Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients

Human Immunology xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

www.ashi-hla.org

journal homepage: www.elsevier.com/locate/humimm

Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients Vivek Jani a, Elizabeth Ingulli b, Kristen Mekeel c, Gerald P. Morris a,⇑ a

Department of Pathology, University of California San Diego, La Jolla, CA, United States Department of Pediatrics, University of California San Diego, La Jolla, CA, United States c Department of Surgery, University of California San Diego, La Jolla, CA, United States b

a r t i c l e

i n f o

Article history: Received 13 July 2016 Revised 11 November 2016 Accepted 14 November 2016 Available online xxxx Keywords: Anti-HLA antibodies Virtual crossmatch Deceased donor kidney transplantation

a b s t r a c t Efficient allocation of deceased donor organs depends upon effective prediction of immunologic compatibility based on donor HLA genotype and recipient alloantibody profile, referred to as virtual crossmatching (VCXM). VCXM has demonstrated utility in predicting compatibility, though there is reduced efficacy for patients highly sensitized against allogeneic HLA antigens. The recently revised deceased donor kidney allocation system (KAS) has increased transplantation for this group, but with an increased burden for histocompatibility testing and organ sharing. Given the limitations of VCXM, we hypothesized that increased organ offers for highly-sensitized patients could result in a concomitant increase in offers rejected due to unexpectedly positive crossmatch. Review of 645 crossmatches performed for deceased donor kidney transplantation at our center did not reveal a significant increase in positive crossmatches following KAS implementation. Positive crossmatches not predicted by VCXM were concentrated among highly-sensitized patients. Root cause analysis of VCXM failures identified technical limitations of antiHLA antibody testing as the most significant contributor to VCXM error. Contributions of technical limitations including additive/synergistic antibody effects, prozone phenomenon, and antigens not represented in standard testing panels, were evaluated by retrospective testing. These data provide insight into the limitations of VCXM, particularly those affecting allocation of kidneys to highly-sensitized patients. Ó 2016 Published by Elsevier Inc. on behalf of American Society for Histocompatibility and Immunogenetics.

1. Introduction Implementation of solid-phase immunoassays (SPI) to detect anti-HLA antibodies [1,2] has revolutionized prediction of immunologic compatibility in organ transplantation. Traditionally, definitive evaluation of compatibility between patients and potential donors relies on cellular crossmatch testing (CXM) to detect antibodies against donor alloantigens [3,4]. Specificity in identifying anti-HLA antibodies afforded by SPI has enabled refinement of algorithms supporting organ allocation. Broad categorization of alloantigen sensitization by panel reactive antibody (PRA) was Abbreviations: cPRA, calculated panel reactive antibody; CXM, crossmatch; FCXM, flow cytometric crossmatch; KAS, revised kidney allocation system; MCS, mean channel shift; MFI, mean fluorescence intensity; PRA, panel reactive antibody; SPI, solid phase immunoassay; VCXM, virtual crossmatch. ⇑ Corresponding author at: Department of Pathology, University of California San Diego, 9500 Gilman Drive, MC 0612, La Jolla, CA 92093, United States. E-mail address: [email protected] (G.P. Morris).

replaced with a calculated PRA (cPRA) determined by the specific anti-HLA antibodies present and the population frequencies of target HLA antigens [5]. This permits prediction of CXM results for potential donors with known HLA genotypes, or VCXM [6,7]. Implementation of specific anti-HLA antibody information and VCXM avoidance of donors against which patients have donorspecific antibodies (DSA) has decreased deceased donor kidney offers declined due to a positive CXM concurrent with an increase in organs transplanted to highly-sensitized (cPRA P 80%) patients [8,9]. To further reduce disadvantage due to sensitization against allogeneic HLA, a revised KAS prioritizing increased allocation of organs to the most highly-sensitized (cPRA 98–100%) patients was implemented in December 2014 [10]. This has increased access for the most highly sensitized patients, with 100% cPRA patients receiving 10.3% of deceased donor kidney transplants in the new KAS era, compared to only 1.0% previously [11,12]. Ensuring equitable transplant access is important, though it comes with

http://dx.doi.org/10.1016/j.humimm.2016.11.003 0198-8859/Ó 2016 Published by Elsevier Inc. on behalf of American Society for Histocompatibility and Immunogenetics.

Please cite this article in press as: V. Jani et al., Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients, Hum. Immunol. (2016), http://dx.doi.org/10.1016/j.humimm.2016.11.003

