De Novo Donor-Specific HLA Antibodies: Biomarkers of Pancreas Transplant Failure



This study assesses the role of posttransplant HLA antibody monitoring in the surveillance of pancreas transplant recipients. Four hundred thirty-three pancreas transplants were performed at the Oxford Transplant Centre 2006–2011 (317 simultaneous pancreas kidney [SPK] and 116 isolated pancreas [IP]). HLA antibody monitoring was performed at 0, 6 and 12 months and annually and during clinical events. There was no association between pancreas graft failure and recipient or donor characteristics. Posttransplant antibody status, available for 354 (81.8%) of recipients, demonstrated that 141 (39.8%) developed de novo HLA antibodies, of which 52 (36.9%) were de novo donor-specific HLA antibodies (DSA) (34 SPK, 18 IP). The development of antibodies to donor HLA, but not to nondonor HLA, was significantly associated with poorer graft outcomes, with 1- and 3-year graft survival inferior in SPK recipients (85.2% vs. 93.5%; 71.8% vs. 90.3%, respectively; log-rank p = 0.002), and particularly in IP recipients (50.0% vs. 82.9%; 16.7 vs. 79.4%, respectively; log-rank p = 0.001). In a multivariate analysis, development of de novo DSA emerged as a strong independent predictor of pancreas graft failure (hazard ratio 4.66, p < 0.001). This is the largest study to examine de novo HLA antibodies following pancreas transplantation and clearly defines a high-risk group in need of specific intervention.


percentage calculated HLA antibody reaction frequency


antibody-mediated rejection


complement-dependent cytotoxicity


cold ischemia time


donors after brainstem death


donors after circulatory death


donor-specific HLA antibodies


flow cytometry


hazard ratio


isolated pancreas transplant


mean fluorescent intensity


pancreas after kidney transplant


pancreas transplant alone


simultaneous pancreas kidney transplant


Pancreas transplantation is an established treatment for patients with advanced or difficult-to-manage diabetes, offering insulin independence and improved quality of life. However, despite improvements in graft survival, medium-term graft attrition rates remain significant [1]. Pancreas graft rejection can be difficult to identify clinically, and by the time deranged blood sugar is evident, damage may be irreversible.

In other solid organ transplantation, studies over the last 10 years have shown that recipients with preformed anti-HLA antibodies have poorer outcomes than those who were unsensitized pretransplant, and that development of de novo HLA antibodies posttransplant has a role in allograft loss [2, 3]. Advancements in solid-phase assay techniques have greatly improved detection and specification of HLA antibody, resulting in greater sensitivity and precision [4, 5]. This has enabled the further discovery that circulating antibodies directed against donor HLA, or donor-specific HLA antibodies (DSA), confer the poorest allograft survival both when preformed before transplant and when formed de novo posttransplant. Several studies have investigated the role of DSA in kidney [6-9] and cardiac [10, 11] transplantation and, to a lesser extent, lung [12, 13] and liver [14, 15] transplantation. However, research into the importance of HLA antibodies in pancreas transplantation has been lacking with only two such series described to date [16, 17].

This study aims to assess the role of routine serial HLA antibody monitoring in identifying grafts at risk of failure after pancreas transplantation.

Materials and Methods

Patient cohort

This is a single-center, retrospective clinical study. The patient cohort included all recipients of deceased donor pancreas transplants performed at the Oxford Transplant Centre between 2006 and 2011: 317 simultaneous pancreas kidney (SPK) and 116 isolated pancreas (IP) transplants, including 68 pancreas transplants alone (PTA), 35 pancreas after kidney (PAK) transplants and 13 second pancreas transplants. The PTA, PAK and second pancreas transplant groups had equivalent graft outcomes. The SPK group had comparatively superior graft outcomes. Pretransplant HLA antibody status was available for all patients. Clinical and HLA antibody data were obtained at 1, 6 and 12 months and annually for the duration of follow-up, as well as at the time of clinical events. Seventy-nine patients were excluded from posttransplant HLA antibody analysis because serum samples were unavailable for analysis.

