Bausell begins with a typology of CAM, classifying it into five kinds: indigenous medical systems such as traditional Chinese or Ayurvedic; nonindigenous medical systems such as osteopathy, naturopathy, homeopathy, and chiropractic; spiritual/energy healing; relaxation-oriented therapies; and extensions of conventional science such as chelation, biofeedback and "natural" drugs.
Placebos moved properly into the realm of science with Henry Beecher's famous 1955 paper "The Powerful Placebo". Bausell considers a number of different definitions, with his favourites including: "any genuine psychological or physiological response to an inert or irrelevant substance or procedure" and "any medical treatment that can have a therapeutic effect only if administered to a patient who is aware that he or she is receiving a medical treatment".
Three chapters are devoted to the problems of making valid causal inferences, looking in turn at the "natural impediments" that affect patients, practitioners and therapists, and scientists. Patients can easily be misled by the natural history of their ailments, with naturally fluctuating levels of pain mistaken for treatment effects. Practitioners are misled by biases in the selection and attrition of patients seeking help. And scientists can be misled by patient natural history, unconscious experimenter bias, outcome selection, regression to the mean, and patient behaviour changing under observation (the Hawthorne effect), among other factors.
The best science involves randomized placebo control groups, but these are not always easy to implement: placebo treatments for manipulation therapies (or drugs with side effects) are difficult to implement and, compounded by the need to obtain informed consent, can result in incomplete blinding and differential attrition. So Bausell considers some general rules for judging the credibility and plausibility of scientific evidence.
After this background, and a brief account of his own experience researching acupuncture, Bausell begins his argument proper. He begins by "proving the obvious", with tests of the placebo effect and of the role of conditioning. From these he concludes:
"The placebo effect is real and is capable of exerting at least a temporary pain reduction effect. It occurs only in the presence of the belief than an intervention (or therapy) is capable of exerting this effect. This belief can be instilled through classical conditioning, or simply by the suggestion of a respected individual that this intervention (or therapy) can reduce pain."
It is probably only responsible for part of placebo effects, but experiments reveal a likely biochemical explanation for placebo pain reduction, involving the release of natural opioids.
"The placebo effect has a plausible, biochemical mechanism of action (at least for pain reduction), and that mechanism of action is the body's endogenous opioid system."
Returning to CAM, Bausell analyses high quality trials — that is, randomised trials with a credible placebo control group, at least fifty participants, and an attrition rate under 25%, published in high-quality peer-reviewed journals. Out of twenty one such trials, there was only one positive result, which is about what you'd expect by chance, testing at a 0.05 level of significance.
An alternative approach is that taken by the Cochrane Collaboration, to systematically review all the trials of a therapy and to analyze them together. In addition to general trial quality concerns, the biggest problem with such reviews is publication bias: positive results are vastly more likely to get published than negative ones. Surveying nearly one hundred of the best quality reviews of CAM therapies reveals some notionally positive results, but these are always qualified by extremely weak support or poor quality evidence. Bausell concludes:
"There is no compelling, credible scientific evidence to suggest that any CAM therapy benefits any medical condition or reduces any medical symptom (pain or otherwise) better than a placebo."
A chapter then considers the putative mechanisms of CAM treatments, which range from the manifestly ridiculous such as homeopathy to the perfectly plausible but unverified such as chiropractic.
"No CAM therapy has a scientifically plausible biochemical mechanism of action over and above those proposed for the placebo effect."
A final chapter considers some of the broader factors behind pathological science, looking at the role of the media and education. Bausell finishes with some tips on how to make best use of the placebo effect in selecting a CAM treatment and therapist.
On the surface Snake Oil Science is, as the title and subtitle suggest, a debunking of CAM, but it is more fundamentally about the nature of scientific inference and methodological issues in the evaluation of medical therapies. It doesn't address drug development or government regulatory regimes at all, but much of it applies just as well to evaluation of clinical studies by big pharmaceutical companies — some of the social issues are different, but the underlying epistemological issues are the same.
Snake Oil Science is unlikely to educate, or even be read by, CAM adherents or practitioners from the more anti-scientific end of the spectrum, but will be useful for those with more empirical approaches. It could also be read by those without any particular interest in CAM, seeking an explanation of why blinded randomised placebo trials are important and how they work.
- Related reviews:
- Trick or Treatment?: Alternative Medicine on Trial
- books about health + medicine
- more popular science
- books published by Oxford University Press