Journal of Clinical Epidemiology 53 (2000) 157–161
`
`j j j
`Why use placebos in clinical trials? A narrative review of the
`methodological literature
`Andrew J. Vickersa,*, and Anton J. M. de Craenb
`aIntegrative Medicine Service and Biostatistics Service, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 11215, USA
`bDepartment of Clinical Epidemiology, Leiden University Medical Centre, Leiden, Netherlands
`Received 20 May 1998; revised 16 June 1999; accepted 23 June 1999
`
`Abstract
`
`Researchers have a number of different options for their choice of control or comparison intervention in randomized trials. We sur-
`veyed the methodological literature looking for reasons why a researcher might choose to administer a placebo (i.e., a mimic interven-
`tion) to control subjects. Two implicit assumptions were that the issue of placebo is pertinent only to drug trials and that the nonplacebo
`effect of a treatment is the “real” or “true” effect. Explicit reasons given in the literature for the use of placebos were facilitating blinding
`and controlling for the placebo effect. The importance of the latter was often inadequately argued. Reasons to avoid placebos in con-
`trolled trials, other than ethics and feasibility, are that placebos do not inform real decisions, and may interfere with accurate estimation
`of effect size and with nonspecific aspects of treatment. Placebo-controlled trials have high internal validity but may be difficult to apply
`to clinical practice; the situation is reversed for trials without placebo control. © 2000 Elsevier Sciences Inc. All rights reserved.
`
`Keywords: Placebos; Research design; Double blind method; Clinical trials; Randomised controlled trials
`
`1. Introduction
`
`Randomized clinical trials involve a comparison between
`at least two interventions. One can usually be termed the
`“experimental” intervention and is the treatment that the re-
`searchers are setting out to test. The other is the “control”
`intervention and is what the experimental intervention is
`tested against.
`Researchers have a number of different options for their
`choice of control intervention, including placebo, usual
`care, or some form of active intervention. More than one
`control intervention may be used in a trial or even in a sin-
`gle arm of a trial. The important distinction, and the focus of
`this article, is between placebo-controlled trials and those
`without placebo control. We define placebo as an interven-
`tion used in a clinical trial that is administered with the in-
`tention of mimicking some other intervention so that a com-
`parison can be made. Though the most common placebo is
`the “sugar pill” in drug trials, placebos can and have been
`used for physical interventions such as ultrasound [1], spi-
`nal manipulation [2, 3], and even surgery [4].
`The question we ask concerns the reasons why a re-
`
`searcher should choose to use a placebo—rather than usual
`care or active treatment—as the control (i.e., comparison)
`intervention in a clinical trial. Though this question has
`wide implications for clinical trial protocols, implications
`ranging as far as good clinical practice and informed con-
`sent procedures, much of the discussion on this issue has
`centered around ethics and feasibility, rather than methodol-
`ogy. For example, Campbell [5] discusses whether it is fea-
`sible to design a placebo that mimics acupuncture treatment
`appropriately. Similarly, Rothman and Michels [6] give nu-
`merous cases of placebo-controlled trials, such as those on
`antidepressants, which they consider unethical because con-
`trol patients were denied treatment of proven value.
`We are concerned with the situation in which it is both
`ethical and feasible to use a placebo in a clinical trial. We will
`argue that there are methodological reasons to use or avoid
`placebos in clinical trials and that these have not always been
`discussed appropriately in the literature on clinical trial meth-
`odology. Furthermore, we will argue that the choice between
`placebo or other control group depends on the research ques-
`tion being asked and may be thought of in terms of the dis-
`tinction between internal and external validity.
`
`2. Reasons to use placebos in clinical trials
`
`* Corresponding author. Tel.: 1-212-639-8629; fax: 1-212-794-5851.
`E-mail address: vickersa@mskcc.org
`
`We examined a number of journal articles [6–16] and
`textbooks [17–32] looking for reasons to use placebo con-
`
`0895-4356/00/$ – see front matter © 2000 Elsevier Science Inc. All rights reserved.
`PII: S0895-4356(99)00139-0
`
`SPRUCE - EXHIBIT 2006
`Neurocrine Biosciences, Inc. v. Spruce Biosciences, Inc.
`PGR2022-00025
`
`

`

`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`158
`
`A.J. Vickers, A.J.M. de Craen / Journal of Clinical Epidemiology 53 (2000) 157–161
`
`trols in clinical trials. We first report on the implicit as-
`sumptions that appear to be made when discussing placebo
`controls in clinical trials. We then discuss the explicit argu-
`ments for and against using placebos in clinical trials.
