The placebo effect describes the improvement of symptoms in a clinical trial of a new therapy, e.g. a new drug, where some patients receive an “inert” pill (tablet, infusion, or other medicines) that does not contain the pharmacological compounds to be investigated. As these studies usually are double-blinded, neither the patient nor the doctor knows who received the drug and who received the placebo.
This is done to control for “unspecific” treatment effects: measurement errors and insecurities, a spontaneous variation of the symptoms, response biases such as the Hawthorn effect, and the results of the patient’s hope/belief that he/she receives effective treatment for his/her symptoms (1). These effects are to separate (subtract) from the drug’s effect to allow a fair assessment of the efficacy of the drug. The same is sometimes done for other medical interventions, e.g., using sham surgery as a control surgery.
While the placebo effect is only dismissed in such a trial, it still raises an important question: Why is it that — depending on the disease, the treatment, and its duration, and the way the treatment is evaluated — patients can substantially improve following a placebo provision (2)?
We have discussed here that even providing such placebos as “open-label” still may be beneficial for the patient. And while the placebo’s effect may be visible in such trials, it is around in everyday medicine as well, literally in every doctor-patient communication, every time someone seeks medical advice, takes medication, or asks a pharmacist or therapist for help.
Not only placebo researchers but also pharmacologists and trialist (experts conducting placebo-controlled trials) would like to know in advance (before starting a study) who will be responding to placebo and who will not — since not all patients are “placebo responders” (3). Leaving them out of the trial would make life easier: fewer patients = lower costs = easier evaluation = better drugs? Therefore, predicting the placebo response by predicting the placebo responder has driven research for quite a while, actually since placebo-controlled trials were “invented” around the middle of the last century (4).
What does sex have to do with it?
The term “sex” describes biological differences between men and women, while “gender” describes a person’s sexual identity, irrespective of his/her biological outfit. In most cases, sex and gender may be identical, but there are deviations from this balance, in both directions. For now, we use both terms synonymously.
Back to the placebo effect: In the old days, placebo responders were seen as highly neurotic (anxious, depressed, hysteric, deviant in general) persons, since they reported symptom improvement after taking a pill that contains “nothing” — so their symptoms cannot be “real,” but imagined. And since personality characteristics such as “neuroticism” were seen as predominantly occurring in women, it was assumed (but never shown) that sex/gender alone or in company with other characteristics would drive the placebo effect in drug trials.
However, the search for a “placebo personality” that could easily be identified by a simple psychometric test has never gained any success: for now, it has been agreed that there is no such thing as a stable characteristic of a placebo responder (3), and neither is there a gene that controls the placebo response (but this will be another story at another time).
How did we find out?
In each single drug trial that may include between 100 and 1000 patients (or more, or less), 50% of the patients are usually females — unless the trial or the clinical condition requires otherwise. If 50% of the patients receive the drug and the other half placebo (on a random basis), a test for the effects of sex on the placebo response seems easy, just by comparing the 25% of female participants on the placebo with the 25% male participants on the placebo, to see whether this makes a difference. Usually, this is not the case, not in the majority of the roughly 30,000 trials conducted every year worldwide.
If you want to estimate the same effect across all clinical trials for the same (or a different) condition, and with the same (or with another) therapy, you may combine them in a meta-analysis and answer the same questions: is the therapy effective across many trials, and is the placebo effect the same? And, of course, you can also ask whether the effect of sex (or age, or another patient characteristic) on the drug or placebo effect is consistent across many trials.
In such an attempt, we extracted (in 2014) 75 systematic reviews, meta-analyses, and meta-regressions (5), including nearly 1,500 randomized clinical trials (RCT) with more than 150,000 patients and more than 40 different diseases, and screened them for whether sex differences had been noted to contribute to the placebo effect. In only 3 such analyses was female sex associated with a higher placebo effect, and in only three diseases (restless leg syndrome, bipolar mania, and schizophrenia). When we updated this analysis in 2018 (6) with another set of papers, it did not change the overall conclusion: There is no evidence for a major contribution of sex to the placebo effect in clinical trials across medicine.
