Chronic Histological Damage in Early Indication Biopsies Is an Independent Risk Factor for Late Renal Allograft Failure

Authors

  • M. Naesens,

    Corresponding author
    1. Department of Microbiology and Immunology, KU Leuven, Belgium, EU
    • Department of Nephrology and Renal Transplantation, University Hospitals Leuven, Leuven, Belgium, EU
    Search for more papers by this author
  • D. R. J. Kuypers,

    1. Department of Nephrology and Renal Transplantation, University Hospitals Leuven, Leuven, Belgium, EU
    2. Department of Microbiology and Immunology, KU Leuven, Belgium, EU
    Search for more papers by this author
  • K. De Vusser,

    1. Department of Nephrology and Renal Transplantation, University Hospitals Leuven, Leuven, Belgium, EU
    2. Department of Microbiology and Immunology, KU Leuven, Belgium, EU
    Search for more papers by this author
  • Y. Vanrenterghem,

    1. Department of Nephrology and Renal Transplantation, University Hospitals Leuven, Leuven, Belgium, EU
    2. Department of Microbiology and Immunology, KU Leuven, Belgium, EU
    Search for more papers by this author
  • P. Evenepoel,

    1. Department of Nephrology and Renal Transplantation, University Hospitals Leuven, Leuven, Belgium, EU
    2. Department of Microbiology and Immunology, KU Leuven, Belgium, EU
    Search for more papers by this author
  • K. Claes,

    1. Department of Nephrology and Renal Transplantation, University Hospitals Leuven, Leuven, Belgium, EU
    2. Department of Microbiology and Immunology, KU Leuven, Belgium, EU
    Search for more papers by this author
  • B. Bammens,

    1. Department of Nephrology and Renal Transplantation, University Hospitals Leuven, Leuven, Belgium, EU
    2. Department of Microbiology and Immunology, KU Leuven, Belgium, EU
    Search for more papers by this author
  • B. Meijers,

    1. Department of Nephrology and Renal Transplantation, University Hospitals Leuven, Leuven, Belgium, EU
    2. Department of Microbiology and Immunology, KU Leuven, Belgium, EU
    Search for more papers by this author
  • E. Lerut

    1. Department of Imaging and Pathology, KU Leuven, Belgium, EU
    Search for more papers by this author

Corresponding author: Prof. Dr. Maarten Naesens, maarten.naesens@uzleuven.be

Abstract

The impact of early histological lesions of renal allografts on long-term graft survival remains unclear. We included all renal allograft recipients transplanted at a single center from 1991 to 2001 (N = 1197). All indication biopsies performed within the first year after transplantation were rescored according to the current Banff classification. Mean follow-up time was 14.8 ± 2.80 years. In multivariate Cox proportional hazards analysis, arteriolar hyalinosis and transplant glomerulopathy were independently associated with death-censored graft survival, adjusted for baseline demographic covariates. Arteriolar hyalinosis correlated with interstitial fibrosis, tubular atrophy, mesangial matrix increase, vascular intimal thickening and glomerulosclerosis. Clustering of the patients according to these chronic lesions, reflecting the global burden of chronic injury, associated better with long-term graft survival than each of the chronic lesions separately. Early chronic histological damage was an independent risk factor for late graft loss, irrespective whether a specific, progressive disease was diagnosed or not, while T cell-mediated rejection did not. We conclude that individual chronic lesions like arteriolar hyalinosis, tubular atrophy, interstitial fibrosis, glomerulosclerosis, mesangial matrix increase and vascular intimal thickening cannot be seen as individual entities. The global burden of early chronic histological damage within the first year after transplantation importantly affects the fate of the allografts.

Abbreviations
ANOVA

analysis of variance

CNI

calcineurin inhibitor

PCA

principal component analysis

PRA

panel reactive antibody

2D

2-dimensional

PAS

periodic acid-Shiff

R

correlation coefficient.

Introduction

In a seminal single-center retrospective study on 153 failed grafts, El-Zoghby et al. showed that the cause of late graft loss is traceable in most cases. Specific disease entities were identified for 95.4% of patients—only in 4.6% of patients the reason for graft loss was not clear [1]. Importantly, it appeared that glomerular disease (36.6%) and rejection phenomena (20.2%), most notably antibody-mediated rejection, were the most frequent causes of graft loss [1], which was then confirmed by others [2-5].

Nonetheless, in previous years, much attention has gone to early chronic histological damage after transplantation. Many transplant centers are currently performing protocol renal allograft biopsies in the first year(s) after transplantation [6-9]. From these protocol biopsy studies, it became clear that chronic histological damage to the tubulointerstitial, vascular and glomerular compartment occurs already in the first years after transplantation [10-12]. In many patients, there is no identifiable cause of this progressive chronic damage, and no targeted therapy is thus available to prevent or revert the progression of nonspecific chronic injury. Importantly, there is some evidence in the literature that early chronic damage, especially in association with inflammatory lesions, associates with graft survival [6, 13-15]. However, the contribution to graft outcome of the individual chronic histological lesions observed early after transplantation, remains largely unclear.

Thus, while retrospective studies on late graft survival illustrated that specific, progressive disease processes cause graft loss, prospective protocol biopsy studies early after transplantation showed a significant impact of early chronic damage and inflammation, often of undefined etiology, on graft survival. To allow for timely and targeted preventative or curative treatment changes, it is necessary to eliminate this apparent discrepancy, and get insight in which early histological lesions are risk factors for late graft loss. Therefore, in the current cohort study, we clarified the impact of the pattern of histological lesions in early indication biopsies on long-term graft survival.

