German Journal of Psychiatry
Roger Lane MD, MPH
Pfizer Inc., 235 East 42nd Street, New York, NY 10017-5755, USA
Administering a simple sugar pill or injecting water can alleviate symptoms or even cure disease – as long as patients believe they are getting a real drug. Much has been made of the “miraculous powers” of placebo. It has even been speculated that antidepressants are little more than sophisticated placebos. However, a distinction should be made between the specific therapeutic response to placebo and the improvement that occurs in a patient who is being treated with placebo. In randomized controlled trials (RCTs) of antidepressants where an “active” drug treatment is compared against inert placebo pill in patients with affective disorders, such as major depression, neither the patient or the treating physician knows the treatment that is being administered. In these studies the response to placebo is often at least half of that to drug and is sometimes so high that the response to drug treatment is not statistically significantly greater than that to placebo. This phenomena has been used to infer that antidepressant drug treatments are not particularly effective, and indeed that they may have no intrinsic activity whatsoever: The perception of side effects by the patient reinforcing the belief that they are receiving active treatment and resulting in an improved response to treatment. Examples, taken mainly from the extensive database of RCTs for the selective serotonin reuptake inhibitor (SSRI) antidepressant sertraline, will be used to provide reassurance as to the efficacy of pharmacotherapy in the treatment of patients with mood and anxiety disorders (German J Psychiatry 1999;2 (3):1-11).
There has been much recent discussion of the placebo-response rate in clinical trials in psychiatry (Kirsh & Sapirstein, 1998; Enserink, 1999). Randomized controlled trials (RCTs) that demonstrate significantly greater treatment effects for active treatment relative to placebo are the most rigorous evidence of efficacy available. Regulatory agencies responsible for the approval of new treatments require evidence from RCTs that a treatment is more effective than placebo and they require this evidence to be replicated. This article adds to these recent discussions and is mainly illustrated with information from the sertraline database of RCTs.
The term placebo response is potentially misleading. The benefit in the placebo group response is not due to inert sugar pill, but a non-specific treatment response, whereas the additional benefit in the experimental group will due to the specifics of active treatment. Spontaneous remission, regression to the mean as patients tend to present for treatment when their symptoms are at their worst, patient and physician expectations of treatment, and the healing effects of the structured environment of the clinical trial setting are just some of the factors that can confound the determination of treatment effects (Ernst & Resch, 1995). Simply enrolling in a trial and receiving help and support, possibly for the first time, from experienced caring health professionals who caringly enquire about the patients life, may result in substantial clinical improvement. The rigorous methodology of a randomized, controlled experiment is required to control for these, and other, confounding variables.
The placebo response is particularly apparent in illnesses that have a strong psychological component, such as anxiety and depression. Moreover, it has been suggested that antidepressants have no specific action in these disorders: The placebo effect being accentuated in patients receiving antidepressants due to the perception of side effects (Kirsch & Sapirstein, 1998). These ‘unblinding’ effects occur when a supposed double-blind design is subverted because the different physiological experiences associated with ingestion of an active drug and an inert placebo lead subjects and assessors to suspect the identity of the medication (Greenberg & Fischer, 1994). It has been confirmed that some antidepressants, among other drugs, can be distinguished from placebo (White et al, 1992). Some research indicates that unblinding can produce spurious positive results in the absence of a real effect (Engelhardt et al, 1969; Toneatto & Sellers, 1992). Placebos containing active substances have been used to address this problem (Moncrieff et al, 1998). However, the milder side effect profiles of newer antidepressants means that it is more difficult to functionally unblind placebo-controlled studies of these agents. Moreover, the development of tolerance to many of their side effects may result in a side effect burden that approaches that seen on placebo in longer term treatment (Doogan & Caillard, 1992). Furthermore, clear cut differences between active treatment and placebo have emerged in studies where patients and physicians have been unable to identify treatments correctly when asked to do so (Malt et al, 1999; Blomholf et al, 1999). Thus, while researchers must remain vigilant to the possibility of functional unblinding in RCTs, due to the occurrence of adverse events attributable to active drug treatment, this may be less of a problem with newer agents and in longer term treatment studies.
