SEARCH

SEARCH BY CITATION

Keywords:

  • Classifier;
  • kidney transplant;
  • microarrays;
  • molecular diagnostics;
  • T cell–mediated rejection

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

We previously developed a microarray-based test for T cell-mediated rejection (TCMR) in a reference set of 403 biopsies. To determine the potential impact of this test in clinical practice, we undertook INTERCOM, a prospective international study of 300 indication biopsies from 264 patients (ClinicalTrials.gov NCT01299168). Biopsies from six centers—Baltimore, Barcelona, Edmonton, Hannover, Manchester and Minneapolis—were analyzed by microarrays, assigning TCMR scores by an algorithm developed in the reference set and comparing TCMR scores to local histology assessment. The TCMR score correlated with histologic TCMR lesions—tubulitis and interstitial infiltration. The accuracy for primary histologic diagnoses (0.87) was similar to the reference set (0.89). The TCMR scores reclassified 77/300 biopsies (26%): 16 histologic TCMR were molecularly non-TCMR; 15 histologic non-TCMR were molecularly TCMR, including 6 with polyoma virus nephropathy; and all 46 “borderline” biopsies were reclassified as TCMR (8) or non-TCMR (38). Like the reference set, discrepancies were primarily in situations where histology has known limitations, for example, in biopsies with scarring and inflammation/tubulitis potentially from other diseases. Neither the TCMR score nor histologic TCMR was associated with graft loss. Thus the molecular TCMR score has potential to add new insight, particularly in situations where histology is ambiguous or potentially misleading.


Abbreviations
ABMR

antibody-mediated rejection

DSA

donor-specific HLA antibody

ISD

immunosuppressive drugs

PRA

panel reactive HLA antibody

TCMR

T cell-mediated rejection.

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Despite advances in immunosuppression, T cell-mediated rejection (TCMR) in kidney transplants remains important in the differential diagnosis of graft dysfunction, as the end point of clinical trials of immunosuppressive drugs (ISDs), and as the prototype for T cell-mediated inflammatory diseases [1]. The biopsy diagnosis of TCMR is based on histologic lesions, guided by the recommendations of the Banff self-organizing group [2]. The lesions (interstitial infiltrate, tubulitis, intimal arteritis) were selected at the 1991 meeting based on opinions derived from the case mix at the time: relatively young donor kidneys with early severe TCMR, the prevalent disease phenotype. It was known that the interstitial inflammation and tubulitis lesions were non-specific and found in other conditions, for example, acute kidney injury (AKI) and progressive renal diseases, creating the potential for false positives [3-9], and difficult to assess in scarred tissue, creating false negatives [10]. Many biopsies are assigned the ambiguous “borderline” designation: In our recent study of 403 biopsies, 35 biopsies were diagnosed TCMR but 40 were called borderline [11]. Interobserver agreement is poor [12, 13], and certain diagnostic rules may be incorrect, for example the designation of all “isolated v-lesions” as TCMR [11, 14]. Central histology assessment is not the solution: It is simply a second opinion using the same standard. Indeed, in a study of independent central readings by three pathologists, when one diagnosed TCMR, another agreed in 45% of biopsies, and three agreed in only 16% [14].

Moreover, time has changed the biopsy case mix. Polyoma virus nephropathy (PVN), which had not been recognized in 1991, renders TCMR lesions of uncertain significance [15], and the 1991 criteria must now be interpreted in biopsies with more aging, injury and scarring. This “data drift”—a shift in disease prevalence and prior probabilities within the target population—is a common occurrence in epidemiological studies, and is to be expected in a diagnostic system in use for over two decades.

