Conclusions: Evidence for induction therapy for lupus nephritis is inconclusive based on treatment effects on all-cause mortality, doubling of serum creatinine level, and end-stage kidne
Trang 1Induction and Maintenance Immunosuppression Treatment
of Proliferative Lupus Nephritis: A Network Meta-analysis
of Randomized Trials
Suetonia C Palmer, MBChB, PhD,1,*David J Tunnicliffe, MIPH,2,3,* Davinder Singh-Grewal, MBBS, PhD,4,5Dimitris Mavridis, PhD,6,7 Marcello Tonelli, MD,8David W Johnson, MBBS (Hons), PhD,9,10 Jonathan C Craig, MBChB, PhD,2,3Allison Tong, PhD,2,3and
Giovanni F.M Strippoli, MD, PhD2,11,12
Background: Intravenous (IV) cyclophosphamide has been first-line treatment for inducing disease
remission in lupus nephritis The comparative efficacy and toxicity of newer agents such as mycophenolate
mofetil (MMF) and calcineurin inhibitors are uncertain.
Study Design: Network meta-analysis.
Setting & Population: Patients with proliferative lupus nephritis.
Selection Criteria for Studies: Randomized trials of immunosuppression to induce or maintain disease
remission.
Interventions: IV cyclophosphamide, oral cyclophosphamide, MMF, calcineurin inhibitor, plasma
ex-change, rituximab, or azathioprine, alone or in combination.
Outcomes: Complete remission, end-stage kidney disease, all-cause mortality, doubling of serum
creatinine level, relapse, and adverse events.
Results: 53 studies involving 4,222 participants were eligible Induction and maintenance treatments were
administered for 12 (IQR, 6-84) and 25 (IQR, 12-48) months, respectively There was no evidence of different
effects between therapies on all-cause mortality, doubling of serum creatinine level, or end-stage kidney
disease Compared to IV cyclophosphamide, the most effective treatments to induce remission in
moderate-to high-quality evidence were combined MMF and calcineurin inhibimoderate-tor therapy, calcineurin inhibimoderate-tors, and
MMF (ORs were 2.69 [95% CI, 1.74-4.16], 1.86 [95% CI, 1.05-3.30], and 1.54 [95% CI, 1.04-2.30],
respectively) MMF was significantly less likely than IV cyclophosphamide to cause alopecia (OR, 0.21; 95%
CI, 0.12-0.36), and MMF combined with calcineurin inhibitor therapy was less likely to cause ovarian failure
(OR, 0.25; 95% CI, 0.07-0.93) Regimens generally had similar odds of major infection MMF was the most
effective strategy to maintain remission.
Limitations: Outcome definitions not standardized, short duration of follow-up, and possible confounding by
previous or subsequent therapy.
Conclusions: Evidence for induction therapy for lupus nephritis is inconclusive based on treatment effects on
all-cause mortality, doubling of serum creatinine level, and end-stage kidney disease MMF, calcineurin
inhibitors, or their combination were most effective for inducing remission compared to IV cyclophosphamide,
while conferring similar or lower treatment toxicity MMF was the most effective maintenance therapy.
Am J Kidney Dis - ( - ): - - - ª 2017 The Authors Published by Elsevier Inc on behalf of the National Kidney
Foundation, Inc This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/
licenses/by-nc-nd/4.0/ ).
INDEX WORDS: Lupus nephritis; immunosuppression; remission; induction therapy; maintenance therapy;
mycophenolate mofetil (MMF); intravenous cyclophosphamide; calcineurin inhibitor; end-stage kidney
disease (ESKD); dialysis; renal failure; toxicity; adverse events; meta-analysis.
From the1Department of Medicine, University of Otago
Christ-church, ChristChrist-church, New Zealand; 2Sydney School of Public
Health, University of Sydney, Sydney, NSW, Australia;3Centre for
Kidney Research, The Children ’s Hospital at Westmead, Westmead,
NSW, Australia;4School of Paediatrics and Child Health, The
Uni-versity of Sydney, Sydney, NSW, Australia;5Department of
Rheu-matology, The Sydney Children ’s Hospitals Network Westmead and
Randwick, NSW, Australia;6Department of Hygiene and
Epidemi-ology, University of Ioannina School of Medicine, Ioannina, Greece;
7 Department of Primary Education, University of Ioannina, Ioannina,
Greece;8Cumming School of Medicine, University of Calgary,
Cal-gary, Canada;9Division of Medicine, Department of Nephrology,
University of Queensland at the Princess Alexandra Hospital,
Bris-bane, QLD, Australia;10Translational Research Institute, Brisbane,
QLD, Australia; 11 Diaverum Medical Scienti fic Office and Diaverum
Academy, Lund, Sweden; and12Department of Emergency and Organ Transplantation, University of Bari, Bari, Italy.
* S.C.P and D.J.T contributed equally to this work.
Received May 16, 2016 Accepted in revised form December 1, 2016.
Address correspondence to Giovanni Strippoli, MD, PhD, Department of Emergency and Organ Transplantation, University
of Bari, Piazza Giulio Cesare, 70124 Bari, Italy E-mail:
gfmstrippoli@gmail.com
2017 The Authors Published by Elsevier Inc on behalf of the National Kidney Foundation, Inc This is an open access article under the CC BY-NC-ND license ( http://creativecommons.org/ licenses/by-nc-nd/4.0/ ).