The placebo response appears to be less stable than the improvement attributable to drugs (Dago & Quitkin, 1995). For example, patients who had responded in a 6-week placebo-controlled study to citalopram were randomized to continue on the same dose of citalopram or to receive placebo for a further 24-weeks (Montgomery et al, 1993). Patients switched to placebo had re-emergence of their depressive symptoms at a significantly higher rate than patients maintained on citalopram. Furthermore, the relapse rate in patients who had responded to placebo during the 6-week acute treatment study, who were continued double-blind with placebo, was similar to the rate in the placebo-substituted patients. Furthermore, the quality of a placebo response may not be as good as an apparently similar response on active treatment. It is widely acknowledged that, in addition to affording symptomatic relief and control of the disease process, a major aim of therapeutic intervention should be to improve quality of life. This is especially true for the treatment of psychiatric disorder. The “true” value of a drug therapy cannot be assessed without including the patient’s own evaluation of the outcome of treatment. Outcome measures commonly used in depression studies, such as the Hamilton symptom rating scale for depression (HAM-D), do not examine the full impact of treatment on broader aspects of patient well-being where important differences between therapies may be apparent. Other domains of functioning (e.g., sense of well-being and life satisfaction), broadly conceptualized as quality of life are increasingly being recognized as valid outcome measures. Data from 351 patients in two placebo-controlled studies of sertraline in panic disorder demonstrated that, although placebo-responders showed similar reductions in panic attack frequency, they experienced significantly less improvement in quality of life compared to sertraline-responders (Rappaport et al, 1998). Similar findings for sertraline responders relative to placebo responders have been found in patients with major depression and premenstrual dysphoric disorder (Endicott et al, 1999).
The tolerability of many of the newer antidepressants in long term treatment approximates that of placebo (Doogan & Caillard, 1992). It is of interest, therefore, that following a period of open treatment (over 12 months in some studies) patients randomized to placebo experienced a significantly greater re-emergence of symptoms of their mood or anxiety disorder than patients continuing to receive sertraline (Doogan & Caillard, 1992, Keller et al, 1998; Koran et al, 1999; Clayton et al, 1999). In these studies patients receiving placebo experienced a 3.5 –fold greater relapse or recurrence of symptoms of major depression (Doogan & Caillard, 1992), a 4.1 -fold greater recurrence of symptoms of chronic/double depression (Keller et al, 1998), a 2.9 – fold greater relapse of symptoms of OCD (Koran et al, 1999), and a 2.3-fold greater relapse of panic disorder symptoms (Clayton et al, 1999).
The apparent re-emergence of symptoms in patients randomized to placebo in relapse studies of the SSRI, paroxetine, has been suggested to be due in part to the misattribution of discontinuation symptoms (Michelson et al, 1998). However, similar to fluoxetine and in significant contrast to paroxetine, sertraline has a low incidence of discontinuation reactions following withdrawal of treatment (Fava et al, 1998). Furthermore, the 11-month placebo-controlled data in the study of Doogan & Caillard (1992) was re-examined for relapse of depressive symptoms in the initial 4 months following switch to placebo and for recurrence of depressive symptoms in the remaining 7 months (Montgomery et al, 1991). The protective effect of sertraline was robustly significant during both time periods. The clinically meaningful effects of sertraline and other SSRIs in preventing the re-emergence of depressive and anxiety symptoms in RCTs using a discontinuation design is a very clear indication of the efficacy of these agents in the long-term management of these chronic disorders. If antidepressants were no more than “sophisticated placebos” then why would significantly more patients see a re-emergence of their symptoms when switched to a real placebo than those continuing to receive active treatment?
In his article in Science, Enserink (1999), made much of the recent announcement by Merck that they were discontinuing development of MK-869, a substance P antagonist, for the treatment of major depression. The reason given was the finding in a recent clinical trial that the drug was not significantly better than placebo. However, MK-869 had already been demonstrated in one published study to be more effective than placebo and at least as effective as an active comparator drug , the SSRI, paroxetine. Knowledgeable researchers in this area are aware that the placebo response in psychiatric clinical studies is often so high as to invalidate the experiment. That the study was a failed trial and not a negative trial was demonstrated by the fact that the SSRI comparator in the study, an approved antidepressant, also failed to differentiate from placebo. A more likely reason for Merck’s abrupt discontinuation of the development of MK-869 is adverse findings in long term toxicity studies that would have precluded its development as a treatment for a chronic condition such as major depression. Merck’s announcement that they are testing a back up compound to MK-869 as an antidepressant means they clearly have not lost faith in the antidepressant properties of this class of compounds.
