Use of sirolimus in liver transplant recipients with renal insufficiency: A systematic review and meta-analysis

Authors


  • This study was supported by a grant from the National Institutes of Health (R01DK-34238 to W. Ray Kim) and by a digestive diseases training grant from the National Institutes of Health (T32 DK07198 to Sumeet K. Asrani).

  • Potential conflict of interest: Nothing to report.

Abstract

Sirolimus is used in patients with renal insufficiency after liver transplantation (LT) and especially in those with calcineurin inhibitor (CNI)–associated nephrotoxicity. We conducted a systematic review of all randomized controlled trials and observational studies to test the hypothesis that the use of sirolimus is associated with an improvement in renal function at 1 year in LT recipients with renal insufficiency [glomerular filtration rate (GFR) < 60 mL/minute or creatinine level ≥ 1.5 mg/dL]. We performed a search of all major databases, conference proceedings, and relevant journals through December 2009 and contacted content experts, corresponding authors, and the pharmaceutical manufacturer. A random effects model was used to determine the pooled estimate of the change in renal function and pooled risk estimates of adverse events that may be associated with sirolimus-based therapy at 1 year. Eleven studies (three randomized controlled trials and eight observational studies) met the final inclusion criteria. A nonsignificant improvement of 3.38 mL/minute [95% confidence interval (CI) = −2.93 to 9.69] was observed in methodologically sound observational studies and controlled trials reporting the primary outcome. In controlled trials, baseline GFR >50 mL/min sirolimus use was associated with an improvement of 10.35 mL/minute (95% CI = 3.98-16.77) in GFR or creatinine clearance. Sirolimus was not significantly associated with death [relative risk (RR) = 1.12, 95% CI = 0.66-1.88] or graft failure (RR = 0.80, 95% CI = 0.45-1.41), although reporting was incomplete. It was associated with a statistically significant risk of infection (RR = 2.47, 95% CI = 1.14-5.36), rash (RR = 7.57, 95% CI = 1.75-32.70), ulcers (RR = 7.44, 95% CI = 2.03-27.28), and discontinuation of therapy (RR = 3.61, 95% CI = 1.32-9.89). Conclusion: Conversion to sirolimus from CNIs is associated with a nonsignificant improvement in renal function in LT recipients with renal insufficiency, although the results are limited by heterogeneity, a risk of bias, and a lack of standardized reporting. (HEPATOLOGY 2010;)

Renal insufficiency in liver transplantation (LT) recipients is associated with progression to end-stage renal disease and a decrease in patient and graft survival.1-4 After LT, calcineurin inhibitors (CNIs) contribute further to the development of chronic renal failure.5-8 Minimizing the nephrotoxicity of immunosuppressive regimens may help to reduce the number of patients developing chronic renal failure and its associated morbidity and mortality after LT.

Sirolimus is used as an alternative to CNI-based immunosuppression, especially because of its nonnephrotoxic side-effect profile. However, outcomes with sirolimus-based conversion in LT recipients have not been systematically studied. Part of the hesitation may be due to the adverse events associated with sirolimus. In 2002, the Food and Drug Administration (FDA) issued a black box warning regarding the risk of hepatic artery thrombosis with de novo sirolimus-based primary immunosuppression. In June 2009, the FDA notified health care professionals of preliminary data suggesting increased mortality in stable LT patients after conversion from a CNI-based immunosuppression regimen to sirolimus.9 Nevertheless, on the basis of data from small single-center studies, sirolimus is actively used in many LT centers for patients with renal insufficiency, often with contrasting results.

We hypothesized that the use of sirolimus in LT recipients with renal insufficiency would be associated with an improvement in renal function at 1 year versus CNI-based therapy or other standard immunosuppression. Additionally, we reviewed reported side effects and adverse events associated with sirolimus.

Abbreviations

AASLD, American Association for the Study of Liver Diseases; ALD, alcoholic liver disease; ATC, American Transplant Congress; CI, confidence interval; CNI, calcineurin inhibitor; CrCl, creatinine clearance; CYC, cyclosporine; DDW, Digestive Disease Week; df, degrees of freedom; EASL, European Association for the Study of the Liver; EDTA, ethylene diamine tetraacetic acid; eGFR, estimated glomerular filtration rate; EMBASE, Excerpta Medica Database; FDA, Food and Drug Administration; GFR, glomerular filtration rate; HCV, hepatitis C virus; ILTS, International Liver Transplantation Society; IQR, interquartile range; LT, liver transplantation; MDRD, Modification of Diet in Renal Disease; MMF, mycophenolate mofetil; MOOSE, Meta-Analysis of Observational Studies in Epidemiology; NA, not available; PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses; RCT, randomized controlled trial; RR, relative risk; RRT, renal replacement therapy; SD, standard deviation; SMD, standardized mean difference; TAC, tacrolimus.

