Designing experiments with Maths saves money and time

A group of statisticians at King’s are working on the design of experiments, known as DOE, to enable their colleagues in health sciences to conduct better experiments and collect data more efficiently.

Senior Lecturer in Statistics in the Department of Mathematics, Dr Kalliopi Mylona is using DOE to optimise pharmaceutical experiments for a new therapy for diabetic wounds - a novel platelet-based therapeutic gel that stimulates wound healing.

She said, “Haphazard experimentation can be very wasteful of resources and lead to needless repetition, poor inference, and it can also be unethical having to retest again and again on animals and humans.

“DOE helps us obtain high quality data, to cut down the experimental costs and the time to complete the experiment. Poor quality data takes time and resources to fix mistakes, so reductions in cost and time have an immediate benefit to our lab-based research.”

The platelet-rich plasma gel at the centre of this experiment is a form of treatment for diabetic ulcers which is being tested to see if it improves the healing rate of the wounds. DOE is helping speed up this testing period.

Using statistical methods before the experiment has even kicked off, the researchers determine how much data needs to be collected from the experiment to get a comprehensive view on the product’s effectiveness. For example, clinical trials can run with fewer participants and still get sufficient results and engineering experiments can run with fewer prototypes which lower costs.

The gel cream experiment aimed to explore how well the gel worked in response to various experimental factors, including how well it stayed on the wounds and how it aided the healing process.

Designed by Dr Mylona and Dr Olga Egorova, the experiment was performed by Aleksandra Olszewska, and Head of Institute of Pharmaceutical Science Professor Ben Forbes. The paper for this research is currently under preparation.

A similar method was used in a recent experiment developing an artificial eye. Dr Mylona designed an experiment involving participants assessing images and how people’s perception was affected by the presence of facial defects.

The number of people in the experiment was set based on the resources available and DOE helped to collect the best data based on those resources and the research questions.

“This project is about studying the perception differences. We obtained a design that would allow for answering multiple research questions efficiently and accounting for restrictions in the experimental setup.

“We gave advice on how particular sample sizes would affect the power of statistical tests and we recommended the minimum size overall.”

Dr Mylona has collaborated with Janssen Pharmaceutical Companies of Johnson & Johnson across several different experiments, including an optimisation experiment of the synthesis of an active pharmaceutical ingredient. She developed a framework for presenting results of statistical analysis of experiments with multiple responses and factors enabling the company to reduce the cost and time involved in experiments.

“This has directly impacted the company’s daily project work by streamlining communication with scientists and presenting the framework to the broader scientific community to share the benefits of the methodology,” Dr Mylona said.