Materials and Methods

Study population

The patient population consisted of all recipients of a renal allograft or a combined kidney–pancreas graft, transplanted at the University Hospitals Leuven between January 1, 1991 and December 31, 2001 (N = 1197). Recipients of kidneys allografts combined with other organ types were excluded. Within this time frame, 54 patients had more than one renal transplant. All data were collected within standard of care and routine clinical practice.

Biopsies and histological scoring

All renal allograft biopsies, performed within the first year after transplantation, were included in this study (N = 963 biopsies in 574 individual transplants). Posttransplant biopsies were all performed for cause, at time of graft dysfunction. No posttransplant surveillance (“protocol”) biopsies were performed in this patient cohort.

For this study, all biopsies were retrospectively reviewed by one pathologist (EL), without knowledge of any clinical information or timing of the biopsy. Slides containing 4–10 paraffin sections (2 μm) were stained with hematoxylin eosin (HE), periodic acid-Schiff (PAS) and silver methenamine (Jones). For this study, an immunohistochemical C4d stain (monoclonal antibody, dilution 1:500, Quidel Corporation, Santa Clara, CA, USA) was performed on frozen tissue. The severity of histological lesions was semiquantitatively scored according to the revised Banff criteria [9, 16, 17]. In addition, the number of sclerosed glomeruli (0 = 0%; 1 = 1–25%; 2 = 25–50%; 3 = >50% glomerulosclerosis) was scored separately, and peritubular capillaritis was scored based on the recently proposed score [9]. Total inflammation was evaluated in the whole cortex using the total i score (total i 0 = <10%; 1 = 10–25%; 2 = 25–50%; 3 = >50%) [9]. C4d deposition in the peritubular capillaries was scored from 0 to 3, with 0 = negative, 1 = <25%, 2 = 25–75% and 3 = >75% of peritubular capillaries positive. Changes suggestive of antibody-mediated rejection were defined as the presence of microcirculation inflammation (glomerulitis or peritubular capillaritis) and/or transplant glomerulopathy, provided that the histological lesions were not secondary to de novo or recurrent primary glomerular disease. No data on donor-specific antibodies were available for this patient cohort.

Data collection and statistical analysis

All clinical data were prospectively collected in electronic clinical patient charts, which were used for clinical patient management as well as were directly linked to the database used in this study. This clinical database was transferred to SAS data files (SAS institute, Cary, NC, USA) at time of analysis.

For variance analysis of continuous variables in different groups, nonparametric Wilcoxon–Mann–Whitney U, nonparametric ANOVA and parametric one-way ANOVA were used, as appropriate. Dichotomous variables were compared using the chi-square test. Multivariate Cox proportional hazards analysis with backward variable selection was performed to evaluate the clinical and histological determinants of death-censored graft survival. For evaluation of ‘long-term’ graft survival, only patients with a functioning graft for at least 1 year posttransplant were included. Repeat transplants within the inclusion time frame were regarded as separate entities in statistical analysis. For the histological lesions, the highest score achieved in the first year after transplantation was recorded for each lesion and each transplant individually, to avoid overrepresentation of patients with repeat biopsies. In multivariate analysis with backward selection, we included all covariates associated at the p = 0.2 level in univariate analysis. Log(−log(survival)) versus log(time) plots showed no nonproportionality of hazards. Kaplan–Meier curves and log-rank testing were used to plot graft survival. Logistic regression analysis and receiver operating characteristics (ROC) curves were used to evaluate the predictive performance of the models.

For evaluation of the correlation of the different histological lesions, spearman correlation analysis was used, as well as unsupervised hierarchical clustering analysis and principal component analysis. From the principal component analysis, a 2D scatter plot was generated to represent the histological lesions according to their score in the loading matrix. K-means clustering was used to determine distinct clusters of patients based on significantly correlated histological lesions.

All tests were two-sided and p values of less than 0.05 were considered to be statistically significant. Analyses were done with SAS (version 9.2; SAS institute, Cary, NC, USA), JMP9.0 (SAS institute) and GraphPad Prism (version 5.00; GraphPad Software, San Diego, CA, USA).

Results

Study population characteristics

Patient and donor demographics, and transplantation-related characteristics of all 1197 recipients are summarized in Table 1. Mean follow-up time posttransplant of patients with a functioning graft at time of data extraction was 14.5 ± 2.80 years posttransplant. One thousand eighty two patients (90.2%) reached at least 1-year graft survival. One hundred fifteen patients (9.6%) did not reach 1-year graft survival, due to death with a functioning graft (N = 41; 3.4%) or due to graft loss and need for restart of dialysis or for a new transplant (N = 74; 6.2%). Within the first year after transplantation, acute rejection was the main cause of graft loss (N = 30/74; 40.5%) (Supporting Table S1). After 1 year posttransplant, 549 (45.9%) additional grafts were lost during follow-up, at a mean time posttransplant of 8.2 ± 4.2 years. Three hundred and twelve patients (26.1%; mean 12.3 ± 4.4 years posttransplant) lost the graft after 1 year due to death with a functioning graft, and 237 (19.8%; mean 7.7 ± 4.3 years posttransplant) patients due to intrinsic loss of graft function. Actuarial overall graft half-life (uncensored for death of the recipients) was 12.8 years after transplantation. Death-censored graft survival was 63% at 20 years after transplantation (Table 1).

Table 1. Demographic and clinical characteristics of transplant donors and recipients (N = 1197)
Donors 
  1. a

    Renal diseases with recurrence potential were primary FSGS, idiopathic membranous nephropathy, IgA nephropathy, membranoproliferative glomerulopathy, nephropathy associated with systemic lupus erythematosis, ANCA-associated vasculitis, anti-GBM disease, Henoch-Schonlein purpura nephropathy, hemolytic uremic syndrome, amyloidosis, monoclonal immunoglobulin deposition disease and primary glomerulonephritis without clear diagnosis.

