Received: 3 January 2020 | Revised: 24 February 2020 | Accepted: 10 March 2020 DOI: 10.1111/ajt.15898 MEETING REPORT The Banff 2019 Kidney Meeting Report (I): Updates on and clarification of criteria for T cell– and antibody-mediated rejection Alexandre Loupy1* | Mark Haas2* | Candice Roufosse3 | Maarten Naesens4,5 | Benjamin Adam6 | Marjan Afrouzian7 | Enver Akalin8 | Nada Alachkar9 | Serena Bagnasco10 | Jan U. Becker11 | Lynn D. Cornell12 | Marian C. Clahsen-van Groningen13 | Anthony J. Demetris14 | 15 1 Duska Dragun | Jean-Paul Duong van Huyen | Alton B. Farris16 | 17 18 19 Agnes B. Fogo | Ian W. Gibson | Denis Glotz | Juliette Gueguen | Zeljko Kikic20 | 1 Nicolas Kozakowski20 | Edward Kraus9 | Carmen Lefaucheur19 | Helen Liapis21 | Roslyn B. Mannon22 | Robert A. Montgomery23 | Brian J. Nankivell24 | 25 26 1 Volker Nickeleit | Peter Nickerson | Marion Rabant | Lorraine Racusen10 | Parmjeet Randhawa14 | Blaise Robin1 | Ivy A. Rosales27 | Ruth Sapir-Pichhadze28 | Carrie A. Schinstock29 | Daniel Seron30 | Harsharan K. Singh25 | Rex N. Smith27 | Mark D. Stegall31 | Adriana Zeevi14 | Kim Solez6 | Robert B. Colvin27 | Michael Mengel6 1 Université de Paris, INSERM U970, University of Paris, Paris Translational Research Centre for Organ Transplantation, Paris and Necker Hospital, Assistance Publique - Hopitaux de Paris, Paris, France 2 Department of Pathology and Laboratory Medicine, Cedars-Sinai Medical Center, Los Angeles, California 3 Department of Immunology and Inflammation, Imperial College London and North West London Pathology, London, UK 4 Department of Microbiology and Immunology, University of Leuven, Leuven, Belgium 5 Department of Nephrology, University Hospitals Leuven, Leuven, Belgium 6 Department of Laboratory Medicine and Pathology, University of Alberta, Edmonton, Canada 7 Department of Pathology, University of Texas Medical Branch, Galveston, Texas 8 Montefiore-Einstein Center for Transplantation, Montefiore Medical Center, Bronx, New York 9 Department of Medicine, Division of Nephrology, Johns Hopkins University School of Medicine, Baltimore, Maryland 10 Department of Pathology, Johns Hopkins University School of Medicine, Baltimore, Maryland 11 Institute of Pathology, University Hospital of Cologne, Cologne, Germany 12 Division of Anatomic Pathology, Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, Minnesota 13 Department of Pathology, Erasmus MC, Rotterdam, the Netherlands 14 Department of Pathology, The Thomas E Starzl Transplantation Institute, University of Pittsburgh, University of Pittsburgh, Pittsburgh, Pennsylvania 15 Department of Nephrology and Critical Care Medicine, Berlin Institute of Health, Charite Universitätsmedizin, Berlin, Germany 16 Emory University, Atlanta, Georgia 17 Departments of Pathology, Microbiology, and Immunology; Medicine and Pediatrics, Vanderbilt University, Nashville, Tennessee 18 Department of Pathology, University of Manitoba, Winnipeg, Canada * These authors contributed equally to this report. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2020 The Authors. American Journal of Transplantation published by Wiley Periodicals LLC on behalf of The American Society of Transplantation and the American Society of Transplant Surgeons 2318 |  amjtransplant.com Am J Transplant. 2020;20:2318–2331. LOUPY et al. | 2319 19 Paris Translational Research Center for Organ Transplantation, Department of Nephrology and Transplantation, Hopital Saint Louis, Université Paris VII, INSERM U 1160, Paris, France 20 Departments of Medicine and Pathology, Medical University of Vienna, Vienna, Austria 21 Department of Pathology & Immunology & Internal Medicine (Renal) Retired, Washington University School of Medicine, Saint Louis, Missouri 22 Division of Nephrology, Department of Medicine, University of Alabama School of Medicine, Birmingham, Alabama 23 New York University Langone Medical Center, NYU Langone Transplant Institute, New York, New York 24 Department of Renal Medicine, Westmead Hospital, Sydney, Australia 25 Division of Nephropathology, Department of Pathology and Laboratory Medicine, The University of North Carolina, Chapel Hill, North Carolina 26 Department of Internal Medicine and Immunology, University of Manitoba, Winnipeg, Canada 27 Department of Pathology, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts 28 Division of Experimental Medicine, Department of Medicine, McGill University, Montreal, Canada 29 William J. von Liebig Center for Transplantation and Clinical Regeneration, Mayo Clinic, Rochester, Minnesota 30 Nephrology Department, Hospital Vall d’Hebron, Autonomous University of Barcelona, Barcelona, Spain 31 Departments of Surgery and Immunology, Mayo Clinic, Rochester, Minnesota Correspondence Alexandre Loupy The XV. Banff conference for allograft pathology was held in conjunction with the Email: alexandre.loupy@inserm.fr annual meeting of the American Society for Histocompatibility and Immunogenetics Mark Haas in Pittsburgh, PA (USA) and focused on refining recent updates to the classification, Email: mark.haas@cshs.org advances from the Banff working groups, and standardization of molecular diagnos- Funding information tics. This report on kidney transplant pathology details clarifications and refinements CareDx; CSL Behring; Elsevier; Eppendorf; to the criteria for chronic active (CA) T cell–mediated rejection (TCMR), borderline, GenDx; Hansa Biopharma; Histogenetics; Immucor; Omion; OneLambda; NanoString; and antibody-mediated rejection (ABMR). The main focus of kidney sessions was Novartis; Takeda; Veloxis; Vitaeris on how to address biopsies meeting criteria for CA TCMR plus borderline or acute TCMR. Recent studies on the clinical impact of borderline infiltrates were also pre- sented to clarify whether the threshold for interstitial inflammation in diagnosis of borderline should be i0 or i1. Sessions on ABMR focused on biopsies showing micro- vascular inflammation in the absence of C4d staining or detectable donor-specific antibodies; the potential value of molecular diagnostics in such cases and recommen- dations for use of the latter in the setting of solid organ transplantation are presented in the accompanying meeting report. Finally, several speakers discussed the capabili- ties of artificial intelligence and the potential for use of machine learning algorithms in diagnosis and personalized therapeutics in solid organ transplantation. KEYWORDS classification systems: Banff classification, clinical decision-making, clinical research/practice, kidney (allograft) function/dysfunction, kidney transplantation/nephrology, molecular biology: mRNA/mRNA expression, pathology/histopathology, rejection, translational research/science 1 | I NTRO D U C TI O N (TCMR) to harmonize the pathologic diagnosis and consequent therapeutic strategies. In addition, consensus of the diagnosis of The XV. Banff Conference for Allograft Pathology was held borderline (suspicious) for acute TCMR was reached. The current lit- September 23-27, 2019, in Pittsburgh, PA (USA), in conjunction with erature on biomarkers and molecular transplant diagnostics was also the Annual Meeting of the American Society for Histocompatibility reviewed, and recommendations of the Banff Molecular Diagnostics and Immunogenetics (ASHI). A total of 1253 delegates from 31 Working Group regarding clinical validation and adoption of the countries attended the conference, including pathologists, immunol- latter into the Banff classification will be presented in a separate ogists, physicians, surgeons, and immunogeneticists as well as rep- meeting report. Continuing a theme from the Banff 2015 meeting,1 resentatives from industry. The focus of kidney sessions at the 2019 results of studies using artificial intelligence (AI) and integrative epi- conference was to clarify diagnostic criteria for antibody-mediated demiological approaches were presented. Their applicability in pre- rejection (ABMR) and chronic active (CA) T cell–mediated rejection diction, prognostication, and clinical trials could change standards of 2320 | LOUPY et al. care for transplant recipients. 