Osteoarthritis and Cartilage (2009) 17, 1255e1262

ª 2009 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.doi:10.1016/j.joca.2009.03.023

InternationalCartilageRepairSociety

ReviewThe ‘‘placebo’’ response in osteoarthritis and its implications forclinical practiceM. Dohertyy* and P. DieppezyUniversity of Nottingham, Nottingham, UKzPeninsula Medical School, Plymouth, UK

Summary

Many observations support a major biological effect from the way in which people interpret the meaning of each component of their medicalexperience and the context in which this occurs. A recent systematic review of randomised controlled trials in osteoarthritis has demonstratedthat the effect size of ‘‘placebo’’ is substantial and is usually greater than that obtained from the specific effect of an individual treatment. In thecontext of a randomised controlled trial, such a large placebo or ‘‘meaning’’ response is considered a nuisance, but in the context of clinicalpractice the optimisation of such meaning and contextual responses, through enhanced ‘‘care’’, could greatly benefit people who suffer fromosteoarthritis.ª 2009 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.

Key words: Placebo, Osteoarthritis, Clinical trials, Meaning response, Context effects.

Introduction

We recently published a systematic review and meta-anal-ysis of placebo response in osteoarthritis (OA) randomisedcontrolled trials (RCTs)1. This confirmed the appreciable ef-fect size (ES) of ‘‘placebos’’ on pain relief and identifiedsome of the factors that may determine the size of this ef-fect in RCTs. These findings clearly are of relevance tothe design of OA RCTs and to the interpretation of thetrue ES of treatments. More importantly, however, these re-sults have important implications for clinical practice. Herewe will: (1) briefly review the history, nomenclature andchanging perspective of ‘‘placebo’’; (2) summarise the find-ings of our meta-analysis; (3) present related key observa-tions taken largely from the pain and mental healthliterature; and finally (4) explain how this knowledge mightinfluence our behaviour as clinicians and the care that weoffer patients with OA.

Historical changes in perspective concerningplacebo

Prior to the mid 20th century it was commonplace for doc-tors to prescribe placebos with the assurance that theywould lead to health and coping benefits2. This deliberateprescription of inert products was described by Thomas Jef-ferson in 1807 as the ‘‘pious fraud’’3 and in the Lancet in1954 as ‘‘humble humbug’’4. However, at a time when therewere few effective treatments such deception was

*Address correspondence and reprint requests to: MichaelDoherty, Academic Rheumatology, Clinical Sciences Building,City Hospital, Nottingham NG5 1PB, UK. Tel: 44-115-8231756;Fax: 44-115-8231757; E-mail: michael.doherty@nottingham.ac.uk

Received 17 March 2009; revision accepted 22 March 2009.

1255

considered both ethical and benevolent2,4,5. Placebomeans ‘‘I will please’’ and up until the 1950s the prevailingview was that products such as bread pills, coloured water,or water injections, would do no harm and by a psychologi-cal trick could comfort patients with incurable conditions, es-pecially those low in intellect or ‘‘inadequate’’ and thus moresusceptible to suggestion4,5.

In the second quarter of the 20th century, however, two par-allel developments changed this perspective. Firstly, the ad-vances in effective drug discovery, and secondly thedevelopments in medical statistics and epidemiology thatled to introduction of the RCT as the gold standard for clinicaltrials. In a landmark publication in 1955 entitled ‘‘The Power-ful Placebo’’6 Henry Beecher described the occurrence inRCTs of real, not imagined, objective benefits from placeboin around 35% of patients. He reasoned that the total benefitfrom a treatment equals the ‘‘real’’ or specific effect of thetreatment plus the non-specific placebo effect, and impor-tantly concluded that response to placebo was irrespectiveof IQ. Subsequent research focused on how this effect mayarise, and amongst mechanisms suggested were the Haw-thorne effect (behaviour change through being observed);patient expectation of improvement; a ‘‘meaning response’’;classic conditioning; relief of anxiety; contextual healing; re-sponse bias; and the patientepractitioner interaction2,7,8.

