Heterogeneity of the Placebo Effect

None of the common drug store remedies cured my warts. I was a 10-year-old kid barely at the age of reason. Our neighbor offered a cure based on a secret ritual. In a dank basement well of his house he gathered some snails and placed them on my finger until the warts were covered in slime. He accompanied this treatment with a mysterious incantation. Within a few days the warts were completely gone! After all these years we still get a good laugh at this at family reunions.

When we talk about the heterogeneity of the placebo effect, we refer to differences we see in clinical effects when comparing two otherwise equal treatment groups. Each group unknowingly receives an inert substance (e.g., sugar pill) with an explanation of some medical benefit we expect to see from the treatment. We do this within an environment that strictly controls for external influences (e.g., randomized, double-blind, compliance-monitored studies). In the end we see a significant difference in both positive clinical effects (placebo) and negative effects (nocebo) from one group to the next.

If I had a few peer-reviewed studies that measured this exact difference I would end this article here. But instead I have to resort to a larger story to show the plausibility of this conjecture.

First, there is much documentation on the reduction to the placebo effect from well-controlled studies over non-controlled studies: upwards to 90% difference. The difference is higher for subjective clinical effects (e.g., migraines) than for physical effects (e.g., broken bones).

Second, there is much documentation on the reduction of the placebo effect due to individual components of well-controlled studies: randomization, blinding of the patient to the treatment, blinding of the investigator to the treatment, etc. It’s clear from a synthesis of these articles that there is still much room (tens of percentage points) for a contribution by the innate placebo tendencies of individuals. This tendency manifests itself in measurable physiological changes, and is not just a figment of the patient's imagination.

Third, this innate tendency is a mixture of dozens, if not hundreds of variables in an individual’s makeup, experience, and current condition. See Vallance for an extensive listing of these variables. At any point in time individuals can be suggestibles, hypochondriacs or over-responders. There is no one placebo effect, rather a constellation of effects that are generalized under this rubric. So any one individual can show an innate placebo tendency that ranges from 0 to 30% or more; and this individual can vary in effect on a day-by-day basis.

Finally, we now need to disentangle the measurable clinical effect of our medicine to be tested from the natural tendency of the placebo effect. Active drugs often heighten the placebo effect over inert substances: individuals sense some physiological change to the active medicine and jump to the conclusion they are not receiving the placebo. This conclusion induces a heightened awareness to any change in physiology.

The conjecture then is that the heterogeneity of the placebo effect means innate placebo tendencies of individuals can contribute 10-20% to the measured clinical differences between the placebo group and the group receiving the actual medical treatment. This is a difference that comes about despite all the industry-standard controls for bias. The innate placebo tendency of individuals is not and cannot be deliberately randomized across the two groups in a comparison study.

These tendencies therefore place a limit on the precision one can expect from any comparison study: upward to a 20% hurdle. If your drug doesn’t show a greater improvement than this over its comparator, then the study should be disregarded merely due to the heterogeneity of the placebo effect. You argue that well-controlled industry studies rarely show 10-20% differences, belying the extent of the heterogeneity; I argue in counter the heterogeneity is highly variable, can go either way, and even at lower amounts swamps and invalidates any purported clinical effects of the tested drugs.

Further Reading