By Robert L. Moore, MD, MPH, MBA, Chief Medical Officer
“From Asclepius through Hippocrates to Galen, and until very recently [about 100 years ago], the history of medical treatment was largely the history of the placebo effect, because all medical treatments, with rare exceptions, were at best placebos, at worst unknowingly deadly.”
– Arthur Shapiro and Elaine Shapiro
Professors of Medicine, Singapore National University Hospital
“Just about everyone on this blog mailing list has been in a leadership role at their PCP practice for fewer than 10 years. For the next few months, as new physicians from Partnership’s 10 new counties join this mailing list, I will be reprinting some of my favorite lead articles from past blogs, with updates as appropriate. This post was first included in the phcprimarycare.org blog in 2012 and was republished in Marin Medicine afterwards. It is part one of a series of three articles on the placebo effect.” Enjoy! –Bob Moore
Case Study: A primary care clinician recently diagnosed a patient with major depression. She prescribed citalopram 20 mg per day. One month later the patient felt less depressed and her PHQ-9 had dropped from 16 (moderate depression) to 12. Since the patient was not in remission, the clinician increased the dose to 40 mg per day. One month later the PHQ-9 score was 8 (mild depression) and two months later, the score was 3 (not depressed). Should the citalopram be continued? Which one of the following answers is most correct?
- Yes. The citalopram worked and the patient is at risk of recurrence if the treatment is stopped.
- Yes. Although the citalopram helped no better than placebo in this patient, stopping it will remove the placebo effect and increase the risk of recurrence.
- No. The citalopram worked and remission is likely to continue without continued treatment.
- No. The citalopram had no effect and the patient became better on her own, so continued treatment is not indicated.
I’ll make it easier for you – 1 and 3 are not correct. A 2008 meta-analysis demonstrated that anti-depressants work well in severe depression (PHQ 9 score greater than 20) but are no better than placebo for mild to moderate depression, as was the case in this patient. Her improvement was not due to the pharmacology of citalopram. She might have improved without taking any pill, or she may have improved taking any pill she believed to be helpful – the placebo effect.
We will return to this case and find which answer is most correct at the end of this post. First, though, please explore with me the implications of some new studies and thoughts on the placebo effect.
All clinicians must have a mastery of the placebo effect, not for nefarious purposes such as pushing sham treatments on patients, but to understand the natural history of disease and the role of the placebo effect in mainstream treatments that sometimes provide benefit.
History of Placebo
The word placebo comes from the Latin “I shall please.” It was used in the 14th Century to refer to sham mourners hired to sob and wail for the deceased at funerals. By 1785 it appeared in the New Medical Dictionary, referring to what were considered (at the time) marginal practices of medicine. In retrospect, of course, we now know that the effect of many treatments available at that time were due to the placebo effect. These included worm secretions for toothache (worked 68% of the time) and powdered Egyptian mummy (formerly available from Merck), which was used by President Lincoln’s physician on the bullet wounds inflicted by John Wilkes Booth. (For more, see the reference by behavioral economist Daniel Ariely.)
The use of placebo in controlled clinical trials became common only in the last 50 years. In controlled trials, the “placebo effect” actually includes two effects: the effect of the natural history of the condition being studied (is it improved even if there is no treatment at all?) and the physiologic placebo effect (requires the patient’s cognitive ability to understand what the treatment is attempting to do). The physiologic placebo effect is influenced by two factors: the expectation or belief that the intervention will work, and classical conditioning.
The Power of Belief
The strength of the placebo effect depends on the strength of the patient’s belief that it will help. The placebo response in studies ranges from 0% to almost 100%, depending on the circumstances. Children tend to have greater responses than adults and patients with Alzheimer’s dementia progressively lose placebo responsiveness.
I would like to highlight five factors that increase the strength of the belief of effectiveness: the invasiveness of the intervention, the confidence with which the prescriber associates the treatment with improvement, advertising/marketing, the cultural background of the patient, and the price of the treatment.