`
`3. Implicit assumptions about placebo
`
`One of the most immediately apparent features of the lit-
`erature is the prevalence of two implicit assumptions. First,
`it is generally assumed in the literature on the methodology
`of clinical trials that the issue of placebo is pertinent only to
`drug trials. An example would be a paper entitled “Scien-
`tific and ethical issues in the use of placebo controls in clin-
`ical trials” [7]. Though placebo is initially defined in gen-
`eral terms as “any therapeutic procedure . . . without
`specific activity” and the experimental group is at first de-
`scribed as “test therapy” or “intervention,” the substantive
`content of the paper is entirely concerned with drug trials:
`within a page, the “effect of a therapeutic intervention” be-
`comes the “pharmacologic effect.”
`The second implicit assumption is that the “true” or
`“real” effect of a treatment and its specific (i.e., nonplacebo)
`effect are the same thing. For example, Ernst and Resch [8]
`define a “true treatment effect” as the “perceived treatment
`effect minus . . . non-specific effects.” Similarly, Hennekens
`and Buring [23] contrast “psychological effects” of a drug
`with “true physiologic benefits” and Gordis [21] states that
`“the placebo plays a major role in identifying . . . the real
`benefits of an agent.”
`We disagree with both of these assumptions. First, the is-
`sue of placebo is clearly not confined to drug trials. Sec-
`ondly, the total effect of a treatment, as experienced by the
`patient, includes both a specific and a nonspecific compo-
`nent. There is no reason to define one part of the total treat-
`ment effect as the “real” or “true” effect.
`
`4. Explicit arguments for the use of placebos
`in clinical trials
`
`A number of explicit arguments for the use of placebos
`in clinical trials can also be found in the literature.
`
`1.
`
`2.
`
`The use of a placebo group controls for the placebo
`
`effect. Rickels [15], for example, argues that placebo
`groups control for suggestibility, “milieu or psycho-
`therapy”; Altman [17] points out that “the act of tak-
`ing some treatment may itself have some benefit to
`the patient.”
`According to
`Placebos facilitate outcome blinding.
`Hennekens and Buring [23], for example, the use of a
`placebo will minimize bias in the “ascertainment of
`both subjective disease outcomes and side-effects.”
`Active controls can also enable blinding, particularly
`in drug trials. So this is primarily an argument about
`the advantages of placebo control over both no treat-
`
`3.
`
`4.
`
`5.
`
`ment control and active treatment control when this is
`unlike the test intervention (e.g., physiotherapy con-
`trol in a chiropractic trial).
`Placebos promote similarity of clinician and patient
`Feinstein [19]
`compliance in each comparison group.
`argues that patients may withdraw from a trial, or be
`less compliant with follow-up procedures, if they are
`aware that they are in the control group. Similarly, cli-
`nicians may be tempted to give those patients not re-
`ceiving the experimental intervention some form of
`compensatory care.
`Placebos control for natural remission and nontrial
`Pledger
`
`et al. [14] claim that placebo
`interventions.
`controls are used “to control for spontaneous symp-
`tom change or remission.”
`Placebo controls have advantages over active con-
`Where an intervention is compared with a stan-
`trols.
`dard therapy, it can be difficult to know whether it has
`activity above that of placebo. If results in both
`groups are equivalent, a claim of greater than placebo
`activity for the test treatment involves an implicit,
`nonrandomized comparison with the placebo arm of
`an earlier study of the control treatment [12,13,16].
`As such, placebos are often demanded by regulatory
`authorities to provide a standard comparison.
`
`Of these five points, the first two clearly hold: placebos
`control for the placebo effect and, by enabling blinding,
`control bias when assessing outcome. Point 3 is really an
`extension of point 2: in this case, blinding allows control of
`performance bias. Point 4 cannot be used to decide whether
`placebos should be used: placebos control for natural dis-
`ease variation but then so do other types of control. Point 5
`is an extension of the first point: active control trials are
`seen as inferior to placebo-controlled trials because they fail
`to exclude the possibility that a test treatment has no greater
`than placebo effects.
`The two arguments for the use of placebos in clinical tri-
`als are therefore that it facilitates blinding and that it con-
`trols for the placebo effect. There is at least some empirical
`evidence that blinding affects the results of randomized tri-
`als [33,34]. However, what are the arguments for control-
`ling for the placebo effect?