There has been much controversy over the meaning ofthe word ‘placebo’ and it is important to distinguish placeboresponses (sometimes called ‘context effects’) from pla-cebo interventions9. Arguably the most useful definition ofplacebo (and nocebo) effects is that provided by Grun-baum in the 1980s10,11; he distinguishes the characteristiceffects of an intervention for the target condition, from allthe non-characteristic (or incidental) effects related to theuse of that intervention. Thus, for example, a sugar pillmight be used as a placebo tablet in a trial of an

1256 M. Doherty and P. Dieppe: ‘‘Placebo’’ response in OA e implications for clinical practice

antidepressant drug since sugar is not thought to have anyspecific (characteristic) effect on depression e but thesame could not be said for diabetes. And if a sugar pill pla-cebo was used in a trial of antidepressant therapy, then itseffects will depend on a huge number of contextual factorssurrounding its administration, as outlined below. Patersonand Dieppe have pointed out that in many complex inter-ventions used for conditions such as OA these non-charac-teristic (context) effects may interact with the characteristiceffects9. This not only makes it more difficult to interpret theresults of trials, it can invalidate the use of the classicalRCT as a way of assessing the value of such interventions(Fig. 1).

At the turn of this century the existence of placebo re-sponse was challenged. In a systematic review of RCTsthat randomised participants to either placebo or no treat-ment (observation only), Hrobjartsson and Gotzsche foundno significant benefit from placebo for binary or objectiveoutcomes, but possible small benefits for continuous sub-jective outcomes and for the treatment of pain12. Althoughthere were many problems with this review and itsmethodology was seriously challenged13,14, this publicationrekindled the debate about placebos. Alternative explana-tions for apparent placebo benefits in RCTs included un-measured concomitant treatments and the patient’sdesire to please the doctor (the ‘‘patient’s placebo’’), butalso impersonal statistical artefacts of measurement suchas natural fluctuations in disease severity and regressionto the mean12,15e17. A key reason for such controversyand confusion is that inclusion of a placebo in RCTs isthe accepted gold standard for the assessment of interven-tion effects, and unless a no-treatment control is also in-cluded (an approach often considered unethical) it isimpossible to ascertain what effect the placebo mighthave e in effect the RCT is a way of deliberately ignoringplacebo responses9. It was because of such controversyand confusion that we decided to examine whether there

treatment factors

interventiongroup

effects in eac

incidentalelements

(placebo response/meaning effect)

unrelated totreatment

characteristicelements

characte(specific)

inciden(placebo)

Characteristic and incidental

effects interact

Fig. 1. The effects of any treatment can be divided into ‘characteristic effand incidental effects (also called non-characteristic effects, context effeccombination of all the other factors affecting the outcome (box on left of dincidental element is often larger than that of the characteristic element.interaction between characteristic elements (e.g., muscle strengthening) acan lead to under-estimation of the effect of an intervention if a classical R

ence between characteristic elements

was evidence for placebo response in OA RCTs, lookingspecifically for trials with no treatment control groups, andto identify possible determinants of the size of any sucheffects1.

Placebo effect in RCTs of OA

We identified by systematic review 198 trials that met ourinclusion criteria. These comprised 193 placebo groups(16,364 patients) and 14 untreated control groups (1167 pa-tients) and examined a range of non-pharmacological, phar-macological and invasive treatments. There was clearevidence to support positive benefits from placebo. Forpain relief the overall ES (the standard mean difference be-tween baseline and endpoint) was 0.51 [95% confidence in-terval (CI) 0.46e0.55] for placebo, but almost 0 foruntreated controls (ES 0.03, 95% CI �0.13 to 0.18). ThisES was even higher (0.77, 95% CI 0.65e0.89) in the threetrials that had head-to-head comparison between placeboand no treatment. The ES for pain relief was also higher(0.71, 95% CI 0.64e0.78) in the 15 trials that gave no res-cue analgesia, perhaps reflecting a greater expectation ofpowerful effect if no rescue was offered. Placebo wasalso effective for other subjective outcomes, for examplestiffness (ES 0.43, 95% CI 0.38e0.49), self-reported func-tion (ES 0.49, 95% CI 0.44e0.54) and interestingly, physi-cian’s global assessment (ES 0.66, 95% CI 0.53e0.78).

With respect to determinants of magnitude, the followingshowed significant independent effects for the primary out-come of pain relief:

The ES of the treatment: the higher the treatment ESthe higher the placebo effect, perhaps explained byhigher expectation of benefit by participants.The baseline level of pain: the higher this was, thegreater the placebo ES.

control group

h group

measureddifference

actual treatment effect

ristic effect

tal effect

ects’ (the specific effects of the intervention on the target disorder),ts, the meaning response or placebo effects), which are a complexiagram). For problems such as pain and depression the size of theIn many complex interventions, such as physiotherapy, there is annd incidental elements (e.g., motivation by the physiotherapist). This

CT is used to test it, as such trial designs only measure the differ-and non-characteristic elements.

