A recent study investigated the effectiveness of the augmented binary method for investigating rare diseases in a small sample size.
Composite endpoints consider a number of individual outcomes to evaluate the efficacy of a treatment overall and are often recommended in research into rare diseases in order to improve power and capture complexity. These composite endpoints often take the form of responder indices, which consist of a mixture of continuous and binary components and analyses of these outcomes are typically treated as binary.
A recent study investigated the effectiveness of the augmented binary method for investigating rare diseases in a small sample size. The study used data from the OSKIRA-1 trial, a phase 3 trial that assessed the use of fostamatinib in patients with a common disease: active rheumatoid arthritis.
Based on the clinical response description and score method in the OSKIRA-1 trial, the researchers established a common responder endpoint as a patient who tolerates treatment, does not receive restricted medications, and demonstrates clinical response.
“If the components are appropriately chosen, composite endpoints that require an event in only  of the components (a or b or c, etc.) may have the ability to improve the power to show a given treatment effect due to the increased number of events,” stated the authors. “These characteristics appeal to rare diseases where many realizations of the diseases are highly variable, and availability of the population may be a binding constraint.”
The researchers applied the standard binary and augmented binary methods to samples ranging from of 30 to 80 in order to determine the power, type I error rate, coverage and average confidence interval width for each of the estimators. Firth’s adjustment for the binary component models was implemented, and small sample variance correction for the generalized estimating equations was applied.
The results of the analysis revealed that, for the log-odds treatment effect, the power of the augmented binary method is 20% to 55% compared to 12% to 20% for the standard binary method, and both methods have about nominal type I error rates. Additionally, both unadjusted methods demonstrate type I error rates of 6% to 8% while the small sample corrected methods have approximately nominal type I error rates.
The decrease in the average confidence interval width when using the adjusted augmented binary method was 17% to 18% on the same on both scales—which is equivalent to requiring a 32% smaller sample size to reach the same statistical power, according to the study.
“In rare diseases where there are few or no available treatments and limited opportunity to test emerging new treatments, the power to detect an effective treatment is of critical importance,” concluded the study. “The augmented binary method with small sample corrections provides a substantial improvement for rare disease trials using composite endpoints.”
The study recommended using the augmented binary method in relevant rare disease trials. It also suggested using this method in addition to other efforts to improve the quality of evidence in rare disease trials.
McMenamin M, Berglind A, Watson JMS. Improving the analysis of composite endpoints in rare disease trials. 2018;13(81). doi: 10.1186/s13023-018-0819-1.