Prospective multicenter clinical trial of immunosuppressive drug withdrawal in stable adult liver transplant recipients


  • Potential conflict of interest: Dr. Larcier was employed by TcLand Expression S.A. Dr. Lerut received grants from Fresenius Biotech. Dr. Daoud consults for Medextens.

  • The study was conducted as part of the European Commission supported RISET consortium (Reprogramming the Immune System for the Establishment of Tolerance). Furthermore, the work was funded by TcLand Expression S.A. and by a grant from the Ministry of Education & Science, Spain (SAF2008-04092) to A.S-F. C.B. was supported by Fundación BBVA and by the Catalan Society of Transplantation. F.B. was supported by a research fellowship of the Deutsche Forschungsgemeinschaft (DfG reference BO 3370/1-1). CIBEREHD is funded by the Instituto de Salud Carlos III Spain.

  • See Editorial on Page 1529


Lifelong immunosuppression increases morbidity and mortality in liver transplantation. Discontinuation of immunosuppressive drugs could lessen this burden, but the safety, applicability, and clinical outcomes of this strategy need to be carefully defined. We enrolled 102 stable liver recipients at least 3 years after transplantation in a single-arm multicenter immunosuppression withdrawal trial. Drugs were gradually discontinued over a 6 to 9-month period. The primary endpoint was the development of operational tolerance, defined as successful immunosuppressive drug cessation maintained for at least 12 months with stable graft function and no histopathologic evidence of rejection. Out of the 98 recipients evaluated, 57 rejected and 41 successfully discontinued all immunosuppressive drugs. In nontolerant recipients rejection episodes were mild and resolved over 5.6 months (two nontolerant patients still exhibited mild gradually improving cholestasis at the end of follow-up). In tolerant recipients no progressive clinically significant histological damage was apparent in follow-up protocol biopsies performed up to 3 years following drug withdrawal. Tolerance was independently associated with time since transplantation (odds ratio [OR] 1.353; P = 0.0001), recipient age (OR 1.073; P = 0.009), and male gender (OR 4.657; P = 0.016). A predictive model incorporating the first two clinical variables identified subgroups of recipients with very high (79%), intermediate (30%-38%), and very low (0%) likelihood of successful withdrawal. Conclusion: When conducted at late timepoints after transplantation, immunosuppression withdrawal is successful in a high proportion of carefully selected liver recipients. A combination of clinical parameters could be useful to predict the success of this strategy. Additional prospective studies are now needed to confirm these results and to validate clinically applicable diagnostic biomarkers. (Hepatology 2013;58:1824–1835)


alanine aminotransferase


aspartate aminotransferase


alkaline phosphatase


classification and regression tree


complement-dependent cytotoxicity


chi-square automatic interaction detection


enzyme-linked immunosorbent assay


gamma-glutamyl transpeptidase


hepatitis C virus


human leukocyte antigen.

Liver transplantation relies on the use of nonspecific immunosuppressive drugs to prevent graft rejection. In a significant proportion of patients these drugs cause severe side effects, such as nephrotoxicity, cancer, metabolic disorders, and opportunistic infections,[1-4] which substantially contribute to morbidity and mortality. Transplantation tolerance, a state in which the graft is specifically accepted without the need of chronic immunosuppression, would be a way to circumvent these problems and thus a highly desirable therapeutic goal.[5] Tolerance can “spontaneously” develop in certain kidney and liver recipients who can discontinue conventional immunosuppressive drugs without incurring in rejection,[6-9] a phenomenon termed spontaneous operational tolerance. In liver transplantation a number of reports have described the feasibility of purposely withdrawing immunosuppression in selected liver recipients.[10-18] However, these were mostly retrospective studies or small single-center experiences lacking standardized entry criteria and follow-up algorithms. Thus, critical clinical aspects of the immunosuppressive drug withdrawal protocol such as prevalence of operational tolerance, stability of the tolerance phenotype, histopathology outcome, and clinical predictors of successful withdrawal have not been adequately defined and quantified in prior studies.

We performed a prospective multicenter clinical trial to assess the safety and applicability of immunosuppression drug withdrawal in stable adult liver transplant recipients, to estimate the prevalence of operational tolerance, and to identify the clinical variables influencing the success of immunosuppression withdrawal.

Patients and Methods

Selection of Patients

Patients were enrolled at Hospital Clinic Barcelona, University Tor Vergata Rome and University Hospitals Leuven. Participating investigators screened all recipients who visited their outpatient clinics for potential enrollment during the inclusion period. Inclusion criteria were: liver transplantation with >3-year posttransplant follow-up; no history of autoimmune liver disease; absence of graft rejection during the previous 12 months; aspartate aminotransferase (AST) and alanine aminotransferase (ALT) <2-fold and alkaline phosphatase (AP) <1.5-fold the local laboratory upper limit of normal; presence of comorbidities related to immunosuppressive drug use (arterial hypertension, diabetes, renal dysfunction), or risk of neoplasm development (defined by history of extrahepatic cancer, history of tobacco or alcohol consumption, or age >60 years).

