Every few days, we will be publishing layman’s abstracts of new articles from our prestigious portfolio of journals in statistics. The aim is to highlight the latest research to a broader audience in an accessible format.
The article featured today is from Statistics in Medicine and the full article, published in issue 38.21, is available to read online here.
Liu, Y, Rybin, D, Heeren, TC, Doros, G. Comparison of novel methods in two‐way enriched clinical trial design. Statistics in Medicine. 2019; 38: 4112– 4130. doi: 10.1002/sim.8288
In clinical trials, the randomized, double-blinded, placebo-controlled design has been considered a gold standard in the assessment of drug efficacy. However, it is well recognized that placebo response is a problem in clinical trials in several therapeutic areas. High placebo response mitigates medication-placebo differences, which makes it more difficult to demonstrate a statistically significant benefit of a putative agent over placebo. In response to this issue, physicians, statisticians, and clinical researchers are exploring methods to identify the predictors of placebo response and to improve the conduct of clinical trials to reduce the impact of placebo response. Many novel clinical trial designs and analysis methods have been introduced in the past ten years.
Two-way enriched design (TED) was originally introduced in 2011, as an extension of the basic sequential parallel comparison design (SPCD). In the published work on TED, the outcomes are assumed independent from stage to stage. However, the repeated measures on the same subject from Stage I to Stage II may be highly correlated. A more comprehensive and robust analysis approach is warranted to incorporate the correlation within the subjects.
In addition, in the existing analytic methods for TED, classification of patients is based solely on the outcome at the end of Stage I. The placebo response and drug response are treated as measurable subject characteristics, without stochastic properties. However, these attributes are not observable directly and may not be measured accurately. Misclassification can be a problem in the estimate of drug efficacy.
In this manuscript, outcomes in TED are considered as repeated measures. Correlation between outcomes from different stages are taken into account in the estimation
of drug efficacy. The placebo non-response is first considered as a measurable binary characteristic (“present” or “absent” status) and the Stage I and Stage II outcomes are modeled by allowing the existence of correlation. Then, the first assumption is relaxed and the placebo non-response is considered as a characteristic that exists in every subject to a certain degree in this study. It is treated as a continuous measure, ranging from 0 to 1, from absolute responder to absolute non-responder. It is proposed to be included in the model as a weight for the placebo group. A repeated measures model will first be applied to binary outcomes and compared with a score test. Then, a repeated measures model and a weighted repeated measures model will be applied to continuous outcomes and compared with the OLS method.
TED has many promising statistical properties, according to the assessment of previous researches. However, this design has yet to be applied, due to the lack of flexible and robust analysis methods. This work provides several novel analysis methods to encourage the use of this design in clinical trials.