Safety of Immunosuppressive Drugs Used as Maintenance Therapy in Kidney Transplantation: A Systematic Review and Meta-Analysis

To evaluate the safety of regimens containing calcineurin inhibitors (CNI), proliferation signal inhibitors (TOR-I) and antimetabolites, we conducted a meta-analysis of randomized clinical trials (RCTs) and observational studies. A total of 4,960 citations were identified in our electronic search and 14 additional articles were identified through hand searching. Forty-eight articles (11,432 participants) from 42 studies (38 RCTs and four cohorts) met the inclusion criteria. Meta-analysis results revealed the following: (i) tacrolimus was associated with an increased risk for diabetes and lower risk of dyslipidemia, compared to cyclosporine; (ii) mycophenolate mofetil (MMF) was associated with increased risk for total infections, abdominal pain, diarrhea and vomiting, compared with azathioprine; (iii) sirolimus was associated with higher risk of anemia, diabetes, dyslipidemia, lymphoceles and withdrawal compared to tacrolimus or cyclosporine, and cyclosporine was associated with an increased risk of CMV infection; (iv) the combination of CNI with antimetabolites was associated with more adverse events than CNI alone; (v) TOR-I was related to more adverse events than MMF. The data observed in this meta-analysis are similar to those describe by others authors; thus, the choice of treatment must be made by the clinical staff based on specific patient characteristics.

terms of the number of patients who presented a specific outcome. All consolidated data were reviewed to avoid typing errors.

Quality Assessment
The quality of the study was independently accessed by two reviewers, and any disagreements were resolved by consensus. Randomized controlled trials were evaluated using the Cochrane Collaboration Tool [19] considering the following items: random sequence generation, allocation concealment, blinding, and incomplete outcome data. Open label studies were not considered risk of bias. Observational studies were evaluated using Newcastle scale [20].

Data Synthesis and Data Analysis
Outcomes were meta-analyzed if they were reported in at least two articles, within the same treatment arm and at the same time of follow-up. Random effect models were employed to estimate the pooled effect sizes across studies [21]. The results are expressed as the relative risk (RR) with 95% confidence interval (95% CI); p value <0.05 was considered significant. To assess heterogeneity I 2 and p values were used (I 2 >50% and p < 0.05 indicated high heterogeneity) Publication bias was accessed using funnel plot. Single analysis (estimated RR from raw data) was performed if the data was not eligible to enter in meta-analysis. All analysis was conducted using Revman 5.1.

Study Characteristics
A PRISMA flow chart describing the publication screening process and the reasons for exclusion is shown in Figure 1A total of 5,875 citations identified by the electronic search, and 16 additional articles were identified via manual searching. A total of 48 articles (11,432 participants), from 42 studies, 38 RCTs  and four cohorts [66][67][68][69] met the inclusion criteria. Table 1 shows the characteristics of included studies organized by treatment strategy, according to the treatment protocol of each study. CsA was the most prevalent drug in all schemes, as it was used in 34 articles. The majority of studies (33%) compared TAC versus (vs.) CsA, usually using an antimetabolite and a steroid in combination. Within each study, differences between groups in terms of gender, race, age and allograft characteristics were not significant, indicating that the allocation of participants into the treatment groups was satisfactory. Twenty five RCTs were multicenter studies, with the number of centers ranging from 2 to 65 centers. Nearly 50% of the studies were conducted in European countries, 21% of the studies were conducted in the United States, and 14% were conducted in Brazil. Of the 38 RCT studies evaluating immunosuppressants only 13 (34%) reported adequate sequence generation and most studies did not report the allocation concealment (73%). One study [34] used the numbers of records for randomization and was classified as high risk for selection bias. All studies, but four [22,23,26,40] used intention to treat or modified intention to treat analysis. A summary of RCT quality is shown in Figure 2. The four included cohorts [66][67][68][69] assigned three or four stars in the selection domain, one star in the comparability of groups and one to two in the exposure, demonstrating good quality. Funnel plots of meta-analyses were all symmetrical, indicating the absence of bias.

