Pretransplant donor-specific anti-HLA antibodies as predictors of early allograft rejection in ABO-compatible liver transplantation

Authors


  • The authors have no conflicts of interest to disclose regarding this article.

Address reprint requests to Alexandru I. Musat, M.D., Department of Medicine, University of Wisconsin School of Medicine and Public Health, 1685 Highland Avenue, 4227 MFCB, Madison, WI 53705-2281. Telephone: 608-263-4034; FAX: 608-265-5677; E-mail: aim@medicine.wisc.edu

Abstract

The significance of preexisting donor-specific HLA antibodies (HLA-DSAs) for liver allograft function is unclear. Our previous studies have shown that humoral alloreactivity frequently accompanies acute cellular rejection (ACR). In the present study, we set out to determine whether pretransplant HLA-DSAs correlate with clinically significant ACR in the first 90 days after transplantation and, if so, to determine their predictive values. Class I HLA-DSAs and class II HLA-DSAs were determined by single-antigen bead flow cytometry for 113 consecutive adult transplants. A statistical analysis was performed for data from 109 consecutive patients with graft survival greater than or equal to 90 days. All patients who developed biochemical graft dysfunction underwent liver biopsy for hematoxylin-eosin and complement component 4d staining. Cox proportional hazards models and associated hazard ratios revealed a significant association of pretransplant HLA-DSAs with clinically significant ACR: this association started with a mean fluorescence intensity (MFI) as low as 300 for both class I (hazard ratio = 2.7, P  < 0.01) and class II (hazard ratio = 6.0, P  < 0.01). Pretransplant HLA-DSAs were associated with an increased risk of ACR: P  < 0.01 for class I (42% versus 18%), P  < 0.001 for class II (37% versus 7%), and P  < 0.001 for either class I or II (36% versus 3%). Class I or II HLA-DSAs with an MFI ≥ 1000 had the best positive predictive value for clinically significant ACR at 46%, whereas class I or II HLA-DSAs with an MFI ≥ 300 had the best negative predictive value at 97.1%. Although our study was based on consecutive patients, it was limited by the relatively low number of single-center subjects. In conclusion, the present study indicates that pretransplant HLA-DSAs, even at low levels of allosensitization, correlate with the risk of clinically significant ACR. Our findings suggest that anti–human leukocyte antigen antibodies could serve as donor-specific markers of immunoreactivity to the liver graft. Liver Transpl 19:1132-1141, 2013. © 2013 AASLD.

Abbreviations
ACR

acute cellular rejection

AMR

antibody-mediated rejection

C4d

complement component 4d

CI

confidence interval

DBD

donation after brain death

DCD

donation after cardiac death

DSA

donor-specific antibody

FFPE

formalin-fixed, paraffin-embedded

HLA

human leukocyte antigen

HLA-DSA

human leukocyte antigen donor-specific antibody

MELD

Model for End-Stage Disease

MFI

mean fluorescence intensity

NPV

negative predictive value

PPV

positive predictive value

ROC

receiver operating characteristic

SAB

single-antigen bead

Confrontation with allogeneic cells will lead to both cellular and humoral immune reactions.[1] In contrast to kidney allografts,[2] the transplanted liver has been traditionally considered relatively resistant to humoral rejection.[3] There is, however, increasing evidence that antibody-mediated alloreactivity mechanisms play a role in the pathogenesis of liver allograft rejection (see Hübscher[4] and references therein). The results of clinical studies looking at the association between pretransplant human leukocyte antigen (HLA) antibodies and the subsequent risk of rejection in liver transplantation have been mixed. Although some studies have found an increased frequency of both acute cellular rejection (ACR) and chronic rejection in liver recipients harboring pretransplant anti-HLA antibodies,[5-16] others have failed to show this association.[17-23] Based on the detection of circulating human leukocyte antigen donor-specific antibodies (HLA-DSAs) by single-antigen bead (SAB) flow cytometry and liver immunolabeling with complement component 4d (C4d), our previous studies have indicated that humoral alloreactivity frequently accompanies both ACR and chronic ductopenic rejection.[24] The current study was aimed at (1) exploring the prevalence of HLA-DSAs among liver transplant recipients; (2) determining whether or not pretransplant HLA-DSAs detected by SAB flow cytometry are associated with a risk of clinically significant ACR (ie, histological rejection in association with biochemical graft dysfunction[25]) in the early postoperative period (≤90 days) and, if they are, gauging their positive predictive values (PPVs) and negative predictive values (NPVs); and (3) determining the lowest mean fluorescence intensity (MFI) for HLA-DSAs at which there is such an association. We believe that this is the first study to determine the sensitivities, specificities, PPVs, and NPVs of pretransplant HLA-DSAs for clinically significant ACR and to show that even a low level of allosensitization affects the risk of clinically significant ACR for the liver allograft. In addition, this study allowed us to estimate the incidence of severe antibody-mediated rejection (AMR) requiring antibody-depleting therapy in the early postoperative period.