Patients and Methods

We used the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and Meta-Analysis of Observational Studies in Epidemiology (MOOSE) recommendations for study reporting.10, 11

Inclusion and Exclusion Criteria.

In primary adult LT recipients, we examined both the use of sirolimus as primary immunosuppression in patients with renal insufficiency and conversion to sirolimus from a reference immunosuppression regimen due to nephrotoxicity. We defined renal insufficiency as either a serum creatinine measurement ≥ 1.5 mg/dL or an estimated glomerular filtration rate (eGFR) < 60 mL/minute/1.73 m2 (this was based on the criteria of the Kidney Disease Outcomes Quality Initiative of the National Kidney Foundation).12, 13 The comparison or reference group could receive a CNI (tacrolimus or cyclosporine at a low dose or the standard dose), mycophenolate mofetil (MMF), or interleukin-2 inhibitors (daclizumab or basiliximab) separately or in combination without sirolimus. All doses of steroids or other secondary immunosuppression (e.g., azathioprine) were eligible.

We included all randomized controlled trials and observational studies with a comparison group. Review articles, editorials, and letters that did not report original data were excluded.

Search Strategy and Selection.

We assembled a team with content (R.H.W. and W.R.K.) and methodological expertise (C.P.W. and M.H.M.). A medical librarian with expertise in conducting systematic reviews (P.J.E.) was consulted to devise a sensitive search strategy. We searched PubMed, Excerpta Medica Database, Scopus, Web of Science, the Cochrane Central Register of Controlled Trials, and the Hepatobiliary Group in the Cochrane Library through December 25, 2009 without restriction on the publication status or the language of publication. We combined database-specific search terms for sirolimus (sirolimus or rapamycin or mammalian target of rapamycin inhibitors or Rapamune) and LT (liver or liver transplantation or hepatic transplantation or hepatic graft or LT or OLT). A hand search of relevant journals and annual meetings was also conducted. Further details regarding the search strategy are provided in Appendix 1 (Supporting Information). Authors of relevant abstracts were contacted to obtain any unpublished data (if available). All reference sections of eligible studies and pertinent reviews were hand-reviewed for potential studies. We also requested information on potentially eligible studies from content experts and contacted the manufacturer of sirolimus for potential trials. We included relevant studies reported in foreign languages for translation into English (J.T.). Finally, authors of the included studies were contacted for data clarification and for input about the accuracy of our extraction.

Two reviewers (S.K.A. and M.D.L.) independently considered the eligibility of potential abstracts and titles. Interrater reliability was calibrated, and retrieval strategies were refined with a smaller set of reports. When there was disagreement about a study or a lack of information for an accurate assessment of eligibility, the study was carried to the full-text stage for evaluation.

Both reviewers worked independently and considered the full-text reports for eligibility. The kappa statistic (a measure of agreement) was 0.83 with 94.4% observed agreement. Disagreements were harmonized by consensus and, when this was not possible, by arbitration (W.R.K.). The excluded studies and the reasons for exclusion are included in Appendix 2 (Supporting Information).

Data Extraction.

All data extraction was performed in duplicate by both reviewers independently. We extracted the inclusion and exclusion criteria, the enrolled participants (patient flow, age, sex, reason for transplantation, previous CNI dose, and length of follow-up), the sirolimus-associated characteristics (timing of intervention, bolus and maintenance dose, and tapering of the previous therapy versus abrupt cessation), and the reference group immunosuppression regimens. Next, we assessed the risk of bias. For all trials, we assessed the loss to follow-up and the way in which missing data were handled. For randomized controlled trials, we applied the Cochrane Collaboration risk assessment tool.14 For assessing the quality of observational studies, we modified criteria suggested by the Newcastle-Ottawa quality assessment tool for observational studies.15 The primary outcome was the change in renal function at 1 year versus the baseline [creatinine or glomerular filtration rate (GFR)]. The secondary outcomes were survival, graft failure, episodes of acute rejection, and the requirement of renal replacement therapy at 1 year. We also examined the following adverse events: dyslipidemia requiring statin use, infection, edema, rash, oromucosal ulcers, proteinuria, poor wound healing, pneumonitis, and hepatic artery thrombosis.

