Predicting Kidney Graft Failure by HLA Antibodies: a Prospective Trial



HLA antibodies have been shown to be associated with late graft loss of organ transplants in prior studies. Recently they were even shown to appear years BEFORE rejection. (1) An international cooperative study of 4763 patients from 36 centers was undertaken to determine the frequency of HLA antibodies in patients with functional transplants. The overall frequency of HLA antibodies among kidney transplant recipients was 20.9%; 19.3% in the liver, 22.8% in the heart, and 14.2% in the lung. Patients treated with CsA-MMF had significantly lower antibodies (9.8%) than those treated with CsA-Aza (18.1%) (0.00008). (2) Second, a prospective trial was performed in 23 kidney transplant centers to determine whether HLA antibodies could predict failures within 1 year. Among the 2278 patients followed up, 91 grafts failed and 34 patients died. Of 500 patients who had HLA antibodies, 6.6% failed compared with 3.3% among 1778 patients without antibodies (p = 0.0007). Among 244 patients who made de novo antibodies, 8.6% failed compared with 3.0% failures among 1421 patients who did not make antibodies (p = 0.00003). Death occurred in 1.5% of patients and was not associated with antibodies. Thus, after 1 year in this prospective trial, patients with HLA antibodies had graft failure at a significantly higher rate than those without antibodies.


For as long as 35 years, it has been known that HLA antibodies occur in the serum of patients after transplantation (1,2). Moreover, they have been strongly associated with subsequent failures (2). However, despite association with failure and demonstration that antibodies occur before rejection in heart (3), lung (4) and kidney transplants (5), reluctance remains in considering use of HLA antibodies for routine monitoring. Systematic monitoring with long-term follow up demonstrating the effectiveness of HLA antibodies in predicting late graft loss was lacking. However, this evidence was supplied in 2002 in the first study covering an 8-year period of annual antibody tests (6). Two significant findings emerged from this long-term systematic study: first, following the detection of antibody, many years may often elapse before the graft is finally rejected, and second, when the antibody is found, it is often not a donor-specific antibody. Probably, donor-specific antibodies were not found because they had been absorbed by the kidney and not spilled into the peripheral blood. However, when the kidney was fully rejected, invariably donor-specific antibodies were found (7). Use of single antigen beads coated with antigens from recombinant lines made it possible to clearly see the donor-specific antibodies postrejection. It also allowed us to observe numerous nondonor-specific antibodies generated by the immunization. No patient had made ‘monospecific’ antibodies directed solely at the mismatched antigen.

The finding that HLA antibodies precede rejection by a few years was recently confirmed in a more extensive study (8). The mean time between appearance of HLA Class I antibodies and failure was 634 days, compared with a mean of 1503 days following the appearance of Class II antibodies. Donor-specific antibodies were found in the patients, but there were 34 additional cases in which donor-specific antibodies were discovered only after the graft failure. In a separate study involving five centers, as many as 96% of 825 patients who had rejected a kidney graft had HLA antibodies (9).

Based upon the two studies (6,8), HLA antibodies apparently are ideal markers for impending graft failure because they appear so long before the failure, providing ample time for intervention. As only approximately 5% of transplants can be expected to fail within 1 year, and some of the failures can be expected to result from nonimmunological causes, it was necessary to initiate a large-scale study with many centers. This study was initiated as part of the 13th International Histocompatibility Workshop. Here we report the antibody test results together with the 1-year follow-up results.


For the antibody frequency studies, from December 1, 2001 to April 30, 2002, a ‘snapshot’ was taken of 4763 patients from 36 centers who had a functioning kidney, heart, liver or lung graft from cadaver and living donors for greater than 6 months. HLA antibody testing was carried out by cytotoxicity, flow cytometry, and ELISA methods.

For the follow-up study, 23 kidney transplant centers provided data. After testing for antibody in 2002, 1 year later, centers were asked to report all deaths or returns to dialysis. Results were accumulated and entered into a computerized file. Data collected were: sample ID, sex, sample date, type of transplant, transplant center, number of prior transplants, PRA pretransplant, transplant date, maintenance immunosuppression, serum creatinine, PRA post-transplant cytotoxicity, PRA post-transplant ELISA, date returned to hemodialysis or re-transplanted, date of death, alive and functioning. Some totals do not match, as not all centers supplied all requested data. Statistical significance was calculated by chi-square analysis.

