Cyclosporine versus Tacrolimus Treated Liver Transplant Recipients with Chronic Hepatitis C: Outcomes Analysis of the UNOS/OPTN Database

Authors


William Irish, birish@ctifacts.com

Abstract

Recurrent hepatitis C virus (HCV) remains a problematic cause of morbidity and mortality for liver transplant patients. Immunosuppression including calcineurin-inhibitors has been implicated in the acceleration of recurrent HCV. Recent studies suggest that outcomes may be better with cyclosporine (CSA-ME) compared to tacrolimus (TAC), but the data are inconclusive. We retrospectively analyzed data received from the United Network for Organ Sharing on 8809 chronic HCV liver transplant recipients receiving either cyclosporine microemulsion (CSA-ME) or tacrolimus (TAC) as maintenance immunosuppression prior to discharge. We analyzed patient death, graft failure, failure due recurrent disease and acute cellular rejection (ACR) for CSA-ME versus TAC treated patients. Three-year unadjusted patient and graft survival rates were 76.8% and 71.5%, respectively, in the CSA-ME group versus 79.9% and 75.0% in the TAC group. Propensity score-adjusted results suggest CSA-ME treated patients are at increased risk of patient death and graft failure [Hazards ratio (HR) = 1.17; 95% CI = 1.01–1.36 and HR = 1.19; 95% CI = 1.04–1.35, respectively] and biopsy-confirmed AR (HR = 2.03; 95% CI = 1.54–2.67) compared to TAC treated patients. These results provide evidence to reconsider the targeted administration of CSA-ME to HCV-infected liver transplant recipients.

Abbreviations: 
ACR

acute cellular rejection

ANOVA

analysis of variance

ATG

anti-thymocyte globulin

BPAR

biopsy-proven acute rejection

CI

confidence interval

CIT

cold ischemia time

CMV

cytomegalovirus

CNI

calcineurin inhibitor

CSA-ME

cyclosporine microemulsion

CV

continuous variable

HBV

hepatitis B virus

HCV

hepatitis C virus

HR

hazard ratio

ICU

intensive care unit

INR

international normalization ratio

MELD

model for end-stage liver disease

MI

maintenance immunosuppression

MMF

mycophenolate mofetil

MPA

mycophenolic acid

OLT

orthotopic liver transplant

OPTN

Organ Procurement and Transplantation Network

OR

odds ratio

PRA

panel reactive antibody

PS

propensity score

SD

standard deviation

TAC

tacrolimus

TAR

treated acute rejection

UNOS

United Network for Organ Sharing

Introduction

Hepatitis C virus (HCV) is the leading cause of cirrhosis and hepatic failure leading to orthotopic liver transplantation (OLT) and accounts for approximately 40% of all liver transplants. Following transplantation, recurrence of HCV is nearly universal. Recurrent HCV is associated with significant complications and may lead to graft loss that requires retransplantation. The severity of recurrent HCV may be altered by the amount and type of immunosuppression. There is clear evidence that the treatment of acute cellular rejection (ACR) has a negative effect on recurrent HCV. Specifically, recurrent bolus therapy of corticosteroids and antibody therapy for ACR are perhaps the most important risk factors for severe recurrent HCV (1,2). However, there is no convincing data that any specific immunosuppressive agent is associated with worsening posttransplantation HCV. In particular, the role of calcineurin inhibitors (CNIs), in the acceleration of posttransplant viral replication of HCV and graft damage is unclear. Recent in vitro data suggests that cyclosporine (CSA-ME) may reduce the severity of recurrent HCV. Nakagawa et al. reported that cyclosporine may have an inhibitory effect on HCV replication in vitro (3). However, there is currently little clinical data to support these findings. In fact, a recent meta-analysis found no difference in severity of recurrent HCV for recipients receiving CSA-ME compared to TAC (4). Currently, there is a randomized trial underway which is specifically designed to assess the potential beneficial effect of cyclosporine on recurrent HCV. Because of the potential beneficial effects of cyclosporine on recurrent HCV, we undertook a study of the United Network for Organ Sharing (UNOS)/Organ Procurement and Transplantation Network (OPTN) database to retrospectively evaluate the long-term clinical outcomes and effects of maintenance immunosuppressive therapy, in particular, the CNIs CSA-ME versus TAC in a large contemporary cohort of patients who underwent OLT for HCV (5).