2

V. Jani et al. / Human Immunology xxx (2016) xxx–xxx

the burden of increased transit time due to regional (increased 42%) and national (increased 50%) sharing, increased rates of delayed graft function, and a 10% increase in discard rates. These costs prompt careful evaluation of the effectiveness of the tools used. The predictive ability of VCXM is highly dependent on SPI for identification of clinically-relevant alloantibodies. In clinical practice, VCXM correctly predicts CXM results in 89–97% of cases [9,13–19]. However, VCXM accuracy can be markedly reduced for highly-sensitized patients [16]. Compared to the overall low 0.4% national rate for deceased donor kidney offer declines due to positive CXM, 6.8% of offers to patients with cPRA 80–97% and 8.3% of offers to patients with cPRA P 98% were refused due to positive CXM [9]. A recent study suggested that as much as 16% of offers to highly-sensitized patients could result in positive CXM [19]. Identifying the limitations of VCXM for highly-sensitized patients is essential for the success in allocation of organs to these difficult to match patients. VCXM limitations are well-recognized and include issues of incomplete donor HLA genotype information [14,17,19,20] and technical factors specific to SPI. Technical issues related to detection of anti-HLA antibodies by SPI include use of appropriate mean fluorescence intensity (MFI) cutoff values [21,22], additive effects of low-concentration DSA [13,22], false positive reactions against the microbeads or non-native HLA epitopes [23,24], and false negative SPI results due to inhibition by interfering substances or prozone effects [25–28]. We hypothesized that all of these factors may disproportionately affect very highly-sensitized patients and impair VCXM approaches relied upon to support the new KAS. Here, we retrospectively analyze VCXM and flow cytometric CXM (FCXM) performed for 645 deceased donor kidney offers at our institution before and after implementation of the revised KAS to identify causes underlying VCXM failures. 2. Materials and methods 2.1. Study population Histocompatibility testing performed at the University of California San Diego Immunogenetics and Transplantation Laboratory for patients with end-stage renal disease being evaluated for kidney transplantation at UCSD or Rady Children’s Hospital between December 2013 and December 2015 were included. Immunologic compatibility for patients presenting with a potential ABOcompatible living donor (n = 179) was tested by FCXM regardless of anti-HLA antibody profile. Patients registered for deceased donor kidney or kidney-pancreas transplantation had HLA serologic specificities against which the patient had clinicallyrelevant alloantibodies listed as avoid antigens in UNET. Immunologic compatibility for deceased donor kidney and kidney-pancreas transplantation was determined by prospective FCXM (n = 645). All laboratory testing was completed as part of standard of care.

alloantibody testing was performed on neat sera, or sera diluted 1:10 in PBS or treated with EDTA (5 ll 6% EDTA solution added to 95 ll serum for samples with control bead MFI values out of range) [29], using LABScreen Single Antigen HLA Class I and Class II Supplements (One Lambda) and LABScreen MICA Single Antigen (One Lambda) assays on a Labscan 100. All data were analyzed using HLA Fusion software (One Lambda). 2.3. Cellular crossmatch testing Donor lymphocytes isolated from peripheral blood, spleen, or lymph node samples by density gradient separation using RosetteSep Lymphocyte Enrichment kit (StemCell Technologies) were treated with 2 mg/ml pronase (Sigma) for 20 min at 37 °C. Donor cells were incubated in duplicate with current (typically <30 days old) and historical peak (maximum cPRA within the last 12 months) recipient serum for 20 min at room temperature. Cells were labeled with anti-CD3 PerCP (SK7, BD Biosciences) anti-CD19 PE (SJ25-C1, BD Biosciences), and goat F(ab’)2 anti-human IgG FITC (Jackson) for 20 min at 4 °C, washed, and analyzed using a FACSCalibur or FACSCanto. Alloantibody binding was determined by calculating mean channel shift (MCS) of cells incubated with patient serum as compared to cells incubated with control normal human serum; MCS P 16 was considered positive for T cell FCXM and MCS P 32 was considered positive for B cell FCXM. 2.4. HLA typing All patients and donors were typed for HLA-A, -B, -C, -DRB1, DRB3/4/5, -DQA, -DQB1, -DPA, and -DPB1. Any donors without complete typing available in UNET were re-typed in the UCSD ITL. Patient and donor typing prior to July 2015 were performed using LABType SSO (One Lambda) on a Labscan 100 and analyzed using HLA Fusion software. After July 2015, deceased donor typing was performed using the LinkSeq real-time pcr assay (Linkage Biosciences) on a Lightcycler 480 (Roche) and analyzed using SureTyper software (Linkage Biosciences). Ambiguities in HLA tying were resolved using MicroSSP Allele Specific trays (One Lambda). 2.5. Statistical analyses Proportional data among groups were analyzed using Fisher’s exact test. Sensitivity and specificity were calculated using receiver-operator curve analyses. Correlation of DSA MFI and FCXM MCS was performed by linear regression and analyzed by goodness of fit. All statistical analyses were performed using Prism 6 (Graph Pad). 3. Results 3.1. Revised KAS increased kidney offers to very highly-sensitized (cPRA P 98%) patients but did not result in statistically significant increases in positive FCXM overall

2.2. Alloantibody testing All patients were tested for anti-HLA antibodies by SPI as part of standard of care treatment. Patients were screened for anti-HLA antibodies by FlowPRA Class I and Class II bead-based flow cytometric assays (One Lambda) using FACSCanto or FACSCalibur instruments (BD). Anti-HLA antibodies detected by FlowPRA were identified using LABScreen Single Antigen HLA Class I and Class II bead assays using a Labscan 100 (Luminex). Normalized mean fluorescence intensity (MFI) values of 500 were used for the limit of detection for SAB. Normalized MFI values of 3000 were used to identify alloantibodies predicted to cause a positive CXM. Extended

We evaluated the impact of the revised KAS by examining offers for deceased donor kidneys to patients at the kidney transplant programs at UCSD and Rady Children’s Hospital San Diego. Policy at our centers is to list HLA antigens as unacceptable in UNET when the average normalized SPI bead MFI for a given antigen is >3000. This leverages UNET to perform VCXM (with the caveat of not accounting for information for HLA-DPB1, -DQA1, and -DPA1 prior to 1/21/2016, or alloantibodies not listed in UNET). Alloantibody profiles for patients on deceased donor crossmatch lists for the first 13 months of the revised KAS (12/4/2014–12/31/2015) were compared with patients receiving an offer (based on FCXM) for