All transplants were performed with systemic venous drainage. Enteric ductal drainage was used in all except three recipients, who received IP transplants with bladder drainage due to a change in center protocol. All recipients received alemtuzumab induction immunosuppression (Campath 30 mg, days 1 and 2, Genzyme Corporation, Boston, MA). Maintenance immunosuppression therapy consisted of tacrolimus, initially at 0.5 mg/kg twice a day (bd) and titrated to maintain trough levels between 8 ng/mL and 10 ng/mL throughout the follow-up and mycophenolate mofetil, at 750 mg bd, with dose adjustments as clinically indicated. Trough serum tacrolimus levels and absolute neutrophil counts were monitored regularly on an outpatient basis, initially at least twice weekly for the first postoperative month, then with reducing frequency to monthly according to clinical status and stability. Mycophenolate levels were not monitored. No steroids were used in maintenance immunotherapy.

Data were entered into an anonymized database and analyzed as below.

HLA antibody analysis

All patients underwent routine pretransplant evaluation. Patients were typed for HLA-A, HLA-B, HLA-C, HLA-DR, HLA-DQ and, where appropriate, HLA-DP using polymerase chain reaction single-specific primer methods [18]. Pancreas transplants were categorized by HLA-A, -B, -DR mismatch: 0, 1 or 2, and by total 0–4 or 5–6 HLA mismatches. Sera were screened pretransplant for the presence of HLA antibodies by Luminex technology using LABScreen® Mixed kits and antibody specification performed using LABScreenPRA® and Single Antigen beads as appropriate (One Lambda Inc., Canoga Park, CA). Sensitization status was determined using the calculated HLA antibody reaction frequency (%cRF), defined as the percentage of HLA-incompatible, blood-group-compatible, donors in a pool of 10 000 UK donors. Recipients were considered sensitized if their %cRF was greater than 5%.

At the time of transplantation, all patients were crossmatched against their allocated donor by complement-dependent cytotoxicity (CDC) and flow cytometry (FC) crossmatch. CDC crossmatching was performed with preincubation of patient serum with dithiothreitol to distinguish between IgG and IgM antibodies. All transplants were CDC IgG negative, although a CDC IgM positive result was not considered a contraindication to transplant [19]. All transplants were FC crossmatch negative.

Routine prospective serial HLA antibody screening for HLA classes I and II antibodies was performed by Luminex technology at 0, 6 and 12 months postoperatively, then annually thereafter, as well as at the time of clinical events, as described above. A mean fluorescent intensity (MFI) value of 1000 was considered positive. Recipients were assessed for changes in their HLA antibody profile and development of DSA.

Clinical data

Demographic and graft outcome data were collected, including donor-recipient- and transplant-related variables. Data were collected regarding kidney rejection episodes, pancreas and kidney graft failures (defined as return to insulin therapy or return to dialysis respectively) and patient survival. Clinically evident and biopsy-proven kidney rejection episodes were treated with pulsed methyl prednisolone in the SPK group. Pancreas graft biopsies were not performed as part of our protocol, and as such pancreas rejection was not defined.

Statistical analysis

Quantitative parametric data were compared between groups using Student's t-test or the Mann–Whitney U test in the case of nonparametric distribution. Cross-tabulated data were analyzed by the chi-squared test or by the Fisher's exact test when the expected count was <5. Patient and death-censored graft survival and incidence of HLA antibodies were assessed using Kaplan–Meier curves and compared with the log-rank test. The Cox proportional hazard regression model, which allows for time-dependent variables, was utilized to estimate the impact of HLA antibodies, DSA and other covariates in graft survival. Variables with a significance level of p < 0.15 on univariate analysis were selected for inclusion in the multivariate model. Values of p < 0.05 were considered statistically significant. Statistical calculations were made using SPSS for Windows software (IBM SPSS Statistics version 20; Chicago, IL).



