Can Synthetic Data Allow for Smaller Sample Sizes in Chronic Urticaria Research?

Area of Science:

• Medical Informatics
• Biostatistics
• Epidemiology

Background:

• Chronic spontaneous urticaria (CSU) research is limited by underrepresentation of elderly and comorbid patients in clinical trials and observational studies.
• Strict trial criteria and small sample sizes hinder robust data analysis and statistical power.
• Generating synthetic patient data can overcome these limitations by reflecting diverse clinical characteristics.

Purpose of the Study:

• To develop and validate a method for generating synthetic patient data for chronic spontaneous urticaria (CSU).
• To assess the accuracy of synthetic data in replicating real-world data (RWD) characteristics and associations.
• To determine the minimum sample size required for high-quality synthetic data generation.

Main Methods:

• A tree-based decision model was employed to generate synthetic data from existing RWD of the Chronic Urticaria Registry (CURE).
• Replication accuracy was evaluated by analyzing descriptive characteristics and variable associations between RWD and synthetic data.
• The minimum sample size threshold for maintaining high synthetic data accuracy was identified.

Main Results:

• The synthetic data closely mirrored patient demographics and clinical characteristics of the RWD.
• Subgroup replication and distributions aligned with RWD, with no significant differences in disease-specific factors and risk factors (p > 0.05).
• High-accuracy synthetic data generation was achieved with as little as 25% of the original RWD, enabling a fourfold population increase.

Conclusions:

• Synthetic data can accurately replicate RWD for CSU patients, even with reduced original population sizes.
• This methodology offers a viable solution for augmenting small patient subgroups in clinical and epidemiological research.
• Synthetic data generation holds significant potential for advancing research in underrepresented populations.