Exclusion criteria were: age <18 years; pregnancy; alcohol or illicit drug abuse; psychiatric or medical illnesses considered incompatible with the safe conduct of the study (e.g., cardiac failure; chronic obstructive pulmonary disease; cancer; major depression); altered baseline liver function tests; baseline liver biopsy exhibiting: (1) signs of acute or chronic rejection according to Banff criteria; (2) portal inflammation in >50% of portal tracts not attributable to chronic hepatitis C virus (HCV) infection; (3) central perivenulitis in >50% central veins; or (4) fibrosis Metavir stage III-IV.

All patients provided written informed consent. The study protocol was approved by the ethics committees of all participating institutions and followed ICH guidelines for Good Clinical Practice in clinical trials. An independent clinical research organization (Medextens, Paris, France) monitored the trial and performed the statistical analysis in collaboration with the clinical investigators. identification: NCT00647283.

Study Design

This was a prospective cohort study of immunosuppressive drug withdrawal in adult stable liver transplant recipients. The participating clinical investigators agreed by consensus on the following working definitions.

Immunosuppressive Drug Withdrawal Regimen

Drug doses were gradually reduced ¼ to ½ every 3 weeks until the minimum feasible daily dose was reached. Doses were then spaced over progressively increasing intervals until one dose per week was reached, after which therapy was discontinued. In patients receiving mycophenolate in combination with a calcineurin inhibitor, mycophenolate was discontinued first over ∼3 months. Target completion time for drug discontinuation was 6 to 9 months.

Patient Follow-up and Surveillance Liver Biopsies

Liver function tests and follow-up visits were performed every 3 weeks during the withdrawal period and monthly during the 12 months after drug cessation. Protocol biopsies were performed at inclusion, 12 months after complete drug cessation, and whenever allograft dysfunction was detected. Following a protocol amendment it was decided to continue follow-up until database lock on 25th December, 2011, with 3-month follow-up visits and additional protocol liver biopsies performed at 36 months after complete drug cessation.

Definition and Management of Allograft Dysfunction

In patients with normal baseline liver function tests, allograft dysfunction was defined as any increase of AST, ALT, or both GGT (gamma-glutamyl transpeptidase) and AP above upper limits. In patients with abnormal baseline liver function tests allograft dysfunction was defined as an increase in AST or ALT >2-fold baseline levels and/or an increase in both GGT and AP >1.5-fold baseline levels. Whenever allograft dysfunction was detected, liver biopsy was immediately performed if AST or ALT increased >2-fold baseline levels and/or AP and GGT jointly increased >1.5-fold baseline levels. In cases of AST or ALT increases <2-fold baseline levels and/or GGT and AP <1.5-fold baseline levels, drug minimization was halted and liver function tests repeated within 14 days. Worsening or persistence of allograft dysfunction constituted an indication for liver biopsy, while improvement allowed the reinstitution of drug minimization without performing a liver biopsy. Isolated increases in GGT were not considered an indication for liver biopsy.

Definition of Rejection

Rejection was diagnosed by the combination of allograft dysfunction and characteristic liver biopsy findings according to Banff criteria.[19] Severe allograft dysfunction occurring during weaning and rapidly responding to immunosuppressive therapy was considered indicative of rejection in the absence of liver biopsy provided that other etiologies of graft dysfunction had been excluded.

Definition of Operational Tolerance

Patients who did not develop rejection were classified as operationally tolerant as long as immunosuppressive drug cessation was maintained for at least 12 months and no histopathologic evidences of acute and/or chronic rejection were observed.

Study Endpoints

The primary endpoint of the study was the development of operational tolerance, as defined above. Secondary endpoints were mortality, graft loss, severity of rejection episodes, time from minimization start to rejection, histological changes following immunosuppression withdrawal, normalization of allograft dysfunction following rejection, and change in adverse effects associated with immunosuppressive drugs 36 months after enrollment.

Histological Examinations

Hematoxylin-eosin and Masson's trichrome-stained sections were examined by local pathologists and subsequently reevaluated by a central pathologist (R.M.) who was blinded to all clinical and biological data. All clinical decisions were made on the basis of the local pathologists' reports only.

Anti-Human Leukocyte Antigen (HLA) Antibody Screening

Anti-HLA antibodies in patient sera were initially detected by Lambda Antigen Tray (One Lambda, Canoga Park, CA) for enzyme-linked immunosorbent assay (ELISA) HLA class I and II. The sera were further tested for HLA antibody specificities by complement-dependent cytotoxicity (CDC) against a panel of peripheral blood cells from 54 different donors in the absence and presence of dithiothreitol (DTT), a reducing agent that breaks pentameric IgM disulfide bonds but has minimal effect on IgG. Specificities were verified and, in some cases, additional specificities were found using flow cytometry, ELISA, or a Gen-Probe Luminex ID kit (Gen-Probe, San Diego, CA). All measurements were centrally performed at Leiden University Medical Center.