Materials and Methods

Study population

All adult patients in the UNOS/OPTN database who received an orthotopic liver transplant from January 1, 2000 to December 31, 2007 due to chronic hepatitis C (HCV cirrhosis or alcoholic cirrhosis with HCV) were included. Patients were excluded if their HCV serology status at time of transplantation was either negative or ‘unknown’ (i.e. missing, not done, unknown), they were less than 18 years of age at time of transplant, had received a previous liver transplant, were recipients of multiple organs (e.g. combined liver–kidney), had a live or foreign donor, or received a partial or split liver transplant. Patients were also excluded if they had both CSA-ME and TAC recorded as maintenance therapy at the time of discharge or neither of the above listed at the time of discharge.

Variable definitions

Risk factors:  Patient outcomes can be impacted by a variety of recipient and donor factors. The UNOS/OPTN database collects data on several of these factors both pre- and posttransplantation. Risk factors evaluated for inclusion in a risk-adjusted multivariable comparative analysis of CSA-ME versus TAC included:

  • • Recipient characteristics (e.g. age of the recipient at time of transplantation, gender, race, diabetes);
  • • Donor characteristics (e.g. age, gender, height [cm], donation after cardiac death, ABO blood type, cytomegalovirus [CMV] status, hepatitis B virus [HBV] status, HCV status);
  • • Transplant-related variables [e.g. cold ischemia time (CIT), total number of HLA mismatches, peak panel reactive antibody (PRA)];
  • • Clinical variables at the time of transplantation [e.g. previous blood transfusions, serum creatinine (mg/dL), bilirubin (mg/dL), international normalization ratio (INR), HBV serology status, history of malignancy and history of other comorbid conditions if data available, medical condition at time of transplant, pretransplant intensive care unit (ICU) hospitalization or hospital stay, patient on life support, previous abdominal surgery and portal vein thrombosis];
  • • Concomitant immunosuppressive drugs administered prior to discharge [e.g. antibody induction, sirolimus, mycophenolate mofetil (MMF)/mycophenolic acid (MPA), azathioprine, corticosteroids].

The Model for End-Stage Liver Disease (MELD) has been validated for purposes of predicting 3-month patient mortality while candidates are on the liver transplant waiting list. MELD scores were also used for purposes of risk adjustment in the comparative analysis of this study. The MELD score was calculated using the following formula:

MELD Score = 0.957 × Loge (creatinine mg/dL) + 0.378 × Loge (bilirubin mg/dL) + 1.120 × Loge (INR) + 0.643 where Loge is the natural logarithm to the base e.

Laboratory values less than 1.0 were set to 1.0 for the purposes of the MELD score calculation. The maximum serum creatinine considered within the MELD score equation is 4.0 mg/dL. If a patient was on dialysis prior to transplantation, then the serum creatinine value was set to 4.0 mg/dL.

Endpoints:  Endpoints for analysis included:

  • 1All-cause primary graft failure and failure due to recurrent hepatitis: All-cause primary graft failure was defined as patient death due to any cause (if no prior retransplantation) or first retransplantation for any reason at any time during follow-up. Graft failure due to recurrent HCV was defined as patient death or first retransplantation that was reported to be due to recurrent disease in the Transplant Recipient Follow-up Form. If the cause of death/graft failure (primary or contributing) was missing, then we ascertained causes of death/graft failure reported as an open text field.
  • 2Primary graft survival: This was calculated from the date of transplantation until date of primary graft failure due to any cause. Patients alive with functioning graft at date of last known follow-up or after 3000 days posttransplantation were right censored.
  • 3Patient survival: Calculated from the date of transplantation until date death due to any cause. Patients alive at date of last known follow-up or after 3000 days posttransplantation were right censored.
  • 4Death-censored primary graft survival (or Time to Retransplantation): This was calculated from the date of transplantation until date of first retransplantation. Patients alive or who died with a functioning graft at date of last known follow-up or after 3000 days posttransplantation were right censored.
  • 5Treated acute rejection (TAR): Defined as treatment for acute rejection as reported on the UNOS/OPTN Transplant Recipient follow-up form.
  • 6Biopsy-proven acute rejection (BPAR): Defined as TAR reported as biopsy-proven on the UNOS/OPTN Transplant Recipient Follow-up Form.

