Simulants: Synthetic Clinical Trial Data via Subject-Level Privacy-Preserving Synthesis.

Area of Science:

• Biomedical Informatics
• Data Science
• Clinical Research

Background:

• Clinical trial data is valuable but inaccessible due to privacy and regulatory hurdles.
• Existing data sharing methods are limited by privacy concerns and data utility.
• Synthetic data offers a potential solution for secure data sharing and research.

Purpose of the Study:

• To introduce Simulants, a novel method for generating research-grade synthetic clinical trial data.
• To evaluate the performance of Simulants against state-of-the-art deep learning synthesizers.
• To demonstrate the adaptability of Simulants for diverse privacy policy requirements.

Main Methods:

• Simulants was developed to generate synthetic clinical trial data from real data sources.
• Performance was assessed using established metrics for data fidelity and privacy preservation.
• Comparison was made against leading deep learning-based synthetic data generation techniques.

Main Results:

• Simulants demonstrated superior performance in both data fidelity and privacy preservation compared to deep learning methods.
• The method effectively captured critical clinical features from the original data.
• Simulants' privacy settings can be tailored to meet specific data sharing policies.

Conclusions:

• Simulants provides a robust solution for generating high-fidelity, privacy-preserving synthetic clinical trial data.
• This approach can unlock the potential of clinical trial data for wider research and drug development.
• The method offers flexibility in privacy control, facilitating compliance with data governance regulations.