Prevention & Treatment, Volume 5, Article 25, posted July 15, 2002
Copyright 2002 by the American Psychological Association

Commentary on The Emperor's New Drugs: An Analysis of Antidepressant Medication Data Submitted to the U.S. Food and Drug Administration

Antidepressants: A Triumph of Marketing Over Science?

David O. Antonuccio
Veterans Affairs Sierra Nevada Health Care System and University of Nevada School of Medicine

David D. Burns
Stanford University

William G. Danton
Veterans Affairs Sierra Nevada Health Care System and University of Nevada School of Medicine

In a meta-analysis of the Food and Drug Administration (FDA) database of controlled trials used in the initial approval for the most popular antidepressants, I. Kirsch, T. J. Moore, A. Scoboria, and S. S. Nicholls (2002) found that antidepressants demonstrated a clinically negligible advantage over inert placebo. These results are surprising, because they come from studies underwritten by the drug manufacturers. This analysis probably overestimates the antidepressant effect because placebo washout strategies, penetration of the blind, reliance on clinician ratings, use of sedative medication, and replacement of nonresponders may penalize the placebo condition or boost the drug condition. These findings do not appear to justify the popularity of antidepressants, which may have been fueled in part by publication bias and outstanding marketing. Psychotherapy may offer an effective alternative with fewer medical risks.

David O. Antonuccio, Veterans Affairs Sierra Nevada Health Care System, Reno, Nevada, and Department of Psychiatry and Behavioral Sciences, University of Nevada School of Medicine; David D. Burns, Department of Psychiatry, Stanford University School of Medicine; William G. Danton, Veterans Affairs Sierra Nevada Health Care System, Reno, Nevada, and Department of Psychiatry and Behavioral Sciences, University of Nevada School of Medicine.

Correspondence concerning this article should be addressed to David O. Antonuccio, University of Nevada School of Medicine, 401 West 2nd Street, Suite 216, Reno, Nevada 89503.

Apparently antidepressants work, but just barely better than inert placebos. Kirsch, Moore, Scoboria, and Nicholls (2002) have gone straight to the heart of the Food and Drug Administration (FDA) database of blinded, randomized, placebo-controlled trials used in the initial approval of the most popular antidepressant medications, including fluoxetine, paroxetine, sertraline, venlafaxine, nefazadone, and citalopram. The results of their analysis are stunning and offer the potential for a paradigm shift in the way we view the efficacy of antidepressant medications. They analyzed 38 studies involving 6,944 patients treated for an average of 6 weeks and found that approximately 80% of the antidepressant response was duplicated by placebo. In the subsample of fluoxetine trials, the placebo response duplicated 89% of the fluoxetine response.

The mean differences between the drug and placebo conditions averaged only 2 points on the Hamilton Rating Scale for Depression (HAM-D; Hamilton, 1960). For a typical depressed patient from the FDA sample, with an average pretreatment HAM-D score of about 25, a 2-point difference is not clinically significant (Jacobson, Roberts, Berns, & McGlinchey, 1999). Kirsch et al. (2002) also found that higher doses of medication did not result in more improvement, raising the question about whether the small differences between drug and placebo reflect a "true" antidepressant effect.

The FDA Database: Science's Trojan Horse

It is hard to think of a database that would offer a more fair opportunity to evaluate antidepressants. These studies were underwritten by the pharmaceutical manufacturers themselves, under conditions most favorable to the active drug condition. They all used a placebo washout procedure prior to randomization (Antonuccio, Danton, DeNelsky, Greenberg, & Gordon, 1999), potentially eliminating both antidepressant nonresponders (i.e., patients already on an antidepressant before the study starts who get better when they are taken off of an antidepressant) and placebo responders (i.e., patients who are not on antidepressants before the study who respond to placebo). The placebo washout procedure may also tend to retain patients who exhibit withdrawal symptoms from a prior antidepressant. Imagine the converse: an antidepressant washout procedure that eliminates all of the antidepressant responders before a study begins. Such a procedure would surely be considered biased (Antonuccio, Burns, Danton, & O'Donohue, 2000).