Four hundred thirty-three pancreas transplants were performed and included in the study. Three hundred seventeen (73.2%) received an SPK transplant, and 116 (26.8%) received an IP transplant. Two hundred ninety-three (92.4%) of SPK recipients received pancreases from donors after brainstem death (DBD) compared to 79 (68.1%) of IP recipients. This difference was statistically significant (p < 0.001) and was principally due to UK organ allocation procedures, where kidneys from donors after circulatory death (DCD) are allocated locally and without the priority to SPK recipients that are applied to DBD donor organs. The groups were otherwise comparable in terms of donor and recipient characteristics, and cold ischemia time (CIT) (Table 1).

Table 1. Demographics and pretransplant characteristics of study cohort by transplant type
  1. CIT, cold ischemia time; %cRF, percentage calculated reaction frequency; DBD, donors after brainstem death; IP, isolated pancreas transplant; SPK, simultaneous pancreas kidney transplant.
n317 (73.2%)116 (26.8%) 
Donor type (DBD)293 (92.4%)79 (68.1%)*<0.001
Donor female155 (48.9%)52 (44.8%)0.453
Donor age (median years)
Donor BMI (kg/m2)
Recipient female121 (38.2%)55 (47.4%)0.083
Recipient age (median years)
Recipient BMI (kg/m2)25.325.40.851
CIT (min)6847050.285
%cRF (>5%)88 (27.8%)26 (22.4%)0.263
HLA mismatch (5–6)98 (30.9%)34 (29.3%)0.748
DR mismatch   
032 (10.1%)23 (19.8%)0.026
1166 (52.4%)55 (47.4%) 
2119 (37.5%)38 (32.8%) 

Pretransplant immunological assessment

Pretransplant HLA antibody screening was performed routinely on all 433 patients.

One hundred fourteen (26.3%) were found to have pretransplant HLA antibodies and were considered to be sensitized (%cRF> 5%): 88/317 (27.8%) SPK, 26/116 (22.4%) IP. Sensitization status between the SPK and IP group was not significantly different, and comparison of pancreas graft outcomes showed that patients who were sensitized pretransplant had outcomes equivalent to unsensitized patients (log-rank p = 0.15; Figure 1A).

Figure 1.

Kaplan–Meier plots of death-censored pancreas graft survival according to pretransplant status. (A) Sensitization status (percentage calculated HLA antibody reaction frequency (%cRF) > 5%); (B) total mismatch; (C) DR mismatch; and (D) complement-dependent cytotoxicity (CDC) crossmatch IgM.

Analysis of donor and recipient HLA mismatching revealed that the SPK and IP transplant groups were comparable in terms of total HLA mismatch; however, the IP group had a lower level of DR mismatches compared to the SPK group (p = 0.026; Table 1). Neither degree of mismatch (0–4 vs. 5–6) nor degree of DR mismatch (0, 1 or 2) was associated with pancreas graft failure (log-rank p = 0.34 and p = 0.25, respectively; Figure 1B and C).

All donor and recipient CDC crossmatches were IgG negative but 100/317 (31.5%) SPK and 30/116 (25.9%) IP recipients had IgM positive results. This was not significantly different between groups (p = 0.253). IgM positive results are likely to represent autologous antibodies and comparison of graft outcomes showed that IgM positivity in the CDC crossmatch was not associated with inferior pancreas graft outcome after SPK or IP transplant (log-rank p = 0.46; Figure 1D).

Posttransplant de novo HLA antibody development

Posttransplant HLA antibody monitoring was performed in 354 (81.8%) patients. Seventy-nine patients did not have posttransplant antibody monitoring because adequate serum samples were not available. Pancreas graft failure occurred in 59/354 (16.7%) pancreas transplant recipients with available posttransplant monitoring.

