The Symphony study showed that at 1 year posttransplant, a regimen based on daclizumab induction, 2 g mycophenolate mofetil (MMF), low-dose tacrolimus and steroids resulted in better renal function and lower acute rejection and graft loss rates compared with three other regimens: two with low-doses of cyclosporine or sirolimus instead of tacrolimus and one with no induction and standard cyclosporine dosage. This is an observational follow-up for 2 additional years with the same endpoints as the core study. Overall, 958 patients participated in the follow-up. During the study, many patients changed their immunosuppressive regimen (e.g. switched from sirolimus to tacrolimus), but the vast majority (95%) remained on MMF. During the follow-up, renal function remained stable (mean change: −0.6 ml/min), and rates of death, graft loss and acute rejection were low (all about 1% per year). The MMF and low-dose tacrolimus arm continued to have the highest GFR (68.6 ± 23.8 ml/min vs. 65.9 ± 26.2 ml/min in the standard-dose cyclosporine, 64.0 ± 23.1 ml/min in the low-dose cyclosporine and 65.3 ± 26.2 ml/min in the low-dose sirolimus arm), but the difference with the other arms was not significant (p = 0.17 in an overall test and 0.077, 0.039 and 0.11, respectively, in pair-wise tests). The MMF and low-dose tacrolimus arm also had the highest graft survival rate, but with reduced differences between groups over time, and the least acute rejection rate. In the Symphony study, the largest ever prospective study in de novo kidney transplantation, over 3 years, daclizumab induction, MMF, steroids and low-dose tacrolimus proved highly efficacious, without the negative effects on renal function commonly reported for standard CNI regimens.
In recent years, progress in immunosuppressive regimens has resulted in fewer acute rejection episodes and excellent short-term outcomes in renal transplantation (1,2), and current research is increasingly targeting factors influencing long-term graft and patient survival (3,4). In this context, there has been considerable interest in immunosuppressive regimens which permit reduction or elimination of calcineurin inhibitor (CNI)-associated and other chronic toxicities, while maintaining adequate immunosuppression (5–10).
The strategies of CNI avoidance, withdrawal and reduced dosing were tested in two large studies (8,11) that were designed at the time when the efficacy of mycophenolate mofetil (MMF) and daclizumab had been established. The use of reduced doses of CNIs was deemed necessary to maintain efficacy and this research line led to the Efficacy Limiting Toxicity Elimination (ELiTE)-Symphony study (9), which was designed to assess whether reduced doses of the adjunct immunosuppressants cyclosporine (CsA), tacrolimus or sirolimus added to the MMF-based regimen containing daclizumab could reduce toxicity (specifically, nephrotoxicity) while maintaining acceptable acute rejection rates. At 1 year posttransplant, the MMF- and daclizumab-based regimen with low-dose tacrolimus and corticosteroids gave superior renal function, graft survival and acute rejection rates compared with regimens containing low-dose CsA or low-dose sirolimus with induction, or standard-dose CsA without induction. Serious adverse events occurred more frequently with low-dose sirolimus than those with the other regimens.
To obtain long-term data on the unique cohort of the Symphony patients, the study had an optional follow-up phase of two additional years. Here, we report the results of the 3-year follow-up of the Symphony study.
Study design and patient population
In brief, the Symphony study included adult patients scheduled to receive a single-organ renal transplant from a living or a deceased donor. Patients receiving a second renal transplant were also eligible, provided that their first graft was not lost due to acute rejection in the first year posttransplant. The study aimed to include patients with a low-to-normal immunological risk. Therefore, exclusion criteria included patients with a current or prior panel reactive antibody (PRA) value of >20%, a positive cross-match, a graft cold ischemia time of >30 h and patients receiving a graft after cardiac death of the donor. Full inclusion and exclusion criteria have been published previously (9).