  2. b

    CNI = calcineurin inhibitor.

Age38.8 ± 15.6
Sex (male)747/1197 (62.4%)
Deceased donation1186/1197 (99.1%)
Stroke as cause of death502/1197 (41.9%)
Recipients 
Age49.1 ± 13.3
Sex (male)698/1197 (58.3%)
Cause of renal failure (glomerulonephritis/interstitial/hereditary/ systemic/malignancy/unknown)324/200/264/237/7/165
Recurrence potential of primary renal diseasea390/1197 (32.6%)
Race/ethnicity (Caucasian)1149/1197 (96.0%)
Repeat transplantation390/1197 (32.6%)
Number of HLA mismatches2.57 ± 1.30
Type of primary immunosuppression 
 Number of immunosuppressants in the regimen2.84 ± 0.61
 Induction therapy168/1197 (14.0%)
 Corticosteroids1197/1197 (100%)
 CNI-based therapy (cyclosporine or tacrolimus)b1179/1197 (98.5%)
  Cyclosporine868/1197 (72.5%)
  Tacrolimus311/1197 (26.0%)
 Use of antimetabolite (azathioprine or mycophenolate)788/1197 (65.8%)
  Azathioprine256/1197 (21.4%)
  Mycophenolate532/1197 (44.4%)
 Use of other immunosuppressive agent61/1197 (5.1%)
Posttransplant follow-up 
Follow-up time posttransplant of patients with a functioning graft (years)14.5 ± 2.80
Delayed graft function166/1197 (13.9%)
Acute T cell-mediated rejection in the first year296/1197 (24.7%)
Acute antibody-mediated rejection in the first year138/1197 (11.5%)
Overall graft survival (not censored for patient death) 
 At 1 year90.2%
 At 2 years86.7%
 At 5 years78.4%
 At 10 years59.9%
 At 20 years35.2%
Death-censored graft survival 
 At 1 year93.7%
 At 2 years91.5%
 At 5 years87.0%
 At 10 years77.9%
 At 20 years63.3%

In the total cohort of 1197 kidney transplants, 963 renal allograft for-cause biopsies were performed within the first year after transplantation in 574 individual transplants. The mean number of biopsies per transplant, of the group with at least one biopsy within the first year after transplantation, was 1.68 ± 0.99. Mean time posttransplantation of the biopsies was 54 ± 77 days (range 2–363 days). Only 49 biopsy samples were of inadequate quality to yield a diagnosis and to evaluate according to the Banff scheme; these biopsies were not included in the analysis. The distribution of the histopathological lesions in the biopsies of sufficient quality (N = 914) and diagnoses is shown in Table 2. Supporting Figure S1 provides an overview of the distribution of patients and biopsies in the different subanalyses.

Table 2. Histopathologic findings in the 963 indication biopsies performed within the first year after transplantation in all kidney recipients transplanted between January 1991 and December 2001 (N = 1197)
Histological diagnosis (primary disease)a   N (%)
  1. a

    In many biopsies, there was concomitant presence of different diagnosis (e.g. interstitial fibrosis and tubular atrophy together with acute T cell-mediated rejection, recurrent disease with borderline changes). Here, only the primary diagnosis is represented, which inevitably oversimplifies the complex histological appearance of the biopsies.

  2. b

    Changes suggestive of antibody-mediated rejection comprise microcirculation inflammation (glomerulitis and/or peritubular capillaritis), and transplant glomerulopathy.

Normal   167 (17.3%)
De novo glomerular disease   6 (0.6%)
Recurrent glomerular disease   16 (1.7%)
Polyomavirus infection   6 (0.6%)
Isolated acute T cell-mediated rejection   131 (13.6%)
  Grade I   49 (5.1%)
  Grade II   82 (8.5%)
Changes suggestive of antibody-mediated rejectionb   385 (40.0%)
  Without acute or borderline T cell   48 (4.99%)
  mediated rejection   32 (3.3%)
  With borderline changes   75 (7.79%)
  With acute T cell-mediated rejection   262 (27.2%)
Isolated borderline changes   93 (9.66%)
  Grade I   99 (10.3%)
  Grade II   9 (0.93%)
  Grade III   2 (0.2%)
No diagnosis (insufficient biopsy quality)   49 (5.1%) 
Individual histological lesion score0123
Tubulitis379 (41.5%)295 (32.3%)160 (17.5%)80 (8.8%)
Interstitial inflammation358 (39.2%)151 (16.5%)142 (15.5%)263 (28.8%)
Intimal arteritis642 (70.2%)239 (26.1%)29 (3.2%)4 (0.4%)
Total i score284 (31.1%)167 (18.3%)156 (17.1%)307 (33.6%)
Glomerulitis592 (64.8%)139 (15.2%)136 (14.9%)47 (5.1%)
Peritubular capillaritis745 (81.5%)148 (16.2%)21 (2.3%)0 (0.0%)
C4d deposition in peritubular capillaries576 (63.0%)59 (6.5%)63 (6.9%)216 (23.6%)
C4d deposition in glomerular capillaries650 (71.1%)25 (2.7%)86 (9.4%)153 (16.7%)
Interstitial fibrosis607 (66.4%)198 (21.7%)86 (9.4%)23 (2.5%)
Tubular atrophy504 (55.1%)365 (39.9%)41 (4.5%)4 (0.4%)
IF/TA grade678 (74.2%)212 (23.2%)22 (2.4%)2 (0.2%)
Transplant glomerulopathy902 (98.7%)6 (0.7%)5 (0.5%)1 (0.1%)
Arteriolar hyalinosis556 (60.8%)240 (26.3%)103 (11.3%)15 (1.6%)
Vascular intimal thickening775 (84.8%)88 (9.6%)39 (4.3%)12 (1.3%)
Mesangial matrix increase764 (83.6%)62 (6.8%)72 (7.9%)16 (1.8%)
Glomerulosclerosis656 (71.8%)211 (23.1%)36 (3.9%)11 (1.2%)