Please cite this article in press as: V. Jani et al., Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients, Hum. Immunol. (2016), http://dx.doi.org/10.1016/j.humimm.2016.11.003

V. Jani et al. / Human Immunology xxx (2016) xxx–xxx

deceased donor kidneys in a comparable prior time period (1/1/2014–11/30/2014) (Fig. 1A). KAS changes resulted in deceased donor offers to patients with cPRA 98–100% increasing from 1.1% to 6.0% (P = 0.001). A similar change was not observed in FCXM performed for evaluating potential living donor transplants (9.2% to 1.9%, P = 0.028), indicating that this increase resulted from the revised KAS rather than a change in our patient population. Concomitantly, we observed a modest increase in the frequency of unexpectedly positive FCXM results for deceased donor kidney offers (13.9% vs. 10.1%, in KAS versus pre-KAS; P > 0.05) (Fig. 1B). Still, the revised KAS policy resulted in an increase in transplants to the most highly sensitized (cPRA 98–100%) patients (n = 0 preKAS vs. n = 5 KAS, P = 0.06) (Fig. 1C). While these increases did not reach statistical significance, they prompted examining which patient populations were most affected. Positive FCXM results were concentrated among highly sensitized patients (cPRA > 80%) in both the pre-KAS and KAS eras (60.7% vs. 51.0%, P > 0.05) (Fig. 1D). The proportion of unexpected positive FCXM results from the most highly sensitized patients (cPRA P 98%) increased from 10.7% to 21.6%, though this is in line with the increase in deceased donor kidney allocation and reflects limits of the VCXM (Fig. 1A). As there did not appear to be an eraspecific effect on the relationship between HLA sensitization and unexpected positive FCXM, we analyzed combined pre-KAS and KAS era data to evaluate likelihood for positive FCXM based on patient HLA sensitization. While the overall incidence of positive FCXM for deceased donors was 12.2%, this varied widely depending upon HLA sensitization (Fig. 1E). Patients with no or moderate HLA sensitization (cPRA 0–79%) were significantly less likely to have an unexpectedly positive FCXM as compared to highly (cPRA 80–97%) and very highly (cPRA 98–100%) sensitized patients (5.8% vs. 60.4% vs. 68% respectively, P < 0.0001). This illustrates challenges specific to identifying immunologically compatible donors for highlysensitized patients.

3

3.2. VCXM accurately predicts FCXM in the majority of patients Performance characteristics of SPI and VCXM were evaluated using living and deceased kidney donor FCXM and SPI data across both pre-and post-KAS eras in aggregate. VCXM has excellent predictive value for FCXM as evidenced by receiver operator characteristic analysis (AUC = 0.859, 95% CI 0.818–0.900) (Fig. 2A). The 3000 MFI cutoff used in our institution provides an overall accuracy of 92.0% for the VCXM, with sensitivity of 47.3% and specificity of 98.6% for identifying anti-HLA antibodies likely to cause positive FCXM (Fig. 2B), supporting VCXM utility as a screening tool for predicting immunologic compatibility. To examine the in-practice performance of VCXM, we compared the predictive power for living and deceased kidney donor-recipient pairs (Fig. 2B). In the living kidney donor population, where no VCXM is performed prior to FCXM for ABO-compatible donor-recipient pairs, this results in a positive predictive value (PPV) of 90.1% and negative predictive value (NPV) of 94.9%. For deceased donor kidney offers, where VCXM is performed by listing HLA antigens where the patients has anti-HLA antibodies with MFI P 3000 as avoids in UNET, we observe a nominal decrease in the predictive power. While the PPV remains high (92.0%) the NPV decreases (88.3%). This likely reflects a change in the composition of the pool of tested donorrecipient pairs; the VCXM algorithm removes pairs with a high likelihood of positive FCXM, leaving only pairs expected to be compatible among which unexpectedly positive FCXM results could occur (depressing the NPV). DSA MFI values were compared with corresponding FCXM MCS values for all samples tested to evaluate potential relationships between DSA MFI and FCXM (Fig. 2C and D). DSA MFI had poor correlation with both T cell (R2 = 0.493) and B cell (R2 = 0.571) FCXM MCS. Positive T cell and B cell FCXM results not predicted by DSA MFI P 3000 were observed across a wide range of MCS values, implying that they included both low-titer and high-titer alloanti-

Fig. 1. FCXM testing for potential living and deceased donor kidney transplantation before and after implementation of revised KAS. (A) Comparison of FCXM performed preKAS (12/2013–11/2014) and post-KAS (12/2014–12/2015) for deceased donor kidney and potential living donor kidney transplantation. (B) Frequencies of positive FCXM among testing for deceased donor kidney transplantation. (C) Transplantation of deceased donor kidneys at our center during the study period. (D) Allocation of positive FCXM results according to patient HLA sensitization by era. (E) Frequency of positive FCXM not predicted by VCXM among all FCXM test results for each cPRA group.