- No pain, no gain, part I. Cutting the skin has a powerful effect. From 1930 to 1955, internal mammary artery ligation was used to treat angina. Doctors opened the chest wall, tied off the internal mammary arteries, and closed the chest wall. Patients noted immediate relief which gradually decreased over time. In 1955, this treatment was compared to placebo: a sham surgery, with the patient put to sleep, the skin was cut and sutured, but the arteries were not ligated. The result rocked the medical world: there was no benefit of internal mammary artery ligation over sham surgery. Tens of thousands of individuals had had open thoracic surgery for an ineffective treatment. The placebo effect was strong because of the invasiveness of the surgery, compounded by the certainty conveyed by the physicians (who really believed in it), and the high price for the surgery.
- No pain, no gain, part II. Have we learned our lesson from the internal mammary artery example? Apparently not. How effective is viscosupplementation injection (using hyaluronic acid) of the knee for painful osteoarthritis? A 2012 meta-analysis of high quality trials of intra-articular hyaluronic acid showed a clinically insignificant benefit over saline injection, and increased side effects. Piercing a knee with a large needle and injecting a liquid works particularly well as a placebo. What is more interesting is the size of the placebo effect: 30-50% improvement in pain and 10-30% improvement in function. Finding the size of the placebo effect can sometimes be a challenge in reading scientific studies; they tend to focus on the “true” effect of the treatment, after adjusting out the placebo effect, even when the placebo effect is far greater in magnitude.
- Is brand name really better? Many patients and even many clinicians believe brand name drugs work better than generic medications. Again, the belief of the patient impacts the strength of the placebo response. In a recent appeal process, a PHC patient stated with great conviction that brand name Concerta helped his symptoms while the generic version did not work. Investigation showed that the same manufacturer produced both the brand name and the generic pills in the same factory, in the same way. The only difference: the brand name pills are labeled Concerta. Thinking through the implications: this patient should receive neither the brand name drug nor the placebo, because the entire benefit of the brand name appears due to its placebo effect, since the exact same medication, when generic, did not work. Please consider this every time a patient says the brand name works better: it is likely placebo effect instead of a true difference in efficacy.
- The effect of culture. Many studies have shown the placebo effect can be stronger in some cultures than others, depending on the condition. This is related to the meaning through which people of different cultures experience illness and treatment. For example: the placebo effect in treating gastric ulcers is low in Brazil, higher in northern Europe, and extremely high in Germany. However, the placebo effect in treating hypertension is lower in Germany then elsewhere.
- The effect of price: higher cost medications work better. Most of us subconsciously use price as a surrogate indicator of quality. Taking this a step further, a higher price increases the belief a treatment will work and increases the strength of the placebo effect. This was best shown in a brilliant little study performed by four economists published in 2008. Volunteers were recruited to study a “new” pain medication. Their baseline pain threshold was established with a series of electric shocks to the wrist. They were then divided into two groups. Each was given one of two identical placebo pills, with one group told this new treatment for pain would cost 10 cents per pill and the other told the treatment would cost $2.50 per pill. Pain was reduced by 55% in the low-price pill group and by 80% in the high-price pill group. Another key finding in this study was related to conditioning from prior use of pain medication, and will be described below. Before considering this, though, think through the implications of this study. What does it mean for a patient who requests a brand name medication, perhaps because of an advertisement seen on television? How should we clinicians interpret our patients’ lack of response to low-cost generics? More on this below.
The Power of Conditioning
The placebo effect in the placebo pain pill study was stronger in patients with a history of chronic pain for which they had taken pain medication in the past. This illustrates the effect of conditioning, the second factor contributing to the effectiveness of the placebo effect. It is well documented that in patients with chronic pain, their pain begins to subside when they know their pain medication is coming, before they even receive the medication. The end result is to reinforce the benefit of the pain medication (or placebo).
