How not to reduce uncertainties in care

Abstract

US Office for Human Research Protections messes up The randomised controlled trial has justifiably been embraced as necessary to the delivery of evidence based medicine. Randomisation reduces confounding by unknown factors, ensures every patient has a fair and equal chance of receiving the best (as yet unknown) treatment option, and is the gold standard approach to identifying effective treatments for future patients. In an ideal world every treatment uncertainty would be dealt with in this way. Recent experience in the United States highlights the unexpected barriers to doing this. Most uncertainties in healthcare relate not to new experimental treatments but to those already in wide use. The administration of oxygen to premature babies is an example of this. A longstanding uncertainty about the treatment—the optimum saturation target—was put to the test of a randomised controlled trial. Preterm babies with respiratory immaturity often need additional oxygen, but too much oxygen is associated with a proliferative retinal vasculopathy—retinopathy of prematurity—a cardinal cause of lifelong visual impairment and blindness. For this reason, the accepted standard of care oxygen saturation range of 85-95% is used to avoid levels …