2 The conference was preceded by a T antigen (pvl1 ≤ 1%, pvl2 > 1%, and <10%, pvl3 ≥ 10%). The pvl and premeeting on “Regenerative Medicine and Digital Pathology” and ci scores were used to define 3 PVN classes (Tables 4 and 5).6 a joint session between the ASHI and Banff key opinion leaders, which presented insights into future tissue and organ engineering and new technologies in solid organ transplantation. Precise detec- 3 | C L A R I FI C ATI O N A N D U PDATE S tion and characterization of anti-HLA or non-HLA donor-specific an- TO TH E BA N FF 2 017 C L A S S I FI C ATI O N tibodies (DSAs) was also revisited. This report summarizes the main REVISIONS outcomes of the Banff 2019 kidney sessions and their impact on the Banff classification. The Banff 2019 kidney sessions focused on reviewing the impact of the changes to the diagnostic criteria for both TCMR and ABMR that were made at the 2017 Banff conference.5 For TCMR, this involved 2 | U PDATE S FRO M TH E BA N FF WO R K I N G mainly the implementation of CA TCMR: The extent to which this cat- G RO U P S egory is being used at different centers around the world, potential confusion regarding the wording of diagnoses within this category, The current active and new Banff working groups (BWGs) and their how frequently patients with CA TCMR were being treated, and the aims, leaders, and progress are listed in Table 1. One new BWG has efficacy of those treatments. The threshold for the interstitial inflam- been formed, the Digital Pathology BWG. The aim of this working mation score (i) in the diagnosis of borderline for acute TCMR, which group is to define standards for digital pathology in the context of has not been applied uniformly by different pathologists7 and been Banff lesion scores in order to standardize diagnostic scoring and the subject of recent investigation8,9 was also discussed. For ABMR, reduce inter-observer variability. In addition, the peritubular capil- discussions centered around potential simplification of the diagnos- laritis (ptc) BWG has been reactivated. Recent studies have shown tic criteria, including clarifying the description of activity and chro- diagnostic and prognostic relevance of reporting the extent of cap- nicity components within the category of CA ABMR—making the illaritis (diffuse vs focal) in addition to the Banff ptc score, in dif- resultant report more intelligible and useful to treating clinicians.10 ferent clinical scenarios including ABMR, mixed ABMR/TCMR, and Standardized reporting for TCMR and ABMR diagnoses were dis- low-grade microvascular inflammation (MVI).3,4 The aim of the BWG cussed, including the inclusion of individual Banff acute and chronic is to validate these findings in a multicenter study, and to evaluate scores (i, t, g, cg, and so on) used in determining the subtype, activ- optimal diagnostic thresholds. ity. and chronicity and explanations of these scores.1,7 The clinical At its meeting during Banff 2019, the electron microscopy (EM) implications of MVI in the absence of (DSAs with and without C4d BWG expanded on its previous recommendations1,5 and agreed to deposition were also presented). Final consensus recommendations guidelines for tissue sampling and performing EM analysis for as- were aided by questionnaires distributed to pathologists, nephrolo- sessment of cg1a and peritubular capillary basement membrane gists, and transplant surgeons attending the 2 kidney-specific ses- multilayering (PTCML). These are summarized in Table 2. The group sions during and immediately after the meeting (Table 6). also recommended the use of specific terminology aligned where possible with ongoing work of the Renal Pathology Society work- ing group developing consensus definitions for individual glomeru- 3.1 | Chronic active TCMR criteria clarifications lar lesions by light microscopy and EM. As specified in Table 3, this includes the following, representing individual components of very To assess whether transplant centers worldwide were applying early lesions of transplant glomerulopathy (cg1a): endothelial cell en- the 2017 Banff classification and diagnosing CA TCMR, we en- largement, subendothelial electron-lucent widening, and subendothelial listed the assistance of The Transplantation Society (TTS) and the neo-densa glomerular basement membrane. Reporting of other endo- CyberNephrology online forum (Dr Kim Solez) in circulating a sur- thelial and basement membrane features in glomeruli and peritubu- vey that was answered by 128 TTS members, including clinicians lar capillaries (PTCs) is left to the discretion of the pathologist as and pathologists, from six different continents before the meet- there are few data to support their use at present. Based on consen- ing (Figure 1). Notably, pathologists at ~90% of surveyed centers sus opinion within the EM BWG, abbreviation for peritubular capil- diagnosed CA TCMR, and at >80% of these centers at least some lary basement membrane multilayering is recommended as PTCML. patients with this diagnosis were treated with steroids or other im- At the 2019 conference, final data from the BWG on the clas- munosuppressive agents. However, those clinicians that do not rou- sification of polyomavirus nephropathy (PVN) were presented. This tinely treat patients with CA TCMR (52/81, 64%) indicated a need BWG studied 192 patients with definitive PVN and identified two for more definitive data on the efficacy of CA TCMR treatment, and histologic markers predictive for graft function and graft survival: the a minority (22/81, 27%) felt that the risk of treatment outweighs the Banff interstitial fibrosis (ci) score and a new score termed “the intra- benefits. Two points that should be emphasized regarding CA TCMR renal polyomavirus load level” or pvl, the latter based on the fraction are that inflammation in areas of the cortex with interstitial fibrosis of tubules with evidence of PV replication by light microscopy OR and tubular atrophy (i-IFTA), while an essential component of CA immunohistochemical staining of epithelial cell nuclei for SV40 large TCMR, are not sufficient to make this diagnosis, which also requires LOUPY et al. | 2321 TA B L E 1 Updates of Banff working groups Working group Leaders Issues to address Group progress/future plans TCMR V. Nickeleit Integration of i-IFTA into classification; A multicenter clinicopathologic study aimed at addressing the P. Randhawa reevaluate thresholds for i and t and listed issues is in progress. To date biopsies and accompanying possible addition of other findings (eg, clinical data from 154 patients with “pure” (no ABMR) acute edema) to TCMR diagnostic criteria TCMR/borderline, 18 with chronic active TCMR, 55 with no rejection. and 31 patients with stable graft function (not biopsied) have been accrued into the study cohort. Data analysis and slide review have commenced, but more cases, especially cases of chronic active TCMR and control cases with complete clinical data for statistical analysis of TCMR/ borderline thresholds, are required. These cases need to have documented absence of DSA and sufficient follow-up in addition to detailed clinical information. Sensitized L. Cornell Define criteria for highly sensitized patients Survey regarding clinical practice related to highly sensitized R. Sapir- (HS), determine consensus for what (HS) patients indicates that clinicians often fail to recognize Pichhadze personnel and facilities are needed for chronic elements of ABMR (eg, cg > 0) and are more likely to E. Kraus centers to perform transplantation in HS consider a diagnosis of chronic active ABMR if C4d is negative, S. Bagnasco recipients, standardize the definitions even if there is no TG, PTCML, or IFTA. The term “acute” is C. Schinstock related to management of sensitized confusing in ABMR, and consequently it was removed from D. Dadhania transplant recipients. Evaluate current the Banff ABMR classification in 2017.3 Further improved practices of centers performing renal communication between pathologists and clinicians regarding transplants in sensitized recipients. reporting of biopsy findings in HS. Follow-up preliminary Evaluate how clinicians interpret and apply survey results indicate that pathologists and clinicians Banff nomenclature, and recommend have similar concerns about biopsy reporting of antibody- changes to wording of classification to mediated injury: Banff classification is complex, changes optimize the use of Banff data in clinical frequently, and does not reflect the entire ABMR disease care of HS patients. spectrum. Comprehensive literature review is in progress to outline patient risk spectrum with respect to antibody and to standardize definitions of incompatible transplants. Molecular M. Mengel Develop consensus discovery gene panel, Launch of a consensus based, commercially available diagnostics R. Colvin which becomes available to the whole (NanoString Inc), standardized Banff Human Organ Transplant transplant community as standardized (B-HOT) discovery gene panel, which can reproducibly be commercially available reagents. Design applied to FFPE samples across organs and in multicenter and conduct collaborative studies to studies. Development of an open source data sharing platform validate clinical utility of molecular to exchange results from studies using the B-HOT panel and to diagnostics in transplant biopsies. allow