Please cite this article in press as: V. Jani et al., Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients, Hum. Immunol. (2016), http://dx.doi.org/10.1016/j.humimm.2016.11.003

4

V. Jani et al. / Human Immunology xxx (2016) xxx–xxx

Fig. 2. SPI accurately identifies DSA causing positive FCXM. (A) Receiver-operator curve analysis of SPI DSA for prediction of FCXM. DSA with maximum MFI was used for samples with multiple DSA. (B) SPI sensitivity and specificity were calculated from all FCXM data using a 3000 MFI cutoff for identifying anti-HLA antibodies by SPI. VCXM predictive values calculated independently for living donor and deceased donor FCXM. (C, D) Relationship between DSA SPI MFI and FCXM MCS for T cell (C) and B cell (D) FCXM.

bodies and were not ‘‘near miss” false negatives with MFI or MCS values near the empirically-determined cutoff values. Conversely, DSA associated with negative FCXM results had MFI values <5000 in nearly all (83/85, 97.7%) cases.

3.3. Technical aspects of SPI underlie majority of VCXM failures To evaluate causes of VCXM failures and determine whether they disproportionately affect highly sensitized patients, we performed root cause analysis of positive FCXM results for deceased donor kidney offers (n = 79). VCXM results were based on analysis of historical peak serum used in performance of FCXM. Four main

causes for false negative VCXM results were identified (Fig. 3); anti-HLA antibodies not listed in UNET (either data entry error or not listed at discretion of transplant program), anti-HLA antibodies with allelic specificity and unable to be listed in UNET, anti-HLADP antibodies (which were not included in UNET for this study period), and false negative SPI results (including antibodies against non-HLA antigens). The first three categories are unique to VCXM for deceased donor organ allocation, as they relate to the use and functionality of UNET for VCXM. Positive FCXM results in these categories highlight the importance of complete and accurate information in VCXM. However, a majority (56.1%) of VCXM failures for deceased donor kidney FCXM occurred due to false negative SPI results, independent of cPRA (P > 0.05). False negative VCXM results for living donor kidney transplantation (n = 6) occurred in all three cPRA groups. Negative FCXM in the setting of positive VCXM results (n = 4 deceased donor, n = 2 living donor) all occurred from low MFI (MFI 3000–6800) anti-HLA class II antibodies (3 anti-DQ, 1 anti-DQA, 1 anti-DR, and 1 anti-DR53). 3.4. Cumulative effects of multiple low-MFI DSA

Fig. 3. Causes of unexpected positive FCXM for deceased donor kidney offers. Percentage of positive FCXM results (n = 79) for all deceased donor kidney offers attributable to 4 primary causes; anti-HLA antibodies detectable by SPI but not listed in UNET, anti-HLA antibodies with allelic specificities that could not be listed in UNET, anti-HLA-DP antibodies that could not be listed in UNET, and antibodies not detected by standard SPI testing. Results are given as percentage of FCXM tests performed for each cPRA group.

Multiple low-titer DSAs can have additive or synergistic effects and result in positive CXM [13,22,30]. To address the possibility for cumulative effects of multiple low-MFI DSA, we re-analyzed SPI results using the sum MFI for all DSA in both living and deceased donor FCXM. We observed 23 cases where max MFI < 3000 and sum MFI P 3000 correlated with a positive FCXM. Conversely, we observed 21 cases of a negative FCXM in the setting of max MFI < 3000 and sum MFI P 3000. Both of these cases were more frequent in highly sensitized patients (Fig. 4A), but also occurred in low cPRA patients. Negative FCXM with positive VCXM using sum MFI occurred about 3-times as frequently when the multiple DSA were restricted to class II HLA antigens as compared to the

Please cite this article in press as: V. Jani et al., Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients, Hum. Immunol. (2016), http://dx.doi.org/10.1016/j.humimm.2016.11.003

V. Jani et al. / Human Immunology xxx (2016) xxx–xxx

5

Fig. 4. Multiple low MFI DSA explains some cases of positive FCXM with negative VCXM. VCXM results for all patients were re-analyzed using total sum MFI of all DSA by SPI as compared to FCXM. (A) Ability of VCXM using sum MFI of all DSA to predict positive FCXM when VCXM based on single maximum DSA MFI did not predict positive FCXM. Results compared to frequency of negative FCXM among positive VCXM using sum MFI of all DSA. Data are shown as percentage of samples tested in each cPRA category. (B) Frequency of negative FCXM with positive VCXM using sum MFI P 3000 for DSA against class I and class II HLA among all FCXM results. (C, D) Relationship between DSA SPI sum MFI and FCXM MCS for T cell (C) and B cell (D) FCXM. (E) Receiver-operator curve analysis of SPI sum DSA MFI for prediction of FCXM. (F) Sensitivity and specificity using sum DSA MFI P 3000 were calculated from all FCXM data. VCXM predictive values calculated independently for living donor and deceased donor FCXM.

cases where anti-class I antibodies were also included (Fig. 4B). Overall, additive/synergistic multiple DSA effect accounted for 27% of VCXM failures in the study period, highlighting the importance of this effect as a limitation of VCXM. Given the significant proportion of VCXM failures attributable to multiple low-titer DSA, we re-analyzed VCXM performance using the sum MFI for all DSA. Comparison of sum DSA MFI values with corresponding FCXM MCS values (Fig. 4C and D), demonstrated poor correlation with both T cell (R2 = 0.453) and B cell (R2 = 0.509) FCXM MCS. ROC analysis (Fig.4E) demonstrated similar overall diagnostic ability for sum MFI in predicting FCXM (AUC = 0.869, 95% CI 0.819–0.901), with an increase in sensitivity