0272-6386
http://dx.doi.org/10.1053/j.ajkd.2016.12.008
Trang 2Systemic lupus erythematosus principally affects
women of child-bearing age Kidney
involve-ment affects 20% to 75% of patients in the first 10
years.1 Although 5-year survival for patients with
systemic lupus erythematosus was ,50% in the
1950s, this has improved to 90%, attributed to
improved immunosuppression and other medical
therapies Therapies have transformed lupus nephritis
from an acute to a chronic illness, in which the longer
term efficacy and adverse effects of treatments may
assume greater importance in medical decision
making
Intravenous (IV) cyclophosphamide combined with
corticosteroids has been first-line therapy to induce
remission from lupus nephritis, but it causes
consid-erable toxicity.2 Existing pairwise meta-analyses
suggest similar efficacy for mycophenolate mofetil
(MMF) and IV cyclophosphamide with lower toxicity
for MMF, but whether MMF or other drugs are
equivalent or superior to IV cyclophosphamide for
induction and maintenance of disease remission is
uncertain.3,4 However, standard pairwise
meta-analysis is only able to compare 2 drug classes that
have already been evaluated in head-to-head trials In
a complex condition with several options for
treat-ment, of which some have not been directly compared
in trials, a network meta-analysis offers the potential
to compare all therapeutic strategies simultaneously
within a single framework and rank treatments per
efficacy and safety Network analysis has been used
to evaluate induction therapy in lupus nephritis, but
results have been inconclusive due to relatively few
included studies5,6or reporting of drug harms only.7
METHODS
Overview
A network meta-analysis was performed within a frequentist
framework The meta-analysis was conducted and reported
ac-cording to a prespeci fied protocol ( Item S1 , available as online
supplementary material) and the PRISMA (Preferred Reporting
Items for Systematic Reviews and Meta-analyses) statement.8
Ethics committee approval was not required for this study design.
Data Sources and Searches
(CENTRAL), MEDLINE, and Embase were searched on July 20,
2016, using a highly sensitive search strategy without language
restriction ( Item S1 ) A Cochrane review and meta-analysis was
also screened for eligible randomized trials.3
Study Selection
Parallel-group randomized trials involving adults, adolescents,
or children 10 years or older with proliferative lupus nephritis and
who received immunosuppression to induce or maintain remission
were included Included trials reported comparisons between 2
immunosuppression strategies, placebo, or usual care Two
re-viewers (S.C.P and D.J.T.) independently screened titles and
ab-stracts of retrieved search records to determine potential eligibility.
Any potentially eligible citation was reviewed in full text by the
same 2 reviewers, who resolved discrepancies through consensus Potentially eligible articles published in languages other than English were translated before full-text assessment.
Data Extraction
Two investigators (S.C.P and D.J.T.) abstracted data indepen-dently into an electronic database The authors cross-checked the data and reached consensus for any discrepancies through discussion.
Risk of Bias
Two independent reviewers (S.C.P and D.J.T.) assessed risks of bias using the Cochrane Collaboration assessment tool.9
Data Synthesis and Analysis
The primary outcomes of interest for induction therapy were complete remission and all-cause mortality Other outcomes were end-stage kidney disease, doubling of serum creatinine level, failure to induce remission, major infection, alopecia, ovarian failure, malignancy, nausea, vomiting, bone toxicity, bladder toxicity, leukopenia, and herpes infection In maintenance therapy trials, relapse after remission was the primary outcome Studies reporting zero events in all arms were excluded from analyses Data from trials principally evaluating induction treatment were analyzed separately from trials evaluating maintenance treatment The clinical setting and participant characteristics were evalu-ated to consider whether the trials were suf ficiently similar that a network meta-analysis approach was appropriate.10 Box plots were generated according to treatment class to explore distribu-tions of key effect modi fiers, including age, sex, serum creatinine level, and date of publication We intended to explore distributions
of treatment classes by ethnicity or race, but these assessments were precluded by insuf ficient data observations.
Random-effects pairwise meta-analysis was then conducted Heterogeneity of treatment estimates between trials in pairwise meta-analysis was assessed usingc2 test and the corresponding I2 statistic I2thresholds of 0% to 40%, 30% to 60%, 50% to 90%, and 75% to 100% were considered to represent heterogeneity that might not be important, that is moderate, that is substantial, and that is considerable, respectively, considering also the magnitude and direction of treatment effects.11
Finally, using a frequentist framework, random-effects network meta-analysis was used to compare all classes of immunosup-pression for each prespeci fied outcome 10,12 We assumed a random-effects model to describe the effects of the base treatment
in each study in each network, with the conventional assumption
of a normal distribution for random effects Comparative treatment effects were calculated as odds ratios (ORs) and 95% con fidence intervals (CIs) The extent of heterogeneity in each network analysis was evaluated using the restricted maximum likelihood method to generate a common heterogeneity variance (tau [s]), which was then compared with an empirical distribution of het-erogeneity variances, considering the range of ORs expected Values of 0.1 to 0.5 were considered low, those 0.5 to 1.0 were considered fairly high, and those 1.0 represented fairly extreme heterogeneity.13To explore for network inconsistency, a loop-speci fic approach was used that compares the estimated treat-ment effects derived from direct and indirect evidence in all triangular and quadratic loops in a network To check the assumption of consistency in the entire analytical network, the design-by-treatment interaction approach was used.14
Drug classes were ranked to generate a hierarchy of treatments for a given clinical end point The relative ranking probability of each treatment being among the “best” treatment was obtained using surface under the cumulative ranking (SUCRA) curves and displayed using rankograms The GRADE (Grading of Recom-mendations Assessment, Development and Evaluation) approach
Trang 3to assess con fidence in estimates of effect associated with specific
drug comparisons in network analyses was used to determine
con fidence in the evidence for all-cause mortality, end-stage
kid-ney disease, and complete remission.15We considered risk of bias,
consistency, imprecision, indirectness, and publication bias
Pair-wise and network meta-analyses were done in Stata, version 13
self-programmed Stata routines.17 We planned metaregression
ana-lyses to explore associations of year of publication, race or
ethnicity, outcome de finition, or age with treatment estimates.