Statistical methods

Continuous data are presented as the mean ± standard deviation (SD) or median and range, and categorical data as counts and percents. Graft and patient survival was estimated using the product-limit method (6).

Propensity scores:  We used a binary logistic regression model (7) to estimate the propensity of a patient to receive CSA-ME versus TAC therapy at time of discharge conditional on their covariates. From this model, individual PS's were calculated to summarize the covariate (donor, recipient, transplant-related, clinical) information as a measure of the likelihood that a person would have been treated with CSA-ME. The PS was then used in the regression models (see below) to simultaneously control for possible treatment-related covariates that are not of primary interest for assessment of posttransplant outcome differences between treatments. This approach was intended to minimize the inherent treatment selection bias of nonrandomized cohorts while keeping the covariate structure of the final model as simple as possible.

Missing data on recipient, donor, transplant-related and clinical variables were incorporated into the binary logistic regression model as follows:

  • 1For continuous variables (CV), two indicator variables (IV1 and IV2) were constructed. If CV = missing then IV1 = 1 or 0 otherwise; and If CV = missing then IV2 = 0 otherwise IV2 = CV. For example, if patient is missing CIT then IV = 1 and IV2 = 0. If patient is not missing CIT then IV1 = 0 and IV2 = CIT.
  • 2For categorical data, missing or unknown were grouped into the ‘no’ or ‘absence’ category. For example, recipients whose diabetes status is unknown were grouped into the ‘no’ category, so that estimates of effect will be ‘yes’ versus ‘no/unknown’.

We adjusted treatment comparisons using the PS stratification method (8–10). Based on the PS model, patients were grouped into five strata determined by each individual's propensity score using the following algorithm:

Step 1: Sort all patients on their propensity for receiving CSA-ME.

Step 2: Partition the patients into five approximately equally sized groups (propensity quintiles) according to their propensity for receiving CSA-ME.

In theory, treatment exposure is considered at random for individuals with the same propensity value, thus any treatment comparison within this group is unbiased. Identifying individuals sharing exactly the same propensity value, however, is infeasible, and so stratification was used to achieve groups where this at least held approximately. As a consequence, treatment effects may be biased if residual confounding remains within strata. To evaluate whether residual confounding existed within strata (11), the degree of balance was examined using one-way analysis of variance (ANOVA). If residual confounding was evident, covariates were included in the hazards regression models.

Hazards regression:  Cox proportional hazards regression model (12) was used to calculate the hazards ratios (HRs) and associated 95% confidence intervals (CIs) on all-cause graft failure, death-censored graft failure (retransplantation), graft failure due to recurrent hepatitis and patient death for CSA-ME versus TAC therapy at time of discharge. Cox model was fit to the data with treatment (CSA-ME vs. TAC) and propensity score strata included as covariates in the model. Additional covariates were added to the model if residual confounding was evident.

Due to the observational nature of the data and UNOS/OPTN reporting procedures, complimentary log–log model (12,13) was used to quantify the relationship between CSA-ME versus TAC and risk of acute rejection (TAR and BPAR). The model is defined as follows

image

where Pij= the probability of occurrence of an event for individual i in jth interval (j = 1,2,3,4); β= log hazards for CSA-ME versus TAC adjusted for propensity score strata (Si); and αj= time-specific baseline log hazards. Additional covariates were added to the model if residual confounding was evident. For purpose of analysis event reporting was restricted to the first four intervals (1, 2, 3 and 4 years) following transplantation. HR and 95% CI were used as measures of strength of association and precision, respectively. The complimentary log–log model was also used to estimate the 3-year unadjusted cumulative incidence of TAR and BPAR.

Sensitivity analyses were performed by including only patients who were transplanted between 2002 and 2007.

All analyses were performed by Dr. William Irish using SAS statistical software (SAS v. 9.1, Cary, NC, USA). A p-value less than 0.05 were considered statistically significant.

Results

Study characteristics

The initial study population analyzed included 46 958 transplants in 43 971 patients inclusive for all types of liver disease. When applying the inclusion/exclusion criteria, we arrived at 9709 (21%) liver transplants in 9709 patients with the diagnosis of HCV cirrhosis or alcoholic cirrhosis with HCV, and who were HCV positive by serology at time of transplantation (Table 1). Of these, 8809 patients had either CSA-ME (n = 717; 8.1%) or TAC (n = 8092; 91.9%) listed as their maintenance CNI immunosuppressive agent at the time of discharge.