Prior to transplantation, patients were randomized in equal proportions to one of four treatment groups: MMF with standard-dose CsA and corticosteroids; MMF with low-dose CsA, daclizumab and corticosteroids; MMF with low-dose tacrolimus, daclizumab and corticosteroids or MMF with low-dose sirolimus, daclizumab and corticosteroids.
Intravenous daclizumab was infused over 15–20 min (2 mg/kg within 24 h before the transplant, followed by four doses of 1 mg/kg every 2 weeks). The first doses of MMF, CsA, tacrolimus and sirolimus were administered within 24 h before or after transplantation. All groups received oral MMF (2 g/day). Dosages of CsA, tacrolimus and sirolimus were adjusted to their predefined target trough levels measured using locally available assays. In the standard-dose CsA group, the target trough level was 150–300 ng/mL for the first 3 months and 100–200 ng/mL thereafter; in the low-dose CsA group, the target trough level was 50–100 ng/mL throughout the study; in the low-dose tacrolimus group, the target trough level was 3–7 ng/mL and in the low-dose sirolimus group, the target trough level was 4–8 ng/mL.
All patients received intraoperative and maintenance corticosteroids according to center practice. The minimum maintenance corticosteroid dosages were 20 mg prednisone (or equivalent) for the first 2 weeks posttransplant, 15 mg from week 3 to week 8, 10 mg from week 9 until the end of month 4, and 5 mg thereafter. Treatment failure was defined as the use of additional immunosuppressive medication, discontinuation of any study medication for more than 14 consecutive days or more than 30 cumulative days, graft loss or patient death.
After conclusion of the core study at 12 months, all patients from centers participating in the long-term follow-up were offered the opportunity to enroll in this optional study phase of two additional years' duration.
Patients were included after provision of written informed consent but regardless of protocol violations or discontinuation of the assigned treatments that occurred during the core study. During the second and third year, there was no mandatory treatment, and regimen modifications did not constitute protocol violations. However, the protocol recommended that the patients remained on the treatment assigned at randomization (either arm) with MMF 2 g/day and unchanged low doses of tacrolimus, sirolimus or CsA. In the standard-dose arm, the trough level of CsA was recommended to be reduced to 100–150 ng/mL. The suggested dose of steroids was 5 mg per day or every other day. Steroid withdrawal was not prohibited; however, centers were recommended to follow a consistent strategy for all their patients.
Data collected during the follow-up included serum creatinine measurements every 3 months and further laboratory values and clinical assessments every 6 months. Adverse events were collected in the case report form on an ongoing basis as in the core study.
The Symphony study was designed to assess whether MMF-based regimens with induction and containing reduced doses of adjunct immunosuppressants (CNI or sirolimus) could reduce toxicity, especially nephrotoxicity, while maintaining acceptable acute rejection rates. The main efficacy parameters were the estimated glomerular filtration rate (GFR) using the Cockcroft–Gault formula, estimated GFR during the course of the study and the incidence rates of acute rejection, graft loss and patient death. Safety parameters included the incidence of adverse events, in particular cardiovascular events, malignancies and opportunistic infections, as well as vital signs and laboratory assessments.
The analysis of this follow-up study was exploratory and employed descriptive statistics and inferential methods without correction for multiplicity. Efficacy evaluations were based on the original group assignment at randomization according to the Intent-To-Treat (ITT) principle. For a post hoc analysis of renal function at month 36, we defined an ‘on-treatment’ analysis population as patients participating in the follow-up who, during the last 6 months of the study, had a trough level measurement of the drug assigned at randomization, received MMF and provided a serum creatinine sample at month 36. Safety evaluations were performed based on the same group assignment as in the core safety population that considered the treatment actually administered to patients during the first 3 days after transplantation. Patients who could not be assigned to any of the predefined groups of the safety population were analyzed as a separate group (‘Group 0’). We performed an analysis of renal function with imputation of missing values, including those due to graft loss, with 10 mL/min and with serum creatinine carried forward from month 18 to month 24 and from month 30 to month 36 when these assessments were not performed, and in addition an analysis without imputation in case of graft loss. In line with the primary analysis of the core Symphony study, we performed global group comparisons of GFR using the Kruskal–Wallis test and pairwise comparisons between treatment groups with the Wilcoxon rank-sum test. Time from randomization to biopsy-proven acute rejection (BPAR), graft loss, death and selected adverse events were analyzed using the Kaplan–Meier method (patients not participating in the follow-up were included in the analysis and censored at their last visit) and group differences were assessed using the log-rank test. In a post hoc analysis, we used multivariate logistic regression to investigate the effect of renal function at month 12, treatment group and other covariates on graft loss in the follow-up phase.