RESULTS
Description of Included Studies
Thirty-eight trials were included from a Cochrane
review3and 15 additional trials were identified through
electronic database searching (Fig 1) Fifty-three
ran-domized trials were eligible involving 4,222
partici-pants (aged $ 10 years; Table S1).18-70 Participants
with active nephritis were randomly allocated to
ther-apy to induce remission in 45 trials (n5 3,623),
whereas participants who had previously achieved
disease remission were randomly allocated to
thera-peutic strategies to maintain remission in 8 trials
(n5 599;Table S2) Of the included trials, 2 reported
outcomes in participants randomly allocated to
in-duction and then subsequently to maintenance therapy
in separate trials within a single cohort.21,22,32,33
The median number of study participants was 47 (range, 6-378), while the mean age of study partici-pants was 30.2 6 4.9 (standard deviation; range, 10.2-40.3) years Induction treatment was continued for a median follow-up of 12 (range, 5.5-84) months, while the median duration of follow-up of mainte-nance therapy was 24 (range, 6-110) months
In the early trials (1972-1984), interventions were oral azathioprine, oral cyclophosphamide, prednisone alone, or plasma exchange The first trial of IV cyclophosphamide was reported in 1986, and trials evaluating calcineurin inhibitors emerged in 1992 The earliest study assessing MMF appeared in 2005, and rituximab has been evaluated in trials since 2009 From 2012 onward, a range of other immunomodu-latory drugs including atacicept, abatacept, laquini-mod, sirukumab, and mizoribine have been evaluated
as induction therapies Twenty-one induction therapy trials (905 participants) evaluated IV cyclophospha-mide (500-1,000 mg/m2 body surface area monthly),19,21,23,25,29,32,37,42,43,48-50,52,56,57,63-67,69 10 trials (516 participants) evaluated MMF (2,000-3,000 mg daily),21,31,42,43,47,56,61,63,65,66 6 trials (123 participants) evaluated oral cyclophosphamide
(1.5-4 mg/kg daily),23,31,39,41,44,58 8 trials (231 partici-pants) evaluated calcineurin inhibitors (cyclosporine,
Unique records identified through updated database searching of MEDLINE, Embase and Cochrane databases through July 20, 2016
(n = 416)
Records identified through existing Cochrane review current to April 2012 (n=50 studies in 168 publications)
Records excluded as ineligible for network meta-analysis (n=22) Not comparing two different drug classes (n=20) Full study publication identified in search update (n=2)
Duplicate records from previous Cochrane review excluded (n = 5)
Records evaluated in title and abstract (n = 411)
Records excluded on title and abstract (n=360) Not people with lupus nephritis (n=226) Not immunosuppression therapy (n=24) Not randomized controlled trial (n=110)
Records evaluated in full text detail
(n = 51)
Records excluded on full text analysis (n=11)
Not people with lupus nephritis (n=1)
Not immunosuppression therapy (n=2)
Not randomized controlled trial (n=6)
Ongoing study (n=2)
Randomized trials included in updated review (n=15 studies in 40 publications)
Randomized trials obtained from Cochrane review (n=38 studies in 146 publications)
53 unique randomized trials in 186 publications included in network meta-analysis (involving 4,222 adults, adolescents, and children)
Induction therapy n=45 studies (3623 adults and children)
Maintenance therapy n=8 studies (599 adults and children)
Figure 1 Summary of evidence search and selection.