What about the nocebo effect, then?
Nocebo effect is another term for side effects (or adverse events, AE) that occur in clinical trials when they are reported in the placebo arm of the trial (as long as it is double-blinded). In this case, these AE cannot be attributed to the chemical content of the drug, but are called “nocebo effects” — they follow similar rules to the placebo effect, but we will deal with it another time.
AE during a drug trial are to be recorded, and serious AE lead to drop-out from the study. Nocebo effect-driven AE reporting and dropout can be substantial (7). Again, the question whether factors can be identified that drive the nocebo effect is important, and testing is easy for single clinical trials as well as for meta-analyses.
We identified 18 meta-analyses covering more than 500 trials including more than 25,000 patients across medicine — mostly, however, for psychiatric, neurological, and pain-associated conditions, presumably because centrally-acting drugs are specifically prone to provoke many AE reports and drop-outs, both in drug and placebo arms. The analyses confirmed our findings for the placebo response, with a few exceptions: There were no systematic effect of sex (or age) on AE reporting and drop-out rates in the placebo arms of these trials (6).
Indeed it is a curious way of coping: to close the play and leaving issues open (Wir stehen selbst enttäuscht und sehn betroffen, den Vorhang zu und alle Fragen offen) (B.Brecht, The Good Women of Setzuan, epilogue)
Is this the final answer to our question? It could have been if we had not have stumbled into sex differences in our research some time before we called it placebo research. Between 1993 and 2005, we performed a number of experimental studies looking into whether and how one could experimentally elicit nausea & vomiting (N&V), in a search for a model to investigate anticipatory N&V in cancer patients (8).
We had chosen a rotation-chair paradigm and conducted a series of experiments, varying conditions (rotation vs pseudo-rotation/vection, upright vs supine position, repetition within hour or days), participants (Caucasians-Chinese, males-females, patients-healthy controls), and interventions (Pavlovian conditioning, latent inhibition and overshadowing and its combinations, verbal manipulations), using gustatory and olfactory stimuli to enhanced or suppress N&V, and measuring rotation tolerance and nausea severity and its myoelectrical and biochemical correlates (e.g. 9-12). And in all these experiments we had consistently found and reported that significant sex differences exist.
What experiments can tell about sex effects that trials cannot
Placebo experiments around the world use only a few paradigms to study the underlying mechanisms: mostly pain models are used (e.g. heat pain at one arm), as they are easy to perform and can easily be exported to brain scanners. Visceral pain, in contrast, is much more difficult and only used in one lab currently. The experimental itch is another rarely used model, and our nausea model is also not widespread. Experimental investigations have, furthermore, used Pavlovian conditioning procedures on the one hand, and verbal (deceptive) induction of placebo responses after applying presumed effective drugs, painkillers, antiemetics, etc. that were placebos, thereby manipulating expectancies.
When screening our literature database, we identified 26 experiments that investigated and reported sex contributions to the placebo effect during different experimental interventions: in 8 experiments, men showed stronger responses than women (5 times with placebo analgesia), 11 showed stronger effects in women compared to men (3 x with nausea, 4 times with pain), and the remaining 7 did not find differences. However, it gets even more complicated.
Sex differences for placebo effects not only exist, but they follow some rules, as it appears:
- Despite higher pain sensitivity in females, placebo analgesia is easier to elicit in males;
- It appears that conditioning is effective specifically to elicit nocebo effects;
- Conditioning works well to elicit placebo and nocebo effects, but only in females;
- Verbal suggestions are insufficient to induce placebo effects in women but work in men.
The apparent difference between men and women is best illustrated by one of our experiments (13) where we showed that women respond to conditioning (of nausea symptoms) much better than men, while men were more susceptible toward verbally-induced symptom provocation. The obvious interpretation of these differences is that for women, learning mechanisms dominate — and previously learned content remains relevant — while in men, an acutely provided information is of higher relevance than past experiences.