PATIENTS AND METHODS

This study was performed using our prospectively maintained transplant database after the study protocol was approved by the institutional review board. The data for 113 consecutive ABO-compatible liver transplants performed for 110 adult patients (including 7 retransplants) at the University of Wisconsin between May 1, 2009 and October 28, 2010 were analyzed. For all patients, circulating HLA-DSAs were determined in serum samples obtained when the donor grafts became available. The HLA-DSA results did not influence the selection of the recipients or the initial immunosuppression. In order to allow a meaningful statistical analysis of the risk for ACR associated with pretransplant HLA-DSAs in the early postoperative period (≤90 days), we excluded 4 patients who lost their grafts during this period, so we performed the statistical analysis for the remaining 109 patients (102 first grafts and 7 retransplants) whose grafts survived for at least 90 days. We included both primary transplant recipients and retransplant recipients in the analysis under the presumption that if there is a relationship between pretransplant donor-specific antibodies (DSAs) and ACR, it should be relevant for both first-time recipients and retransplant recipients. All patients underwent induction with intravenous dexamethasone intraoperatively, and this was generally followed by triple-immunosuppressive therapy with calcineurin inhibitors (mainly tacrolimus), mycophenolate, and prednisone in decreasing doses. Four patients had cyclosporine substituted for tacrolimus. Twenty-nine patients with renal failure received basiliximab in the first postoperative week to allow recovery of kidney function before the introduction of tacrolimus. None of the study patients were treated empirically for rejection.

Detection of HLA-DSAs

Each recipient's serum was analyzed for HLA antibodies with LABScreen single-antigen class I and class II beads (One Lambda, Inc., Canoga Park, CA) on a Luminex instrument. All donors were routinely typed for HLA-A, HLA-B, HLA-C, HLA-DRB1, HLA-DQB1, HLA-DQA1, and HLA-DPB1. Recipients were typed for HLA-A, HLA-B, HLA-C, HLA-DRB1, and HLA-DQB1; typing for HLA-DQA1 and HLA-DPB1 was performed as needed to confirm anti-DQA1 and anti-DPB1 antibody specificities. High-resolution allele retesting was performed for donors to confirm allele specificity as required. Donor specificity for HLA antibodies was determined through a comparison of donor/recipient-mismatched HLAs with the anti-HLA antibody profile of each recipient. For the purposes of this study, strict, normalized MFI cutoffs were used to define DSAs. MFI cutoff thresholds were defined for each analysis and started as low as 300. The maximum single HLA-DSA MFI for each class, defined as the highest ranked donor-specific SAB, was used for the statistical analysis.

Liver Biopsy

Liver biopsy was performed for all patients who developed biochemical graft dysfunction as evidenced by deterioration in liver-related biochemical tests, including bilirubin, aminotransferases, alkaline phosphatase, and gamma-glutamyl transpeptidase. Patients whose liver biochemical tests had been normalizing satisfactorily or had already normalized did not undergo biopsy, and this was concordant with the previously defined “absence of biochemical graft dysfunction” in the early posttransplant period.[25] Liver core biopsy samples were obtained under ultrasound guidance with an automatic needle (BioPince, Medical Device Technologies, Inc., Gainesville, FL). All patients underwent a Doppler ultrasound examination of the liver to evaluate the patency of the bile ducts and liver vessels at the time of liver biopsy.