Clinical determinants of death-censored graft survival

In multivariate Cox proportional hazards analysis of baseline demographics and death-censored graft survival (N = 1192 patients with full clinical data; five patients were excluded because of missing data), four baseline demographic variables were significant and independent risk factors for long-term graft loss, adjusted for transplant year and type of immunosuppression (Supporting Table S2): younger recipients age (40–59 vs. <40 years adjusted HR 0.72 [0.55–0.93], p = 0.01; ≥60 vs. <40 years adjusted HR 0.67 [0.48–0.94], p = 0.02); older donor age (40–59 vs. <40 years adjusted HR 1.56 [1.22–1.99], p = 0.0004; ≥60 vs. <40 years adjusted HR 2.77 [1.93–396], p < 0.0001), higher number of HLA mismatches (1–2 vs. 0 adjusted HR 1.91 [1.14–3.20], p = 0.01; 3–4 vs. 0 adjusted HR 2.09 [1.26–3.46], p = 0.004; 5–6 vs. 0 adjusted HR 1.92 [0.91–4.06], p = 0.09) and repeat transplantation (adjusted HR 1.47 [1.07–2.01], p = 0.02).

Next, also clinical posttransplant variables like acute T cell-mediated rejection, potentially progressive diseases and delayed graft function were included in the multivariate Cox proportional hazards model for death-censored graft survival >1 year (N = 1077 patients; 115 recipients with full clinical data lost their graft during the first year after transplantation and were not included in this analysis). “Progressive disease” was defined as changes suggestive of antibody-mediated rejection, de novo or recurrent glomerular disease and polyomavirus nephropathy. These diagnoses were taken together in this analysis in one single category, as the number of patients with de novo or recurrent glomerular disease (N = 16) and with polyomavirus nephropathy (N = 4) was too low to allow for robust statistical interpretation in the multivariate models. Again, younger recipient age, older donor age, higher number of HLA mismatches and repeat transplantation were significantly and independently associated with long-term graft survival, adjusted for transplant year and primary immunosuppression, next to the diagnosis of a potentially progressive disease within the first year after transplantation (presence vs. absence of a potentially progressive disease HR 1.69 [1.28–2.23], p = 0.0002) (Figure 1, Supporting Table S3).

Figure 1.

Kaplan–Meier estimates of death-censored graft survival, stratified respectively according to recipient age, donor age, number of HLA mismatches, repeat transplantation, changes suggestive of antibody-mediated rejection and de novo or recurrent glomerular disease. The p-values are calculated with the log-rank test.

Diagnosis of at least one episode of T cell-mediated rejection within the first year after transplantation was significantly associated with long-term graft survival after the first year in univariate analysis (adjusted HR 1.40 [1.06–1.84], p = 0.02), but no longer after backward selection in the multivariate model. Therefore, T cell-mediated rejection was not considered a “progressive disease” in the analyses.

Finally, if performing a biopsy in the first year after transplantation was entered in the multivariate model, this variable was identified as the most significant independent risk factor for death-censored graft loss after the first year posttransplant (HR 1.83 [1.45–2.30], p < 0.0001, adjusted for donor age, repeat transplantation, transplant year and type of immunosuppression).

Histological determinants of death-censored graft survival

In Cox proportional hazards analysis of the association between individual histological lesions in the first year after transplantation and death-censored graft survival in patients with at least 1 year graft survival (N = 491), interstitial fibrosis, tubular atrophy, arteriolar hyalinosis, mesangial matrix increase, glomerulosclerosis and transplant glomerulopathy were significantly associated with long-term death-censored graft survival, adjusted for baseline demographic factors (transplant year, type of immunosuppression, donor and recipient age) (Supporting Table S4, Figure 2). None of the individual acute inflammatory lesions was associated with death-censored graft survival after the first year, except glomerulitis. Vascular intimal thickening was significant in univariate analysis, unadjusted for baseline demographic variables, but no longer significant after adjustment for the baseline demographic variables (Supporting Table S4, Figure 2).

Figure 2.

Association of death-censored graft survival and selected individual histological lesions in indication biopsies obtained the first year after transplantation, by Kaplan–Meier survival analysis. The p-values are calculated with the log-rank test.

In the final multivariate Cox proportional hazards model for death-censored graft survival after the first year after transplantation, only transplant glomerulopathy and arteriolar hyalinosis remained significant, adjusted for baseline demographic factors (donor and recipient age, transplant year and primary immunosuppression) (Supporting Table S4). All other chronic histological lesions (tubular atrophy, interstitial fibrosis, mesangial matrix increase, glomerulosclerosis), albeit highly statistically significant when assessed individually, were not retained in the multivariate model after backward elimination.