Side effects may confound the determination of efficacy in RCTs. The HAM-D is inherently biased towards heterocyclic antidepressant drugs, in that typical side effects such as sedation, somnolence and increased appetite/body weight are rated as improvements in depression. This may be the reason why a faster onset of therapeutic effect has rarely been demonstrated for a SSRI relative to a TCA despite the usual initiation of the former on a therapeutic dose and the need to titrate the latter to a therapeutic dose (Fournier et al, 1997). In addition, it has been suggested that certain acute adverse events associated with the SSRIs, such as nausea and gastrointestinal disturbances, insomnia or other sleep disturbances, asthenia, nervousness and agitation may act to mimic depressive symptomatology (Walczak et al, 1996).
Clearly there is a potential confounding of the key efficacy assessment by antidepressant side effects in acute treatment. It is possible, therefore, that the efficacy of SSRIs relative to TCAs may have been underestimated in acute treatment studies. In the 6-month study of Malt et al (1999) in 372 depressed patients the response to mianserin was initially greater than that of sertraline, perhaps due to the favorable effects of mianserin side effects on sleep and appetite. However, response in the mianserin group, similar to the placebo group, appeared to plateau after 12 weeks, in contrast to continuing improvement in the sertraline. At the end of treatment mianserin was not significantly different from placebo, in contrast to sertraline. In addition, many of the acute treatment studies performed during the initial development of SSRIs employed rapid upward dose titration designs, allowing TCAs to be titrated to therapeutic doses, but resulting in doses of SSRIs and incidences of side effects that were higher than necessary (Preskorn & Lane, 1995). Many of the confounding SSRI side effects appear to be dose related and tend to decline with continued treatment.
Despite almost identical changes of the HAM-D, some studies comparing SSRIs and TCAs/MAOIs in patients with affective disorders have shown significant differences on patient-rated quality of life measures in favor of SSRI treatment (Lydiard et al, 1998; McEntee et al, 1996; Sogaard et al, 1999). In addition, amitriptyline responders in the 8-week RCT of Lydiard et al (1998) in patients with major depression did not show significant quality of life improvements over placebo responders, in contrast to sertraline responders (Endicott et al, 1999). Furthermore, subsequent analysis of nortriptyline and sertraline-treated elderly patients (> 60 years) with major depression meeting response criteria (n = 102, 59% response rate) in a 12-week study, showed patients receiving sertraline had significantly greater quality of life improvements in endpoint and completers relative to nortriptyline responders (Endicott et al, 1999). Similarly, sertraline responders demonstrated significantly greater improvement in quality of life scores than desipramine responders in a placebo-controlled study in premenstrual dysphoric disorder (Endicott et al, 1999). This difference was mirrored by assessments of occupational functioning, family responsibilities and social activities which evidenced significantly greater improvement in sertraline responders.
The natural history of many psychiatric disorders is to wax and wane. Careful patient selection for patients that have more unremitting and severe forms of the illness may lessen the likelihood of failed RCTs due to a high response rate in the placebo group. However, such patients are often the hardest to recruit as they may be relatively rare relative to patients with mild forms of the illness. In addition, an experimental treatment study, that includes a 50% chance of receiving a placebo treatment for several months, may be relatively less attractive to more severely ill patients and the clinicians who treat them. Reluctance to participate is particularly likely when effective treatment is widely available.
Early placebo-controlled studies with clomipramine showed a large treatment effect with a minimal placebo response (Collaborative Clomipramine Study, 1991). At the time of these initial studies there were many severely ill patients who had been under the care of clinicians with no effective treatment available. Expectations of treatment efficacy were low. However, the population of patients entering OCD clinical studies has changed over time. More recent patient populations may be less severely ill and have higher expectations of treatment. In addition, later SSRI trials had significant numbers of patients who had previously failed trials of clomipramine or one of the other SSRIs. More recent studies with SSRIs have demonstrated greater placebo responses and consequently smaller relative treatment effects, although all SSRIs have been demonstrated to be significantly better than placebo (Bisserbe and Lane, 1996). Efficacy data from four double-blind placebo-controlled trials of sertraline in the treatment of OCD were combined to determine if there were significant differences in outcome for patients who had previously received trials of SSRIs or clomipramine (Rasmussen et al, 1997). Over 450 patients on sertraline were included in the combined analysis. A significantly (p < .001) greater response was found by week 4 in therapeutically naïve patients compared to patients having previous treatment, on both primary outcome measures (CGI-I responders and Yale Brown Obsessive Compulsive scale mean score). The imputed week 10 effect size for sertraline was approximately equal to the week 10 endpoint effect size in the Collaborative Clomipramine Study (1991) in which all patients were, by definition, serotonin reuptake inhibitor antidepressant naïve.
