Criteria for considering studies for this review
Types of studies
Randomized controlled trials (RCTs) or controlled clinical trials (CCTs), that is quasi-randomized clinical trials.
Types of participants
Patients with SLE who meet the American College of Rheumatology classification criteria for SLE (Hochberg 1997).
Types of interventions
Belimumab, alone or in combination with other immunosuppressive drug (such as azathioprine, cyclosporine, mycophenolate mofetil) or another biologic, compared to placebo or other disease-modifying anti-rheumatic drugs (DMARDs), DMARD combinations, or biologics.
Types of outcome measures
Outcomes as assessed by:
change in SLE scores on validated disease activity indices, the Systemic Lupus Erythematosus Disease Activity Index SELENA Modification (SELENA-SLEDAI) (Petri 2005), modified SELENA-SLEDAI Flare Index (SFI) (Petri 2005; Petri 1999); British Isles Lupus Assessment Group index (BILAG) (Hay 1993; Isenberg 2000); or other similar validated indices;
composite responder rate, as defined with the newly validated measure the Systemic lupus Erythematosus Responder Index (SRI), where a responder is defined as a patient with 1) a ≥ 4-point reduction in SELENA-SLEDAI score, 2) no new BILAG A or no more than one new BILAG B domain score, and 3) no deterioration from baseline in the physician's global assessment by ≥ 0.3 points (Furie 2009);
quality of life, assessed by the Short-Form 36 (SF-36) or similar assessments;
withdrawals due to adverse events;
serious adverse events, number of serious adverse events (SAEs) or number of patients with ≥ 1 SAE;
Physician Global assessment (PGA), assessed on a 0 to 3 scale (0 indicating inactive disease and 3 severe disease) (Petri 2005)
Reduction in glucocorticoid dose, defined as mean reduction of prednisone equivalent dose in mg/day, cumulative dose, % reduction or below a certain threshold (< 10 mg/day or < 7.5 mg/day), or % patients off glucocorticoids
Disease remission, based on SLEDAI (Mosca 2006) or other similar indices
Change in biomarkers including, but not limited to, complement levels, anti-double stranded DNA levels, and other novel biomarkers
Number of lupus flares
Time to first lupus flare
Number of severe flares
End-organ damage or failure, proportion with significant renal insufficiency (doubling of creatinine or halving of glomerular filtration rate (GFR), progression to dialysis, worsening of proteinuria), stroke, myocardial infarction, respiratory failure etc.
Number of adverse event (AEs), defined as the total number of adverse events. In the event this is not presented, we will take the number of patients with any (≥ 1) adverse events in a study for this outcome
Number of specific adverse events (SAEs) including infusion reactions, malignancy AEs, etc.
Withdrawals for any reason, a combination of withdrawals due to adverse events or inefficacy, or other reasons
Damage as measured by an index such as the Systemic Lupus International Collaborative Clinics (SLICC) (Gladman 1992)
Individual organ response in SLE as indicated by improvement in SLE activity scores within an organ, such as BILAG As aggregated across an organ system
Search methods for identification of studies
The Trials Search Coordinator (TSC) will carry out the searches of The Cochrane Library, MEDLINE, EMBASE, CINAHL, Web of Science, and the World Health Organization (WHO) International Clinical Trials Registry Platform. There will be no language or date restrictions in the search for trials and the databases will be searched from inception to present date. The search will be updated before the completion of the review to ensure inclusion of new trials in the intervening period.
We will use the MEDLINE search strategy as in Appendix 1, also adapted for other databases (The Cochrane Library, Appendix 2; EMBASE, Appendix 3; CINAHL, Appendix 4; Web of Science, Appendix 5; WHO International Clinical Trials Registry Platform, Appendix 6).
Searching other resources
We will search the reference lists of included studies to find additional articles of relevance.
Data collection and analysis
Selection of studies
Two review authors (JS, SN) will independently review titles and abstracts to find eligible studies. Disagreements will be discussed and resolved by consensus.
Data extraction and management
Two independent review authors (JS, SN) will extract data from the included studies using standardized data extraction forms. The extracted data will include study characteristics, study population characteristics, details about the interventions, funding sources, and outcomes of interest. When we need more information, we will contact the authors of the studies. We will extract data from the published reports including mean and standard deviation for continuous outcomes and number of events and people at risk for dichotomous data. Whenever possible, we will extract the data based on an intention-to-treat analysis. Risk of bias will be assessed using the methods sections of all identified trials, as described below.
