Methodological issues of corticosteroid use in SLE clinical trials

Objective: To describe and analyze how investigators handled concurrent steroid intake in SLE clinical trials; to identify remaining methodological problems and to discuss potential solutions. Methods: A review of the literature was performed in Medline to identify randomized controlled trials (RCT) in lupus published during the past decade. A set of criteria was defined a priori to analyze these trials covering different aspects of steroid use in RCTs with regard to eligibility, randomization, post-randomization steroid use, analysis and reporting. Results: Seventeen trials met the inclusion criteria and were analyzed. Median sample size was 30, ranging from 10-71. In three trials corticosteroid application was the study intervention. Three reports did not address steroid use among study subjects at all. In seven trials steroid use was part of the eligibility criteria in some way. of the trials that allowed concurrent steroid use, seven checked for equal distribution of baseline steroids. In some cases clinically relevant differences did not reach statistical significance as sample sizes were very small. In only one study was randomization stratified for steroid use. Instructions for post-randomization steroid use were specified in only 3 of 14 trials where steroids were not part of the intervention. In 11 RCTs concurrent steroid use was free or no information was provided. Steroid dosing during the intervention was reported as a secondary outcome measure in four trials. Conclusions: As sample sizes are often small, simple randomization is not reliable to avoid relevant between-group differences in baseline steroids. The power of statistical testing for group differences is limited for the same reason. Stratified randomization is rarely used. Free concurrent steroid use during the intervention is frequently allowed in lupus RCTs despite the risk of bias. Better data and accepted standards for rt priori specified steroid regimens in RCTs are highly warranted.