multicenter and multiorgan validation of clinical utility of molecular diagnostics in transplant biopsies. Details given in a separate meeting report. HIV+/HIV+ renal S. Bagnasco Compare kidney transplants from HIV+ Based on preliminary data from the pilot phase of the US transplants and HIV-negative donors to HIV+ HOPE prospective multicenter trial for HIV+ to HIV+ kidney recipients with regard to graft function transplantation, TCMR is predominant, rejection appears to be and graft survival, · incidence, type and unrelated to donor HIV status and may possibly be related to pathologic features of allograft lesions the type of immunosuppression. Additional data will become including rejection, recurrent and de novo available with ongoing trial enrollment in upcoming years. HIV-related/-unrelated renal disease; The results of protocol/indication biopsies (archived as digital injury associated with antiretroviral and images), combined with comprehensive clinical information will immunosuppressive treatment. be analyzed to expand these preliminary observations. Electron C. Roufosse Recommendations for the sampling and Consensus document produced at Banff 2019 meeting. microscopy (EM) H. K Singh processing of tissue for EM and reporting Results of initial survey of interobserver variability were poor of EM findings in transplant biopsies. (presented at Banff 2017); a new comprehensive teaching Harmonize terminology with a list of module will focus on the most problematic lesions gleaned definitions for cg1a and PTCML. Reduce from the 2017 survey; Goal is to improve the uniform use of inter-observer variability in the use of currently published criteria and the inter-observer variability. uniform guidelines for the evaluation Research proposal presented at Banff 2019: pilot test case of TG and PTCML Multicenter study data collection in 2020; final study design to be agreed upon of the natural history, associations and before initiation of the full study in 2020-21. Goal is the the predictive value of glomerular and multicenter validation of ultrastructural features through peritubular capillary ultrastructural correlation with outcomes. features (endothelium; basement membrane; electron dense deposits). (Continues) 2322 | LOUPY et al. TA B L E 1 (Continued) Working group Leaders Issues to address Group progress/future plans Thrombotic M. Afrouzian Establish uniform diagnostic criteria for Using Delphi methodology, two phases were designed: Phase microangiopathy H. Liapis TMA. Determine the frequency with I (with pathologists) and Phase II (with clinicians). Phase I (6/9 (TMA) which TMA occurs in renal allograft biopsy rounds completed) involving 23 centers; 1. We collected 37 specimens. Determine if there are specific transplant biopsies, and after 6 Delphi rounds, narrowed down features of TMA in renal allografts that 331 criteria to 61; 2. We classified the criteria into 4 classes help resolve the differential diagnosis of (Pathology, Clinical, Laboratory and Differential Diagnosis) and the TMA when the cause is not readily 12 categories reflecting positive and negative criteria for LM, apparent from clinical history, DSA, C4d, IF, EM, clinical, laboratory and genetic criteria; 3. Participants etc validated the 37 cases in 2 validation rounds. Results showed that 75% of the participants reached consensus in over 75% of cases, using the 61 criteria. 4. We identified the reasons for over- and under-diagnosis of TMA. Results of rounds 7, 8 and 9 are pending. Phase II: Recruitment of participants for the clinical Delphi starting January 2020. Future efforts will focus on completing the 3 remaining Delphi rounds (Phase I), soliciting nephrologists’ input (Phase II), combining Phase I and II results, dissecting diagnostic issues that emerged during Phase I, such as ABMR mimicking TMA and completing the molecular studies to generate the final consensus guidelines. The WG also plans to collaborate with groups that are working on native kidney TMA to compare results and generate a consensus document. Recurrent N. Alachkar Establish pathologic guidelines for early Biopsy specimens and clinical data are now being collected glomerular S. Bagnasco recurrence of glomerular diseases, from 10 international centers. Preliminary results confirm disease including FSGS, IgA nephropathy, IgA nephropathy and FSGS as the most prevalent recurrent membranous nephropathy, MPGN/ diseases. Combine all data to create a posttransplant GN C3GN. What are frequencies, clinical registry. Future directions: manifestations, and pathologic • Correlate GN findings with Banff classification characteristics of recurrent/de novo • Correlate pathology with biomarkers, DSA, etc glomerular diseases, and can any of these • Determine if outcomes of posttransplant GN can be predict recurrence and/or graft outcomes? improved with early detection, prevention and therapy. Are Understand the pathologic changes of there clinical and/or pathologic features of the native disease recurrent glomerular diseases occurring that predict likelihood of recurrence? Are there clinical and/ concurrently with rejection and other or pathologic features of the recurrent disease in the allograft transplant-associated lesions. that predict graft loss? Which pathologic analyses are needed for optimal and early diagnosis of recurrent disease? Is the apparent association of recurrent glomerular disease with acute rejection related to biopsy bias (ie incidental discovery of recurrent disease in biopsies done to rule out rejection), under- immunosuppression, or both? Surrogate A. Loupy Define valid surrogate endpoints and New WG in progress endpoints M. Naesens invasive or noninvasive biomarkers for clinical trials and how histology and lesions related to tissue injury or scarring have to be integrated in such systems. Insure Banff active involvement and interactions with agencies such as FDA/ EMA and societies (AST, ESOT, TTS). Banff rules and J. U. Becker Elaboration of diagnostic algorithms Collation of all Banff content for kidney allograft pathology dissemination C. Roufosse for the Banff Diagnostic Classes 1-6. up to and including Banff 2017 (Transplantation 2018; Determination of the needs of the 102:1795-1814). Future plans include survey of the transplant transplant community for dissemination of community regarding dissemination needs, podcast based on Banff content. Predominantly web-based the 2018 review article in collaboration with TTS, development dissemination of Banff content. and implementation of web-based tools and dissemination of relevant Banff content including the content of this 2019 update. (Continues) LOUPY et al. | 2323 TA B L E 1 (Continued) Working group Leaders Issues to address Group progress/future plans Digital pathology K. Solez Digital automation of pathology practice: Standardization of practices, classification for studies using A. B. Farris computing, artificial intelligence, integrative approaches, IFTA scoring, inflammation scoring, nanotechnology, machine learning, slide algorithms to fit to the classification and decrease inter- numerization. observer variability. Archetypes to be validated across multiple institutions. Delivery of precision diagnostic, molecular pathways and therapeutics. Peritubular I. W.Gibson Evaluation whether a combined view of ptc Multicenter validation cohort study is under preparation, for capillaritis Z. Kikic score and ptc extent is superior compared collection, centralization, assessment and circulation of a N. Kozakowski to the currently applied standard of care subset of cases from different diagnostic groups for inter- (only ptc score) in distinguishing the role observer agreement is planned for 2020-2021. Prognostic of ptc in the following diagnostic settings: significance of a peritubular capillaritis grade incorporating active and chronic active ABMR, acute both ptc score and extent will be investigated and chronic active TCMR/borderline, low-grade MVI and mixed ABMR/TCMR. Test the reproducibility and prognostic significance of the combined of ptc score and extent. Gene expression analysis of peritubular capillaritis in different diagnostic settings. Comparison of cortical ptc vs medullary ptc (vasa recta). Abbreviation: FFPE, formalin-fixed, paraffin-embedded. (1) at least a moderate degree of total cortical inflammation and (2) therapy. These data support the 2017 Banff criteria for CA TCMR, moderate tubulitis involving cortical tubules other than severely but also favor indicating the level of active inflammation in non- atrophic tubules (Table 4). i-IFTA itself is not a specific lesion and is scarred cortex. seen in the context of tissue injury due to many causes other than TCMR, including BK virus nephropathy and