Statistical Analysis.

The primary analysis determined the mean change in the renal function from the baseline to 1 year in the sirolimus group versus the reference group. We generated relative risk (RR) estimates for the secondary dichotomous outcomes (e.g., acute rejection episodes). A random effects model for pooled estimates and associated confidence intervals (CIs) was used. Heterogeneity was explored with the I2 test.16 We calculated separate pooled estimates for randomized controlled trials.

We also calculated a standardized mean difference (SMD) to obtain a pooled estimate for all controlled trials and observational studies. The SMD is used as a summary statistic in meta-analyses when studies measure the same outcome (i.e., a change in renal function) but measure it in a variety of ways (e.g., GFR, creatinine, or creatinine clearance). The SMD expresses the size of the intervention effect with respect to the variability or the standard deviation observed in that study.14 Because serum creatinine and eGFR have an inverse relationship, the mean values in studies reporting serum creatinine were multiplied by −1 to obtain the SMD. After calculation of the SMD, it was back-transformed and re-expressed in familiar scales (eGFR) by multiplication of the SMD by the baseline standard deviation of a representative observational study.17

To explore causes of inconsistency and subgroup-treatment interactions, a priori subgroup analyses were performed on the basis of the trial design (randomized trials versus observational studies), interventions (tapering versus abrupt cessation of CNIs, bolus versus no sirolimus bolus, timing of intervention, and CNIs versus no CNIs), and patient characteristics (>50% male). A sensitivity analysis was performed to examine whether the results changed when borderline-eligible articles were excluded. We used RevMan 5 to conduct the analyses.

Missing Information.

We encountered two main types of missing information; these involved either the precision of the estimates of the change in renal function (no standard deviation provided) or an incomplete description of all adverse events of interest. To reconcile this, we provided all authors with a data extraction sheet and requested the entry of missing items. When information was still lacking, standard deviations were imputed (Appendix 3, Supporting Information). We also included information from trials reporting no events in either arm over the follow-up period. Although the inclusion of these zero-event trials may have driven the results toward the null, they increase the precision of the estimate and allow for the inclusion and evaluation of the totality of the evidence.18 To ensure that our imputations did not change our conclusion, we conducted sensitivity analyses in which the assumptions used in the imputations were varied.

Results

Ascertainment of the Studies.

Figure 1 summarizes the retrieval process used for the systematic review.

Figure 1.

Flowchart of the included studies. Abbreviations: AASLD, American Association for the Study of Liver Diseases; ATC, American Transplant Congress; DDW, Digestive Disease Week; EASL, European Association for the Study of the Liver; EMBASE, Excerpta Medica Database; ILTS, International Liver Transplantation Society; RCT, randomized controlled trial.

There were 16 eligible studies: 2 reports were excluded because they characterized the same study,19, 20 1 study was excluded because it was a non-English study indexed only by the National Library of Medicine,21 and 1 study was excluded because Wyeth Pharmaceuticals was unable to share unpublished data.22 One study was available only in abstract form, and no further data regarding renal function were available despite author contact.23 The sponsor of the study, Genentech, was unable to share unpublished data.

In all, 11 studies (3 randomized controlled trials24-26 and 8 observational studies17, 27-33) were included for the systematic review and meta-analysis. For two of these studies, the authors provided recently accepted unpublished manuscripts.24, 29 Only abstract data were available for two other studies.30, 31 Authors of 7 of the 11 studies provided further information.

Study Characteristics.

Table 1 lists the characteristics of the included studies; further details regarding patient selection, comparison groups, and outcomes are provided in Appendix 4 (Supporting Information). Most were single-center studies, and the total number of patients per study ranged from 16 to 148. Eighty-six patients were included in the three controlled trials. A majority of the patients were male and underwent transplantation for either alcoholic liver disease or hepatitis C.

Table 1. Characteristics of the Studies
Authors/CountrySirolimus (n)Reference (n)Age in Years [Mean (SD)]Female (%)ALD/HCV (%)Baseline Renal Function [Mean (SD)]Measurement of Renal Function
SirolimusReference
  • Abbreviations: ALD, alcoholic liver disease; CrCl, creatinine clearance; EDTA, ethylene diamine tetraacetic acid; HCV, hepatitis C virus; MDRD, Modification of Diet in Renal Disease; NA, not available; SD, standard deviation.