Patients and treatments
In the core Symphony study, a total of 1645 patients were randomized to treatment at 83 sites in 15 countries between November 2002 and November 2004. The ITT population comprised 1589 patients. Patients were primarily male (65%) and Caucasian (93%), and had a median age of 47 years (range: 18.1–75.8 years). Most kidneys were obtained from deceased donors (64%; 22% with expanded donation criteria); 29% and 6% of donors were living related and unrelated, respectively. Treatment groups were well balanced with respect to demographics, cause of end-stage renal disease, donor characteristics and transplantation risk profile (9).
A total of 958 patients (60% of the core ITT population) from 67 centers (79% of the centers of the core study) in 14 countries participated in the 2-year extended follow-up and provided valuable data for the main analysis—these patients constituted the ITT and safety study populations (Table 1). Considering only patients who were potentially eligible for participation (i.e. ITT patients from centers contributing patients to the follow-up and who completed the year 1 visit), the inclusion rate in the follow-up was 77% (73% for low-dose sirolimus and 78% for the other groups). At 3 years postrandomization, serum creatinine data from 710 patients (74% of the included patients, 45% of the core ITT population) were available. A total of 56% of the follow-up ITT patients were part of the on-treatment analysis population (65% of the low-dose tacrolimus and 40% of the low-dose sirolimus patients) (Table 1). The countries contributing the largest number of patients were Spain (218, or 21% of the total), followed by Turkey (14%), Brazil (13%) and Germany (11%).
Table 1. Patient populations and characteristics
aTwo patients in the follow-up safety population were assigned to ‘Group 0’ of the safety population of the core study. This group included patients who did not receive the randomized treatment in the first 3 days and could not be assigned to any of the other groups.
bEligible patients were ITT patients from centers participating in the extended follow-up and who had completed the 1 year follow-up visit.
cPatients on treatment as randomized including mycophenolate mofetil and cyclosporine, tacrolimus or sirolimus and providing a serum creatinine value at 3 years.
Core study ITT population
Follow-up safety population
956 + 2a
Follow-up ITT population
% of core study ITT population
% of eligibleb ITT patients
Patients with serum creatinine data at 3 years
% of core study ITT
% of follow-up ITT
On-treatment population at 3 yearsc
% of core study ITT
% of follow-up ITT
Baseline characteristics of follow-up ITT population
Recipient age (years)
Type of donor (%)
Donor age (years)
Expanded donor criteria for deceased donors (%)
Baseline characteristics of follow-up patients (Table 1) matched well with those of the ITT population of the core study and there was no noteworthy imbalance among the treatment groups.
To examine the presence of a patient selection bias leading to the inclusion or exclusion of patients in relation to their renal function at the end of the core study, we compared patients who were included in the follow-up to those who were excluded although potentially eligible. Included patients had a GFR of 66 ± 22 mL/min, which was better than that in the remaining potentially eligible but not included patients (60 ± 24 mL/min, p < 0.0001) and in the remaining eligible patients when only centers participating in the follow-up were considered (62 ± 23 mL/min, p = 0.0019). This indicates a selection bias favoring the enrollment of patients with better outcomes in the follow-up. This bias was more marked in the low-dose sirolimus and the standard-dose CsA groups (difference between patients included and not included in follow-up: 8 and 7 mL/min, respectively) than that in the low-dose CsA and low-dose tacrolimus groups (5 and 4 mL/min, respectively; p < 0.0001 in an overall comparison of the four groups).