Double-blind studies directly comparing SSRIs and clomipramine give a clearer picture of the relative efficacy and tolerability of the different treatments. The results of a 16-week double-blind study in 168 patients with OCD comparing sertraline (50-200mg/day) and clomipramine (50-200mg/day) showed that sertraline was significantly more effective than clomipramine in improving study efficacy assessments (Yale-Brown Obsessive Compulsive Scale, National Institute of Mental Health Global Obsessive Compulsive Scale, Clinical Global Impression of Severity, Clinical Anxiety Scale) in the intent-to-treat, last observation carried forward analysis (p < .05) (Bisserbe et al, 1997). Treatments were similarly effective in the completer analysis. The greater efficacy for sertraline in the intent-to-treat population was attributable to significantly fewer premature treatment discontinuations for adverse events in the sertraline group (10.5%) than in patients who received clomipramine (26.6%).
Although improvement in the experimental group may be statistically greater than in the placebo group, the clinical relevance and the generalizability of such results to routine clinical practice has been questioned.
Conventional RCTs are studies of treatment efficacy, where the blinding of patients and providers assures that any differences in outcome reflect essential differences between alternative drugs, such as pharmacological differences. While randomized, double-blind clinical trials are the “gold standard” for comparison of treatment efficacy, conventional RCTs may not reflect the relative effectiveness of treatments under “real world” conditions. Key elements of the conventional randomized trial are designed to isolate differences in efficacy due to treatment assignment (i.e., strict selection criteria, placebo “washout”, blinding of patients and providers, and specific treatment protocols). Patients seen in primary care or community psychiatric practice may differ systematically from clinical trial patients and may be less clinically homogenous, less severely ill, less compliant, more likely to show spontaneous remission, frequently suffer from comorbid medical conditions and receive concomitant medications. Thus, attempts to control the variability in treatment process and outcome may limit the generalizability of RCT results to usual clinical practice (Simon, Wagner & Von Korf, 1995). “Real world” effectiveness depends not only on a treatment’s essential therapeutic properties but also on ease of administration, acceptability to patient and provider and impact on overall health care resource utilization.
The traditional TCAs, with their relatively complex dosing regimens and their greater tolerability problems, may perform better under controlled trial conditions than in everyday practice. These treatments may benefit from the higher levels of clinician training, more intensive follow up, and higher levels of compliance typically seen in clinical trials. In contrast, the advantages of SSRIs, such as improved tolerability and convenient once daily dosing without the need for titration to a therapeutic dose, may be minimized in the highly standardized environment of a controlled clinical trial. A blinded comparison would not observe the significant “real world” advantages, including the economic advantages, of a simpler treatment regimen, as blinding requires that patients assigned to different treatments receive indistinguishable care. Many of the design elements of clinical trials systematically favor more complex treatment that is better provided by specialists over simpler treatment more suited to delivery in primary care.
Concern over the generalizability of randomized clinical trials have led to increased use of observational data to compare patient outcomes. However, while observational data sources include representative samples of patients treated by representative practitioners, treatment comparisons using such data must be interpreted cautiously. Clinician’s decisions regarding alternative treatments are not random, creating significant opportunity for bias. However, RCTs, mainly using the SSRI, sertraline, have been designed that are close to the reality of good clinical practice (Malt et al, 1999; Doogan & Langdon, 1994; Blomholf et al, 1999; Steiner et al, 1998; Moon et al 1994). For example, the “naturalistic” design of the study of Malt et al (1999) included a 6 month duration of treatment (in keeping with consensus treatment guidelines), was in a primary care setting (where over 90% of depressed patients are treated), administered a standardized regimen of emotional support and counseling (approximating the reality of good primary care practice), and included all patients with depressive symptoms thought to require antidepressant treatment (regardless of whether symptoms fulfilled diagnostic criteria for a major depressive episode). These more naturalistic-type studies demonstrate significant efficacy for sertraline relative to placebo, but heterocyclic (mianserin) and TCA (dothiepin) comparators failed to differentiate from placebo at study endpoint.