1257Osteoarthritis and Cartilage Vol. 17, No. 10

The sample size: the larger the study numbers thehigher the placebo effect. This might be explained bythe power requirements to differentiate the relativelysmall additional benefit of treatments over placebo.Similarly, the funnel plot of ES for pain relief showeda skewed distribution, suggesting that trials withsmaller placebo effects are more likely to be published,perhaps because it is easier to demonstrate superiorityof a treatment when this occurs.The route of delivery: RCTs involving repeated nee-dling (acupuncture, intra-articular injection of hyalur-onan) had the highest placebo effects (Fig. 2).Interestingly, the one steroid injection study that gaveserial injections had a higher placebo ES than studiesinvolving single injection. RCTs of topical NSAID alsohad higher than average placebo ES.

Such determinants provide reassurance of the validity ofour main finding. Therefore, we found unequivocal evidenceof a large placebo response in OA, particularly for pain re-lief. Notably the ES of placebo (0.5e0.7) is much largerthan that achieved with most of our conventional therapies,such as analgesic and anti-inflammatory drugs (0.2e0.3)18.Recent publications showing similar results in depressiontrials19,20 attracted widespread publicity, leading one scien-tific author to suggest that the effectiveness of antidepres-sants was a myth21 and some Press to wrongly concludethat conventional therapy was of little or no value.

The determinants of ‘‘placebo responses’’

Although hardly studied in OA, there is a large literatureon placebo and the non-specific effects of care in otherareas of medicine. Examples, especially from pain andmental health studies, are worth examining here sincethey are largely generic and therefore relevant to OA.Many observations point to a major biological effect fromthe way people interpret the meaning of each componentof their medical experience and the context in which it oc-curs. The term ‘‘meaning response’’ encapsulates these ex-planatory mechanisms, and placebo is just one example ofthe responses that are integral to any clinical encounter ortreatment8,22. Although most examples relate to drugs, theprinciples probably extend to any intervention.

COLOUR AND NUMBER

In a classic experiment medical students were given ei-ther one or two blue or pink tablets and told that one wasa stimulant and the other a sedative, though in fact bothwere inert. Overall, pink tablets caused stimulant effectsand blue tablets sedation, and two tablets had more effectthan one23. Clearly these effects are nothing to do with re-gression to the mean or disease fluctuation, but are ex-plained by the meaning associated with the colour (pinkfor hot, blue for cool) and the expectancy that a double ismore potent than a single dose. A systematic review in1996 concluded that green and blue may have more seda-tive, and red and orange more stimulant effects, and thatthis difference is reflected in the colour of marketed drugs,perhaps boosting their efficacy24. In an Italian study of sed-ative placebos, however, blue tablets had the usual seda-tive effect in women but the opposite effect in men25,perhaps because blue is the national soccer team colourand associates in men with feelings of excitement. Thus itis not the colour per se that influences the outcome, butthe meaning attributed to it by an individual.

BRANDING AND COST

Having the confidence that the treatment you are receiv-ing is reputable and ‘‘high class’’ makes a difference to out-come. In a placebo-controlled trial of aspirin for relief ofheadache, patients were randomised to one of four treat-ments: aspirin labelled with a well-known brand; plain un-branded aspirin; placebo marked with the same brand; orplain placebo. Aspirin worked better than placebo, butbranded tablets worked better than their unbranded coun-terparts26. A recent US placebo analgesic study alsoshowed that patients who were told that their tablets wereobtained at full cost ($2.50) obtained better pain reliefthan those who were told their tablets were discounted atjust 10 cents27.

METHOD OF DELIVERY AND FREQUENCY

In general, the larger the intervention the higher the pla-cebo effect. As in our systematic review1, greater placebo ef-fects have been reported with injections than with oralmedication in many conditions varying from pain relief in mi-graine28 to control of hypertension29 (Fig. 2). Also greater pla-cebo benefit from a higher frequency of administration hasbeen observed not just for sedation or stimulation23 but forother conditions including healing of peptic ulcer disease30.

Surgery, of course, is one of the ultimate treatments interms of magnitude. In OA, the report that invasive arthro-scopic lavage, with or without debridement, is no betterthan sham procedure31 caused considerable controversy32

since it implied that the benefit to the patient resulted fromthe expectancy and experience of the intervention ratherthan to whatever was done inside the joint. Understandably,for ethical reasons, there are very few placebo-controlledstudies of surgery. However, in 195933 and 196034 twosmall controlled studies examined the efficacy of bilateralinternal mammary ligation (BIMAL) for treating angina.This procedure was widely practiced at the time; the ratio-nale being that ligation of these arteries would redirectmore blood through the coronary arteries. The combined re-sults of the two studies showed that 73% of patients (24/33)undergoing BIMAL had significant improvement with re-duced angina, better exercise tolerance, and reduced ni-trate consumption. However, 83% of patients (10/12) whoreceived sham surgery (anaesthesia, chest incision, arteryexposed but not ligated) showed similar substantial im-provement. Because of prevailing ethical considerations,whether many currently performed technical proceduresgive any benefit above the experience of surgery itself willremain open to question.