Gene expression in biopsies is emerging as a new diagnostic system for TCMR [14] and other diseases such as inflammatory bowel disease [16]. While many investigators have analyzed molecular changes in biopsies with histologic rejection, using either RT-PCR [17, 18] or microarrays [19-21], these studies failed to distinguish TCMR from antibody-mediated rejection (ABMR), in part because the criteria for diagnosing ABMR, particularly C4d-negative ABMR, are evolving and remain controversial [22, 23]. Because of this, we established the Alberta Transplant Applied Genomics Centre (ATAGC) Reference Standard histology system (http://atagc.med.ualberta.ca/) in a prospective cohort of 403 indication biopsies (the BFC403 cohort). By setting out criteria for C4d-negative ABMR, the ATAGC Reference Standard classification permits the first rigorous separation of TCMR from ABMR [14, 24]. This allowed us to develop algorithms to assign TCMR scores that reflect the probability of TCMR in each biopsy. The TCMR scores correlated with the histologic TCMR lesions and diagnoses, and differentiated borderline biopsies into those with and without TCMR. The accuracy of the TCMR score for the histologic diagnosis was 89%, with most discrepancies between the TCMR score and histologic TCMR occurring in situations known to be challenging for histology, such as when the tissue is inflamed due to AKI, damaged by renal disease, or has extensive scarring. The TCMR score confirmed previous concerns [25, 26] about diagnosing TCMR on the basis of intimal arteritis (isolated v-lesions): 19/24 biopsies with isolated v-lesions had low negative TCMR scores. We concluded that many biopsies with low TCMR scores and histologic TCMR were histology false positives due to inflammation induced by AKI or renal diseases, and that some biopsies with high TCMR scores and histologic assessment indicating no TCMR were histology false negatives, that is, TCMR obscured by scarring or complicated by PVN, which can make pathologists reluctant to diagnose TCMR [14].

The present prospective multicenter INTERCOM study explored the potential impact of the TCMR score by comparing it to local diagnoses in six established programs in five countries. Using microarray measurements and an algorithm developed in the earlier BFC403 reference set [14], we assigned TCMR scores to 300 new indication biopsies (INT300). Our goals were to examine the histology-molecular relationships in INT300, and to compare them to the histology-molecular relationships in the BFC403 reference set. The hypothesis arising from the BFC403 reference set was that the TCMR score would predict histologic TCMR defined by these highly experienced local centers; that the relationship between the TCMR score and conventional assessment (e.g. accuracy) in INT300 would be similar to the reference set; and that the discrepancies would be concentrated in situations where histology was potentially misleading, as was the case in the reference set. Thus the INTERCOM study should provide an estimate of the potential error rate in histology assessment in experienced transplant centers.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Patient population, specimens and data collection

INTERCOM was an international study involving six established centers. The design was a prospective, open, uncontrolled, nonrandomized and observational study (ClinicalTrials.gov NCT01299168). The centers provided 320 consecutive biopsies over 18 months. Twenty were excluded as unsuitable for microarray testing due to degradation and/or low yield, leaving 300 biopsies from 264 patients suitable for analysis. This study was approved by all institutional review boards (University of Alberta, Edmonton; University of Minnesota, Minneapolis; University of Maryland, Baltimore; Hospital Vall d'Hebron, Barcelona; Hannover Medical School and Manchester Royal Infirmary). The inclusion criterion was any kidney transplant recipient ≥18 years of age undergoing a kidney biopsy for clinical indications and able and willing to give informed consent. However, the centers submitted two biopsies from individuals under age 18 (ages 10 and 17), and these were processed in the study.

In addition to the 300 biopsies in this study, six normal kidney specimens from native nephrectomies for renal carcinoma were used as baseline controls. These samples were not included in any analyses.

Data collection

Data collected included patient clinical information, data related to time of biopsy and local human leukocyte antigen (HLA) antibody screen and specificity testing. All data were stored in a Scientific Data Management System, supported by Genologics, Inc., and operated behind high level firewall systems of the University of Alberta.

Classification of the local center pathology reports

The renal biopsy histologic assessment (lesions and diagnoses) was performed by the local pathologists, and was provided to the ATAGC in the form of a report. To compare the local assessment with our previous publications, the local diagnoses were assembled into categories for analysis by a pathologist associated with ATAGC, Dr. M. Mengel, without knowledge of the molecular assessments (Table S1).

Data analysis

The study required one additional needle biopsy core beyond the usual two to three biopsy cores for local diagnosis by histology. The core for the study was immediately stabilized in RNAlater® to prevent mRNA degradation and stored at −4°C for 24 h then at −20°C until processing. Samples from distant centers were shipped in batches to the ATAGC on dry ice. One 5-mL serum sample was collected for HLA antibody testing locally, at the time of the biopsy. The INT300 microarray results were normalized using Robust Multiarray Averaging as implemented in the Bioconductor “RefPlus” package. BFC403 was normalized as a separate batch as described previously [14], also using RefPlus. Because Affymetrix changed their labeling kit after the BFC403 biopsies were processed, the BFC403 expression values were adjusted for this batch effect using the Ratio-G method [27] after normalization. All analyses and graphics were done using the “R” software package, version 2.12.1 (64-bit), with various libraries from Bioconductor 2.8.

