Transplanting the Highly Sensitized Patient: The Emory Algorithm

Authors


*Corresponding author: Howard M. Gebel, hgebel@emory.edu

Abstract

Renal transplant patients sensitized to HLA antigens comprise nearly one-third of the UNOS wait-list and receive 14% of deceased donor (DD) transplants, a rate half that of unsensitized patients. Between 1999 and 2003, we performed 492 adult renal transplants from DD; 120 patients (∼25%) had a panel reactive antibody (PRA) of >30%, with nearly half (n = 58) having a PRA of >80%. Our approach is based upon high-resolution solid-phase HLA antibody analysis to identify class I/II antibodies and a ‘virtual crossmatch’ to predict compatible donor/recipient combinations. Recipients are excluded from the United Network for Organ Sharing match run if donors possess unacceptable antigens. Thus, when sensitized patients appear on the match run, they have a high probability of a negative final crossmatch. Here, we describe our 5-year experience with this approach. Five-year graft survival ranged from 66% to 70% among unsensitized (n = 272), moderately sensitized (PRA < 30%, n = 100) and highly sensitized (>30% PRA; n = 120) patients, equal to the average national graft survival (65.7%). The application of this approach (the Emory Algorithm) provides a logical and systematic approach to improve the access of sensitized patients to DD organs and promote more equitable allocation to a highly disadvantaged group of patients awaiting renal transplantation.

Introduction

Following the landmark studies of Patel and Terasaki, donor-directed HLA antibodies have been recognized as a major risk factor among patients awaiting a renal allograft (1). Standard practice now calls for HLA laboratories to collect and monitor these antibodies in potential transplant recipients on a routine basis. Referred to as panel reactive antibody (PRA) analysis, this approach is used, prospectively, to determine whether a patient possesses antibodies to HLA antigens, and if so, to which specificities. The PRA value, i.e. the percentage of targets (such as HLA-typed lymphocytes) reactive with the patient's serum, is generally considered to represent a patient's ‘transplantability index’. A PRA of 80% suggests the patient's crossmatch will be positive with 4 of 5 donors. Thus, the higher the PRA value, the greater the likelihood of a positive crossmatch with a random donor and the lower the likelihood of receiving a transplant. Currently, >15 000 patients awaiting deceased donor (DD) renal transplantation in the United States are sensitized (defined as >20% PRA) and almost half of those patients (∼7500) have a PRA of >80%. If considered as an independent category, sensitized allograft candidates represent the most disadvantaged group of patients awaiting renal transplantation. The vast majority of DD kidneys are transplanted into patients with a current PRA of <20%. In fact, nationally, only 14% of DD kidneys are transplanted into sensitized patients (2).

There are two distinct approaches to transplanting sensitized patients; one is pharmacological and the other is biological. The principal pharmacological approaches utilize intravenous immunoglobulin ± plasmapheresis as a preemptive tactic to eliminate/inactivate donor-reactive HLA antibodies (3,4). Typically, these therapeutic strategies are limited to sensitized recipients with living donors and relatively low titer (<1:16) antibodies (5). The costs of these approaches can be significant and logistics have largely prevented their application in DD transplantation. Clearly, desensitization protocols cannot be applied to all potential recipients. Alternatively, patients can be transplanted across a positive crossmatch due to what many centers consider to be weak antibodies (e.g. HLA antibodies detectable only by flow cytometry) with little risk of hyperacute rejection; such patients are far from risk free. Such transplants are significantly more likely to experience early antibody-mediated rejection, acute rejection and graft loss (6–9; Campbell, et al: http://cnserver0.nkf.med.ualberta.ca/Banff/2001/Movies/Campbell1.mov). Thus, in current practice and depending on the transplant center, a positive crossmatch due to HLA antibodies may be considered a contraindication to transplantation or, at the very least, a risk factor requiring therapeutic intervention pre- and/or post-transplantation (10).

While the optimal allograft is from a HLA identical donor (such as a sibling), the likelihood of identifying such a donor is limited. For recipients without a living donor, ∼15% of all DD transplants are from zero-antigen mismatched (HLA-A, B and DR) donors. Furthermore, recipients of a zero-antigen mismatch tend to have common HLA haplotypes, meaning that patients with uncommon haplotypes (e.g. minority groups) are unlikely to be transplanted with a well-matched kidney (2). Not surprisingly, the vast majority of allografts occur among HLA nonidentical donors and recipients.

