameters on disease progression at 3 months are shown in Table 5. On univariate analysis, triple immunosuppression maintenance therapy and severe interstitial inflammation (grade i3) on index biopsy were significantly associated with worse allograft function at 3 months. Conversely, antibody induction and PCR screening (performed in eras II and III) were associated with better preservation of graft function. Urinary and plasma BKV load at the time of biopsy, BKVN class, and histologic viral load (expressed as percentage of BKV positive tubules) did not influence outcome. On multivariate logistic regression analysis only severe interstitial inflammation (OR=3.92; 95%CI 1.28–12.80) was as an independent predictor of progressive rise in serum creatinine beyond 3 months (p=0.02).

Table 5.

Univariate and multivariate logistic regression analysis^* of factors associated with worse graft function 3 months after the diagnosis.

Models	Univariate analysis		Multivariate analysis
	OR (95% CI)	p value	OR (95% CI)	p value
Clinical parameters
Recipient age (per year)	0.99 (0.96–1.02)	0.7	1.00 (0.96–1.04)	0.9
Recipient sex (male)	0.59 (0.21–1.64)	0.3	0.55 (0.16–1.89)	0.3
Recipient race (White)	0.63 (0.19–2.04)	0.4	---
Donor source (Deceased donor)	2.15 (0.67–7.65)	0.2	---
Donor age (per year)	0.99 (0.97–1.02)	0.7
Donor sex (male)	1.66 (0.65–4.31)	0.3	---
Cold ischemia time	2.22 (0.40–13.33)	0.4	---
HLA mismatch (per mismatch)	1.05 (0.76–1.47)	0.8
Antibody induction	0.20 (0.06–0.58)	0.006	0.69 (0.10–4.85)	0.7
Maintenance immunosuppression	3.08 (1.17–8.46)	0.03	1.70 (0.47–6.12)	0.4
PCR testing at biopsy (Eras II, III)	0.25 (0.09–0.68)	0.007	0.33 (0.06–1.54)	0.2
PCR screening (Era III)	0.35 (0.07–1.34)	0.1	---
Previous acute rejection	0.96 (0.37–2.44)	0.9	---
Baseline serum creatinine	1.16 (0.54–2.52)	0.7	---
Creatinine rise at diagnosis	0.90 (0.39–1.96)	0.8
Log plasma BKV load at diagnosis	1.47 (0.02–67.37)	0.8	---
Log urine BKV load at diagnosis	2.96 (0.12–71.58)	0.5	---
Pathological parameters
Banff Working Proposal 2009
ClassA	1.19 (0.20–7.03)	0.9	---
Class B	1 (Reference)
Class C	1.32 (0.45–3.92)
UMD/AST Classification		0.4
ClassA	NA**		---
Class B	1 (Reference)
Class C	1.35 (0.47–3.95)
BKV positive tubules
<1%	1 (Reference)	0.4	---
1–10%	2.69 (0.85–9.44)
>10%	1.50 (0.79–2.95)
Banff score
i score (i3)	4.08 (1.49–11.97)	0.008	3.82 (1.23–12.82)	0.02
t score (t3)	1.27 (0.48–3.44)	0.6	---
ci score (ci3)	1.30 (0.45–3.78)	0.6	---
ct score (ct3)	1.30 (0.45–3.78)	0.6	---
cv score (cv2 and 3)	0.76 (0.22–2.44)	0.6	---

^*Unit odds ratios (OR) and confidence interval (CI) were calculated for continuous variables. Categorical data was treated as binary variables. Maintenance immunosuppression was analyzed as presence or absence of a triple immunosuppressive regimen (defined as Tacrolimus + Azathioprine + prednisone, or Tacrolimus + mycophenolate mofetil + prednisone, or Tacrolimus + Sirolimus + prednisone).

^**Odds ratio could not be calculated as there is a single observation in this category.

Long-term graft outcome evaluated by death censored graft survival was the worst in class C disease (Figure 5). This was true irrespective of which histologic classification was applied. The UMD/AST analysis had insufficient evaluable data in class A, since there was only 1 patient who had graft loss within 2 years of transplantation. Patients with UMD/AST class B disease fared better than class C consistent with a lower score of interstitial fibrosis (Figure 5B). Within UMD/AST class B, class B3 behaved worse than B2 as one might expect, but the patients in class B1 did not have the best outcome as one might have expected (Figure 5C). The reason for the latter observation is not clear, but may pertain to small sample size and unaccounted confounding factors. For example, this group had the highest proportion of cases with diabetes mellitus (58.3%) and >25% luminal occlusion in the arteries (27.2%). On univariate analyses using Cox’s regression model (Table 6), factors associated with graft loss were male donor, magnitude of increase in serum creatinine at time of diagnosis, BKVN class, severe interstitial fibrosis (Banff “ci” score=3), and severe tubular atrophy (Banff “ct” score=3). Severe interstitial inflammation (Banff “i” score=3) showed borderline significance (p=0.07). On multivariate analysis, donor sex and magnitude of increase in serum creatinine remained significant, while severe interstitial fibrosis was associated with a trend towards graft loss (p=0.09). Banff ci score=3, ct score=3, and class C disease were found to be multicollinear variables.