Correlation between the individual histological lesions

The reason for disappearance of interstitial fibrosis, tubular atrophy, mesangial matrix increase and glomerulosclerosis score in multivariate analysis after backward elimination was found in the very significant correlation between these histological lesions. The correlation between the different maximum histological lesions is illustrated in the heatmap plot (Figure 3A, Supporting Table S5). The unsupervised hierarchical clustering analysis identified two different clusters of histological lesions. First, a cluster that included inflammatory lesions like interstitial inflammation, tubulitis, intimal arteritis, total i score, peritubular capillaritis and C4d deposition in glomerular and peritubular capillaries. Microcirculation inflammation (glomerulitis and peritubular capillaritis) correlated best with C4d deposition. The second cluster consisted of interstitial fibrosis, tubular atrophy, arteriolar hyalinosis, mesangial matrix increase, vascular intimal thickening, glomerulosclerosis and transplant glomerulopathy. All chronic histological lesions that highly significantly correlated with each other, except transplant glomerulopathy, which was a rare finding in this early stage after transplantation (1.2% of biopsies), and correlated only marginally with the other chronic lesions (Figure 3A, Supporting Table S5). Transplant glomerulopathy correlated very weakly but significantly with glomerulitis, vascular intimal thickening and mesangial matrix increase (respectively r = 0.13, p = 0.01; r = 0.10, p = 0.03; r = 0.29, p < 0.0001), but not with any other lesion.

Figure 3.

(A) Spearman correlation heatmap of individual histological lesions in 914 indication biopsies of 563 renal allograft recipients who had at least one posttransplant biopsy within the first year after transplantation. The dendrogram is obtained by unsupervised hierarchical clustering analysis of the histological lesions. The order of the lesions in the y axis equals the order in the x-axis. (B) Principal component analysis using the individual histological lesions in indication biopsies of 491 renal allograft recipients with at least one posttransplant biopsy within the first year after transplantation and at least 1-year graft survival. ci, interstitial fibrosis; ct, tubular atrophy; ah, arteriolar hyalinosis; mm, mesangial matrix increase; gs, glomerulosclerosis; cv, vascular intimal thickening; cg, transplant glomerulopathy; t, tubulitis; i, interstitial inflammation; ti, total i score; v, intimal arteritis; C4d ptc, C4d deposition in the peritubular capillaries; C4d ptc, C4d deposition in glomerular capillaries; ptc, peritubular capillaritis. (C) K-means clustering of 491 patients based on chronic histological damage. The clustering was based on the chronic histological lesions that significantly associated with long-term graft survival in univariate analysis (interstitial fibrosis, tubular atrophy, arteriolar hyalinosis, mesangial matrix increase and glomerulosclerosis). (D) Death-censored graft survival according to K-means clustering of 491 patients based on chronic histological damage. The p-value is calculated with the log-rank test.

Looking at the maximum interstitial fibrosis score reached within the first 3 months after transplantation, and comparing this with the maximum interstitial fibrosis score reached in months 3–12 (67 patients with both a biopsy in the first 3 months and in months 3–12), we observed that only 6 of 67 (9%) patients had regression of the interstitial fibrosis score. 28 patients (42%) had the same ci score in the second interval compared to the first interval. The other patients (N = 33/67, 49%) had increase in the ci score within the first year posttransplantation.

Principal component analysis of the individual histological lesions

Principal component analysis (PCA) was applied to the maximum histological lesions in posttransplant biopsies within the first year after transplantation, and the summary plot is represented in Figure 3B. Principal component 1 explained 23.9% of the total variance in the cohort, while principal component 2 explained 18.9% of the variance. There are at least two groups of features, similar to the results of the unsupervised hierarchical clustering analysis: a cluster with chronic histological lesions (interstitial fibrosis, tubular atrophy, arteriolar hyalinosis, mesangial matrix increase, vascular intimal thickening and glomerulosclerosis), and a cluster with inflammatory/immunity-related lesions (interstitial inflammation, tubulitis, intimal arteritis, peritubular capillaritis, total i score, glomerulitis and C4d deposition in glomeruli and peritubular capillaries). These findings illustrate that acute inflammatory lesions are very often present together in the same biopsy. On the other hand, histologically very distinct chronic lesions in different renal compartments track together.

K-means clustering to identify two distinct patient groups based on total chronic injury burden

To assess the impact of the global burden of chronic damage, and not the individual Banff lesions separately, the patient population was divided into two clearly distinct groups of patients using K-means clustering (Figure 3C), based on those chronic histological lesions that significantly associated with long-term graft survival in univariate analysis (see above: interstitial fibrosis, tubular atrophy, arteriolar hyalinosis, mesangial matrix increase and glomerulosclerosis). The differences between both patient clusters are presented in Supporting Table S6, which illustrates that the clusters clearly differ in terms of chronic histological damage, but also in donor age and stroke as reason for donor death (both significantly higher in the cluster with extensive chronic histological damage). In multivariate logistic regression analysis with backward elimination, donor age and stroke as the reason of donor death were the only determinants of the K-means clustering result. In the final model, acute T cell-mediated rejection or progressive diseases were not withheld as independent risk factors for the global burden of chronic histological damage.

In univariate analysis, the division of the patients in two clusters by K-means clustering of chronic histological lesions within the first year after transplantation was a strong risk factor for death-censored graft loss in patients with at least 1-year graft survival (Figure 3D). In multivariate analysis with backward elimination, with baseline demographics, all histological lesions, and the two-groups division of the K-means clustering entered into the model, two histological variables remained in the final model, independent of baseline demographics (donor and recipient age, transplant year and primary immunosuppression): transplant glomerulopathy (adjusted HR 5.80 [2.57–13.1], p < 0.0001) and the K-means cluster of total chronic histological damage (adjusted HR 2.18 [1.50–3.17], p < 0.0001). Taking into account all mutually correlating chronic histological lesions in the multivariate model is thus stronger than each of the contributing individual lesions separately. Although both K-means clusters clearly differ from each other in terms of donor characteristics (age and cause of death), the association between this clustering and graft outcome was independent of these donor characteristics, which illustrates that the association between the global burden of chronic damage and graft survival cannot be solely attributed to donor graft quality. Of note, there was no association between the extent of chronic damage (K-means clustering result) and patient survival (HR 1.08 [0.80–1.46], p = 0.62).