In the United States, the principal strategy to identify acceptable donors for potential kidney transplant recipients has been via a negative crossmatch. Typically, HLA laboratories maintain current sera from acceptable transplant candidates to perform preliminary cytotoxic crossmatches. Sera are plated into microtiter trays and tested against lymphocytes from ABO-compatible donors. At the very least, patients must have a negative cytotoxic crossmatch with T cells from the DD on this so-called regional organ procurement (ROP) tray before they will be considered as a recipient for an organ from that donor. Even when a recipient's serum is negative with donor T cells on the ROP trays, the patient may still not be offered the organ because of a subsequent positive crossmatch with donor B cells and/or after testing by the more sensitive flow cytometric crossmatch (FCXM). Clearly, additional information permitting a ‘virtual crossmatch’, (i.e. accurate crossmatch prediction) would be highly desirable.

The recent introduction of methodologies using solid-phase matrices coated exclusively with HLA class I or class II antigens has provided the transplant community with more sensitive tools to detect HLA antibodies. Importantly, these new technologies permit HLA antibody specificity determination that far exceeds cell-based assays (11,12). In this study, we demonstrate the feasibility of applying a simple algorithm that increases the access to and allocation of DD renal allografts to sensitized patients.

Materials and Methods

Patients

Excluding recipients of six antigen-matched/zero-antigen mismatched donors, the subjects included in this study consisted of 492 consecutive recipients (96% primary, 4% regrafts) transplanted with renal allografts from DDs between 1999 and 2003. Patients were transplanted only when flow cytometric T- and B-cell crossmatches were negative with current (day 0 of transplant) serum and at least two additional sera within the previous 12 months. Immunosuppressive protocols in use at Emory University from 1999 to 2003 were based on triple-drug therapy including glucocorticoids (prednisone), calcineurin inhibitor (cyclosporine or tacrolimus) and an antiproliferative agent (mycophenolate mofetil [MMF]). More specifically, glucocorticoids were administered as follows: 125 mg intravenously on call to the operating room and rapidly tapered over the next 3 days to 20 mg orally each day. Over the next 2 months, the dosage was gradually reduced to a final dose of 10 mg/day. For patients receiving cyclosporine, an initial dose of 5 mg/kg/12 h was administered orally and adjusted to target a 12-h trough level of 300–450 ng/mL as determined by monoclonal antibody assay. For those patients receiving tacrolimus, the initial dose ranged from 0.075 to 0.1 mg/kg/12 h to target a 12-h trough level of 10–15 ng/mL. More than 90% of the patients received MMF administered 2.0–3.0 g/day in a divided dose. The remaining patients were treated with either azathiopine or sirolimus. Induction therapy (basiliximab or antithymocyte globulin) was used in less than 5% of transplant recipients. Graft failure was defined as return to dialysis or impending failure (i.e. GFR < 15 mL/min/1.73 m3).

Donors

Deceased donor organs were allocated according to standard UNOS allocation system (13).

HLA typing

HLA typing was performed by sequence specific priming (14) using commercial SSP kits (One Lambda, Canoga Park, CA; Pel Freez, Brown Deer, WI).

Antibody detection

FlowPRA® screening, specificity and single antigen bead assays (One Lambda, Inc.,) were used (11,12). Briefly, 5 μL of class I- and class II-coated microparticles was incubated with 50 μL of patient serum and washed. Microparticles were then stained with fluorescein isothiocyanate-conjugated anti-human IgG (FITC-anti-Ig; Fc fragment specific, Jackson ImmunoResearch Laboratories, West Grove, PA). Particle fluorescence was then assessed by flow cytometry. Positive samples were reflexed to specificity analysis using flow specificity and/or single antigen beads coated with either a complete HLA phenotype or a single HLA allele. A control bead devoid of HLA antigens was also included. Fluorescence of 5000 events was analyzed on a flow cytometer as described (14).