Histopathology

The liver tissue cores were fixed for rapid processing in 10% neutral buffered formalin for approximately 1 hour before paraffin embedding [ie, they were formalin-fixed, paraffin-embedded (FFPE)]. FFPE blocks were cut into 4-μm sections and stained with hematoxylin-eosin. Liver rejection was diagnosed with the Banff criteria.[26] The FFPE sections were subjected to C4d-specific immunohistochemistry with a 1:50 dilution of rabbit polyclonal C4d antibodies (American Research Products, Belmont, MA) and an automated immunostainer (Benchmark XT, Ventana Medical Systems, Tucson, AZ). For antigen retrieval, the FFPE tissue underwent deparaffinization and mild heat-induced retrieval with Cell Conditioning 1, a high-pH buffer from Ventana. A kidney with humoral rejection was used as a positive control for every case of C4d staining. A semiquantitative evaluation of C4d immunolabeling was performed through an assessment of the staining of the endothelium of the portal capillaries, portal vein, sinusoids, terminal hepatic venule, and hepatic artery. Nonspecific C4d deposition in the arterial elastic lamina was not counted. Diffuse staining was defined as C4d deposition in the portal microvasculature of 50% or more of the portal tracts. Portal C4d immunolabeling of less than 50% of the tracts was considered focal. The biopsy samples were read blindly by a single pathologist (R.M.A.).

Statistical Analysis

Continuous variables were summarized as means and standard deviations and were compared between groups with the Wilcoxon rank-sum test. Categorical variables were summarized as percentages and were compared between groups with Fisher's exact test.

ACR rates were estimated with Kaplan-Meier analysis and were compared between groups with the log-rank test. To assess factors potentially associated with this outcome, Cox proportional hazards models were performed. Patients with graft survival of at least 90 days were noted to have had (or not had) clinically significant ACR, which was defined as histological rejection in association with biochemical graft dysfunction, during this time frame. The comparison group comprised patients who underwent biopsy but did not have histological ACR plus patients who did not require biopsy because they did not develop biochemical graft dysfunction and, therefore, were assumed to not have clinically significant rejection. In order to determine whether pretransplant HLA-DSAs are an independent risk factor for ACR, the following factors were entered into a multivariate analysis: recipient age at transplant, sex, pretransplant blood transfusions, previous organ transplantation, donor age, liver type [donation after cardiac death (DCD) or donation after brain death (DBD)], cold ischemia time, poor kidney function at transplant (creatinine > 2 mg/dL or dialysis), a calculated (physiological) Model for End-Stage Disease (MELD) score > 15, autoimmune disease (autoimmune hepatitis, primary sclerosing cholangitis, or primary biliary cirrhosis) versus others, and posttransplant immunosuppression. The sensitivities, specificities, PPVs, and NPVs associated with different MFI values were estimated, and the resultant receiver operating characteristic (ROC) curve was plotted. P values < 0.05 were considered significant. SAS 9.1 statistical software (SAS Institute, Inc., Cary, NC) was used.

RESULTS

Comparisons of the characteristics of patients with pretransplant HLA-DSAs and patients without them are summarized in Table 1. Female sex and higher MELD scores were overrepresented in the HLA-DSA–positive group.

Table 1. Characteristics of 109 Recipients Who Had a Graft for at Least 90 Days
CharacteristicDSA-Negative (n = 35)DSA-Positive: MFI ≥ 300 (n = 74)P Value
  1. a

    The data are presented as means and standard deviations.

  2. b

    The blood transfusion status was unknown for 3 patients.

  3. c

    Creatinine > 2 mg/dL or dialysis.

  4. d

    Autoimmune hepatitis, primary sclerosing cholangitis, and primary biliary cirrhosis.

Age (years)a54.77 ± 7.8854.23 ± 10.70.79
Sex: female/male [n (%)]3 (9)/32 (91)28 (38)/46 (62)0.001
Prior blood transfusion [n (%)]b25 (71)62 (84)0.06
Primary transplant/previous transplant [n (%)]31 (89)/4 (11)71 (96)/3 (4)0.21
Poor kidney function [n (%)]c12 (34)33 (45)0.41
Calculated MELD score > 15 [n (%)]22 (63)66 (89)<0.01
Donor age (years)a43.51 ± 16.3341.74 ± 15.760.59
Donor type: DBD/DCD/living donation [n (%)]32 (91)/3 (9)/0 (0)63 (85)/10 (14)/1 (1)1.00
Cold ischemia time (hours)a7.67 ± 2.057.81 ± 2.660.79
Warm ischemia time (hours)a2.42 ± 12.741.42 ± 6.370.66
Primary liver disease [n (%)]   
Hepatitis C virus/hepatocellular carcinoma10 (29)/523 (31)/80.82
Alcoholic liver disease/hepatocellular carcinoma13 (37)/217 (23)/30.16
Nonalcoholic steatohepatitis/hepatocellular carcinoma3 (9)/014 (19)/20.26
Autoimmune hepatitis0 (0)0 (0) 
Primary sclerosing cholangitis/cholangiocarcinoma4 (11)/44 (5)/10.27
Primary biliary cirrhosis0 (0)6 (8)0.17
Alpha-1-antitrypsin deficiency2 (6)2 (3)0.59
Hepatitis B virus/hepatocellular carcinoma1 (3)/03 (4)0.42
Acute liver failure0 (0)2 (3)1.00
Cryptogenic0 (0)1 (1)1.00
Miscellaneous/hepatocellular carcinoma2 (6)/02 (3)/10.59
Autoimmune diseasesd4 (11)10 (14)1.00
Immunosuppression [n (%)]   
Steroids35 (100)74 (100)1.00
Tacrolimus/cyclosporine34 (97)/1 (3)70 (95)/3 (4)1.00
Mycophenolate35 (100)72 (97)1.00
Basiliximab9 (26)20 (27)1.00
Steroids, tacrolimus, and mycophenolate25 (71)50 (68)0.83
Steroids, basiliximab, tacrolimus, and mycophenolate9 (26)18 (24)1.00

