PADRAIC 7: Expert elicitation of weight loss on the risk of cancer, using the SHELF method
In collaboration with the University of Sheffield, we plan to undertake an expert elicitation using the SHELF method, available here: https://shelf.sites.sheffield.ac.uk
Methods
A two-step in-silico trial has been designed, with prior distributions taken from real-world clinical trials, to simulate an individual’s weight loss depending on the intervention. The individuals weight loss is then assigned a probability of an incident cancer during the duration of a clinical trial. The prior distribution representing the relationship between individual weight loss and cancer risk is very poorly defined in the literature with a high degree of uncertainty. We compare the effect on trial assurance of the selection of 3 plausible prior distributions for this effect demonstrated in figure 1. We then undertake expert elicitation of this distribution using the SHELF framework, comparing the impact on assurance as a single value and a distribution of probability.
Results
A two-arm in-silico trial comparing a weight loss drug (mean 15% body weight loss) to a control group (mean 2% body weight loss) ,using a total sample size of 1500 randomised equaly between arms with a binary primary endpoint indicating cancer incidence at 10 years.. High and low plausible extremes gave an unconditional probability of observing a statistically significant result of, 99.9% and 37.44% respectively. This highlights the impact of variation of this value, motivating the SHELF elicitation exercise planned for February 12th 2025 and involving 8 experts.
PADRAIC 