Articles

The Placebo Gene?

A study recently published in PLOS one (Catechol-O-Methyltransferase val158met Polymorphism Predicts Placebo Effect in Irritable Bowel Syndrome) purports to have found a gene variant that correlates strongly with a placebo response in irritable bowel syndrome (IBS). The study is small and preliminary, but the results are interesting and do raise important questions about placebo responses.

Researchers are increasingly trying to tease apart the various components of “the placebo effect.” In reality we should use the term “placebo effects” as it is demonstrably multifactorial. “The placebo effect” really refers to whatever is measured in the placebo arm of a clinical trial – everything other than a physiological response to an active intervention. Within that measured response there are many potential factors that would cause an outcome from a fake treatment to be different from no treatment at all. These include statistical effects like regression to the mean and the natural course of symptoms and illness, reporting bias on the part of the subject, and a non-specific response to the therapeutic interaction with the practitioner.

It is also critical to realize that placebo responses vary greatly depending on the disease or symptom that is being treated and the outcome that is being measured. Placebo response is greatest for subjective symptoms of conditions that are known to be modified by things like mood and attention, while it is virtually non-existent for objective outcomes in pathological conditions. So there is a substantial placebo response for pain and nausea, but nothing significant for cancer survival.

In the current study researchers performed a retrospective analysis of an existing study population in which there was a no-treatment arm (waitlist control), placebo acupuncture with minimal practitioner interaction, and placebo acupuncture with enhanced emotional support (the “warm and fuzzy” arm). The study population are those with IBS, a syndrome known to have a substantial emotional component. Further, the outcome was the subjective report of symptoms from the subjects. In other words – the syndrome and outcomes were amenable to maximal placebo responses. Prior studies have consistently shown that for subjective symptoms such as this the interaction with the practitioner is the single most important factor in reporting a subjective improvement in symptoms from a placebo intervention.

In this study subjects were also genetically typed for the COMT gene – an enzyme that breaks down dopamine. The met variant of this gene is associated with decreased activity of the gene, and therefore greater availability of dopamine. The authors report:

The methionine isoform has reduced thermostability, resulting in a three to four-fold decrease in activity relative to the ancestral valine isoform. This functional polymorphism has been correlated with variations in memory function, cognition, attentional processing, affect, confirmation bias, pain processing and sensitivity.

They found:

In this study, we demonstrated that IBS patients homozygous for the COMT val158met methionine allele (met/met) were the most responsive to placebo treatment. Heterozygous (val/met) patients showed an intermediate response, and homozygous valine (val/val) patients showed essentially no placebo mediated symptom improvement.

Patients with the gene variant associated with increased dopamine activity in both alleles (copies of the gene) showed more of a placebo response, especially to the warm and fuzzy intervention. While again I have to emphasize the preliminary nature of this research, the results are plausible and do make sense. IBS is particularly susceptible to suggestion and a warm interaction with a practitioner, interventions that reduce anxiety and improve mood. Anxiety, mood, and the emotional response to pain and discomfort are all brain phenomena, and dopamine is an important brain neurotransmitter involved with emotion and reward, so it’s not surprising.

Assuming this result is reproducible, what are the implications of this finding? First, 20 out of 104 subjects had the met/met genotype –  the type that responded to placebo interventions. That means for about one fifth of the population placebo responses for subjective symptoms, especially those modifiable by anxiety and mood, are likely to be significant. About 52% had an intermediate response, and another 28% had essentially no response. This fits the typical overall 30% placebo response for subjective symptoms seen in clinical trials.

There isn’t much of a downside to being caring and supportive to patients (other than the time investment) and this can enhance the patient’s sense of well-being, reduce anxiety, encourage compliance with treatment and good lifestyle choices, etc. But the effect size is perhaps only significant for a minority of patients, around 20%.