RESPONSE EXPECTANCY AND CONCEALMENT

The expectation of what a treatment will do is an impor-tant determinant of the outcome. In a study of aerobic exer-cise one group was told that their aerobic capacity shouldincrease after 10 weeks of the programme, and the othergroup was told that their fitness would improve but also theirwellbeing would be enhanced. Both groups equally im-proved aerobic capacity but only the second group reportedfeeling better35. In a study of the physiological effects of cof-fee, one group was given either regular or decaffeinatedcoffee in a double blind design, while a second group wastold they would all receive regular coffee, but in fact wasgiven decaffeinated. Increase in pulse rate, alertness andblood pressure changes were greater in the second group(who were all expecting such effects) than in those givendecaffeinated coffee in the first group (who believed they

Fig. 2. Treatments delivered by needling, such as acupuncture (shown above, with burning moxa) and intra-articular injection (shown below)seem to associate with greater expectancy, higher placebo effects, and more ‘‘magic’’.

1258 M. Doherty and P. Dieppe: ‘‘Placebo’’ response in OA e implications for clinical practice

had only a 50:50 chance of having active coffee)36. Such re-sponse expectancy can even reverse the pharmacologicaleffect of a drug. For example, ipecachuana (a potentemetic) can improve nausea if patients are told that theyare receiving an anti-emetic37.

Knowing that a treatment is being given is important. Ina post-operative analgesia study, patients who could seemorphine being administered into their line obtained rapidand effective pain relief, whereas patients who received mor-phine by concealed administration obtained much slowerbenefit, implying that the initial relief (from our strongest anal-gesic) is largely placebo effect. Equally, open discontinuationof morphine led to rapid return of pain whereas covert discon-tinuation did not38. Other studies of pain, anxiety and Parkin-son’s disease have similarly shown a decrease in treatmentefficacy when the patient is unaware that they are receivingthem38. This difference between open and covert administra-tion is equivalent to the ‘‘placebo’’ component of the treat-ment, and investigating this difference has been suggestedas an alternative strategy, at least for some treatments, toa classic placebo RCT design.

BEHAVIOURAL CONDITIONING

In a study of the immunosuppressive effects of cyclospor-ine A, the drug was given to healthy volunteers togetherwith a novel green liquid drink. The expected degree of

immunosuppression was observed. After ‘‘conditioning’’the subjects in four sessions over three consecutive days,the experiment was repeated the following week with pla-cebo tablets given with the same drink. The repeatexposure, with dummy tablets, resulted in immunosupres-sion39. The authors of this work point out that althoughthe mechanism of such responses remains to be eluci-dated, it might be possible to enhance the effects of immu-nomodulatory and other drugs by behavioural conditioninginterventions e an approach that could have widespreadimplications in medicine.

HOW ‘NEW’ THE INTERVENTION IS?

In studies of interventions for various conditions, includ-ing depression and peptic ulcer healing, there are data tosuggest that drugs or other interventions that are ‘new’ re-sult in larger placebo responses than those which are es-tablished22,40, suggesting that meaning is enhanced ifpatients and practitioners believe that the intervention is‘modern’.

PROVIDER EFFECTS

Patients are influenced by the expectation and behaviourof the practitioner. In a double blind dental pain study pa-tients were told that they would receive either fentanyl

1259Osteoarthritis and Cartilage Vol. 17, No. 10

(which might reduce pain), the opioid antagonist naloxone(which might increase pain) or a placebo. However, earlyin the study the investigators, but not the patients, weretold that due to administrative problems fentanyl was un-available so patients would only be randomised to the nal-oxone or placebo arms (i.e., only two ‘‘negative’’treatments). Later in the study investigators were informedthat fentanyl was now available and included in the adjustedrandomisation procedure. Patients receiving placebo in thesecond phase of the study experienced significant pain re-lief but patients on placebo in the first phase did not41. Thusthe clinician’s optimism or pessimism concerning treatmenthas a dynamic effect on the patient’s outcome.