Figure 5.

Kaplan-Meier estimates of death censored graft survival using the Banff Working Proposal 2009 (A) and UMD/AST Classification schema (B and C).

Table 6.

Univariate and multivariate Cox regression analysis^* of factors associated with death censored graft loss in BKVN patients.

Models	Univariate analysis		Multivariate analysis
	HR (95% CI)	p value	HR (95% CI)	p value
Clinical parameters
Recipient age (per year)	1.00 (0.98–1.03)	0.8	1.02 (0.99–1.05)	0.2
Recipient sex (male)	0.69 (0.37–1.36)	0.3	0.79 (0.40–1.62)	0.5
Recipient race (White)	0.62 (0.30–1.44)	0.2	---
Donor source (Deceased donor)	1.34 (0.65–3.11)	0.4	---
Donor age (per year)	0.99 (0.98–1.01)	0.7	---
Donor sex (male)	2.33 (1.26–4.43)	0.007	2.24 (1.16–4.49)	0.02
Cold ischemia time	1.19 (0.41–3.36)	0.7	---
HLA mismatch (per mismatch)	1.08 (0.89–1.33)	0.4	---
Antibody induction	0.88 (0.40–1.76)	0.7	---
Maintenance immunosuppression	1.09 (0.59–2.04)	0.8	---
PCR testing at biopsy (Eras II, III)	1.43 (0.77–2.64)	0.3	---
PCR screening (Era III)	1.80 (0.61–4.33)	0.1	---
Previous acute rejection	1.03 (0.56–1.97)	0.9	---
Baseline serum creatinine	1.18 (0.71–1.86)	0.5	---
Creatinine rise at diagnosis	2.03 (1.26–3.24)	0.006	1.84 (1.08–2.98)	0.03
Log plasma BKV load at diagnosis	1.18 (0.17–4.75)	0.8	---
Log urine BKV load at diagnosis	1.17 (0.09–9.09)	0.9	---
Pathological parameters
Banff Working Proposal 2009
ClassA	1.47 (0.34–4.35)	0.03
Class B	1 (Reference)
Class C	2.45 (1.25–4.71)
UMD/AST Classification
Class A	NA**	0.02
Class B	1 (Reference)
Class C	2.34 (1.21–4.38)
BKV positive tubules
<1%	1 (Reference)	0.8
1–10%	0.78 (0.33–1.86)
>10%	1.01 (0.67–1.55)
Banff score
i score (i3)	1.76 (0.96–3.22)	0.07	1.36 (0.76–2.62)	0.4
t score (t3)	0.99 (0.54–1.94)	1.0	---
ci score (ci3)	2.35 (1.22–4.41)	0.01	1.86 (0.91–3.71)	0.09
ct score (ct3)	2.35 (1.22–4.41)	0.01	---
cv score (cv2 and 3)	1.06 (0.47–2.19)	0.9	---

^*Unit hazard ratios (HR) and confidence interval (CI) were calculated for continuous variables. Categorical data was treated as binary variables. Maintenance immunosuppression was analyzed as presence or absence of a triple immunosuppressive regimen (defined as Tacrolimus + Azathioprine + prednisone, or Tacrolimus + mycophenolate mofetil + prednisone, or Tacrolimus + Sirolimus + prednisone).

^**Odds ratio could not be calculated as there is a single observation in this category.

BKV DNA PCR follow-up following the diagnosis of BKVN was only performed in eras II and III: 32 of 35 patients received 3 or more PCR tests (median 14 times, range 3–86 times) performed a median of 17 months (range 3–71months) after the index biopsies. Twenty-six patients (81.3%) with clearance of viremia on follow-up showed better graft survival than the 6 patients (18.7%) without viral clearance (p=0.02, Figure 6A). The median time to clearance of viremia was 12 weeks (range 2–76 weeks). Graft survival in patients with and without clearance of viremia was 71.5% and 33.3% respectively at 3 years. Five years post-transplant, the corresponding graft survivals were 46.0% in patients and 16.7%. Patients with plasma viral load in the highest quartile at time of diagnosis of BKVN tended to have worse 5 year survival (25.0%) compared to patients in the lowest quartile, graft survival 56.3%). Most patients (27/32, 84.4%) continued to excrete virus in the urine and clearance of viruria as well as viremia could be documented in only in 5 cases. The median time to clearance of viruria was 48 weeks (range 7–72weeks). Graft survivals were not different between the patients with and without clearance of viruria (Figure 6B).

