The
Zen of Science
Evidence
By MARC RINGEL, MDAs
I love writing for Nexus readers, addressing a smart, literate
audience that's concerned about health. These columns are a real
stretch for me, though. Usually when I talk about healthcare,
I'm the unconventional one in the conversation. With the Nexus
crowd, I play straight-guy doctor. The consistent aim of the essays
I've written for this magazine has been to discuss matters of
health and illness in a way that balances the truths of science
with the insights of spirit. In this installment, however, I will
concentrate almost solely on the science part. My goal is to inform
those readers who may not be conversant with the scientific method
as it applies to medicine, about how my profession arrives at
the (sometimes short-lived) objective truths on which we stake
your life.
A basic standard for clinical research is called the randomized
trial. With this research design, study subjects are randomly
assigned to either an intervention group or to a control group.
If the researchers wish to examine the effect of a drug on a particular
disease, a group of patients with that diagnosis is gathered.
Intervention subjects get the drug or treatment being evaluated
and control subjects get a placebo, an inert substance. Scientists
then compare the outcomes of the two groups. If a large enough
fraction of the patients who received the treatment did better
or worse than those who got the placebo (and if the groups were
big enough to make the results statistically convincing), the
researchers can conclude, to a precisely defined degree of certainty,
whether patients such as these will be better or worse off when
given the drug in question. The randomized clinical trial is the
workhorse of medical research.
When I explored the history of medical research as I prepared
this article, I was surprised to find that the first randomized
medical trial wasn't reported until 1946, when the British Research
Council, using statistical techniques borrowed from agronomy,
studied the effectiveness of streptomycin in treating tuberculosis.
Streptomycin was the first modern anti-tuberculosis drug. In 1946
the antibiotic was expensive to make and in short supply. So the
British government sponsored a study to show definitively if the
drug was worth the investment. It was.
The next step in sophistication of experimental design, and
in potential reliability of results, is called double blinding.
This technique earns its peculiar name by keeping both subjects
and researchers in the dark as to who gets the real treatment
and who gets the placebo. If you want to measure the objective
effect of a medication, you need to separate how much subjects
improve because they expect to get better from how much they improve
because of the drug's actual physical effects. Likewise, you want
to filter out how much patients improve because their doctor expects
them to get better. Hence, double blind: blind patient and blind
doctor.
There's a whole discussion about the utility of the placebo
effect, which some scientists estimate accounts for, on average,
about half of drugs' effectiveness. (For a more ample examination
of placebos, see "The Zen of Science" in the March/April 2004
Nexus.) Briefly, why should I care if my patient feels better
because I prescribed the right medication or because she thinks
I did? Just so long as she's getting better.
Then there's the potential benefit of professionals' expectations.
From the world of education research comes a study in which teachers
were told that certain children were expected, based on objective
tests, to raise their IQs during the next school year. Sure enough,
those kids from whom the teachers had been led to expect big improvements,
did indeed perform better on IQ tests. In actuality the predictive
tests were a complete fiction, with children randomly assigned
to the improvement and to the control groups. You might ask the
same question about blinding the researcher as you ask about blinding
the subject. So what? If the kids performed better wasn't that
the point?
Medical practice is different from medical research. I count
on the placebo effect to help virtually every patient I see. No
matter how well chosen my prescription, each time I place that
small page of hieroglyphics into a patient's hand I seek to enhance
its effect by expressing my confidence that the medication will
help. On the other hand, when I review medical research, I prefer
it be double blinded.
In the early 1990s, the term "evidence-based practice" started
appearing in medical journal articles. The authors of these pieces
set out principles for doctors to use in weighing research reports
and translating them into practice. As if this were a new idea!
Ever since 1910, when the report commissioned by the Carnegie
Foundation and penned by Abraham Flexner, entitled "Medical Education
in the United States and Canada," was published, American medical
education has been 100 percent devoted to an academic, scientific
conception of the healing endeavor. Every medical student learns
from the get-go how to read and interpret research reports.
There are uncountable ways that research can go wrong: too few
subjects; non-comparable intervention and control groups; broken
blinding; inappropriate statistical analysis; unsupported conclusions,
etc. It takes lots of skill to separate reliable from unreliable
studies.
No matter how well they were designed and executed, the conclusions
of some medical studies are reversed with regularity. Just look
at how recommendations for post-menopausal hormone supplementation
have taken a complete 180 degree turn in the last couple of years,
with both pro, and now con, positions based on well-done research
performed on thousands of women.
Finally, the most reliable study conclusions may not apply to
a certain patient. For example, until recently researchers into
coronary artery disease had been hugely biased toward studying
men, casting considerable doubt on the applicability to women
of the findings of their investigations.
As a graduate of an American medical college, I have been well-trained
in the art of evaluating medical research. Busy practicing doctor
that I am, though, there is no way I have nearly enough time to
apply those evaluation skills to even a tiny fraction of the studies
pouring forth every day that might be pertinent to my practice.
So, how do I make what I do with patients as evidence-based as
possible?
I call my approach to the medical literature the "bottle of
wine method." I am no oenophile. Don't desire to become one. Don't
have the money to spend on all those expensive vintages I'd need
to taste to educate my palate, nor the time. So I learn who are
the good vintners, read the labels with the stars on them under
the bins on the shelves of liquor store and ask a trusted salesperson
for advice. When it comes to medical literature I rely on well-written
articles that review the latest research published in the most
reputable journals. And I ask for advice from my expert colleagues
who are specialists in the fields I need help in. Fortunately,
a number of excellent websites have sprung up that are dedicated
to presenting the best of the best of medical evidence and translating
it into clear recommendations for evidence-based practice.
Still, it's estimated that at most 30 percent of what a doctor
does is supported by well-vetted research. The rest is grounded
in experience, reasoning from what we know of anatomy and physiology,
the best studies available (though these may be second-rate) and
gut instinct.
Which is why I get along so well with the Nexus folks. I have
no illusion that I, nor the discipline to which I've devoted my
adult life, has all the answers. Nor does anybody else, not even
at Nexus.
Marc Ringel, MD, is a family practitioner and writer based in
Greeley, Colorado.