A recent audit has shown that intentional prescribing of‘‘placebo treatments’’, such as vitamins and over the counterproducts regarded as inert, is commonplace amongst US in-ternists and rheumatologists, and that usually these are de-scribed to the patient as a potentially beneficial treatmentnot usually used for their condition42. Placebos are even com-mercially available in the US as ‘‘obecalp’’ (placebo speltbackwards), ‘‘cebocap’’ (placebo capsules) and ‘‘jujubes’’(for parents to administer to their children). However, irre-spective of the ethical issues of not being transparent witha patient, it would appear that placebo is less likely to givebenefit when it is knowingly administered.

BIOLOGICAL EFFECTS, ADHERENCE AND DEATH

There is growing evidence that context effects and mean-ing responses can effect alterations in body systems (suchas immune function, mentioned above39) in addition tochanges in psyche and behaviour. For example, in somepain studies placebo effects are abolished by naloxone, im-plicating endogenous opioid release as one mechanismthat can convert expectancy to ‘‘real’’ analgesia43e45.

Fig. 3. The importance of environment. A ward with a bleak view (left) elic

Neuroimaging studies have made major contributions tothe field recently46. For example, positron-emission tomog-raphy with [11C] carfentanil has been used to providegraphic evidence that placebo analgesia causes endoge-nous opioid activity in the m-opioid receptor rich regionsthat play a central role in pain and affect47.

The ultimate ‘‘reality’’ of placebo and meaning response isexemplified by studies of adherence and death. It is logical toexpect that people who adhere to a treatment usually do bet-ter than non-adherers, but the same difference occurs withplacebo. In a study of antibiotic prophylaxis following chemo-therapy48 non-adherers to placebo had twice the infectionrate of adherers (64% vs 32%), and in a study of chlorprom-azine for schizophrenia49 non-adherers to placebo had twicethe relapse rate of adherers (80% vs 40%). This differentialoutcome from adherence to placebo extends to mortality. Ina large RCT examining the effect of the cholesterol-loweringagent clofibrate on 5-year survival following myocardial in-farction, participants who took >80% of study medicationsdid better than non-adherers, specifically: 15% vs 25% mor-tality for those on clofibrate; and 15% vs 28% for those on pla-cebo50. In addition to heart disease, reduced mortality inadherers to placebo is reported in studies of diabetes, immu-nosuppression following transplantation, and anti-retroviraltreatment for AIDS51. Therefore adherence to what you be-lieve to be a beneficial treatment may prevent developmentof disease and even death.

ENVIRONMENT AND CONTEXT

The environment in which you receive treatment canmodify your outcome. Matched patients who had under-gone gall bladder surgery were allocated to one of twowards which were similar apart from their outlook e onehad a pleasant green-field view and the other over-lookeda high wall and parking lot. The patients with the natural

its worse outcomes than a ward with a positive natural view (right).

Fig. 4. The difference between a positive encounter (above) and a negative encounter (below) with a practitioner can influence the outcome forthe patient. It is our professional responsibility to address such aspects.

1260 M. Doherty and P. Dieppe: ‘‘Placebo’’ response in OA e implications for clinical practice

view had fewer post-operative complications, required lessanalgesia, made fewer complaints and were dischargedearlier than those with the bleak view52 (Fig. 3).

In an attempt to dissect out separate components ofplacebo Kaptchuk et al. recently undertook a study ofsham acupuncture in patients with irritable bowel syn-drome. Patients were randomised to one of three groups:waiting list (assessment and observation); sham acupunc-ture with only limited patientepractitioner interaction (thetherapeutic ‘‘ritual’’); or sham acupuncture with moreusual patientepractitioner interaction augmented bywarmth, attention and confidence. At 3 weeks the propor-tions in each group with adequate pain relief were 28%(observation), 44% (ritual) and 62% (augmented interac-tion). A similar trend was observed for quality of life(3.6, 4.1, 9.3) and all other measured outcomes53. Clearlyall three components contributed to benefit but the mostrobust was patientepractitioner interaction.

The importance of the patientepractitioner interactionwas also the focus of a study in general practice in which200 patients with symptoms but no abnormal signs wererandomly assigned to either: (1) a ‘‘positive’’ consultation,in which the patient was given a confident diagnosis and re-assured that things would soon improve; or (2) a ‘‘negative’’consultation, where the doctor admitted ‘‘I cannot be certainof what is the matter with you’’. In both groups the doctorcould also prescribe thiamine tablets as ‘‘placebo’’ medica-tion. Two weeks later, 64% of patients who had the positive

consult were better, compared to 39% of those who had thenegative encounter. Receiving a prescription made no dif-ference. The overriding factor appeared to be the patient re-sponse to the certainty of diagnosis and reassuranceconcerning prognosis54. Other contextual aspects that aresuggested to encourage a beneficial meaning response in-clude: when the patient regards the practitioner as experi-enced and competent55,56; when the clinician is optimisticthat the treatment will help55,56 and when the practitionerwishes to see the patient again to monitor progress57.

