Impact of sirolimus and tacrolimus on mortality and graft loss in liver transplant recipients with or without hepatitis C virus: An analysis of the Scientific Registry of Transplant Recipients Database †
Kymberly D. Watt,
Division of Gastroenterology and Hepatology, Rochester, MN
A recent analysis of the Scientific Registry of Transplant Recipients (SRTR) database has demonstrated that liver disease secondary to a hepatitis C virus (HCV) infection will continue to be the most common indication for liver transplantation (LT) in the United States for at least the decade.1 The recurrence of HCV infections is universal after LT and is associated with worse patient and graft survival after LT.2 The relative impact of immunosuppressive agents on posttransplant outcomes is incompletely understood. Although they were not explicitly designed for the purpose, prospective, multicenter randomized controlled trials powered to determine the comparative impact on the outcomes of recipients with HCV infections are available only for tacrolimus versus cyclosporine3 and for mycophenolate mofetil versus azathioprine.4 Although there have been randomized controlled trials and retrospective analyses of other immunosuppressive agents commonly used after LT, they have been too small in number and/or too short in follow-up to meaningfully determine the effects of immunosuppression choices specifically on HCV-infected patients. Reports of the comparative impact of immunosuppressive agents on key posttransplant outcomes of patients with HCV have been highly variable. Absent large randomized controlled studies of comparative efficacy and in the setting of conflicting results from retrospective analyses, the choice of posttransplant immunosuppression is largely empirical and varies substantially between transplant centers.
Using a large, prospectively obtained national database (the SRTR database), we sought to determine comparative risk factors (including primary immunosuppression) predictive of death and graft loss in patients with or without HCV. Although the SRTR database does not provide detailed information about the histological features of HCV recurrence, its data about the most important posttransplant outcomes—death and graft loss—are comprehensive and may provide important insights into the impact of immunosuppression on posttransplant outcomes.
CI, confidence interval; HCV, hepatitis C virus; HR, hazard ratio; LT, liver transplantation; MELD, Model for End-Stage Liver Disease; mTOR, mammalian target of rapamycin; SRTR, Scientific Registry of Transplant Recipients.
PATIENTS AND METHODS
The SRTR database was analyzed and included all adult recipients (≥18 years old) who underwent primary, solitary LT between 2000 and 2006. Transplant patients with an underlying HCV-related liver disease (with or without hepatocellular carcinoma) were compared to patients undergoing transplantation for non–virus-related liver diseases. We excluded the small number of hepatitis B patients in the SRTR database from the comparison group to remove any potential confounding effect of immunosuppression on hepatitis B virus and the substantial changes in hepatitis B virus antiviral therapies during the course of the analysis; this left a true nonviral comparator cohort. The human immunodeficiency virus status was not known to us for the patients in the HCV cohort or the nonviral cohort. Patients undergoing transplantation for fulminant hepatic failure or patients undergoing multiorgan transplantation or retransplantation were also excluded. Immunosuppression was examined at the time of discharge from the hospital and as a time-dependent variable (data were available 6 months after transplantation and annually thereafter) in the survival analysis. Immunosuppression medications that were investigated at the baseline included tacrolimus, cyclosporine, and sirolimus-based regimens. Patients receiving sirolimus may or may not have been receiving low-dose tacrolimus or cyclosporine. All results were stratified by the transplant center in the analysis.
The primary endpoint of the analysis was patient survival. The secondary endpoint of graft loss or death was also analyzed. Death reports are mandated and are reported to the SRTR by transplant centers. These data are overseen by the Organ Procurement and Transplantation Network, and deaths are corroborated by the US Social Security Administration. The database, however, lacks sufficient granularity to determine HCV recurrence in allografts or histological progression over time. In addition, only death dates are available, and causes of death cannot be reliably determined and thus may be unrelated to HCV infections.
