Methodological issues of corticosteroid use in SLE clinical trials

Objective: To describe and analyze how investigators handled concurrent ste roid intake in SLE clinical trials; to identify remaining methodological pr oblems and to discuss potential solutions. Methods: A review of the literature was performed in Medline to identify ra ndomized controlled trials (RCT) in lupus published during the past decade. A set of criteria was defined a priori to analyze these trials covering di fferent aspects of steroid use in RCTs with regard to eligibility, randomiz ation, post-randomization steroid use, analysis and reporting. Results: Seventeen trials met the inclusion criteria and were analyzed. Med ian sample size was 30, ranging from 10-71. In three trials corticosteroid application was the study intervention. Three reports did not address stero id 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 concurren t steroid use, seven checked for equal distribution of baseline steroids. I n some cases clinically relevant differences did not reach statistical sign ificance as sample sizes were very small. In only one study was randomizati on stratified for steroid use. Instructions for post-randomization steroid use were specified in only 3 of 14 trials where steroids were not part of t he intervention. In 11 RCTs concurrent steroid use was free or no informati on was provided. Steroid dosing during the intervention was reported as a s econdary outcome measure in four trials. Conclusions: As sample sizes are often small, simple randomization is not r eliable to avoid relevant between-group differences in baseline steroids. T he power of statistical testing for group differences is limited for the sa me reason. Stratified randomization is rarely used. Free concurrent steroid use during the intervention is frequently allowed in lupus RCTs despite th e risk of bias. Better data and accepted standards for rt priori specified steroid regimens in RCTs are highly warranted.