Microarray expression files for BFC403 (GSE36059) and INT300 (GSE48581) are posted on the Gene Expression Omnibus website.

To obtain measures of association between the TCMR scores and histologic lesions, gamma rank correlations [28] were used, because of the large number of “tied” values.

Molecular measurements

TCMR scores were assigned using a classifier algorithm derived in the BFC403 reference set [14]. Since INT300 is an independent test set, the full BFC403 dataset was used to train the classifier, rather than using repeated 10-fold cross-validation as in the original paper. The classifier output is a score between 0.0 and 1.0, reflecting the probability that a biopsy has TCMR.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

The demographics of the INT300 population are presented in Table S2, compared with the BFC403 reference set [14]. The case mix in INT300 was similar to BFC403, as expected, since both are unselected indication biopsy populations. The median follow-up time of working grafts in the more recent INT300 (384 days) is shorter than in BFC403 (1135 days). The distribution of local diagnoses in INT300 was similar to the BFC403 reference set (Table 1). The category of “possible ABMR” used in BFC403 did not prove useful, and was eliminated in INTERCOM, with such biopsies assigned to other relevant categories.

Table 1. Histologic diagnoses
Histology-DSA diagnosisINTERCOM (INT300)1: interpreted from local biopsy reportsReference set (BFC403): reference standard classification
Early (<1 year)Late (>1 year)All (% of total)Early (<1 year)Late (>1 year)All (% of total)
  • 1

    The local assessment of ABMR and suspicious for ABMR was interpreted from the pathology reports, and included such terms as probable, suspicious for and rule out.

  • 2

    Others (n = 8) included “acute tubular necrosis” (n = 1), C4d staining with no pathology (n = 1), histological diagnosis not defined (n = 1), pyelonephritis (n = 1) and diabetic nephropathy (n = 4).

TCMR-related
TCMR26632 (11)27835 (9)
Mixed TCMR and ABMR156 (2)22022 (5)
Borderline242246 (15)291342 (10)
ABMR-related
ABMR (C4d+ and C4d−)33740 (13)65965 (16)
Transplant glomerulopathy21820 (7)134 (1)
Polyoma virus nephropathy (PVN)9413 (4)10212 (3)
Glomerulonephritis33740 (40)63541 (10)
Acute kidney injury14014 (5)50050 (12)
No major abnormalities241943 (14)344276 (19)
Atrophy-fibrosis83038 (13)63440 (10)
Others2682 (3)11516 (4)
Total116184300 (100)182221403 (100)

Initially, it was intended to include central review of the histology as an add-on but this proved to be impossible: Some centers were unwilling to submit their slides, and there was no way to resolve disagreements between the local center and the central reviewer.

By histology, 32 biopsies had TCMR and 6 had mixed rejection, that is, histologic criteria for both ABMR and TCMR. Biopsies with no rejection, borderline changes, or specific diseases were classified as AKI (n = 14) if they were before 42 days posttransplant. Those more than 42 days posttransplant were designated as either “no major abnormalities” (NOMOA) if they had minimal interstitial fibrosis (ci < 2, n = 43), or “atrophy-fibrosis of unknown significance” (IFTA) if there was scarring (ci > 1; n = 38). PVN (n = 13) was diagnosed by local standard of care (blood or urine polymerase chain reaction [PCR], immunohistochemistry, in situ hybridization and/or electron microscopy).

The majority of biopsies were late (>1 year posttransplant) in INT300 (61%), similar to the BFC403 reference set (55%). TCMR was diagnosed in 9% of BFC403 and 11% of INT300, mostly early; borderline in 10% of BFC403 and 15% of INT300 and mixed rejection in 5% of BFC403 biopsies and 2% of INT300 biopsies.

Relating TCMR scores to histologic lesions

TCMR scores in INT300 biopsies correlated with interstitial inflammation (r = 0.79), tubulitis (r = 0.83) and intimal arteritis (r = 0.85) (Table 2). The correlations in INT300 were similar to those in the BFC403 reference set [14]. There was a weak correlation with peritubular capillaritis (r = 0.28), which sometimes occurs in TCMR as well as in ABMR [29]. There were negative correlations with glomerular double contours, an ABMR lesion; with arterial fibrous intimal thickening and arteriolar hyalinosis; and with time posttransplant. The TCMR score was not associated with DSA positivity.

Table 2. Gamma rank correlations between TCMR scores1 and histologic lesions in the 300 INTERCOM biopsies and the reference set
Histologic lesionsCorrelation
INTERCOM (INT300)Reference set (previously published)
  • 1

    For purposes of analysis, the TCMR scores were split into high (>0.1) vs. low (≤0.1).