Implications for clinical practice

OA is a condition in which the pathology does not alwayscorrelate with the symptoms58. Thus someone may havesevere joint damage but no pain, or severe pain but onlyminor joint pathology. As outlined above, placebo re-sponses can result in big effects on symptoms (the illness),but we have not found evidence that they affect X-rays andpathology (the disease). The fact that placebo responsesare best documented in other painful conditions and inmental health disorders, suggests that the value of placebointerventions is mainly for patient symptoms and distress.These are the principal treatment targets in patients withOA. We believe that we should be more aware of thepower of the placebo response to relieve pain and sufferingin people with OA, and that we should learn how to use itbetter.

1261Osteoarthritis and Cartilage Vol. 17, No. 10

From the above it is apparent that in the context of a clin-ical encounter many facets of the patient’s experience de-termine their outcome from any intervention. In audits ofcare, including OA59, common patient complaints are:

that the doctor was ‘‘too busy’’ to listen,that the doctor undertakes only a superficial examina-tion, or no examination at all,that the doctor didn’t address key concerns,that the doctor did not give a follow-up appointment.

Given these circumstances it is obvious that practitionersshould capitalise on the impact of context effects to en-hance the benefits to their patients, as a professional re-sponsibility. Many of these aspects are nothing to do withthe ‘‘ritual’’ of a drug or physical treatment. When a practi-tioner gives a patient their focused and unhurried attentionin a comfortable environment, listens to them, undertakesa thorough examination, explains in understandable termswhat is happening, addresses concerns and providesa way forward, the patient experiences a ‘‘good deal’’.This is in itself a treatment (Fig. 4). The patient often leavesboth looking and feeling better, long before they cash in any(adjunctive) prescription they have just received. The con-verse is also true. Recent studies of the ‘‘nocebo’’ effect(expectation of a negative effect leading to worse outcome)have shown, for example, that negative words that increaseanxiety can result in increased pain60.

The use of the term ‘meaning response’ to cover what wehave here called non-characteristic or context effects8,22 em-phasises the importance of the meaning of a consultation andany suggested intervention to the patient. The meanings ofinterventions are culturally determined and depend in parton the health beliefs of the patient. Exorcism may work betterthan aspirin if a patient firmly believes that their pain is causedby their possession by evil spirits61. Health beliefs can be eli-cited in a clinical encounter, and measured for research pur-poses62. Furthermore, in many cases practitioners can helpthe patient towards a health belief that is more in tune withthe interventions that are available to help them.

Practitioners of complementary and alternative medicine(CAM) often do this very well, and seem ahead of manyof us more traditional physicians. We often pride ourselveson being over-worked and too busy to spend time on such‘‘peripheral’’ matters, and feel more useful and importantwhen we exercise our licence to prescribe potentially dan-gerous drugs. We often label practitioners of CAM as char-latans and explain their treatment success as ‘‘just placeboeffect’’, apparently oblivious of the surprisingly large ES ofsuch ‘‘non-treatment’’ benefits. But if we did learn from theresearch literature, from practitioners of CAM, and from sim-ple observation, and optimise these meaning responses8,22

in our clinical practice, the benefits of such ‘‘contextual heal-ing’’7 to the population of people with OA would be huge.

Conflict of interest

The authors have no conflict of interest to declare.

References

1. Zhang W, Robertson J, Jones A, Dieppe PA, Doherty M. The placeboresponse and its determinants in osteoarthritis e meta-analysis ofrandomised controlled trials. Ann Rheum Dis 2008;67:1716e23.

2. Kaptchuk TJ. Powerful placebo: the dark side of the randomised con-trolled trial. Lancet 1998;351:1722e5.

3. Jefferson T. In: Ford PL, Ed. The Writings of Thomas Jefferson. NewYork: GP Putnam’s; 1898.

4. Anon. The humble humbug. Lancet 1954;ii:321.5. Pepper OHP. A note on the placebo. Am J Pharm 1945;117:409e12.6. Beecher HK. The powerful placebo. JAMA 1955;159:1602e6.7. Miller FG, Kaptchuk TJ. The power of context: reconceptualising the pla-

cebo effect. J R Soc Med 2008;101:222e5.8. Moerman D. Meaning, Medicine and the ‘‘Placebo Effect’’. Cambridge,

UK: Cambridge University Press; 2002.9. Paterson C, Dieppe P. Characteristic and incidental (placebo) effects in

complex interventions such as acupuncture. BMJ 2005;330:1202e5.10. Gunbaum A. The placebo concept. Behav Res Ther 1981;19:157e67.11. Gunbaum A. The placebo concept in medicine and psychiatry. Psychol

Med 1986;16:19e38.12. Hrobjartsson A, Gotzsche PC. Is the placebo powerless? An analysis of

clinical trials comparing placebo with no treatment. N Engl J Med2001;344:1594e602.