ABMR.11 According to Banff 2017,5 the tubulitis allowed for the diagnosis of CA TCMR may 3.2 | Borderline (suspicious) for acute TCMR criteria be within or outside of scarred areas. However, for clarity and future clarifications investigation we suggest that tubulitis be independently scored (pro- viding that severely atrophic tubules are not scored) both in areas of The Banff ’97 criteria for diagnosis of borderline for acute preserved cortex (Banff t score) and within (Banff t-IFTA score) areas TCMR required inflammation in at least 10% of nonscarred cor- of cortical IFTA, as defined in Table 5. This may enable investiga- tex (Banff i1) in addition to tubulitis.13 However, subsequent tors and clinicians to determine the impact on treatment response iterations of the Banff classification (1) accepted tubulitis with and graft survival of these different forms of tubulitis. An important minimal inflammation (isolated t; i0 t > 0) as sufficient for a di- concern expressed by some clinicians and pathologists is that the agnosis of borderline, and a recent survey showed variability Banff 2017 diagnosis of CA TCMR does not discriminate between among pathologists with regard to which threshold was used for cases where all or nearly all inflammation is limited to i-IFTA vs those this diagnosis.7 More recently, longitudinal studies of Nankivell instances where coexistent i-IFTA and inflammation within non- et al9 showed no effect of isolated t lesions on graft outcomes, scarred areas meets criteria for additional borderline acute TCMR concluding that the original Banff ’97 threshold for borderline or acute TCMR grade IA or IB diagnoses. Questionnaire responses lesions should be applied, and 82% (57/69) of pathologists and from meeting attendees clearly replied that reporting of “mixed” clinicians (nephrologists and transplant surgeons) responding CA TCMR with grade IA or IB acute TCMR could be improved by to the questionnaire distributed at the 2019 Banff conference indicating separately when biopsies simultaneously met criteria for agreed (Table 6). Consensus was therefore achieved that the di- borderline for acute TCMR or acute TCMR grade IA or IB (Table 6). agnostic threshold for borderline for acute TCMR should read Likewise, participants agreed that when intimal arteritis was present as follows: “interstitial inflammation involving 10%-25% of non- in addition to CA TCMR, this should also be specifically reported sclerotic cortex (Banff i1) with at least mild tubulitis (t > 0).” The as a separate reported diagnosis. These suggestions pertain only to minimum lesion for a borderline diagnosis is thus i1t1. As with CA TCMR reporting and the actual histologic criteria for this diag- previous versions of the Banff classification,1,13 lesions charac- 12 nosis remain unchanged (Tables 4 and 5). Nankivell et al reported terized by mild tubulitis (Banff t1) with inflammation involving that both development of i-IFTA and clearance of inflammation from >25% of nonsclerotic cortex (Banff i2/3) and no intimal arteritis scarred areas were dependent of the efficacy of immunosuppressive are also included in the borderline category (Table 4). However, 2324 | LOUPY et al. TA B L E 2 Banff recommendations for electron microscopy in TA B L E 3 Definitions of the individual components of early renal transplant biopsies lesions of transplant glomerulopathy Recommendations for taking of a sample for EM • Endothelial cell enlargement is defined as glomerular endothelial Take a sample in all cases if possible, fixed and embedded as a resin enlargement with associated increase in cytoplasm, decrease in block. At a minimum, this should be done if there is any suspicion the capillary luminal area and loss of fenestrations. This should of glomerular disease. While EM can be performed on samples be used rather than less precise or incorrect terminology such recovered from paraffin blocks or frozen tissue to examine for “endothelial swelling”, “reactive endothelial changes”, “endothelial electron dense deposits and to some extent the degree of foot activation”, “endothelial injury” or “endothelial cell degeneration”. process effacement, this is not recommended for assessment of • Subendothelial electron-lucent widening is defined as expansion glomerular basement membrane (GBM) thickness, cg1a or PTCML. of space between the glomerular endothelium and lamina densa In all cases, light microscopic (LM) examination of semi-thin stained of the glomerular basement membrane (GBM) by electron-lucent sections should be performed. material. This should be used instead of “lamina rara interna expansion” or “subendothelial rarefaction”. Recommendations for performing ultrastructural analysis • Subendothelial neo-densa glomerular basement membrane is Recommended in cases with: (a) Clinical, light microscopic and/or defined as circumferential or discontinuous, single or multiple immunohistochemical suspicion of glomerular disease or of other basement membrane layer(s) clearly distinct from the original diseases where EM may assist in diagnosis; (b) Patients at risk for GBM. antibody-mediated rejection (ABMR): patients who are sensitized, have documented DSA at any time posttransplant, and/or who have had a prior biopsy showing features of ABMR (C4d staining, biopsies with tubulitis (t > 0) but without inflammation in at least glomerulitis and/or peritubular capillaritis). 10% of nonsclerotic cortex (i0) are no longer included in this cat- EM can also be useful: (a) To detect early recurrence in patients with biopsy-proven glomerular disease as the cause of native kidney failure; egory, and neither are cases with inflammation in the absence of (b) In for-cause biopsies ≥3 mo posttransplant, and in all biopsies tubulitis (i > 0, t0). These are simply reported in the microscopic ≥6 mo posttransplant, to determine if early changes of transplant description of the biopsy. glomerulopathy (cg1a) are present, prompting testing for DSA. In all cases, undertake an assessment for glomerular disease as in a native renal biopsy. If performing EM in the setting of possible ABMR, assess for cg1a 3.3 | ABMR criteria clarifications and PTCML as recommended below. Recommendations for preparing the block for ultrastructural Since the introduction of ABMR into the Banff classification in examination 2003,14 the criteria for diagnosis of ABMR have become more Select an area in the resin block(s) of viable (non-necrotic) cortex, complex, with major modifications in 201315 and 2017, 5 which preferentially with minimal tubular atrophy/interstitial fibrosis. have improved diagnostic sensitivity and predictive value for graft Aim to include ≥2 full glomeruli (at least 1). Do not taper the block to a “cone” or “pyramid” by removing cortical tissue around the outcomes.16 A major residual issue within the classification is that glomeruli and aim to have at least 10 peritubular capillaries (PTCs) it still subclassifies ABMR into active or chronic active and chronic for examination. inactive (ie, transplant glomerulopathy without MVI and with cur- Guidelines for ultrastructural assessment and reporting of cg1a rent or historic DSA; Table 4) subtypes as opposed to representing Exclude globally or partly sclerosed glomeruli, and severely ischemic the diverse morphologic and molecular lesions at different post- glomeruli. transplant time points in antibody-mediated tissue injury.17,18 In Examine 1 complete glomerulus as a minimum; where possible particular, the category of CA ABMR encompasses lesions with examine 2 or more. Examine the capillary loops at high magnification (≥5000x). severe activity and mild chronicity (eg, g3 ptc3 cg1 ci0 ct0 C4d3), cg1a is defined as: (a) No GBM double contours on LM; (b) those with mild activity and severe chronicity (eg, g1 ptc1 cg3 ci3 ≥3 capillary loops (in a single glomerulus) each showing: ct3 C4d0), and intermediate cases. Thus a simple diagnosis of CA subendothelial neo-densa glomerular basement membrane ABMR gives treating clinicians limited information for appropri- (circumferential or not, single or multiple layers) AND endothelial cell enlargement and/or subendothelial electron-lucent widening. ate treatment options. Although the addition of individual Banff The report should state how many glomeruli were examined, and lesion scores to the diagnostic line of the report is interpretable how many loops show cg1a. by experienced clinicians, it may confuse some clinicians who are Guidelines for ultrastructural assessment and reporting of PTCML unfamiliar with subtleties of the Banff system. This concern was Assess for PTCML in the cortical peritubular capillaries between the highlighted by the recent work from the BWG on highly sensitized glomeruli. patients