  • *

    p < 0.05.

  • **

    p < 0.01.

  • ***

    p < 0.001.

TCMR lesions
Interstitial inflammation (i)0.79***0.87***
Tubulitis (t)0.83***0.90***
TCMR and ABMR lesions
Intimal arteritis (v)0.85*0.76**
ABMR lesions
Peritubular capillaritis (ptc)0.280.36*
Glomerulitis (g)−0.200.09
Glomerular double contours (cg)−0.58***−0.62***
Atrophy-scarring related
Interstitial fibrosis (ci)−0.05−0.04
Tubular atrophy (ct)−0.060.11
Arterial fibrous intimal thickening (cv)−0.29*−0.20
Arteriolar hyalinosis (ah)−0.65***−0.55***
Time posttransplant: early (<1 year) vs. late (>1 year)−0.31*−0.51**

Agreement between the TCMR score and histology diagnosis

TCMR scores were assigned by the classifier generated in the BFC403 reference set (Figure 1). For comparison with local histology assessments, TCMR scores were divided into high or low using the same cutoff of 0.1 as in the BFC403 reference set.

image

Figure 1. Relationship between the TCMR score and the diagnoses based on local center assessment in INT300. The order within each diagnostic category is random. The horizontal line shows our threshold of 0.1 for defining high versus low TCMR scores. The different symbols represent time posttransplantation: early (<1 year: empty triangles) and late (>1 year: solid triangles). ABMR, antibody-mediated rejection; TG, transplant glomerulopathy; M, mixed; TCMR, T cell-mediated rejection; Bord., borderline; PVN, polyoma virus nephropathy; GN, glomerulonephritis; AKI, acute kidney injury; NOMOA, no major abnormalities; Oth., others; Neph., nephrectomies.

Download figure to PowerPoint

We studied the relationship between TCMR scores and histologic diagnoses of TCMR in INT300, and compared this to the histology-molecular relationship in the BFC403 reference set (Figure 2). The area under the receiver operating characteristic curve (AUC) of the TCMR score for a primary histologic diagnosis of TCMR or mixed was 0.85, similar to that in BFC403, with borderline considered not TCMR. When borderline biopsies were excluded, the AUCs rose to 0.86 in INT300 and 0.85 in BFC403 (Table 3). Thus the relationship between the TCMR and the conventional assessment in INT300 was virtually identical to that in BFC403.

image

Figure 2. Prediction of TCMR and mixed histologic diagnosis by TCMR score. Receiver operating characteristic (ROC) curve and area under curve (AUC) for the ability of the TCMR score to predict the local primary diagnosis of TCMR or mixed rejection in INT300 (n = 300) and in BFC403 (n = 403).

Download figure to PowerPoint

Table 3. Agreement of the TCMR score with histologic diagnosis
TCMR scorePrimary histologic diagnosisTotal
TCMRDetailed diagnosis in non-TCMR biopsies
TCMRMixedTCMR and mixedABMR (C4d+ and C4d−)TGPVNGNAcute kidney injuryNo major abnormalitiesAtrophy fibrosisOtherAll non-TCMR except borderlineBorderline
>0.12112250611012015845
<0.1115163520739144236820138255
Total3263840201340144338821646300
 Prediction of the local diagnosis of TCMR or mixed rejection by the TCMR score
 nAccuracySensitivitySpecificityPPVNPVAUC
  • Accuracy, sensitivity, specificity, PPV and NPV were calculated based on a TCMR score cutoff of 0.1.

  • PPV, positive predictive value; NPV, negative predictive value; AUC, area under the receiver operating characteristic curve.

  • 1

    One biopsy diagnosed as PVN had a secondary diagnosis of TCMR recorded in the local center, and is considered to be non-TCMR for the purposes of calculating accuracy, etc.

TCMR score in all biopsies
INT3003000.870.580.910.490.940.85
BFC4034030.890.510.950.620.920.84
TCMR score after leaving out borderline biopsies
INT3003000.880.580.930.590.920.86
BFC4034030.890.510.960.720.910.85

High TCMR scores in INT300 biopsies were mostly in biopsies with histology diagnoses of TCMR or borderline: of 45 biopsies with high TCMR scores, 22 had histology diagnoses of TCMR (n = 21) or mixed (n = 1) and 8 were classified as borderline by histology (Table 3). Most biopsies with histologic TCMR or mixed rejection had elevated TCMR scores (22/38), whereas biopsies with IFTA or AKI or other histologic diagnoses (e.g. ABMR) did not. Of 46 histologic borderline biopsies, 38 were TCMR score negative, confirming previous conclusions [11].