Statistical Analysis

The overall estimated rate of successful immunosuppressive drug withdrawal in liver transplant recipients is 20%.[10-18] Given that this percentage is still uncertain, depending on the sampling frame and the screening and follow-up criteria employed, sample size estimation was driven by the reestimation of its confidence interval. With a precision of ±10%, we estimated the need to include at least 80 patients. In order to confirm that this sample size would be sufficient to identify transcriptional biomarkers of tolerance in ancillary microarray studies, we conducted a permutation-based reanalysis of a previously reported retrospective cross-sectional study that identified transcriptional markers associated with tolerance.[20, 21] We estimated that at least 12 tolerant recipients would be required to identify tolerance-related gene expression markers with a false-discovery rate (FDR) <0.05 and that at least 80 patients would be required to enable a sample of at least 60 patients with usable data. Medextens performed data analyses according to a preestablished analysis plan. The primary endpoint was estimated on an intention-to-treat basis taking into account all enrolled patients, and on a per-protocol basis excluding withdrawn patients. All other endpoints were analyzed on a per-protocol basis excluding withdrawn patients. Dichotomous variables were compared employing the Pearson chi-square test or Fisher's exact test, while continuous variables were compared by means of Student's t test or the nonparametric Mann-Whitney rank-sum. Histological data were analyzed comparing individual scores using paired t test and repeated-measures analysis of variance (ANOVA). The probability of rejection over time was explored using the Kaplan-Meier method. Factors independently associated with the probability of being tolerant were explored in a multivariate logistic regression analysis. The contribution of each variable was estimated with the adjusted odds ratio (OR) and their 95% confidence interval (CI). Variables independently associated with operational tolerance were selected with automated backward stepwise regression, where variables explaining a statistically significant proportion of the variance were maintained in the model. All tests were two-sided and P < 0.05 was employed as the threshold for statistical significance without adjustment for multiple tests. To elaborate a classification rule to identify subgroups of recipients with high and low likelihoods of reaching the primary endpoint we performed a classification and regression tree (CART) analysis using the CHAID (CHi-square Automatic Interaction Detection) algorithm.[22] Ten-fold internal cross-validation was performed to select the most robust predictive model. Analyses were performed on available data without imputation for missing data. All data handling and statistical analyses were performed in STATA 10 software (StataCorp, College Station, TX) by means of auditable and re-runnable script files.


Study Patients

Between April 2006 and December 2008, participating clinical investigators screened 500 consecutive outpatient liver recipients for study eligibility (Fig. 1). Screening protocol liver biopsies were conducted on 122 patients and a total of 102 patients were enrolled (Table 1). Out of the 102 patients included in the intention-to-treat analysis, 41 (40.2%) successfully discontinued immunosuppressive drugs and were classified as tolerant, whereas 57 (59.8%) developed rejection and were classified as nontolerant. Four recipients were withdrawn from the study (due to noncompliance in two, withdrawal of informed consent in one, and biliary stricture in one) and classified as nontolerant. The 98 recipients who remained in the study throughout the entire follow-up period were included in the per-protocol analysis. Forty-one (41.84%) were classified as tolerant and 57 (58.16%) as nontolerant (Fig. 2). The mean follow-up period (time between initiation of drug weaning and database lock in December 2011) was 48.9 months (standard deviation [SD] 7.7 months).

Figure 1.

Screening and enrollment of patients.

Table 1. Baseline Characteristics of Evaluable Patients (Per-Protocol)
CharacteristicTotal Cohort (N = 98)Tolerant Group (N = 41)Nontolerant Group (N = 57)P Value
  1. Data are expressed as means ± SD. HCV denotes hepatitis C virus, HBV hepatitis B virus.

  2. a

    12 out of the 24 patients transplanted for HCV cirrhosis were RNA-HCV positive at the time of enrolment. The remaining 12 patients cleared HCV-RNA after a course of posttransplant antiviral therapy.

  3. b

    During the study period all patients with HBV cirrhosis were on prophylaxis with lamivudine and none were receiving anti-HBV immunoglobulin.

Age at weaning start (years)56±1161±1052±10.8<0.0001
Age at transplantation (years)47±1050±1045±110.05
Gender (% male)6883580.009
BMI (kg/m2)26.5±4.427.2± 4.726±4.20.196
Indication for liver transplantation (no. of patients)   0.154
HCV cirrhosisa24915 
HBV cirrhosisb231112 
Alcoholic cirrhosis261511 
Cryptogenic cirrhosis752 
Amyloidotic polyneuropathy505 
Fulminant hepatitis514 
Other causes808 
Time from transplant to minimization start (months)103±47131±4383±40<0.0001
Donor age (years)43±1940±1844±200.859
Previous episodes of rejection (%)2829260.824
Active RNA-positive HCV infection (no. of patients)a12750.738
Total HLA-A,B,DR mismatches4.8±1.14.6±1.24.8±1.10.343
Immunosuppressive therapy at weaning start (no. of patients)   0.003
Cyclosporin A261214 
Tacrolimus + mycophenolate541 
Cyclosporin A + mycophenolate1037 
Calcineurin inhibitor-based immunosuppression at weaning start (%)7056790.005
Immunosuppressive drug trough levels at weaning start (ng/mL)    
Cyclosporin A50±41.542.9±41.754.7±41.60.370
Liver function tests at weaning start    
Aspartate aminotransferases (U/L)28±1028±1228±91
Alanin aminotransferases (U/L)29±2132±2828±120.338
Gamma-glutamyl transpeptidase (U/L)36±5150±7027±280.0269
Alkaline phosphatase (U/L)171±64176±69168±620.556
Bilirrubin (mg/dL)0.9±0.590.99±0.690.79±0.490.102
Renal function tests at weaning start    
Creatinine (mg/dl)1.3±0.71.3±11.1±0.30.17
Glomerular filtration rate (MDRD) (mL/min/1.73m2)62±2561±3163±180.836
Figure 2.