The FDA studies incorporate another bias as well. The double blind in these studies is likely to be unintentionally penetrated because of the pattern of side effects in the active and inactive drug conditions (Greenberg & Fisher, 1997). Research clinicians routinely educate patients about potential side effects as part of the standard informed consent process. Further, these studies rely on measures by clinicians who often have a major allegiance or stake in the outcome. Efforts to ensure the integrity of the blind tend to diminish drug efficacy. For example, a recent review of the Cochrane database of antidepressant studies using "active" placebos (making side-effect differences more difficult to detect) found very small or nonsignificant outcome differences, suggesting that trials using inert placebos may overestimate drug effects (Moncrieff, Wessely, & Hardy, 2001).

Kirsch et al. (2002) noted that at least six studies allowed replacement of nonresponders. This may also inflate estimates of the drug response if it is subtly related to side effects and limited primarily to the subjects in the active drug condition (Moncrieff, 2001). A separate analysis excluding these studies might be illuminating. Also, most studies allowed the prescription of a sedative, consistent with findings from another recent meta-analysis of antidepressant trials (Walsh, Seidman, Sysko, & Gould, 2002). If patients in the drug condition were more likely to take sedatives or antidepressants with sedative properties, results could be distorted because there are at least 6 points on the HAM-D that favor medications with sedative properties (Moncrieff, 2001).

Kirsch et al. (2002) noted that the overall active drug effects from the paroxetine, sertraline, and citalopram trials may have been inflated because means were not reported from several studies that found nonsignificant differences. In these trials with incomplete means, the placebo response duplicated less of the antidepressant response than in studies with complete data (72% vs. 82%), suggesting the presence of bias due to selective reporting. Perhaps the FDA could ask the pharmaceutical companies to resubmit the original data in order to have the most complete and accurate database possible.

Given that patients tend to report smaller differences than clinicians (Moncrieff, 2001), one is left to wonder how the pattern of results would look using self-report measures like the Beck Depression Inventory (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). Another question that remains unanswered has to do with what happens to these patients over longer periods, especially after the medications are withdrawn.

In spite of the research design flaws that may favor the drug condition, there is a huge advantage to the FDA database from a scientific perspective. The database includes all of the data from initial trials, published or not, and therefore it is not subject to the usual "file drawer" and publication biases. It would be interesting to know how many of these studies actually were published, to get an estimate of the publication bias in this literature. An indirect estimate of publication bias is possible by looking at the percentage of studies that resulted in significant differences. Antidepressants are significantly more effective than inert placebos in about two thirds of published studies (Thase, 1999). In the FDA database, Kirsch et al. (2002) found the study medication had a significant advantage over inert placebo less than half the time (in 20 of 46 trials), similar to findings from other independent analyses of the same database (Khan, Khan, & Brown, in press; Laughren, 2001). Such a pattern would be consistent with a failure to publish about 23% of antidepressant trials, presumably those showing no advantage to the antidepressant, slightly more than has been estimated in the fluoxetine literature (Gram, 1994). This is a serious problem, because clinical practice guidelines based on the published literature fail to take this distortion of the literature into account (Gilbody & Song, 2000).

Klein (2000) and Quitkin (1999) have argued that because antidepressants have been established as effective in the treatment of depression, trials that do not find a statistical advantage of antidepressants over placebo lack "assay sensitivity" (i.e., the ability to detect specific treatment effects). In other words, they argue that something is wrong with the sampling or methodology of such trials, and the results should be discounted or discarded. If that logic were applied to this meta-analysis, more than half of the studies would have to be discarded, a strategy that would seriously distort the overall results. This example graphically illustrates the hazards of using established "fact" as a reason to throw out newer contradictory data (see Otto & Nierenberg, in press).

Some might suggest that comparing antidepressants and placebos after only 6 weeks of treatment is unfair to the active drug condition and that longer-term outcome will surely favor the medication. However, data from the National Institute of Mental Health Collaborative Depression Study suggest otherwise. The 18-month follow-up data from that multisite trial (Shea et al., 1992) found patients assigned to placebo (plus clinical management) had intent-to-treat outcome comparable to that in the active drug condition (plus clinical management). Other research suggests that longer treatment does not lower relapse following drug withdrawal (Viguera, Baldessarini, & Friedberg, 1998). Even with maintenance antidepressant treatment, the published literature suggests that up to 33% of remitted patients experience a return of depressive symptoms (Byrne & Rothschild, 1998).