In summary, compared to histology, the TCMR score reclassified 77/300 biopsies (26%): 15 histologic TCMR negative biopsies were TCMR score positive; 16 biopsies with histologic TCMR (including five mixed) were TCMR score negative and the 46 borderline biopsies were reclassified as TCMR score positive (n = 8) or negative (n = 38).

Discrepancy analysis

Discrepancies between the TCMR score and the primary diagnosis were of two types: score positive/histology negative and score negative/histology positive (Figure 3). The distribution of the two classes of discrepancies over time was different, both from each other and from that of the biopsies where both systems agreed TCMR was present.

image

Figure 3. Smoothed frequency distribution (conditional probability) for TCMR/mixed over time. Score − Histology+: TCMR score negative and TCMR or mixed histologic diagnosis positive; Score + Histology+: TCMR score positive and TCMR or mixed histologic diagnosis positive; Score + Histology−: TCMR score positive and TCMR or mixed histologic diagnosis negative. Comparison of Score − Histology+ and Score + Histology− over time: Chi-square p-value <0.05.

Download figure to PowerPoint

The 15 score positive/histology negative biopsies were predominantly late—12 were beyond 1 year—and none was earlier than 6 weeks posttransplant (Figure 3). However, despite having other primary diagnoses, 9 of the 15 (Table 4) had secondary diagnoses or text descriptions of lesions meeting the criteria for TCMR or borderline changes, and thus represent under-reporting of TCMR-like changes, not discrepancies (including six with PVN as discussed below). Of the remaining six, three were late with severe scarring (ci3); one had a TCMR score of 0.11 and may represent a false positive TCMR score at the cutoff used (0.10) and two were unexplained.

Table 4. Score positive–histology negative: TCMR score positive biopsies with primary histology diagnosis other than TCMR or mixed or borderline (n = 15)1
Biopsy identificationHistologic diagnosisGFRBorderline secondary diagnosis presentTCMR secondary diagnosis presentBorderline lesions present2TCMR lesions present3Fibrosis ci > 1Proposed interpretation, based on categories proposed in reference set
  • tNA, histologic t-score not available; PVN, polyoma virus nephropathy.

  • 1

    Eight biopsies with primary diagnosis of “borderline” were excluded.

  • 2

    Borderline lesions: v = 0 and t > 0 and i < 2, or v = 0 and t = 1 and i > 1.

  • 3

    TCMR lesions: i ≥ 2 and t ≥ 2.

48C4d + ABMR    XTCMR obscured by scarring (ci = 3)
220C4d + ABMR42     Unexplained
58C4d − ABMR42X X XSecondary diagnosis of “borderline”
103C4d − ABMR52   X Meets criteria for “TCMR”
178C4d − ABMR23X X XSecondary diagnosis of “borderline”
5GN31X X  Secondary diagnosis of “borderline”
52IFTA    XTCMR obscured by scarring (ci = 3)
168IFTA43    XTCMR obscured by scarring (ci = 3)
7NOMOA31     Unexplained
113PVN19   XXMeets criteria for “TCMR”
32PVN31 X X Meets criteria for “TCMR”
205PVN18   X Meets criteria for “TCMR”
249PVN   i3, tNA Meets i-score criteria for “TCMR”
67PVN<20   XXMeets criteria for “TCMR”
217PVN25  X  Meets criteria for “borderline”

Most of the 16 score negative/histology positive biopsies were either very early (six <42 days) or late (six >1 year) (Table 5; Figure 3). Three of the six early biopsies may reflect interstitial inflammation and tubulitis induced by AKI, as described in protocol biopsies [30], and three were diagnosed exclusively on the basis of the questionable isolated v-lesion criterion, raising the possibility that endothelial injury at the time of transplantation can produce isolated v-lesions. It is possible that at least some of the other nine reflect tubulitis and interstitial inflammation due to parenchymal damage induced by diseases such as ABMR, since these features are common in primary renal diseases [6-9]. One was pretreated before the biopsy, which may eliminate molecular changes despite persistent histologic lesions [26].

