The Neurobiology of Placebo Responses
Placebo responses are the bane of psychiatric treatment research, for they obscure the specific effects of treatment—it’s hard to show a particular psychotherapy or medication is effective if the control group has a high rate of improvement. For clinicians, they can elicit a false impression of the effectiveness of their own treatment methods. On the other hand, patients who improve because of the appearance, rituals, and relationships of treatment do actually improve, and there is evidence that many stay well. Placebo responses are the norm not only in psychiatric treatments but throughout medicine and surgery as well.
The neuroscientist Fabrizio Benedetti and his colleagues at the University of Turin in Italy have published a fascinating review of the psychobiological processes that occur when a patient responds to a placebo. Placebo responses may involve one or more of several mechanisms, including reduction of anxiety, promotion of hope, and learning to associate treatment with improvement. The authors also discuss the role of genes, the interesting possibility that some drugs may change elements of the placebo response itself, and how some diseases and medications may interfere with placebo responses by impairing prefrontal executive control.
Benedetti and colleagues note that improvement after administration of an inert pill or injection or a sham physical treatment may result from factors other than psychobiological placebo responses. These include spontaneous remission of the disease, selection bias in clinical trials, and the effects of unidentified co-interventions which accompany treatment. Their review focuses on the processes which occur in the patient’s brain when treatment is administered in the setting of sensory and social stimuli that lead the patient to expect recovery. They present evidence of genetic predisposition to placebo responses, which is not surprising, since we would expect brain systems, which utilize enzymes, receptors, and neuropeptides, all of which are coded by genes, to be subject to genetic variability. The neural systems involved include anxiety, reward, and conditioned learning, social learning, and reinforced expectations.
Anxiety about symptoms or disease can be reduced by telling a patient in advance that a treatment will reduce be effective. Benedetti and colleagues reference two studies which appear to use patient reports of anxiety levels, but much of the research on anxiety they cite used benzodiazepines or benzodiazepine antagonists to manipulate anxiety levels. In addition to human studies, they describe a study using social-defeat models of anxiety in rats. They separate anxiety about pain, which appears to involve cholecystokinin, a gastrointestinal peptide which has receptors in the central nervous system, from anxiety about external stresses, which may involve activation of the endogenous opioid system.
Benedetti and colleagues are not always clear about the distinction they (and others) make between conscious control of anxiety and the nonconscious threat responses which occur in conditioning experiments. Expectation-related anxiety appears to be a top-down phenomenon involving thoughts about future distress. One of the experiments the authors use to illustrate this appears to be a conditioning experiment. And some of the work they describe seems to imply a pharmacologic definition of anxiety as that which is reduced by benzodiazepines or increased by benzodiazepine antagonists. I am not sure that assumption would hold up to either clinical or neurobiological examination.
A number of imaging studies in humans have implicated the dopamine reward system, which I discussed in my post about addiction,in the expectation of benefit from placebos. In particular, in a study comparing responses of hospitalized men with depression to placebo and fluoxetine, they found activation of the ventral striatal and orbital frontal reward areas in the first week of treatment with both fluoxetine and placebo, suggesting activation of the reward system before the pharmacodynamic effects of the antidepressant, which they observed later in treatment in other areas of the brain, occurred.
Benedettti and colleagues note an additional neurochemical circuit involved in placebo reduction of pain. An opioid-based pathway which inhibits pain responses descends from the cortex to the hypothalamus, periaqueductal gray, rostroventromedial medulla, and spinal cord. This can be antagonized by cholecystokinin at several levels. So-called nocebo responses, in which pain is aggravated rather than relieved by an inert substance, involves deactivation of both mu-opioids and dopamine receptors.
Conditioned learning is also central to placebo responses. Patients learn to associate not only pills and syringes, but also hospitals, medical equipment, and medical staff with symptom relief. Thus placebos administered after an active drugs produce more symptom relief than the placebo given alone for the first time. The authors note that both nonconscious and conscious learning may be involved in such conditioned associations. They give interesting examples of conditioned responses to placebo of the human immune and endocrine systems. They also cite an experiment showing that altering conscious expectation by informing subjects that a placebo was inert eliminated a placebo response. Other experiments have provided additional evidence for separate effects of conditioning and expectation.
While most placebo research involves perception of pain, placebo effects have been demonstrated on motor output in patients with Parkinson’s disease as well as on muscle strength in normal subjects. In both cases, cognitive learning appears to be involved.
Benedetti and colleagues also describe social learning of placebo responses, in which observing someone else experience reduced pain after a cue produced reduced pain in the experimental subjects after the same cue. This socially learned response was as powerful as a control group’s classically conditioned placebo response.
Finally, Benedetti and colleagues present evidence that prefrontal executive control is necessary for placebo responses. In patients with Alzheimer’s disease, which involves frontal lobe degeneration, reduction in placebo response is correlated with cognitive impairment. A fMRI study showed that the opioid antagonist naloxone blocks placebo analgesia by interfering with descending pain-control circuits. This raises the question of possible unintended effects on placebo responses when opioid antagonists such as naltrexone are used to treat alcoholism and opioid addiction. Similarly, there is evidence that inactivation of dorsolateral prefrontal regions by repetitive transcranial magnetic stimulation (rTMS) similarly blocks placebo analgesia. If it also blocks the placebo components of improvement resulting from rTMS treatment of depression, which targets the same areas, this could have implications for clinical use of rTMS.
Benedetti and colleagues note that some drugs, particularly opioids, in theory could alter placebo responses by direct pharmacodynamics effects on the brain circuits which mediate placebo responses. This has been tested in experiments which deliver drugs without the subject’s knowledge, eliminating expectation effects. They summarize a study of a cholecystokinin antagonist which induced analgesia better than placebo, but a hidden injection of the same drug produced no analgesia. They believe the drug’s action was to enhance placebo-activated release of endogenous opioids. They pose the question of whether such a drug that acts not by blocking pain but by expectation-activated mechanisms should be considered an effective analgesic.
Benedetti and colleagues have summarized a large number of human and animal experiments. A number of results are intriguing. Psychobiological placebo responses should be distinguished from other reasons patients in control groups improve. Several brain mechanisms are involved. The rituals of treatment administration, including the sensory characteristics of the treatment, the treatment setting, and the patient relationship with the treating clinician, can themselves produce improvement by reducing the patient’s anxiety as well as by enhancing hope and the expectation of benefit. Through conditioned learning, social learning, and learned reinforcement of expectations, treatments become associated with improvement. Genes and the effects of both early life and present day experiences on their expression contribute to individual variation in placebo responsiveness. And finally, process which interfere with prefrontal executive control, including Alzheimer’s disease, other psychiatric disorders, medication such as opioid antagonists, and treatments such as rTMS can interfere with placebo responsiveness.
The issue of placebo responses to antidepressant medication is a fascinating question; it may make sense to consider the neurobiology of depression and antidepressants’ mechanisms of action before looking at that.
Further reading: Fabrizio Benedetti, Elisa Carlino, and Antonella Pollo, How Placebos Change the Patient's Brain.
Neuropsychopharmacology 2011 Jan; 36(1): 339–354.