Synthetic Samples

Our synthetic sample generation approach ensures accurate representation of real-world populations for modelling clinical outcomes/intervention effectiveness.

Case Studies

The challenge was to develop a synthetic population that resembles the population of the Scandinavian Simvastatin Survival Study (4S). Using a stochastic resampling technique we generated a semi-random sample of individuals matching the control group’s characteristics in 4S, across 10 risk factors such as age, BMI and systolic blood pressure. We presented the algorithm’s success at ISPOR, demonstrating the tool’s potential for aiding in clinical trial planning and preparation, and empowering evidence-based decision-making in drug development and healthcare research.

Crystallise’s synthetic sample generation approach for the pharmaceutical industry is a cutting-edge solution that uses Spearman Rank Correlation and Cholesky decomposition techniques. With this method, Crystallise has created a sample of 1 million individuals that closely mirrors the characteristics of a real-world population to within 0.5% of the original population from the Health Survey for England.

By using Crystallise’s synthetic data, pharmaceutical companies can confidently model the potential impact of new interventions on clinical outcomes such as mortality or morbidity, where directly comparable clinical data is sparse or non-existent. This approach identifies indications where a new product is unlikely to be cost-effective, to help clients decide whether to proceed with expensive phase 3 clinical trials.

Get in Touch

17 High Street,
SS17 0HD



Company No: 7980921
Data Protection Act Registration No: Z3363643
VAT No: 190 8750 82

© Crystallise 2023

Crystallise Policy

Crystallise Carbon Policy