Flow cytometric crossmatches

A three-color prospective FCXM was performed according to previously established techniques (15) using pretransplant sera from each recipient (obtained on the day of transplant) and at least two additional sera acquired within the previous 12 months were admixed with fresh lymphocytes from their respective donors. A positive B and/or T FCXM was a contraindication to transplantation, independent of the complement-dependent crossmatch.

Glomerular filtration rate calculations

All subjects in this study were over 18 years old and glomerular filtration rate (GFR) analysis was calculated by the modification of diet in renal disease formula (MDRD) (16; http://www.nephron.com/cgibin/MDRD_GFR.cgi). GFR < 15 mL/min/1.73m3 was considered as impending graft failure.

Statistical analysis

Kaplan-Meier/Cox Proportional Hazards survival analysis was performed as previously described (17).

Results

Recipient demographics

Among 492 recipients receiving DD kidneys at Emory University Hospital between January 1999 and December 2003, 42% were Caucasian, 51% were African American, 3% were Hispanic, 3% were Asian-Pacific Islanders and 1% were either Native American or not clearly identified. Sixty-four percent of the recipients were male. Our patients were unsensitized (n = 272), or had PRA levels of 1–29% (n = 100), 30–49% (n = 32), 50–79% (n = 30) and 80–100% (n = 58). Patients whose PRA was <30 had an average wait time of 493 days from listing until transplant, while the average wait time for patients with a PRA of ≥30% was 1047 days. For this evaluation, if patients had both class I and class II antibodies, only the higher PRA was considered for analysis.

HLA antibody identification

Sera from transplant candidates that tested positive for class I and/or class II HLA antibodies by FlowPRA® screening beads were subsequently tested with FlowPRA® specific and single antigen beads to identify which HLA antibodies were present. A representative example is shown in Figures 1 and 2. As can be seen, (Figure 1A,B), the patient has 87% class I PRA and 96% class II PRA. In Figure 2, representative dot plots of FlowPRA® specific and single antigen beads were used to assign class I (2A and 2B) and class II (2C and 2D) antibody specificities. For the patient illustrated in Figures 1 and 2, the HLA type was A*02, A*03; B*07, B*14(64); Cw*07, *08; DRB1*14, DRB1*16; DQB1*05, –. The () indicates the serological equivalent of the molecular HLA type. Complete analysis of this patient's serum revealed the following HLA antibody specificities: Class I: A1, 11, 23, 24, 25, 26, 29, 33, 34, 66: B8, 13, 38, 39, 44, 45, 49, 50, 51, 52, 53, 57, 58, 62, 63, 70: Class II: DRB1*04, *07, *09; DRB4*01 (DRw53), and DQB1*02, *0301(7), *0302(8), *0303(9), *04, *06. Interestingly, bead #6 on plot 2A showed a weak positive reaction most likely due to antibodies against HLA Cw*15 or Cw*17. FlowPRA® single antigen Cw locus-specific analysis was not performed. Antigens corresponding to the above specificities were entered into UNET (the secure Internet-based transplant information database developed by UNOS) as ‘unacceptable’, meaning that any potential donor organ whose HLA type contained one or more of the unacceptable antigens would not be offered to that patient. This algorithm was applied to each recipient whose screening PRA was positive, allowing a virtual crossmatch to be performed. Specifically, based on the knowledge of unacceptable antigens and the HLA type of the potential donor, a decision was made whether to proceed to a final crossmatch. Preliminary crossmatches were not performed.

Figure 1.

Illustrates the class I (A) and class II (B) FlowPRA® results from a representative patient. The cursor labeled ‘M1’ marks the position of the fluorescence peak in the negative control serum. For both 1A and 1B, cursor ‘M2’ indicates the fluorescence that is considered ‘positive’. For both histograms, the percent PRA is calculated as the number of ‘positive’ events (beads) divided by the total number of events (beads) collected. Approximately 10 000 total events were collected for each class of beads.

Figure 2.

Figure 2.