It is important to note that the results of Kirsch et al. (2002) are not entirely new or isolated, even though they may surprise many readers. The modest advantage over inert placebo replicates results from many other studies (e.g., Bech et al., 2000; Khan et al., in press; Khan, Warner, & Brown, 2000; Kirsch & Sapirstein, 1998; Laughren, 2001; Moncrieff et al., 2001; Preskorn, 1997; Thase, Entsuah, & Rudolph, 2001). The lack of a dose-response relationship has also come up repeatedly (e.g., Amsterdam et al., 1997; Freemantle, Anderson, & Young, 2000; Preskorn, 1997). Until there are data to support it, increasing an unresponsive patient's dose of an antidepressant would not seem justified. The 63% completion rate in the drug conditions from the FDA database is consistent with an intent-to-treat response rate of about 42% (assuming 66% of completers are also classified as responders), far below expectations but quite typical of antidepressant trials (Thase, 1999). On the basis of the usual inclusionary and exclusionary criteria, only about 15% of patients in a typical outpatient primary care clinic would even qualify to participate in most antidepressant trials (Zimmerman, Mattia, & Posternak, 2002), suggesting that the real world outcome may be even lower. However, despite lack of evidence (Shean, 2001), beliefs about correcting "chemical imbalances" and more powerful antidepressant effects are likely to persist (Antonuccio et al., 2000; Double, 2002; Muņoz, 2000).

In Search of "True" Antidepressant Effects

Kirsch et al. (2002) have proposed the balanced placebo design as a way to untangle the "true" influence of the drug and placebo. This 2 × 2 factorial design manipulates what drug patients are told they are taking (antidepressant or placebo) with what drug they are actually taking (antidepressant or placebo). As an example, this design proved useful in a smoking cessation study where it was shown that the actual drug condition had less impact on withdrawal symptoms and relapse than the believed drug condition (Gottlieb, Killen, Marlatt, & Taylor, 1987).

However, the blind is still subject to penetration in the balanced placebo design. In other words, patients and clinicians may still see through the deception. A more practical solution that doesn't involve as much deception and does not require designs more complicated than those currently in use may be to simply ask clinicians and patients to guess the actual condition (Hughes & Krahn, 1985). Then the outcome impact of blind penetration can be estimated by comparing those patients who are truly blind with those who are not. This strategy can simulate a balanced placebo design and can be implemented right now in ongoing studies, with very little additional cost. At the very least, such a procedure would help settle the debate about the extent of blind penetration and its impact. To address concerns that outcome may tip off the condition, clinicians and patients could make their guesses early in the study before treatment effects take hold (e.g., within the first 2 weeks). It has been demonstrated that independent raters were able to fairly accurately guess treatment conditions in a placebo-controlled antidepressant study without any outcome data and while relying only on side-effect profiles (White, Kando, Park, Waternaux, & Brown, 1992). It is our contention that no study should be able to claim double blind status and pass peer review without testing the integrity of the blind (Antonuccio et al., 2000; Piasecki, Antonuccio, Steinagel, & Kohlenberg, in press). Statistical strategies for estimating clinician rater bias (Petkova, Quitkin, McGrath, Stewart, & Klein, 2000) are laudable but not a substitute for an actual test of the blind.

There are some important questions that this database won't help answer. How is it that, with such a weak advantage over placebo, antidepressants have apparently become the most popular treatment for depression in the United States (Langer, 2000; Olfson et al., 2002) and in many parts of the world? Does the small advantage of antidepressants over placebo justify the risks and side effects associated with these medications? How have we come to think of antidepressants as powerful, even "life-saving" treatments in the face of such weak outcome data (see Beutler, 1998)?

