Ever notice how yesterday’s urgent medical advice is today’s critical error? Kayt Sukel covers the reasons in NewScientist (27 August 2016, paywall), and this particular reason caught me by surprise:
Among them is the fact that you’re often working on a moving target. “Diseases aren’t static – they can and do change over time,” says Gerstein. “Diabetes today is not the same as it was 50 years ago.” The number of people diagnosed with type 2 diabetes has more than doubled in the past two decades, and the age of onset has dropped. He says today’s type 2 diabetes is more likely to come with additional health issues like cancer, kidney problems and heart disease, compared with 50 years ago. “We call it by the same name but it behaves differently,” says Gerstein.
Other reasons include enthusiasm for new therapies from sufferers, poor early study design, commercial influences, and unfortunate use of “surrogate outcomes”, or what I tend to call proxies, rather than actual outcomes themselves. While there has been wide encouragement that all studies be published, regardless of outcome, and that all data be published, along with methodologies of data collection and analysis, I have not noticed anyone suggesting pre-study analyses. That is, every proposed study should also be published, once it reaches a certain point in the process of doing a study. The purpose would be to give peers the opportunity to critique the study’s various facets before they are executed. The advantages would be to eliminate weak and invalid methodologies.
However, this might not fly. Researchers are already extremely busy, and being asked to critique a proposed study may end up at the bottom of the priority list; and people qualified to critique are probably mostly doing research, so it’s unlikely that there’s a cohort of highly qualified scientists and techs who could perform this work.
But it might be worth making it work.
(I’ve also wondered if grad students are taught good study design, or if they’re expected to just pick it up from their advisors. Anyone know?)
