The lay abstract featured today (for Group Sequential Test for Two-Sample Ordinal Outcome Measures by Yuan Wu, Ryan A. Simmons, Baoshan Zhang and Jesse D. Troy) is from Statistics in Medicine with the full article now available to read here.

How to cite

Wu, Y., Simmons, R.A., Zhang, B. and Troy, J.D. (2025), Group Sequential Test for Two-Sample Ordinal Outcome Measures. Statistics in Medicine, 44: e70053. https://doi.org/10.1002/sim.70053

Lay Abstract

This abstract was developed with assistance from ChatGPT 4.0. The chat session is available here.

Clinical trials help determine whether new medical interventions—such as drugs, medical devices, vaccines, and behavioral therapies—are safe and effective. Many clinical trials evaluate new medical interventions by comparing how well people do after receiving the new intervention versus an already approved treatment (or a placebo, if no treatment currently exists). This type of study is called a randomized controlled trial. Most randomized controlled trials run from start to finish before results are analyzed. In some cases, this means it can take several years before researchers know whether a new medical intervention is safe and more effective than the standard treatment or placebo.

A modification to the standard randomized controlled trial—called a group sequential trial—allows researchers to evaluate how well the new medical intervention is performing as the trial progresses, rather than waiting until the end. This approach has the potential to shorten the time it takes to get an answer about the new intervention. Researchers can use group sequential methods to stop a trial early if results suggest the new intervention is unlikely to be better than the standard treatment, avoiding unnecessary time and resources. Likewise, the use of group sequential methods can lead to early stopping of clinical trials when there is overwhelming evidence of benefit, in which case proceeding to the end of the study may not be necessary to demonstrate how well the new intervention works.

Group sequential methods for clinical trials are commonly used to study things like drug effects on blood pressure, whether treating cancer patients with chemotherapy in addition to radiation will help them live longer, or whether vaccines can prevent infections like COVID-19. However, many clinical trials collect ordinal data, where response to an intervention is measured in ordered categories (such as “none,” “mild,” “moderate,” or “severe” symptoms). Standard statistical methods for group sequential designs do not always perform well for these types of outcome measures. This research develops a new statistical method that extends group sequential designs to work effectively with ordinal data.

The method highlighted in this article is an extension of a statistical approach commonly applied to ordinal data: the Mann-Whitney-Wilcoxon test. This work includes simulation studies that demonstrate the new method controls the risk of false positives (incorrectly concluding a treatment works) while maintaining the ability to detect true treatment effects. Included among these simulation studies are comparisons to other group sequential methods for ordinal data previously proposed by researchers. Results show that the new method has advantages over existing approaches when the number of participants in the analysis is small. A real-world example is also provided, showing how this approach can be applied to future clinical trials. The example is based on the Modified Rankin Scale, which is a seven-category ordinal scale used to measure disability after a stroke. These results expand the toolkit available to researchers conducting clinical trials with ordinal outcomes, making it easier to design efficient and ethical studies that are able to identify interventions that will benefit patients.