13. Spiegel D, Kramer H, Carlson R. Is the placebo powerless? N Engl JMed 2001;345:1276.

14. Miller FG, Rosenstein DL. The nature and power of the placebo effect. JClin Epidemiol 2006;59:331e5.

15. McDonald CJ, Mazucca SA, McCabe GP. How much of the placebo‘effect’ is really statistical regression? Stat Med 1983;2:417e27.

16. Kienle GS, Kiene H. The powerful placebo: fact or fiction? J Clin Epide-miol 1997;50:1311e8.

17. Gotzsche PC. Is there logic to the placebo? Lancet 1994;344:925e6.

18. Osteoarthritis e National Clinical Guideline for Care and Management inAdults. NICE and Royal College of Physicians Guidelines on Osteoar-thritis, <www.NICE.org>; 2008.

19. Turner EH, Mathews AM, Linardatos E, Tell RA, Rosenthal R. Selectivepublication of antidepressant trials and its influence on apparent effi-cacy. N Engl J Med 2008;358:252e60.

20. Kirsch I, Deacon JB, Huedo-Medina TB, Scobaria A, Moore TJ,Johnson BT. Initial severity and antidepressant benefits: a meta-anal-ysis of data submitted to the Food and Drug Administration. PLoSMed 2008;5(2):e45.

21. Ionnidis J. Effectiveness of antidepressants: an evidence myth con-structed from a thousand randomized trials? Philos Ethics HumanitMed 2008;3:14.

22. Moerman DE, Jonas WB. Deconstructing the placebo effect and findingthe meaning response. Ann Intern Med 2002;136:471e6.

23. Blackwell B, Bloomfield SS, Buncher CR. Demonstration to medical stu-dents of placebo responses and non-drug factors. Lancet 1972;1(763):1279e82.

24. de Craen AJM, Roos PJ, de Vries LA, Kleijnen J. Effect of colour ofdrugs: systematic review of perceived effect of drugs and of their ef-fectiveness. BMJ 1996;313:1624e6.

25. Cattaneo AD, Lucchilli PE, Filippucci G. Sedative effects of placebotreatment. Eur J Clin Pharmacol 1970;3:43e5.

26. Branthwaite A, Cooper P. Analgesic effects of branding in treatment ofheadaches. BMJ 1981;282:1576e8.

27. Waber RL, Shiv B, Carmon Z, Ariely D. Commercial features of placeboand therapeutic efficacy. JAMA 2008;299:101607.

28. de Craen AJM, Tijssen JG, de Gans J, Kleijsen J. Placebo effect in theacute treatment of migraine: subcutaneous placebos are better thanoral placebos. J Neurol 2000;247:183e8.

29. Grenfell RF, Briggs AH, Holland WC. A double-blind study of the treat-ment of hypertension. JAMA 1961;176:124e8.

30. de Craen AJ, Moerman DE, Heisterkamp SH, Tytgat GN, Tijssen JG,Kleijnen J. Placebo effect in the treatment of duodenal ulcer. Br JClin Pharmacol 1999;48:853e60.

31. Moseley JB, O’Malley K, Petersen NJ, Menke TJ, Brody BA,Kuykenall DH, et al. A controlled trial of arthroscopic surgery for oste-oarthritis of the knee. N Engl J Med 2002;347:81e8.

32. Zhang W, Moskowitz RW, Nuki G, Abramson S, Altman RD, Arden N,et al. OARSI recommendations for the management of hip and kneeosteoarthritis. Part II: OARSI evidence-based, expert consensusguidelines. Osteoarthritis Cartilage 2008;16:137e62.

33. Cobb L, Thomas GI, Dillard GH, Merendino KA, Bruce RA. An eval-uation of internal mammary artery ligation by double blind technic.N Engl J Med 1959;260:1115e8.

34. Dimond EG, Kittle CF, Crockett JE. Comparison of internal mammary li-gation and sham operation for angina pectoris. Am J Cardiol 1960;5:483e6.

35. Desharmais R, Jobin J, Cote C, Levesque L, Godin G. Aerobic exerciseand the placebo effect: a controlled study. Psychosom Med 1993;55:149e54.