Medical knowledge comes from a number of sources, not only RCTs. However, a well conducted RCT investigating a clinically relevant question is a powerful tool that can provide something near to certainty. A placebo- or “non-specific treatment-” response rate is to be expected in such studies. This does not invalidate the efficacy of active pharmacological treatment if such treatment is shown to produce a response that is significantly greater than placebo. Rather a high placebo response rate illustrates that other factors, such as emotional support, counseling, hope, contact with health professionals, etc, are also important to achieve an optimal clinical response. However, information from SSRI studies clearly demonstrate that the stability and quality of the “response” on placebo is not as great as a “response” on, in the examples used, sertraline. This may also be true of other antidepressants. Studies employing a relapse design where patients who have responded or remitted on pharmacological therapy are randomized under double-blind conditions to continue active treatment or to receive placebo provide powerful evidence of the efficacy of antidepressant drug treatment.
The essential rubric of the RCT does not have to be abandoned to answer questions pertinent to routine clinical practice. The use of more representative samples of patients treated by more usual practioners in more typical contexts, and for a sufficient duration of time, may increase the generalizability of results to routine clinical practice. In addition, the narrow focus on symptom-based surrogate markers of clinical improvement in conventional RCTs may not adequately capture the full spectrum of treatment response, and the impact on quality of life and perceived functioning. Outcome measures assessing the broader aspects of response should be incorporated into clinical studies to assess the full impact of treatment on patient well being.
Bisserbe J-C, Lane R (1996) Effect size, placebo response, and tolerability in clinical studies in obsessive-compulsive disorder [abstract]. Eur Neuropsychopharmacol 6: 112
Bisserbe JC, Lane RM, Flament MF, and the Franco-Belgian OCD Study Group (1997) A double-blind comparison of sertraline in outpatients with obsessive-compulsive disorder. European Psychiatry 12: 82-93
Blomholf S, Haug TT, Humble M, Hallstrom K, Madsbu HP, Wold JE, Holme I. Treatment of generalized social phobia. New Research Abstracts 152nd Annual Meeting American Psychiatric Association, Washington, DC, May 15-20, 1999: NR650
Clayton A, Lydiard RB, Wolkow R, Rubin A, Hackett E. Efficacy of sertraline in long-term treatment of panic disorder: Results of a multicenter study. Poster presented at 152nd Annual Meeting of the American Psychiatric Association, Washington , DC May 15-20, 1999
Clomipramine Collaborative Study Group. Clomipramine in the treatment of patients with obsessive-compulsive disorder. Arch Gen Psychiatr 1991; 48: 730-738
Dago PL, Quitkin FM. Role odf the placebo response in the treatment of depressive disorders. CNS Drugs 1995; 4: 335-340
Doogan DP, Caillard V. Sertraline in the prevention of depression. Br J Psychiatry 1992;160:217-22.(92-01)
Doogan DP, Langdon CJ. A double-blind, placebo-controlled comparison of sertraline and dothiepin in the treatment of major depression in general practice. Int Clin Psychopharmacol 1994; 9: 95-100
Endicott J, Rapaport MH, Clary MC, O’Sullivan RL, Lane RM. Sertraline and quality of life across mood and anxiety disorders. Poster presented at 152nd Annual Meeting of the American Psychiatric Association, Washington , DC May 15-20, 1999
Engelhardt DM, Margolis RA, Rudorfer L, et al (1969) Physician bias and the double-blind. Archives of Psychiatry, 20: 315-320
Enserink M (1999) Can placebo be the cure? Science 284: 238-240
Ernst E, Resch KL. Concept of true and perceived placebo effects. BMJ 1995; 311: 551-3
Fava M et al (1998) A comparison of symptoms following treatment interruption: Evidence from a randomized double-blind trial with fluoxetine, sertraline and paroxetine. Poster resented XX1st CINP Congress, Glasgow, Scotland, July 12-16,1998.
Fournier J-P, Lane RM, Chouinard G, Watson DB, Amin M, Remick RA, Thorpe LU (1997) The efficacy and safety of sertraline versus imipramine in outpatients with major depression: a six-month, parallel, double-blind, multicenter study. Human Psychopharmacology 12: 203-215
Greenberg RP, Fischer S (1994) Suspended judgement. Seeing through the double-masked design: a commentary. Controlled Clinical Trials 15: 244-246
Keller MB, Kocsis JH, Thase ME, Gelenberg AJ, Rush AJ, Koran L, Schatzberg A, Russel J, Hirschfeld R, Klein D, McCullough JP, Fawcett JA, Kornstein S, LaVinge L, Harrison W, for the Chronic Depression Study Group. Maintenance phase efficacy of sertraline for chronic depression: A randomized controlled trial. JAMA 1998; 280: 1665-1672
Kirsch I, Sapirstein G. Listening to Prozac but Hearing Placebo. Prevention and Treatment 1998. Article 0002a (journals.apa.org/prevention).