who found that the classification was interpreted in a Exclude scarred cortex and necrotic or hemorrhagic areas. highly variable way with implications for resultant therapeutic de- Examine at least 10 PTC at high magnification (≥ 5000×). Count layers of basement membrane (BM) in the 3 worst affected cision making.10,19 Potential improvements in the wording of the PTCs, for each counting in the PTC segment with the most layers pathologic diagnoses in standard reports were discussed, which (worst affected area along the circumference). remained within the context of Banff 2017 to optimally convey The report should state the number of PTC examined; the number the activity and chronicity levels of CA ABMR cases. For example, of BM layers in the most affected PTC; and whether the Banff 2013 threshold for chronic ABMR (ptcml1; 1 PTC with ≥7 layers +2 in lupus nephritis, 20 the previous designations of active (A), active PTC with ≥5 layers) is met. and chronic (A/C), and chronic (C) were replaced with a revised LOUPY et al. | 2325 TA B L E 4 Updates of 2019 Banff classification for ABMR, TA B L E 4 (Continued) borderline changes, TCMR, and polyomavirus nephropathy. All C4d staining without evidence of rejection; all 4 features must be updates in boldface typea present for diagnosisc Category 1: Normal biopsy or nonspecific changes 1. Linear C4d staining in peritubular capillaries (C4d2 or C4d3 by IF Category 2: Antibody-mediated changes on frozen sections, or C4d > 0 by IHC on paraffin sections) Active ABMR; all 3 criteria must be met for diagnosis 2. Criterion 1 for active or chronic active ABMR not met 1. Histologic evidence of acute tissue injury, including 1 or more of 3. No molecular evidence for ABMR as in criterion 2 for active and the following: chronic active ABMR • Microvascular inflammation (g > 0 and/or ptc > 0), in the absence 4. No acute or chronic active TCMR, or borderline changes of recurrent or de novo glomerulonephritis, although in the Category 3: Borderline (Suspicious) for acute TCMR presence of acute TCMR, borderline infiltrate, or infection, ptc ≥ 1 alone is not sufficient and g must be ≥ 1 Foci of tubulitis (t1, t2, or t3) with mild interstitial inflammation (i1), • Intimal or transmural arteritis (v > 0)b or mild (t1) tubulitis with moderate-severe interstitial inflammation • Acute thrombotic microangiopathy, in the absence of any other (i2 or i3) cause No intimal or transmural arteritis (v = 0) • Acute tubular injury, in the absence of any other apparent cause Category 4: TCMR 2. Evidence of current/recent antibody interaction with vascular Acute TCMR endothelium, including 1 or more of the following: • Linear C4d staining in peritubular capillaries or medullary vasa Grade IA: Interstitial inflammation involving >25% of non-sclerotic recta (C4d2 or C4d3 by IF on frozen sections, or C4d > 0 by IHC cortical parenchyma (i2 or i3) with moderate tubulitis (t2) involving on paraffin sections) 1 or more tubules, not including tubules that are severely atrophicd • At least moderate microvascular inflammation ([g + ptc] ≥2) in the Grade IB: Interstitial inflammation involving >25% of non-sclerotic absence of recurrent or de novo glomerulonephritis, although in cortical parenchyma (i2 or i3) with severe tubulitis (t3) involving 1 the presence of acute TCMR, borderline infiltrate, or infection, or more tubules, not including tubules that are severely atrophicd ptc ≥ 2 alone is not sufficient and g must be ≥1 Grade IIA: Mild to moderate intimal arteritis (v1), with or without • Increased expression of gene transcripts/classifiers in the biopsy interstitial inflammation and/or tubulitis tissue strongly associated with ABMR, if thoroughly validated Grade IIB: Severe intimal arteritis (v2), with or without interstitial 3. Serologic evidence of circulating donor-specific antibodies (DSA inflammation and/or tubulitis to HLA or other antigens). C4d staining or expression of validated transcripts/classifiers as noted above in criterion 2 may substitute Grade III: Transmural arteritis and/or arterial fibrinoid necrosis for DSA; however thorough DSA testing, including testing for involving medial smooth muscle with accompanying mononuclear non-HLA antibodies if HLA antibody testing is negative, is strongly cell intimal arteritis (v3), with or without interstitial inflammation advised whenever criteria 1 and 2 are met and/or tubulitis Chronic active ABMR; all 3 criteria must be met for diagnosis Chronic active TCMRe 1. Morphologic evidence of chronic tissue injury, including 1 or more Grade IA: Interstitial inflammation involving >25% of sclerotic of the following: cortical parenchyma (i-IFTA2 or i-IFTA3) AND > 25% of total Transplant glomerulopathy (cg > 0) if no evidence of chronic TMA cortical parenchyma (ti2 or ti3) with moderate tubulitis (t2 or or chronic recurrent/de novo glomerulonephritis; includes changes t-IFTA2) involving 1 or more tubules, not including severely evident by electron microscopy (EM) alone (cg1a) atrophic tubulesd ; other known causes of i-IFTA should be ruled Severe peritubular capillary basement membrane multilayering out (ptcml1; requires EM) Grade IB: Interstitial inflammation involving >25% of sclerotic Arterial intimal fibrosis of new onset, excluding other causes; cortical parenchyma (i-IFTA2 or i-IFTA3) AND > 25% of total leukocytes within the sclerotic intima favor chronic ABMR if there cortical parenchyma (ti2 or ti3) with severe tubulitis (t3 or t-IFTA3) is no prior history of TCMR, but are not required involving 1 or more tubules, not including severely atrophic 2. Identical to criterion 2 for active ABMR, above tubulesd ; other known causes of i-IFTA should be ruled out 3. Identical to criterion 3 for active ABMR, above, including strong Grade II: Chronic allograft arteriopathy (arterial intimal fibrosis recommendation for DSA testing whenever criteria 1 and 2 with mononuclear cell inflammation in fibrosis and formation of are met. Biopsies meeting criterion 1 but not criterion 2 with neointima). This may also be a manifestation of chronic active or current or prior evidence of DSA (posttransplant) may be stated chronic ABMR or mixed ABMR/TCMR as showing chronic ABMR, however remote DSA should not be Category 5: polyomavirus nephropathyf considered for diagnosis of chronic active or active ABMR PVN Class 1 Chronic (inactive) ABMR pvl 1 and ci 0-1 1. cg > 0 and/or severe ptcml (ptcml1) PVN Class 2 2. Absence of criterion 2 of current/recent antibody interaction with pvl 1 and ci 2-3 OR the endothelium pvl 2 and ci 0-3 OR pvl 3 and ci 0-1 3. Prior documented diagnosis of active or chronic active ABMR and/or documented prior evidence of DSA PVN Class 3 pvl 3 and ci 2-3 (Continues) (Continues) 2326 | LOUPY et al. TA B L E 4 (Continued) at this point, the clinical implications of these findings remain unclear. a Individual Banff lesion scores are defined in Table 5, and it is Recent evidence has suggested that patients with MVI on biopsies recommended that these be included in the biopsy report. b performed early posttransplant in the absence of DSAs had signifi- It should be noted that these arterial lesions may be indicative of cantly better graft survival than those with anti-HLA DSAs (mainly ABMR, TCMR, or mixed ABMR/TCMR. “v” lesions and chronic allograft arteriopathy are only scored in arteries having a continuous media with low-level, preexisting DSAs), regardless of the presence or absence ≥2 smooth muscle layers. of PTC C4d staining on the biopsy. 