September 4, 2013
Basics for Evaluating Medical Research Studies
A new book, “Basics for Evaluating Medical Research Studies: A Simplified Approach,” according to its authors, “represents a summation of our decades of work in this field.”
The authors are Sheri Ann Strite and Michael E. Stuart, MD – two people I’ve had occasional communication with over the past year or so. You can read about them and their work on their Delfini.org website.
While they say the book is written for physicians and other health care professionals, they also intended it to be written in terms “that even the layperson can understand and put to use.” Journalists should check it out, along with the collection of tools on the Delfini website.
The book is only 112 pages. It won’t overwhelm you.
But it will educate you on many of the themes we touch on so often on this website as we analyze media messages about medical research studies. Themes such as:
- problems with observational research
- limitations of surrogate markers…surrogate endpoints…intermediate markers…proxy markers
- bias, confounding and chance in studies
- intention-to-treat analysis
- absolute versus relative risk
- number-needed-to-treat (NNT)
- confidence intervals
Strite and Stuart also provide examples to help educate. For example:
- “What if we told you that we believe that the misreading by many doctors of two sentences in a published research study abstract may have contributed to an estimated 27,785 heart attacks and sudden cardiac deaths over a four year period of time? … Somehow many physicians managed to overlook the data on heart attack that were right there in the abstract. Unfortunately (amazingly!) physicians didn’t seem to notice that it was possible that Vioxx was actually causing heart attacks in one out of 103 people taking the agent.” (The New England Journal of Medicine 2000 paper on the VIGOR trial of Vioxx.)
- “The encainide/flecainide example is instructive not only about the risk of assuming the relatedness of an intermediate marker to a clinical outcome, but also about what can happen when treatments become established through simple observations or through pathophysiological reasoning instead of through high quality science.”
- “Blinding can even matter when studying objective outcomes such as mortality. The RECORD trial (Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycemia in Diabetes) is an example of an unblinded or open-label trial with blinded assessment in which lack of blinding led to a suspected biased mortality assessment that appears to have significantly altered results.”
The book is available electronically or in paperback.
Follow us on Twitter:
and on Facebook.