Modelling Clinical Trials
London Euston station is a child waking up. Shutters raising. Bleary-eyed passengers staring at the information board. I am catching an early train, for which I’m early.
WH Smith bright and inviting. On the shelf the US edition of Wired. What's inside? Yes: The Body Synthetic. Something to enjoy on the journey.
In 1997 David Eddy applied a model he designed called Archimedes to predict the result of a clinical trial named Collaborative Atorvastatin Diabetes Study or CARDS.
Archimedes is like a Treasury Model of human physiology: knowledge from epidemiology, clinical trials and clinical interviews encoded into differential equations.
Using broad patient data, Eddy and his team constructed a comparable trial to CARDS in 2 months and issued the results. When the 7-year clinical trial reported, it turns out the findings of the model were remarkably close.
It is also remarkable to read the criticisms of the model: a self contained ant farm that has no bearing on reality, a black box. Interesting remarks, because the whole of medical practice is based on abstractions, assumptions about biological reality. Do we fully understand the complexity of human physiology? We do not.
Medicine is often quick to point to the mote in the eyes of others (black box, lack of evidence), while forgetting the beam in its own (black box, lack of evidence).
Nonetheless, it’s refreshing to read about people prepared to face criticism and innovate with IT. To open eyes. Like shutters raising in the morning.