Intervention Bias

Why is medicine prone to overdiagnosis and overtreatment?

Your first thought might be “follow the money.” Not a bad thought, but in fact there’s another force that’s more insidious: information.

Health care practitioners, whether conventional or alternative, get their information about medical practices from professional journals and colleagues, either directly or at professional meetings. You’ll notice that “from patients” isn’t on the list. There’s nothing mysterious here. It’s something we do ourselves: rely on trusted sources of information for the decisions we make in our professional lives.

How does that stream of information affect decisions practitioners make?

I spend a lot of time critically evaluating research because part of what I do is provide a countervailing force to received wisdom, both on this show and in my health education practice. In other words, part of my job is to encourage critical thinking. On the other hand, doctors and other health care practitioners use information from research to guide what they do to patients. Yes, I said, “what they do to patients” and not “what they do for patients.” As I said before, that information comes from professional journals and colleagues, who, of course, get their information from leading journals and colleagues, and so on. They do not evaluate this information critically so much as practically—what to do to patients.

Where does that information come from?

It comes from researchers toiling away in various domains, looking for practices that improve medical outcomes. This is where it becomes insidious because the channels through which research finds its way to practitioners is inherently biased toward intervention—that is, favors overdiagnosis and overtreatment. A recent study in the journal BMC Medical Research Methodology reveals how this bias in the structure of information works.

For decades, a group of researchers has studied what’s called “publication bias.” This means that what tends to get published is research that shows positive results. For example, you will not be surprised to learn that a study of influenza interventions showed that industry-funded studies tend to show positive results more often than independent studies and as a consequence tend to be published in professional journals more frequently. But even without industry loading, many other forces are at work to create publication bias—for example, career enhancing research needs to show results.

The BMC researchers examined a large number of meta-analyses, each of which looked for publication bias. (A meta-analysis is an examination of many studies intended to reveal the consistency and validity of the individual studies’ conclusions. In our book Too Much Medicine, Not Enough Health we discuss how this works, for example, with regard to why mammograms don’t save lives.)

Anyway, the BMC researchers put each meta-analysis in one of four groups. The first consisted of studies that followed research from when it first started through publication—that is, the projects entire life span. The second consisted of studies that followed research from the time it was presented to a regulatory agency—that is, allowing the project to be buried after it started. The third consisted of studies that started with presentation at a conference—even more of an opportunity to bury a project. And the fourth consisted of studies that simply looked at published studies—obviously having the greatest opportunity to bury research.

What the researchers found was that publication bias was most pronounced in the first two phases—that is, in the two phases that are furthest from the light of day. In other words, research that doesn’t show positive results is weeded out before anyone presents the results to colleagues at professional meetings or has them accepted for publication in a professional journal—the two places practitioners rely on.

Let me put this in a less abstract way. A urologist sees an article in the Journal of the American Medical Association about the benefits of contesting testosterone with some drug to treat prostate cancer. That information becomes part of the urologist’s repertoire. What that doctor doesn’t see is the research where the drug doesn’t work or the research that shows it might even make things worse. A naturopath might go through a similar process reading the Townsend Letter, substituting “herb” for “drug.”

The bias in this information should be called an intervention bias instead of a publication bias. That’s because the information that’s favored is information that tells practitioners how great it is to use that drug or that surgery or that herb or that nutraceutical or whatever the practitioner’s therapeutic proclivities might be. This has nothing to do with the practitioner’s good intentions or your good feelings for them. It has to do with an informational aura that shifts the guiding principle of care from “First do no harm” to “First do something.”

Permit me to remind you that our guiding principle of care is “Do the simplest things first.”

These issues are discussed further in our book Too Much Medicine, Not Enough Health.