`
`5. Why control for the placebo effect in clinical trials?
`
`For the purposes of this article, we will define placebo
`effects as those which are attributable to factors other than
`the characteristic content of the intervention under trial [35].
`Such factors have been categorized by Feinstein [19] into
`three components: passive, active, and psychic. The passive
`component refers to the natural course of disease. The ac-
`tive component refers to co-interventions accompanying the
`main intervention: an example would be such as dietary
`therapy provided alongside a hypoglycemic drug in diabe-
`tes. The psychic (or psychological) component refers to the
`
`

`

`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`A.J. Vickers, A.J.M. de Craen / Journal of Clinical Epidemiology 53 (2000) 157–161
`
`159
`
`patient’s anticipation of benefit. The main argument in the
`literature about placebo effects concerns this third, psycho-
`logical, component of the placebo effect.
`Just as with the use of words like “true” and “real” to de-
`scribe specific effects, methodologists seem to take it as ob-
`vious that the psychological component of the placebo ef-
`fect should be eliminated in clinical trials. Klerman [11], for
`example, describes how placebos control for “attention,”
`“the patient’s expectations and hope for improvement,” and
`even “the cognitive reframing of the meaning given by the
`patient to his or her illness.” These are said to “obscure and
`confound the demonstration of drug efficacy.” However, it
`is not stated why we should be interested in “efficacy”
`rather than in the total effects of treatment including the
`psychological component of the placebo effect.
`Fletcher
`
`et al. [20] are rare in giving an explicit reason
`why controlling the placebo effect is important: “it is also
`useful to know what part of the total effect is specific and
`what is nonspecific so as to avoid dangerous, uncomfortable
`or costly interventions when relatively little of their effect
`can be attributed to their specific actions.” This argument is,
`however, dubious. Even if an intervention is “dangerous,
`uncomfortable and costly,” patients and clinicians will want
`to know first, the overall benefits relative to costs and risks,
`and second, how this compares to other possible interven-
`tions. They should choose the intervention with the largest
`benefit to cost and risk ratio, irrespective of the proportion
`of benefit that is attributable to the psychological compo-
`nent of placebo.
`The underlying assumption of Fletcher
`et al.’s argument
`
`is that if an intervention is a placebo, it can be replaced by
`some other placebo intervention to the same effect. This in
`turn involves assuming that there is a single, relatively con-
`stant “placebo effect” that can be elicited readily and the
`size of which is constant across different interventions. The
`evidence on the placebo effect suggests, conversely, that it
`varies enormously and that it is modified by numerous pa-
`tient and clinician variables, including practitioner expecta-
`tion [36], patient expectation [37], the color of medication
`[38], and previous exposure to treatment [39]. The size of
`an intervention’s nonspecific component is particular to that
`intervention. As such, two interventions may have different
`effects on patient outcome even though both were equiva-
`lent to placebo in clinical trials.
`We believe that there are good reasons to control for the
`placebo effect in clinical trials, even though these have
`never been made explicit in the literature. First, it is worth
`asking whether certain therapies have more than placebo ac-
`tion in order to plan research strategically. Research com-
`paring different dosages and formulations of a drug is not
`appropriate until it has been shown, in a placebo-controlled
`trial, that dosage and formulation had anything to do with
`the drug’s therapeutic effect: for example, trials comparing
`high- and low-dose nicotine gum took place only after the
`superior effectiveness of nicotine gum over placebo had
`been demonstrated [40].
`
`A second reason for controlling the placebo effect in
`clinical trials is that the size of the specific effect of an inter-
`vention arguably is more stable than that of its placebo com-
`ponent. The psychological effects of a treatment are likely
`to vary considerably depending on the outlook of the indi-
`vidual patient and clinician, current therapeutic fashions,
`media hype, and so on. New interventions announced in a
`blaze of excitement and publicity (be they laetrile or
`Prozac) can result in large placebo effects that may dissipate
`over time.
`
`6. Reasons to avoid the use of placebos in clinical trials
`
`In our survey of the methodological literature, surpris-
`ingly few authors described reasons other than ethics or fea-
`sibility to avoid placebos in clinical trials.
`
`1.
`
`2.
`
`Placebo-controlled trials may interfere with accurate
`
`
`Fletcher et al. [20] argue that:
`estimation of effect size.