Part of the answer to the last question probably lies in widely acknowledged publication biases (e.g., Blumenthal, Campbell, Anderson, Causino, & Louis, 1997; Callaham, Wears, Weber, Barton, & Young, 1998; Chalmers, 2000; Gilbody & Song, 2000; Lancet, 2001; Misakian & Bero, 1998; Rennie, 1999; Wise & Drury, 1996), often related to conflicts of interest (Campbell, Louis, & Blumenthal, 1998; Cech & Leonard, 2001; DeAngelis, Fontanarosa, & Flanagin, 2001; Fava, 2001; Lo, Wolf, & Berkeley, 2000) that favor pharmaceutical industry products. In fact, these biases have so eroded the credibility of the medical literature (Quick, 2001) that new proposals call for stringent accountability guidelines (e.g., Davidoff et al., 2001; Moses & Martin, 2001) aimed at ensuring researcher independence in study design, access to data, and right to publish. It remains to be seen whether these new guidelines will have the desired effect of improving the quality and the credibility of the literature. Unfortunately, the New England Journal of Medicine has recently had to relax its strict policy against financial conflicts of interest by editorial and review authors (Drazen & Curfman, 2002), because it could not find enough experts without financial ties to the drug industry, an issue that has been highlighted in the antidepressant literature (Angell, 2000).

Yet, publication biases that may inflate the perceived efficacy of antidepressants cannot fully explain the popularity of some of the newer medications (Gram, 1994), especially given the reduced cost and equivalent efficacy of older antidepressants (Geddes, Freemantle, Mason, Eccles, & Boynton, 2001). A second factor in the popularity of antidepressant medications generally, and selective serotonin reuptake inhibitors (SSRIs) in particular, undoubtedly has to do with extremely effective marketing practices, often framed as education.

Continuing Medical Advertising

The pharmaceutical industry is at least a $250 billion annual business worldwide, with the United States accounting for about one third of all pharmaceutical sales (Louie, 2001). It is America's most profitable industry, number one in return on revenues, return on assets, and return on equity (Fortune, 2000). According to IMS Health, a pharmaceutical consulting firm, the industry spent more than $19 billion in 2001 in U.S. advertising alone. The pharmaceutical industry had a combined lobbying and campaign contribution budget of $197 million in 1999 and 2000, larger than any other industry (Wayne & Peterson, 2001). The industry has 625 registered lobbyists, more than there are members of congress (Wayne & Peterson, 2001). Industry underwrites about 70% of all clinical drug trials in the United States (DeAngelis et al., 2001).

The pharmaceutical industry does an outstanding job of marketing its products, and antidepressants are no exception. In 1999, 3 of the top 10 best-selling pharmaceuticals were the SSRIs Prozac, Paxil, and Zoloft, accounting for combined revenues of $6.7 billion (Louie, 2001). Revenues generated by SSRIs are growing by about 25% each year (Kroenke et al., 2001). Recent polling data suggested that as many as 1 in 8 adult Americans had taken an antidepressant in the past 10 years, and 3.5 billion doses of SSRIs were consumed in 1999 alone (Langer, 2000). About 60% of those who had taken antidepressants indicated they had taken them for more than 3 months, whereas 46% indicated they had been taking them for a year or more.

Consumer Reports has documented marketing strategies (Miracle or Media Drugs, 1992; Pushing Drugs to Doctors, 1992) that include but are not limited to the following: (a) giving free samples and information to doctors, (b) advertising in medical journals, (c) using "ask your doctor" media advertisements aimed directly at the consumer (see Wolfe, 2002), (d) sponsoring promotional dinner meetings with substantial gifts or even cash provided for attendees, (e) paying consultants to speak at scientific meetings where it is possible to circumvent FDA guidelines that require disclosure of side effects, (f) funding only those research projects that have a high likelihood of producing favorable results for a particular drug company's product (see Bodenheimer, 2000), (g) terminating negative studies before they are ready for publication, (h) involving large numbers of physicians in studies not intended to yield publishable information but simply designed to yield maximum product exposure, (i) including "look alike" publication supplements in professional journals, (j) offering to pay journalists to cover their products, (k) offering prepackaged information for journalists in the form of video news releases that give the appearance of having been independently developed, and (l) helping to fund patient advocacy and other public interest groups so the consumer group appears to be publicly carrying the banner of a particular drug.

