EconTech

Apparently, it doesn’t get funding:

Why Medicine Should Care Less About ‘Sick,’ More About ‘Normal’

If you had died 50 years ago, your body would have stood a pretty good chance of serving science. In the 1960s, autopsy rates at US hospitals exceeded 50 percent. Pathologists weren’t necessarily looking for what killed people — they were taking advantage of the fact that a body was available and ready for inspection. There was still much to learn about normal human biology, the thinking went, so every corpse was an educational opportunity.

These days, autopsy rates have fallen below 10 percent, a decline that’s symptomatic of a larger deficiency. Medicine has become all about finding a problem — a tumor, a heart attack, a failing kidney — and deploying advanced treatment technologies. In the process, we seem to have given up on measuring and tracking what constitutes normal. That’s an alarming — and potentially dangerous — trend.

What’s normal matters because we’re entering a new era of health care, one in which we look not for causes of illness but for risks. It’s called predictive medicine, and its primary tool is the screening test. A good screening test should provide a range of results, distinguishing between a condition within normal parameters — which doesn’t require intervention — and an anomaly, which demands it. That’s how most blood tests work, for instance. But for all sorts of conditions, there’s often no definition of normal. In heart disease, for example, CT screening tests can spot abnormalities in arterial plaque — but no research exists on whether that information is actually predictive of heart disease or stroke. “We need to know normal variation,” says Pat Brown, a professor of biochemistry at Stanford University School of Medicine. “It’s really underappreciated as a part of science.” In too many areas, Brown argues, we’re too quick to jump at any blip without understanding whether it’s a true red alert or just normal background noise.

Consider prostate cancer. Right now, about two-thirds of men diagnosed with the disease get treated with surgery or radiation (both of which carry a significant risk of impotence or incontinence). But in February, researchers at the Cancer Institute of New Jersey found that 80 percent of men over age 66 with detectable prostate cancer who do nothing (so-called watchful waiting) will likely die of something else. In other words, most of those who get treatment — and could be impotent as a result — should have gone without it. “We’re way overtreating the disease,” says Peter Nelson, an oncologist at the Fred Hutchinson Cancer Research Center in Seattle. “Really, you only want to know about the ones that are potentially fatal.”

Ironically, this problem is brought on by technology. Imaging and scanning tools are now so good at peering inside our bodies, they’ve surpassed our capacity to interpret the results. Many findings are what doctors call “incidentalomas,” smudges that look like cancer but turn out — often after surgery — to be benign. Though new detection technologies like proteomics have made great progress in associating particular biomarkers with certain cancers or diseases, we still don’t know how often those same markers turn up in nondisease situations.

It seems like it would be easy just to step back and survey the broad picture. But research costs money, and studying what’s normal is generally considered trivial, dismissed as mere butterfly collecting. At the National Institutes for Health, for instance, all grants are given a “priority score,” an indication of a project’s novelty, originality, and “scientific merit.” Normal need not apply.

But in these data-rich days, studying what’s normal could be a project of startling originality and merit. With petabytes of storage and ample processing power at hand, there’s an opportunity to create a sort of Normal Human Project — a macro understanding of human biology on a micro scale. Or, as Stanford’s Brown describes it, a “comprehensive, quantitative molecular and cellular characterization of the normal human.’” That may sound daunting, but complementary projects are already under way. Seattle’s Allen Institute for Brain Science, having completed a 3-D model of the mouse brain in 2006, is now aiming to model the human brain in its normal state. Even postmortem examinations are coming back into vogue, via high-volume autopsy centers, which can add their results to resources like Johns Hopkins’ online autopsy database.

The annals of medicine are full of tidy explanations of how the body works, from Dr. Atkins all the way back to Hippocrates. Inevitably, though, someone comes along and shows that there’s a little more to it. It would be wise, as Brown suggests, “to accept the fact that we don’t know a tremendous amount about things we think we know. We could learn some humility.” That, however, may be asking too much of science.

I have mentioned that I am a big proponent of electronic medical records a number of times, though with the caveat that there are serious privacy concerns that software makers have thus far failed miserably at addressing properly (not to mention the interface issues which medical personnel complain about). This is yet another area where they could shine. By having a full medical profile linked up with various causes of death, you can get a good idea of what a normal functioning X organ system looks like when it is an isolated case of Y killing the person.

I have seen a number of articles in the past (though nothing specific comes to mind) showing that many of aggressive approaches to these ‘blips’ in the body’ background noise leads to a higher number of expected deaths than simply waiting and watching. This is a solid piece of evidence in favor of the need for evidence-based medicine. [Did some Republican hatchet man come up with that name? It's not like current medicine is not based on evidence, just that this new school, more or less, wants a more formal approach to the data.]

Depressingly, not only may the rate of innovation in the medical sector be causing negative externalities with regard to health care costs and the provision of health care to the poor and the capability of our physicians to provide care without the use of high tech capital, but the response from the information from that high tech capital may be killing people.

On a related note, a lot of studies on the need to decrease the variation in medical care due to the cost savings such a reduction could provide seem to fail to take into account the endogenous learning processes that move knowledge upstream from the clinic to the research lab as different physicians see different treatments work in various ways. A model that takes this into account and gets data run through it would be most interesting. If anyone has seen such a study, definitely pass it on.