Table 1.  Study population disposition
 Number of transplantsNumber of patients
  1. *HCV cirrhosis or alcoholic cirrhosis with HCV.

Year: 2000–200746 95843 971
Single organ44 03541 312
Deceased donor41 22838 810
Age ≥18 years37 75735 655
Whole liver37 15735 122
First transplant34 12234 122
Orthotopic34 07134 071
Cirrhosis–HCV*11 82311 823
Positive HCV serology status at time of transplantation  9709  9709
CsA-ME or Tacrolimus maintenance IM  8809  8809

Recipient and donor characteristics are summarized in Table 2. The mean recipient age was 51.8 ± 7.1 years, 75% were male and 8.6% were African American. Less than 5% of the recipients received organs from donors after cardiac death with the majority of donors being ABO identical with the recipient.

Table 2.  Recipient and donor characteristics
 Statistics or categoryPrior to dischargeOverall (N = 8809)
CSA-ME (N = 717)Tacrolimus (N = 8092)
Recipient variable
 DiagnosisAlcoholic Cirrhosis with HCV125 (17.43%)1560 (19.28%)1685 (19.13%)
HCV Cirrhosis592 (82.57%)6532 (80.72%)7124 (80.87%)
 Age at transplant (years)Mean (SD)51.19 (7.13)51.85 (7.11)51.80 (7.11)
 SexMale535 (74.62%)6067 (74.98%)6602 (74.95%)
 Race/EthnicityCaucasian571 (79.64%)5923 (73.20%)6494 (73.72%)
African American58 (8.09%)703 (8.69%)761 (8.64%)
Hispanic75 (10.46%)1184 (14.63%)1259 (14.29%)
Other13 (1.82%)282 (34.85%)295 (33.49%)
 Diabetes pretransplantYes15.24%17.23%17.07%
Status Unknown9 (1.26%)138 (1.71%)147 (1.67%)
Missingn = 2n = 4n = 6
Donor variable
 Age (years)Mean (SD)37.58 (15.40)40.49 (16.58)40.25 (16.50)
 SexFemale295 (41.14%)3054 (37.74%)3349 (38.02%)
 DonorBrain death710 (99.02%)7787 (96.23%)8497 (96.46%)
 Recipient–Donor ABO matchIdentical681 (94.98%)7656 (94.61%)8337 (94.64%)
Compatible34 (4.74%)401 (4.96%)435 (4.94%)
Incompatible2 (0.28%)35 (0.43%)37 (0.42%)

Concomitant medications

Antibody induction was in used in 20.1% of all patients (31.4% in the CSA-ME group vs. 19.1% in the TAC group) with the majority of CSA-ME patients receiving basiliximab and the majority of TAC patients receiving rabbit anti-thymocyte globulin (ATG) (Figure 1). Antimetabolite therapy was administered in 70.1% of patients in the CSA-ME group and 60.0% in the TAC group (Figure 1), most of which was MMF. Corticosteroids for induction were used in 59.4% of patients who received CSA-ME and 61.9% in patients who received TAC while corticosteroids in maintenance were used in over 80% of patients in both treatment groups. Alternative therapy with sirolimus was infrequently given in both treatment groups (CSA-ME = 6.7% and TAC = 3.8%).

Figure 1.

Concomitant immunosuppression at time of discharge.

Factors associated with CSA-ME versus TAC use prior to discharge

Results of the PS analysis of CSA-ME versus TAC prior to discharge are presented in Table 3. Propensity score analysis showed that patients who received CSA-ME were significantly (p<0.05) more likely to have received: a younger [Odds ratio (OR) per 1 year increase = 0.988; 95% CI = 0.984–0.993], female (OR = 1.29; 95% CI = 1.09–1.52) and brain-death (OR = 3.30; 95% CI = 1.54–7.09) donor than TAC patients. CSA-ME treated patients were also more likely to be Caucasian (OR vs. non-Caucasian = 1.45; 95% CI = 1.19–1.76), have prolonged CIT (OR per 1 h increase = 1.04; 95% CI = 1.02–1.06), higher MELD score (OR per unit increase = 1.02; 95% CI = 1.01–1.03), not to be hospitalized (OR vs. hospitalized in ICU = 1.46; 95% CI = 1.06–2.02) and have had previous abdominal surgery (OR = 1.26; 95% CI = 1.07–1.49).

