The placebo effect: powerful, powerless or redundant?
- 1The George Institute for Global Health, Musculoskeletal Division, University of Sydney, Sydney, New South Wales, Australia
- 2The EMGO+ Institute, VU University, Amsterdam, The Netherlands
- Correspondence to Dr Steven J Kamper, The EMGO+ Institute, VU University, Vander Boechorststraat 7, Amsterdam 1081BT, The Netherlands,
- Received 14 June 2012
- Revised 19 July 2012
- Published Online First 14 August 2012
Over the last 200 years, the placebo effect has cast a large and persuasive shadow over the medical field. In that time it has been by turn; harmless charade, charlatan's ruse, therapeutic device, methodological tool, ethical dilemma, research theme and source of controversy. Despite popular recognition, pervasive problems underlie conceptualisation of placebos and placebo effects. With medicine now firmly entrenched in the age of evidence-based practice, there is a question as to whether it is time to leave the old placebo behind us. The idea of a magical black box from which unexplained therapeutic effects spring up is archaic and also unhelpful from a scientific point of view. If there really is an effect, surely we are best served by directly investigating what is responsible for it. This knowledge can be used in the clinic to improve treatment effectiveness and in research to inform study design.
The view and role of placebos have developed over time. Modern medical understanding of placebos dates back to the late 1700s1, and since then, placebo interventions appear to have been used fairly commonly in medical practice alongside other treatments.2 Up until midway through the 20th century, the prevailing opinion appears to have been that placebo interventions had no effect on pathophysiology. Placebos were used only to bring comfort to the patient, ‘a camouflage behind which to watch nature takes its course’1 (p. 512).
From the 1950s onwards, there was a shift towards regarding placebos as having genuine effects of their own. Beecher's well-known article ‘The Powerful Placebo’3 apparently confirmed the claims of widespread and important therapeutic effectiveness of placebo interventions. This period also saw the start of rapid growth in the use of placebos as a control intervention for testing the efficacy of other treatments. Thus, as research generally, and randomised controlled trials (RCTs) in particular, grew in stature and volume, placebo interventions also gained a more prominent face in the medical literature. Nowadays, exploration of placebo effects represents a field of research on its own.4
The view that placebos have a powerful therapeutic effect remained largely unchallenged until the 1990s. Around this time, authors began to question the scientific basis upon which the claims of large therapeutic placebo effects were made.5 ,6 Critically, it was pointed out that longitudinal changes in a group of patients who have received a placebo intervention are not only due to placebo effects. From the point of onset (or clinical contact), many conditions improve with the passage of time independent of clinical management, due to the natural history of the condition. Improvements over time are likely to be further enhanced by regression to the mean which is a statistical artefact whereby scores that are high at one point in time, ie, at study entry, are always more likely to be lower at a later time. Thus, when a group of patients receive a placebo intervention the changes afterwards may be due to placebo effects plus changes due to natural history plus the effects of regression to the mean (figure 1). In order to calculate the size of a placebo effect, a concurrent control group that receives no intervention is required. This observation was important because it highlighted the need to partition out change due to natural course (and other factors) from purported placebo effects. Authors contended that the placebo effects reported in trials could be substantially or completely explained by natural history along with issues related to observer and subject biases, regression to the mean and errors in analysis.6 ,7 A key development in this argument was the publication of Hrobjartsson et al’8 systematic review and meta-analysis titled ‘Is the Placebo Powerless?’ which found little evidence for the existence of large placebo effects in clinical trials.