Trang 41-5 mg/kg, daily or tacrolimus, 0.05-0.1 mg/kg,
daily),19,24,32,37,52,56,58,61 2 trials (201 participants)
evaluated MMF (1,000 mg daily) combined with
tacrolimus (4-8 mg daily),25,57 6 trials (90
partici-pants) evaluated plasma exchange,34,35,40,54,64,70 1
trial (40 participants) evaluated MMF (1,000 mg
daily) combined with IV cyclophosphamide (400 mg/
m2/body surface area monthly),69 and 2 trials (81
participants) evaluated rituximab (commencing at
1,000 mg on days 1 and 15).55,59
Risks of Bias
Risks of bias are shown in Figs S1and S2
Spe-cifically, 15 (28%) studies were at low risk of bias in
generation of the random sequence, whereas 16 (30%)
studies were at low risk of bias for allocation
concealment Nine (17%) studies reported blinding of
participants and investigators to allocated treatment
and 5 (9%) reported blinded outcome assessment
Forty-five (85%) studies were at low risk of
incom-plete outcome reporting, whereas 31 (58%) were at
low risk of selective outcome reporting Overall, 36
(68%) studies were at low risk of other sources of
bias Risks of bias in treatment estimates for complete
remission, all-cause mortality, and end-stage kidney
disease proportional to contributions of head-to-head
trials to the network estimates are summarized in
Exploration of Network Structure, Heterogeneity, and Consistency
When participant characteristics (age, sex, kidney function, disease definition, and racial origin), in-terventions (doses and duration), and study design (duration of follow-up) in the included trials were eval-uated per treatment class, the trials were deemed suffi-ciently similar for the key interventions that a network analysis was reasonable (Fig S4) There was evidence of different publication eras for therapies: trials evaluating
IV cyclophosphamide, MMF, and calcineurin inhibitors tended to be published more recently than those evalu-ating corticosteroids alone, oral cyclophosphamide, and azathioprine Network heterogeneity was frequently very low (heterogeneitys, 0.001 for most networks) consistent with low statistical power in the networks to detect heterogeneity (figs a-p ofItem S2)
Pairwise and network meta-analysis estimates were similar in magnitude (Tables S4 and S5) and testing did not reveal evidence of inconsistency between direct and indirect treatment effects, although CIs were frequently wide (Table S6) There was no evidence of global inconsistency in any network (Table S7) Outcomes
Overview
The raw event data for each outcome per trial are shown in Table S8 Trial end points defined by
Table 1 Network Treatment Estimates for Efficacy of Induction Therapies for Disease Remission in Proliferative Lupus Nephritis
Treatment Strategy
Complete Remissiona
All-Cause Mortalityb ESKDb Doubling Scrb
Treatment Failureb
Calcineurin inhibitor 1.74 (1.09-2.79)c 0.83 (0.27-2.56) 2.08 (0.23-18.9) 3.26 (0.25-42.0) 0.28 (0.12-0.65)c
Oral cyclophosphamide 0.57 (0.23-1.40) 2.86 (0.82-10.0) 1.34 (0.31-5.88) 1.85 (0.48-7.22) 1.70 (0.24-12.5)
No of studies; no of participants
in network
Note: Treatment estimates are shown as odds ratio (95% confidence interval) derived from network meta-analysis for all treatments compared with IV cyclophosphamide (reference) Data are shown in order of efficacy surface under the cumulative ranking (SUCRA) curve for induction of complete disease remission.
Abbreviations: ESKD, end-stage kidney disease; IV, intravenous; MMF, mycophenolate mofetil; Scr, serum creatinine.
a
Odds ratio 1 favors active drug class.
b
Odds ratio , 1 favors active drug class.
There were insufficient observations to calculate estimated treatment effects from network analysis for rituximab for any efficacy outcome The heterogeneity tau ( s ) values in the network analyses were: complete remission, s , 0.001 (low heterogeneity); all-cause mortality, s 50.18 (low heterogeneity); end-stage kidney disease, s , 0.001 (low heterogeneity); doubling of Scr level, s , 0.001 (low heterogeneity); and treatment failure, s , 0.001 (low heterogeneity) Extreme low P values for network heterogeneity were consistent with low statistical power in the networks to detect heterogeneity.
c Statistically significant.
Trang 5investigators are described in Table S9 Network analysis results are summarized inTables 1and2and
mortality, complete remission, and end-stage kidney failure are provided in Table 3 Treatment networks for primary outcomes are shown in Fig 2, and for other outcomes are shown infigures a-c ofItem S3
Primary Outcomes
Disease remission Compared to IV cyclophos-phamide, the most effective treatments to induce remission in moderate- to high-quality evidence were MMF combined with a calcineurin inhibitor, calci-neurin inhibitor alone, or MMF alone (ORs of 2.69 [95% CI, 1.74-4.16], 1.74 [95% CI, 1.09-2.79], and 1.44 [95% CI, 1.00-2.06], respectively) The combi-nation of MMF plus calcineurin inhibitor ranked as the best treatment to induce remission (Fig 3) All-cause mortality There was no evidence of different effects between immunosuppression strate-gies on all-cause mortality in generally moderate- to low-quality evidence (Table 1), and treatment rank-ings were uncertain (Fig 3)
Secondary Outcomes