Table 3.  Factors associated with CSA-ME versus tac at time of discharge: propensity score model
 95% CI for OR
Odds ratio (OR)LowerUpperp-Value
Recipient
 HCV only versus alcoholic with HCV1.3120.6170.9420.0118
 Age (per 1 year increase)0.9980.9861.010.7265
 Gender (male vs. female)1.120.9241.3570.2475
 Race
   Black versus White0.8750.6541.1720.3705
   Hispanic versus White0.6380.4930.8270.0007
   Other versus White0.4600.2570.820.0085
 White versus non-White1.4491.1901.7640.0002
 Diabetes at time of registration
   Yes versus no0.9510.7631.1850.654
   Unknown versus no0.7820.4121.4850.4521
Donor
 Age (per 1 year increase)0.9880.9840.993<0.0001
 Gender (female vs. male)1.2891.0921.5200.0028
 Brain death versus DCD3.3001.5437.0920.0021
 ABO match—Not identical versus identical1.020.7091.4680.9131
 CMV status—Positive versus other0.940.7961.110.4672
 HCV antibody status—Positive versus other1.3520.9651.8940.0797
 HBV core antibody—Positive versus other1.0090.711.4330.9622
Clinical
 On dialysis—Yes versus no0.5570.3430.9040.0178
 HBV core antibody—Positive versus other1.0220.8581.2170.8108
 HBV surface antigen—Positive versus other0.760.3631.5940.4683
 Previous malignancy—Yes versus no0.8840.6021.2970.5287
 Medical condition
   Hospitalized not in ICU versus in ICU1.2120.8431.7420.2998
   Not hospitalized versus in ICU1.461.0562.0190.022
 Previous upper abdominal surgery—Yes versus no1.2641.0741.4880.0048
 On mechanical, ventilated or organ-perfusion support—Yes versus no1.4990.6693.3580.3257
 On ventilator—Yes versus no0.6930.2811.7060.4248
 Portal vein thrombosis—Yes versus other0.7610.4891.1830.2254
 Ascites—Yes versus other1.1310.821.5610.4515
 Encephalopathy—Gr 1–4 versus None/NR0.8590.6591.120.262
 Cold ischemia time—Missing yes versus no1.190.8851.6020.2498
 Cold ischemia time (per 1 h increase)1.0381.0171.0580.0003
 MELD score—Missing yes versus no1.7561.2532.4610.0011
 MELD score—Per unit increase1.0221.0091.0350.0008
 Days on liver waiting list (per 1 day increase)1110.2601
Concomitant immunosuppression
 Any antibody induction—Yes versus no2.0321.6912.442<0.0001
 Any antimetabolite—Yes versus no1.9691.6442.357<0.0001
 Rapamycin—Yes versus no1.9831.4142.781<0.0001
 Steroids induction—Yes versus no0.8210.6970.9670.018
 Steroids maintenance—Yes versus no1.1090.8791.40.3833
Transplant year
 2001 versus 20000.5410.4150.706<0.0001
 2002 versus 20000.3050.2080.448<0.0001
 2003 versus 20000.2990.1990.447<0.0001
 2004 versus 20000.3570.240.53<0.0001
 2005 versus 20000.3040.2020.456<0.0001
 2006 versus 20000.1890.1220.293<0.0001
 2007 versus 20000.2730.1820.409<0.0001

Graft and patient survival

The median time of follow-up was 5.0 years for the CSA-ME group and 4.0 years for the TAC group, with an overall median follow-up of 4.0 years. Of the 8809 patients, 25.0% died (29.7% of CSA-ME treated patients and 24.5% of TAC treated patients) and 6.4% were retransplanted (7.5% in the CSA-ME treated group and 6.3% in the TAC treated group). The 1-year graft and patient survival rate in the CSA-ME treated group was 84.6 ± 1.3% and 88.3 ± 1.2%, respectively and 86.5 ± 0.4% and 89.9 ± 0.3% in the TAC treated group. A divergence between treatment groups could be seen in graft and patient survival by 3 years posttransplantation. The 3-year graft and patient survival rate in the CSA-ME treated group was 71.5 ± 1.7% and 76.8 ± 1.7%, respectively, and 75.0 ± 0.5% and 79.9 ± 0.5% in the TAC treated group (Figures 2a-b). The cumulative (unadjusted) hazards of graft failure due to recurrent HCV disease demonstrated that patients receiving CSA-ME were more likely than TAC patients to lose their allograft due to recurrent disease by approximately 2 years posttransplant (Figure 3).