The publication of Hrobjartsson's review in the New England Journal of Medicine ignited the debate. In particular researchers investigating potential mechanisms of placebo effects, mainly in experimental settings disputed the conclusion that placebo effects were small and clinically unimportant.9 Experimental pain researchers pointed out the apparent discrepancy between the very small placebo effect sizes reported in clinical studies and the much larger effects in studies investigating placebo mechanisms.9 It may be that this discrepancy is due to different designs of placebos for these two purposes. In placebo mechanism studies, researchers design ‘maximised’ placebos to have the greatest positive impact on the patient's expectation of benefit. Whereas clinical researchers strive for ‘comparable’ placebos to match the expectation of benefit attached to the placebo intervention with that in the index intervention.10 This debate has continued for the best part of the last decade.11 ,12
Emerging ethical issues
Along with the developing view of the role of placebos, the last half century has seen a shift towards acceptance of the evidence-based medicine paradigm. These two factors have led to increasing recognition of ethical and legal concerns associated with the administration of placebos in practice. The issues are difficult and consensus as to the appropriateness of placebos in clinical practice has not been achieved.13 At the heart of the matter lies the question of whether deliberate deception of the patient is acceptable in the course of their treatment. These difficulties are illustrated by the American Medical Association's treatment of the issue in their Ethics Policy.14 They provide guarded permission for placebo interventions, but ‘only if the patient is informed of and agrees to its use’.
An underlying controversy concerns the theoretical conceptualisation, or indeed the very logic, of the placebo.5 The crux of this controversy is the logical paradox of something inert (a placebo) having an effect (a placebo effect). The well-recognised failure of researchers to find agreed definitions for placebos and placebo effects likely stems from this fundamental contradiction.15
Implicit in the definition of placebo effects is that they are psychologically mediated. Whether the placebo is an ingested agent such as a sugar or vitamin pill, or a procedural intervention such as non-penetrating acupuncture, sham electrotherapy or an inert topical formulation, it must be physically inert. However, there follows the issue of disentangling inert from non-inert psychological interventions. The problem is, if there is a (psychological) mechanism, can the psychotherapeutic part of the intervention still be said to be inert? One could argue that if we knew more about the condition (and the intervention), we would be able to identify the mechanism of action and hence the effects would no longer be due to placebo.
The notion of ‘inertness’ of the placebo must also be attached to whatever the condition the researcher is interested in. For example, a sham shoulder taping procedure may not result in biomechanical changes to joint movement or provide structural support but it may provide tactile feedback that cues a patient to avoid a certain behaviour. So the sham taping may be a placebo for one patient group, for example, football players with recurrent dislocations but an active treatment for another, for example, older adults with rotator cuff tendinopathy. Along a similar line, researchers have proposed the term ‘nocebo’ to describe undesirable effects of placebos. For example, studies have described immunosuppressive effects of placebo interventions16 which may be undesirable (a nocebo) for one patient group, for example general surgery patients but desirable (a placebo) for another group, for example, organ transplant recipients. Thus, placebos can be nocebos and vice versa depending on who gets them.
While researchers have made attempts to resolve this problem by adding qualifiers such as ‘inherent’ or ‘specific’ mechanism, none have provided a universally accepted solution. If we accept that a placebo effect is a response to something other than that which we hypothesised, as implied by the use of terms like ‘inherent’ and ‘specific’, whether or not something is a placebo then rests on the theory of the researcher. One man's meat is another man's placebo.
Further, an intervention assumed to have an effect at one point in time could subsequently be shown to have no physiological effect and be used as a placebo. For example, therapeutic ultrasound has enjoyed a history of use for musculoskeletal pain and sporting injuries but more robust research conducted recently suggests its specific physiological effects are negligible or absent.17 It is not beyond the realms of possibility to suggest that ultrasound may make the transition from treatment in the 20th century, to placebo in the 21st century.
In some ways, it is not surprising a definition for the placebo effect has proven so elusive. Conceivably, the same agent might be a placebo, a nocebo or an active intervention depending on who gives it, who gets it or when it is given.
The paradox of an effect without a mechanism weighs heavily on any conceptualisation of placebo effects. It seems to have become a heuristic that has outlived its usefulness and from the point of view of the advancement of the healthcare field, rethinking the idea is warranted. One option is to abandon the black box named placebo and where therapeutic effects of so-called placebo treatments are observed, attempt to identify and directly assess their cause. Supposing such effects are due to something ‘inert’ is illogical and gets us no closer to understanding the processes involved.