`“what is important to clinicians and their patients is
`the total effect of the intervention beyond what would
`have otherwise occurred in the course of disease with-
`out treatment.” Placebo-controlled trials measure only
`the specific effect of treatment. In the case of thera-
`pies for which the psychological component of pla-
`cebo is large, placebo-controlled trials may show only
`a small effect size even when the total effect, includ-
`ing placebo, is of significant clinical value. We have
`examined this issue at greater length elsewhere [41]
`as has Feinstein [19], who argues that the role of “ia-
`trotherapy” is often overlooked in the evaluation of
`therapeutic techniques. The importance of nonspecific
`effects may also be of particular importance in drug
`cessation trials. For example, in a recent study of an-
`kle edema that compared open withdrawal of diuretic
`drugs to continued treatment, the authors argued that:
`“for general practitioners it is more relevant to know
`the total effect of stopping drugs openly” than to “sep-
`arate the pharmacological effect and the placebo ef-
`fect” [42]. Furthermore, important outcomes can
`sometimes depend on knowledge of treatment assign-
` [43], for instance, randomized 716
`ment. Little
`et al.
`primary care patients with sore throats to a prescrip-
`tion for antibiotics, no prescription, or a prescription
`that could be collected if symptoms persisted for 3
`days. In addition to measuring duration of illness, the
`study looked at outcomes such as intention to consult
`a doctor for similar attacks in the future, satisfaction
`with care, and antibiotic use. The authors argued that
`these outcomes could best be assessed if a trial “mim-
`icked perceptions and behaviour of everyday prac-
`tice.”
`Placebo-controlled trials interfere with nonspecific
`Numerous aspects of the clini-
`aspects of treatment.
`cian–patient relationship have been identified as mod-
`ifying the placebo response. Giving a prescription
`
`

`

`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`160
`
`A.J. Vickers, A.J.M. de Craen / Journal of Clinical Epidemiology 53 (2000) 157–161
`
`3.
`
`with confidence, for example, has been shown to im-
`prove outcome [44]. Increased treatment effects have
`also been linked to patient expectation of benefit
`[45,46]. However, it seems probable that a placebo-
`controlled trial will reduce both clinician confidence
`and patient expectation: it is difficult to give a double-
`blind prescription with enthusiasm.
`Placebo-controlled trials do not inform real deci-
`Clinical practice does not normally involve a
`sions.
`decision between giving a treatment or a placebo.
`Typically, the decision that needs to be made is which
`of several treatments should be given or whether any
`treatment is preferable to no intervention. It is this
`second set of choices that should therefore determine
`the control groups in clinical trials. For instance,
`knowing that antidepressant A is better than placebo
`and that antidepressant B is also better than placebo
`may be useful, but does not help a psychiatrist choose
`which drug to prescribe.
`
`7. Discussion
`
`Placebo-controlled trials facilitate blinding and control
`for the placebo effect. However, they do not give informa-
`tion, such as effect size or comparative effectiveness of real-
`world choices, which is useful for clinicians in the field.
`They therefore have high internal validity, or rigor, and low
`external validity, or generalizability. A trial without a pla-
`cebo control may provide information that is more immedi-
`ately applicable to clinical practice. However, except in cir-
`cumstances such as a comparison between two drugs,
`blinding will not be possible and the trial will be more prone
`to bias. Trials without placebo control therefore have higher
`external validity at the possible expense of internal validity.
`Like so many other aspects of research, the choice of
`control group will ultimately depend on the question being
`asked. If researchers are interested in whether an interven-
`tion has effects over and above those of placebo, a placebo-
`controlled trial is clearly indicated. However, researchers
`need to think carefully about why it is important to know
`the extent of an intervention’s nonplacebo effects. We have
`argued that the relevance of this question should not be as-
`sumed to be self-evident. If researchers are interested apply-
`ing their results to clinical practice, a usual care or active
`treatment control group may be more appropriate. In such a
`comparison, researchers will need to consider carefully how
`they will minimize bias. In some cases, the risks of bias will
`outweigh the disadvantages of placebo control, especially
`where the use of placebo is unlikely to interfere signifi-
`cantly with usual clinical care.