Table 5. Score negative–histology positive: histology diagnosis of TCMR with TCMR score <0.1 (n = 16)
Biopsy IDHistologic diagnosisTime from transplantGFRTreatmentIndicationIsolated v-lesions present1TCMR lesions present2Suggested interpretation, based on categories proposed in reference set
  • Dx, diagnosis; DRF, deterioration of renal function; SIRF, stable but impaired renal function; Neg, negative; Pos, positive; Ser, serum; GFR, glomerular filtration rate; AKI, acute kidney injury.

  • 1

    Isolated v-lesions: i < 2 or t < 2 and v > 0.

  • 2

    TCMR lesions: i ≥ 2 and t ≥ 2.

42TCMR1646NoDRFX Histology false positive due to isolated v-lesions
158TCMR1060NASIRF XHistology false positive: inflammation due to AKI
256TCMR24740NoDRF XHistology false positive: persistent parenchymal damage in old donor kidney (donor age 75)
302TCMR3343NoDRF XHistology false positive: inflammation due to AKI
76TCMR16027NoDRF XHistology false positive: persistent parenchymal damage in old donor kidney (donor age 72)
125TCMR20987NoInvestigate proteinuria XUnexplained
240TCMR101103NoDRF XUnexplained; history of DGF
224TCMR106370NoDRF XUnexplained
151TCMR14NAYesDGFX Histology false positive due to isolated v-lesions; pretreated
81TCMR517NoDGF XHistology false positive: inflammation due to AKI
121TCMR50077YesUnknown BorderlinePretreated
37Mixed101172NoDRF XUnexplained
137Mixed324262NoInvestigate proteinuria XHistology false positive: inflammation/tubulitis due to severe parenchymal damage from ABMR
127Mixed2821NANoDRF XHistology false positive: inflammation/tubulitis due to severe parenchymal damage from ABMR
282Mixed248241NADRF XHistology false positive: inflammation/tubulitis due to severe parenchymal damage from ABMR
294Mixed2628YesDRFX Histology false positive due to isolated v-lesions; pretreated

Thus the discrepancies confirmed the predictions of the BFC403 study that the disagreement between the TCMR score and histology is not random but is concentrated in situations where histology has known potential for false negatives and false positives.

The number of times the pathologists indicated biopsy inadequacies with respect to TCMR features was as follows: 8 of the 300 lacked t, 7 lacked i and 21 lacked v. These included some biopsies with multiple missing lesions, and there were 25 biopsies in total with missing i/t/v data. Of these 25 biopsies (all called non-TCMR by histology), only 1 had a TCMR score >0.1, so their contribution to the number of discrepancies was minimal.

Polyoma virus nephropathy

The six biopsies diagnosed locally as PVN with high TCMR scores were between 4 and 12 months (Table 6) and all had histologic criteria for TCMR or borderline. One actually had a secondary histologic diagnosis of TCMR, and five had TCMR lesions recorded but not reported as TCMR, reflecting the uncertainty over the significance of these lesions in PVN [31]. Two other PVN biopsies with inflammation (i- or t-lesions) had very low TCMR scores, confirming the suspicion that injury from PVN can induce lesions that mimic TCMR [31].

Table 6. TCMR scores and TCMR histologic lesions in biopsies diagnosed as PVN nephropathy in INT300
Biopsy identificationSecondary diagnosisTCMR scoreitvgptccitictcvcgahmm
  1. i, Banff i; mm, mesangial matrix expansion; ti, total i.

32TCMR0.99331
2050.99330001100
2490.99300110
670.99230003331010
2170.9621001110000
113GN0.91330003331000
400.10230001211000
164IFTA0.0922000321012
1790.01130003232001
2180.010000011000
187IFTA<0.0012100011000
118<0.00100000110000
143<0.001000002020000

Relationship of TCMR to graft survival

Patients with TCMR (TCMR score or histology) in any biopsy were compared to all other patients. Survival in the TCMR group was calculated as the time after the first biopsy showing TCMR. There was no association of the TCMR score or histologic TCMR with graft survival in either INT300 or BFC403 (Table 7).

Table 7. Univariate analysis of the effect of TCMR on death-censored graft survival using a Cox proportional hazards regression model
VariableFailures as a fraction of total patients with TCMR1Hazard ratio for TCMR in the whole population (95% confidence interval)P-value
  • Survival in the TCMR group was calculated as the time after the first biopsy showing TCMR.

  • 1

    Patients with TCMR (TCMR score or histology) in any biopsy were compared to all other patients.