(62.6%) but a concurrent decrease in specificity (97.2%) (Fig. 4F). This significantly reduced the overall clinical utility for predicting compatibility, with the modest improvements in NPV offset by large decreases in PPV in both living and deceased donor crossmatch settings. 3.5. Prozone effect false negative SPI results limit VCXM The sandwich immunoassay format of SPI is subject to false negative results due to prozone effects caused by analyte concentrations greatly exceeding the testable range of the assay [28,30]. We hypothesized that highly-sensitized patients may be particu-

Please cite this article in press as: V. Jani et al., Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients, Hum. Immunol. (2016), http://dx.doi.org/10.1016/j.humimm.2016.11.003

6

V. Jani et al. / Human Immunology xxx (2016) xxx–xxx

larly affected by this. To evaluate the impact of prozone effect, we tested 102 samples by SPI using 1:10 dilution of serum. Anti-HLA antibodies not identified from testing neat sera were detected at MFI P 3000 in 22 (21.6%) diluted serum samples (Fig. 5A). Prozone effect false negative results disproportionately affected samples with high cPRA values, with 25.0% and 41.7% of samples from patients with 80–97% and 98–100% cPRA respectively demonstrating novel anti-HLA antibodies by SPI using 1:10 serum dilution, compared to only 8.7% of samples with cPRA < 80% (P = 0.005). Samples with prozone effect demonstrated 6.6 ± 6.7 (range 1– 25) new anti-HLA antibodies/sample at 1:10 serum dilution (Fig. 5B). False negative SPI antibodies were found across HLA groups, though anti-HLA-A (n = 63), anti-HLA-B (n = 39), and anti-HLA-DQ (n = 25) accounted for a majority of the 152 prozone antibodies identified (Fig. 5C). Novel DSA detected by SPI at 1:10 serum dilution but not with neat serum explained 8 (21.0%) of the VCXM failures observed during the study period; 2/6 (33.3%) living donor VCXM and 6/32 (18.8%) deceased donor VCXM failures attributed to SPI. 3.6. Standard SPI does not detect all alloantibodies present capable of causing positive FCXM While the highly multiplexed nature of SPI enables detection of antibodies recognizing a wide range of HLA antigens covering identified serologic specificities, genetic diversity of HLA precludes testing all HLA alleles. To determine if antibodies against HLA alleles not represented in standard SPI assays contributed to VCXM failure to predict FCXM, we analyzed 19 serum samples associated with unexpectedly positive deceased donor kidney FCXM using extended HLA antigen SPI panels. Six (31.6%) samples demonstrated novel anti-HLA allele reactivity not predicted by standard SPI (Fig. 6A). Notably, all novel anti-HLA reactivity was detected in high cPRA (P80%) samples, with 5/8 (62.5%) of 80–97% cPRA samples and 1/2 (50.0%) of 98–100% cPRA samples demonstrating additional reactivity. Extended HLA SPI identified between 1 and 6 new anti-HLA specificities/sample against HLA-A, -B, -C, -DR, and DP alleles (Fig. 6B). Importantly, 2 of these novel anti-HLA antibodies explained positive FCXM results not predicted by standard SPI. While the majority of histocompatibility testing is directed toward identifying and avoiding antibodies against HLA alloantigens, antibodies against non-HLA antigens can cause allograft rejection. In particular, antibodies against MICA alloantigens have been demonstrated to cause positive CXM and AMR in the absence of anti-HLA antibodies [31]. We examined 19 serum samples associated with positive FCXM but without identified anti-HLA DSA for the presence of anti-MICA antibodies. Only 1/19 (5.3%) samples

tested demonstrated anti-MICA antibodies (Fig. 6C). Additional testing revealed an antibody against HNA3a in 1 sample. Both of these non-HLA antibodies were in samples with low cPRA (<55%) values. While only 2 samples tested demonstrated non-HLA alloantibodies, they were associated with 11 positive FCXM results not predicted by VCXM. Though we cannot definitively attribute positive FCXM results for these samples to non-HLA alloantibodies due to a lack of information of donor genotype, these results indicate that non-HLA antibodies are present among patients awaiting deceased donor kidney transplantation and may limit VCXM prediction of compatibility. 4. Discussion VCXM approaches are essential to support the revised KAS prioritizing difficult-to-match patients. However, SPI and VCXM are imperfect, as evidenced by a persistence of kidney offers ultimately rejected or re-directed due to immunologic incompatibility. These limitations have important implications for deceased donor organ allocation strategies; restrictive VCXM criteria ensuring immunologic compatibility could unnecessarily exclude potentially compatible transplants (particularly for highly-sensitized patients). Conversely, less restrictive VCXM approaches have higher risk of unpredicted incompatibility and resulting organ redirection, discard, or rejection. Not unexpectedly, it is more difficult to accurately predict CXM results for highly-sensitized patients [9,16,19]. Given the considerable effort put forth to increasing organ access for these patients, we performed a critical evaluation of VCXM. While the revised KAS resulted in an increase in transplantation for the most highly sensitized patients, we observed a modest increase in unexpected positive FCXM results disproportionately concentrated among highly-sensitized patients (Fig. 1). Root cause analysis identified false negative SPI results as the underlying cause for a majority of positive FCXM results (Fig. 2). SPI is highly accurate for identification of anti-HLA antibodies capable of causing positive FCXM, though cutoff value selection can affect interpretation [21,22]. Consensus guidelines suggest selection of cutoff values that differentiate biologically relevant alloantibodies from among those detectable by highly-sensitive SPI [32]. Our SPI cutoff value of 3000 MFI for identifying antibodies expected to cause positive FCXM is designed to do this, minimizing unexpectedly positive FCXM results without unnecessarily limiting potential organ offers. An additional complication may be differences between HLA loci as targets for alloantibodies causing positive cellular CXM, based on expression on the cell surface. Our study is unable to address this specific limitation, as we did not have enough positive FCXM results attributable to single DSA to

