antidepressant trials and why it was so interesting and in the process learned that the lead author - Seetal Dodd had a paper being reviewed on the nocebo effect in bipolar trials. That paper finally came out and I had the pleasure of reading it and presenting it here.
As a brief refresher, the nocebo effect is an adverse drug experience or worsening based on taking placebo or inactive medication in a clinical trial. At the clinical level, it can also be an unrealistic reaction to a medication based on similar response. Any researcher who has participated in clinical trials that all breaking the protocol and allow the subject to be informed about whether they were taking active drug or placebo has probably observed this effect. Clinicians commonly see it as an improbable reaction to a medication or in some cases multiple medications. It is an important phenomenon because it blurs the results of clinical trials by making it seem that there is less difference between placebo and the active drug being studied. It also may lead to the rate of actual adverse events due to the study drug being underestimated.
For the purpose of this study the nocebo response is defined as any a treatment emergent adverse events (TEAE) or clinical worsening in people treated with a placebo. That involves collecting data on both TEAEs and rating mania and depression.
The data used in this study was from a randomized placebo controlled clinical trial of olanzapine monotherapy for bipolar disorder versus placebo and several comparators (haloperidol, valproate, olanzapine-fluoxetine combination [OFC]). The trials occurred between 1996 and 2007. There were a total of 9 studies and 7 were published.
Only data for the patients randomized to placebo (N = 1185/4680) were used for the purpose of this study and meta-analysis. 866/1185 or 68% experienced (TEAE) and 4.6% discontinued the study due to a TEAE. Typical rating scales for mania and depression were used to rate symptoms. TEAEs were significant ranging from 3% to 11.8% of the placebo treated group. Headache, insomnia , somnolence, anxiety, nausea, diarrhea, irritability, and agitation all occurred in over 5% of the placebo treated patients. Median time to report the TEAE was 16 days with the longest time of 37 days.
Apart from the TEAEs, a significant number of the placebo treated patients experienced clinical worsening as noted on the rating scales for depression (321 or 27%), mania (585 or 49%), or global function (278 or 23%).
The 806 patients reporting TEAEs reported a total of 1,119 nocebo events.
TEAEs were associated with not being treatment naive, being obese, being located in the US, and participating in an earlier study. There were no significant difference based on gender or smoking status.
A major limitation of the study is that it is not possible to detect if clinical worsening is a nocebo effect or and the effect of worsening illness.
In their discussion the authors point out that the evidence is that the nocebo effect is significant in clinical trials. Of all of the possible correlates they studied they di not find any that were useful to predict who might be a nocebo responder. They discussed some psychological theories and in a couple of additional papers discuss the putative neurobiological underpinnings of both the placebo and nocebo effect. Certainly any effect that leads to 4.5% or placebo treated patients discontinuing the study is significant. Until the nocebo effect is better characterized we probably do not have an adequate estimate of the side effect profiles. The authors describe is as similar to the placebo response as "an expected consequence of exposure to therapy."
The nocebo response may be less clear in looking at the therapeutic effects of medication. I think it is good to remember that true nocebo/placebo effects are estimated on the placebo response. In other words a 20% placebo response rate means that 20% of the actively treated group also responded to placebo. Hence, if 68% of the placebo treated group experience TEAEs and 4.5% discontinue the study on that basis - what happened when those number are applied to the actively treated group? It would generally mean less subject carried forward in an intent-to-treat analysis and and a more favorable side effect profile if the nocebo responders could be accurately identified. A good place to start to look for more accurate numbers and methods of identification may be in large scale medicine studies looking at an identifiable quantitative endpoint like blood pressure. The nocebo effect is easier to sort out with a medication that has a clear effect on a more easily measured parameter.
Clinicians are left with estimating the likelihood of a TEAE or nocebo response based on the likelihood of a patient exhibiting a particular side effect. A low likelihood of a particular presentation is fairly frequent in clinical practice and continuing the medication in those circumstances often comes down to clinical necessity. As an example, the patient has a vague TEAE, but learns from the physician that there are no other medications that can be prescribed for the problem.
Nocebo responses are certainly out there. Clinically we can be more specific until there is better research guidance on what to do about it. I have had conversations with many people who were interested in the placebo response or why a particular pattern of responding to a therapeutic intervention may have been established.
George Dawson, MD, DFAPA
1: Dodd S, Walker AJ, Brnabic AJM, Hong N, Burns A, Berk M. Incidence and characteristics of the nocebo response from meta-analyses of the placebo arms of clinical trials of olanzapine for bipolar disorder. Bipolar Disord. 2018 Jun 21. doi: 10.1111/bdi.12662. [Epub ahead of print] PubMed PMID: 29926533.
Figure at the top is stock photo from Shutterstock per their licensing agreement by kasezo entitled:
Stock illustration ID: 284558927 conceptual 3d design of false pill.( placebo and nocebo effect.red and green colored version)