Populations, Individuals, and What the Doctor Tells the Patient
We are awash in medical advice. Our doctors quote the latest studies on the virtues or hazards of exercise, alcohol, fruits, coffee, breads, statins, proton-pump inhibitors, antidepressants, sleep, broccoli, and almost every other matter of medication and lifestyle. Much of this comes from well-designed randomly controlled trials (RCT, “the gold standard”), often published in the top journals. Doubts have been raised about the validity of many of these studies, including a built-in “publication bias.” But even if we accept the validity of the trial, the correctness of the analysis, and the wisdom of the editorial board, how are we to interpret population statistics in terms of individual patients? The response of a person to a treatment depends on many things, most of which are not known to the investigator, and in any case can not be controlled by either the investigator or the individual.
I will suggest a simple thought experiment as a way to bound the probabilities of significant benefit and significant harm resulting from the treatment of an individual. I will apply these bounds to treatments shown to produce positive outcomes at the population level, and reported on in our best medical journals. I will argue for new metrics to help doctors and patients make rational choices.
Stuart Geman received his B.S. in physics from the University of Michigan, his M.S. in neurophysiology from the Dartmouth Medical School, and his Ph.D. in mathematics from the Massachusetts Institute of Technology. He is currently the James Manning Professor of Applied Mathematics at Brown University.