When I wrote two weeks ago about depression, two genes stood out--the serotonin transporter gene and two associated with catecholamine metabolism. Interestingly, variants in these and related genes turn out to be associated with placebo responses. That led me to wonder if an overlap between the mechanisms antidepressant effects and the biology of placebo responses might account for the high placebo response rates in clinical trials of antidepressant medication and the notorious difficulty of demonstrating efficacy of antidepressants in randomized controlled trials.
For a first take on the question, I will look at fascinating review of genetics and placebo effects by Kathyrn Hall, Joseph Loscalzo, and Ted Kaptchuk of the Program in Placebo Studies and Therapeutic Encounter at Beth Israel Deaconess Medical Center in Boston. I will focus mainly on antidepressant placebo responses.
The authors note that almost all modern randomized controlled trials lack a no-treatment control, which makes it difficult to separate genuine biological placebo responses to the interpersonal aspects of the clinical encounter from spontaneous remission, regression to the mean, and natural variations in symptoms. I would add that even no-treatment controls may activate some placebo effects, since the procedures involved in evaluating patients for and monitoring symptoms themselves have biological effects in the brain.
Hall and her colleagues cite research showing involvement of µ-opioid and endocannabinoid mechanisms in placebo analgesia, the dopamine reward pathway in expectation or anticipation of placebo responses, and serotonergic pathways in placebo antidepressant responses. I would add that we should not rule out the possibility that opioid and endocannabinoid pathways contribute to antidepressant placebo responses--µ-opioid agonists have been shown to relieve depression, though their addictive properties severely limit their clinical use. Candidate gene studies, in which variants in one or more specific genes are correlated with clinical phenomena, have looked at these pathways. In addition, the newer technology of genome-wide association studies can scan an individual's entire genome for genes associated with placebo responses. Unfortunately, the largest genome-wide study of response to antidepressant medication, the STAR*D study, did not include a placebo arm.
One of the few studies which included a no-treatment control group looked at the gene for catechol-O-methyltransferase (COMT), a dopamine-metabolizing enzyme. A variant associated with slower dopamine metabolism and thus higher dopamine levels was associated with higher placebo responses to sham acupuncture in patients with irritable bowel syndrome.
Another study looked at variants in the gene for monoamine oxidase A (MAO-A), which metabolizes dopamine, norepinephrine, and serotonin. Depressed individuals with the variant coding for the less active form of the enzyme, and thus higher synaptic levels of these monoamines, had higher placebo responses in trials of antidepressants. In that study, however, COMT variants were not associated with placebo responses.
A study of depressed patients’ responses to placebo bupropion looked at thirty-four candidate genes. A different variant in the MAO-A gene was associated with placebo response, and there was suggestive evidence for involvement of several variants in the COMT gene. In addition, variants of the genes for dopamine beta-hydroxylase, an enzyme which converts dopamine to norepinephrine, the serotonin transporter gene, and the gene for a serotonin 2A receptor were associated with placebo responses.
A small but very interesting study used both genetic and PET scan technology to look at placebo responses in social anxiety disorder. Placebo-induced reductions in anxiety were accompanied by decreases in stress-related amygdala activity in patients homozygous for two serotonin pathway gene variants: the tryptophan hydroxylase-2 gene promoter, which triggers synthesis of the enzyme which makes serotonin, and the serotonin transporter gene.
These intriguing findings, Hall and colleagues note, raise the question of drug interactions with biological placebo responses. If the placebo and active drug arms of a study contained differents proportion of patients genetically predisposed to placebo responses, for example, this would compromise the fundamental assumption of placebo-controlled trials, that the rate of drug response is the rate of response of drug-treated patients minus the rate in placebo-treated patients. Hall and colleagues also raise the question of “gene-drug-placebo” interactions, giving examples of a Parkinson’s disease trial in which the homozygosity for the low-activity variant of COMT resulted in patients responding better to placebo than active treatment, and similar findings in a study of primary prevention of cardiovascular disease.
The authors note that “sophisticated network analyses” may be needed to untangle the genetic, proteomic, metabolic, and neural circuit effects of placebos, emphasizing the likely biological complexity of placebo-related phenomena. They briefly discuss the substantial implications of this new neuroscience of placebo responses for clinical trials. They also write that genetic information about a patient’s tendencies to placebo responses could help clinicians determine drug doses, though I am not sure how. If a patient is likely to have a strong placebo response to an antidepressant, it might suggest that a lower dose would be adequate, but this hinges on whether the placebo effects are likely to persist—though there is evidence that many do. This brings up the authors’ other clinical point, which is that open-label “honest placebo” treatments may be indicated, and they cite a trial their group conducted for major depression.
These genetic findings provide a foundation for understanding my experience treating hundreds of depressed patients with various antidepressants and psychotherapy. I am convinced that both placebo and pharmacological effects contribute importantly to patients’ recoveries, and that nocebo effects, i.e. interpersonal aspects of the clinical interaction that interfere with symptom relief, contribute to treatment resistance. I have vivid memories of dramatic responses to tricyclic antidepressants, SSRI’s, SNRI’s, and nefazodone, often when I pushed the doses to high levels. I do not recall such responses to psychotherapy for severe depression, or to trazodone or augmentation with lithium. Of course selection bias in my personal sample may have contributed to failures in treatment-resistant patients who received such third-line pharmacotherapies, and my assessment and memory may well be distorted by influences like drug advertising and psychiatric group-think. But I am inclined to believe tricyclics, SSRI’s, SNRI’s, nefazodone, and probably MAOI’s have pharmacologic effects greater than what can be demonstrated in double-blind trials as they are now conducted. And I also believe, as Jerome Frank argued long ago, that the interpersonal aspects of clinical encounters and healing relationships contribute importantly to patients’ recoveries, in addition to the technical aspects of the psychodynamic and cognitive-behavioral therapies I employ.