Discussion
There is widespread agreement that accurate virtual crossmatches that do not eliminate acceptable donors are required for KPD to reach its full potential. However, considerable controversy remains regarding the best practices for accomplishing this goal. This report demonstrates that virtual crossmatches can be used to reliably predict acceptable cell-based crossmatches and a large number of transplants can be achieved by allowing each transplant center to establish criteria for assigning unacceptable antigens and acceptable crossmatches.
The NKR serves an extremely sensitized patient population; as of December 2012, 68% of patients had CPRA ≥80%. During the period of this investigation (between March 1, 2011 and December 20, 2012) the NKR facilitated 307 transplants including 96 (31%) to highly sensitized patients (≥80% CPRA). For comparison, the UNOS deceased donor waiting list in 2011 had 17% highly sensitized patients (>80% PRA) and 5% of new listings in 2011 were highly sensitized (2011 HRSA:SRTR data report; http://srtr.transplant.hrsa.gov/annual_reports/2011/default.aspx). During 2011, 18% deceased donor recipients and 7% living donor recipients had CPRA >80%.
We believe that one of the factors contributing to the large number of NKR transplants involving highly sensitized patients was a 91% accuracy rate for virtual crossmatches, which was achieved by allowing participating centers to set individual thresholds that were appropriate for their criteria for an acceptable final crossmatch. Importantly, this high rate of accurate virtual crossmatches was achieved without conservative strategies that have been proposed by others, such as establishing common parameters for unacceptable antigens and acceptable crossmatches or employing a central laboratory.
Other approaches have achieved high success rates for virtual crossmatches for KPD but these conservative approaches can exclude acceptable donors. For example, the Australian National Exchange program used 2000 MFI to assign unacceptable antigens, but no compatible donors were identified in four quarterly match runs. The likelihood of a match was improved by increasing the threshold to 8000 MFI, but patients with even higher levels of DSAs can be successfully transplanted. It is likely that any standardized approach using MFI to define unacceptable antigens will prevent transplants that might be successful because this strategy does not consider differences in HLA antibodies that might affect rejection risk (e.g. different levels of HLA expression) or modified transplant protocols. The study reported here shows that transplant centers in consultation with their histocompatibility laboratories can develop center-specific criteria to achieve virtual crossmatches for reliably selecting donors that meet each center's criteria for transplantation.
Comprehensive HLA typing of donors is important for virtual crossmatching because sensitized patients have antibodies against the products of all HLA loci (Figure 2). During the interval of this study, donors were routinely typed for HLA-A, -B, -Bw4/6, -C, -DRB1, -DRB3/4/5, -DQB1 and -DPB1. In situations where antibodies frequently distinguished the subtypes, there were additional requirements for subtyping (e.g. DPB1 was subtyped into 04:01 and 04:02). Since 26% of NKR patients listed HLA-DP unacceptable antigens, HLA-DP typing of donors likely played an important role in accurate virtual crossmatching for this highly sensitized population. HLA-DQA1 typing and additional allele-level typing could have eliminated five of the unexpected crossmatches observed during this interval.
In June 2013, the NKR made HLA-DQA1 typing donors mandatory and this is expected to further increase virtual crossmatch accuracy. HLA-DQA1 typing offers two advantages: (1) automatic elimination of donors when the patient has antibodies that appear to be specific for HLA-DQA1 and (2) an opportunity to exclude donors based upon antibody epitopes formed by a particular combination of HLA-DQ alpha and beta chains. The ability to consider combinations of DQA and DQB chains is important because a high percentage of HLA-DQ antibodies are against epitopes created by a particular combination of DQA and DQB chains and the component HLA-DQA and -DQB types would exclude compatible donors who have the same HLA-DQA and -DQB types but in different combinations that do not contain the relevant antibody epitopes. With HLA-DQA1 typing, specific combinations of DQA1 and DQB1 types can be individually considered before moving forward with cell-based crossmatching. Precise HLA-DQ typing is particularly important because many patients have HLA-DQ antibodies (53% of patients had HLA-DQB unacceptable antigens) and epitopes do not always correspond to the classic serological types.
There are additional modifiable factors that could further improve virtual crossmatch accuracy. Four virtual crossmatch failures (6% of failed virtual crossmatches) were attributed to errors in data entry. To prevent such errors, the NKR has recently implemented a requirement for laboratories to audit histocompatibility data. A major cause of virtual crossmatch failures was changes in HLA antibodies (n = 14, 21% of failed virtual crossmatches). At least some of these could be eliminated by more frequent antibody testing. A minimum of quarterly testing was recommended by a consensus conference, but this might be insufficient for highly sensitized patients who might experience changes in antibody levels from any inflammatory event, including infections. Another potentially modifiable factor is non-HLA antibodies that cause positive cell-based crossmatches but are not a contraindication to transplant. Autoantibodies are one example of this situation that could be addressed by performing autologous crossmatches, particularly for patients at high risk for autoantibodies such as those with autoimmune disease.