During the follow-up, 230 patients (24% of the included patients) prematurely withdrew from the study and further 10 patients did not provide a serum creatinine sample at month 36. Of the withdrawals, 8% were due to graft loss, 5% due to death, 19% due to ‘failure to return’, 7% due to ‘consent withdrawal’ and for 61%, the reason was indicated as ‘other’.
At inclusion into follow-up and at 3 years, 47% and 42% of patients, respectively, received a regimen containing CsA (at standard or low dose), 34% and 36% tacrolimus, and 17% and 17% sirolimus (Figure 1). Many follow-up patients had been switched to tacrolimus in the first year, including 24% of patients randomized to sirolimus. In this way, the relative proportion of patients treated with CsA, tacrolimus and sirolimus or none of these drugs in the patient population shifted from 50%:25%:25%:0% at randomization to 47%:34%:17%:2% at year 1 and 42%:36%:17%:5% at year 3.
The proportion of patients who concluded the study and were treated with a triple regimen of MMF, steroids plus either CsA, tacrolimus or sirolimus varied between 62% and 69% in the four groups (including patients who changed the immunosuppressive drug assigned at randomization). Steroid withdrawal was permitted by the protocol and at 3 years, 28% of patients were steroid free. The median daily steroid dose in treated patients was 5 mg prednisone equivalents in all groups; the average was 5.9 ± 3.5 mg and varied between 6.3 ± 5.0 mg in the standard-dose CsA and 5.4 ± 3.0 mg in the low-dose tacrolimus arm (Table 2). Most patients were maintained on MMF: 95% of the patients who concluded the observational follow-up. These patients received on average 1518 ± 615 mg/day. The MMF dose ranged from 1709 ± 467 mg/day in the low-dose CsA group to 1371 ± 699 mg/day in the low-dose sirolimus group. At the end of the study, the average CsA trough level was 114 ± 56 ng/mL and 103 ± 71 ng/mL in the standard and low-dose CsA groups, respectively. In low-dose tacrolimus and low-dose sirolimus patients, the levels of the respective drugs at year 3 were 6.5 ± 2.3 ng/mL and 7.0 ± 3.2 ng/mL.
Table 2. Exposure to study-specific drugs in treated patientsa in the follow-up safety population
aPatients not receiving the specific drug are excluded from the analysis.
MMF dose [mg/day]− mean ± SD
1755 ± 520
1798 ± 514
1531 ± 618
1556 ± 647
1662 ± 588
1623 ± 597
1709 ± 467
1353 ± 611
1371 ± 699
1518 ± 615
Trough level of CsA/tacrolimus/sirolimus (ng/mL) only in the respective group − mean ± SD
142 ± 64
101 ± 71
6.4 ± 2.4
7.5 ± 3.1
114 ± 56
103 ± 71
6.5 ± 2.3
7.0 ± 3.2
Corticosteroids dose (mg/day, prednisone equivalents) − mean ±SD
7.4 ± 5.3
6.9 ± 2.9
6.8 ± 3.7
7.5 ± 3.7
7.1 ± 4.3
6.3 ± 5.0
6.1 ± 3.0
5.4 ± 3.0
5.8 ± 2.9
5.9 ± 3.5
Overall, renal function was stable. The average GFR change over the second and third year was between +1 and −3 mL/min in the four arms. Low-dose tacrolimus still had a slightly higher GFR than the other arms (69 vs. 64–66 mL/min; overall p = 0.17, pairwise comparison against low-dose CsA: 0.0386; Table 3). The renal function over time (Figure 2) showed consistently superior GFR for low-dose tacrolimus-treated patients at all assessment time points, but with a difference compared to the other groups that was reduced by time after transplantation. The analysis based on the on-treatment principle indicated that in the fraction of patients who could be kept on the originally assigned drug combination, the GFR differences among treatment arms were small and nonsignificant (p = 0.17). In this analysis (Table 3), the low-dose sirolimus group had the best GFR (71 mL/min) followed by low-dose tacrolimus (70 mL/min).