Shows sample dot plots from a representative patient. (A) Shows a dot plot from the class I FlowPRA® Specific Beads (Group 1) while (B) shows the dot plot from the class I FlowPRA® Single Antigen Beads (Group 1) from the corresponding serum sample. Results were scored as negative (−), weak positive (+/−) or positive (+) based on their level of green fluorescence (anti-IgG-FITC; x-axis) as compared to the negative control serum. For all dot plots, the vertical line indicates the approximate position of the beads when incubated with a negative control serum. The beads are numbered 1–8 based on there level of red fluorescence (y-axis). The corresponding HLA antigen(s) that are expressed on the surface of each bead are also listed. The bead labeled ‘control bead’ is an identical microparticle that has been coated with human albumin. (C) Shows a dot plot from the class II FlowPRA® Specific Beads (Group 2) with probable specificities indicated by an underline (__). (D) Shows a representative dot plot from the class II FlowPRA® Single Antigen Beads (Group 3) from the corresponding serum sample.

Figure 2.

Figure 2.

Shows sample dot plots from a representative patient. (A) Shows a dot plot from the class I FlowPRA® Specific Beads (Group 1) while (B) shows the dot plot from the class I FlowPRA® Single Antigen Beads (Group 1) from the corresponding serum sample. Results were scored as negative (−), weak positive (+/−) or positive (+) based on their level of green fluorescence (anti-IgG-FITC; x-axis) as compared to the negative control serum. For all dot plots, the vertical line indicates the approximate position of the beads when incubated with a negative control serum. The beads are numbered 1–8 based on there level of red fluorescence (y-axis). The corresponding HLA antigen(s) that are expressed on the surface of each bead are also listed. The bead labeled ‘control bead’ is an identical microparticle that has been coated with human albumin. (C) Shows a dot plot from the class II FlowPRA® Specific Beads (Group 2) with probable specificities indicated by an underline (__). (D) Shows a representative dot plot from the class II FlowPRA® Single Antigen Beads (Group 3) from the corresponding serum sample.

Flow cytometric crossmatches

When donors and recipients were predicted to be ‘acceptably mismatched’ by the virtual crossmatch approach described above, a flow cytometric T- and B-cell crossmatch was performed with current and historic sera to confirm (or refute) the anticipated result. A negative T- and B-cell FCXM was obtained in >80% of these cases (data not shown). When a crossmatch between a given recipient/donor pair that was predicted to be negative actually was positive (T and B cell or B cell only), the transplant was not performed with that recipient. Instead the kidneys were transplanted into ‘backup’ candidates (sensitized or unsensitized) whose crossmatches were negative.

Transplant outcome

Five-year actuarial graft survival is shown in Figure 3. Overall graft survival was similar between unsensitized patients (70%), patients with PRA between 1% and 29% (69%) and patients >30% PRA (66%). Interestingly, graft survival among the subgroup of patients transplanted with a PRA >80% was identical to that of unsensitized patients. The differences among groups were not statistically significant (p > 0.05).

Figure 3.

Kaplan-Meier survival curve illustrating the 5-year follow-up for patients classified as ‘sensitized’ (i.e.: PRA > 30%) or ‘unsensitized’ (PRA < 30%). There was no statistical difference between the groups (p > 0.05). Boxed data show the distribution (numbers and percentages) of ‘sensitized’ patients within selected PRA groups.

Discussion

This study addresses one of the most compelling issues in DD renal transplantation, namely, equity among candidates awaiting a transplant. Here we show that combining solid-phase HLA antibody identification methodologies with virtual crossmatching significantly increases the allocation of DD organs to highly sensitized patients (defined in this study as >30% PRA). Our data also suggest that long-term graft survival among these recipients is virtually identical to that of unsensitized patients. Certainly, a more detailed analysis of outcome variables such as episodes of acute and chronic rejection, number/types of viral infections, varying immunosuppressive regimens, etc., between the different patient groups will be needed to confirm these preliminary observations. Nonetheless, our findings clearly demonstrate that, through a comprehensive evaluation of HLA antibodies, unacceptable HLA antigens can be identified before patients are listed for a DD organ. Application of this information allows for selection of only those recipients whose final crossmatches have a high probability to be negative. For the past 5 years, we have been transplanting highly sensitized patients at a rate that appears to be twice the national average (∼25% vs. 12%). However, not all patients on the national list were tested with sensitive solid phase detection formats as done in this study, meaning that the number of sensitized patients in our program is likely higher than the national average based solely on our antibody detection techniques (11). Thus, assuming a proportional relationship, if we had twice the number of sensitized patients than the national list, we would be predicted to transplant twice as many patients. In fact, a review of our wait-list reveals that Emory has ∼43% more sensitized patients than the UNOS wait-list. Assuming the same proportional relationship as above, the number of sensitized patients receiving a DD transplant should have increased by 43% (from 12% to 17%). In actuality, 25% of our sensitized patients were transplanted, corresponding to an increase 47% greater than expected. This value likely underestimates the impact of applying the Emory Algorithm to sensitized patients. Since offers of DD organs are independent of how antibodies are defined, increasing the number of sensitized patients vying for the same organs should be a disadvantage to transplantation. As shown in this study, the Emory Algorithm provides a rationale for selecting the most appropriate patients to crossmatch with a given donor.