Drugs reported to be used at the time of antibody testing were: cyclosporinee (CsA), prednisone, mycophenolate mofetil (MMF), azathiaprine (Aza), and tacrolimus.

Funding sources were from Terasaki Foundation Laboratories and One Lambda Inc.


Frequency of antibodies post transplantation

Figure 1 shows the frequency of HLA antibodies found in patients with functioning allografts for greater than 6 months following a kidney, heart, liver, or lung transplant in 36 centers. Results from each center are listed in descending order of number of patients tested at that center. The average percent post-transplant antibody in kidney transplant patients was 20.9%; 19.3% in 278 liver, 22.8% in 393 heart, and 14.2% in 113 lung transplant patients. The frequency of antibodies according to the test method was: 21.0% for 2496 patients tested by cytotoxicity; 19.5% for 3804 patients tested by ELISA; and 25.2% for 968 patients tested by flow cytometry. Patients tested by two methods were counted for their results using separate methods. Although both ELISA and flow cytometry are more sensitive than cytotoxicity, the overall frequencies were similar possibly because IgG antibodies had been searched for in the more sensitive techniques, whereas both IgG and IgM antibodies had been detected by cytotoxicity. In addition, the frequency of antibodies at any given center was influenced by the type of immunosuppression used, as shown in Table 1. Thus if the center that used flow cytometry had used a drug which resulted in a lower antibody formation, this center may have the same frequency of antibody as a center using cytotoxicity testing together with a less effective drug. The percentage of patients who have antibodies in a given center may be more dependent on various factors such as the particular sampling of patients, their treatment, etc. than the particular PRA test used.

Figure 1.

Frequency of HLA antibodies found post-transplantation. Patients with functioning kidney, liver, heart and lung transplants from 36 centers were examined for antibodies between December 1, 2001 and April 30, 2002. The centers are listed in descending order of number of transplants studied. The first center listed for kidneys had tested 628 patients.

Table 1. Immunosuppressive drugs and occurrence of post-transplant antibody in kidney transplant recipients. Only patients who did not have pre-transplant antibodies are included
Drug combinations# patients w/drug% antibodyp value
CsA, Aza53618.1%Reference
CsA, MMF5339.8%0.000082
Tac, Aza6713.4%0.34
Tac, MMF23312.9%0.073

Data regarding whether or not pretransplant HLA antibodies were present was available in 2659 patients. Among 2185 nonsensitized patients, 14.7% developed antibodies. In contrast, among the 474 patients with antibodies before transplantation, 52.7% had post-transplant antibodies, meaning that in the remainder of the patients, antibodies had been lost by the time the patient was tested.

Among the 2278 kidney transplant patients from whom we obtained follow-up data on graft failures, the frequency of antibodies according to the duration of the transplant is shown in Table 2. Interestingly, the frequency of antibodies found was relatively constant for patients transplanted from 1 to as many as 40 years ago. Among the 525 patients transplanted more than 10 years ago, the frequency of antibody was the highest at 26.9%.

Table 2. Frequency of HLA antibodies in kidney transplant patients transplanted in various years and then tested in year 2002
Number of patientsYear of transplantPercent with antibody
2278 (total)

Influence of immunosuppressive drugs on antibodies

Only kidney transplant recipients who were not presensitized before transplantation were considered for this analysis (Table 1). For two drug combinations, if CsA and azathiaprine were taken as the reference, 18.1% of patients had antibodies. In comparison, CsA with MMF resulted in 9.8% of patients with antibody (p = 0.00008), and tacrolimus with MMF resulted in 12.9% of patients with antibody (p = 0.07).

One-year follow-up analysis of kidney transplant patients

Table 3 documents the results obtained from 23 kidney transplant centers. The patients had been divided into two groups: those with HLA antibody and those without antibody and registered a year earlier. Antibody results were based on PRA testing, and no attempt was made to obtain data on donor-specific antibody. Follow-up data sent by the centers were re-verified individually by the centers before publication of the table. At each individual center, the number of failures was small, and it was difficult to draw conclusions regarding the effect of antibodies.

Table 3. Patients studied for post transplant antibody and follow-up at 1 year
 TotalPost tx with antibodyPost tx, no antibody
DirectorCountry# patients# patients# fail# patients# fail
  1. 1Deceased.