Fig. 5. Prozone effect can cause false negative SPI results inhibiting VCXM. Prozone effect was examined in 102 samples by testing serum diluted 1:10 by SPI. (A) Incidence of prozone effect false negative SPI results, defined as detection of anti-HLA antibodies in 1:10 diluted serum samples that were not detected by SPI using neat serum. Results shown as samples negative or positive for detecting prozone effect antibodies in each cPRA group. (B) The number of new antibodies detected in each sample positive for prozone effect. (C) Anti-HLA specificity of new antibodies identified in all samples with prozone effect.

Please cite this article in press as: V. Jani et al., Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients, Hum. Immunol. (2016), http://dx.doi.org/10.1016/j.humimm.2016.11.003

V. Jani et al. / Human Immunology xxx (2016) xxx–xxx

7

Fig. 6. Extended SPI testing identifies alloantibodies against HLA and non-HLA antigens associated with positive FCXM. Nineteen samples associated with positive FCXM and negative VCXM were tested for alloantibodies with extended SPI. (A) Identification of new anti-HLA antibodies using SPI with extended HLA phenotype panels. Results shown as samples negative or positive for detecting prozone effect antibodies in each cPRA group. (B) SPI reactivity against standard and extended phenotype HLA panels for samples with new anti-HLA antibodies detected by extended phenotype testing. Molecular genotype of recombinant HLA antigen used in assay shown. Reaction patterns found in multiple samples indicated by ‘‘x2”. New anti-HLA antibodies representing DSA causing positive FCXM not predicted from standard SPI indicated with arrows. (C) Identification of anti-MICA antibodies using SPI. Results shown as samples negative or positive for detecting prozone effect antibodies in each cPRA group.

enable this analysis. While there is agreement that antibodies detected by SPI causing positive CXM present a barrier to transplantation [32], correlation between SPI results and eventual AMR is unclear. Reports have described antibodies detectable by SPI but not CXM that do not appear to impair allograft outcomes [14,33,34], and conversely anti-HLA antibodies not causing positive CXM but detectable by SPI predictive of inferior long-term kidney transplant outcomes [15,35,36]. Considerable debate exists as to methods for differentiating alloantibodies that present a barrier to transplant from those that do not. While several groups have demonstrated some utility in modified SPI including a functional component (ie. binding complement components), stratification of anti-HLA antibodies based on SPI MFI is a robust method for identifying antibodies causing positive FCXM and precluding transplantation [28,37–39]. Beyond MFI cutoff value determination, technical limitations of SPI, including effects from additive/synergistic low-titer DSA, false negative results due to prozone effect and interfering substances, and incomplete representation of all possible alloantigens, have been described [13,22–28]. In retrospective examination of the 44 deceased donor positive FCXM results not predicted by our VCXM approach, we were able to resolve the underlying cause in 42 (95.5%) cases. The largest source of difficulty in predicting FCXM was additive/synergistic DSA, which accounted for 52.2% of VCXM failures (Fig. 4). Other groups have proposed using sum MFI of DSA to reduce this source of false negative VCXM prediction [13,22]. This information is certainly useful when interpreting FCXM, though our results suggest that the benefit of this approach in VCXM is limited due to an unacceptable increase in VCXM incorrectly predicting positive FCXM, which would unnecessarily limit transplantation of highly-sensitized patients. Of samples tested by extended SPI, antibodies against HLA antigens not included in standard SPI were the most common finding (31.6%) (Fig. 6A), with prozone effect antibodies (21.6%) (Fig. 5A), and non-HLA antibodies (10%) (Fig. 6C) also identified. While we could not definitively demonstrate that antibodies against nonHLA antigens were responsible for positive FCXM (due to not having donor genotyping), we found antibodies against non-HLA antigens were associated with 25.0% of unexplained positive FCXM. It is noteworthy that these limitations disproportionately affected highly-sensitized (cPRA P 80%) patients. The relatively high incidence of prozone effect and antibodies against HLA antigens not in standard SPI panels supports consideration of extended SPI testing in standard practice. While larger studies are necessary to determine optimal utility of extended SPI testing and its inclusion in VCXM algorithms, our data demonstrating a disproportionate

effect of these phenomenon on highly-sensitized patients suggests an approach focusing on extended testing for these patients may be appropriate. We have implemented extended SPI testing for high-risk (cPRA P 80%) patients in our program. VCXM is an important and powerful tool for organ allocation, and is increasingly used by transplant programs as a primary means for prospective assessment of immunologic compatibility [17,38]. However, the potential impact of inaccurate VCXM cannot be underestimated. Although the revised KAS has increased access for difficult-to-match patients, it has the expense of increased organ sharing and transit time with associated increases in delayed graft function and increased organ redirection or discard rates [11,12]. Our data demonstrates the potential limitations of this approach and suggests thorough consideration of limitations in SPI and VCXM are essential before considering to proceed with transplantation in the absence of prospective CXM. Disclosure The authors of this manuscript have no conflicts of interest to disclose. Authorship VJ, LI, KM, and GPM were responsible for study concept, design, data analysis, and manuscript preparation. VJ and LI performed extended SPI testing. All authors reviewed and approved the final version of the manuscript. Funding No funding was required for this study. Acknowledgements The authors would like to thank Barry Falch, Mitra YazdaniBuicky, and Lain Tan for assistance in performing additional serologic testing. References [1] R. Pei, J. Lee, T. Chen, S. Rojo, P.I. Terasaki, Flow cytometric detection of HLA antibodies using a spectrum of microbeads, Human Immunol. 60 (1999) 1293– 1302.

