State-of-the-art statistical design and analysis for maximal success with minimal patient burden in Epidermolysis bullosa trials (EBStatMax)

Short lay summary

A particular dataset from Epidermolysis Bullosa Simplex (EBS) research forms the basis of the EBStatMax project, aiming to reanalyze the data using various state-of-the-art methodologies, design recommendations for future trials, devise computational tools for practitioners, and provide educational material.

Based on the simulation studies of the original data, various methods could be considered as alternatives to conventional methods. The results revealed that there is not one single best method, since a trade-off between achieving high power, accounting for period, cross-over and carry-over effects should be made. To better communicate this to non-statisticians (e.g. patients, clinicians), educational materials were designed in a modular structure, and using a blended learning concept. A shiny web application was developed to perform statistical analysis in a user-friendly way using the recommended approaches of this project.

Scientific summary

EBStatMax is a “demonstration project” funded by the European Joint Programme on Rare Diseases, aimed at bridging the gap between challenges arising from clinical practice and their potential statistical solutions. An international consortium of statisticians with different overlapping and complementing areas of expertise, were

• reanalyzing and simulating data using different state-of-the- art methodologies, trying to exploit the longitudinal nature of the data as much as possible,
• investigating the impact that certain characteristics of the trial have on the statistical analysis,
• developing strategies and designing recommendations for future trials in this area, but also transferable to other rare diseases,
• devising computational tools that can be used by practitioners in order to implement the above-mentioned methodologies in concrete trial planning and analysis, and
• preparing educational material to ensure transferability and high dissemination of the results.

Neutral comparison studies of parametric (model averaging), semiparametric (generalized estimating equations type [GEE-like]) and nonparametric (generalized pairwise comparisons [GPC] and nparLD) methods revealed that the statistical power and the results of the tested methodological approaches depend strongly on the level of measurement of the outcome variable (count, binary and ordinal).

In principle, it could be shown there is no uniformly best method for the aforementioned types of outcome variables, but in particular situations, there are methods that perform better than others. Especially if maximizing power is the primary goal, the prioritized unmatched GPC method was able to achieve particularly good results, besides being appropriate for prioritizing clinically relevant time points. Overall, a balance must be found between achieving high power, accounting for cross-over, period, or carry-over effects, and prioritizing clinically relevant time points.

Strategic relevance

The proposed approach to innovative longitudinal statistical analysis significantly contributes to optimizing the exploitation of longitudinally collected data from a study sample limited in size. This in turn increases the validity of the treatment recommendations derived from the results and enables efficacy assessments based on high-quality evidence despite small sample sizes.