Patient and graft survival estimates were computed with Kaplan-Meier methods. Cox models were fitted; they were stratified by the transplant center and included adjustments for pretransplant variables known to affect patient and graft survival (eg, creatinine and donor and recipient ages). A P value < 0.05 was considered statistically significant. Many patients managed with a sirolimus-based regimen would have had either renal insufficiency or a history of malignancy before transplantation, which conceivably could have affected patient survival. To adjust for this, we performed a propensity score analysis by modeling the probability of having a sirolimus-based regimen at discharge with a number of factors via a logistic regression. The probability of having sirolimus was estimated for each patient and was entered into subsequent Cox models for survival outcomes. The following variables were included in the computation of the propensity scores: hepatocellular carcinoma, previous malignancy (yes, no, or missing), hemodialysis (yes, no, or missing), logarithm of the creatinine level, logarithm of the bilirubin level, logarithm of the international normalized ratio, donor and recipient sex, time from listing to transplantation, recipient age, diabetes, sex mismatch, donor age, donor body mass index, recipient blood type (A, B, AB, or O), donor body mass index, and partial/split liver.
In all, 26,414 transplant recipients were included in the analysis, and 12,589 had an underlying HCV infection. Hepatocellular carcinoma was documented in 3850 patients, and sirolimus was prescribed as part of the baseline immunosuppression in 1685 patients. Sirolimus was given at the time of discharge from the hospital after LT to 6.9% (259/3777) of the patients with hepatocellular carcinoma, to 6.5% (795/12,269) of the patients with HCV, and to 10.3% (75/730) of the patients who were on hemodialysis before transplantation. Patients on sirolimus were statistically different (P < 0.001) from the cohort of patients without sirolimus with respect to the creatinine level before transplantation (1.5 ± 1.1 versus 1.3 ± 0.9 mg/dL), the Model for End-Stage Liver Disease (MELD) score before transplantation (20.4 ± 9.1 versus 18.7 ± 8.2), and the donor age (38.6 ± 17.0 versus 40.9 ± 17.3 years). The demographics of the 2 investigated cohorts (the HCV and nonviral groups) are outlined in Table 1.
Table 1. Demographics and Immunosuppression for the Study Population at the Baseline
Non-HCV Patients (n = 13,825)
HCV Patients (n = 12,589)
NOTE: Immunosuppression percentages were calculated for the patients with data available.
The data are presented as means and standard deviations.
After transplantation, 6.5% of the patients in the HCV group were discharged on a sirolimus-based immunosuppression regimen. Sirolimus information was available for 10,033 HCV patients at the baseline and 1 year after transplantation: 3.5% of these patients remained on sirolimus 1 year after transplantation, 2.7% were discharged on sirolimus but were not on sirolimus at 1 year, and 7.7% were started on sirolimus at or after 1 year of follow-up. In the nonviral group, 6.6% of the patients were discharged on a sirolimus-based regimen. Sirolimus information was available for 11,182 nonviral patients at the baseline and 1 year after transplantation: 4.1% remained on sirolimus at 1 year, 2.4% stopped taking sirolimus by 1 year, and 9.0% started to take sirolimus at or after 1 year of follow-up. Sirolimus was never used by 86.1% of the HCV patients and 84% of the nonviral patients.
Within the first 3 years after transplantation, 5398 patient deaths and 6463 graft losses/deaths occurred. Patient survival was significantly better in the nonviral group versus the HCV group (P < 0.001). The Kaplan-Meier estimates for 1- and 3-year patient survival were 86% and 76%, respectively, for the HCV group and 88% and 81%, respectively, for the nonviral group. The survival rates for the cohort with hepatocellular carcinoma were 86% and 72%, respectively. Survival according to sirolimus exposure is depicted in Fig. 1.
Univariate risk factors for death within 3 years for all patients are shown in Table 2. The model estimates for the outcome of death or graft loss were similar. Univariate risk factors specific for patients with HCV included cyclosporine [hazard ratio (HR) = 1.10, confidence interval (95% CI) = 0.96-1.27, P = 0.16], tacrolimus (HR = 0.70, 95% CI = 0.62-0.79, P < 0.0001), and sirolimus (HR = 1.50, 95% CI = 1.29-1.74, P < 0.0001); factors specific for patients undergoing transplantation for a nonviral etiology included cyclosporine (HR = 1.31, CI = 1.12-1.52, P = 0.0005), tacrolimus (HR = 0.51, CI = 0.45-0.58, P < 0.0001), and sirolimus (HR = 1.46, 95% CI = 1.24-1.72, P < 0.0001).