36. Kirsch I, Weixel LJ. Double-blind versus deceptive administration ofa placebo. Behav Neurosci 1988;102:319e23.

37. Wolf S. Effects of suggestion and conditioning on the action of chemicalagents in the human e the pharmacology of placebos. J Clin Invest1950;29:703e9.

1262 M. Doherty and P. Dieppe: ‘‘Placebo’’ response in OA e implications for clinical practice

38. Colloca L, Lopiano L, Lanotte M, Benedetti. Overt versus covert treat-ment for pain, anxiety and Parkinson’s disease. Lancet Neurol2004;3:679e84.

39. Goebel M, Trebst A, Steiner J, Xie Y, Exton M, Frede S, et al. Behaviou-ral conditioning of immunosuppression is possible in humans. FASEBJ 2002;16:1869e73.

40. Walsh B, Seidman S, Sysko R, Gould M. Placebo responses in studiesof major depression: variable, substantial and growing. JAMA 2002;287:1840e7.

41. Gracely RH, Dubner R, Deeter WR, Wolskee PJ. Clinicians’ expecta-tions influence placebo analgesic. Lancet 1985;1:43.

42. Tilburt JC, Emanuel EJ, Kaptchuk TJ, Curlin FA, Miller FG. Prescribing‘‘placebo treatments’’: results of national survey of US internists andrheumatologists. BMJ 2008;337:a1938.

43. Amanzio M, Benedetti F. Neuropharmacological dissection of placeboanalgesia: expectation-activated opioid systems versus conditioning-activated specific subsystems. J Neurosci 1999;19:484e94.

44. Amanzio M, Pollo A, Maggi G, Benedetti F. Response variability to an-algesics: a role for non-specific activation of endogenous opioids.Pain 2001;90:205e15.

45. Gracely RJ, Dubner R, Wolskee PJ, Deeter WR. Placebo and naloxonecan alter post-surgical pain by separate mechanisms. Nature 1983;306:264e5.

46. Lidstone SC, Stoessl AJ. Understanding the placebo effect: contribu-tions from neuroimaging. Mol Imaging Biol 2007;9:176e85.

47. Wager TD, Scott DJ, Zubieta JK. Placebo effects on human mu-opioidactivity during pain. Proc Natl Acad Sci USA 2007;104:11056e61.

48. Pizzo PA, Robichaud KJ, Edwards BK, Schumaker K, Kramer BS,Johnson A. Oral antibiotic prophylaxis in patients with cancer: a dou-ble-blind randomised placebo-controlled trial. J Paediatr 1983;102:125e33.

49. Hogarty GE, Goldberg SC. Drug and sociotherapy in the aftercare ofschizophrenic patients. One year relapse rates. Arch Gen Psychiatry1973;28:54e64.

50. Coronary Drug Project Research Group. Influence of adherence to treat-ment and response of cholesterol on mortality in the coronary drugproject. N Engl J Med 1980;303:1038e41.

51. Simpson SH, Eurich DT, Majumdar SR, Padwal RS, Tsuyuki RT,Varney J, et al. A meta-analysis of the association between adher-ence to drug therapy and mortality. BMJ 2006;333:15e8.

52. Ulrich RS. View through a window may influence recovery from surgery.Science 1984;224:420e1.

53. Kaptchuk TJ, Kelley JM, Conboy LA, Davis RB, Kerr CE, Jacobson EE,et al. Components of the placebo effect: randomised controlled trial inpatients with irritable bowel syndrome. BMJ 2008;336:999e1003.

54. Thomas KB. General practice consultations: is there any point in beingpositive? BMJ 1987;294:1200e2.

55. Thomas KB. The placebo in general practice. Lancet 1994;344:1066e7.56. Barford TS. Placebo therapy in dermatology. Clin Dermatol 1999;17:

69e76.57. Bernstein CN. Placebos in medicine. Semin Gastrointest Dis 1999;10:3e7.58. Dieppe P, Lohmander S. Pathogenesis and management of pain in os-

teoarthritis. Lancet 2005;365:967e73.59. OA Nation, <www.arthritiscare.org.uk.OANation>;; 2004.60. Benedetti F, Lanotte M, Lopiano L, Collaco L. When words are painful:

unravelling mechanisms of the nocebo effect. Neuroscience 2007;147:260e71.

61. Kleinman A. The Illness Narratives: Suffering, Healing and the HumanCondition. London, UK: Basic Books (Harper Collins); 1988.

62. Moss-Morris R, Weinman J, Petrie K, Horne R, Cameron L, Buick D. Therevised Illness Perception Questionnaire (IPQ-R). Psychol Health2002;17:1e16.