  • *

    Duplicate study by Willcocks et al.20

  • Median age and range.

  • Including cryptogenic cirrhosis.

  • §

    Including hepatitis B cirrhosis.

  • Duplicate study by Kniepeiss et al.19

  • Abstract data only.

  • #

    Abstract data only.

  • **

    A smaller number of participants were evaluated for renal function: 43 in the sirolimus arm and 40 in the reference arm.

Controlled trials        
 Watson et al.26 (2007)/United Kingdom*131759 (54-66)4130/050.0 (9.3)47.2 (13.4)GFR: 6-point EDTA clearance
 Shenoy et al.25 (2007)/United States202059 (7)5220/1364 (18)60 (12)24-hour CrCl
 Eisenberger et al.24 (2009)/Switzerland8856 (9)2531/31§55.4 (24)66.9 (17)CrCl: Cockcroft-Gault formula
Observational studies        
 Montalbano et al.28 (2004)/United States12853NA55/1043.6 (NA)55 (NA)CrCl
 Kniepeiss et al.27 (2005)/Austria223654 (26-70)2431/291.7 (NA)1.4 (NA)CrCl
 Smallwood et al.30 (2005)/United States273458 (10)NANA1.97 (1.26)2.58 (0.78)Serum creatinine
 Zaghla et al.32 (2006)/United States2810154 (10)359/402.3 (1.5)1.3 (0.82)Serum creatinine
 Dubay et al.17 (2008)/Canada575752 (21-70)459/3337.0 (12.1)37.3 (10.5)CrCl: Cockcroft-Gault formula
 Thompson et al.31 (2008)/United States#4258NANA20/3140.3 (NA)33.3 (NA)eGFR: MDRD
 Zhang et al.33 (2008)/China201950 (3)285/2.5Early: 35.33 (5.93) Late: 49.53 (12.69)Early: 37.2 (9.37) Late: 52.4 (11.73)CrCl
 Rogers et al.29 (2009)/United States72**65**52 (10)NANAEarly: 40.7 (24.7) Late: 30.6 (12.6)73.4 (34.24)eGFR: MDRD

Table 2 shows details regarding the immunosuppression regimen. Two studies examined de novo sirolimus use at the time of transplantation,32, 34 and nine studies examined patients whose immunosuppression regimens were converted to sirolimus. Most studies restricted their analysis to the late initiation of sirolimus (>6 months after transplantation). Two studies had separate subgroups of patients who underwent early initiation (<6 months).29, 33 The comparison group received only a CNI,17, 24-26, 28, 29 a CNI and MMF,27, 33 or only MMF.30, 31 Zaghla et al.'s study32 had two sirolimus arms: (1) sirolimus alone with or without MMF and (2) sirolimus and a CNI with or without MMF. The former sirolimus group was compared to the reference arm. The studies differed in the method of switching from a CNI to sirolimus (abrupt cessation versus tapering) and in the administration of sirolimus (bolus versus no bolus, target sirolimus level, and dose).

Table 2. Immunosuppression in the Studies
Authors/CountryImmunosuppression in the Sirolimus Arm Before Initiation or Conversion (n)Method of Switching from the CNI in the Sirolimus ArmTime to Sirolimus Initiation*Sirolimus Bolus/Dose/TroughCNI in the Sirolimus Arm After Initiation or ConversionImmunosuppression in the Reference Group at Initiation or Conversion (n)
  • Abbreviations: CYC, cyclosporine; IQR, interquartile range; NA, not available; SD, standard deviation; TAC, tacrolimus.

  • *

    The mean time is stated unless otherwise specified.

  • The reference group received only MMF and did not receive CNIs.

  • The target was the lower limit of the range for TAC (10-12 ng/mL) or CYC (250-300 ng/mL).