BowMA, USA1583411245
LeforFL, USA147584892
duToit/TaljaardS. Africa107242831
Milford/CapenterMA, USA107301777
McAlackPA, USA79151646
Crowe/LangoneTN, USA2970221
Orosz/PelletierOH, USA2450190
SaidmanMA, USA2340190
FuggleOxford, UK1741130
Mahoney1/BreggiaMA, USA20020
Total 227850033 (6.6%)117858 (3.3%)

When the data from the kidney transplant patients was pooled (Table 4), among the antibody-positive patients, 6.6% failed, compared with 3.3% of those who did not have antibodies. This difference was highly significant (p = 0.0007). Among patients already sensitized at the time of transplantation, graft failure for those with post-transplant antibodies was 6.0%, which was significantly higher than the 2.1% failure rate among those who had lost their antibodies (p = 0.05) (Table 4). This difference was greater for patients who were unsensitized before transplantation. Patients with de novo antibodies had an 8.6% graft failure rate, compared with 3.0% for those without antibodies (p = 0.00003). Deaths accounted for 27.2% of total 125 failures. No significant difference was noted among those with or without antibodies (Table 4). Among recipients with antibodies, 1.6% died, compared with 1.5% of those without antibodies (p = 0.76). An effect of antibodies on death was not apparent, even in patients who developed de-novo antibodies.

Table 4. Kidney transplant patients with 1-year follow-up data. Graft failures and deaths among kidney patients with or without antibody
 # PatientsGraft failure% FailpDeath% Deathp
 post ab pos500336.6% 81.6% 
 post ab neg1778583.3%0.0007261.5%0.76
 total2278914.0% 341.5% 
Pre-transpl antibody, pos
 post ab pos167106.0% 53.0% 
 post ab neg19042.1%0.0521.1%0.19
 total357143.9% 72.0% 
Pre-transpl antibody, neg
 post ab pos244218.6% 20.8% 
 post ab neg1421433.0%0.00003151.1%0.80
 total1665643.8% 171.0% 


We first established that approximately 20% of patients with a well functioning transplant will have HLA antibodies. The present prospective trial confirmed the findings of 23 prior studies (2), showing that approximately 20% of patients with functioning grafts had HLA antibodies (Figure 1). Moreover, we confirmed that the presence of antibodies occurs not only in transplanted kidneys, but also in heart, lung, and liver transplants. The fact that HLA antibodies were present in 20% of patients with functioning transplants shows that the antibodies do not have an obvious immediate deleterious effect on function.

As shown earlier, antibodies are continuously present for many years before graft rejection occurs (6,8). Presumably, intimal thickening of vessels walls resulting from a cycle of damage and repair occurs slowly over the course of many years (10).

Long-term loss of kidneys post transplantation (after 6 months) occurs at a constant loss rate of approximately 5% per year, which is equivalent to a 14-year half-life. In this prospective trial, 5.5% of the 2278 patients actually failed during the 1-year trial period. This suggests that underreporting of failures did not occur in this prospective trial. Death occurred in 1.5% and graft failure in 4.0%. The significant finding was that among patients who did not have antibodies before transplantation, 8.6% of the failures occurred in patients who developed HLA antibodies, compared with 3.0% of failures in those who did not produce antibodies (p = 0.00003).

At this point, we should note that antibody detection by the PRA test is not completely standardized among laboratories. Aside from the problem that today, three different tests are commonly used, namely, cytotoxicity, ELISA, and flow cytometry, standardization of test methods for these three tests has not been agreed upon by all labs. Most importantly, the cut-off between positive and negative is not completely objective in all three tests. With cytotoxicity, it is the subjective reading of reactions in a microscope, indicating that there are slightly more dead cells in the test well than in the control wells. In ELISA tests, some laboratories use two times the standard deviation of the mean as the cut-off for positivity, whereas different manufacturers of kits may use another formula. Also there is often a problem with some sera showing a high background. For flow cytometry, the cut-off is often determined visually from the shape and ‘tail’ of the curve in the control tube. There is also the problem of the number of test cells used for the panel testing, which varies between laboratories. These problems relate to the weak sera. Approximately 90% of sera give strong positive or clean negative results by all tests. Despite center variables in PRA testing, the significant finding reported here (Table 4) is that patients who were identified as having antibodies by the PRA test 1 year previously had a higher loss of graft function than those without antibodies.