The major cause of unexpected positive cell-based crossmatches is equivocal virtual crossmatches caused by DSA levels that cannot be used to reliably predict the result of a cell-based crossmatch (MFI near the threshold for an unacceptable crossmatch or the cumulative effects of multiple weak DSAs that individually would be acceptable). Improved technology and more experience with virtual crossmatching are likely to diminish these somewhat, but virtual crossmatching is unlikely to reach 100% accuracy for predicting the results of cell-based crossmatching because there is technical variation for solid-phase antibody tests and cell-based crossmatches. Exclusion of donors-based upon equivocal virtual crossmatches is not recommended because this might eliminate the only option for transplant for some patients. KPD programs could take measures to diminish the impact of equivocal virtual crossmatches. For example, preliminary cell-based crossmatches could be performed when virtual crossmatches are equivocal. A progressive option, which was utilized by two of the most active centers in the NKR, is to utilize desensitization protocols designed for patients with positive cell-based crossmatches. It has been reported that using desensitization protocols for transplantation of highly sensitized patients with positive crossmatches improves patient survival relative to those who are not transplanted. Longer follow-up of a large cohort is needed to determine if this advantage will also be realized by patients currently being transplanted through exchange programs.
To ensure that assessment of sensitization was not substantially influenced by the relatively small donor pool used to calculate CPRA, CPRA was also determined. CPRAs were remarkably similar when there were no antibodies against HLA-C and HLA-DP that are not considered in the OPTN calculation. This suggests that the size of the donor pool used for determining PRA is not very important for approximating the percentage of incompatible donors. This comparison also illustrates how the current limitations of the OPTN CPRA calculator can affect CPRA values, which in turn influence organ allocation. When the CPRA is extremely high, there are often many antibodies against HLA-A, -B, -DR and -DQ, which create a high CPRA independent of HLA-C and HLA-DP antibodies. However, when the CPRA is lower, the contributions from HLA-C and -DP antibodies can be substantial. Until this is remedied, these patients will be disadvantaged in the national allocation system.
There are limitations to use of registry data. The NKR database records unacceptable antigens, not DSA levels. Therefore, DSA levels could not be correlated with transplant outcomes. Although only a few grafts have failed (2.9%) and serum creatinine levels at 6 months and 1 year were generally acceptable, follow-up times are still relatively short and we were unable to evaluate rates of antibody-mediated rejection or pathological changes in kidneys transplanted to sensitized and highly sensitized patients. Only three failures were reported to be directly or indirectly due to immunological rejection and all occurred in recipients who were unsensitized at the time of transplant. Another limitation of this study is that it is too early to know if long-term outcomes for NKR patients will be acceptable.
The recommendations for histocompatibility testing for KPD (e.g. more HLA typing requirements, auditing data entry, more frequent testing for HLA antibodies and testing for non-HLA antibodies) can increase costs for managing listed patients. However, the costs of additional testing prevent wasting of other financial and human resources related to unexpected positive crossmatches and disrupted chains. Further, reducing chain failures should diminish frustration of staff, patients and donors that can discourage participation in KPD. In addition, accurate and efficient virtual crossmatching requires close communication between the lab and the transplant center. A dedicated and passionate KPD team is important for managing patients and ensuring close monitoring of HLA antibodies. KDP involves financial and professional investment, but for some patients, KPD is the best or only option for transplant and transplants made possible by KPD can improve their survival.
In summary, the NKR, which serves as a highly sensitized patient population, achieved a 91% accuracy rate for virtual crossmatches. Data presented here show that highly predictive virtual crossmatches, which are extremely important for efficient identification of chains of compatible donor-recipient pairings, can be achieved without using a standardized approach for assigning unacceptable antigens. By allowing centers to set their own thresholds, centers have the opportunity to transplant more sensitized patients based upon their risk tolerance, but long-term follow-up is still needed to establish the efficacy of higher-risk transplants. Analysis of virtual crossmatch failures can guide development of new policies that will further improve accuracy. Based upon the results reported here, the NKR has developed a donor preview function that allows centers to exclude donors if a patient has multiple DSAs that fall below the unacceptable threshold, but in combination, might be unacceptable. Analysis of NKR data also resulted in a new requirement for HLA-DQA1 typing for donors. Options to request additional HLA typing (for alleles) and an exploratory crossmatch prior to accepting a potential donor have been added for evaluating complex virtual crossmatches. These new developments should further improve virtual crossmatch accuracy, ultimately resulting in transplantation of more highly sensitized patients and reduced waiting time.