Controlled trials      
 Watson et al.26 (2007)/United KingdomCYC: 3AbruptMedian: 3.1 years (IQR = 0.88-7.4)None/2 mg/5-15 ng/mLNoneCYC: 2
TAC: 10TAC: 12
 Shenoy et al.25 (2007)/United StatesCYC: 15Overlap (1-2 weeks)4.4 years5 mg/3 mg/6-10 ng/mLNoneCYC: 16
TAC: 5TAC: 4
 Eisenberger et al.24 (2009)/SwitzerlandCYC: 6AbruptMedian: 50 months (range = 6-123)10-15 mg/3-5 mg/6-16 ng/mLNoneCYC: 5
TAC: 2TAC: 3
Observational studies      
 Montalbano et al.28 (2004)/United StatesNot applicableNot applicableAt transplantMean dose: 8.7 ng/mLTAC: 12TAC: In all?
 Kniepeiss et al.27 (2005)/AustriaCYC: 1NA59 months (range = 1-126)NA/NA/4-10 ng/mLCYC: 1CYC: 31
TAC: 14TAC: 14TAC: 5
MMF: 20MMF: 20MMF: 29
 Smallwood et al.30 (2005)/United StatesNANA4.5 years (SD = 3)4 mg?/9.1 mg (4.3 mg)/NANoneMMF: 34
 Zaghla et al.32 (2006)/United StatesNot applicableNot applicableAt transplantNone/5-7 mg/NANoneNA
 Dubay et al.17 (2008)/CanadaCYC: 38Overlapping and abruptMedian: 45 months (range = 3-204)None/1-2 mg/5-15 μg/dLLow-dose CNI: 5CYC: 36
TAC: 19TAC: 21
 Thompson et al.31 (2008)/United StatesNANA47.8 monthsNANoneMMF: 58
 Zhang et al.33 (2008)/ChinaCYC/TAC: 20Taper (3 days)0-9 months after LTNA/2 mg/NANoneCYC/TAC: 19
MMF: AllMMF: All
 Rogers et al.29 (2009)/United StatesNATaperEarly: 50 days (20) Late: 309 days (SD = 292)NA/NA/3-8 ng/dLNoneCYC: NA
TAC: NA

Risk of Bias.

We evaluated the risk of bias in the three published randomized controlled trials (Table 3). Allocation sequence generation was either described or confirmed by author contact for all studies. Allocation concealment was clearly described only by Watson et al.26 None of the randomized controlled trials were blinded to the observer or the patient. Neither Eisenberger et al.24 nor Shenoy et al.25 used intent-to-treat analyses. Although Watson et al.'s study was meant to be an intent-to-treat analysis, randomization was not preserved because three patients who were randomized to a CNI were not analyzed. Also, the trial was stopped during the interim analysis and was not powered for the primary outcome.

Table 3. Risk of Bias in the Published Controlled Trials
AuthorsAllocation Sequence DescribedAllocation ConcealmentBlindingSelective Outcome Reporting*Other Potential BiasHandling of Missing Data
PatientPersonnelAssessor
  • *

    A discrepancy in the findings was reported in the final article versus the preceding conference proceedings.

  • Three patients who were randomized to CNIs were not analyzed.

  • The author provided the information.

Watson et al.26 (2007)YesYesNoNoYesNoNot intention-to-treat, not powered, stopped during the interim analysisUnclear
Eisenberger et al.24 (2009)YesNoNoNoNoNoNot intention-to-treatUnclear
Shenoy et al.25 (2007)YesNoNoNoNoNo Last value forward

The risk of bias was evaluated by modification of the Newcastle-Ottawa scale (Table 4) for the six observational studies. Only abstract data were available for the other two studies.30, 31 The comparison groups in all observational studies consisted of LT recipients who did not receive sirolimus. The comparability of the sirolimus and comparison groups was determined by restriction in three of the six studies.17, 28, 32 No other method for ensuring comparability (adjustments for confounders, stratification, or multivariable analysis) was described for the other studies. Selection bias may have influenced the outcome in all trials because the reasons for allocation to the sirolimus group or the reference group may have varied. Allocation may have been at the physician's discretion in most studies, although Rogers et al.29 reported that a protocol was in place for sirolimus initiation. Data about the primary outcome at 1 year was not available for all patients. Methods for handling missing data were not adequately described in any study. Patient flow was adequately described in one study.17

Table 4. Risk of Bias in the Observational Studies Using Ottawa-Newcastle Rules and Other Criteria
AuthorsRepresentative Cohort/ReferenceExposure AscertainmentComparability of Sirolimus Group and ReferenceOutcome AssessmentSufficient Duration?Follow-UpSelection Bias*Missing Data and Other
  • Abbreviation: RRT, renal replacement therapy.