End-stage kidney disease and doubling of serum creatinine level There was no evidence that MMF or calcineurin inhibitors, alone or in combination, had different effects on end-stage kidney disease or doubling of serum creatinine level compared to IV cyclophosphamide or each other (Table 1;Item S2) in low-quality evidence Ranking of treatments by ef fi-cacy for kidney outcomes was not informative because drug strategies had similar ranking probabilities (Fig 3) Treatment failure Compared to IV cyclophos-phamide, calcineurin inhibitors and MMF had lower risks for treatment failure (ORs of 0.28 [95% CI, 0.12-0.65] and 0.51 [95% CI, 0.29-0.90], respec-tively) and were ranked the best treatments (Table 1) Adverse events Compared with calcineurin in-hibitors, MMF had higher odds of major infection (OR, 2.16; 95% CI, 1.05-4.44), although neither drug class had significantly different odds of major infec-tion compared to IV cyclophosphamide (Table 2) Compared to IV cyclophosphamide, MMF had lower odds of alopecia (OR, 0.21; 95% CI, 0.12-0.36), whereas the odds with calcineurin inhibitors were not significantly different; calcineurin inhibitors were ranked similarly to MMF
Compared to oral cyclophosphamide, IV cyclo-phosphamide, MMF, and calcineurin inhibitors conferred similarly lower odds of ovarian failure (ORs of 0.13 [95% CI, 0.01-1.42], 0.09 [95% CI, 0.01-0.99], and 0.04 [95% CI, 0.00-0.40], respec-tively), although the result for IV cyclophosphamide was not significant
a Stati
Trang 6Table 3 Summary of Confidence in Network Treatment Estimates for Complete Disease Remission, All-Cause Mortality, and End-Stage Kidney Disease Associated With Immunosuppression Treatment to Induce Disease Remission in Lupus Nephritis
Outcome & Treatment
Strategy
Confidence
in Evidence Reasons for Downgrading Confidence in Evidencea,b
Network Treatment Estimate vs
IV Cyclophosphamidec
Complete Remission
MMF 1 calcineurin inhibitor High
CCCC
CCCC
IV cyclophosphamide 1 MMF Low
CCBB
Downgrade 2 levels in confidence based on study limitations ( 21) and imprecision (21)
1.48 (0.62-3.53)
CCCB
Downgrade 1 level in confidence based on study limitations ( 21)
1.44 (1.00-2.06) d
CCBB
Downgrade 2 levels in confidence based on study limitations ( 21) and imprecision (21)
0.57 (0.23-1.40)
CCBB
Downgrade 2 levels in confidence based on study limitations ( 21) and imprecision (21)
0.57 (0.23-1.40)
CBBB
Downgrade 3 levels in confidence based on study limitations ( 22) and imprecision (21)
0.29 (0.08-1.11)
All-cause mortality
CCCB
Downgrade 1 level in confidence based on imprecision ( 21)
1.20 (0.59-2.44)
CCBB
Downgrade 2 levels in confidence based on study limitations ( 21) and imprecision (21)
2.86 (0.82-10.0)
CCCB
Downgrade 1 level in confidence based on imprecision ( 21)
0.83 (0.27-2.56)
CCBB
Downgrade 2 levels in confidence based on study limitations ( 21) and imprecision (21)
1.52 (0.52-4.46)
MMF 1 calcineurin inhibitor Moderate
CCCB
Downgrade 1 level in confidence based on imprecision ( 21)
1.00 (0.02-52.8)
IV cyclophosphamide 1 MMF Moderate
CCCB
Downgrade 1 level in confidence based on imprecision ( 21)
0.92 (0.06-15.3)
CBBB
Downgrade 3 levels in confidence based on study limitations ( 22) and imprecision
8.21 (0.22-304)
CCCB
Downgrade 1 level in confidence based on imprecision ( 21)
2.01 (0.69-5.86)
End-Stage Kidney Disease
CCBB
Downgrade 2 levels in confidence based on study limitations ( 21) and imprecision (21)
1.34 (0.31-5.88)
CBBB
Downgrade 3 levels in confidence based on study limitations ( 22) and imprecision (21)
1.79 (0.56-5.70)
CBBB
Downgrade 3 levels in confidence based on study limitations ( 22) and imprecision (21)
2.08 (0.23-18.9)
CCCC
CCBB
Downgrade 2 levels in confidence based on study limitations ( 21) and imprecision (21)
2.60 (0.36-18.7)
CCBB
Downgrade 2 levels in confidence based on study limitations ( 21) and imprecision (21)
2.92 (0.31-27.8)
Abbreviations: IV, intravenous; MMF, mycophenolate mofetil.
a The confidence in the evidence was adjudicated according to GRADE (Grading of Recommendations Assessment, Development and Evaluation) criteria 15 based on study limitations (risks of bias including sequence generation, allocation concealment, blinding, attrition from follow-up, selective reporting of outcomes, and other sources of bias), consistency in treatment effects between studies, directness of the evidence to likely clinical setting, evidence of small study effects (smaller studies with systematically different results from larger studies), and precision of the estimate (imprecision was considered to be present when the confidence interval favored either of the compared treatments) Study limitations were calculated as proportional to the risks of bias from the contributions of direct (head-to-head) evidence to each network estimate for complete remission, all-cause mortality, and end-stage kidney disease The risks of bias proportional to the contributions of head-to-head trials evidence in each network estimate are shown in Fig S3 b
High confidence in the overall evidence meant that additional studies were not likely to have an impact on our confidence in the treatment effects and moderate confidence in the overall evidence meant that additional studies were likely to have an important impact on our confidence in treatment effects and may change the estimate Low confidence means that additional studies are very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
c
Values given as odds ratio (95% confidence interval).
d
Statistically significant.