Actuarial probability of being free from rejection after the initiation of immunosuppression weaning. Data shown correspond to the intention-to-treat analysis.

Tolerant Patients

In tolerant patients drugs were discontinued over a mean of 8.02 (SD 4.64) months. Mean time between cessation of immunosuppressive therapy and data extraction on December 2011 was 41.8 (SD 9.2) months. Twenty-three tolerant patients underwent one episode of transient allograft dysfunction not attributable to allograft rejection. In four patients who met the prespecified criteria to undergo liver biopsy, histological analysis excluded rejection. All episodes spontaneously resolved without increases in immunosuppression doses. No significant changes in liver chemistry were noted when baseline and end-of-follow-up tests were compared. Two tolerant patients died with normal graft function 16 and 26 months after complete drug cessation due to postoperative complications of colon cancer resection and metastatic ovarian cancer, respectively. None of the tolerant recipients reaching the primary endpoint developed rejection during follow-up.

Nontolerant Patients

In the 57 nontolerant patients acute rejection occurred a mean of 6.44 months after the start of drug minimization (SD 4.37, range 1.28-21.35; Fig. 2). Mean liver function tests at the time of rejection were AST 130 IU/L (range 30-590), ALT 181 IU/L (range 45-769), AP 271 U/L (range 38-795), GGT 140 IU/L (range 9-683), and bilirubin 0.9 mg/dL (range 0.3-4.8). All nontolerant patients were still receiving immunosuppressive drugs at the time of rejection (67% were on daily doses, 13% received medication every 48 hours, 7% once every 72 hours, and 3% less than once every 72 hours). Liver biopsies were obtained in 89% of cases and graded by the central pathologist as indeterminate (46%), mild (33%), moderate (16%), or severe (5%) rejection according to Banff criteria (Table 2). No cases of chronic rejection were observed. In 21 patients rejection was treated with reinstitution of baseline immunosuppression. In the remaining patients reinstitution of baseline immunosuppression was combined with low-dose steroids (20 mg/day over 4-6 weeks; 30 patients), moderate-dose steroids (40-60 mg/day over 4-6 weeks; four patients), and steroid boluses (one patient; Table 2). In one patient baseline immunosuppression was substituted by sirolimus plus mycophenolate. Treatment of rejection required hospitalization in two patients. Allograft dysfunction due to rejection fully normalized in 55 recipients (96.5%) a mean of 5.6 (SD 5) months after diagnosis. At the end of the follow-up period the remaining two recipients had normalized transaminase and bilirubin levels and exhibited gradually improving GGT and AP levels (in one of these two patients resolution of rejection was confirmed by a follow-up liver biopsy and the persistent cholestasis attributed to low-grade biliary stricture). No grafts were lost due to rejection. The 11% of nontolerant recipients in whom liver biopsies were not obtained at the time of rejection had similar clinical outcomes to the remaining 89% of patients (e.g., type of rejection treatment and time until resolution).

Table 2. Characteristics of Rejection Episodes According to Treatment Received
 Baseline IMSBaseline IMS + Prednisone 20 mg/dayBaseline IMS + Prednisone 40-60 mg/daySteroid BolusesP Value
  1. Data are expressed as median (range).

  2. a

    At the time of database lock this patient had already normalized transaminase levels but displayed mildly increased gamma-glutamyl transpeptidase and alkaline phosphatase levels.

Number of patients213041 
Age (years)54 (27-67)58 (30-71)42 (25-56)590.235
Gender (% male)52%70%50%100%0.489
Time from transplatation to weaning start (months)72 (37-234)93 (29-187)41 (39-47)840.108
Time from weaning start to rejection (months)4.8 (1.4-21)6.2 (1.2-17)3 (1.6-4.3)6.40.119
Severity of rejection (number of patients)     
Banff score4 (3-7)3 (2-7)7 (6-9)60.004
Highest AST level (UI/L)84 (37-449)80 (30-447)357 (101-590)1040.000
Highest ALT level (UI/L)113 (57-769)120 (48-662)447 (181-690)1760.012
Highest bilirubin level (mg/dL)0.8 (0.2-3.5)0.7 (0.4-1.5)1.5 (0.2-2.2)33.90.000
Highest gamma-glutamyl transpeptidase levels (UI/L)86 (28-533)107 (9-362)339 (213-416)3180.020
Highest alkaline phosphatase levels (UI/L)222 (38-580)210 (127-612)537 (273-795)3370.114
Time until rejection resolution (months)3.4±2.75.1±2.918±10Persistent low grade cholestasis a0.000

Comorbidities and Adverse Effects

There were no significant changes between baseline and 36-month follow-up timepoints in the prevalence and severity of comorbidities influenced by immunosuppressive therapy for either tolerant or nontolerant recipients (Table 3). The overall number of adverse events other than rejection was also similar for the two groups of patients.