Sensitized patients (especially those with PRA >80%) are one of the most disadvantaged group of recipients on the national waiting list. In the United States, sensitized patients (even those with a PRA as low as 30%) are transplanted at approximately half the rate of their unsensitized counterparts (2). Patients with >80% PRA, arguably the most disadvantaged group, constitute 12% of the national wait-list, receive <5% of DD organs and are equally likely to die awaiting transplantation as they are to receive a compatible organ from a DD. Recognizing this inequity, UNOS policy attempts to give priority to these patients. Specifically, patients with a PRA value >80% are awarded four additional points in the allocation system for DD kidneys, wherein the priority of patients is determined by a point system (13). Unfortunately, even when a medically suitable patient becomes the number one candidate on the UNOS wait-list and is offered an organ, transplantation is still not a certainty. Since 1969, the final crossmatch has served as the foundation for organ allocation (1) and for highly sensitized patients; a positive final crossmatch remains the single greatest barrier to transplanting the sensitized patient. Typically, sera from potential recipients are pretested against lymphocytes from ABO-compatible donors as a form of triage. Only patients with negative preliminary crossmatches (i.e. first-phase testing) are given further consideration. A final crossmatch then is performed on selected patients who passed the ‘first phase’. The transplant recipients are then selected from those patients whose final crossmatches are negative. There are several problems associated with the above approach. First, the method is resource intense and time-consuming. Large centers may crossmatch several hundred patients as part of their first phase testing. A second issue is one of false-positive reactions. In these situations, sera from patients without HLA antibodies have preliminary positive crossmatches (e.g. due to autoantibodies or carryover from positive patients) thereby inappropriately excluding them from consideration. A third point is one of target cell selection. Preliminary crossmatches are typically restricted to T lymphocytes or total lymphocytes which can detect antibodies to class I HLA antigens, but are poorly suited or unable to detect clinically relevant antibodies against class II HLA antigens. Finally, assay sensitivity is also a concern. Preliminary crossmatches are generally performed using some form of complement-dependent cytotoxicity assay. If the ‘second phase’ (i.e. final) crossmatch is performed with a more sensitive technique such as flow cytometry, final crossmatches may be unexpectedly positive (18). Alternatively, when complement-dependent assays are used for both preliminary and final crossmatches, clinically relevant antibodies may be missed (6,10).

When the serum of a potential renal recipient has HLA antibodies, specificities can be identified to determine which HLA antigens should be considered as unacceptable for that patient. By entering this information into the UNOS database (UNET), a patient will not be offered a kidney from a donor who expresses an incompatible HLA phenotype (i.e. antigens to which the recipient has antibodies). The corollary is that such sensitized patients (even those with >80% PRA) would have HLA antibodies that would not be donor directed and would have a high probability of a negative crossmatch. By identifying unacceptable HLA antigens, it should be possible to predict a negative crossmatch with donors who possess HLA antigens that would be acceptably mismatched. In fact, this approach has been successful in Eurotransplant, (the organization responsible for organ allocation in several European countries) where 450 highly sensitized patients have been transplanted since the inception of their acceptable mismatch program (reviewed in 19). Eurotransplant recently combined its acceptable mismatch program with HLA Matchmaker (20). In its simplest form, the HLA Matchmaker computer program predicts donor/recipient compatibility by comparing linear amino acid sequences of their HLA antigens (21). Relevant epitopes are defined as polymorphisms residing on the alpha helices and beta loops of the HLA molecule. While extremely creative, HLA Matchmaker does not consider that polymorphisms on/in the floor of the HLA molecule contribute to its immunogenicity. This oversimplifies the complex biology of a globular protein. Additionally, by considering only its linear sequence, HLA Matchmaker ignores the role of three-dimensional conformations in defining epitopes on the HLA molecule.