Prevalence of HLA-DSAs

The mean maximum normalized MFI was 1528 ± 3540 (range = 0-20,286) for pretransplant class I HLA-DSAs and 1687 ± 3605 (range = 0-20,653) for class II HLA-DSAs. Figure 1 shows the prevalence of HLA-DSAs at the MFI thresholds used in our analysis.

Figure 1.

Prevalence of pretransplant HLA-DSAs in adult liver transplant recipients stratified by the maximum SAB MFI.

Clinically Significant ACR

Forty grafts (37%) underwent biopsy at least once, and this was prompted in all cases by the deterioration of liver-related biochemical tests. Twenty-eight of the 109 patients (26%) experienced 1 or more episodes of ACR in association with biochemical graft dysfunction that fulfilled the definition of clinically significant ACR during the first 90 days after transplantation. For 27 of these 28 patients (96%), the first episode of clinically significant ACR was detected in the initial indication biopsy sample 19.4 ± 17.8 days after transplantation. The remaining 69 patients (63%) did not require biopsy because they did not sustain deterioration in their liver biochemical tests, and it was therefore assumed that they did not have clinically significant ACR during the study period. These patients (n = 69) along with those who underwent indication biopsy because of the development of graft dysfunction but were found to not have histological ACR (n = 12) formed the comparison group (n = 81).

Endothelial C4d deposition was detected in 93% of the first biopsy samples showing ACR in the following patterns: diffuse portal (13/28 or 46%), focal portal (13/28 or 46%), and sinusoidal (12/28 or 43%). Sinusoidal C4d deposition was seen only in association with portal deposition and was present in 9 of the 12 samples (75%) with diffuse portal deposition and in 3 of the 12 samples (25%) with focal portal deposition. There was no correlation between the degree of C4d deposition and the Banff score for cellular rejection. Four patients, all positive for pretransplant HLA-DSAs, developed steroid-resistant rejection. The C4d deposition in these 4 cases along with the HLA-DSA status at the time of biopsy included the following: focal portal only (DSA-negative), diffuse portal only (DSA-positive), focal portal plus sinusoidal (DSA-positive), and diffuse plus sinusoidal (DSA-positive). The last 2 cases progressed to severe AMR associated with progressive increases in C4d deposition (discussed later).

Association of Pretransplant HLA-DSAs With the Risk of Clinically Significant Rejection

The pretransplant HLA-DSA status and the mean and median MFIs for patients with ACR and patients without ACR are summarized in Table 2.

Table 2. Pretransplant HLA-DSA Status and Mean and Median MFIs for Patients With ACR and Patients Without ACR
 ACR-Negative [n = 81 (74%)]ACR-Positive [n = 28 (26%)]P Value
  1. a

    The data are presented as means and standard deviations.

  2. b

    The data are presented as medians and ranges.

