This is the next installment of our look at the increasingly high placebo response that is plaguing clinical trials in analgesia and psychiatry. Check out the other posts here.
Over the past few weeks, we’ve discussed the psychological, neurobiological, and genetic mechanisms responsible for the placebo response. Today, we turn to the study designs used in this research and their associated benefits and drawbacks. But first, an important disclaimer: the trial designs we’ll discuss today are only tools used by researchers to study the placebo response and not for use in randomized clinical trials for support of approval of drugs by FDA or EMA.
Clinical trial designs
A common clinical trial design used in placebo research is a parallel group design in which patients are split between three groups: active drug, placebo, and a no treatment control group. The no treatment group allows placebo researchers to tease out the contribution of trial-related factors from two other trial-independent factors that are rarely mentioned in discussions of the placebo response. The first is the natural tendency of diseases to improve over time, and the second is the statistical phenomenon of regression to the mean, where outliers in a data set subsequently move towards the mean. We’ll cover these extrinsic, trial-independent factors in more detail in a later post in the series.
By comparing the difference in response between the placebo and no treatment group, researchers can assess the true placebo response, while ruling out these extrinsic factors. For ethical reasons, few clinical trials designed to test a novel drug or other treatment utilize a no treatment group—in fact, no treatment arms are never included in clinical trials we conduct—and mainly consist of one or more active treatment arms compared to a placebo arm. Therefore, it is challenging to obtain rigorous estimates of the magnitude of the placebo response in most clinical trials. Meta-analyses of the placebo arms of multiple clinical trials can shed some light, but they are subject to a number of limitations and caveats.
Another trial design that can potentially be used to study placebo responses is a cross-over design, in which each subject serves as his or her own control and participates in both the active treatment and placebo groups at different phases of the trial. However, the utility of this design in placebo research is limited due to confounding learning effects—prior exposure to the drug or placebo affects subsequent placebo or drug responses, respectively. As we learned in our third post, learning and past experiences are major components of the psychological mechanisms that underlie the placebo response.
Classical conditioning also plays a significant role in the generation of the placebo response. Two common conditioning paradigms are used in placebo research to maximize the placebo response in order to study the underlying brain activity or pharmacological contributors that we discussed in our psych post.
In a response conditioning design, verbal instructions are used to condition subjects. For example, two topical creams are applied to participants: one they are told is an effective analgesic and one that supposedly has no pain-relieving properties. Subjects are then given a painful stimulus at the site where the control cream was applied and a less intense stimulus where the “analgesic” was applied, but told both stimuli are equivalent. In the third phase of the experiment, an equally painful stimulus is applied to both sites. The placebo response is measured as the difference in pain levels between the two sites. This response conditioning design is the most common one used in neuroimaging studies.
In a pharmacological conditioning design, subjects are conditioned with a drug. In half of participants an active drug is paired with verbal and nonverbal cues over several days; the other half of subjects receive the same cues but without the drug. The cues are then presented alone and the placebo response is measured by the difference between the drug-paired and non-paired conditions. Many studies that have measured the role that various neurotransmitter systems such as endogenous opioids and endocannabinoids play in the placebo response have done so by adding an additional phase to the pharmacological conditioning design in which subjects are given drugs that inhibit or potentiate these systems. The blockade or enhancement, respectively, of the conditioning becomes evidence that these systems are involved.
Join us next week for a look at how drug and placebo effects interact. Is the response of the drug group in a trial simply the pharmacological action of the drug plus the placebo response (the additivity hypothesis), or is the relationship between the two more complicated? We’ll examine whether the additivity hypothesis, first proposed in 1955, still holds up, and discuss the implications for the interpretation of clinical trial data.