Outcomes
All adverse events reported in the included articles were collected, and the most prevalent events were included in the synthesis. The following outcomes were included: abdominal pain, anemia, bacterial infections (all definitions), cytomegalovirus (CMV) infections, diabetes mellitus (new-onset diabetes mellitus, post-transplant diabetes, and use of hypoglycemic drugs were considered), diarrhea, dyslipidemia (hypercholesterolemia, hypertriglyceridemia and hyperlipidemia were considered), gastritis, total infections (as reported in the study), hypertension (use of antihypertensive drugs was also considered), leukopenia, lymphoceles, malignancies (all types), nausea, vomiting, thrombocytopenia, urinary tract infection (UTI) and withdrawal (discontinuation and crossover of study medication were considered).
For data synthesis and analysis, the comparable schemes in each study were classified into the following groups: CNI vs. CNI; AMETAB vs. AMETAB;TOR-I vs. CNI; CNI + AMETAB vs. CNI; TOR-I vs. AMETAB; and AMETAB vs. CNI. In some studies, it was possible to compare more than one group, such as studies that included the treatment protocol of CNI + AMETAB vs. CNI + AMETAB vs. CNI + AMETAB (it was possible to compare CNI vs. CNI and AMETAB vs. AMETAB).
The results of 13 articles, two cohorts and 11 RCTs, were meta-analyzed and are displayed in Table 2. Both the cohort and RCT pooled results indicate that TAC was associated with an increased risk for diabetes ( Figure 3). This association was also found at 120 months follow-up in one cohort that was not included in the pooled analysis (n = 192; RR = 2.10; 95% CI: 1.17, 3.77; p = 0.01) [67]. The risk of dyslipidemia was reduced in TAC regimens, as shown in the meta-analysis and in two single studies: a cohort of 36 months (n = 506; RR = 0.74; 95% CI: 0.57, 0.97; p = 0.03) [62] and a RCT of 60 months (n = 76; RR = 0.62; 95% CI: 0.40, 0.95; p = 0.03) [31].
Although the other outcomes had no statistical significance in the pooled results, studies in single analysis showed that TAC was associated with a higher risk of withdrawing the treatment at 120 months of follow-up (n = 192; RR = 11.21; 95% CI: 2.50, 50.23; p = 0.002) [67] and that CsA presented a greater risk for hypertension at 36 months of follow-up (n = 89; RR = 0.67; 95% CI: 0.48, 0.94; p = 0.02) [26].  [32]. Sensitivity analysis showed much reduced heterogeneity (p = 0.23, I2 = 29%) when these trials were removed from the analysis. * The study of Ekberg is a RCT with 12 months of experimental data and 36 months of observational data, so that the article containing results at 36 months results was included in cohort analysis.
All five articles were included in meta-analysis and the results are shown in Table 3. MMF was associated with an increased risk for total infections and gastrointestinal discomfort, including abdominal pain, diarrhea and vomiting. The sensitivity analysis for heterogeneity indicated that participants who were taking MMF had a higher risk of withdrawing from the treatment.

TOR-I vs. CNI
Eight articles from five RCTs with 12 months of follow-up and one RCT of 24 months were included in this group: two accessing SRL vs. TAC [47,65], four accessing SRL vs. CsA [43][44][45][46] and two accessing both [61,62]. In these studies, SRL was used at low (4-8 ng/mL) and standard (10-20ng/mL) doses, as well as CsA, whereas TAC was administered only at low dose. SRL was applied as experimental drug and CsA and TAC served as controls.
Five articles were included in the meta-analysis (Table 4). When compared with any of the CNIs, the use of SRL presented a higher risk for anemia, dyslipidemia, lymphoceles and withdrawal. The association between SRL and anemia was also reported by one cohort of 36 months, which was not included in the pooled analysis, compared with CsA (n = 476; RR = 1.30; 95% CI: 1.05, 1.60; p = 0.02) and TAC (n = 472; RR = 1.29; 95% CI: 1.05, 1.60; p = 0.02) [62]. A similar result was observed for dyslipidemia when SRL was compared with TAC (n = 472; RR = 1.42; 95% CI: 1.15, 1.76; p = 0.001) [62]. Single analysis of a RCT comparing SRL and TAC at 24 months showed similar results with the SRL vs. TAC meta-analysis: SRL was associated with increased risk of anemia, dyslipidemia, lymphoceles and withdrawal, and had no difference for malignancy, infections or hypertension [65]. Moreover, the risk of diabetes was higher for SRL, and the risk of CMV infection was higher for CsA.  [44], which compared SRL vs. CsA, and reported outcomes as mouth ulcers, acne, UTI, malignancies, diabetes and withdrawal due to adverse event, but none showed difference between groups in single analysis [45].

