Snake Oil Science: The Use of Placebos in Research
This morning we presented a post from R. Barker Bausell, author of Snake Oil Science: The Truth About Complementary and Alternative Medicine , in which he argues that the placebo effect has as much healing power as alternative medicine. Below, in an excerpt from Bausell’s book, we learn about the history of the use of placebos in scientific research.
…The placebo effect itself escaped serious scientific scrutiny until 1955, having largely been considered prior to that time to be more a part of medical lore (or physician mystique) than a documented clinical entity. In that year, Dr. Henry K. Beecher published an article in the Journal of the American Medical Association titled ‘‘The Powerful Placebo’’ and, in the process, moved the placebo effect from the realm of myth into mainstream science.
Beecher’s research was relatively simple. He reanalyzed the results of fifteen clinical trials that had employed a placebo control group and concluded that 35 percent of the patients who had received a placebo responded positively. The article became an instant classic, and while a number of scientists have justifiably taken issue with his methods, Beecher’s study proved quite influential in changing medical research practices because of the logical question it raised:
If patients participating in a clinical trial can improve simply because they believe they are receiving an effective medical intervention, how can anyone have any confidence in the results of any clinical trial that did not employ a placebo control group?
Taken to its logical extreme, this single finding implies that any nonharmful therapy evaluated by a trial without a placebo control group has the potential to produce positive results due to the placebo effect. In fact, even trials that employ other types of control groups (such as routine medical care) would also be suspect because the new therapy (effective or not) will produce a placebo effect, while patients who know they aren’t receiving this innovative therapy will not have the benefit of a placebo effect.
For Beecher, the solution was patently clear: when a clinical trial is being conducted, patients must not be allowed to know whether or not they are receiving the experimental therapy. And since physicians are experts in producing placebo effects by carefully engendering positive expectations in their patients, it is important not to let them know, either.
The most effective way to avoid letting these cats out of their bags, Beecher reasoned, is to employ placebo control groups and to allow neither patients nor their providers to know who receives the experimental therapy and who receives the placebo therapy. Called ‘‘double blinding,’’ this strategy soon became a scientific requirement for serious clinical trials and has since been employed for almost all types of medical interventions. Of course, creating credible placebo control groups is a lot easier in investigations of drug interventions (since a placebo capsule can be manufactured that is indistinguishable in taste, smell, and appearance from the active drug under evaluation) than is the case for other types of therapy. Still, though it is more difficult to guarantee the credibility of nonpharmaceutical placebos, these have now been successfully employed in everything from surgery (where a sham incision is performed under anesthesia) to acupuncture (where pointless needles appear to be inserted, or real needles are inserted in irrelevant locations).
It is almost solely within the scientific experimental arena, in fact, that the placebo effect possesses any importance in modern medicine. By comparison, the active use of placebos in patient care has declined, which may be partly due to a medical ethicist named Sissela Bok who, not long after Beecher’s study, published an almost equally influential article in Scientific American titled ‘‘The Ethics of Giving ] Placebos.’’ Bok’s conclusion was that their use in clinical practice was indeed unethical, and so today few if any hospital pharmacies stock placebo pills for anything other than their clinical trials.
Just because physicians no longer have sugar pills in their medicine cabinets, however, does not mean that they do not freely take advantage of the placebo effect in their day-to-day practices. A recent article in a peer reviewed journal that I have edited for thirty years indicates that 86 percent of Danish general practitioners admitted they had employed a placebo treatment at least once during the past year, and 48 percent had used them ten times or more.6 The most common uses (and remember, the practitioners wrote prescriptions or recommended these therapies knowing full well that they would have no effect on the conditions for which they were prescribed) were:
1. Antibiotics for viral infections
3. Sedatives for conditions other than depression
4. B vitamins for conditions such as multiple sclerosis and hair loss
5. Saline injections
There is little question, though, that the use of placebos as a treatment option has definitely fallen from favor in the United States over the past several decades. It is primarily to research specifically targeting the placebo effect, therefore, that we owe much of our rapidly growing knowledge about this effect and the conditions under which it occurs.