21 DSA+ patients were more fre- c The clinical significance of these findings may be quite different in quently retransplant recipients and had C4d+ biopsies. By contrast, grafts exposed to anti–blood group antibodies (ABO-incompatible studies of patients biopsied much later with more severe MVI and allografts), where they do not appear to be injurious to the graft and may represent accommodation. However, with anti-HLA antibodies, frequent transplant glomerulopathy (TG) showed no difference in such lesions may progress to chronic ABMR, and more outcome data graft survival in the presence or absence of anti-HLA DSAs, which are needed. in these studies were de novo DSAs. 22,23 Thus, at this point, we are d Severely atrophic tubules are defined as having each of 3 features: unable to draw significant conclusions regarding the significance diameter <25% of unaffected or minimally affected tubules in the same of biopsies showing (g + ptc >2) in the absence of anti-HLA DSAs, biopsy, an undifferentiated-appearing, cuboidal or flattened epithelium, and pronounced wrinkling and/or thickening of the tubular basement other than to comment that the biological behavior of such changes membrane. may well differ depending on whether these are seen relatively early e It was felt by the majority of Banff 2019 meeting attendees that posttransplant at a time when many DSAs represent persistent or reporting of chronic active TCMR should be accompanied by a second recurrent/memory DSAs vs those seen later when most DSAs de- diagnosis of borderline acute TCMR or acute TCMR (with appropriate tected are de novo DSAs. In addition, the differential diagnosis of grade) when criteria for both diagnoses are met. f MVI with negative DSA and negative C4d should include T cell–me- pvl scores are defined in Table 5 and in detail in ref. 4. An adequate sample for such scoring should include 2 biopsies cores and contain diated endothelial injury (ptc), membranoproliferative glomerulone- medulla. PVN can coexist with ABMR or with TCMR grades 2 or 3. phritis (g), and thrombotic microangiopathy (g). Furthermore, it has been shown recently in large populations that non-HLA antibodies version of the old National Institutes of Health (NIH) activity and could be independently associated with risk of allograft loss. 24 These chronicity indices in a more granular fashion. However, the appli- findings emphasize the importance of testing for clinically relevant cability of this approach to transplantation remains to be validated. non-HLA antibodies in patients with biopsies showing MVI. Questionnaire respondents were also asked to select the optimal It is also hoped that implementation of molecular diagnostics could wording of the current Banff 2017 ABMR diagnosis from several help classifying these cases in a more pathogenesis-driven way, either options detailing activity and chronicity information, and how to employing the Molecular Microscope Diagnostic System (MMDx)25 or best incorporate relevant Banff lesion scores into the final report. the NanoString platform that can be used with formalin-fixed, paraf- The option preferred by 56% of respondents was to word the di- fin-embedded tissue,26 providing greater insight into factors predictive agnosis line as follows: “(chronic) active ABMR with [mild, moder- of the biological behavior of MVI without detectable anti-HLA DSAs, ate, severe] activity and [mild, moderate, severe] chronicity” and as well as cases with C4d staining without MVI.27 Current recommen- include the Banff lesion scores in the text of the diagnostic line for dations of the Banff Molecular Diagnostics Working Group in these the biopsy report or in a table (eg, Tables 5 and 6). However, the and other circumstances are detailed in a separate meeting report. definitions and optimal thresholds for mild, moderate, and severe Presently, however, available data do not warrant making changes to still need to be validated in future studies, and as such, it remains the ABMR classification specifically relevant to these cases, beyond at the discretion of individual pathologists whether to adopt such those made in Banff 2017.5 wording at this time or simply provide the Banff lesion scores. It should also be emphasized that these suggestions pertain only to the CA ABMR reporting language, and the actual histologic crite- 4 | RO LE O F A RTI FI C I A L I NTE LLI G E N C E , ria for diagnosis of CA ABMR, as specified in Banff 2017, remain DATA I NTEG R ATI O N , A N D M AC H I N E unchanged. 5 LE A R N I N G I N TH E BA N FF C L A S S I FI C ATI O N During the 2019 Banff meeting, many projects using artificial intelli- 3.4 | Microvascular inflammation in the absence of gence (AI), machine learning (ML), as well as deep learning (DL) were detectable DSAs presented, with a heavy focus on applying the Banff classification in an automated fashion, clustering of patients, and digital pathology. At the Banff 2017 conference, updates to the ABMR classification However, the field is still poorly understood by most clinicians and pa- were made to help with decision-making in cases where a biopsy thologists. Several speakers at the 2019 Banff meeting discussed the showed at least moderate MVI ((g + ptc) >2) but no detectable anti- capabilities of AI and the potential for use of ML algorithms in diagnosis bodies against the graft.5 However, there are considerable numbers and personalized medication in organ transplantation (Table 7). of biopsies with (g + ptc) >2 in the absence of detectable anti-HLA Computer-assisted digital pathology consists of analyzing whole DSAs or C4d, and with molecular ABMR testing not widely adopted slide digital images of biopsies by a machine, and remains a challenge LOUPY et al. | 2327 TA B L E 5 Banff reporting standardization scheme: Individual Banff scores used in grading of acute and chronic active ABMR and TCMR. Inclusion in the biopsy report of table similar to the below or a simple listing of individual scores in a comment is advised Acute & chronic Banff Acute Banff scores Grading (0, 1, 2, 3) Chronic Banff scores Grading (0, 1, 2, 3) scores Grading (0, 1, 2, 3) i ci ti t ct i-IFTA v cv t-IFTA g cg pvl ptc ptcml C4d Abbreviations: i, inflammation in non-scarred cortex, scored as 0 (absent/minimal, <10% of non-scarred cortex inflamed), 1 (mild, 10%-25%), 2 (moderate 26%-50%), 3 (severe, >50%). The subcapsular cortex should not be considered; t, tubulitis in cortical tubules within non-scarred cortex, scored as 0 (none), 1 (mild, 1-4 mononuclear leukocytes per tubular cross-section or 10 tubular epithelial cells in most severely involved tubule), 2 (moderate, 5-10 mononuclear leukocytes), 3 (severe, >10 mononuclear leukocytes). Although the tubulitis score (1-3) is based on the single most severely involved tubule, at least mild tubulitis must be present in ≥2 cortical foci for a t score >0 to be assigned. Severely atrophic tubules should NOT be scored; v, endarteritis (intimal arteritis), scored as 0 (none), 1 (mild, 1 or more leukocytes directly beneath the endothelium of 1 or more arteries; endothelial cells typically appear enlarged with associated subendothelial edema; <25% luminal occlusion), 2 (moderate, as grade 1, but with ≥25% luminal occlusion), 3 (severe, with arterial fibrinoid necrosis or transmural inflammation). Note that only arteries (minimum 2 smooth muscle layers) are scored; g, glomerulitis, scored as 0 (none), 1 (mild, with ≥1 leukocyte AND associated endothelial swelling occluding >50% of 1 of more capillary lumina in at least one but <25% of glomeruli), 2 (moderate, these changes involving 25%-75% of glomeruli), 3 (severe, involving >75% of glomeruli). Ischemic, collapsed glomeruli and glomeruli with >50% sclerosis should not be scored; ptc, peritubular capillaritis, scored as 0 (minimal, with <3 leukocytes in the most severely involved cortical PTC and/or leukocytes in <10% of cortical PTCs), 1 (mild, with ≥1 leukocyte in ≥10% of cortical PTCs AND 3-4 leukocytes in the most severely involved PTC), 2 (moderate, as grade 1 but with 5-10 leukocytes in most severely involved PTC), 3 (severe, as grades 1-2 but with >10 leukocytes in most severely involved cortical PTC). The extent of PTC inflammation should be documented as focal (10%-50% of cortical PTCs) or diffuse (>50%). Note that medullary capillaries are NOT scored; C4d, linear staining in PTCs or medullary vasa recta by immunofluorescence (IF) on frozen sections of fresh tissue or immunohistochemistry (IHC) on formalin-fixed, paraffin-embedded tissue, scored as 0 (none), 1 (minimal, staining in >0 but <10% of PTCs), 2 (focal, 10%-50% of PTCs), 3 (diffuse, >50% of PTCs). By IF on frozen sections scores ≥2 are considered positive; by IHC on paraffin sections all scores >0 are considered positive; ci, interstitial fibrosis in cortex, scored as 0 (minimal, ≤5%), 1 (mild 6%-25%), 2 (moderate, 26%-50%), 3 (severe, >50%). The subcapsular cortex should not be scored; ct, tubular atrophy in cortex, scored as 0 (none), 1 (mild, 1%-25%), 2 (moderate, 26%-50%), 3 (severe, >50%). The subcapsular cortex should not be scored; cv, arterial intimal fibrosis (fibrointimal thickening), scored as 0 (none), 1 (mild, present but with ≤25% narrowing of luminal area in the most involved artery), 2 (moderate, 26%-50% luminal narrowing), 3 (severe, >50% luminal narrowing); cg, chronic glomerulopathy (transplant glomerulopathy), scored as 0 (none, no GBM double contours by light microscopy [LM] or EM), 1a (early mild, no GBM double contours by LM but subendothelial neo-densa in ≥3 glomerular capillaries by EM with associated endothelial cell enlargement and/or subendothelial electron-lucent widening, 1b (mild, GBM double contours by LM in 1%-25% of glomerular capillaries by LM in the most severely involved glomerulus), 2 (moderate, double contours by LM in 26%-50% of capillaries), 3 (severe, double contours by LM in >50% of capillaries). Ischemic, collapsed glomeruli and glomeruli with >50% sclerosis should not