Noninferiority Trials Provide Reliable Evidence of Drug Effects in RA

Noninferiority trials do not provide direct evidence of superiority to placebo but can provide evidence to support efficacy of RA therapies.

Noninferiority trials of active comparator drugs are feasible and may provide reliable evidence of drug effects in rheumatoid arthritis (RA), according to research results published in Arthritis & Rheumatology.

Investigators examined the potential specifications and considerations for noninferiority trials that would assess the effects of drugs on common symptomatic, functional, and radiographic end points, with tumor necrosis factor inhibitors (TNFis) used as the active comparators.

To determine an appropriate and feasible noninferiority margin for noninferiority studies for both symptomatic and functional RA improvement, researchers reviewed placebo-controlled clinical trials that included patients with active RA and performed a meta-analysis to examine the treatment effect of 5 TNFis that have been approved for RA treatment: adalimumab, infliximab, etanercept, certolizumab, and golimumab.

The researchers focused on the absolute difference metric between the treatments given so that the effects of all 5 TNFIs were generally consistent with regard to response difference, response ratio, and odds ratio. In addition, the absolute difference metric captures the potential effect on public health of replacing a comparator with a new treatment.

The use of a single fixed noninferiority margin across different TNFi active controls was deemed to be appropriate for the American College of Rheumatology 20% response end point, owing to the large and reasonably consistent effect sizes across this class of drugs. Slight differences in the effect sizes may be due to either chance or minor differences in study design or analysis.

A noninferiority margin of 12% was chosen by the researchers based on multiple considerations. Investigators had a high level of confidence that this margin would rule out substantial efficacy losses and would correspond to an acceptable percentage of conservative estimates of the effect of active controls, which are needed to establish treatment efficacy compared with placebo.

The researchers’ review of historic, placebo-controlled trials identified 4 drugs — adalimumab, certolizumab, golimumab, and infliximab — that had available placebo-controlled data at 52 weeks. Treatment effect estimates for these drugs ranged from -0.7 to -5.7, with 95% upper confidence bounds ranging from 0.4 to -2.8. These results led investigators to note that for radiographic progression, noninferiority margins should be specifically defined for each drug.

Sample size for these trials would depend on the choice of noninferiority margin, significance level, power, and the underlying variability in the trial end point. With a margin of 12% for noninferiority, sample sizes would need to be in the range of those used for placebo-controlled RA drug trials.

In considering possible ways to increase the sensitivity of noninferiority studies, researchers suggested enriching the study population for radiographic progression studies or incorporating different strategies for reading x-rays.

Finally, investigators reviewed inclusion and exclusion criteria, baseline population measurements, and key elements in study design in order to determine a noninferiority margin for symptomatic and radiographic end points. For all noninferiority studies, settings should mimic historical studies as similarly as possible in order to ensure sensitivity.

There is one primary limitation to this approach: noninferiority trials do not provide direct evidence of superiority to placebo. However, the investigators noted that adequately designed noninferiority trials do provide strong evidence to support efficacy standards for RA therapies.

“[W]e recommend pursuing phase 3, active-controlled [noninferiority] trials in drug development in RA,” the researchers concluded. “Active-controlled superiority trials may also be considered if the new drug is believed to be more effective than the active control… Trials that include active comparators are critically needed in this current RA drug development landscape.”


Rothwell R, Nikolov NP, Maynard JW, Levin G. Non-inferiority trials to evaluate drug effects in rheumatoid arthritis [published online March 17, 2020]. Arthritis Rheumatol. doi:10.1002/art.41257