INT300 (n = 264 patients with 33 failures)
Positive TCMR score8/431.79 (0.80–4)0.15
Histologic diagnosis of TCMR or mixed3/350.57 (0.17–1.90)0.36
BFC403 (n = 315 patients with 80 failures)
Positive TCMR score11/421.1 (0.58–2.07)0.76
Histologic diagnosis of TCMR or mixed11/460.83 (0.44–1.58)0.58

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

To assess the potential clinical impact of molecular diagnosis of TCMR, this prospective multicenter study generated microarray-based TCMR scores in a population of 300 recent indication biopsies of kidney transplants (INT300), using an algorithm developed in the BFC403 reference set. We assessed the ability of the TCMR score to predict local histologic standard-of-care diagnoses, established the relationship between the molecular and histologic variables in INT300, and compared these findings to the BFC403 reference set. The case mix in INT300 was similar to that in the reference set. The scores from the previously generated TCMR classifier correlated with histologic TCMR lesions and TCMR diagnoses by the local centers, in a pattern similar to the reference set in terms of accuracy and AUC, with discrepancies in the same situations. The TCMR score changed the assessment of 26% of biopsies, particularly in situations known to be challenging for histology such as PVN, scarring, and when the tissue had AKI or had been damaged by renal diseases, as predicted by the BFC403 study. Neither the TCMR score nor histologic TCMR was associated with increased risk of failure in INT300, consistent with the findings in BFC403. The results establish the feasibility and utility of central microarray-based molecular measurements to estimate disease states in transplant biopsies, and support the potential of molecular testing of biopsies used in conjunction with histology to increase our understanding of these diseases.

As the first multicenter study of a previously developed biopsy diagnostic system in transplantation, INTERCOM had to invent a design to compare the TCMR score to the actual diagnoses applied in leading transplant centers. We followed the lead of the ISD trials, in which the primary endpoint was local assessment. We did not prescribe what experienced centers “should” be doing because there was no evidence to indicate what choices would be more correct than local opinion in experienced centers. Histology diagnosis is an expert opinion-based discipline, in which the guidelines of the self-organizing Banff group and other groups are useful but their application is voluntary, and the actual degree of application in experienced transplant centers is not known. For example, we had previously discovered that some pathologists who nominally subscribe to the Banff published guidelines actually use personal variants that significantly affect the diagnosis of TCMR [14].

The TCMR score offers an opportunity to improve our understanding of PVN, where histology struggles to distinguish TCMR from immune reactions to viral products and inflammatory reactions to virus-induced epithelial injury. While one can speculate that the T cell response to virus proteins could give positive TCMR scores, this is in our opinion unlikely, since in most cases of PVN in the INT300 and in BFC403 the TCMR score was negative, even in some cases where the biopsy was inflamed. Our results argue that PVN biopsies with high TCMR scores actually have TCMR, perhaps triggered by ISD reduction. This is supported by our finding that some biopsies in both BFC403 and INT300 were inflamed but did not have positive TCMR scores. The presence of TCMR lesions but no TCMR diagnosis in some PNV biopsies may reflect reluctance of the pathologists to diagnose TCMR when other diseases are present, given the known nonspecificity of these lesions in AKI and in primary kidney diseases. Resolution of this issue must await further prospective trials, and ultimately criteria that truly distinguish allorecognition from immune recognition of viral products.

Although TCMR had no impact on subsequent graft survival, whether diagnosed by the TCMR score or by histology, further study is needed to define the outcomes for kidneys with TCMR score positive biopsies currently being called negative by histology. Late “rejection” is a known risk for graft loss [32], but these studies had no way of distinguishing TCMR from ABMR reliably. Late ABMR carries a very poor prognosis [23], but late TCMR may also be significant, for example as a signal that the immunosuppressive regimen is inadequate, perhaps reflecting nonadherence [33] or “minimization” of ISD doses. The present findings raise the possibility that the TCMR component of these complex late biopsies may often be missed by histology, and the issue must now be revisited.

While it is not possible to explain the discrepancies with certainty, the present results confirm the earlier conclusion that discrepancies are not random but are in situations where histology is known to be limited, including those in which TCMR lesions are difficult to assess (severe scarring) or when TCMR lesions may be induced by other diseases. The relationship of histology diagnosis to the TCMR score (AUC) was similar to that in the BFC403 reference set. As predicted [14], score positive–histology negative biopsies often occurred in PVN and in late biopsies with scarring. Score negative–histology positive biopsies often occurred in kidneys that had other potential causes of parenchymal damage, including early AKI, persistent parenchymal dysfunction in old donor kidneys, or active nephron damage in ABMR. Since tubulitis and interstitial inflammation are well known in primary renal diseases, their presence in some kidney transplants due to processes other than TCMR is expected. This would explain why the finding of these lesions in protocol biopsies fails to predict progression [30] and or benefit from treatment [34].