Table 3. Evolution of Recipient Comorbidities Over the 36 Months Following the Initiation of the Study (Per Protocol)
ComorbidityEvolutionTolerant (n = 41)Nontolerant (n = 53)P Value
  1. a

    Improvement was defined as a reduction in the number or dose of medications required to control a preexistent comorbidity. Worsening was defined as the onset of a comorbidity or as the need to increase the dose or the number of medications required for its treatment. Results were based on the blinded evaluations of a single investigator.

  2. b

    Cardiovascular complications included ischemic heart disease and/or stroke.

  3. c

    Any infection requiring in-hospital admission and treatment.

  4. d

    GFR was calculated using MDRD-4.

HypertensionaImprovement6 (11%)7 (17%)0.7
 Stabilization30 (73%)40 (76%) 
 Worsening4 (10%)7 (13%) 
DiabetesaImprovement2 (5%)3 (6%)0.7
 Stabilization37 (90%)49 (93%) 
 Worsening2 (5%)3 (6%) 
HypercholesterolemiaaImprovement3 (7%)1 (2%)0.4
 Stabilization36 (88%)49 (92%) 
 Worsening2 (5%)3 (6%) 
HypertriglyceridemiaaImprovement4 (10%)2 (4%)0.5
 Stabilization36 (88%)50 (94%) 
 Worsening1 (2%)1 (2%) 
Cardiovascular ComplicationsbNo37 (92%)49 (92%)0.6
 Yes3 (8%)4 (8%) 
Bone fracturesNo40 (98%)50 (94%)0.4
 Yes1 (2%)3 (6%) 
InfectionscNo41 (100%)51 (96%)0.3
 Yes02 (4%) 
NeoplasmsNo40 (98%)53 (100%)0.4
 Yes1 (2%)0 
Renal function (change in glomerular filtration rate between enrolment and 36 months after drug withdrawal or rejection; ml/min/1.73 m2)d + 4.9 ± 10.9+ 4.5 ± 10.10.8

Histological Analyses

All enrolled recipients underwent liver biopsy before initiating drug minimization. Follow-up liver biopsies were performed 12 months after drug cessation in 40 recipients. Additional liver biopsies performed 36 months after drug cessation were available from 33 out of the 38 (87%) tolerant recipients reaching this timepoint (3-year postweaning biopsies performed following database lock were also included in this analysis). Tolerant and nontolerant recipients exhibited similar liver histopathologic findings at baseline. Analysis of the three sequential liver samples obtained from tolerant recipients employing a repeated-measures ANOVA revealed a significant increase in lobular inflammation over time (Table 4). This increase was mild (not more than 1 grade in the scoring scale) and correlated with development and/or worsening of macrovesicular steatosis (which increased from a mean of 6% at baseline, to 10% and 20% at 1 and 3 years after drug withdrawal, respectively; P < 0.001). A paired comparison between baseline and 1-year postwithdrawal biopsies showed increase in portal inflammation, interface hepatitis, and lymphocytic cholangitis. These changes, however, were mild and could no longer be observed 3 years after drug discontinuation. In the small subgroup of HCV RNA-positive recipients, tolerant and nontolerant recipients displayed no significant differences at baseline, and only portal fibrosis significantly worsened over time (Table 4).

Table 4. Central Review of Liver Histopathology Findings Obtained During the Course of the Study (Per Protocol)
Histological EvaluationBaselineRejectionP ValueBaseline12.Month Postweaning36-Month PostweaningP Value
  1. Data are expressed as median (range).

  2. Lobular inflammation: 0 = no; 1 = mild (sinusoidal cells and/or mild focal necrosis); 2 = moderate (multiple necro-inflammatory foci); 3 = marked (confluent or bridging necrosis). Central perivenulitis (with or without endothelitis): 0 = no; 1 = mild (patchy, focal perivenular inflammation); 2 = moderate (perivenulitis in most central veins); 3 = marked (confluent or bridging hepatocellular necrosis). Portal inflammation: 0 = no; 1 = mild (small groups of inflammatory cells); 2 = moderate (>50% of portal tracts, expansive); 3 = marked. Interfase hepatitis: 0 = no; 1 = mild; 2 = moderate; 3 = severe. Bile duct lesions: 0 = no; 1 = minimal (intraepithelial inflammatory cells or abnormal colangiocytes); 2 = moderate (epithelial lesions in most of portal tracts, no destruction); 3 = marked. Bile duct loss: 0 = no; 1 = <50%; 2 = ≥ 50%. Portal vein branches: 0 = present in all portal tracts; 1 = absent in a minority of portal tracts; 2 = absent in most of the portal tracts. Portal vein endothelitis: 0 = no; 1 = mild (minority of portal veins); 2 = mild (most of the portal veins); 3 = marked. Portal fibrosis: 0 = no; 1 = minimal (minority of portal tracts); 2 = moderate (most of the portal tracts, periportal expansion); 3 = bridging fibrosis; 4 = cirrhosis. *P < 0.05 compared to baseline biopsy.