Table 3. Main efficacy endpoints
CG = Cockcroft–Gault; MDRD = Modification of Diet in Renal Disease; SD = standard deviation. BPAR = biopsy-proven acute rejection.
aIn the ‘all patients’ column, p-values refer to a global test of the overall hypothesis of no difference among the four groups. p-values are from the Kruskal-Wallis and Wilcoxon rank-sum test (comparisons among four groups and between two groups, respectively) for GFR endpoints and the log-rank test for all other endpoints.
bValues are computed taking pair-wise differences, i.e. only considering patients with values at both Month 12 and Month 36.
cPatients on treatment as randomized including mycophenolate mofetil and cyclosporine, tacrolimus or sirolimus, and providing a serum creatinine at 3 years.
dp-values are from log-rank tests in an analysis including the period from transplantation to year 3.
Month 12: Mean CG-GFR (±SD) (mL/min) without imputation
64.7 ± 19.5
64.6 ± 20.9
70.3 ± 23.1
64.2 ± 23.1
66.0 ± 21.9
Month 24: Mean CG-GFR (±SD) (mL/min) without imputation
65.5 ± 22.9
63.6 ± 23.0
69.7 ± 24.2
66.1 ± 27.3
66.2 ± 24.4
Month 36: Mean CG GFR (±SD) (mL/min) without imputation
65.9 ± 26.2
64.0 ± 23.1
68.6 ± 23.8
65.3 ± 26.2
66.0 ± 24.8
p value versus low-dose tacrolimus
Mean CG-GFR change between Months 12 and 36 (mL/min)b
1.2 ± 16.9
−2.5 ± 14.6
−1.9 ± 13.8
1.1 ± 13.6
−0.6 ± 14.8
Month 36: Mean CG-GFR (±SD) (mL/min)] with imputation
65.3 ± 26.8
61.7 ± 24.7
67.6 ± 24.5
63.5 ± 27.7
64.5 ± 26.0
Month 36: mean Calculated GFR based on abbreviated MDRD (±SD) (mL/min) without imputation
54.8 ± 22.3
53.6 ± 18.2
57.5 ± 19.1
54.6 ± 21.1
55.1 ± 20.2
Month 12: Mean CG-GFR (±SD) [mL/min], on-treatment populationc
65.7 ± 19.2
67.9 ± 19.9
71.1 ± 22.7
69.8 ± 22.4
68.7 ± 21.1
Month 36: Mean CG-GFR (±SD) (mL/min), on-treatment population
67.1 ± 26.5
65.6 ± 22.7
69.6 ± 23.7
71.1 ± 25.4
68.1 ± 24.5
BPAR (excluding borderline) at 12 months (% of pts)
BPAR (excluding borderline) at 24 months (% of pts)
BPAR (excluding borderline) at 36 months (% of pts)
P value versus low-dose tacrolimusd
BPAR: increase between Month 12 and Month 36
Death-censored graft survival at Month 12 (%)
Death-censored graft survival at Month 24 (%)
Death-censored graft survival at Month 36 (%)
P value versus low-dose tacrolimusd
Uncensored graft survival at Month 12 (%)
Uncensored graft survival at Month 24 (%)
Uncensored graft survival at Month 36 (%)
P value versus low-dose tacrolimusd
Patient survival at Month 12 (%)
Patient survival at Month 24 (%)
Patient survival at Month 36 (%)
Over the second and third year, all groups had a low incidence of BPAR (1–3%). Because there was already a large between-group difference at 1 year, low-dose tacrolimus remained clearly superior at the end of the follow-up (14% vs. 27–39% in the other arms in Kaplan–Meier estimates, p < 0.001 in an overall log-rank test).