The INTERCOM study provides an estimate of the potential error rate in histologic diagnosis of TCMR in leadership centers using the international standard-of-care, and offers a solution to reducing these errors through new prospective studies focusing on the scenarios where the discrepancies occurred. The TCMR score offers new insight in biopsies called “borderline”, an ambiguous category that is a weakness of the consensus guidelines [3] and has the potential to cause both over- and under-treatment. The TCMR score will also be welcome in kidneys with scarring that can obscure TCMR lesions. Moreover, in AKI and in progressive parenchymal diseases such as ABMR or GN, where nephron damage can induce tubulitis and interstitial inflammation that mimics TCMR lesions, the TCMR score can potentially identify which cases actually need treatment of TCMR.

With microarray assessment of biopsies, renal transplantation joins other areas of medicine moving toward precision diagnostic systems as standard of care for serious illnesses [35], and is a step in the direction of evidence based practice, moving from expert opinion to formal prospective studies (for levels of evidence see the US Preventive Services Task Force [http://www.uspreventiveservicestaskforce.org/]). Molecular assessment of cancer tissue is proving useful in tests such as Oncotype DX [36, 37] and Mammaprint [38, 39], but application of molecular testing to noncancer biopsies with inflammatory diseases is only beginning. For example, we are proceeding to assess the TCMR score in other organ transplants such as heart and lung, where histology assessment is even more problematic than in kidney transplants, and in nontransplant biopsies diseases such inflammatory bowel disease [40]. The TCMR score was reliable in operation, providing answers in all 300 biopsies that had been stabilized properly for analysis, with a turnaround of 30 h after receiving the sample. Moreover, this “first generation” TCMR score has considerable potential for refinement, for example, by incorporating other analytical approaches such as ensemble methods [41].

The low sensitivity and positive predictive value of the TCMR score for a histology diagnosis of TCMR suggest that we temper our interpretations until further experience is available. However, these values are similar to the sensitivity and positive predictive values (average 0.45) between expert pathologists [14]. This reflects the problem of “reference standard-related bias”: a new test, whether based on a molecular classifier or the opinion of a second pathologist, will not agree perfectly with the existing gold standard (the opinion of a different pathologist) when there is a high degree of subjectivity involved in the assignment of the gold standard. “Noise” within histology is intrinsic, due to the poor kappa values for TCMR lesions, for example, tubulitis 0.17, where complete agreement is 1.0 and random is 0.0 [12, 13]. For the present, the clearest value of the TCMR score will be in problematic cases, and its limited ability to provide information about diagnoses such as recurrent diseases must be acknowledged.

While the present study was designed to compare the TCMR score to the standard of care without distorting what the best centers are actually doing, the study could have been improved by imposing a supplementary research checklist for reporting, which we now recommend for future studies. Our design was to reflect actual reporting, following the principle of “intention-to-treat” analysis established in multicenter drug trials. We hope that our design offers lessons for future studies in which new tests are compared to classic practices, but always ensuring that such additions do not require deviation from the standard of care in these centers.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

We are grateful to Dr. M. Mengel for reviewing the biopsy reports from the centers to help categorize them. We would like to thank Richard Ugarte at the University of Maryland for help with organizing samples and clinical data. This research has been supported by funding and/or resources from Novartis Pharma AG, and in the past by Genome Canada, the University of Alberta Hospital Foundation, Roche Molecular Systems, Hoffmann-La Roche Canada Ltd., the Alberta Ministry of Advanced Education and Technology, the Roche Organ Transplant Research Foundation and Astellas. Dr. Halloran held a Canada Research Chair in Transplant Immunology until 2008 and currently holds the Muttart Chair in Clinical Immunology.

Disclosure

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

P.F. Halloran holds shares in Transcriptome Sciences, Inc., a company with an interest in molecular diagnostics. The other authors have no competing financial interests.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. Disclosure
  9. References
  10. Supporting Information

Additional Supporting Information may be found in the online version of this article at the publisher's web-site.

FilenameFormatSizeDescription
ajt12387-sm-0001-suppTab.pdf99K

Table S1: Summary diagnosis rules

Table S2: Demographics

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.