Conditioning can also affect prescribers! When multiple patients state with great conviction that the expensive brand name medication works better than the inexpensive generic equivalent, the clinician may start to believe this also and discount FDA studies showing bio-equivalence. Even worse, the clinician’s belief can affect their prescribing pattern, leading them to initiate treatment with more expensive medications. It can also influence the degree to which they reassure their patients that the medication they prescribe will help them.
The over-use of expensive medications by prescribers because of conditioning leads to the important and disturbing conclusion that the placebo effect impacts not just the patient being treated, but also affects the clinician recommending the treatment. Here are some examples of the placebo effect on prescribers:
- Generic citalopram contains both the active levo-isomer and the inactive dextro-isomer of the medication. Brand name Lexapro contains only the l-isomer. While there is a small possibility of an unfavorable side effect profile from the d-isomer, the presence of the d-isomer should have no effect on the efficacy of the l-isomer. If the l-isomer dose is the same, the efficacy should be equivalent. Only the price and the brand designation could explain increased perceived efficacy.
- This same principle applies to proton pump inhibitors. When prescribers select a more expensive agent like Nexium, they are not helping their patients, but only contributing to the high cost of health care. Brand name medications are on average about 15 times more expensive than equivalent generic medications.
- Patients with chronic, non-malignant pain on greater than 120 mg of morphine per day will often, with time, develop more pain and request ever higher doses of narcotics. It is a vicious cycle, with higher doses temporarily alleviating pain. But with time, the pain becomes more severe and disabling. As narcotic overdoses have surpassed auto accidents as a cause of mortality in California in the decade prior to 2010, the medical community is now aware of the danger high dose narcotics pose for patients and the community, without really alleviating pain or improving function of the patients taking this therapy. Counseling patients on this vicious cycle is difficult because of their belief in the power of narcotic medications and the conditioning that changes the physiology of patients taking these medications chronically. This belief is reinforced by withdrawal symptoms when a dose is missed or a clinician begins reducing the dose. Any drug with withdrawal symptoms will be harder for a patient to stop, because they often deeply believe that only that medication is actually controlling the disease. This is true of narcotic medications, benzodiazepine anxiolytics, many antidepressants, muscle relaxants, and even NSAIDS. In all these cases, prolonged use of the medication produces long-term changes in the synapses. These changes cause patients to feel symptoms when the medication is withdrawn. Thus, withdrawal symptoms strengthen the placebo effect! In fact, if the patient can get through the withdrawal, they may feel the same or better than they did on chronic therapy, but it is hard to get the patient there. Trust between the clinician and patient and detailed education are keys to success, but may not be enough in some cases. For this reason, we should avoid initiating prolonged use of medications which can cause withdrawal symptoms until all other options are exhausted.
- When reviewing evidence for effectiveness of a given treatment, journal articles often compare only the effectiveness relative to placebo because this is the standard for FDA approval. I recommend always digging into the article to look at the treatment’s effect on the control/placebo group; it gives valuable information. For example, early studies comparing amoxicillin to placebo for treatment of acute otitis media (AOM) in children showed the control group improved in 82% of cases (presumably most cases were viral and did not need antibiotics), while about 90% of the amoxicillin group improved. In this case, the mechanism of the placebo effect is not related to belief or conditioning, but on the natural course of the disease. Nonetheless, use of antibiotics in pediatric otalgia conditions parents to expect antibiotics for treatment of all ear pain, since their child improved with treatment the last time antibiotics were prescribed. Looking closely at the placebo effect in studies helps clinicians interpret marketing for new products. For example, should we preferentially prescribe a more expensive, more powerful antibiotic that is reported to improve 50% more cases of AOM than amoxicillin (i.e. a 12% benefit over placebo compared to 8%)? Perhaps we should just recommend ibuprofen for a few days (which would work 82% of the time) and reserve antibacterial treatment for cases where this conservative approach fails. This approach has been recommended by the American Academy of Pediatrics as an option since 2004, but clinicians in the U.S. have been slow to adopt it due to clinician conditioning.