It is difficult to think of any arena involving information about medications that does not have significant industry financial or marketing influences. Industry financial influences extend to federal regulatory agencies (e.g., Cimons, 1999; Horton, 2001), professional organizations and their journals (Pellegrino & Relman, 1999), continuing medical education (Christensen & Tueth, 1998; Coyle, 2002; Romano, 2001; Ross, Lurie, Wolfe, & Public Citizen's Health Research Group, 2000; Wazana, 2000), scientific researchers (Angell, 2000; Bodenheimer, 2000; Brennan, 1994; Choudry, Stelfox, & Detsky, 2002; Drazen & Curfman, 2002; Elliot, 2001; Flanagin et al., 1998; Krimsky & Rothenberg, 2001; Krimsky, Rothenberg, Stott, & Kyle, 1998), media experts (Moynihan et al., 2000; Steinbrook, 2000; Woloshin & Schwartz, 2002), and consumer advocacy organizations (e.g., Silverstein, 1999). Respected psychiatric researchers like Marks (Marks et al., 1993) and Fava (1998) have warned that such widespread corporate interests may result in self-selecting academic oligarchies influencing clinical and scientific information. In fact, those who produce data contrary to industry interests may find themselves vulnerable to legal, professional, or even personal attack, directly or indirectly financed by the industry (e.g., Boseley, 2002; Deyo, Psaty, Simon, Wagner, & Omenn, 1997; Healy, 2002; Marks et al., 1993; Monbiot, 2002; Rennie, 1997).

It is entirely reasonable for the pharmaceutical industry to market its products as effectively as possible. The question can be raised, however, about whether effective marketing sometimes clashes with good science and with the public interest. Company-sponsored experts, whether they are researchers or educators, are by definition company employees. They will be retained only if they offer consistently favorable treatment to the company's products. It could be argued that their efforts on behalf of antidepressants often fit more properly under the rubric of marketing or advertising, not science or education.

Psychotherapy for Depression: No Stronger Medicine

For clinicians who do not wish to expose their patients to the health risks and side effects associated with antidepressants, what is the alternative? Placebo treatment faces an inherent paradox. For treatment to be effective, the patient should believe that the treatment will remedy the problem. Kirsch et al. (2002) pointed out that the deception inherent in the administration of true placebos inhibits their use. Of course, this problem is obviated when the clinician and the patient both believe that the medication is effective. Perhaps this practical self-deception explains why so many otherwise well-informed clinicians persist in defending the power of antidepressant efficacy. Until someone demonstrates a truly powerful and safe antidepressant medication, or finds a way to deliver a nondeceptive placebo, psychotherapy is a viable alternative.

Some might suggest that because there appears to be a small advantage over placebo, antidepressants should still be considered a first line of treatment for depression. However, there may be several reasons why nondrug alternatives ought to be considered first. First, most of the drug only conditions in studies of antidepressants use supportive weekly visits adjunctively (e.g., Elkin et al., 1989) and do not adequately evaluate the efficacy of medication alone. The resultant therapeutic alliance (e.g., Blatt, Sanislow, Zuroff, Pilkonis, 1996) may enhance the efficacy of the medication and give an inaccurate picture of the effectiveness of these medications in a managed care environment where antidepressants are often delivered in conjunction with infrequent visits to a mental health professional.

Another reason to consider psychotherapy as a preferred first line of treatment is that psychotherapy (particularly cognitive therapy, behavioral activation, and interpersonal therapy) compares favorably with medications in the short term, even when the depression is severe (e.g., DeRubeis, Gelfand, Tang, & Simons, 1999), and appears superior to medications when long term follow-up is considered (Antonuccio, Danton, & DeNelsky, 1995; Hollon, Shelton, & Loosen, 1991). In accepting that psychotherapy has efficacy comparable with that of antidepressants, it is not necessary to give up the biological model completely, because there is evidence of metabolic changes associated with improvement in psychotherapy that are similar to those seen with antidepressants (Brody et al., 2001; Martin, Martin, Rai, Richardson, & Royall, 2001). In fact, distinct metabolic (Mayberg et al., 2002) and electroencephalogram changes (Leuchter, Cook, Witte, Morgan, & Abrams, 2002) have also been documented in patients who respond to placebo.