Table 2. Univariate Models for Patient Mortality Within 3 Years for All Patients
HR (95% CI)
Three categories were considered for the dialysis variable: yes, missing, and no (reference); data for yes versus no are shown. Estimates for the missing category are not shown.
Because of the multiple factors involved in choosing immunosuppression regimens, a multivariate analysis was performed. A multivariate analysis of 3-year mortality is depicted in Table 3 for the HCV and nonviral groups. Similar results were obtained with graft loss or death as the endpoint, except that sirolimus was less predictive in the HCV group (HR = 1.14, 95% CI = 0.98-1.33, P = 0.09) and the nonviral group (HR = 1.08, 95% CI = 0.92-1.27, P = 0.34).
Table 3. Multivariate Models for Patient Mortality Within 3 years
HR (95% CI)
HR (95% CI)
NOTE: Because of missing values, 752 HCV patients and 858 nonviral patients were not included in these models.
Recipient age (decades)
Donor age (decades)
Creatinine at LT (log)
Tacrolimus at the baseline
Sirolimus at the baseline
To adjust for the probability of a patient being started on sirolimus, we performed a propensity analysis (Table 4). This model was adjusted for any factor that might be associated with increased mortality from our analysis (including hepatocellular carcinoma, previous malignancy, hemodialysis, creatinine, bilirubin, international normalized ratio, sex, time from listing to transplantation, recipient age, diabetes, sex mismatch, donor age, donor body mass index, blood type, body mass index, and partial/split liver) as well as factors known to be related to patient outcomes. According to this analysis, HCV patients who received sirolimus immunosuppression at discharge had an increased risk of 3-year mortality despite adjustments for all other potential influences on mortality (HR = 1.29, 95% CI = 1.08-1.55, P = 0.0053), whereas nonviral patients did not have an increased risk related specifically to sirolimus use (HR = 1.09, 95% CI = 0.89-1.33, P = 0.40).
Table 4. Analysis of Independent Risks Associated With 3-Year Mortality Accounting for the Probability of Being Treated With Sirolimus at Discharge (Propensity Score)
HR (95% CI)
HR (95% CI)
NOTE: Because of missing values, 3059 HCV patients and 3476 non-HCV patients were not included in these models.
Recipient age (decades)
Donor age (decades)
Creatinine at LT (log)
Tacrolimus at the baseline
Sirolimus at the baseline
To better understand the effects of changing immunosuppression over time, we fit stratified Cox models for 5-year mortality; we considered a time-dependent effect of active sirolimus use (yes or no) at the baseline and 6 months, 1 year, and 2 years after transplantation and adjusted for the same factors previously outlined in the propensity analysis. The 5-year mortality HRs for the time-dependent sirolimus variable (at the baseline and 6 months, 1 year, and 2 years after transplantation) were 1.43 (95% CI = 1.26-1.63, P < 0.0001) for the HCV patients and 1.42 (95% CI = 1.24-1.62, P < 0.0001) for the nonviral patients.
The following results were obtained from the models comparing sirolimus changes from the baseline to 1 year after transplantation with 5-year mortality.
With respect to mortality between 1 and 5 years, the HR comparing HCV patients on sirolimus at both times and HCV patients not on sirolimus at either time was 1.35 (95% CI = 1.01-1.80, P = 0.042). The HR comparing those who were not on sirolimus at the baseline but were on sirolimus at 1 year and those who were not on sirolimus at either time was 1.50 (95% CI = 1.25-1.81, P < 0.0001). The HR comparing those who discontinued sirolimus by 1 year and those not on sirolimus at either time was 1.26 (95% CI = 0.93-1.72, P = 0.14).