Number of patients       
HCV-RNA negative5249 342929 
HCV-RNA positive55 764 
Number of complete portal tracts       
HCV-RNA negative7 (1-34)9 (2-22)0.987 (1-22)6 (2-18)8 (1-18)0.766
HCV-RNA positive6 (4-15)10 (6-17)0.087 (2-8)8 (7-11)8 (3-11)0.562
Number of central veins       
HCV-RNA negative5 (0-19)7 (1-13)0.2644 (0-18)5 (0-13)6 (2-12)0.877
HCV-RNA positive5 (2-11)6 (0-7)0.063 (1-7)5 (0-7)2 (1-5)0.775
Lobular inflammation       
HCV-RNA negative0 (0-2)1 (0-2)<0.0010 (0-2)1 (0-2)*1 (0-2)*<0.001
HCV-RNA positive1 (1-2)1 (1-2)11 (0-1)1 (1-2)1 (0-1)0.44
Central perivenulitis       
HCV-RNA negative0 (0-2)0 (0-2)0.0480 (0-2)0 (0-2)0 (0-2)0.404
HCV-RNA positive0 (0-2)2 (1-3)0.2700 (0-1)1 (0-1)0 (0-1)1
Portal inflammation       
HCV-RNA negative1 (0-2)1 (1-3)<0.0011 (0-2)1 (0-2)*1 (0-2)0.069
HCV-RNA positive2 (1-2)2 (1-3)11 (1-2)1 (0-2)1 (1-3)0.577
Interface hepatitis       
HCV-RNA negative0 (0-2)1 (0-3)<0.0010 (0-2)0 (0-2)*0 (0-1)0.095
HCV-RNA positive1 (0-2)2 (1-3)0.1820 (0-1)0.5 (0-1)1 (0-2)0.250
Bile duct lesions       
HCV-RNA negative0 (0-3)1 (0-3)<0.0010 (0-1)0 (0-1)*0 (0-1)0.056
HCV-RNA positive1(0-1)1(1-2)0.3910.5 (0-1)0 (0-1)00.423
Bile duct loss       
HCV-RNA negative0 (0-1)0 (0-1)0.5700 (0-1)0 (0-1)0.157
HCV-RNA positive0 (0-1)00.39100 (0-1)0 
Portal vein branches       
HCV-RNA negative0 (0-2)0 (0-2)0.4740 (0-1)0 (0-1)0 (0-2)0.072
HCV-RNA positive0 (0-2)1 (0-1)0.3910 (0-2)0 (0-2)1 (1-2)0.192
Portal vein endothelitis       
HCV-RNA negative0 (0-1)1 (0-3)<0.0010 (0-1)0 (0-1)0 (0-1)0.321
HCV-RNA positive0 (0-1)1(0-2)0.7180 (0-1)0 (0-2)00.500
Portal fibrosis       
HCV-RNA negative0 (0-2)1 (0-3)<0.0010 (0-3)0 (0-2)0 (0-2)0.929
HCV-RNA positive1.5 (0-2)3 (0-3)0.0581 (0-2)2 (1-2)30.05

Clinical Parameters Associated With Successful Weaning

Time from transplantation to enrollment was significantly higher in the tolerant group (Fig. 3A). Successful drug withdrawal was also associated with male gender (OR 3.53; 95% CI 1.34-9.3; P = 0.009), older recipient age at transplantation (OR 1.041; 95% CI 1-1.084; P = 0.05; Fig. 3B), and absence of calcineurin inhibitors in the immunosuppressive drug regimen employed at enrollment (OR 0.2; 95% CI 0.06-0.6; P = 0.005). In the multivariate analysis, only time since transplantation (OR 1.353; 95% CI 1.166-1.570; P < 0.0001), age at transplantation (OR 1.073; 95% CI 1.018-1.130; P = 0.009), and gender (OR 4.375, 95% CI 1.315-14.412; P = 0.016) were independently associated with the probability of being tolerant (C-statistic 0.816; 95% CI 0.733-0.9). Time after transplantation was identified as the strongest predictor of outcome in a classification and regression tree analysis (Fig. 3C). Patients enrolled more than 10.6 years after transplantation had a 79% rate of successful withdrawal, while in those transplanted for more than 5.7 years withdrawal was successful in 38% of cases. In these two subgroups of recipients with high and intermediate likelihood of tolerance, incorporation of age or gender did not improve the model's predictive capacity. In contrast, among recipients enrolled 3-5.7 years after transplantation, age at transplantation helped to identify subgroups of recipients with very low (0%) or intermediate (30%) likelihood of success.