There are several promising lines of investigation that may help to elucidate what is responsible for placebo effects. In particular, the roles of patient expectations and conditioning appear promising in explaining outcomes.18 Researchers have also investigated the influence of patient–practitioner interactions and explored patient interpretation of the meaning of different clinical situations and encounters. Whether these factors produce a biochemical effect, work by altering mood or emotions, act via patient satisfaction or change the cognitive or subconscious evaluation of symptoms may all prove fruitful lines of research. Irrespective of the proposed putative mechanism, research should identify, a priori, the factors of interest and measure their influence as directly as possible.7
Considering the placebo effect in this manner has benefits for clinicians and for researchers. From a clinical perspective, understanding what specific factors (aside from the index intervention) are likely to maximise the treatment effect assists in provision of optimal care and also ameliorates some ethical issues with placebo use. With regard to research, abandoning the amorphous placebo of old in favour of consideration of the contextual, meaning, expectation, conditioning, etc factors that influence treatment effect size guides design and selection of appropriate control interventions for clinical trials.
Implications for research and practice
It must be recognised that all interventions, real or sham, in clinical practice or research, are delivered in the context of a treatment ritual. Commonly, researchers will want to control for any effects of the treatment ritual, in which case it must be standardised across the index and control groups. For example, a RCT might seek to test the effect of a topical anti-inflammatory, the control intervention might involve a gel that does not contain the anti-inflammatory compound. Obviously, the inert gel should look, smell and feel like the real gel but the contexts in which the two gels are administered should also be identical. This would include; how it is administered; the treatment rooms; the training, demeanour and presentation of the clinicians; verbal interaction with the patient; information or advice given; preparation of the patient; contact time; etc.
Only by understanding what elements a sham intervention actually controls for can we determine what may be responsible a difference in treatment effect between groups. Consider a study investigating the effect of a resistance exercise programme for knee pain, delivered by a sports scientist twice weekly in individually supervised, 1-h sessions in a gym. A control (placebo) intervention might involve the same exercises in the same format by the same practitioner but with no load applied. An alternative control (also placebo) intervention could involve detuned ultrasound, administered for 30 min three time per week by a research assistant in a hospital treatment room. Both interventions control for resisted exercises but the former example also controls for other potentially important elements. Hence, while a between group difference in a study using the former control is very likely due to the resistance training itself, other elements could be responsible for a difference observed in a study employing the latter control.
It is noted that placebo controls are not the only form of control treatment used in RCTs. Whether they are the most appropriate for any particular study is entirely dependent on the specific research question. As outlined above, different placebo interventions control for different aspects of the treatment of interest. Similarly, whether researchers choose a ‘usual care’ control, a ‘wait list’ control or an alternative intervention, comparison between the groups addresses different research questions. In any case, designation of a control intervention as ‘placebo’ is unnecessary and unhelpful.
The question of how to make use of placebo effects in the clinic is a difficult one. Currently, theories of expectancy and conditioning underpin our understanding of placebo effects; however, research into how these can be used to optimise patient outcome is far from conclusive. Key components of expectations include process expectations (what the patient thinks treatment will involve) and outcome expectations (what they think will happen with respect to their condition). Clinicians should elicit their patients’ process and outcome expectations and if these expectations do not align with their own expectations then the reasons should be explored and discussed. Second, in cases where more than one roughly equivalent treatment options are available a clinician might decide in favour of an intervention with which the patient has experienced a successful outcome previously.
The conception of placebos and placebo effects has developed significantly over the past 200 years in concert with changes in the nature and practice of healthcare as a whole. In the past 20 years, questions have arisen as to the clinical significance of placebo effects and perhaps more importantly, the logical basis of the concept itself. As things stand, enquiry into placebo effects faces the nonsensical situation of attempting to explain an effect that has no mechanism. Abandoning the placebo black box in favour of theory-directed research that specifically targets and investigates the cause of therapeutic effects offers several advantages. Effective techniques can be manipulated and incorporated in clinical practice to produce better outcomes for patients, and better control interventions can be designed to produce more robust clinical research.
The authors acknowledge the contribution of colleagues for helpful comments on earlier versions of the manuscript, particularly Dr James McAuley and Dr Amanda Hall.
Contributors SJK and CMW planned, wrote and revised the manuscript.
Funding SJK and CMW are both supported by the National Health and Medical Research Council of Australia.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed