Measuring treatment effect during a clinical trial is often the source of much debate, particularly during rare disease trials that must stimulate investigations using small samples. Unlike statistically significant results, for which there are many tests, meaningful measures of treatment effect are still under development (Kieser 2012). Cytel statistician Ursula Garczarek wonders whether this holds true in the realm of small samples and small target populations. After all, does the summary statistic in such a small trial rely on many assumptions that might not correlate with reality?
Dr. Garczarek’s recent talk at Cytel’s New Horizons series motivated a conversation on whether simulations for rare disease clinical trials might offer more insights than a summary statistic. Even within Cytel, many scientists argue for asymptotic or non-asymptotic exact methods, such as those available in StatXact and LogXact.
During the webinar, Dr. Garczarek walked the audience through a case study where meta-analysis on historical datasets suggested that there would be two clusters of responders to a drug to enhance visual acuity. This added the complexity of heterogeneity into an already difficult statistical problem.
Watch the webinar below to participate in the discussion.
The New Horizons Series aims to galvanize discussion on the newest trends and innovations in statistics for clinical trial research. The next webinar in this series is by Cytel’s Alind Gupta where he will be presenting case studies from applying machine learning for predictive analysis and evidence generation from randomized clinical trials and real-world data. Click the button register.
 Kieser, Meinhard, Tim Friede, and Matthias Gondan. "Assessment of statistical significance and clinical relevance." Statistics in Medicine 32.10 (2013): 1707-1719.