Figure 2.

(A) Kaplan-Meier (unadjusted) patient survival; (B) Kaplan-Meier (unadjusted) graft survival.

Figure 3.

Cummulative hazards of graft failure due to recurrent disease.

Propensity score adjusted results of CSA-ME versus TAC are presented in Table 4. Adjusting for differences in recipient-, donor- and transplant- and clinical-related characteristics and concomitant immunosuppression at time of discharge, CSA-ME treated patients were at increased risk of primary graft failure, death with primary graft and failure due to recurrent disease. There was no evidence of a significant effect of CSA-ME versus TAC on rate of retransplantation.

Table 4.  Propensity score-adjusted results
Response variableHazard ratio195% Confidence intervalp-Value
  1. 1Adjusted for propensity score strata.

Primary graft failure1.1861.042–1.3500.0100
Retransplantation1.2370.927–1.6510.1482
Death with primary graft1.1731.015–1.3570.0310
Primary graft failure due recurrent disease1.4111.098–1.8150.0072

We evaluated whether the effect CSA-ME versus TAC on outcome was differentially affected by year of transplantation. This was accomplished by including a two-way interaction term for CNI (CSA-ME vs. TAC) by year of transplantation in the propensity score adjusted Cox proportional hazards model. None of the two-way interactions evaluated were statistically significant (p-value range = 0.2051–0.9205).

Results of the sensitivity analysis were similar to the results presented for the entire cohort. Propensity score adjusted HR (CSA-ME vs. TAC) for primary graft failure was 1.26 (95% CI = 1.06–1.50); for retransplantation, HR = 1.10 (95% CI = 0.73–1.67); for death with primary graft, HR = 1.30 (95% CI = 1.07–1.58) and for failure due to recurrent disease, HR = 1.66 (95% CI = 1.19–2.31).

Acute rejection

The cumulative (unadjusted) incidence of TAR and BPAR at 3 years posttransplant was 26.8% and 19.9%, respectively, in the CSA-ME treated group versus 12.4% and 9.0% in the TAC treated group. The propensity score adjusted HR (CSA-ME vs. TAC) for TAR was 2.17 (95% CI = 1.71–2.75; p < 0.0001) and for BPAR, HR = 2.03 (95% CI = 1.54–2.67; p < 0.0001). There was no evidence of a differential effect of CSA-ME versus TAC on acute rejection by year of transplantation (two-way interaction p-value for TAR = 0.9223 and for BPAR, p-value = 0.5922).

Results of the sensitivity analysis were similar to the results reported for the entire cohort: For TAR the PS adjusted HR = 2.15 (95% CI = 1.68–2.75) and for BPAR, HR = 2.02 (95% CI = 1.52–2.70).

Discussion

The results suggest that liver transplant recipients receiving CSA-ME have an increased risk of death and graft loss compared to those receiving TAC. These results were quite surprising given past results, which have largely reported no difference in outcomes as well as the current consideration of beneficial effects of CSA-ME on recurrent HCV. The explanation for worse outcomes is not known, but may be related to the higher rate of ACR and steroid-resistant ACR in the CSA-ME group. Higher rejection rates could require multiple treatments of corticosteroid boluses both of which are associated with more severe posttransplant HCV recurrence as described above. While these results are surprising, a more careful examination of the available data suggests that they are not novel. The studies evaluating the effects of immunosuppression focus primarily on 1-year outcome as the longest follow-up endpoint (4). In these studies, there has been no evidence of differences in patient or graft survival in HCV-infected recipients treated with CSA-ME or TAC. However, the natural history of posttransplant HCV does not become completely manifest until at least 1 year after transplant (14). That is, significant differences in graft loss are not apparent until at least 1 year. Consequently, such a short interval follow-up may be insufficient to determine the true effect of a particular immunosuppressive agent on the natural history of posttransplant HCV. One study that reported worse outcomes associated with CSA-ME had follow-up for more than 1 year. Wiesner et al. reported significantly worse patient and graft survival in patients transplanted with CSA-ME compared to TAC at 5 years in the tacrolimus registration trial (15). Hepatitis C patients receiving TAC had a significantly lower 5-year mortality 21.1% compared to CSA 39.5%, p = 0.041.