Unsupervised hierarchical clustering analysis

Unsupervised Ward hierarchical clustering analysis was applied to the data set, to evaluate the overall pattern of all maximum histological lesions of the patients as individuals. From this hierarchical clustering analysis, six clearly distinct patient groups were identified (Figure 4A): (1) inflammation with C4d deposition, (2) inflammation without C4d deposition, (3) no inflammation, no chronic damage (‘normal’), (4) chronic damage without inflammation, (5) chronic damage with inflammation (6) transplant glomerulopathy.

Figure 4.

(A) Unsupervised Ward hierarchical clustering analysis. The clustering was performed on the histological lesions in 914 good-quality renal allograft for-cause biopsies obtained within the first year after transplantation in 563 individual transplants. (B) Death-censored graft survival according to unsupervised Ward hierarchical clustering analysis of 491 patients based on the clusters identified in (A). The p-value is calculated with the log-rank test.

The result of the unsupervised hierarchical clustering correlated highly significantly with death-censored graft survival after the first year posttransplant (Figure 4B). This analysis confirmed the significantly worse death-censored graft survival for patients with transplant glomerulopathy compared to any other patient group (adjusted HR 12.3 [4.77–31.5] compared to “normal”, p < 0.0001; adjusted for donor and recipient age, transplant year and primary immunosuppression). Chronic histological damage with inflammation was associated with worse death-censored graft survival compared to “normal” (adjusted HR 2.80 [1.50–5.21], p = 0.001) and compared to patients with active inflammation without chronic damage (adjusted HR 2.43 [1.38–4.27], p = 0.002 for inflammation without C4d deposition, and 2.07 [1.12–5.32], p = 0.021 for inflammation with C4d deposition). Also patients with chronic histological damage in the absence of inflammatory lesions had significantly worse death-censored graft survival compared to “normal” histology (adjusted HR 2.05 [1.07–3.91], p = 0.030), but not compared to patients with active inflammation without chronic damage.

Importantly, in the absence of chronic damage, death-censored graft survival of patients with active inflammation, either with or without C4d deposition, did not differ from survival in patients with normal histology (respectively adjusted HR 1.35 [0.73–2.49], p = 0.34 and 1.15 [0.67–1.98], p = 0.62). After 15 years posttransplantation, there was a trend toward decreased death-censored graft survival in the two patient groups with inflammation without chronic damage compared to patients with normal histology in the first year after transplantation, although low numbers obviated robust statistical analysis and interpretation of this finding.

Relative Impact of Chronic Damage Versus Progressive Diseases

Finally, multivariate Cox proportional hazards analysis was performed to evaluate the relative impact of progressive diseases (de novo or recurrent glomerular diseases, changes suggestive of antibody-mediated rejection and polyomavirus nephropathy) versus individual histological lesions on long-term death-censored graft survival (excluding glomerulitis and transplant glomerulopathy, which were contained in the definition of progressive disease) in 491 renal allograft recipients with at least one biopsy within the first year, and at least 1-year graft survival.

Both progressive disease (adjusted HR 1.56 [1.10–2.22], p = 0.01) and IF/TA grade (adjusted HR grade 1 vs. 0: 2.04 [1.39–2.99], p = 0.0003; adjusted HR grade 2/3 vs. 0: 4.12 [2.17–7.83], p < 0.0001) were significant and independent risk factors for late graft failure, adjusted for transplant year, primary immunosuppression, donor age and recipient age (Figure 5A). In the absence of IF/TA (grade 0), there was no difference in long-term outcome between patients with and without a progressive disease (adjusted HR 1.32 [0.84–2.09]; p = 0.23). In the presence of mild IF/TA (grade 1), the occurrence of a progressive disease comprised significantly worse long-term death-censored graft survival (adjusted HR 2.14 [1.21–3.78]; p = 0.009). With moderate to severe IF/TA (grade 2–3), there was no difference in outcome between patients with versus without progressive diseases (adjusted HR 0.94 [0.26–3.33]; p = 0.92). Importantly, both in the absence of progressive diseases (IF/TA 1 vs. 0 adjusted HR 1.56 [0.90–2.70], p = 0.11; IF/TA 2/3 vs. 0 adjusted HR 5.20 [1.80–15.0], p = 0.002) as in the presence of progressive diseases (IF/TA 1 vs. 0 adjusted HR 2.53 [1.53–4.19], p = 0.0003; IF/TA 2/3 vs. 0 adjusted HR 3.69 [1.64–8.31], p = 0.002), higher IF/TA grade was significantly associated with worse death-censored graft survival (Figure 5A).

Figure 5.

(A) Death-censored graft survival in 491 patients, according to the presence/absence of a progressive disease within the first year after transplantation, and according to the maximum IF/TA grade reached in indication biopsies obtained within the first year after transplantation. (B) Death-censored graft survival in 491 patients, according to the presence/absence of a progressive disease within the first year after transplantation, and according to the K-means clustering based on chronic histological damage. The p-values are calculated with the log-rank test.