Figure 3.

Influence of recipient age and time since transplantation on the probability of reaching the primary endpoint. (A,B) Relationship between the estimated probability of reaching the primary endpoint and the time elapsed since transplantation, and between the estimated probability of reaching the primary endpoint and recipient age at transplantation, respectively. A prognostic analysis generated by classification and regression tree analysis is shown (C).

Anti-HLA Antibodies

Anti-HLA antibodies were measured in baseline preweaning serum samples from 86 recipients and in serum samples collected 36 months after enrollment in 40 recipients. ELISA experiments revealed that the prevalence of anti-HLA class I and class II antibodies at baseline was similar in tolerant and nontolerant recipients (11% and 14% for class I (P = 0.791), 31% and 20% for class II (P = 0.162), respectively). None of the 86 recipients displayed detectable complement binding anti-HLA antibodies as assessed by CDC. When baseline measurements were compared with those obtained at the end of follow-up, de novo anti-HLA antibodies were identified in 5 out of 23 nontolerant recipients and in 1 out of 15 tolerant (P = 0.238). On the other hand, in three nontolerant recipients anti-HLA antibodies present at baseline were no longer detectable at the end of the study.


In liver recipients surviving more than 1 year after transplantation mortality is still substantially higher than what would be expected in the general population and has not significantly improved over the past 20 years.[23, 24] Aside from recurrent HCV infection, the major causes of late mortality in the adult liver transplant population are malignancies and metabolic and cardiovascular complications.[4, 25] Immunosuppression has a negative impact on the development of these comorbidities. In an attempt to improve long-term outcomes most liver transplant programs minimize exposure to immunosuppressive therapy by gradually decreasing drug dosages. In addition, some centers occasionally attempt to withdraw all immunosuppression from selected liver recipients, a strategy that is facilitated by the unique tolerogenic microenvironment of the liver. Initial reports described a 5.6% to 38% rate of successful drug withdrawal in selected liver transplant recipients.[10-18] Due to methodological shortcomings, however, these prior studies failed to provide the medical community with rigorously collected and generalizable knowledge on the outcome and safety of drug withdrawal. More recently, Feng et al.[26] described the results of the first multicenter drug withdrawal trial in pediatric recipients of parental living donor liver transplants. Although the number of enrolled patients was small, they were all treated according to a very rigorous protocol and meticulously monitored with sequential protocol liver biopsies. The study concluded that an unexpectedly high proportion (60%) of carefully selected pediatric liver recipients could remain off immunosuppression while maintaining normal graft function and histology. We report here the results of the first multinational trial of complete immunosuppressive drug withdrawal in stable liver transplant recipients. In all, 102 recipients were enrolled, and the trial was successful in 41.84% of evaluable recipients. Although 58.16% of recipients rejected, rejection episodes tended to be mild and could be resolved in all cases (two nontolerant patients still exhibited mild cholestasis at the end of follow-up but this was gradually improving). Furthermore, in tolerant recipients no clinically significant progressive detrimental histological changes were observed in sequential liver biopsies conducted as long as 3 years after complete drug withdrawal. Over a relatively short follow-up period the overall strategy was therefore safe for both tolerant and nontolerant recipients.

The identification of clinical parameters associated with a high likelihood of successful drug withdrawal is critical in order to select patients in whom the potential benefits of drug discontinuation outweigh the risks of rejection. Time from transplantation to enrollment was the most powerful baseline clinical predictor of successful immunosuppressive drug withdrawal. The strength of this variable had not been well recognized in the adult literature before. This is probably due to the fact that in the initially reported studies proportionately few long-term surviving recipients were enrolled.[10-18] Our observation is consistent with the well-known but incompletely understood phenomenon of host-graft adaptation,[1, 30] which is thought to account for the decrease in graft immunogenicity and reduced risk of rejection that occurs following the first few posttransplantation months. The observation that older recipients had higher odds of being tolerant is consistent with the lower rates of rejection observed in elderly kidney and liver graft recipients,[31] and suggests that immunosenescence might be involved in the drug-free acceptance of liver allografts. While the explanation for the increased prevalence of tolerance in male as compared to female recipients is unclear, the observation is consistent with results from two previous studies in kidney and in liver transplantation[32, 33] and needs therefore to be further investigated. While the use of non-calcineurin inhibitor (CNI)-based immunosuppression at enrollment was more frequent in tolerant than in nontolerant recipients, this variable was found to be colineal with recipient age and time after transplantation and was not independently associated with drug withdrawal outcome. Our study provides no evidence, therefore, indicating that specific immunosuppressive regimens are more effective than others at promoting the development of operational tolerance, although the effects of therapeutic strategies such as mammalian target of rapamycin (mTOR) inhibition or costimulation blockade remain to be explored. Incorporation of clinical variables independently associated with tolerance into a predictive model allowed the identification of subgroups of recipients with high, low, and intermediate probability of successful drug withdrawal. These results should provide guidance for future clinical trials of drug withdrawal. For instance, they could serve to select the subgroups of patients most likely to benefit from novel tolerance-promoting therapeutic strategies and/or diagnostic biomarkers of tolerance.