be scored; ptcml, peritubular capillary basement membrane multilayering (requires EM), scored as 1 (≥7 basement membrane layers in the most affected PTC AND ≥5 layers in two additional PTCs), 0 (not meeting these criteria), or NA if EM is not performed; ti, total cortical inflammation, including scarred and non-scarred cortex, scored as 0 (absent/minimal, <10%), 1 (mild, 10%-25%), 2 (moderate, 26%-50%), 3 (severe, >50%); i-IFTA, inflammation in scarred cortex, scored as 0 (absent/minimal, <10% of non-scarred cortex inflamed OR if the extent of cortical IFTA is <10%), 1 (mild, 10%-25% of scarred cortex inflamed), 2 (moderate 26%-50% of scarred cortex inflamed), 3 (severe, >50% of scarred cortex inflamed). The subcapsular cortex should not be considered; t-IFTA, tubulitis in tubules within scarred cortex, scored as 0 (none), 1 (mild, 1-4 mononuclear leukocytes per tubular cross-section or 10 tubular epithelial cells in most severely involved tubule), 2 (moderate, 5-10 mononuclear leukocytes), 3 (severe, >10 mononuclear leukocytes). Severely atrophic tubules should NOT be scored; pvl, intrarenal polyomavirus load level, defined by the overall fraction of tubules in the entire biopsy (cortex and medulla) with at least 1 epithelial cell showing a viral inclusion body by light microscopy OR nuclear staining for SV40 large T antigen by IHC. As IHC is more sensitive, proper pvl scoring requires IHC. Scored as 0 (none), 1 (mild, positive cells in ≤1% of tubules), 2 (moderate, >1% and <10%), 3 (severe, ≥10%). The pvl score in combination with the ci score is used to define the PVN class (Table 4). even with most advanced image analyzing algorithms such as con- personalized medicine has been a major focus in recent years, utiliz- volutional neural networks. At this point DL-based histopathologic ing algorithms and archetype analysis has the potential to combine assessment of kidney tissue shows the ability to distinguish glom- biopsy results with clinical and laboratory data and to provide a more eruli, proximal and distal tubules, and areas of fibrosis. 28-30 The cur- complete pathologic and prognostic picture, especially in the era of rent approach is promising but still shows high variability and poor expanding electronic medical records.31 Furthermore, A. Loupy pre- correlation with clinical outcomes. A BWG on Digital Pathology has sented the capability of natural language processing (NLP) to help such been formed at the 2019 Banff meeting to standardize scanning and algorithms assist clinicians and reduce human errors. NLP could be used analytical practices through facilitating studies focusing on IFTA and for an otherwise time-consuming meta-analysis or accelerated reading inflammation scoring while using algorithms fitting the most cur- of large numbers of pathology reports. Finally, since the Banff rules rent Banff classification to decrease inter-observer variability. Because are becoming increasing complex with numerous possible scenarios, 2328 | TA B L E 6 Survey results on reporting of chronic active T cell–mediated rejection (CA TCMR), chronic active antibody-mediated rejection (CA ABMR), and Banff i threshold for borderline (BL) lesions Preferred diagnosis wording CA TCMR CA TCMR + borderline or CA TCMR +active component meeting CA TCMR Acute TCMR+ moderate inflammed Other acute TCMR criteria for BL or acute TCMR IFTA i1 t2 ti2 iIFTA3 ci2 ct2 46% 33% 20% 0% 1% +noMVI i2 t2 ti2 iIFTA3 ci2 ct2 61% 21% 5% 12% 1% + noMVI i1 t2 ti2 iIFTA3 ci2 ct2 70% 12.5% 6.5% 10% 1% v1 + noMVI CA ABMR Current wording: CA ABMR CA ABMR +specify g and ptc scores CA ABMR +mild/moderate/severe ABMR +mild/moderate/severe Other activity/chronicity activity/chronicity 14% 26% 27% 29% 4% Preferred threshold Borderline (suspicious) for Inflammation in <10% of Inflammation in 10%-24% of Inflammation in 5%-24% of acute TCMR nonsclerotic cortex (i0 t > 0) nonsclerotic cortex (i1 t > 0) nonsclerotic cortex (with t > 0) 15% 82% 3% Note: This table summarizes the answers obtained from a survey distributed to Banff attendees during or immediately after the meeting. These results rely on answers from 69 pathologists and clinicians regarding preferred wording in situation of CA TCMR, and CA ABMR and regarding preferred threshold for borderline for acute TCMR. Abbreviations: BL, borderline for acute TCMR; CA ABMR, chronic active antibody-mediated rejection; IFTA, interstitial fibrosis and tubular atrophy; MVI, microvascular inflammation. LOUPY et al. LOUPY et al. | 2329 When do you treat CA TCMR ? Routinely Under condition Rarely 36% 49 % 15% - No severe IFTA (47%) - Lack of data (64%) - Depending on i and v (49%) - Risks outweighs benefits or - If for cause biopsy (4%) experience of ineffectiveness (27%) - Pathologists do not diagnose CA TCMR (9%) 30 % How do you treat ? Which response rate do you observe? 25 20 23% 15 <50% of cases 10 46% >50% of patients usually partia 14% >50% sometimes complete 5 rarely 0 Steroids Steroids + Increase ATG if no Steroids +/- Other 17% ATG if IB baseline IS response to ATG if IB v>0 steroids F I G U R E 1 Responses to TTS survey regarding the management of chronic active (CA) TCMR: of the 128 respondents to this survey, 47 (37%) were from Europe, 26 (20%) from Asia, 23 (18%) from the United States and Canada, 18 (14%) from Latin America, 9 (7%) from the Middle East and Africa, and 5 (4%) from Australia and New Zealand. Sixty-five percent of centers performed <100 renal transplantations annually; 12% performed >200. At all but 12% of centers, biopsies were read by a renal or transplant pathologist TA B L E 7 Feasible applications of artificial intelligence in the Banff classification Type Fields of usage Popular algorithms Image recognition Digital pathology CNN, ResNet, VGG, etc NLP Meta-analysis SyntaxNet, transfer learning, SVM, naive Bayes classifier, etc Text mining Meta-analysis, automated report, report/web k-means clustering, naive Bayes classifier, scraping KNN, SVM, etc Cluster pattern recognition Gene expression, personalized medicine Linkage algorithms, k-means, DBSCAN, archetypal analysis, etc Class prediction Graft lost, response to therapy Random forest, multinomial logistic regression, SVM, neural network, etc Abbreviations: CNN, convolutional neural network; DBscan, description density-based spatial clustering of applications with noise; KNN, K nearest neighbors; NLP, natural language processing; SVM, support vector machine; VGG, visual geometry group. there is increasing demand for an automated coding of Banff rules. 5 | CO N C LU S I O N S This would require integration of experienced pathologists to decode the Banff rules and computer/data scientists to create an algorithm to The international consensus-building framework provided by the apply the Banff rules. Key to achieving quality results from ML-based Banff process is ongoing. Much has been done in the field of diag- diagnosis are the integration of multiple centers providing quality input nostic assessment of allograft pathology, but many issues remain to data to construct a standardized and reproducible analytical process. be studied further. Although our experience in continually updating 2330 | LOUPY et al. the diagnostic criteria for renal allograft rejection and related le- ORCID sions has improved diagnostic accuracy and clinicopathologic cor- Alexandre Loupy https://orcid.org/0000-0003-3388-7747 relations,16 this experience has also helped clarify the limits of Candice Roufosse https://orcid.org/0000-0002-6490-4290 histology and immunohistology in renal allograft biopsy interpre- Benjamin Adam https://orcid.org/0000-0003-1908-1739 tation and emphasized the need for development of additional Marian C. Clahsen-van Groningen https://orcid.org/0000-0003-0565-9560 diagnostic modalities, including molecular diagnostics. Current rec- Anthony J. Demetris https://orcid.org/0000-0002-9582-3733 ommendations regarding the further clinical validation and use of Jean-Paul Duong van Huyen https://orcid.org/0000-0002-3051-2141 the latter are detailed in a separate meeting report; this and other Alton B. Farris https://orcid.org/0000-0001-5534-7763 themes will also be addressed at the XVI. Banff meeting, which Roslyn B. Mannon https://orcid.org/0000-0003-1776-3680 will be held jointly with the Canadian Society of Transplantation Brian J. Nankivell https://orcid.org/0000-0002-2300-0287 in Banff, Canada, October 4-7 2021, celebrating the 30-year an- Peter Nickerson https://orcid.org/0000-0002-7393-7799 niversary of the Banff classification at the original location of its Marion Rabant https://orcid.org/0000-0001-5696-6478 inception. Parmjeet Randhawa https://orcid.org/0000-0002-7523-3437 Carrie A. Schinstock https://orcid.org/0000-0002-3274-3277 Harsharan K. Singh https://orcid.org/0000-0002-3123-5027 AC K N OW L E D G M E N T S Mark D. Stegall https://orcid.org/0000-0001-9157-5405 Members of the EM BWG: Amanda Kan, Anne Raisanen-Sokolowski, Michael Mengel https://orcid.org/0000-0002-7222-3356 Chris Bellamy, Catherine Horsfield, Deb Schady, David Buob, Finn P Reinholt, George Liapis, Ibrahim Batal, Bela Ivanyi, Johan Molne, REFERENCES Marlene Praet, Paisit Paueksakon, Patricia Revelo, Praveen Chander, 1. Loupy A, Haas M, Solez K, et al. The Banff 2015 Kidney Meeting Rafael Maldonado, Surya V. Seshan, Verena Broecker, Yasemin Ozluk, Report: current challenges in rejection classification and pros- pects for adopting molecular pathology. Am J Transplant. Nicholas Kozakowski, Linda Moran, Yong Mee Cho, Luan Truong, 2017;17(1):28-41. Christin Vanbeek, John Brealey, Antonella Barreca, Amit Dinda, 2. Loupy A, Aubert O, Orandi BJ, et al. Prediction system for risk of Geetika Singh, Joris Roelofs, Alexei Mikhailov, Shana Coley. The 2019 allograft loss in patients receiving kidney transplants: international Banff meeting received sponsorship from: CareDx, CSL Behring, derivation and validation study. BMJ. 2019;366:l4923. 