Trang 7There was no evidence that MMF, calcineurin in-hibitors, or IV cyclophosphamide had significantly different odds of leukopenia and herpes infection
cyclophospha-mide, MMF was more likely to cause diarrhea (OR, 2.70; 95% CI, 1.61-4.53) Compared to IV cyclo-phosphamide, MMF and calcineurin inhibitor therapy incurred lower odds of nausea (ORs of 0.21 [95% CI, 0.12-0.34] and 0.14 [95% CI, 0.02-0.92], respec-tively) There was no evidence of different odds of bladder toxicity between oral and IV cyclophospha-mide (OR, 0.11; 95% CI, 0.01-2.25])
Disease relapse When considering maintenance therapy to prevent disease relapse, MMF was superior
to azathioprine (OR, 0.57; 95% CI, 0.35-0.93), whereas the treatment effects of calcineurin inhibitors did not differ significantly from MMF (OR, 1.13; 95%
CI, 0.35-3.69) and were not significantly better than azathioprine (OR, 0.64; 95% CI, 0.22-1.88;Table 4;
therapy were not estimated due to the relatively few studies reporting these treatment strategies Sensitivity Analysis
Treatment effects were imprecise in sensitivity analyses restricted to trials with follow-up of 24 months or longer and in trials in which allocation was adequately concealed (Table S10) There were insufficient observations to perform reliable metaregression analyses accounting for year of pub-lication, race or ethnicity, outcome definition, or age
DISCUSSION
Considering benefits and harms of therapy, this network meta-analysis indicated that when added to corticosteroids, MMF or calcineurin inhibitors or their combination were superior to IV cyclophosphamide for inducing remission However, despite 53 trials, the longer term effects of immunosuppression on risks for mortality or end-stage kidney disease remain uncer-tain, in part due to the relative rarity of these events and the short duration of existing studies Compared
to IV cyclophosphamide, MMF incurred lower risks for alopecia, nausea, and vomiting, whereas calci-neurin inhibitors incurred lower risks for nausea Combined MMF and calcineurin inhibitor therapy had considerably lower average odds of ovarian fail-ure than IV cyclophosphamide, but this did not reach statistical significance for either therapy alone MMF was the best treatment to maintain disease remission and was superior to azathioprine
Based on these results and weighing the balance of benefits and harms, no treatment approach demon-strably lowers long-term risks for end-stage kidney failure or death Based on surrogate outcomes, MMF with or without calcineurin inhibitor treatment would
A Complete disease remission
B All-cause mortality
MMF
Oral cyclophosphamide Prednisone
Calcineurin
inhibitor
MMF +
calcineurin
inhibitor
Intravenous cyclophosphamide
Azathioprine
Intravenous cyclophosphamide
Oral cyclophosphamide MMF
Prednisone
Calcineurin
inhibitor
Oral
cyclophosphamide +
azathioprine
Azathioprine + prednisone
Oral cyclophosphamide +
prednisone
Mizoribine
Standard therapy
Azathioprine + heparin
IV cyclophosphamide + MMF
Oral cyclophosphamide + prednisone
Plasma
exchange
MMF +
calcineurin
inhibitor
IV cyclophosphamide + MMF
1/26
C End-stage kidney disease
Intravenous cyclophosphamide
Oral cyclophosphamide MMF
Plasma
exchange
Prednisone
Calcineurin
inhibitors
Oral
cyclophosphamide +
azathioprine
Azathioprine + heparin
Azathioprine + prednisone
Oral cyclophosphamide + prednisone Azathioprine
Standard therapy
2/91
2/75 2/104
Figure 2 Graphic representation of treatment comparisons for
efficacy and safety of induction immunosuppression treatment for
proliferative lupus nephritis Lines represent trials comparing 2
clas-ses of drug or drugs for (A) complete remission of lupus nephritis,
(B) all-cause mortality, and (C) end-stage kidney disease Numbers
on connecting lines represent the number of studies/number of
par-ticipants in trials directly comparing the two treatments The nodes
indicate the drug treatments assessed in existing trials The size of
the node is proportional to the number of studies evaluating the
treatment For example, the most commonly evaluated treatment
for complete remission of lupus nephritis is intravenous
cyclophos-phamide Abbreviation: MMF, mycophenolate mofetil.
Trang 8be a reasonable first-line agent for inducing and
maintaining complete remission in patients with
pro-liferative lupus nephritis However, patients should be
made aware of the lack of long-term data for
treat-ment effects on kidney function and longer term
toxicity
The finding that MMF and calcineurin inhibitors
alone or in combination have a higher probability of
inducing remission than IV cyclophosphamide
contrasts with a 2012 updated Cochrane review that found no differences between these 2 treatments for complete remission of proteinuria,3 although it
Table 4 Summary Network Estimates of Drug Regimens as
Maintenance Treatment on Disease Relapse Compared to
Azathioprine
Drug(s) Comparison for
Relapse
Network Meta-analysis Estimate vs
Azathioprine
Note: Based on 6 studies (570 participants) in network Values
are given as odds ratio (95% confidence interval) derived from
network meta-analysis Odds ratio , 1 favors active drug class.
The heterogeneity tau ( s ) value in the network analysis for
treatment relapse was s , 0.001 (low heterogeneity), consistent
with the possibility there was insufficient statistical power in the
network to detect heterogeneity Treatment estimates are shown
in order of efficacy per the surface under the cumulative ranking
(SUCRA) curve.
Abbreviations: IV, intravenous; MMF, mycophenolate mofetil.
a Statistically significant.
Best Worst
Intravenous
Calcineurin
Best Worst Best Worst
Best Worst
Figure 4 Rankings for efficacy of immunosuppression as maintenance therapy to prevent disease relapse in lupus nephritis Graph displays distribution of probabilities for each outcome Ranking indicates probability that drug class is first “best,” second best, etc For example, mycophenolate mofetil (MMF) is among the best for preventing disease relapse during maintenance ther-apy, while intravenous cyclophosphamide is among the worst.