De novo HLA antibodies developed in 134/354 (37.9%) transplant recipients with no difference between transplant groups (97 SPK vs. 37 IP). Kaplan–Meier comparison showed poorer pancreas graft survival in patients who developed de novo HLA antibodies, and this was statistically significant in the IP group (log-rank p = 0.011). Further analysis showed there was no association between the development of non-DSA and graft failure. However, the development of de novo DSA was significantly associated with poorer pancreas graft outcomes for both SPK and IP transplants (log-rank p = 0.002 and p = 0.001, respectively). De novo DSA developed in 52/354 (14.7%) patients, of which 34 were SPK and 18 IP transplants. Inferior 1- and 3-year graft survival rates were achieved in SPK recipients who developed de novo DSA compared to those who did not (1 year graft survival, 85.2% vs. 93.5%; 3-year survival 71.8% vs. 90.3%; log-rank p = 0.002), but the differences were more pronounced in the IP group (1 year graft survival, 50.0% vs. 82.9%; 3-year survival 16.7% vs. 79.4%) (Figure 2).

Figure 2.

Death-censored pancreas graft survival and development of de novo HLA antibodies after simultaneous pancreas kidney transplant (SPK; left panel) and isolated pancreas transplant (IP; right panel).

Biopsy-proven kidney rejection was significantly more common (11/34 or 32.4% vs. 17/226 or 7.5%; p < 0.001) and kidney graft survival was also significantly poorer in SPK recipients who developed de novo DSA (log-rank p > 0.001; Figure 3). Antibody-mediated rejection (AMR) was reported in two recipients with DSA. The remaining showed acute cellular rejection, with evidence of arteritis more common in recipients who developed DSA (2/9 vs. 2/17). Patients who developed DSA also had a significantly higher chance of losing both pancreas and kidney grafts (8/34 or 23.5% vs. 8/226 or 3.5%; p < 0.001). In the absence of pancreas biopsies, pancreas rejection was not easily characterized in our cohort and was not analyzed.

Figure 3.

Kidney graft survival after simultaneous pancreas kidney transplant stratified by development of de novo HLA antibodies.

Early nonimmunological graft failure (thrombosis or pancreatitis) was more common in those who did not develop DSA. Patients who developed DSA were more likely to suffer later, insidious, presumed immunological, pancreas graft failure (19/24 or 79.2% vs. 18/35 or 51.4%, p = 0.05).

Multivariate analysis demonstrated that donor and recipient factors were not predictive of pancreas graft failure, perhaps reflecting the narrow acceptance and listing criteria at our center. Recipients of IP transplants had poorer pancreas graft survival compared to those receiving SPK transplants (hazard ratio [HR] 2.51, p = 0.007). However, the development of de novo DSA emerged as the most predictive independent risk factor for pancreas graft failure (HR 4.66, p < 0.001) (Table 2).

Table 2. Multivariate Cox regression survival analysis for predictors of pancreas graft survival
  1. CI, confidence interval; CIT, cold ischemia time; %cRF, percentage calculated reaction frequency; DBD, donors after brainstem death; DSA, donor-specific antibodies; HR, hazard ratio; SPK, simultaneous pancreas kidney transplant.
Donor typeDBD1.010.45–2.270.989
Donor genderFemale0.710.37–1.360.301
Donor age (years) 1.021.00–1.040.125
Donor BMI (kg/m2) 0.990.96–1.020.592
Recipient genderFemale1.660.83–3.350.154
Recipient age (years) 1.000.96–1.030.868
Recipient BMI (kg/m2) 1.070.98–1.170.134
CIT (min) 1.001.00–1.000.869
HLA mismatch0–41.200.63–2.310.463
De novo DSANo DSA4.662.40–9.05<0.001

De novo DSA

Analysis of DSA specificities showed all loci to be represented, although recipients who developed both class I and class II DSA were more likely to have pancreas or pancreas and kidney graft failure than recipients who developed either class I or class II only (Table 3).