CNI + AMETAB vs. CNI
Seven studies were included in this group: one cohort with 240 months of follow-up comparing CsA+AZA vs. CsA [69], one RCT with 12 months of follow-up comparing CsA + MMF vs. CsA [33], three RCTs with 12 and 36 months of follow-up comparing TAC + AZA vs. TAC [36][37][38], one RCT with 12 months of follow-up comparing TAC + MMF vs. TAC [39] and one RCT with 12 months of follow-up comparing CsA + AZA vs. TAC [34]. CsA and MMF were administrated at both low and standard doses, whereas TAC and AZA were used only at the standard dose.
Two studies comparing TAC + AZA vs. TAC at 12 months of follow-up were meta-analyzed and the results are displayed on Table 5. The meta-analysis revealed that the combination of TAC and AZA was associated with a greater risk of leukopenia and withdrawal compared with TAC alone. The RCT with 36 months of follow-up confirmed this result for leukopenia (n = 475; RR = 5.60; 95% CI: 2.39, 13.08; p < 0.01) [37]. In a single analysis, the combination of TAC and AZA was associated with anemia at 12 months of follow-up (n = 475; RR = 1.55; 95% CI: 1.06, 2.28; p = 0.02) [36]. The combination of TAC and MMF at 12 months was associated with a greater risk of gastritis (n = 135; RR = 1.92; 95% CI: 1.18, 3.14; p = 0.009) and leukopenia (n = 153; RR = 3.00; 95% CI: 1.13, 8.01; p = 0.03) [39], compared with TAC alone.

TOR-I vs. AMETAB
Seven RCTs were included in this group: four comparing EVL vs. MMF, at 12 months [54,59], 24 months [60], and 36 months of follow-up [55,56] and three comparing SRL vs. MMF at 12 months [57,58] and 24 months of follow-up [63]. SRL and MMF were used at standard doses, and EVL was used at low (1.5 mg/day) and high (3 mg/day) doses. The studies comparing EVL and MMF were meta-analyzed in subgroups of 12 and 36 months of follow-up and divided based on low-and high-dose EVL (Table 6). Additionally, the studies comparing SRL and MMF at 12 months of follow-up were meta-analyzed. No differences were observed when comparing low-and high-dose EVL. Independent of dosing, EVL was associated with an increased risk of dyslipidemia and withdrawal. MMF presented a higher risk of CMV infection when compared with both doses of EVL. The single analysis for 24 months of follow-up showed similar results: both doses of EVL were associated with increased risk of withdrawal and lower risk of CMV infection and leukopenia when compared with MMF [60]. In the same study, high-dose EVL was associated with a greater risk of diabetes (n = 833; RR = 1.96; 95% CI: 1.18, 1.87; p = 0.01) [60]. Compared with MMF, SRL showed increased risk of withdrawal in meta-analysis and no significant results were found in single-analysis.

AMETAB vs. CNI
Three studies compared an antimetabolite with a CNI: one RCT with 63 months of follow-up that compared MMF vs. TAC [39] and two studies comparing AZA vs. CsA, a cohort with 240 months of follow-up [69] and a RCT with 36 months of follow-up [40]. All drugs were used at standard doses.

Conclusions
Six different groups of immunosuppressant drugs were evaluated and compared. Evaluating the safety of immunosuppressive drugs is complex because kidney transplantation requires the simultaneous use of multiple classes of drugs at varying doses.
The majority of the studies included here showed a low risk of bias, and only one study revealed a high risk of bias for allocation order generation and allocation confidentiality [34]. Based on the parameters described in the Cochrane Handbook [19], the quality of most studies was compromised by a lack of sufficient information to judge the randomization and allocation concealment.
However, this quality assessment did not invalidate the results of the meta-analysis. Overall, the heterogeneity of the treatment-efficacy results was low, indicating small inter-study variability. In general, the observational studies did not show selection bias, and the majority of these studies used the same time in and out of treatment, allowing for a comparison of populations.