DSA class I or II [n (%)]46 (57)27 (96)< 0.001
Class I MFIa1080 ± 25473028 ± 54340.08
Class II MFIa1148 ± 24483350 ± 56650.06
Class I MFIb125 (0-11,836)388 (0-20,286) 
Class II MFIb388 (0-12,876)656 (0-20,653) 

A Cox proportional hazards regression showed a statistically significant association of pretransplant HLA-DSAs with the risk of clinically significant ACR (P = 0.0007 when the MFI was treated as a continuous variable; Table 3). In order to adjust for factors that differed between groups positive and negative for HLA-DSAs (Table 1), we performed a multivariate analysis accounting for potential confounding factors known to lead to HLA sensitization (ie, sex, transfusions, and prior transplantation) and factors that could increase the risk of rejection. The results of the multivariate analysis indicated that HLA-DSAs were an independent risk factor for ACR (P = 0.016) after we controlled for the following: recipient age (P = 0.21), sex (P = 0.45), pretransplant blood transfusions (P = 0.70), prior organ transplantation (P = 0.20), poor kidney function (P = 0.69), a physiological MELD score ≤ 15 (P = 0.24), autoimmune disease (P = 0.69), donor age (P = 0.32), liver type (DBD or DCD; P = 0.42), cold ischemia time (P = 0.73), and posttransplant immunosuppression regimen combinations [steroid, tacrolimus, and mycophenolate (P = 0.06) or steroid, basiliximab, tacrolimus, and mycophenolate (P < 0.99)].

Table 3. Association of Pretransplant HLA-DSAs at Different MFI Values With the Risk of Cellular Rejection According to Cox Proportional Hazards Modeling
ClassMFIACR Patients/Patients With MFI < Threshold (n/n)ACR Patients/Patients With MFI ≥ Threshold (n/n)Hazard Ratio95% CI
I≥30012/7016/392.71.3-5.8
≥40014/7414/352.41.2-5.1
≥50016/7912/302.21.0-4.6
≥60016/8112/282.41.1-5.1
≥70017/8311/262.31.1-4.9
≥80017/8311/262.31.1-4.9
≥90017/8411/252.51.2-5.3
≥100017/8511/242.61.2-5.6
≥200020/918/182.21.0-5.1
≥500022/956/142.00.8-5.1
II≥3003/4125/686.01.8-19.8
≥4006/4822/613.41.4-8.3
≥50010/5518/542.11.0-4.6
≥60011/6317/462.51.2-5.4
≥70015/7213/371.90.9-4.0
≥80016/7612/331.90.9-4.1
≥90017/8311/262.41.1-5.1
≥100017/8511/242.71.3-5.9
≥200020/928/172.51.1-5.8
≥500022/986/113.01.2-7.5
I or II≥3001/3527/7415.32.1-112
≥4003/4025/695.71.7-19
≥5005/4523/643.91.5-10.2
≥6006/5222/574.11.7-10.1
≥70010/6118/482.61.2-5.7
≥80011/6517/442.61.2-5.5
≥90012/7116/382.91.4-6.1
≥100012/7416/353.31.6-7.1
≥200016/8212/272.61.2-5.5
≥500020/918/182.31.0-5.2

Because we were interested not only in exploring the association of pretransplant HLA-DSAs with the risk of ACR but also in determining whether low levels of allosensitization (MFI < 1000) affected the risk of ACR, we stratified the patients as follows: we began with a lower MFI threshold of 300 and increased the MFI cutoffs in increments of 100. We found that there was an increasing risk of clinically significant ACR in patients with pretransplant HLA-DSAs, and this started with an MFI of 300 for class I HLA-DSAs, class II HLA-DSAs, and class I or II HLA-DSAs (Table 3).

According to the Kaplan-Meier analysis of patients with graft survival of at least 90 days (n = 109), pretransplant HLA-DSAs were significantly associated with a higher rate of ACR at an MFI ≥ 300: log-rank P < 0.01 for class I (42% versus 18%), log-rank P < 0.001 for class II (37% versus 7%), and log-rank P < 0.001 for class I or II (36% versus 3%; Fig. 2). The association was maintained for an MFI ≥ 1000 as follows: log-rank P < 0.01 for class I (46% versus 20%), log-rank P < 0.01 for class II (46% versus 20%), and log-rank P < 0.001 for class I or II (46% versus 16%).

Figure 2.

Kaplan-Meier analysis showing a significant association between pretransplant HLA-DSAs and lower ACR-free graft survival: (A) class I, (B) class II, and (C) class I or II. Red lines indicate patients with a maximum single-bead MFI ≥ 300 for pretransplant HLA-DSAs, whereas blue lines indicate patients with an MFI < 300.