Table 3. Class of donor-specific HLA antibodies (DSA) formed after pancreas transplantation in the overall cohort, patients with pancreas graft failure and patients with failure of both transplant organs
 Overall n = 52Pancreas failed n = 24Both organs failed n = 11
Class I only21 (40.4%)2 (8.3%)2 (18.2%)
Class II only14 (26.9%)3 (12.5%)2 (18.2%)
Both17 (32.7%)19 (79.2%)7 (63.6%)
HLA-A22 (42.3%)16 (66.6%)7 (63.6%)
HLA-B29 (55.8%)15 (62.5%)5 (45.5%)
HLA-C7 (13.5%)3 (12.5%)2 (18.2%)
HLA-DR18 (34.6%)12 (50.0%)4 (36.4%)
HLA-DQ21 (40.4%)13 (54.2%)5 (45.5%)
HLA-DP1 (1.9%)0 (0%)0 (0%)

Comparing those patients who developed DSA to those who did not, there were statistically significant differences between the groups for donor age and recipient BMI. Notably, patients who were sensitized pretransplant were more likely to develop de novo HLA antibodies (51/134, 38.1% vs. 45/220, 20.5%) (Table 4).

Table 4. Demographics of groups according to posttransplant HLA antibody status
 DSA positive n = 52Nondonor HLA antibodies n = 82No de novo HLA antibodies n = 220p-Value
  1. ischemia time; %cRF, percentage calculated reaction frequency; DBD, donors after brainstem death; DSA, donor-specific HLA antibodies; SPK, simultaneous pancreas kidney transplant.
Transplant (% SPK)34 (65.4%)63 (76.8%)169 (76.8%)0.212
Donor type (DBD)47 (90.4%)73 (89.0%)189 (85.9%)0.592
Donor female19 (36.5%)44 (53.7%)102 (46.4%)0.152
Donor age (years)37.040.435.50.021
Donor BMI (kg/m2)23.524.924.90.820
Recipient female22 (42.3%)40 (48.8%)81 (36.8%)0.387
Recipient age (years)42.445.343.40.090
Recipient BMI (kg/m2)25.426.525.10.029
CIT (min)7256596940.164
%cRF (>5%)22 (42.3%)29 (35.4%)45 (20.5%)0.001
HLA mismatch (5–6)19 (36.5%)21 (25.6%)65 (29.5%)0.401
Pancreas failures (%)24 (46.1%)8 (9.8%)27 (12.3%)<0.001
Kidney failures (%)9 (17.3%)4 (4.9%)13 (5.9%)0.002

Analysis of the timing of DSA emergence in relation to graft failure did not reveal any temporal relationship. DSA emerged between 1 and 35 months posttransplant in both the patients who went on to pancreatic graft failure and those who did not. However, where pancreas graft failure did occur this occurred within 9 months of the appearance of DSA in all but two cases.


The prognostic significance of de novo DSA on transplant outcomes has been demonstrated in solid organ transplantation including kidney, heart and lung; however, evidence in pancreas transplantation has been limited [13]. This study is the largest series of its kind to date and uniquely describes a cohort of pancreas transplants performed in the modern era with homogeneous pretransplant assessment, immunosuppressive and operative management and postoperative monitoring. Previous to this analysis, there have only been two studies examining HLA antibodies after pancreas transplantation, both involving much smaller cohorts and with conflicting results [16, 17].

First, we have shown that de novo DSA developed in 14.7% of pancreas transplant recipients, similar to previous reports. In our large homogenous cohort, we have demonstrated that the development of DSA conferred significantly poorer pancreas and kidney graft survival. Furthermore, de novo DSA was a significant independent predictor of pancreas graft failure in a multivariate model. These findings are consistent with data from kidney transplantation [6, 7, 20, 21], with similar findings having been observed recently after heart transplantation [10, 11], lung transplantation [12] and possibly in liver transplantation [15].

Second, we have shown for the first time that the association between DSA and pancreas graft failure is more pronounced in the IP group. The reasons for this are not clear but may be important with respect to the known inferior long-term outcomes of IP compared to SPK transplants [1]. The hypotheses proposed to explain these include: (i) the greater immunocompetence of IP recipients in the absence of uremia [22]; (ii) an immunological effect of transplanting both organs simultaneously and (iii) the advantages of using the kidney in an SPK recipient as an early surrogate indicator of rejection [23]. Although kidney rejection is not a robust indicator of pancreas rejection [24], the lack of a kidney graft in the IP group may result in the under-treatment of some subclinical rejection episodes, contributing to poorer longer-term outcomes. The importance of acute rejection in association with the development of DSA has been noted in kidney transplantation [25] and it is notable that, in our study, SPK patients who developed DSA also experienced significantly more kidney rejection episodes.