  • *

    The physician decided the arm allocation, and no protocol was present.

Montalbano et al.28 (2004)Yes/Same patient baseMedical recordsRestriction to early initiation (matching ambiguous)Record linkageYesDetails providedPossible bias in allocation to the sirolimus groupUnclear if only positive reports are presented
Kniepeiss et al.27 (2005)Yes/Same patient baseUnclearUnclearUnclearYesUnclearPossible bias in allocation to the sirolimus groupUnclear
Zaghla et al.32 (2006)Yes/Same patient baseMedical recordsRestriction to early initiation and exclusion of patients not on sirolimus plus a CNIRecord linkageYesUnclear: large dropout and uncertainty about handling of RRTPossible bias in allocation to the sirolimus groupUnclear
Dubay et al.17 (2008)Yes/Same patient baseMedical recordsRestriction to late initiation and matching on gender, year of transplant, and creatinineRecord linkageYesUnclearPossible bias in allocation to the sirolimus groupUnclear if data are available for all patients at 12 months
Zhang et al.33 (2008)Unclear/Same patient baseUnclearNo restriction or matching (controls were reported, but matching was not described)UnclearNoNo 12-month data availablePossible bias in allocation to the sirolimus groupUnclear
Rogers et al.29 (2009)Yes/Same patient baseMedical recordsNo restriction or matchingRecord linkageYes>3-month renal follow-up for only 42 of 72Protocol for sirolimus initiationUnclear

Changes in Renal Function over One Year.

Because of the variability in the reporting of the outcome (four studies reported the change in creatinine, three studies reported either the change in creatinine clearance or GFR, and four studies reported both), we calculated the SMD.

Figure 2 shows that when all controlled trials and observational studies with complete data were considered, sirolimus use was associated with a nonsignificant improvement in renal function at 1 year (SMD = 0.30, 95% CI = −0.26 to 0.86). For example, after transformation into the eGFR scale, this implies that sirolimus was associated with an improvement of 3.38 mL/minute (95% CI = −2.93 to 9.69).

Figure 2.

SMDs of changes in renal function from the initiation of sirolimus to 1 year. The test for subgroup interaction was not significant (P = 0.17). After back-transformation into the eGFR scale, sirolimus was associated with an improvement of 3.38 mL/minute (95% CI = −2.93 to 9.69). Borderline-eligible observational studies are not included (see Appendix 6, Supporting Information). Smallwood et al.30 and Zaghla et al.32 reported only creatinine. The mean values of the studies reporting serum creatinine were multiplied by −1 to make the scale uniform. Abbreviations: df, degrees of freedom; SD, standard deviation.

All controlled trials had late initiation of sirolimus. In all three trials, the mean eGFR at the baseline was >50 mL/minute (Table 1). When we limited the analysis to the three controlled trials (Fig. 3), sirolimus use was associated with an improvement of 10.38 mL/minute (95% CI = 3.98-16.77, I2 = 0%) in renal function.24-26

Figure 3.

Pooled estimates of changes in GFR or creatinine clearance (mL/minute) from the initiation of sirolimus to 1 year in controlled trials. Abbreviations: df, degrees of freedom; SD, standard deviation.

Imputation for the mean change and standard deviation of the change was required in most studies. Additionally, it is unclear whether patients who progressed to renal replacement therapy (and therefore deterioration in renal function) were included in the analysis of the change in renal function. Data could not be included from Thompson et al.31 (the endpoint was not well specified) or Montalbano et al.28 or Kniepeiss et al.27 (no estimate of variability for the baseline or 1 year was provided). Furthermore, only 6-month data were available from Zhang et al.,33 and for the reference group of Rogers et al.,29 only eGFR values before transplantation (rather than before sirolimus conversion) and at the end of follow-up (rather than 1 year after sirolimus conversion; mean = 616 days) were available. However, the inclusion of these borderline-eligible observational studies yielded a more favorable result (SMD = 0.89, 95% CI = 0.23-1.55) or an improvement of 10.03 mL/minute (95% CI = 3.27-17.47; Appendix 5, Supporting Information).

Secondary Outcomes.