With respect to mortality between 1 and 5 years, the HR comparing nonviral patients on sirolimus at both times and nonviral patients not on sirolimus at either time was 1.51 (95% CI = 1.11-2.06, P = 0.008). The HR comparing those who were not on sirolimus at the baseline but were on sirolimus at 1 year and those who were not on sirolimus at either time was 1.75 (95% CI = 1.43-2.13, P < 0.0001). The HR comparing those who discontinued sirolimus by 1 year and those not on sirolimus at either time was 1.75 (95% CI = 1.22-2.50, P = 0.002).
Lastly, nonviral patients who discontinued sirolimus at 1 year were also statistically similar to those not on sirolimus at either time (HR = 1.27, 95% CI = 0.77-2.10, P = 0.34).
That immunosuppression affects the natural history of HCV infections is undoubted, and an overall proviral effect results in progressive histological graft injury for many patients.5–9 A greater understanding of the relative impact of available immunosuppressive agents on key posttransplant outcomes is one of the most pressing needs of the LT community. Our current analysis of the SRTR database has led to several potentially important findings. The most notable observation is the highly significant association of sirolimus use with an increased risk of mortality and graft loss in LT recipients with HCV infections. The basis for and the implications of this observation merit detailed consideration.
The number of LT recipients in the SRTR database who received sirolimus was 1685, and the proportions were nearly identical for HCV recipients (6.5%) and non-HCV recipients (6.6%). More than two-thirds of the patients receiving sirolimus were taking a calcineurin inhibitor concomitantly. The SRTR does not record the rationale for immunosuppression choices, and this means that we must speculate about why patients were selected to receive sirolimus. The most obvious potential explanation for the observed increased risk of mortality and graft loss in LT recipients with HCV infections who received sirolimus is that patients who were chosen to receive sirolimus were in some way more likely to die or experience graft loss before sirolimus use. To explore this possibility, we conducted a propensity analysis. Propensity scores provide a statistical method for estimating treatment effects when the treatment assignment is not random. In the propensity analysis, after corrections for parameters known to affect posttransplant survival (including malignancy and renal function), sirolimus use remained a significant predictor of long-term mortality in the HCV cohort but not in the nonviral group. The disease-specific nature of the predictivity of sirolimus use for mortality and graft loss suggests that the interaction between sirolimus itself and an HCV infection may account for the difference in outcomes. Our observation of a duration-of-exposure effect adds to concerns about an effect of sirolimus on posttransplant outcomes: patients who remained on sirolimus longer than 1 year experienced worse survival than patients who remained on sirolimus for less than 1 year. This effect was also seen despite a lower donor age in the sirolimus group. The time-dependent analysis demonstrated a worse prognosis for both HCV patients and nonviral patients who started sirolimus later in their posttransplant course and suggested that the reasons for changing to sirolimus led to a poorer prognosis.
Ideally, we would be able to compare the results of this SRTR database analysis to prospective studies of the outcomes of patients with HCV receiving sirolimus versus patients receiving standard immunosuppression. Unfortunately, 1 of the 3 large randomized controlled studies of the use of mammalian target of rapamycin (mTOR) inhibitors after LT has been published only in abstract form,11 and another has not been published at all. A large randomized study of everolimus with cyclosporine after transplantation demonstrated similar overall survival for patients who received everolimus with cyclosporine and patients treated with cyclosporine alone, but it did not look specifically at HCV outcomes.10 A relatively large (n = 163), multicenter randomized controlled trial of sirolimus with cyclosporine versus tacrolimus as de novo immunosuppression, which was presented orally at the American Transplant Congress in 2003 and has been published only in abstract form, reported a 1-year frequency of 20% for the composite endpoint of death and graft loss in patients with HCV who were randomized to sirolimus and a 1-year frequency of 0% in patients randomized to tacrolimus.11 This study was stopped before its completion because of excess deaths and adverse events in the sirolimus arm. There are 2 recently published reports of retrospective analyses of associations of sirolimus with posttransplant outcomes in recipients with HCV infections.12, 13 In an analysis of a combined total of 67 patients, Wagner et al.12 reported that patient survival was better in the sirolimus group (n = 39). An unusual aspect of the study by Wagner et al. was the low 3-year survival rate (56%) in the nonsirolimus group. In a retrospective analysis of patients receiving de novo immunosuppression with sirolimus that was underpowered to detect less than a very large treatment effect (total number of subjects = 88), Asthana et al.13 reported no association between sirolimus use and the time to recurrence, fibrosis scores at the time of recurrence, or overall survival.13 Two more retrospective studies reported that higher serum levels of mTOR inhibitors were associated with earlier histological recurrence13 and clinical flares of hepatitis.14
On balance, according to the results of our current analysis of the SRTR database and the preliminary results of the multicenter randomized controlled trial, there should be a high level of suspicion that sirolimus may confer an increased risk of mortality and graft loss to LT recipients with HCV infections.