3. Kozakowski N, Herkner H, Böhmig GA, et al. The diffuse extent of Elsevier, Eppendorf, GenDx, Hansa Biopharma, Histogenetics, peritubular capillaritis in renal allograft rejection is an independent Immucor, Omion, OneLambda, NanoString, Novartis, Takeda, Veloxis, risk factor for graft loss. Kidney Int. 2015;88:332-340. and Vitaeris. 4. Kozakowski N, Herkner H, Eskandary F, et al. An integrative ap- proach for the assessment of peritubular capillaritis extent and score in low-grade microvascular inflammation – associations with transplant glomerulopathy and graft loss. Nephrol Dial Transplant. D I S C LO S U R E 2019;34:166-217. The authors of this manuscript have conflicts of interest to dis- 5. Haas M, Loupy A, Lefaucheur C, et al. The Banff 2017 Kidney close as described by the American Journal of Transplantation. Meeting Report: revised diagnostic criteria for chronic active T Michael Mengel received honoraria from Novartis, CSL Behring, cell-mediated rejection, antibody-mediated rejection, and pros- pects for integrative endpoints for next-generation clinical trials. and Vitaeris. Mark Haas received consulting fees from Shire Am J Transplant. 2018;18(2):293-307. ViroPharma, AstraZeneca, Novartis, and CareDx, and honoraria 6. Nickeleit V, Singh HK, Randhawa P, et al. The Banff Working from CareDx. Denis Glotz received honoraria from Sanofi and CSL Group classification of definitive polyomavirus nephropa- Behring and consulting fees from BMS and Atara. Roslyn Mannon thy: morphologic definitions and clinical correlations. JASN. 2018;29(2):680-693. has received grant funding from CSL Behring, CareDX, Vitaeris, 7. Becker JU, Chang A, Nickeleit V, Randhawa P, Roufosse C. Banff Astellas, and Transplant Genomics, and honoraria from Hansa, borderline changes suspicious for acute T cell-mediated rejection: Novartis, CSL Behring, and Vitaeris. Robert Colvin is a consult- where do we stand? Am J Transplant. 2016;16(9):2654-2660. ant for Shire ViroPharma, CSL Behring, Alexion, and eGenesis. 8. Nankivell BJ, Agrawal N, Sharma A, et al. The clinical and patho- logical significance of borderline T cell-mediated rejection. Am J AJ Demetris is consultant for Novartis and TransMedics, Robert Transplant. 2019;19(5):1452-1463. Montgomery received honoraria, consulting fees, or travel ex- 9. Nankivell BJ, P’Ng CH, Chapman JR. Does tubulitis without intersti- penses from Alexion, Hansa Medical, CSL Behring, Sanofi, tial inflammation represent borderline acute T cell mediated rejec- Novartis, Viela Bio, Vitaeris Bio, Terasaki Foundation, and Shire/ tion? Am J Transplant. 2019;19(1):132-144. 10. Schinstock CA, Sapir-Pichhadze R, Naesens M, et al. Banff survey Takeda. on antibody-mediated rejection clinical practices in kidney trans- plantation: diagnostic misinterpretation has potential therapeutic DATA AVA I L A B I L I T Y S TAT E M E N T implications. Am J Transplant. 2019;19(1):123-131. The authors confirm that the data supporting the findings of 11. Halloran PF, Matas A, Kasiske BL, Madill-Thomsen KS, Mackova M, Famulski KS. Molecular phenotype of kidney transplant indica- this study are available within the article or its supplementary tion biopsies with inflammation in scarred areas. Am J Transplant. material. 2019;19(5):1356-1370. LOUPY et al. | 2331 12. Nankivell BJ, Shingde M, Keung KL, et al. The causes, significance 24. Lefaucheur C, Viglietti D, Bouatou Y, et al. Non-HLA agonistic and consequences of inflammatory fibrosis in kidney transplanta- anti-angiotensin II type 1 receptor antibodies induce a distinctive tion: the Banff i-IFTA lesion. Am J Transplant. 2018;18(2):364-376. phenotype of antibody-mediated rejection in kidney transplant re- 13. Racusen LC, Solez K, Colvin RB, et al. The Banff 97 working classifi- cipients. Kidney Int. 2019;96(1):189-201. cation of renal allograft pathology. Kidney Int. 1999;55(2):713-723. 25. Reeve J, Böhmig GA, Eskandary F, et al. Generating automated kid- 14. Racusen LC, Colvin RB, Solez K, et al. Antibody-mediated rejection ney transplant biopsy reports combining molecular measurements criteria - an addition to the Banff 97 classification of renal allograft with ensembles of machine learning classifiers. Am J Transplant. rejection. Am J Transplant. 2003;3(6):708-714. 2019;19(10):2719-2731. 15. Haas M. The revised (2013) Banff classification for antibody-me- 26. Adam B, Afzali B, Dominy KM, et al. Multiplexed color-coded diated rejection of renal allografts: update, difficulties, and future probe-based gene expression assessment for clinical molecular considerations. Am J Transplant. 2016;16(5):1352-1357. diagnostics in formalin-fixed paraffin-embedded human renal al- 16. De Serres SA, Noël R, Côté I, et al. 2013 Banff criteria for chronic lograft tissue. Clin Transplant. 2016;30(3):295-305. active antibody-mediated rejection: assessment in a real-life set- 27. Dominy KM, Willicombe M, Al Johani T, et al. Molecular assessment ting. Am J Transplant. 2016;16(5):1516-1525. of C4d-positive renal transplant biopsies without evidence of rejec- 17. Bagnasco SM, Zachary AA, Racusen LC, et al. Time course tion. Kidney Int Rep. 2019;4(1):148-158. of pathologic changes in kidney allografts of positive cross- 28. Marsh JN, Matlock MK, Kudose S, et al. Deep learning global glo- match HLA-incompatible transplant recipients. Transplantation. merulosclerosis in transplant kidney frozen sections. IEEE Trans Med 2014;97(4):440-445. Imaging. 2018;37(12):2718-2728. 18. Haas M. The relationship between pathologic lesions of active 29. Lituiev DS, Cha SJ, Chin A, et al. Automated localization and seg- and chronic antibody-mediated rejection in renal allografts. Am J mentation of mononuclear cell aggregates in kidney histological Transplant. 2018;18(12):2849-2856. images using deep learning. medRxiv. 2019;19002634. https://doi. 19. Mannon RB. The Banff schema for antibody-mediated rejection: org/10.1101/19002634 lost in translation? Am J Transplant. 2019;19(1):9-10. 30. Hermsen M, de Bel T, den Boer M, et al. Deep learning-based 20. Bajema IM, Wilhelmus S, Alpers CE, et al. Revision of the histopathologic assessment of kidney tissue. J Am Soc Nephrol. International Society of Nephrology/Renal Pathology Society clas- 2019;30(10):1968-1979. sification for lupus nephritis: clarification of definitions, and mod- 31. Aubert O, Higgins S, Bouatou Y, et al. Archetype analysis identifies ified National Institutes of Health activity and chronicity indices. distinct profiles in renal transplant recipients with transplant glo- Kidney Int. 2018;93(4):789-796. merulopathy associated with allograft survival. J Am Soc Nephrol. 21. Senev A, Coemans M, Lerut E, et al. Histological picture of anti- 2019;30(4):625-639. body-mediated rejection without donor-specific anti-HLA anti- bodies: clinical presentation and implications for outcome. Am J Transplant. 2019;19(3):763-780. How to cite this article: Loupy A, Haas M, Roufosse C, et al. 22. Sablik KA, Clahsen-van Groningen MC, Looman CWN, et al. The Banff 2019 Kidney Meeting Report (I): Updates on and Chronic-active antibody-mediated rejection with or without do- nor-specific antibodies has similar histomorphology and clinical clarification of criteria for T cell– and antibody-mediated outcome - a retrospective study. Transpl Int. 2018;31(8):900-908. rejection. Am J Transplant. 2020;20:2318–2331. https://doi. 23. Parajuli S, Redfield RR, Garg N, et al. Clinical significance of micro- org/10.1111/ajt.15898 vascular inflammation in the absence of anti-HLA DSA in kidney transplantation. Transplantation. 2019;103(7):1468-1476.