Similarly, when the global burden of chronic damage (K-means clustering result) was evaluated, progressive disease (1.48 [1.05–2.10], p = 0.03) and K-means clustering result (adjusted HR 2.36 [1.65–3.38], p < 0.0001) were significant and independent risk factors for late graft failure (Figure 5B). The global burden of chronic damage (the K-means clustering result) was significantly associated with death-censored graft survival, both in the presence of a progressive disease (adjusted HR 2.57 [1.62–4.08], p < 0.0001) as in the absence of a progressive disease (adjusted HR 2.01 [1.17–3.47], p = 0.01). In the absence of chronic damage, the presence of a progressive disease did not lead to impaired graft outcome (adjusted HR 1.33 [0.86–2.07], p = 0.20), while diagnosis of a progressive disease process impaired graft outcome in the presence of chronic damage, although this association was only borderline significant (adjusted HR 1.71 [0.98–2.97], p = 0.058) (Figure 5B).

Although the association between chronic histological damage and outcome was highly statistically significant, the predictive performance of the early histological appearance for long-term graft survival was insufficient to use early histology as sole predictor of long-term outcome, with an AUC under the receiver operating characteristic (ROC) curve for prediction of 10-year graft function of 0.60 for the combination of “progressive disease” and global burden of chronic damage (K-means clustering result of the chronic histological lesions). Adding recipient and donor characteristics (donor age, recipient age, number of HLA mismatches, repeat transplantation) only marginally improved this predictive performance to an AUC under the ROC curve of 0.67.

Discussion

The current study clarifies the impact of histological lesions early after renal transplantation on long-term graft survival. Long-term graft survival is highly significantly associated with the global burden of chronic histological damage of all renal compartments (tubulointerstitial, glomerular and vascular) in the first year after transplantation, independent of baseline demographic factors. In addition, the early presence of a specific progressive disease is an important risk factor for graft loss, affecting outcome already in the first years after detection of this lesion. Nevertheless, we demonstrate that the amount of chronic damage in early biopsies partly determines the outcome of specific disease processes diagnosed early after transplantation. Interestingly, T cell-mediated rejection in itself, and treated with standard antirejection therapies, is not an independent risk factor for graft failure after the first year posttransplant. Finally, the current study shows that chronic damage without clear etiology is also an important and independent risk factor for graft loss.

Using unsupervised principal component analysis and unsupervised hierarchical clustering analysis, we confirm that interstitial fibrosis, tubular atrophy, glomerulosclerosis, arteriolar hyalinosis, mesangial matrix increase and vascular intimal thickening very often collate in the same biopsies, as was previously shown [18]. Thus, when analyzing the impact of histological damage on renal transplant outcome, one should not evaluate the impact of each lesion separately, but take into account the significant entanglement of the different chronic lesions. Different chronic histological lesions occur simultaneously, and we show that the global burden of chronic damage associates stronger with graft survival than the individual chronic histological lesions. Studying single histological lesions, without considering the closely correlating other histological lesions (as has been done largely in the past, e.g. in the study of risk factors and impact of ‘chronic allograft nephropathy’), is too simplistic, could lead to erroneous conclusions, and should thus be avoided.

Our data thus show that early chronic histological damage, of whatever cause [19-23], should be recognized as a major contributor to graft loss, independent of baseline demographic factors like donor and recipient age, transplant year and primary immunosuppression. While the etiology of chronic histological damage in association with a specific or progressive disease might be clear, we demonstrate that many patients also have chronic damage of apparently unknown cause. In order to improve the outcome of renal transplantation, it will be important to elucidate the underlying injury processes that lead to chronic damage, e.g. by using molecular screening tools [22, 23].

The question that arises, especially in the light of recent data suggesting that specific disease processes cause graft loss [1, 2, 5], is whether the association between the global burden of chronic histological damage and long-term graft survival is causal, and if so, how. Several hypotheses for the significant association of early chronic damage on long-term graft survival can be put forward, and it is possible that these phenomena cooccur. First, it is possible that chronic damage culminates in graft loss by inducing a snowball effect [24]. Our data in a large patient cohort show that even the lowest degrees of chronic histological lesions (grade 1 tubular atrophy, arteriolar hyalinosis and mesangial matrix increase) are significantly associated with impaired long-term graft outcome, and should therefore not be regarded as insignificant. Second, it is plausible that preexisting chronic histological damage significantly decreases the adaptive capacity of the transplanted kidneys to specific disease processes. This decreased adaptive capacity and significantly worse outcome of kidneys with chronic damage, independent of the specific (glomerular) disease process, is well known in the native kidney glomerulonephritis literature [25-27]. Finally, it could be hypothesized that kidneys with preexisting histological damage are more susceptible to new injury processes. Kidneys with chronic damage have a higher antigenic potential, which could induce alloreactive immune responses like T cell-mediated and antibody-mediated rejection [28, 29], and damaged kidneys could be more susceptible to viral infections like polyomavirus-associated nephropathy [30]. Each of these hypotheses needs to be tested in well-powered prospective protocol biopsy studies with sufficient data on the long-term subclinical histological evolution, and in studies that correlate the final cause of graft loss with the graft's early histological appearance. However, even then it will be difficult to prove that chronic damage is in itself deleterious: chronic damage can be just an innocent secondary reaction without intrinsic harmful effects. Only studies with specific inhibitors of fibrogenesis will be able to elucidate the exact pathogenic role of chronic damage in the outcome of kidney transplantation.