Figure 6.

Kaplan-Meier analyses of death censored graft survival in patients with and without clearance of viremia (A) and viruria (B) in era II and III.

Discussion

The unmet need to accurately predict the course of BKVN in individual patients led to formulation of the Banff Working Proposal 2009 for staging this disease (10). Independent scoring by four pathologists recently showed that this system has moderate reproducibility for overall class (11). The clinical utility of this system was not addressed in this study, although it was noted that Class B was associated with higher serum creatinine than class A. There were few cases C for meaningful analysis.

In our series the proportion of cases showing partial and complete responses 3 months after diagnosis tended to be higher in Banff class A compared to Banff class B and C, but the difference did not reach statistical significance. No conclusions could be reached about the better prognosis of UMD/AST class A since only 1 patient fell in this category. Class C, which is defined in an identical manner by both the Banff and UMD/AST systems, was associated with a greater hazard of graft loss compared to class B (HR=2.45, Table 3). In general, graft loss occurred in the majority of patients in this series who were followed up for a sufficient length of time. Paradoxically, Banff class A was associated with a greater risk of graft loss using Banff class B as the reference (HR=1.47). We believe this reflects the fact that Banff histologic classes are quite heterogeneous with respect to associated inflammation and tubulitis. Thus, many class A and B biopsies showed moderate to severe interstitial inflammation, severe tubulitis, and mild to moderate interstitial fibrosis. These parameters are known to have a strong influence on graft outcome (18).

On multivariate logistic regression analysis of individual histologic parameters, only severe interstitial inflammation was as an independent predictor of progression at 3 m. Thus, a strong case can be made for incorporating interstitial inflammation into the Banff Working Proposal 2009. This conclusion is in line with the known utility of inflammatory infiltrates as a prognostic parameter in other diseases affecting the renal allograft including antibody mediated rejection, mixed T-cell and antibody mediated rejection, and glomerulonephritis (12).

The primary importance currently given to extent of tubular injury in the Banff Working Proposal 2009 deserves comment. In our experience, assessment of tubular injury by standard light microscopy does not correlate well with the serum creatinine even outside the setting of BKVN, for example in patients with delayed graft function. Furthermore, it has also been shown that tubular injury signals measured by gene expression profiling do not correlate with several histologic features commonly scored in the Banff schema for allograft pathology (19). By contrast, in previously published studies using the UMD/AST system, class B disease characterized by inflammation has worse prognosis compared to class A disease characterized only by tubular injury alone (6, 20).

Histologic viral load expressed as percentage of BKV positive tubules did not influence outcome in this series. Prior studies that have evaluated this parameter report variable results: it was predictive in some studies (5, 7), but not in others (6, 20). These discrepancies can be at least partly explained by the focal nature of BKV infection in the human kidney (21). Another potential contributing factor is that the criteria used to grade histologic viral load have not been uniform. Some studies have arbitrarily defined cut offs for the percentage of tubules infected (Banff), while others have counted the number of tubular cross sections with viral cytopathic effect (5), the total number of infected cells (18), or the % of biopsy areas with infected tubules (9). Technical variables are also important: in an international multicenter immunohistochemistry trial we have observed that morphologic quantitation of viral load in biopsy tissue has poor reproducibility (unpublished data).

Limitations of this study include its retrospective, single center nature, and paucity of UMD/AST class A cases. Long-term follow-up was not consistently available for all patients. Sampling of blood and urine for BKV PCR did not follow a consistent protocol. The immunosuppressive regimens used in era II and III featured tacrolimus monotherapy. This may explain why our incidence of BKVN is at the lower end of the 1–10% range reported in the literature. It also indicates that additional studies are needed before our conclusions on prognosis and outcome can be generalized to all transplant patients.

In summary, we present data suggesting that inflammation and fibrosis are important prognostic parameters in determining the outcome of BKVN, whereas BKV induced tubular injury and histologic viral load are not informative. The Banff Working Proposal 2009 needs to be modified to incorporate the degree of inflammation into the morphologic criteria used for staging this disease. If the negative data with respect to viral cytopathic effect can be confirmed in other studies, a case could be made to revert to the use of the UMD/AST Classification wherein class A, B and C are distinguished entirely by presence or absence of inflammation and the extent of fibrosis.

Abbreviations

AST

American Society of Transplantation

BKV

Polyomavirus BK

BKVN

BKV nephropathy

UMD

University of Maryland