Koran LM, Robinson DG, Hackett E, Rubin A, Wolkow R. Efficacy of sertraline in long-term OCD treatment: Preliminary results of a multicenter study. New Research Abstracts 152nd Annual Meeting American Psychiatric Association, Washington, DC, May 15-20, 1999: NR645
Kramer MS, Cutler N, Feighner J, Shrivastava, Carman J, Sramek JJ, Reines SA, Liu G et al. Distinct mechanism for antidepressant activity by blockade of central substance P receptors. Science 1998; 281: 1640-5
Lydiard RB, Stahl SM, Hertzman M, Harrison WM (1998) A double-blind, placebo-controlled study comparing the effects of sertraline versus amitriptyline in the treatment of major depression. J Clin Psychiatry 1997; 58: 484-491
Malt UF, Robak OH, Madsu H-P, Bakke O, Loeb M. The Norwegian naturalistic treatment study of depression in general practice (NORDEP) – I: randomized double blind study. 1999; 318: 1180-4
McEntee W, Ko G, Richter E Sertraline and nortriptyline: heart rate, cognitive improvement and quality of life in the depressed elderly. Eur Neuropsychopharmacol 1996; 6 (suppl 3): 36
Michelson D, Lydiard RB, Pollack MH, Tamura RN, Hoog SL, Tepner R, Memitrack MA, Tollefson GD, and Fluoxetine Panic Disorder Study Group. Outcome Assessment and Clinical Improvement in Panic Disorder: Evidence from a randomized controlled trial of fluoxetine and placebo. Am J Psychiatry 1998; 155: 1570-1577
Moncrieff J, Wessely S, Hardy R (1998) Meta-analysis of trials comparing antidepressants with active placebos. Br J Psychiatry 172: 227-231
Montgomery SA, Rasmussen JGC, Tanghoj P (1993) A 24-week study of 20 mg citalopram, 40 mg citalopram, and placebo in the prevention of relapse of major depression. Int Clin Psychopharmacol 8: 181-188
Moon CAL, Jago W, Wood K et al. A double-blind comparison of sertraline and clomipramine in the treatment of major depressive disorders and associated anxiety in general practice. J Psychopharmacol 1994; 8 (3):171-6. (94-216)
Preskorn SH, Lane RM (1995) Sertraline 50 mg daily – the optimal daily dose in depression. International Clinical Psychopharmacol 1995; 10(3):129-41.
Rapaport MH, Wolkow R, Clary CM (1998) Quality of life differences in sertaline and placebo responsive panic disorder patients. Poster presented at XXIst Congress of the Collegium Internationale Neuro-psychopharmacologicum, Glasgow, Scotland, July 12-16, 1998: PW14043
Rasmussen SA, Baer L, Shera D. Previous SRI treatment and efficacy of sertraline for OCD: combined analysis of four multicenter trials. Biological Psychiatry 1997; 42: 26S
Simon GE, Wagner E, Von Korf M (1995) Cost-effectiveness comparisons using “real world” randomized trials: The case of new antidepressant drugs. J Clin Epidemol;48:363-373
SÆgaard J, Latimer P, Behnke K, Christiansen PE, Nielsen B, Ravindran AV, Reesal RT, Goodwin DP, Lane RM (1999) A 12-Week Study Comparing Moclobemide and Sertraline in the Treatment of Outpatients with Atypical Depression. J Psychopharmacol (in press)
Steiner M, Browne G, Roberts J, Gafni A, Byrne C, Bell B, Dunn E. Sertraline and IPT in dysthymia: one year follow up. Eur Neuropsychopharmacol 1998; 8(suppl 2): S202
Toneatto T, Sellers EM (1992) Perception of whether drug or placebo has been administered is a determinant of drinking reduction. In: Sellers EM, Naranjo CA (eds) Novel Pharmacological Interventions for Alcoholism. Berlin: Springer Verlag
Walczak DD, Apter JT, Halikas JA, Borison RL, Carman JS, Post GL, Patrick R, et al. The oral dose-effect relationship for fluvoxamine: a fixed-dose comparison against placebo in depressed outpatients. Annals of Clinical Psychiatry 1996; 8: 139-151
White K, Kando J, Park T, et al (1992) Side effects and the “blindability” of clinical trials. American Journal of Psychiatry 149: 1730-1761