Intravenous
cyclophosphamide
cyclophosphamide
inhibitor
Complete disease remission All-cause mortality End-stage kidney disease
Prednisone
Intravenous cyclophosphamide
cyclophosphamide
inhibitor
Major infection Alopecia Ovarian failure
Prednisone
+ calcineurin inhibitor
Intravenous cyclophosphamide + MMF
+ calcineurin inhibitor
Intravenous cyclophosphamide + MMF
Figure 3 Rankings for efficacy and safety of immunosuppression treatment to induce disease remission in lupus nephritis The graphs display the distribution of probabilities of treatment ranking from best through worst for each outcome Ranking indicates the probability that drug class is first “best,” second “best,” etc For example, the ranking suggests that oral cyclophosphamide treat-ment posed the highest risk for incurring ovarian failure (worst), while prednisone incurred the lowest probability of ovarian failure (best) Mycophenolate mofetil (MMF) and calcineurin inhibitors were among the best treatments for inducing disease remission, while intravenous cyclophosphamide and prednisone alone provided the lowest probability of disease remission (worst).
Trang 9concluded that there was similar uncertainty, as
observed in the current analysis for treatment odds of
death and kidney function Similarly, a 2009
meta-analysis observed similar efficacy between MMF
and IV cyclophosphamide for inducing kidney
dis-ease remission, with comparable risks for death and
end-stage kidney disease.4 A Bayesian network
meta-analysis published in 2014 concluded that there
was insufficient evidence to determine whether
mycophenolate or tacrolimus was superior to
cyclo-phosphamide for inducing proteinuric remission or
normal serum creatinine levels at 6 months.5 A
network analysis involving 9 trials in 972 patients
concluded that tacrolimus was superior to
cyclo-phosphamide for inducing complete or partial disease
remission, whereas MMF was comparable to
cyclo-phosphamide treatment.6 The different conclusions
drawn by these meta-analyses and the present study
are likely due to the larger amount of information
available in the present review to permit more precise
inferences about competing treatments for lupus
nephritis
The findings of this network meta-analysis were
consistent with those of a recent randomized trial
comparing combined tacrolimus and MMF therapy
versus IV cyclophosphamide in biopsy-proven lupus
nephritis, which showed that patients who received
combined therapy had a higher probability of
com-plete or partial disease remission at 6 months (hazard
ratio [HR], 1.72; 95% CI, 1.34-2.21), although there
were more serious adverse events and greater
with-drawal with tacrolimus plus MMF therapy, driven
primarily by infection-related events.57This raises the
possibility that dual therapy might incur greater
toxicity than each individual treatment class alone
Although there have been no previous head-to-head
trials of dual versus monotherapy, when the 2
tri-als25,57 evaluating combined MMF plus tacrolimus
versus cyclophosphamide as induction therapy were
included in the network for complete remission in this
meta-analysis, dual MMF plus tacrolimus therapy was
significantly better than MMF alone (OR,1.87; 95%
CI, 1.06-3.29) but was not superior to tacrolimus
alone (OR, 1.54; 95% CI, 0.81-2.94) Based on these
promising results and to test the balance of benefits
and harms of dual therapy further, a trial comparing
tacrolimus monotherapy against combined MMF plus
tacrolimus might be considered a priority, including
careful documentation of long-term patient-centered
treatment efficacy and harms
The findings of this network analysis were
consistent with a 2015 study evaluating comparative
effects of cyclophosphamide, azathioprine, MMF, and
prednisone alone on maintenance of disease remission
in lupus nephritis.71In that meta-analysis, MMF was
ranked as the best therapy for preventing kidney
failure during maintenance treatment, although due to
a small number of events, the estimated treatment effects were very imprecise The similarfindings be-tween that previous review and the present study despite differing statistical approaches and end points strengthen the conclusions of both studies that MMF might be the best treatment for maintaining remission
of lupus nephritis
Although the strengths of this systematic review included a comprehensive literature search without language or date restriction, evaluation of the as-sumptions of consistency among included trials before generating treatment estimates, and including standardized approaches to assess the confidence that might be held in the results, the meta-analysis has limitations that might be considered First, the anal-ysis was limited by the data in the primary trials and the methods of reporting data For example, complete remission was a heterogeneous outcome with variable
definitions in existing studies Notwithstanding the differences in end point definitions among trials, there was evidence of low heterogeneity in the pooled analysis, indicating that a network meta-analysis was appropriate, although the statistical power within networks to detect heterogeneity was likely to be low Second, there were few deaths (n5 235) and patients progressing to end-stage kidney disease (n5 164) during trial follow-up Nearly 40% of studies did not report clinically relevant outcomes, especially end-stage kidney failure, because these were only likely
to have accrued over several years of treatment and specific outcomes for trials have not been mandated This has led to considerable uncertainty in treatment effects on these patient-relevant outcomes and has resulted in an inability of patients and clinicians to weigh the relative balance of benefits and harms of treatments It remains uncertain whether biochemical remission of disease based on proteinuria and/or serum creatinine level is a valid predictor of end-stage kidney disease because existing trials generally do not follow up patients long enough Similarly, there was insufficient precision in treatment effects on doubling
of serum creatinine level, although azathioprine or corticosteroids alone were clearly inferior to IV cyclophosphamide Third, although treatment classes were derived from similar study populations (age, sex, and serum creatinine level), there were secular trends in the publication era for differing treatments