Please cite this article in press as: V. Jani et al., Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients, Hum. Immunol. (2016), http://dx.doi.org/10.1016/j.humimm.2016.11.003

8

V. Jani et al. / Human Immunology xxx (2016) xxx–xxx

[2] R. Pei, L.H. Lee, N.J. Shih, M. Chen, P.I. Terasaki, Single human leukocyte antigen flow cytometry beads for accurate identification of human leukocyte antigen antibody specificities, Transplantation 75 (2003) 43–49. [3] R. Patel, P.I. Terasaki, Significance of the positive crossmatch test in kidney transplantation, N. Engl. J. Med. 280 (1969) 735–739. [4] Y. Iwaki, M. Lau, D.J. Cook, S. Takemoto, P.I. Terasaki, Crossmatching with B and T cells and flow cytometry, Clin. Transpl. 1 (1986) 277–284. [5] J.M. Cecka, Calculated PRA (CPRA): the new measure of sensitization for transplant candidates, Am. J. Transplant. 10 (2010) 26–29. [6] S.D. Zangwill, T.M. Ellis, J. Zlotocha, R.D. Jaquiss, J.S. Tweddell, K.A. Mussatto, et al., The virtual crossmatch–a screening tool for sensitized pediatric heart transplant recipients, Pediatr. Transplant. 10 (2006) 38–41. [7] R.A. Bray, J.D. Nolen, C. Larsen, T. Pearson, K.A. Newell, K. Kokko, et al., Transplanting the highly sensitized patient: the emory algorithm, Am. J. Transplant. 6 (2006) 2307–2315. [8] J.M. Cecka, A.Y. Kucheryavaya, N.L. Reinsmoen, M.S. Leffell, Calculated PRA: initial results show benefits for sensitized patients and a reduction in positive crossmatches, Am. J. Transplant. 11 (2011) 719–724. [9] L.A. Baxter-Lowe, A. Kucheryavaya, D. Tyan, N. Reinsmoen, CPRA for allocation of kidneys in the US: more candidates P98% CPRA, lower positive crossmatch rates and improved transplant rates for sensitized patients, Human Immunol. 77 (2016) 395–402. [10] A.J. Matas, J.M. Smith, M.A. Skeans, B. Thompson, S.K. Gustafson, D.E. Stewart, et al., OPTN/SRTR 2013 annual data report: kidney, Am. J. Transplant. 15 (S2) (2015) 1–34. [11] A.B. Massie, X. Luo, B.E. Lonze, N.M. Desai, A.W. Bingaman, M. Cooper, et al., Early changes in kidney distribution under the new allocation system, J. Am. Soc. Nephrol. 17 (2015). [12] D.E. Stewart, A.Y. Kucheryavaya, D.K. Klassen, N.A. Turgeon, R.N. Formica, M.I. Aeder, Changes in deceased donor kidney transplantation one year after KAS implementation, Am. J. Transplant. 16 (2016) 1834–1847. [13] A.R. Tambur, D.S. Ramon, D.B. Kaufman, J. Friedewald, X. Luo, B. Ho, et al., Perception versus reality?: virtual crossmatch–how to overcome some of the technical and logistic limitations, Am. J. Transplant. 9 (2009) 1886–1893. [14] G.P. Morris, D.L. Phelan, M.D. Jendrisak, T. Mohanakumar, Virtual crossmatch by identification of donor-specific anti-human leukocyte antigen antibodies by solid-phase immunoassay: a 30-month analysis in living donor kidney transplantation, Human Immunol. 71 (2010) 268–273. [15] D. Bielmann, G. Honger, D. Lutz, M.J. Mihatsch, J. Steiger, S. Schaub, Pretransplant risk assessment in renal allograft recipients using virtual crossmatching, Am. J. Transplant. 7 (2007) 626–632. [16] A.W. Bingaman, C.L. Murphey, J. Palma-Vargas, F. Wright, A virtual crossmatch protocol significantly increases access of highly sensitized patients to deceased donor kidney transplantation, Transplantation 86 (2008) 1864– 1868. [17] L.A. Baxter-Lowe, M. Cecka, M. Kamoun, J. Sinacore, M.L. Melcher, Centerdefined unacceptable HLA antigens facilitate transplants for sensitized patients in a multi-center kidney exchange program, Am. J. Transplant. 14 (2014) 1592–1598. [18] C.P. Johnson, J.J. Schiller, Y.R. Zhu, S. Hariharan, A.M. Roza, D.C. Cronin, et al., Renal transplantation with final allocation based on the virtual crossmatch, Am. J. Transplant. 16 (2016) 1503–1515. [19] A.R. Tambur, K.M. Haarberg, J.J. Friedewald, J.R. Leventhal, M.F. Cusick, A. Jaramillo, et al., Unintended consequences of the new national kidney allocation policy in the United States, Am. J. Transplant. 15 (2015) 2465–2469. [20] A. Piazza, G. Ozzella, E. Poggi, D. Caputo, A. Manfreda, D. Adorno, Virtual crossmatch in kidney transplantation, Transplant. Proc. 46 (2014) 2195–2198. [21] C. Liu, L. Wetter, S. Pang, D.L. Phelan, T. Mohanakumar, G.P. Morris, Cutoff values and data handling for solid-phase testing for antibodies to HLA: effects