Whether one subscribes to the Hippocratic dictum "first do no harm" or takes a cost-benefit approach to treatment, it is impossible to ignore the fact that antidepressants are not medically benign treatments. The short-term side effects of SSRIs are well-established and commonly include agitation, sleep disruption, gastrointestinal problems, and sexual dysfunction (Antonuccio et al., 1999). Side effects and medical risks (including risk of dying) increase when SSRIs are combined with other medications (Dalfen & Stewart, 2001), as is often the case (Antonuccio et al., 1999). In addition, the withdrawal symptoms of SSRIs are substantial for many, if not most, patients (Coupland, Bell, & Potokar, 1996; Fava, 2002; Rosenbaum, Fava, Hood, Ashcroft, & Krebs, 1998). Although the data are mixed and somewhat controversial, other potential risks that warrant further investigation include the association (although not an established causal relationship) of SSRIs with breast cancer (e.g., Cotterchio, Kreiger, Darlington, & Steingart, 2000), acts of deliberate self-harm (e.g., Donovan et al., 2000; Healy, 2002), manic episodes (e.g., Preda, MacLean, Mazure, & Bowers., 2001), and the possibility of irreversible biochemical changes predisposing some susceptible patients to chronic depression (e.g., Baldessarini, 1995; Fava, 1995, 2002). When medical cost offset (Hunsley, in press), relapse, and side effects are considered in a cost-benefit analysis, psychotherapy can be very cost-effective, particularly in a psychoeducational (e.g., therapist-assisted bibliotherapy) or group format (Antonuccio, Thomas, & Danton, 1997).

Some might argue that antidepressants are important to stave off suicide in very depressed patients. However, there is no evidence that antidepressants reduce the risk of suicide or suicide attempts in comparison with a placebo in randomized controlled trials (Khan et al., 2000; Storosum, van Zwieten, van den Brink, Gersons, & Broekmans, 2001). Finally, most patients seem to prefer a psychotherapeutic intervention when given the choice (Chilvers et al., 2001; Hall & Robertson, 1998; Jorm, 2000; Paykel, Hart, & Priest, 1998; Priest, Vize, Roberts, Roberts, & Tylee, 1996). Despite this repeated finding, it is often concluded that consumers just need more "education" about the benefits and safety of antidepressants to bring their preferences in line with accepted practice (e.g., Paykel et al., 1998; Priest et al., 1996).

A Bitter Pill to Swallow

Kirsch et al. (2002) have convincingly demonstrated, using pharmaceutical industry data, that the image of antidepressants as powerfully more effective than placebo is not supported by the data. The small advantage over inert placebo credited to antidepressants is quite possibly a methodological artifact (Moncrieff, 2002). It could be argued that the patients randomly assigned to placebo are the lucky ones, because they derive a benefit virtually comparable with the medication condition without the associated medical risks.

The silver lining in these results for psychiatry is that the psychiatrist, or at least something about the psychiatric relationship, and not the pill, appears to facilitate improvement in depression. Figuring out ways to enhance the therapeutic alliance, originally pioneered but recently marginalized by organized psychiatry, may prove more fruitful than modifying the selectivity of antidepressants. There is no doubt that antidepressants have a biochemical impact on the brain, but the valence of that impact is open for considerable debate, and whether it corrects a chemical imbalance is in grave doubt. It may take another decade, but when the logical arguments and the advertising blitz fail to discredit the FDA data and when we begin to fully appreciate the cost of side effects and long-term health risks associated with antidepressants, American medicine's preference for these medications as the primary treatment for depression may change. Other popular medical practices have certainly proven ineffective or too risky over time. One day we may look back and marvel at the stroke of marketing genius that led to calling these medications antidepressants in the first place. Kirsch et al. (2002) have demonstrated that just because a pill is called an antidepressant, it doesn't necessarily make it so.


Amsterdam, J. D., Fawcett, J., Quitkin, F. M., Reimherr, F. W., Rosenbaum, J. F., Michelson, D., et al. (1997). Fluoxetine and norfluoxetine plasma concentrations in major depression: A multicenter study. American Journal of Psychiatry, 154, 963-969.

Angell, M. (2000). Is academic medicine for sale? New England Journal of Medicine, 342, 1516-1518.