In addition, the current study shows that transplant glomerulopathy is the single most powerful risk factor of renal allograft failure. Although the prevalence of this lesion in the first year after transplantation was low in the current study, this corresponds to findings in protocol biopsy studies [10, 11]. The highly significant association between transplant glomerulopathy and graft survival has been described previously [1, 14, 31]. Transplant glomerulopathy has been related to antibody-mediated rejection, donor-specific antibodies, C4d deposition and microcirculatory inflammation [32-34]. In our study, it is interesting to note that transplant glomerulopathy also correlates significantly (albeit weakly) with mesangial matrix increase and vascular intimal thickening, but not with C4d deposition. Moreover, peritubular capillaritis and C4d deposition, and diagnosis of antibody-mediated rejection in the first year after transplantation are not associated with long-term graft survival. This finding warrants further study on the exact nature and impact of early antibody-mediated changes and C4d deposition, as was also the conclusion of the most recent international Banff consensus meeting [17].

The finding that early acute inflammation has an additive effect on graft survival if the inflammation is associated with chronic histological injury has previously been shown in protocol studies [14, 15], and this is confirmed by the current study. We indeed show that the cluster of patients with chronic damage and inflammation has worse graft survival compared to all other patient clusters, except patients with transplant glomerulopathy. Moreover, the presence of chronic damage portends a significantly worse outcome of a progressive disease process like antibody-mediated rejection, polyomavirus nephropathy or de novo/recurrent glomerular disease. Importantly and unexpectedly however, in the absence of chronic damage, diagnosing of a progressive disease does not lead to impaired long-term graft outcome. This illustrates that not only the specific disease process, but also the extent of chronic damage associated with this process should be taken into account when evaluating transplant kidney biopsies, for prognostication and likely for therapeutic decision making. Conversely, as the presence of a progressive disease significantly impacts on outcome in patients with chronic histological damage, it is essential to identify the underlying progressive disease. Evaluating chronic damage without evaluation of the underlying disease process, withholds the patients necessary therapies to halt or slow the injury process.

Furthermore, it is interesting to note that acute T cell-mediated rejection within the first year after transplantation does not associate independently with decreased long-term graft survival. Within the first year after transplantation however, acute rejection was a significant contributor to graft loss, as 2.5% of the patients lost their graft within the first year due to acute rejection phenomena. From this, it is clear that T cell-mediated rejection forms an acute risk for graft loss, but if overcome by successful antirejection treatment, T cell-mediated rejection in itself has no long-lasting effect on graft outcome. This corroborates with recent findings in other patient cohorts and with the idea that T cell inflammation in renal allografts can be seen as a response to wounding [5, 35]. Moreover, it is interesting to note that the incidence of acute rejection is higher in the current study than previously reported (0.4%) [1]. This difference is most likely due to the earlier era in which the current study was performed, the important differences in the immunosuppressive regimen used, and the use of deceased donor kidneys in 99% of patients in the current study, versus 28% in the previously studied population [1].

The histological lesions in the current study were evaluated in biopsies for cause, per definition at time of graft dysfunction. Therefore, the contribution of donor graft pathology and of subclinical but progressive disease processes to the early chronic histological lesions is not evaluated in the current study. Moreover, not all patients included in the current study had biopsies performed, and the exclusive use of for-cause biopsies represents inherent selection bias. The current study results can thus not be extrapolated to kidneys with perfectly stable graft function. This is illustrated by the apparent discrepancy between our study, where we see a steep increase in chronic damage within the first year after transplantation, and previous protocol biopsy studies that showed that biopsies with only mild fibrosis at 1 year did not progress to histological changes at 5 years, in patients treated with modern immunosuppression [36]. The need for repeat indication biopsies within the first year after transplantation is a clear manifestation of a troubled kidney, which is often associated with rapidly progressive chronic damage. Conversely, in patients with stable graft function, without need for indication biopsies, renal prognosis is much better, and there is not always an ongoing injury process that culminates in progressive chronic damage. This is further exemplified by our finding that performing a for-cause biopsy is the most important risk factor for graft loss, irrespective of its intrinsic histological appearance.

However, given the highly significant impact of subtle chronic damage on graft outcome in the current study, it is very well plausible that chronic histological damage in protocol biopsies could be used as additional prognostic marker, as was also suggested previously [6, 13-15]. Nevertheless, the clinical use of renal allograft chronic histological lesions as prognostic tools for individual patients is inevitably hampered by sampling error issues: chronic histological injury could be overestimated by accidental sampling of scarred tissue, while minimal injury or patchy processes can be underdiagnosed, and thus lead to falsely optimistic prospects. None of these problems can be overcome, as sampling error is inherent to the use of needle biopsies as a clinical tool for follow-up of renal allograft recipients.

It has to be acknowledged that modern-type antibody testing and screening for polyomavirus nephropathy was not available in the current study cohort. The underlying disease process could thus have been missed in some patients, and the impact of antibody-mediated phenomena and polyomavirus nephropathy on the occurrence of chronic damage and on graft outcome could be underestimated in the analyses. Studies with current immunosuppressive regimens and present-day antibody and virus screening strategies are warranted to elucidate the true contribution of these diseases to progressive chronic histological damage, and to decipher the pathophysiology of chronic damage of “unknown etiology”. In addition, it should be noted that not all patients were treated with the same immunosuppressive regimen. This could be of importance, as it was previously shown that different immunosuppressants have different effects on renal histology [36]. Finally, the kidney grafts in the current study were almost exclusively from deceased donors, and donation after cardiac death was not performed at the time of inclusion. It will therefore be necessary to validate our findings in patient cohorts more representative for current clinical practice and treated with current immunosuppressive protocols.

Acknowledgments

The authors had full access to the data and take responsibility for its integrity. All authors have read and agree with the manuscript as written. We thank the centers of the Leuven Collaborative Group for Renal Transplantation, the clinicians and surgeons, nursing staff and the patients.

Disclosure

The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.

Ancillary