[22]

[23]

[24] [25]

[26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

[38] [39]

on listing unacceptable antigens for thoracic organ transplantation, Human Immunol. 73 (2012) 597–604. T.M. Ellis, J.J. Schiller, A.M. Roza, D.C. Cronin, B.D. Shames, C.P. Johnson, Diagnostic accuracy of solid phase HLA antibody assays for prediction of crossmatch strength, Human Immunol. 73 (2012) 706–710. S. Pereira, S. Perkins, J.H. Lee, W. Shumway, W. LeFor, M. Lopez-Cepero, et al., Donor-specific antibody against denatured HLA-A1: clinically nonsignificant?, Human Immunol 72 (2011) 492–498. E.K. Jacob, S.R. De Goey, M.J. Gandhi, Positive virtual crossmatch with negative flow crossmatch results in two cases, Transpl. Immunol. 25 (2011) 77–81. A.A. Zachary, D.P. Lucas, B. Detrick, M.S. Leffell, Naturally occurring interference in Luminex assays for HLA-specific antibodies: characteristics and resolution, Human Immunol. 70 (2009) 496–501. G. Guidicelli, G. Anies, T. Bachelet, V. Dubois, J.F. Moreau, P. Merville, et al., The complement interference phenomenon as a cause for sharp fluctuations of serum anti-HLA antibody strength in kidney transplant patients, Transpl. Immunol. 29 (2013) 17–21. E. Schwaiger, M. Wahrmann, G. Bond, F. Eskandary, G.A. Bohmig, Complement component C3 activation: the leading cause of the prozone phenomenon affecting HLA antibody detection on single-antigen beads, Transplantation 97 (2014) 1279–1285. A.R. Tambur, N.D. Herrera, K.M. Haarberg, M.F. Cusick, R.A. Gordon, J.R. Leventhal, et al., Assessing antibody strength: comparison of MFI, C1q, and titer information, Am. J. Transplant. 15 (2015) 2421–2430. M. Schnaidt, C. Weinstock, M. Jurisic, B. Schmid-Horch, A. Ender, D. Wernet, HLA antibody specification using single-antigen beads–a technical solution for the prozone effect, Transplantation 92 (2011) 510–515. M. Wahrmann, G. Hlavin, G. Fischer, L. Marinova, E. Schwaiger, W.H. Horl, et al., Modified solid-phase alloantibody detection for improved crossmatch prediction, Human Immunol. 74 (2013) 32–40. Y. Zou, P. Stastny, C. Susal, B. Dohler, G. Opelz, Antibodies against MICA antigens and kidney-transplant rejection, N. Engl. J. Med. 357 (2007) 1293– 1300. B.D. Tait, C. Susal, H.M. Gebel, P.W. Nickerson, A.A. Zachary, F.H. Claas, et al., Consensus guidelines on the testing and clinical management issues associated with HLA and non-HLA antibodies in transplantation, Transplantation 95 (2013) 19–47. P. Amico, G. Honger, M. Mayr, J. Steiger, H. Hopfer, S. Schaub, Clinical relevance of pretransplant donor-specific HLA antibodies detected by single-antigen flow-beads, Transplantation 87 (2009) 1681–1688. T.B. Dunn, H. Noreen, K. Gillingham, D. Maurer, O.G. Ozturk, T.L. Pruett, et al., Revisiting traditional risk factors for rejection and graft loss after kidney transplantation, Am. J. Transplant. 11 (2011) 2132–2143. J.M. Gloor, J.L. Winters, L.D. Cornell, L.A. Fix, S.R. DeGoey, R.M. Knauer, et al., Baseline donor-specific antibody levels and outcomes in positive crossmatch kidney transplantation, Am. J. Transplant. 10 (2010) 582–589. H.G. Otten, M.C. Verhaar, H.P. Borst, R.J. Hene, A.D. van Zuilen, Pretransplant donor-specific HLA class-I and -II antibodies are associated with an increased risk for kidney graft failure, Am. J. Transplant. 12 (2012) 1618–1623. M. Crespo, A. Torio, V. Mas, D. Redondo, M.J. Perez-Saez, M. Mir, et al., Clinical relevance of pretransplant anti-HLA donor-specific antibodies: does C1qfixation matter?, Transpl Immunol. 29 (2013) 28–33. S. Schaub, G. Honger, M.T. Koller, R. Liwski, P. Amico, Determinants of C1q binding in the single antigen bead assay, Transplantation 98 (2014) 387–393. P. Amico, P. Hirt-Minkowski, G. Honger, L. Gurke, M.J. Mihatsch, J. Steiger, et al., Risk stratification by the virtual crossmatch: a prospective study in 233 renal transplantations, Transpl. Int. 24 (2011) 560–569.

Please cite this article in press as: V. Jani et al., Root cause analysis of limitations of virtual crossmatch for kidney allocation to highly-sensitized patients, Hum. Immunol. (2016), http://dx.doi.org/10.1016/j.humimm.2016.11.003