Predictive Values of HLA-DSAs for Clinically Significant Cellular Rejection

The sensitivities, specificities, PPVs, and NPVs of HLA-DSAs for clinically significant ACR are summarized in Table 4. The highest PPV for ACR was observed for class I HLA-DSAs with an MFI ≥ 1000 and for class II HLA-DSAs with an MFI ≥ 5000 (46% and 54.6%, respectively). An absence of pretransplant HLA-DSAs (MFI < 300) had an excellent NPV for clinically significant rejection, and it appeared better for class II (93%) versus class I (83%). The NPV for having neither class I nor class II with an MFI ≥ 300 was 97%. Our analysis indicates that there was no clear-cut MFI value that maximized both the NPV and the PPV, with the PPV increasing with an increasing MFI at the expense of a decreasing NPV. The ROC analysis and the associated areas under the curve for class I and II HLA-DSAs as predictors of clinically significant ACR are presented in Fig. 3. In comparison with class I HLA-DSAs, class II HLA-DSAs appeared to nonsignificantly better predict ACR (P = 0.08).

Table 4. Sensitivities, Specificities, PPVs, and NPVs of HLA-DSAs for Cellular Rejection
ClassMFISensitivity (%)Specificity (%)PPV (%)NPV (%)
I≥30057.171.641.082.9
≥60042.980.342.980.3
≥100039.384.045.880.0
≥200028.687.744.478.0
≥300025.088.943.877.4
≥500021.490.142.976.8
II≥30089.346.936.892.7
≥60060.764.237.082.5
≥100039.383.945.880.0
≥200028.688.947.178.3
≥300021.491.446.277.1
≥500021.493.854.677.6
I or II≥30096.442.036.597.1
≥60078.656.838.688.5
≥100057.176.545.783.8
≥200042.781.544.480.5
≥300032.185.242.978.4
≥500028.687.744.478.0
Figure 3.

ROC curves for class I HLA-DSAs, class II HLA-DSAs, and class I or II HLA-DSAs as predictors of cellular rejection. The areas under the curve were as follows: 0.67 for class I HLA-DSAs (95% CI = 0.55-0.78), 0.69 for class II HLA-DSAs (95% CI = 0.57-0.80), and 0.74 for class I or II HLA-DSAs (95% CI = 0.53-0.78).

Incidence of Severe AMR Requiring Antibody-Depleting Therapy

Although most ACR cases were associated with C4d deposition indicative of humoral alloreactivity (Fig. 4), 2 patients fulfilled all the recently proposed restrictive diagnostic criteria[22] for full-blown AMR leading to severe cholestasis[5, 13, 22, 27] and requiring therapy specifically directed toward antibody depletion. On the basis of these data, the incidence of severe AMR in the early postoperative period was estimated to be 1.8% (2 cases among 109 recipients who had an allograft for at least 90 days) at our institution.

Figure 4.

Examples of C4d immunolabeling from 2 AMR cases. Note the strong C4d immunolabeling of the endothelium of the portal microvasculature and sinusoids. Panel number 1 and 2 from severe AMR case 1 and 2 respectively.

DISCUSSION

The present study has estimated the prevalence of pretransplant HLA-DSAs, has indicated that even a low level of allosensitization is associated with an increased risk of clinically significant ACR, and has provided insight into the PPVs and NPVs of pretransplant HLA-DSAs with low and high MFIs for clinically significant ACR in ABO-compatible liver transplantation. In addition, it has allowed us to estimate the incidence of severe AMR of the liver allograft requiring antibody-depleting therapy in the first 3 months.

The prevalence of HLA-DSAs in our study was comparable to the prevalence reported in another study using a similar methodology for HLA-DSA determination at a given MFI value.[28] That study, comprising 90 patients (including 10 retransplants) and using an MFI positivity cutoff of 2000, reported a prevalence of pretransplant HLA-DSAs of 22%, which was comparable to the prevalence of 25% at the same MFI in our cohort of 113 consecutive transplants. We examined, however, the frequency and distribution of pretransplant HLA-DSAs also on the basis of different maximum single-bead MFIs. In doing so, we found that the prevalence of pretransplant HLA-DSAs depends on where the MFI cutoff for positivity is set (Fig. 1). In our study, we stratified the patients with MFI cutoffs in increments of 100 for maximum single-bead MFIs less than 1000, and we found that MFIs as low as 300 had a clinical impact and were associated with an increased risk of clinically significant ACR (ie, histologically proven ACR accompanied by biochemical graft dysfunction). The MFI positivity cutoff of 300 is also concordant with recent studies in kidney recipients.[29, 30]