As expected, azathioprine, oral cyclophosphamide, and prednisone alone were principally evaluated in earlier decades, whereas IV cyclophosphamide, MMF, and calcineurin inhibitors were assessed in more recent trials Although this difference might threaten the assumed consistency required to generate
a single analytical network and confound treatment comparisons due to differing epidemiologic patterns
Trang 10of disease and treatments over time, notably there was
low heterogeneity observed in networks for the
pri-mary outcomes, and the key treatment comparisons
(cyclophosphamide, MMF, and calcineurin
in-hibitors) were drawn from trials published more
recently Fourth, different outcomes and responses to
treatment were observed among people of different
racial origins in lupus nephritis, and therefore it might
be hypothesized that treatment effects might be
different based on ethnicity However, there were
insufficient data for race or ethnicity in the original
trial reports to perform metaregression analyses to
explore this possibility Fifth, we did not include trials
with zero events This approach may have resulted in
effect estimates moving toward the null, although for
all clinical efficacy end points including mortality and
end-stage kidney disease, this approach likely
over-estimated the magnitude of effect Finally, the
inconsistent end point definitions and imprecision in
treatment estimates for mortality and end-stage
kid-ney disease have implications for future trial design
In future studies, longer term end points and larger
study populations might be achieved through more
efficient study design (eg, registry-based randomized
trials72) in which important patient-centered outcomes
such as death and end-stage kidney disease are
captured automatically during long-term routine
follow-up within registry databases Standardization
of both safety and short- and long-term efficacy
out-comes in trials evaluating therapies for lupus
nephritis, as has been generated in rheumatology,
might facilitate better understanding about the
bene-fits and harms of therapy.73
Based on the potential benefits of calcineurin inhibitors and MMF on
short-term outcomes in this analysis, future head-to-head
trials comparing the benefits and harms of these
treatments alone or in combination might be
prioritized
In conclusion, evidence for induction therapy for
lupus nephritis is inconclusive based on treatment
effects on all-cause mortality, doubling of serum
creatinine level, and end-stage kidney disease
Compared to IV cyclophosphamide, the most
effec-tive therapies for inducing remission were MMF,
calcineurin inhibitors, or their combination while
conferring similar or lower treatment toxicity The
most effective maintenance therapy was MMF
ACKNOWLEDGEMENTS
Support: There were no speci fic funders of this study Dr
Palmer is supported by a Rutherford Discovery Fellowship Mr
Tunnicliffe is funded by a postgraduate scholarship from the
Sydney Medical School, The University of Sydney Dr Mavridis is
supported by the European Research Council (IMMA 260559) Dr
Tonelli received funding from Alberta Innovates Health Solutions.
Financial Disclosure: Dr Palmer received a research grant from
the Royal Society of New Zealand during the study and has
received research funding from Amgen Dompé Dr Johnson has received consultancy fees from Baxter, Fresenius, Gambro, Amgen, Janssen-Cilag, Roche, Genzyme, Shire, Sigma, Sano fi-Aventis, Boehringer-Ingelheim, Lilley, Merck Sharpe & Dohme, Bristol-Myers Squibb, and Novartis; speaker ’s honoraria from Baxter, Fresenius, Gambro, Amgen, Janssen-Cilag, Roche, Servier, Shire, Merck Sharpe & Dohme, Boehringer-Ingelheim, and Bristol Myers Squibb; research grants from Baxter Extra-mural, Fresenius, Roche Foundation for Anaemia Research (RoFar), Amgen, Janssen-Cilaz, P fizer, and Abbott; and travel sponsorships from Baxter, Fresenius, Gambro, Amgen, Janssen-Cilag, Roche, and Shire Dr Tonelli has received honoraria for a lecture series on management of dyslipidemia of chronic kidney disease from Merck; all honoraria were donated to charity Dr Strippoli received a research grant from Agenzia Italiana del Farmaco during the study and has received personal fees for consultancy from Servier Laboratories.
Contributions: Research idea and study design: SCP, GFMS; data acquisition: DJT, SCP; data interpretation: DJT, DS-G, DM,
MT, DWJ, JCC, AT, GFMS; statistical analysis: SCP; supervision
or mentorship: DM, GFMS.Each author contributed important intellectual content during manuscript drafting or revision and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved GFMS takes responsibility that this study has been reported honestly, accu-rately, and transparently; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.
Peer Review: Evaluated by 2 external peer reviewers, a Statis-tical Editor, a Co-Editor, and Editor-in-Chief Levey.
SUPPLEMENTARY MATERIAL Table S1: Characteristics of included studies.
Table S2: Description of randomized treatment strategies for induction or maintenance of remission and nonrandomized treat-ment to maintain remission.
Table S3: Funding sources in included trials.
Table S4: Pairwise and network estimates of ef ficacy end points associated with immunosuppression for induction treatment of proliferative lupus nephritis.
Table S5: Summary pairwise and network estimates of safety end points associated with immunosuppression for induction treatment of proliferative lupus nephritis.
Table S6: Evaluation of loop-speci fic consistency in triangular and quadrilateral treatment loops for each binary outcome network.
Table S7: Evidence of global heterogeneity within analyses Table S8: Summary raw data for duration of follow up, no of events, and no of participants at risk for each outcome reported in network.
Table S9: De finitions of primary and secondary end points in trials.
Table S10: Pre-speci fied sensitivity analyses for primary outcome of complete disease remission.
Figure S1: Risk of bias summary: judgments about each bias item for each study.
Figure S2: Risk of bias summary graph.
Figure S3: Study limitations for each network estimate for immunosuppressive strategies vs IV cyclophosphamide Figure S4: Summary study-level characteristics according to drug class.
Item S1: Study protocol.
Item S2: Network treatment estimates.
Item S3: Networks of treatment comparisons for other outcomes.