Reporting on secondary outcomes was incomplete across all studies. We assessed the RRs of adverse outcomes occurring within 1 year of the initiation of sirolimus (Fig. 4). Specific details regarding the studies included for each outcome are provided in Appendix 6 (Supporting Information). At 1 year, the RRs of patient death (RR = 1.12, 95% CI = 0.66-1.88), graft failure (RR = 0.80, 95% CI = 0.45-1.41), and rejection (RR = 0.88, 95% CI = 0.44-1.76) were not significantly increased in the sirolimus arm versus the reference arm. However, sirolimus use was associated with a nonsignificant RR of renal replacement therapy (RR = 1.71, 95% CI = 0.83-3.51) and the need for statin therapy (RR = 2.93, 95% CI = 0.52-16.55). It was associated with a statistically significant risk of infection (RR = 2.47, 95% CI = 1.14-5.36), rash (RR = 7.57, 95% CI = 1.75-32.70), edema (RR = 2.49, 95% CI = 1.09-5.66), and oromucosal ulcers (RR = 7.44, 95% CI = 2.03-27.28). The rates of proteinuria and poor wound healing were similar for the sirolimus and reference groups. Insufficient details were provided to obtain the RR for either pneumonitis or hepatic artery thrombosis.

Figure 4.

Pooled RR estimates and 95% CIs of side effects and adverse events from the initiation of sirolimus to 1 year. Only studies reporting 1-year data are presented.

Rates of discontinuation due to tolerability and potential side effects of therapy were significant in the sirolimus arm (RR = 3.61, 95% CI = 1.32-9.89). In addition, three studies reported discontinuation rates for the sirolimus arm but did not provide data regarding the reference arm. In these studies, discontinuation rates in the sirolimus arm ranged from 33% to 55%.17, 28, 29

Subgroup and Sensitivity Analyses.

We conducted prespecified subgroup and sensitivity analyses to explore heterogeneity on the basis of the study design, patient population, sirolimus administration, and comparison group. Pooled estimates by trial design are provided in Fig. 2.

In certain cases, subgroup analysis was not possible because of the limited number of trials or missing information. For ease of interpretation, estimates expressed as SMDs were transformed into the eGFR scale. Subgroup analysis by early initiation was not possible because of incomplete information. All controlled trials had late initiation of sirolimus. Sirolimus was associated with a nonsignificant GFR change of 4.62 mL/minute (95% CI = −1.01 to 10.26) when the analysis was limited to studies with late initiation of sirolimus versus a CNI-only reference group.17, 24-26 Sirolimus was associated with a nonsignificant GFR improvement of 1.35 mL/minute (95% CI = −10.81 to 13.52) in studies in which a sirolimus bolus was used24, 25, 30 and with an improvement of 9.58 mL/minute (95% CI = 2.37-16.69) when CNI cessation was abrupt.24, 26 Results were consistent across sensitivity analyses for the imputation of standard deviations and for studies with no events in either arm at the end of the follow-up period (Appendix 3, Supporting Information).

Discussion

Reducing the nephrotoxicity associated with current immunosuppressive regimens, especially in patients with underlying renal insufficiency, is important in optimizing the long-term outcome of LT recipients. We conducted a systematic review and meta-analysis to assess the role of sirolimus in improving renal function after LT. Our study has several important findings: First, there is a paucity of well-controlled trials and studies examining this issue. Current studies were heterogeneous, compared different interventions to sirolimus use, and were susceptible to bias. Second, the initiation of sirolimus is associated with a nonsignificant improvement in renal function. Third, the risks of developing various infections, rash, oromucosal ulcerations, and discontinuation of therapy were higher for patients treated with sirolimus. Finally, the reporting of outcomes of renal function and adverse events is nonuniform, and this severely limits the conclusions.

A recent manufacturer-sponsored randomized open-label trial of conversion from a CNI to sirolimus treatment versus continued CNI treatment was completed.23 Patients with a GFR > 40 mL/minute were included. At the end of 12 months, the baseline-adjusted GFR was similar in patients receiving sirolimus (62 mL/minute) and those maintained on a CNI (63 mL/minute). However, details regarding the baseline GFR and percentage change were not available to the authors despite contact with Wyeth Pharmaceuticals. Whether these results will be published remains to be seen. This trial formed the basis of a recent FDA-sponsored alert regarding the elevated risk of death seen in the sirolimus arm. The number of deaths, though not statistically significant, was higher in the sirolimus arm (3.8% versus 1.4%). Therefore, the use of sirolimus must be tempered with the risk of adverse events. In our meta-analysis, the risk of death (seven studies) or graft failure (six studies) at 1 year was not significantly higher in the sirolimus arm. However, the inherent weaknesses of the individual studies as well as the lack of complete reporting of the outcomes may be responsible for the nonsignificant effect seen in the meta-analysis.