Assignment of acceptable mismatches is predicated on the comprehensive identification of the HLA antibody specificities possessed by the prospective recipient. Using serological testing, and until recently, even solid-phase technologies, HLA antibody specificities were identified with targets that expressed the entire class I or class II phenotype of the donor. Because of antigen overlap and linkage, it was not always possible to identify a unique specificity (e.g. HLA-A1) independent of another (e.g. HLA-B8). In fact, depending on how many different target cells were included in the panel and the constellation of HLA antigens expressed, certain class I antibody specificities could be masked or missed entirely. Moreover, for cytotoxicity assays, the use of T cell targets that lack expression of class II antigens does not permit the detection of class II antibodies. Thus, the inability to completely define antibody specificities among a genetically diverse population of renal transplant candidates has rendered crossmatch prediction in the U.S. inconsistent. As a result, DD kidney allocation has been predominantly driven by the crossmatch, an approach which unintentionally favors moderately sensitized and nonsensitized patients. For example, when, for a given DD, a highly sensitized patient has inappropriately reached the top of a UNOS match run (because of failure to identify all donor directed HLA antibodies) and is brought in for a final crossmatch (which will be positive) another highly sensitized recipient (without donor directed antibodies) may have been denied the opportunity to be transplanted with a kidney from that donor. Moreover, significant resources have been expended and time wasted crossmatching the inappropriate donor/recipient pair.

At our center, we focus on using flow cytometry-based solid-phase HLA antibody detection assays to specifically and comprehensively identify class I and class II HLA antibodies, both of which we consider to be clinically relevant (22). When a donor expresses class I and/or class II HLA antigens to which the recipient possesses corresponding antibodies, the recipient is considered incompatible for that donor. The Emory Algorithm (Figure 4) is a systematic approach that benefits the highly sensitized patient. Utilizing information from comprehensive antibody screening, it begins with a virtual crossmatch of highly sensitized patients that triages out potential recipients who possess donor directed antibodies. As a result, only those sensitized patients whose HLA antibodies are not donor directed (and therefore have a high probability of a negative final crossmatch) appear on the match run. In essence, the Emory Algorithm rearranges the priority of highly sensitized patients on any given match run according to the specific HLA antigens expressed by the donor. While quite successful, virtual crossmatching is still short of perfect; the predicative probability of our algorithm is ∼80%. There are several explanations for not achieving 100% predictability such as: (1) solid-phase assays do not provide targets to every known HLA antigen/allele (now >1800); (2) nontraditional HLA antibodies such as those to HLA-DP and DQ-alpha polymorphisms (23,24) could be responsible for an unexpected positive crossmatch; (3) the positive crossmatch is due to non-HLA antibodies.

Figure 4.

Emory algorithm for classifying patients based on HLA sensitization. Sera from potential transplant recipients are initially screened for the presence/absence of HLA antibodies. Patient's sera that are ‘positive’ for HLA antibody are subjected to detailed antibody specificity analysis using HLA-specific microparticle technology. Well-defined HLA specificities are then listed in UNET as ‘unacceptable’ antigens. When a donor organ is available, patient's unacceptable antigens are used to select potential recipients with the highest probability of a negative crossmatch. Sensitized patients who are listed on the UNOS match run are selected for final crossmatching by flow cytometry. Only flow cytometric crossmatch negative patients were transplanted.

In conclusion, the Emory Algorithm prioritizes sensitized patients based on a predicted probability of a negative crossmatch with a given donor. In effect, it efficiently identifies donor/recipient pairs that appear to be acceptably mismatched. By applying this algorithm nationally, we believe it may be possible to achieve equity (or at least significantly reduce inequity) between sensitized and unsensitized recipients regarding access to and allocation of DD organs. Furthermore, while the data are preliminary, it appears that 5 year transplant outcome may not be compromised by high PRA levels as previously reported (25,26). Additional studies will be needed to confirm/refute these initial observations.

Ancillary