It is important to consider the potential biological basis for negative (or positive) associations of a specific therapeutic agent with outcomes. Sirolimus has been described as having antifibrotic properties. Animal studies of bile duct–ligated rats, for example, have found reduced hepatic stellate cell activation and subsequent fibrosis regression with a short-term, low-dose sirolimus treatment15 and with a high-dose sirolimus treatment.16 Neef et al.15 also looked at a thioacetamide-induced cirrhotic rat model (nonbiliary cirrhosis), and they found similar antifibrotic effects of a low-dose, short-term sirolimus treatment, although the effect was lost with longer term treatment. In addition to possible antifibrotic properties, the overall impact of mTOR inhibition is likely, however, to be proviral because mTOR is known to affect cell (and viral) protein synthesis and proliferation and cell survival.17–19 An important consequence of mTOR inhibition is the loss of mTOR-dependent interferon activity. Sirolimus is a potent inhibitor of interferon-α–induced antiviral activity at multiple levels.19–21 In addition, the inhibition of mTOR reduces the antiviral effects of ribavirin.22
We also noted a strong association between tacrolimus and superior patient and graft survival for both HCV and non-HCV recipients. This finding has been observed from the results of a multicenter randomized controlled trial3 and in recent analyses of the SRTR database.23, 24 The basis for improved outcomes with tacrolimus as the baseline and maintenance immunosuppression is probably related to the greater efficacy of tacrolimus versus cyclosporine in preventing acute cellular rejection.25 In addition to being proviral, the treatment of acute cellular rejection with corticosteroids has been consistently associated with increased mortality and graft loss in LT recipients with HCV infections.18, 19
There are several limitations to our study. Although the SRTR database affords the ability to perform multivariate analyses and determine propensity scores for a large number of patients, it limits the endpoints that can be measured. Important limitations of SRTR analyses include the frequency of missing data (eg, immunosuppression drugs, acute cellular rejection, and cytomegalovirus infections) over time and the lack of information on HCV disease severity. It is possible that sirolimus use may be the result of disease severity in ways not apparent from SRTR data. In the absence of histological data, no amount of adjusting can exclude this alternative interpretation. Thus, although the mortality rate is higher, we cannot establish a cause and effect relationship because the cause of death is not reliably or consistently available in the SRTR database. There are no data about HCV-specific outcomes, including viral loads, histological outcomes, antiviral therapy, and the attributability of death and/or graft loss to recurrent HCV. Prospectively obtained data from large randomized controlled trials with protocol biopsy samples assessing the comparative histological severity of recurrence would be needed to definitively determine whether the increased mortality and graft loss associated with sirolimus use are directly attributable to sirolimus. In the context of a Food and Drug Administration black-box warning against sirolimus use in LT recipients, such a study is highly unlikely to take place.
In conclusion, this study of a large national database demonstrates that sirolimus is associated with worse patient and graft survival among LT recipients with HCV infections. Absent the publication of results from prospective randomized controlled trials and/or in vivo studies refuting or confirming a negative impact of mTOR inhibition on posttransplant outcomes, the results of this analysis suggest that mTOR inhibitors should be used with great caution in LT recipients with HCV infections.
The authors thank Rachel Pedersen for her assistance with the initial statistical analysis.