Assessment of risk of bias in included studies
Two review authors (JS, SN) will independently assess the risk of bias for each included trial using the Cochrane Collaboration recommendations. We will assess the risk of bias on each of the following criteria: presence of blinding (participants, personnel, and outcome assessors) in the studies, allocation concealment, random sequence generation, incomplete outcome data, and selective outcome reporting (Higgins 2011). The risk of bias will be assessed as recommended: low risk, high risk, or unclear risk (either lack of information or uncertainty over the potential for bias). Disagreements will be resolved by discussion between the review authors.
Measures of treatment effect
Treatment effect in each study will be assessed using risk ratios (RRs) for dichotomous outcomes and mean differences for continuous outcomes; the 95% confidence interval (CI) will be calculated. For continuous measures, we chose mean differences over standardized mean difference (SMD), since mean difference (MD) is easier to interpret. If different scales measuring the same construct are found, then an SMD will be calculated and then transformed back to a scale familiar to clinicians to aid in the interpretation. We plan to use the random-effects model as a conservative approach.
Unit of analysis issues
We do not anticipate any unit of analysis issues. Most studies use the patient as the unit of analysis. If a study has more than one treatment arm for belimumab, we will use the treatment arm with the standard, approved dose of belimumab. If there are multiple comparator arms, we will split the numbers from the belimumab dose arm for comparison as appropriate (split into half for two comparator arms etc.).
Dealing with missing data
In the case of significant missing data, we plan to send email queries to the authors requesting the missing data. When data for variability statistics (as standard deviations) cannot be obtained, we will use the baseline standard deviation. We will describe when missingness is more than 20%. We do not plan to impute any missing data.
Assessment of heterogeneity
We will assess for clinical homogeneity by the following characteristics: age, gender, race, disease duration, number of immunosuppressives used or failed previously, and types of control interventions. To assess statistical heterogeneity quantitatively, we will perform a Chi2 test and consider P < 0.10 as an indicator of potentially significant heterogeneity. To examine statistical heterogeneity qualitatively, we will examine the I2 statistic. In interpreting the I2 statistic, we will use the recommendation in the Cochrane Handbook for Systematic Reviews of Interventions, which indicates that 0% to 40% might not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; and 75% to 100% considerable heterogeneity (Deeks 2011).
Assessment of reporting biases
We will make a funnel plot to assess reporting biases when performing an analysis on 10 or more studies. The funnel plot is a scatter plot with sample size on the y-axis and the treatment effect on the x-axis. An asymmetry in the funnel plot is indicative of reporting bias or other biases related to small study size.
When feasible, based on the lack of substantial or considerable heterogeneity, we will pool data using meta-analysis. The random-effects model will be our default model for pooling outcomes in the meta-analysis. We will calculate the number needed to treat to benefit (NNTB) or harm (NNTH) as the inverse of the absolute risk difference. We will calculate the number needed to treat using the Visual Rx NNT calculator (Cates 2004) for dichotomous outcomes. The NNT for continuous measures will also be calculated with help from the Cochrane Musculoskeletal Group (CMSG) editorial office using the Wells calculator.
Subgroup analysis and investigation of heterogeneity
The planned subanalyses are the following, if data are available.
By disease severity: patients stratified based on baseline lupus disease activity corresponding to mild versus moderate versus severe lupus (as per different cut-offs on various scales). If data are scarce, we will collapse the moderate and severe categories.
By patient demographics: age (< 65 versus 65 years and above), race or ethnicity.
By concomitant prednisone use: dose of prednisone < or ≥ 7.5 to 10 mg/day equivalent.
We plan to carry out sensitivity analyses by allocation concealment and blinding.
Summary of findings tables
We will make a 'Summary of findings' table, included in RevMan 5.2, in order to communicate the key outcomes of the review, as recommended by The Cochrane Collaboration (Schunemann 2011a). We will use GRADE Profiler software to provide an overall grading of the quality of the evidence in the 'Summary of findings' table (Schunemann 2011b).