Overall, graft loss was low. At 3 years, death-censored graft survival in the ITT population was better in the low-dose tacrolimus group (93% in Kaplan–Meier estimates, Table 3 and Figure 3) compared to the other groups (89–91%), but the overall difference was not significant (p = 0.11). However, in pairwise comparisons, low-dose tacrolimus was superior to low-dose sirolimus (p = 0.019) and also to all other patients together (p = 0.03).
Patient survival at 36 months was 95% in the whole patient population and between 94% and 95% in the four groups (Table 3). The incidence of uncensored graft loss, including patient death, over the second and third year was low (approximately 2% per year), ranging between 3% and 6% in the four treatment groups. Deaths contributed to roughly half of all graft losses. Uncensored graft survival at 3 years was 88% in the ITT population and varied between 90% in the low-dose tacrolimus and 85% in the low-dose sirolimus groups (p = 0.035 for the comparison of these two groups but p = 0.13 in an overall comparison of the four groups).
Logistic regression identified renal function at 1 year as the only variable significantly associated with graft loss (excluding patient death) between years 1 and 3 (odds ratio: 2.1 for a 10 mL/min GFR decrease, p < 0.0001), in a model including in addition treatment group, patient and recipient age, gender, baseline diabetes and acute rejection in the first year (p > 0.1 for all these factors).
During the follow-up phase, adverse events were experienced by 73% of the safety population and between 71% and 76% of patients in the four groups (crude incidence rates, Table 4). The most frequently affected System Organ Class (SOC) was ‘Infections and Infestations’ (40% of the safety population, between 45% in low-dose sirolimus and 35% in low-dose CsA patients) and the most common AE was urinary tract infection reported by 11% of patients. In addition, opportunistic infections were reported by 3% of patients. The next most frequently affected SOC was ‘Gastrointestinal Disorders’ (21%, between 24% in low-dose tacrolimus and 7% in low-dose sirolimus groups, respectively) with diarrhea, experienced by 9% of patients, being the most common event. Next, 18% of patients experienced ‘Metabolism and Nutrition Disorders’ and 4% had a report of hyperlipidemia (crude incidence rates). New-onset diabetes after transplantation (NODAT) occurred in less than 1% of patients.
Table 4. Main safety endpoints
AE = adverse event; SAE = serious adverse event.
aThe additional group of two patients assigned to a separate safety group for analysis in the core study is not displayed in this table. Therefore, the number of events under ‘All groups’ can be larger than the sum over the four groups.
bPatients experiencing events in more than one sub-category are counted once here.
cp-values are from log-rank tests in an analysis including the period from transplantation to year 3.
Number of patients experiencing at least one event during the follow-up (years 2 and 3, crude incidence rates)
Peripheral vascular events
Kaplan–Meier incidence rates of selected AEs at 3 years
Increment years 1–3
Increment years 1–3
New-onset diabetes after transplantation
Increment years 1–3
Increment years 1–3
Increment years 1–3
Increment years 1–3
Serious AEs were reported by 28% of patients, with the highest incidence (32%) occurring in the low-dose sirolimus group (Table 4). ‘Infections and Infestations’ were the most common events (affecting 11% of all patients, and between 9% in low-dose CsA and low-dose tacrolimus and 17% in low-dose sirolimus). In addition, 1% of patients experienced serious opportunistic infections. Urinary tract infections and gastroenteritis were the most common serious infections. Other SOCs were affected in 3% of patients or less.
Although data on nonfatal significant cardiovascular events (cardiac, cerebrovascular, peripheral vascular events) were specifically solicited during the follow-up period (with a questionnaire to be filled out at each visit), their incidence was low, in particular in the low-dose tacrolimus group (2% vs. 5–6%, Table 4).
At 3 years, proteinuria more frequently affected low-dose sirolimus (28%) and standard-dose CsA patients (24%) than low-dose tacrolimus or low-dose CsA patients (19% for both arms). The differences among the arms were already present at entry into the follow-up, and the incidence of proteinuria in the follow-up period was 15–16% for all arms (Table 4).