Since there were significantly fewer patients on CSA-ME at time of discharge versus TAC, these patients may be representative of only a few centers in which CSA-ME is used as maintenance immunosuppression in liver transplantation. As such, the treatment effect observed in this study may be due in part to center differences and not directly related to treatment. Unfortunately, center identifiers are not available to researchers in the UNOS/OPTN database and therefore, center effects cannot be accounted for in the analysis. This is a potential limitation of the data and will need to be explored in future studies.

Changing treatment practices over time can influence results if not appropriately controlled for in the design or analysis of retrospective studies. This is particularly important in transplantation as immunosuppression protocols are continually being modified (e.g. CNI minimization or avoidance protocols) as a means to optimize long-term outcomes and reduce the occurrence of adverse events (e.g. renal injury). This type of information is not explicitly captured in the UNOS/OPTN database, and therefore is not accounted for in the propensity score-adjusted analysis. One approach to minimizing the impact of changing treatment practices in a nonrandomized observational study is to limit cohort selection to patients who received a liver transplant over a relatively short time period. In our analysis, we included patients who were transplanted between 2000 and 2007, although a disproportionate number of patients who received CSA-ME were transplanted between 2000 and 2001. One could argue that patients who received CSA-ME in 2000 and 2001 represent a unique study population whereby CSA-ME exposure was administered under different immunosuppression protocols compared to the later transplant period. Results of the sensitivity analysis, however, were similar to those reported for the entire cohort.

The limitations of this analysis include those common to any large database analysis: the limitations of data reporting from the participating sites, quality of the data and changing treatment practices. Survey reports to transplant registries are usually more brief and concise to reduce the burden of reporting (16). However, there is no reason to suspect a differential reporting bias between treatment groups, given the large number of patients included in the analysis. Two studies in transplantation were performed to evaluate the integrity of data submitted to UNOS/OPTN on immunosuppressive drug exposure. The authors found that immunosuppressive medications reported to UNOS/OPTN are consistent with those electronically captured by Medicare billing claims, although agreement with respect to steroid use was limited (17,18). There are limitations to this study that could not be accommodated in the design or analysis. For example, the study included patients who received CSA-ME or TAC at the time of hospital discharge and does not take into account dose changes, treatment switches or discontinuation from therapy. Moreover, the UNOS/OPTN database does not capture data on CNI dose and levels, either at time of discharge or at posttransplant survey time points. As noted above, data collected on immunosuppressive therapies are episodic and do not include dose information. This critical piece of information would be important, especially when evaluating the long-term impact of treatment. Relating drug usage at time of discharge with long-term outcome may be inadequate to capture the true effect of the drug on long-term outcomes.

Another important limitation is inherent in the use of the PS method, which is used to minimize potential treatment selection bias in nonrandomized comparison of treatments. The PS method can only control for factors that are observed and available to researchers. As such, important factors not captured and not available to researchers (e.g. HCV genotype) are not controlled for in the comparative analysis.

In summary, the administration of CSA-ME in HCV patients, compared to TAC, is associated with a significant increase in patient and graft loss. The explanation for this difference is not clear, but may be related to the higher rate of rejection with CSA-ME. These results need to be interpreted with caution due to the retrospective nature of this data analysis and the limitations discussed, but practitioners should consider the avoidance of CSA-ME in their HCV patients.

Acknowledgments

This study was funded by Novartis Pharmaceuticals Corporation, East Hanover, NJ.

Disclosure

The authors W. Irish, D. Bowers and S. Arcona of this manuscript have conflicts of interest to disclose as described by the American Journal of Transplantation. W. Irish and D. Bowers are employees of CTI Clinical Trial and Consulting Services (Cincinnati, OH) and received research funding from Novartis Pharmaceuticals Corporation. S. Arcona is currently an employee of Novartis Pharmaceuticals Corporation (East Hanover, NJ).

The author J. Trotter of this manuscript has conflicts of interest to disclose as described by the American Journal of Transplantation. He has received grants from Eisai Incorporated (Woodcliff Lake, NJ) and Vital Therapies Incorporated (San Diego, CA) and consultancy fees from Novartis Pharmaceuticals Corporation (East Hanover, NJ).

Ancillary