Antonuccio, D. O., Burns, D., Danton, W. G., & O'Donohue, W. (2000). The rumble in Reno: The psychosocial perspective on depression. Psychiatric Times, 17, 24-28. Available on the World Wide Web:

Antonuccio, D. O., Danton, W. G., & DeNelsky, G. (1995). Psychotherapy vs. medication for depression: Challenging the conventional wisdom with data. Professional Psychology: Research and Practice, 26, 574-585.

Antonuccio, D. O., Danton, W. G., DeNelsky, G. Y., Greenberg, R. P., & Gordon, J. S. (1999). Raising questions about antidepressants. Psychotherapy and Psychosomatics, 68, 3-14.

Antonuccio, D. O., Thomas, M., & Danton, W. G. (1997). A cost-effectiveness analysis of cognitive behavior therapy and fluoxetine (Prozac) in the treatment of depression. Behavior Therapy, 28, 187-210.

Baldessarini, R. J. (1995). Risks and implications of interrupting maintenance psychotropic drug therapy. Psychotherapy and Psychosomatics, 63, 137-141.

Bech, P., Cialdella, P., Haugh, M. C., Hours, A., Boissel, J. P., Birkett, M. A., & Tollefson, G. D. (2000). Meta-analysis of randomized controlled trials of fluoxetine v. placebo and tricyclic antidepressants in the short-term treatment of major depression. British Journal of Psychiatry, 176, 421-428.

Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561-571.

Beutler, L. E. (1998). Prozac and placebo: There's a pony in there somewhere. Prevention & Treatment, 1, Article 3. Available on the World Wide Web:

Blatt, S. J., Sanislow, C. A., Zuroff, D. C., & Pilkonis, P. A. (1996). Characteristics of effective therapists: Further analyses of data from the National Institute of Mental Health Treatment of Depression Collaborative Research Program. Journal of Consulting and Clinical Psychology, 64, 1276-1284.

Blumenthal, D., Campbell, E. G., Anderson, M. S., Causino, N., & Louis, K. S. (1997). Withholding research results in academic life science: Evidence from a national survey of faculty. Journal of the American Medical Association, 277, 1224-1228.

Bodenheimer, T. (2000). Uneasy alliance-Clinical investigators and the pharmaceutical industry. New England Journal of Medicine, 342, 1539-1544.

Boseley, S. (2002). Bitter pill. Guardian. Retrieved May 21, 2002, from,5500,718916,00.html

Brennan, T. A. (1994). Buying editorials. New England Journal of Medicine, 331, 673-675.

Brody, A. L., Saxena, S., Stoessel, P., Gillies, L. A., Fairbanks, L. A., Alborzian, S., et al. (2001). Regional brain metabolic changes in patients with major depression treated with either paroxetine or interpersonal therapy: Preliminary findings. Archives of General Psychiatry, 58, 631-640.

Byrne, S. E., & Rothschild, A. J. (1998). Loss of antidepressant efficacy during maintenance therapy: Possible mechanisms and treatments. Journal of Clinical Psychiatry, 59, 279-288.

Callaham, M. L., Wears, R. L., Weber, E. J., Barton, C., & Young, G. (1998). Positive-outcome bias and other limitations in the outcome of research abstracts submitted to a scientific meeting. Journal of the American Medical Association, 280, 254-257.

Campbell, E. G., Louis, K. S., & Blumenthal, D. (1998). Looking a gift horse in the mouth: Corporate gifts supporting life sciences research. Journal of the American Medical Association, 279, 995-999.

Cech, T. R., & Leonard, J. S. (2001, February 9). Science and business. Conflict of interest-Moving beyond disclosure. Science, 291, 989-990.

Chalmers, I. (2000). Current controlled trials: An opportunity to help improve the quality of clinical research. Current Controlled Trials in Cardiovascular Medicine, 1, 3-8.

Chilvers, C., Dewey, M., Fielding, K., Gretton, V., Miller, P., Palmer, B., et al. (2001). Antidepressant drugs and generic counseling for treatment of major depression in primary care: Randomised trial with patient preference arms. British Medical Journal, 322, 772-775.

Choudhry, N. K., Stelfox, H. T., & Detsky, A. S. (2002). Relationships between authors of clinical practice guidelines and the pharmaceutical industry. Journal of the American Medical Association, 287, 612-617.