This may also explain the higher susceptibility of men to verbally-induced placebo analgesia, despite their lower overall pain sensitivity and the opposite response pattern in women. And it explains why the male-over-female response for verbally-induced placebo analgesia is reversed when trust is enhanced by oxytocin, known to be specifically effective in females (14).
The apparent discrepancy between trials and experiments requires an explanation
The best explanation that we can provide today is referring to the different nature of experiments on the one hand and clinical trials on the other. In a well-planned experiment, the separation of expectancy manipulation and learning/conditioning — the two main underlying mechanisms of the placebo response — can be achieved, and the relative contribution of either can be explored. This allows us to directly compare the relative potency of the two and of any intervention on them.
In a randomized placebo-controlled trial, in contrast, the amount and effect of factors relating to patients’ learning (medical history, previous therapies and their success and/or failure, duration of knowing the treating doctor, etc.) and to expectancies delivered and associated with the treatment (informed consent and AE reports, symptom diaries, number and intensity of doctor-patient contacts etc.) is neither known nor balanced, and may vary from patient to patient as well, e.g. in relation to his/her social environment and the “placebo by proxy” influences (15). Under these circumstances, it is conceivable that any existing difference in placebo responsiveness between the sexes are averaged out in clinical trials and result in equally-sized placebo effects in men and women, as we have seen.
This is part 5 of a series covering “placebo” provided by Paul Enck and Sibylle Klosterhalfen from the Tübingen University Hospital. Continuous updates on placebo research can be found at www.jips.online.
- Enck P, et al. The placebo response in medicine: minimize, maximize or personalize? Nat Rev Drug Discov. 2013;12(3):191-204.
- Elsenbruch S & Enck P. Placebo effects and their determinants in gastrointestinal disorders. Nat Rev Gastroenterol Hepatol. 2015;12(8):472-85.
- Kaptchuk TJ, et al. Do “placebo responders” exist? Contemp Clin Trials. 2008;29(4):587-95.
- Weimer K, et al. Placebo effects in psychiatry: mediators and moderators. Lancet Psychiatry. 2015;2(3):246-57.
- Weimer K, et al. Age and sex as moderators of the placebo response – an evaluation of systematic reviews and meta-analyses across medicine. Gerontology. 2015;61(2):97-108.
- Enck P & Klosterhalfen S. Does sex differentially contributes to the placebo effects in clinical trials and in experimental studies? Frontiers in Psychiatry 2018 (in preparation)
- Mitsikostas DD, et al. Nocebo in fibromyalgia: meta-analysis of placebo-controlled clinical trials and implications for practice. Eur J Neurol. 2012;19(5):672-80.
- Stockhorst U, et al. Anticipatory nausea in cancer patients receiving chemotherapy: classical conditioning etiology and therapeutical implications. Integr Physiol Behav Sci. 1993;28(2):177-81
- Stockhorst U, et al. Effects of overshadowing on conditioned nausea in cancer patients: an experimental study. Physiol Behav. 1998;64(5):743-53.
- Klosterhalfen S, et al. Pavlovian conditioning of taste aversion using a motion sickness paradigm. Psychosom Med. 2000;62(5):671-7.
- Klosterhalfen S, et al. Latent inhibition of rotation chair-induced nausea in healthy male and female volunteers. Psychosom Med. 2005;67(2):335-40.
- Klosterhalfen S, et al. Effects of ethnicity and gender on motion sickness susceptibility. Aviat Space Environ Med. 2005;76(11):1051-7.
- Klosterhalfen S, et al. Gender and the nocebo response following conditioning and expectancy. J Psychosom Res. 2009;66(4):323-8
- Colloca L, e al. Vasopressin Boosts Placebo Analgesic Effects in Women: A Randomized Trial. Biol Psychiatry. 2016;79(10):794-802.
- Grelotti DJ & Kaptchuk TJ. Placebo by proxy. BMJ. 2011;343:d4345.