Adapting our clinical practice and communication with patients to account for the placebo effect may be the most important skill we develop. We must use our scientific training to inform the art of medicine. As clinicians, we owe it to our patients and society to account for the placebo effect and use only low cost, relatively safe treatments when the major effect is likely to be placebo.
If the patient really wants a pill – when education about the nature of the condition doesn’t seem to work – what should we do? Because it is ethically problematic to give a treatment the provider knows is placebo, clinicians tend to resort to medically acceptable treatments (especially pharmaceuticals), even if they know or strongly believe their benefit is most likely due to the placebo effect. One way around this conundrum is to consider lifestyle interventions. Examples include: recommending gentle physical activity for low back strain, headaches, depression, etc.; working with the patient to decrease fat intake or make other beneficial dietary changes for gastrointestinal symptoms, depression, anxiety, etc.; or recommending a low cost multivitamin-with-mineral supplement to boost the immune system’s ability to heal. If there is not time or resources for the more in-depth support required to encourage dietary and physical activity changes, the vitamin option may be more attractive.
An important word of caution: the placebo effect does not prove that a serious condition is not present. Patients with life threatening conditions can get some temporary relief from a placebo treatment.
Going back to our original case involving citalopram for moderate depression: There is likely some effect of placebo, in which case the relapse may occur no matter what choice is selected. If a relapse of moderate depression happens after citalopram is stopped, it may be attributed to stopping the medication. If it happens while taking citalopram, it may be attributed to the medication not working as well. In the first case the clinician perhaps “stopped the medication too early” and will restart it. If a recurrence occurs on citalopram, however, the clinician may then move to another medication that might “work” better, perhaps a nice expensive brand name product. The clinician gives four weeks of Lexapro samples, which the patient reports works very well (the placebo effect enhanced by the use of expensive, brand name samples). The clinician then requests that Medi-Cal cover the medication. Suppose Medi-Cal denies the request and the patient appeals (Actually, since the pharmacy benefit has been carved out since 2022, MediCal prefers brand name medications due to rebates, but set that consideration aside for the moment to think about the principle). This puts the PCP in the position of supporting the selection of an expensive brand name medication whose effectiveness is entirely due to the placebo effect.
An alternative: use your influence with the patient to increase their belief that a particular low-cost treatment will work and is the safest medication for them. Use your word choice to increase the patient’s perception that the treatment is effective and lessen their belief that high cost or brand name equates to more effective treatment.
The correct answer (in 2012)? Option 4 (stop the medication) is the most cost-effective option. It avoids escalation of treatments with no proven efficacy. For the best option of all, based on an understanding of the placebo effect, consider trying non-pharmacological treatment: brief intervention counseling or referral for cognitive behavioral therapy if you have these available in your health center or in your community.
Ariely, Dan. The Power of Price: Why a 50-Cent Aspirin Can Do What a Penny Aspirin Can’t. Predictably Irrational: The Hidden Forces that Shape our Decisions. HarperCollins Publishers, 2008.
Lieberthal et al. Clinical Practice Guideline: The Diagnosis and Management of Acute Otitis Media. Pediatrics. 2013; 131:e964-e999. American Academy of Pediatrics.
Moerman DE. Cultural variations in the Placebo Effect: ulcers, anxiety, and blood pressure. Med Anthropology Quarterly 14:51-72.
Rutjes, et al. Viscosupplementation for Osteoarthritis of the Knee: A Systematic Review and Metaanalysis. Ann Intern Med 2012;157(3):180-191.
Venekamp et al. Antibiotics for acute otitis media in children (Cochrane Review).The Cochrane Library 2013 Issue 1.
Waber, et al. Commercial Features of Placebo and Therapeutic Efficacy. JAMA March 5, 2008, 299(9), p. 1016-1017.