Our study shows that pretransplant class I or II HLA-DSAs with an MFI as low as 300 are associated with an increased risk of clinically significant ACR in the first 90 days after transplantation. Our results differ from the results of a recent study by Taner et al.,[28] who found no increase in ACR in patients with pretransplant HLA-DSAs. One important potential explanation for these divergent results is that this group examined only day 7 protocol biopsy samples and probably detected a significant number of subclinical histological ACR episodes of little clinical significance. This contention is supported by previous large studies showing that 32% of patients with histological ACR detected by protocol biopsy in the first few postoperative weeks have no associated biochemical graft dysfunction and require no adjuvant immunosuppression (subclinical rejection) because only a minority of these patients with histological rejection and no biochemical dysfunction will later develop abnormal biochemical graft dysfunction requiring additional immunosuppression.[25] Our study was based on indication biopsies that were prompted by the development of biochemical graft dysfunction and, therefore, captured all cases of clinically significant ACR that occurred within the first 90 days after transplantation, including the aforementioned minority of patients who might have progressed from subclinical rejection to biochemical graft dysfunction. The results of our study suggest that preformed HLA-DSAs may contribute to the development of clinically significant ACR; by extension, Taner et al.'s data may indicate that HLA-DSAs may not be important in subclinical ACR. Our results are concordant with those reported for kidney allografts in a study of comparable design and statistical analysis regarding the risk of rejection in crossmatch-negative recipients harboring low-MFI HLA-DSAs before transplantation[31] and with a recent study reporting a significantly higher incidence and severity of ACR for intestinal grafts within 90 days of transplantation in recipients with preexisting HLA-DSAs.[16] It is important to note that the use of lower MFI thresholds may create difficulty in distinguishing true antibody positivity from the assay background. On the other hand, it is clear from our own experience and that of others[30] as well as recent consensus guidelines[32] that higher cutoffs may exclude low-level allosensitization. Our studies indicate that even low levels of pretransplant DSAs confer a significantly increased risk of rejection.

The finding that pretransplant HLA-DSAs may predict clinically significant cellular rejection suggests that they may serve as donor-specific markers of immunoreactivity to the graft. Our study also shows that there is no clear-cut MFI value that maximizes both the PPV and the NPV (Table 4). For example, having either class I or II HLA-DSAs has the best PPV value of 46% for an MFI ≥ 1000, but having neither class I nor II with an MFI ≥ 300 has the best NPV of 97%. The PPV value of HLA-DSAs for ACR will probably improve further with the determination of HLA-DSA immunoglobulin G subclasses, as recent studies would indicate.[33] For practical reasons, we chose the maximum single-bead MFI for analysis. It is possible that a future analysis using the sum MFI may also lead to an improvement in the PPV for HLA-DSAs.

We believe that an important finding of the present study is that both class I HLA-DSAs and class II HLA-DSAs have an excellent NPV for rejection, and this suggests that the determination of HLA-DSAs may be useful when we are pondering the risk of rejection during weaning from immunosuppression. This finding is in line with studies suggesting that operationally tolerant liver transplant recipients lack donor-specific alloantibodies.[34, 35] In light of the excellent NPV for rejection (>97%), HLA-DSA determination has the potential of adding a significant safety layer to biopsy findings conducive to lowering immunosuppression, as recently proposed by the Banff working group,[36] in the immunosuppression minimization and weaning algorithms.

Our study had several limitations. Although it was based on consecutive patients, it was limited by the relatively low number of subjects from a single center. Incomplete data for HLA-DSAs contemporaneous with liver biopsy samples did not allow the determination of the magnitude of the persistence of preformed HLA-DSAs after transplantation or of the development of de novo anti-HLA antibodies, except for the few cases of steroid-resistant rejection and severe AMR that we reported. For the same reason, we could not determine in all patients the presence of humoral alloreactivity (ie, the simultaneous detection of both C4d deposition and posttransplant HLA-DSAs), which was approximated in this study with the detection of C4d deposition. The lack of protocol biopsy samples prevented the determination of possible subclinical cellular and/or humoral alloreactivity to provide a more complete picture of the consequences of harboring pretransplant HLA-DSAs.

In conclusion, pretransplant HLA-DSAs correlate with the risk of clinically significant ACR. On the basis of their predictive values, HLA-DSAs may serve as donor-specific markers of immunoreactivity to the graft for assessing the risk of rejection.

Ancillary