Third, we have shown that pretransplant immunological factors including sensitization status and degree of HLA mismatch were not predictive of pancreas graft outcome. Preformed DSA have been shown to have a deleterious effect in kidney transplantation, even at very low levels [9]. No transplants included in this analysis were performed in the presence of preformed DSA, and all antibodies contributing to the recipient's pretransplant sensitization status were directed against nondonor HLA. Although the degree of HLA mismatch has been shown to continue to be important in kidney transplantation [7], there is no evidence in the literature that HLA mismatch impacts on graft survival after pancreas transplantation. This may be due to a high percentage of pancreas transplant recipients receiving relatively poorly matched grafts [1]—in our study, only 30/433 (6.9%) recipients received a transplant with fewer than two mismatches. This study is, therefore, not powered to demonstrate a statistical difference between well and poorly matched recipients.

Studies in other organ types have suggested the greater importance of class II DSA in the pathological processes leading to graft failure [7]. In this study, class II was not more represented and it appears that concurrent de novo DSA against both classes I and II HLA conferred poorer graft survival. This group appeared to be at very high risk and frequently lost both pancreas and kidney grafts. In this cohort, repeat transplantation and pregnancy were rare and did not significantly predict for the development of HLA antibodies. A total of 197 of 354 recipients (55.6%) received a blood transfusion postoperatively; however, this was not significantly different between those who developed HLA antibodies and those who did not, and did not predict for de novo DSA. We did observe that recipients developing HLA antibodies posttransplant were more likely to have been sensitized pretransplant. This may represent a more immune-reactive recipient; the presence of transient HLA antibodies or pretransplant sensitization status may be related to future DSA by epitope spreading.

The authors recognize that there are limitations to this paper. First, this is a retrospective analysis and so suffers the problems with bias and missing data associated with this type of study. Second, sample acquisition was possibly not frequent enough to determine a temporal relationship between formation of DSA and graft loss, or to clearly define the persistent or transient nature of de novo antibodies. In future studies, more frequent sampling may be informative. Given there is no clinical intervention of proven efficacy, antibody testing at 3-monthly intervals and at the time of clinical events is appropriate and practicable. Defining causation is difficult and, although we show that DSA are strongly associated with allograft failure, unknown variables may also have an impact on survival. Available data relating to drug levels do not suggest that compliance with immunosuppression was a significant problem in this cohort. Relatively, small numbers make it difficult to stratify failure groups or allow further more complex analyses. Third, a commonly used MFI cut-off was used for the purposes of this analysis. We recognize that there is debate surrounding a clinically relevant cutoff, and that limitations of the assay may affect MFI values. Loupy et al demonstrated that the presence of complement-binding HLA antibodies conferred the poorest kidney graft outcomes [26]. Such testing was not performed in this study, although this may be a valuable investigation. Fourth, pancreatic histology would have enabled analyses relating DSA emergence to findings of AMR [27]. The presence of C4d staining in combination with DSA has been correlated to poorer pancreas graft outcomes than DSA alone [28, 29]. It is not known whether additional histological data would inform management decisions in the presence of DSA, with or without clinical dysfunction, and this question should be addressed in future research. Having defined a sub-group of patients at high risk of immunological graft loss, it is clearly essential to define an effective clinical response to mitigate this risk. Intravenous immunoglobulins, anti-CD20 antibodies and protease inhibitors have been suggested as treatment options [30], but there is no strategy of proven efficacy and this remains, therefore, an area of vital research. In conclusion, this is the largest study to date to examine the association between de novo HLA antibodies following pancreas transplant and graft outcomes, and clearly demonstrates a strong association between development of DSA and subsequent pancreas graft failure.


The authors acknowledge the NIHR Biomedical Research Centre Oxford, who fund the first author's Clinical Research Fellowship, and the clinical team at the Oxford Transplant Centre.


The authors of this manuscript have no conflicts of interest as described by the American Journal of Transplantation.