In line with other safety parameters, the incidence of adverse events of special interest such as infections, NODAT, diarrhea, hyperlipidemia and anemia was lower in the follow-up than that in the core study (Kaplan–Meier incidence rates; Table 4) and the differences among the study groups observed at 1 year remained approximately stable.
The number of deaths in patients included in the follow-up was 12 (1.3% of follow-up patients). There were seven deaths due to cardiovascular events (three cerebrovascular accidents, two myocardial infarctions and two cardiac arrests), two due to fatal malignancies (one adenocarcinoma of the lung and liver, and one nonsmall cell lung cancer), and one each due to sepsis, bowel infarction and an unknown cause.
The Symphony study was the largest ever prospective study in de novo solid organ transplant patients, designed to provide robust data on the relative benefits of the four regimens tested. In the 1-year Symphony core study, the regimen including daclizumab induction, 2 g/day MMF, low-dose tacrolimus and steroids resulted in better renal function, less acute rejections and less graft losses compared to the other regimens. Possibly due, in part, to insufficient target exposure to sirolimus, the CNI-free arm had the worst outcomes (9).
In the second and third posttransplant year, we observed only small changes in the main efficacy parameters, with mean GFR decline, annual incidence of graft loss and death, all at about 1%. The yearly BPAR rate during the follow-up was also approximately 1%. The graft survival results are equivalent or better compared with other de novo studies with a similar follow-up (12–16), or with registry data (4).
Low-dose tacrolimus continued to be the group with the best outcomes, mainly due to differences developing over the first 3–6 months after transplantation and remaining approximately stable thereafter. Looking in more detail at between-group differences, we observed slightly smaller treatment effects, i.e. reduced differences among the original treatment groups, for most safety and efficacy measures between the first and the third year posttransplant. The differences among treatment groups were often no longer significant at 3 years. However, the follow-up study was not powered to address differences in the endpoints at 3 years: for instance the statistical power to detect the observed GFR differences with the sample size at 3 years would be approximately 35%. It is very likely that the changes in treatment regimens that occurred over the entire 3 years contributed to the blurring of treatment effects. A very common treatment change was a switch from sirolimus to tacrolimus following a rejection episode, mostly during the first year. The impact of the various and unsystematic treatment modifications occurred during the study period is difficult to tackle from the analytical point of view and our main analysis, being based on the ITT principle, did not address this issue explicitly. It should be noted that the small fraction of patients who could be maintained on MMF and sirolimus for the entire study period (i.e. excluding patients who lost their graft or had a regimen change for safety or efficacy reasons) had numerically slightly better average renal function values at 3 years and a more favorable development over the follow-up phase than patients remaining on other combinations. A concurrent explanation for the decreased between-group differences could be the preferential inclusion of patients with good prognosis in the follow-up: an indication for the presence of a certain selection bias is that the estimated GFR was better in patients included in the follow-up than in those not included, in particular affecting, that is improving, the GFR at 3 years in the two groups with worse renal function at 1 year.
In the literature (17–19), lower renal function at 1 year is a risk factor associated with graft loss, although with a limited predictive value (20). Our data here indicate that renal function at 1 year was the only variable significantly associated with graft loss. This corroborates the validity of renal function as the primary endpoint of the Symphony study (9). On the other hand, in relation with the aforementioned patient selection bias, the impact of renal function on graft survival suggests that our figures may slightly underestimate the graft loss rates in the original study population.
As for the main efficacy parameters, the incidence of the main adverse events was low in the second and third year and the overall safety profile of the treatment groups was compatible with that in the core study, but with smaller and often nonsignificant differences among the groups. Cardiovascular events were rather infrequent in the second and third posttransplant year but, in line with the fact that cardiovascular events are the leading cause of death with a functioning graft, they replaced infections as the primary cause of death. NODAT was also reported only in a handful of patients during the follow-up, despite the fact that patients were exposed for another 2 years compared to 1 year in the core study. Given the fact that most NODAT patients did not require long-term antidiabetic medication (9), new-onset diabetes does not seem to occur at a clinically worrisome rate in the study. Our figures are similar or lower than those in other reports with a similar follow-up (12,14,15). However, a limitation of the Symphony study is that diabetes mellitus was assessed based on AE reports and neither explicitly solicited in the case report form nor diagnosed according to American Diabetes Association/World Health Organization criteria, and therefore underreporting may have occurred.