Undoubtedly a considerable amount of the impetus for this work can be attributed to Henry Beecher, because in science once a technique has been developed for reliably measuring something, it is as sure as death and taxes that that phenomenon will be the subject of further inquiry— and the placebo effect is no exception. There have been a plethora of studies performed on the phenomenon since Beecher’s time, most of which have been designed to directly investigate the conditions under which the effect occurs, the methods by which the effect can be magnified, and the physiological mechanisms of action within the body that cause the set of reactions that we call placebo effects.
Ironically, in science, an almost universal result of conducting research important enough to encourage more work in the same area of inquiry is that the initial investigators will later be proved wrong, whether the subject matter is the placebo effect or quantum mechanics. It is even said that the ultimate compliment that any scientist can receive is to have done work that is important enough to encourage someone else to attack it.
By this standard, Henry Beecher’s work was indeed important. Reanalyses of his methods and conclusions have demonstrated that his landmark study was so flawed that it did very little to actually establish the existence of the placebo effect, much less its proposed 35 percent effect size. One reason for this failing is the still relatively common misconception that any gains or improvements observed by a placebo control group represent nothing but a placebo effect. Instead, what such gains represent are the effects of that entire class of logical and psychological impediments to logical thought that I alluded to earlier. True, the placebo is a prominent member of this community, but so too are artifacts such as the natural history of a disease (that is, the tendency for people to get better or worse during the course of an illness irrespective of any treatment at all), the fact that people behave differently when they are participating in an experiment than when they are not, a desire to please the experimental staff by providing socially desirable answers, and numerous other villains that I will discuss in detail later. What this means from a research perspective, therefore, is that as a general rule patients enrolled in clinical trials will witness some improvement over time regardless of whether the therapy being investigated works or not.
What Beecher apparently did not understand was that the only way to prove the existence of a placebo effect and to measure its size (at least in research designed to measure the effectiveness of a medical therapy) is to employ two control groups in the same experiment: one in which participants receive a placebo and one in which they receive no treatment at all.
While this is a relatively rare strategy, it is occasionally used, and the science of systematically locating such trials (and reanalyzing their combined results) has come a long way since Henry Beecher’s time. Some years ago two Danish researchers (one of whom, Asbjorn Hrobjartsson, published the survey that I just discussed) were able to locate 114 clinical trials in which both types of controls were employed in the same study. The authors, reporting their results in the New England Journal of Medicine, concluded that they had found ‘‘little evidence in general that placebos had powerful clinical effects’’ except for one notable exception: pain relief. Here, in the twenty-seven trials that involved this outcome (and encompassing a total of 1,602 patients), Hrobjartsson and Peter Gotzsche found that the difference between patients receiving no treatment at all and those receiving a placebo was indeed statistically significant in favor of the patients receiving placebos—hence confirming Beecher’s original contention (at least for pain) by employing a considerably more rigorous methodology. (I like to think that this finding supplies a happy footnote to Beecher’s distinguished scientific career. His research was important enough to generate more work, and the results of this work yielded the almost unprecedented benign judgment that his conclusions were largely correct—just for the wrong reasons.)
As is common practice in this field, Hrobjartsson and Gotzsche’s conclusions were almost immediately attacked on the basis that they had underestimated the potency of the placebo effect by ignoring some studies with controls that actually offered some substantive formof treatment rather than no treatment at all. All this ironically gives some credence to the conclusion reached by another set of researchers, Gunver Kienle and Helmut Kiene: ‘‘The placebo topic seems to invite sloppy methodological thinking.’’ But let’s not quibble. While there is some excellent additional research suggesting that placebos can influence a number of conditions other than pain (such as depression, which—like pain—just happens to be measured by self-reports), the evidence is as close to I controvertible for pain as we customarily get in science.