Grafts from tolerant patients exhibited increased inflammatory infiltrates 12 months after drug withdrawal. These changes, however, were mild and could no longer be detected 24 months later. The observation that tolerant recipients develop transient subclinical and spontaneously resolving inflammatory infiltrates is reminiscent of what has been described in experimental animal models,[34] and could represent a sign of tolerogenic immune-engagement. This phenomenon could also explain why 54% of tolerant patients developed episodes of mild transient allograft dysfunction during weaning, albeit in our study no significant correlation between the two events was found (data not shown). Careful long-term histological follow-up will be critical to exclude progressive low-grade graft damage and to elucidate the clinical significance of macrosteatosis in these patients. In our study 20 recipients were excluded at screening due to abnormalities in the baseline liver biopsy, mostly related to the presence of idiopathic chronic hepatitis.[27-29] These observations emphasize the need to include baseline and follow-up protocol liver biopsies as an essential element in any drug withdrawal strategy. Patient enrollment was intentionally performed employing a relatively loose set of histological eligibility criteria. The use of more stringent criteria would likely further reduce the number of eligible recipients. It should be noted, however, that in our study baseline histology characteristics were not predictive of drug withdrawal outcome.

Development of antibodies against donor HLA molecules markedly influences immunologic outcomes in kidney and heart transplant recipients. In our study, in contrast, no differences in anti-HLA antibody titers were noted between tolerant and nontolerant recipients. Furthermore, in the tolerant cohort drug withdrawal was not associated with the development of new anti-HLA antibodies. Our data suggest that humoral alloreactivity is unlikely to be involved in the development of liver allograft tolerance. However, a more precise elucidation of the relevance of this mechanism incorporating measurement of donor-specific anti-HLA antibodies and liver tissue C4d deposition is warranted.

Our study has a number of limitations. The main one is the possibility of a selection bias derived from the fact that out of the 341 recipients with more than 3-year posttransplant follow-up at screening only 30% were enrolled in the study. This low rate of applicability was due to the high level of cautiousness with which the study was designed and conducted. It should be noted, however, that the reasons why almost two-thirds of the 239 excluded patients were not enrolled were either extrahepatic medical conditions or logistical issues inherent to the very restrictive clinical protocol employed (Fig. 1). In daily clinical practice no formal contraindications would exist for the possibility of weaning immunosuppression from these patients. Indeed, only 83/239 screened but not enrolled patients were excluded due to conditions known to increase the risk of rejection following drug withdrawal (autoimmunity, recent episode of rejection, grossly abnormal histology and/or liver function tests). Whether our selected cohort is truly representative of the population of nonautoimmune, non-HCV, and Caucasic liver recipients, however, remains to be formally demonstrated. Inclusion of long-term surviving recipients could also have resulted in the enrollment of patients with predominantly low immunologic risk. This seems unlikely, however, given that graft losses due to immunologic events are uncommon in liver transplantation.[23, 24] A longer follow-up is needed, given that tolerant recipients can occasionally reject many years after drug discontinuation.[35, 36] Demonstrating whether weaning results in reduced immunosuppression-related comorbidities will also necessitate prolonged follow-up and much larger sample sizes considering that these patients were already receiving low-dose immunosuppression at enrollment. Ultimately, to prove that immunosuppression withdrawal results in clinical benefits, a randomized controlled trial will need to be undertaken. Even then, benefits might not be apparent unless drug withdrawal is achieved early after transplantation, before long-term exposure to immunosuppression-related toxicities has occurred. Despite these limitations, our data confirms in the adult liver transplant population the recent results by Feng et al.,[26] and challenges the current paradigm that lifelong immunosuppression is required for the health and well-being of all liver transplant recipients. These two studies will serve as a guideline on how to select recipients for future drug withdrawal trials and on the specific timeframe after transplantation when such studies ought to be implemented.

In conclusion, our findings indicate that the proportion of adult liver recipients who can discontinue immunosuppressive drugs is higher than previously estimated, particularly in selected non-HCV and nonautoimmune patients in whom drug withdrawal is performed a long period of time after transplantation. In this group of recipients immunosuppressive drug weaning is a safe procedure if performed under close medical supervision and, at least during the first 3 years after drug discontinuation, is not associated with clinically significant detrimental histological changes. Additional prospective studies are now needed to confirm the generalizability of this approach, to determine its safety and efficacy compared to maintenance immunosuppression, and to refine and validate accurate biomarkers of tolerance.[37]


We thank Marta Martínez and Anna Rodríguez from the IDIBAPS Liver Transplant Immunology Laboratory, and Andres Cárdenas, Miquel Navasa, Xavier Forns, David Calatayud, and Juan Carlos García-Valdecasas from the Hospital Clinic Liver Unit for scientific and technical assistance.