`In conclusion, whatever might be implicit in the method-
`ological literature, the choice of a placebo control group in a
`clinical trial is not simple or unproblematic. A placebo-con-
`trolled trial is useful in determining whether an intervention
`has a specific effect. However, trials without placebo con-
`trol may be required to provide information of clinical
`
`value, such as the effect size of an intervention relative to an
`alternative treatment strategy. Such trials provide, in the
`words of Knottnerus and Dinant [47], “medicine-based evi-
`dence,” the obvious prerequisite for evidence-based medi-
`cine. We suggest that a wider debate needs to take place on
`the use of placebo controls and, more generally, on the na-
`ture of questions to be asked in clinical trials.
`
`Acknowledgments
`
`The authors thank Asbjorn Hrobjartsson of the Nordic
`Cochrane Centre for his comments on a draft version of the
`manuscript.
`
`References
`
`[1] Ho KH, Hashish I, Salmon P, Freeman R, Harvey W. Reduction of
`post-operative swelling by a placebo effect. J Psychosom Res 1988;
`32:197–205.
`[2] Morgan JP, Dickey JL, Hunt HH, Hudgins PM. A controlled trial of
`spinal manipulation in the management of hypertension. J Am Osteo-
`path Assoc 1985;85:308–13.
`[3] Kokjohn K, Schmid DM, Triano JJ, Brennan PC. The effect of spinal
`manipulation on pain and prostaglandin levels in women with pri-
`mary dysmenorrhea. J Manipulative Physiol Ther 1992;15:279–85.
`[4] Moseley JB, Wray NP, Kuykendall D, Willis K, Landon G. Arthro-
`scopic treatment of ostearthritis of the knee: a prospective, random-
`ized, placebo-controlled study. Am J Sports Med 1996;24:28–34.
`[5] Campbell A. Hunting the snark: the quest for the perfect acupuncture
`placebo. Acupunc Med 1991;9:83–4.
`[6] Rothman KJ, Michels KB. The continuing unethical use of placebo
`controls. N Engl J Med 1994;33:394–7.
`[7] Clark PI, Leaverton PE. Scientific and ethical issues in the use of pla-
`cebo controls in clinical trials. Annu Rev Public Health 1994;15:19–38.
`[8] Ernst E, Resch KL. Concept of true and perceived placebo effects.
`BMJ 1995;311:551–3.
`[9] Haegerstam G, Huitfeldt B, Nilsson BS, et al. Placebo in clinical drug
`
`
`trials—a multidisciplinary review. Methods Find Exp Clin Pharmacol
`1982;4:261–78.
`[10] Keating JC. The placebo issue and clinical research. J Manipulative
`Physiol Ther 1987;10:329–32.
`[11] Klerman GL. Scientific and ethical considerations in the use of pla-
`cebo controls in clinical trials in psychopharmacology. Psychophar-
`maco Bull 1986;22:25–29.
`[12] Leber P. The placebo control in clinical trials (a view from the FDA).
`Psychopharmaco Bull 1986;22:30–2.
`[13] Makuch RW, Johnson MF. Dilemmas in the use of active control
`groups in clinical research IRB Rev Hum Subjects Res 1989;11:1–5.
`[14] Pledger GW, Hall D. Active control trials: Do they address the effi-
`cacy issue?. Proc Am Stat Soc 1986:1–10.
`[15] Rickels K. Use of placebo in clinical trials. Psychopharmacol Bull
`1986;22:19–24.
`[16] Temple R. Government viewpoint of clinical trials. Drug Inform J
`1982;16:10–7.
`[17] Altman DG. Practical statistics for medical research. London: Chap-
`man and Hall 1991.
`[18] Campbell MJ, Machin D. Medical Statistics: a Commonsense Ap-
`proach. Chichester: John Wiley and Sons, 1993.
`[19] Feinstein AR. Clinical Epidemiology. The Architecture of Clinical
`Research. Philadelphia, PA: W.B. Saunders Company, 1985.
`[20] Fletcher RH, Fletcher SW, Wagner EH. Clinical Epidemiology. The
`Essentials (3rd ed.). Baltimore, MD: Williams & Wilkins, 1996.
`[21] Gordis L. Epidemiology. Philadelphia: W.B.G. Saunders Company,
`1996.
`
`

`

`
`
`
`
`
`
`A.J. Vickers, A.J.M. de Craen / Journal of Clinical Epidemiology 53 (2000) 157–161
`
`161
`
`[22] Greenberg RS. Medical Epidemiology. Norwalk, CT: Appleton &
`Lange, 1993.
`[23] Hennekens CH, Buring JE. Epidemiology in Medicine. Boston, MA:
`Little, Brown and Company, 1987.