It is also unclear if this result will extrapolate to other conditions. Perhaps dopamine activity is particularly important for gastrointestinal symptoms. It remains to be seen for how many other conditions the met polymorphism will be predictive of placebo response.

None of this implies that placebo responses involve a real biological “healing” mechanism. We are still dealing with the reporting of subjective outcomes. Dopamine may even predict bias in reporting symptoms rather than what the patient experiences (similar to the association with confirmation bias found previously). We may be seeing a suggestibility effect more than anything else.

This research also brings up another interesting possibility – screening out placebo responders from clinical trials. The pharmaceutical industry in particular has been interested in the idea of finding some way to eliminate likely placebo responders from  clinical trials so that the effect of the intervention (medication) would be easier to detect statistically. This could result in smaller and less expensive trials, or even detecting a positive effect that would have been obscured by placebo responders.

This study therefore raises the possibility of having a met/met genotype for the COMT gene (or other genetic typing) be an exclusionary criteria for a clinical trial. Will this increase or decrease the reliability of the study findings and its applicability to the general population?

It also raises the possibility of targeting specific therapies at those with the met/met genotype or other identified types that predict placebo responses. This kind of information could play a role in personalized medicine – targeting therapies to individual patients based upon their genetic makeup. We may be seeing a gene variant that predicts response (or lack of response)  to certain kinds of psychological interventions. I wonder if such gene variants would predict response to cognitive behavioral therapy for certain conditions.

This is a good study in that it raises more questions than answers.

Posted in: Clinical Trials

Leave a Comment (7) ↓

7 thoughts on “The Placebo Gene?

  1. rork says:

    Thanks for pointing this paper out.
    1) I’m not sure regression to the mean or natural course are statistical effects differing between placebo and no treatment. We probably get what you meant anyway.
    2) Rather than remove met/met people or even heterozygotes, we could just use everyone, and measure the gene and put it in the model. Ideally it could be part of the basis for your randomization too. Still model it though.
    3) Figure 2 made me think heterozygotes might be different enough.

    About the paper:
    a) I think they don’t cough up the actual data, nor do they plot it. Why do people plot the mean and standard deviation in graphs where every single reader wanted to see all the data? This is a serious question for biologists that I have been fighting for years. The answer: argument from popularity. The truth: the practice is common exactly because it hides data – it is popular because it is bad. So by following others you are sinning, and opening the door for others to sin (my John Donne for today).
    b) They aren’t perfectly clear what the model for treatment “dose” looks like. Do you think it was 3 levels, or did they recode the doses -1,0,1 (or equivalently, 0,1,2 or 1,2,3)? I think they assigned numbers to the doses, since that is simpler and might have given just enough power for them to squeak under p=.05.
    c) They didn’t really test if hets are different, so the question in my point 3 above, was not addressed. They only show that gene dose correlated (that’s using all of the data). Want to fit your own models? You’ll have to ask for data I think.

  2. BillyJoe says:

    Rork: “We probably get what you meant anyway.”

    We probably would…if we read the second paragraph of the article. (;

  3. Badly Shaved Monkey says:

    Imagine that this result is validated: placebo responses are mediated by a genetic variant that gives the subjects a kick of dopamine. So, a Big Pharma company licenses a dopaminergic drug to confer this benefit to more patients. The slogan could be, “Placebin. It can’t fix your disease but it can make you feel better about being ill.” All the benefits of altie medicine with none of the tedious inconvenience of being stuck with needles or spending endless hours with a loopy homeopath.

  4. Badly Shaved Monkey says:

    One reason for engaging in that little thought experiment was to consider what the SCAMsters’ response would be. It would be for them final proof that Big Pharma is intent on controlling symptoms instead of producing true cures. Which, ahem, is all that the SCAMsters can show they do. Oh, the irony.

  5. DugganSC says:

    I agree that it would be a bad idea to just screen out people with the gene complex. After all, we don’t really know all of the effects. Genes are seldom lightswitches that only turn one thing on or off.

Comments are closed.