Statistics c Study
Compared with CsA, treating kidney transplant patients with TAC resulted in a higher risk for diabetes, whereas those taking CsA had a greater risk of developing dyslipidemia. A retrospective study of risk factors for new-onset diabetes after transplantation (NODAT) found that higher tacrolimus concentrations were an independent predictor of NODAT [70]. In a meta-analysis comparing TAC vs. CsA as the primary immunosuppressant for kidney transplant recipients, TAC-treated patients were two to three times more likely to develop new diabetes mellitus that required insulin. However, the adverse events associated to CsA (constipation, hirsutism, and gingival hyperplasia) were different from those that we found, likely due to the time of use [16].
In regard to MMF vs. AZA, the majority of studies reported a larger number of adverse events for the groups treated with MMF at 12 months of follow-up. The meta-analyses of total infections, vomiting, diarrhea, and abdominal pains statistically favored treatment with AZA. The results of the present meta-analysis agree with the findings of a systematic review conducted in 2009 [71], which found that MMF-treated patients had a greater risk of diarrhea, whereas the risks of CMV infection, anemia, leukopenia and malignancy were not significant.
Our results showed that the use of SRL was associated with higher risk for anemia, dyslipidemia, lymphoceles and withdrawal compared with any CNI. There were no significant differences for infections, UTI, leukopenia, hypertension, or malignancies. These results agree with the findings of a multicenter study which used TAC in combination with different doses of SRL and showed that the incidence of dyslipidemia (hypercholesterolemia) was associated with higher doses of SRL [72]. Another study comparing TOR-I versus CNI found an increased risk of bone marrow suppression outcomes (leukopenia, thrombocytopenia, and anemia), lymphoceles and dyslipidemia for patients taking SRL [73]. Compared with CsA, SRL presented a higher risk for diabetes and reduced risk of CMV infection. Although regimens containing SRL have a higher risk of post-transplant diabetes than regimens without SRL [74], TAC has a higher risk for diabetes than CsA, thus the difference of risk between TAC and SRL may have no significance. Johnston et al. compared SRL with TAC and with CsA and found that patients treated with CsA had the lowest incidence of diabetes (15.6%), followed by SRL (17.8%) and then TAC (19%) [74]. Sirolimus in combination with CNI may increase clinically significant adverse events, such as CNI-related nephrotoxicity and dyslipidemia. Other outcomes include hematologic side effects and a higher incidence of lymphoceles [75]. Furthermore, the use of SRL combined with TAC might increase the risk of post-transplant diabetes mellitus [17,76].
As the majority of studies comparing TOR-I with CNI assess CNI minimization or elimination thru conversion from CNI to TOR-I, the number of studies with such comparison included in the present review was limited, once conversion of drugs was considered exclusion criteria.
Independent of the dose, EVL was associated with increased risk of dyslipidemia and withdrawal. MMF presented higher risk of CMV infection compared with both doses of EVL, but there was no difference in bone marrow suppression (leukopenia and anemia), hypertension, lymphoceles and infections. According to one study, SRL and MMF are associated with similar incidences of both leukopenia and thrombocytopenia [77]. This study reported similar incidences of leukopenia with the combination of MMF and SRL and MMF alone, and similar incidences of thrombocytopenia were observed between their combination and SRL alone, indicating no difference in the risk of these outcomes [77]. The definitions of diabetes and other diseases, such malignancies and dyslipidemia, vary a lot between studies, thus the interpretation of results regarding these diseases should consider the differences between definitions. Many of the clinical trials included were funded by pharmaceutical industries, limiting the interpretation of results, as these companies may benefit from reporting only favorable findings.
Current immunosuppressive protocols use combinations of immunosuppressive agents with different mechanisms of action to maximize efficacy and minimize the toxicity of each drug. The appearance of new immunosuppressive agents and tolerance protocols emerge shows potential as a means to deliver immunosuppression without long-term toxicity. In this regard, belatacept is a second-generation costimulation blocker that in phase 3 trials was to provide effective immunosuppression while avoiding the toxicities associated with calcineurin inhibitors [78].
Modifications are still being introduced in immunosuppressant protocols to take advantage of the drugs' beneficial actions and to reduce the adverse events. Although safety information alone is not enough to base decision making in health, together with reliable information about the long-term efficacy of immunosuppressants, the results of the present review might assist healthcare professionals and managers in choosing the best immunosuppressant regimen. We concluded that the data examined in this meta-analysis are similar to those describe by others authors. Adverse reactions were observed in all classes of immunosuppressive drugs; thus the choice of treatment must be made by the clinical staff based on specific patient characteristics.