`[24] Hulley SB, Cummings SR. Designing Clinical Research. Baltimore,
`MD: Williams & Wilkins, 1988.
`[25] Knapp RG, Miller III MC. Clinical Epidemiology and Biostatistics.
`Baltimore, MD: Williams & Wilkins, 1992.
`[26] Kramer MS. Clinical Epidemiology and Biostatistics. Berlin/Heidel-
`berg: Springer-Verlag, 1988.
`[27] Meinert CL. Monographs in Epidemiology and Biostatistics. Clinical
`Trials. Design, Conduct, and Analysis (vol. 11). New York: Oxford
`University Press, 1986.
`[28] Morton RF, Hebel JR, McCarter RJ. A Study Guide to Epidemiol-
`ogy and Biostatistics (3rd ed.). Gaithersberg, MD: Aspen Publish-
`ers, Inc., 1989.
`[29] Pocock SJ. Clinical trials: A Practical Approach. Chichester: John
`Wiley and Sons, 1983.
`[30] Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical Epidemiol-
`ogy. Boston, MA: Little, Brown and Company, 1991.
`[31] Spilker B. Guide to Clinical Studies and Developing Protocols. New
`York: Raven Press, 1984.
`[32] Weiss NS. Monographs in Epidemiology and Biostatistics. Clinical
`Epidemiology: The Study of the Outcome of Illness (vol. 11). New
`York: Oxford University Press, 1986.
`[33] Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of
`bias. JAMA 1995;273:408–412.
`[34] Colditz GA, Miller JB, Mosteller F. How study design affects out-
`comes in comparisons of therapy. I: Medical. Stat Med 1989;8(4):
`441–54.
`[35] Turner JA, Deyo RA, Loeser JD, Von Korff M, Fordyce WE. The im-
`portance of placebo effects in pain treatment and research. JAMA
`1994;271:1609–14.
`
`[36] Gracely RH, Dubner R, Deeter WR, Wolskee PJ. Clinicians’ expecta-
`tions influence placebo analgesia. Lancet 1985;i:43.
`[37] Pastorello EA, Codecasa LR, Gerosa A, Buonocore E, Sillano V, Za-
`nussi C. The role of suggestion in asthma. I. Effects of inactive solu-
`tion on bronchial reactivity under bronchoconstrictor or bronchodila-
`tor suggestion. Ann Allergy 1987;59:336–8.
`[38] de Craen AJM, Roos PJ, de Vries AL, Kleijnen J. Effects of colour of
`drugs: systematic review of perceived effect of drugs and of their ef-
`fectiveness. BMJ 1996;313:1624–6.
`[39] Voudouris NJ, Peck CL, Coleman G. The role of conditioning and
`verbal expectancy in the placebo response. Pain 1990;43:121–8.
`[40] Silagy C, Mant D, Fowler G, Lancaster T. Nicotine replacement ther-
`apy for smoking cessation. In: Lancaster T, Silagy C, Fullerton D (ed-
`itors) Tobacco Addiction Model of The Cochrane Database of Sys-
`tematic Reviews. Available in The Cochrane Library [database on
`disk and CDROM] The Cochrane Collaboration; Issue 4. Oxford:
`Update Software, 1998.
`[41] de Craen AJM, Vickers AJ, Tijssen JGP, Kleijnen J. The number-
`needed-to-treat and placebo-controlled trials. Lancet 1998;351:310.
`[42] de Jonge JW, Knottnerus JA, van Zutphen WM, de Bruijne GA, Stru-
`ijker Boudier HAJ. Short term effect of withdrawal of diuretic drugs
`prescribed for ankle oedema. BMJ 1994;308:511–3.
`[43] Little P, Williamson I, Warner G, Gould C, Gantley M, Kinmonth
`AL. Open randomised trial of prescribing strategies in managing sore
`throat. BMJ 1997;314:722–7.
`[44] Thomas KB. General practice consultations: is there any point in be-
`ing positive? BMJ 1987;294:1200.
`[45] Gryll SL, Katahn M. Situational factors contributing to the placebo
`effect. Psychopharmacology 1978;47:253–61.
`[46] Skovlund E. Should we tell trial patients that they might receive pla-
`cebos? Lancet 1991;337:1041.
`[47] Knottnerus JA, Dinant GJ. Medicine based evidence, a prerequisite
`for evidence based medicine. BMJ 1997;315:309–10.
`
`