It is unlikely that HLA antibodies are epiphenomenon, as in the current prospective trial, patients with antibodies had a higher 1-year failure rate than those without antibodies. In addition, long-term follow-up studies have shown that many years later, all patients with antibodies will fail (6,8). The theoretical basis for the use of HLA antibodies to monitor graft failure is that it is postulated that these antibodies are directly related to the failure. If they are the cause, it would be reasonable to use them as predictors of impending failure.

We recently reviewed 35 studies showing an association of HLA antibodies with rejection (11). There is evidence for the humoral theory of transplantation, which states that antibodies, not cells, destroy allografts (11). In the evidence cited, antibodies were clearly the cause of hyperacute rejection (12). An increasing fraction of acute rejections was recently shown to be antibody-mediated by C4d staining (13). Recent adoption of C4d as a means of identifying immunologic humoral rejection is a major advance in clearly identifying an immune response-generated injury (14,15). C4d provides evidence of complement activation, sometimes by other pathways, but most predominantly by antibody. When C4d deposits are found in kidney biopsies, they are accompanied by the presence of antibodies in the serum, as shown by five different groups of investigators (14,16–19). The C4d test is ultimately a more accurate indication of humoral rejection than the PRA test, as it is a complement degradation product, whereas the PRA test is dependent upon detection of antibody circulating in the peripheral blood.

The antibodies that produce rejection are, of course, directed against the specific mismatch of the donor. Recent testing confirms that these donor-specific antibodies are sometimes not present in the circulation, presumably because they are absorbed by the kidney. Fortunately, for purposes of response detection, HLA antibodies are produced against cross-reactive epitopes of other specificities (7), and can be detected by routine panel reactive antibody (PRA) testing. These antibodies were produced de novo in response to the transplant, and are different from those that may have existed before transplantation, and unrelated to the particular negative crossmatch donor that was utilized for transplantation. This same tendency to produce extraneous reactivities was noted earlier from the common experience of finding almost no mono-specific typing serum from transplanted patients, even when a single HLA antigen had been mismatched.

If, from the foregoing, HLA antibodies are causally related to humoral rejection, then patients can be treated to inhibit antibody formation. The simplest treatment is conversion to a new drug. From the current findings of the incidence of antibodies relative to drugs used, MMF appears to be effective for reducing the incidence of HLA antibodies (Table 1). In addition to these findings, tacrolimus and MMF were utilized to reduce antibody levels in four kidney patients who had chronic rejection with concurrent donor-specific antibodies (20). Other methods of reducing antibody levels have been reviewed (11).

In conclusion, we showed in this prospective trial after a 1-year period, HLA antibodies predicted kidney graft failure. By identifying those patients undergoing chronic rejection, known effective steps can be taken to reduce antibodies. Although these tests were described for kidney transplants, the fact that post-transplantation antibodies are also present in similar frequencies in heart, liver, and lung transplants (Figure 1) suggests that HLA antibodies can also be used to monitor immunologic graft failure in these organs. The mechanism of immunologic rejection is probably the same for all solid organs: HLA incompatibilities stimulate the production of HLA antibodies, which are associated with graft failure.


PIT designed the experiment and wrote the paper, and MO communicated with the centers and created a computerized file from which analysis was performed.

Conflict of interest statement

Paul Terasaki is chairman and major stockholder of One Lambda Inc, one of the companies that produces HLA antibody testing kits.


We thank all those listed in Table 3 for their patient follow-up information, and the following investigators for the antibody frequency data shown in Figure 1. Larry Chan, MD, Violet Esquenazi, PhD, Christina Glehn, PhD, Nancy Higgins, S. Kankonkar, MD, Po-chan Lee, MD, Elissaveta Naumova, MD, PhD, B. Perichon, MD, Zafirios Polymenidis, MD, Elaine Reed, PhD, Nancy Reinsmoen, PhD, Lemonia Skoura, MD, Caner Suesal, MD, and Adriana Zeevi, PhD. Thanks also to Dr John Hansen for supporting the initiation of these studies as part of the International Histocompatibility Workshop, Dr Daniel Cook for help in the early phases of these studies, Rene Castro for database support, and Su-Hui Lee for secretarial assistance.