Keeping the limitations of the individual studies and trials in mind, we found that sirolimus use was associated with an improvement in renal function in the three controlled trials. When this was combined with methodologically sound observational studies, the improvement in renal function was no longer significant. All three trials examined late conversion (>6 months) from CNI-based therapy to sirolimus. However, the mean GFR in the study was >50 mL/minute, and hence it is unclear whether this degree of improvement will be seen in patients with worse renal function. Furthermore, only 86 patients were analyzed in the three trials. In addition, the absence of differences in the open-label trial sponsored by Wyeth and the unavailability of these data lend uncertainty to the true benefit of sirolimus in improving renal function. Whether the marginal improvement in renal function is sustained over the long term (beyond 1 year) is also unknown. This is especially important because the rate of discontinuation for sirolimus was high.

Alternatively, there may be a renal threshold or a point of no return beyond which sirolimus initiation may not provide a benefit. Indeed, among patients assigned to a sirolimus arm, a higher percentage progressed to requiring renal replacement therapy, although this was not significant. This was primarily seen in observational studies and suggests a selection bias. The increased rate of renal replacement therapy may imply that sirolimus was used in persons identified by their providers as having rapidly declining renal function prompting active changes to the immunosuppression regimen. Sirolimus may have been used too late in patients who may have derived a potential benefit from early initiation when renal function was preserved. Hence, progression to renal replacement therapy could not be prevented. Indeed, as reported by Dubay et al.,17 patients who were exposed to a CNI for more than 5 years or those with an initial creatinine clearance of less than 30 mL/minute who were converted to sirolimus did worse than control patients maintained on low-dose CNIs.17 Among the three controlled trials (n = 86), there was only one instance of renal replacement therapy in the sirolimus arm.

Therefore, despite our rigorous methodology, the inherent limitations of the individual studies and trials preclude a stronger conclusion and further emphasize the need for large, well-designed studies with longer follow-up and well-characterized endpoints.

This study has several strengths. We conducted an extensive literature search and provided the most up-to-date information. Using two reviewers, we ensured interrater reliability and assessed the inclusion of articles and extracted data independently. We aggressively contacted authors of the included studies and content experts to accurately represent our findings. We improved interrater reliability by creating standard data extraction sheets and conducting pilot searches. We minimized publication bias by the inclusion of non-English studies, unpublished data, and conference proceedings over a large time period. We applied known criteria to judge the quality of the trials and observational studies. We incorporated the MOOSE and PRISMA guidelines for the reporting of our systematic review. Finally, we contacted all authors of eligible trials. We assessed the appropriateness of our imputations in dealing with missing data. We were able to examine a multitude of adverse events.

However, our study may have limitations. Only 50% of the authors who were contacted provided information. Nonetheless, author contact is rarely performed for meta-analyses and is a significant strength of our methodology. Imputations were required for many of the studies. However, we conducted a sensitivity analysis to ensure that the imputations did not alter the results. We were unable to include two unpublished reports, and their authors could not be reached. The results of at least one of them would have enhanced our primary outcome. Finally, pooling may not have been appropriate in all cases because of the heterogeneity. However, we felt that providing an estimate with a statement of the limitations of the primary studies would provide useful information to the reader. Because of the small number of studies, we did not feel that a funnel plot would provide useful information. Through the rigorous and sensitive search strategy outlined previously, any publication bias should have been averted. We were unable to conduct a subgroup analysis of patients with early initiation of sirolimus (<6 months) because of incomplete data. Whether a larger benefit would have been seen with earlier initiation remains to be seen. On the other hand, the incremental improvement in GFR may not be sustained in patients with worse renal function. However, stage-specific information was not available for the trials.

In summary, the use of sirolimus was associated with a nonsignificant improvement in GFR. However, sirolimus use was also associated with the development of oral ulcers, rash, various infections, and discontinuation of therapy. A high-quality randomized controlled trial limited to the early initiation of sirolimus, a standardized definition of renal insufficiency, and the identification of the degree of renal dysfunction beyond which sirolimus may not be beneficial is needed to better define the role of sirolimus in the treatment of these patients.

Ancillary