There are methodological limitations that cannot be ignored when interpreting our results. First of all, there is the fact that only approximately half of the randomized patients concluded the observational follow-up and, in relation to this, the evidence pointing to a preferential selection of better-performing patients at inclusion into the follow-up phase. Second is the observational design of the follow-up study with uncontrolled treatment modifications and the limited power of the statistical comparisons. Also, the Symphony study included only low-to-normal risk patients. Therefore, extrapolation of our results to patient populations not included in this study, in particular African Americans and sensitized patients, should be done only with caution and more studies are needed to validate these results in broader or high-risk patient populations.
In conclusion, follow-up results based on approximately half of the core Symphony study population indicate that during the second and third year, renal function remains stable and the incidence of acute rejection and graft loss is low. Partly explained by substantial transitions from one treatment to another and by the preferential inclusion of better-performing patients, the differences among the original ITT treatment groups in terms of graft loss and renal function were smaller than those at 1 year and did not reach statistical significance. However, in line with the 1-year results, at 3 years, the group of patients initially assigned to a regimen with daclizumab induction, 2 g/day MMF, low-dose tacrolimus and steroids had the best renal function and graft loss rate. The results of the Symphony study at 3 years after transplantation support the use of this immunosuppressive regimen in de novo kidney transplantation.
We thank Christa Silberbauer, the EliTE–Symphony project leader, for her great contributions to the study, and Peter Wijngaard, Carolyn Sutter, Jana Nöldeke and Marie-Claire Eger for their support (all from F. Hoffmann-La Roche). We also thank Anne-Marie Stephani (Wolters Kluwer Health) for editorial assistance under the guidance of Henrik Ekberg. Thanks are extended to Sir Peter Morris, Daniel Abramowicz and Gerhard Opelz (Chair) of the Data Safety and Monitoring Board. We also thank the study investigators and coordinators from the 15 participating countries for all their hard work in making this study possible. Funding for this study was provided by F. Hoffmann-La Roche Ltd, Basel, Switzerland.
Investigators who participated in the ELiTE-Symphony study 3-year follow-up are as follows:
Australia: R. Walker, G. Russ, J. Eris, P. O'Connell; Austria: F. Mühlbacher, R. Margreiter; Belgium: Y. Vanrenterghem, P. Peeters; Brazil: D. Carvalho, R. Esmeraldo, H. Tedesco-Silva, L. Saber, G. Majella, G. Alves Filho, R. Gonçalves; Canada: S. Gourishankar, P. Daloze; Czech Republic: V. Treska, J. Dedochová, M. Kuman, P. Navrátil, S. Vítko; Germany: U. Frei, M.W. Büchler, J. Steinhoff, H-H. Neumayer, E. Nagel, C. Hugo, F. Eitner, R. Viebahn, J. Klempnauer, U. Ott; Israel: M. Shabtai, A. Yussim, R. Nakache; Mexico: R. Reyes Acevedo, J.M. Barron Sixth; Poland: M. Durlik, M. Klinger; Spain: J. Grinyó, M. Arias, J.M. Morales, F. Oppenheimer, J. Sanchez Plumed, F. Valdes, M. Angel Gentil, A. Osuna, D. Hernandez, M. Tabernero, F. Ortega, M. Rivero Sanchez, R. Marcen, M. Gonzalez Molina, R. Lauzurica; Sweden: H. Ekberg, L. Mjörnstedt; Turkey: E. Ok, E. Akoglu, A. Gurkan, A. Demirbas, M. Sukru Sever, U. Erken, B. Erbay, K. Dilek, I. Titiz; UK: M. Nicholson.