Christensen, R. C., & Tueth, M. J. (1998). Pharmaceutical companies and academic departments of psychiatry: A call for ethics education. Academic Psychiatry, 22, 135-137.

Cimons, M. (1999). NIH opens conflict-of-interest investigation. Nature Medicine, 5, 129-130.

Cotterchio, M., Kreiger, N., Darlington, G., & Steingart, A. (2000). Antidepressant medication use and breast cancer risk. American Journal of Epidemiology, 151, 951-957.

Coupland, N. J., Bell, C. J., & Potokar, J. P. (1996). Serotonin reuptake inhibitor withdrawal. Journal of Clinical Psychopharmacology, 16, 356-362.

Coyle, S. (2002). Physician-industry relations. Annals of Internal Medicine, 136, 396-406.

Dalfen, A. K., & Stewart, D. E. (2001). Who develops stable or fatal adverse drug reactions to selective serotonin reuptake inhibitors? Canadian Journal of Psychiatry, 46, 258-262.

Davidoff, F., DeAngelis, C. D., Drazen, J. M., Nicholls, M. G., Hoey, J., Hojgaard, L., et al. (2001). Sponsorship, authorship, and accountability. New England Journal of Medicine, 345, 825-827.

DeAngelis, C. D., Fontanarosa, P. B., & Flanagin, A. (2001). Reporting financial conflicts of interest and relationships between investigators and research sponsors. Journal of the American Medical Association, 286, 89-91.

DeRubeis, R. J., Gelfand, L. A., Tang, T. Z., & Simons, A. D. (1999). Medications versus cognitive behavior therapy for severely depressed outpatients: Mega-analysis of four randomized comparisons. American Journal of Psychiatry, 156, 1007-1013.

Deyo, R. A., Psaty, B. M., Simon, G., Wagner, E. H., & Omenn, G. S. (1997). The messenger under attack-Intimidation of researchers by special-interest groups. New England Journal of Medicine, 336, 1176-1179.

Donovan, S., Madeley, R., Clayton, A., Beeharry, M., Jones, S., Kirk, C., et al. (2000). Deliberate self-harm and antidepressant drugs: Investigation of a possible link. British Journal of Psychiatry, 177, 551-556.

Double, D. (2002). The limits of psychiatry. British Medical Journal, 324, 900-904.

Drazen, J. M., & Curfman, G. D. (2002). Financial associations of authors. New England Journal of Medicine, 346, 1901-1902.

Elkin, I., Shea, T., Watkins, J. T., Imber, S. D., Sotsky, S. M., Collins, J. F., et al. (1989). National Institute of Mental Health Treatment of Depression Collaborative Research Program: General effectiveness of treatments. Archives of General Psychiatry, 46, 971-982.

Elliot, C. (2001). Pharma buys a conscience. American Prospect, 12, 1-9.

Fava, G. A. (1995). Holding on: Depression, sensitization by antidepressant drugs, and the prodigal experts. Psychotherapy and Psychosomatics, 64, 57-61.

Fava, G. A. (1998). All our dreams are sold. Psychotherapy and Psychosomatics, 67, 191-193.

Fava, G. A. (2001). Conflict of interest in special interest groups: The making of a counter culture. Psychotherapy and Psychosomatics, 70, 1-5.

Fava, G. A. (2002). Long-term treatment with antidepressant drugs: The spectacular achievements of propaganda. Psychotherapy and Psychosomatics, 71, 127-132.

Flanagin, A., Carey, L. A., Fontanarosa, P. B., Phillips, S. G., Pace, B. P., Lundberg, G. D., & Rennie, D. (1998). Prevalence of articles with honorary authors and ghost authors in peer-reviewed medical journals. Journal of the American Medical Association, 280, 222-224.

Fortune. (2000). How the industries stack up. Retrieved April 17, 2000, from

Freemantle, N., Anderson, I. M., & Young, P. (2000). Predictive value of pharmacological activity for the relative efficacy of antidepressant drugs: Meta-regression analysis. British Journal of Psychiatry, 177, 292-302.

Geddes, J. R., Freemantle, N., Mason, J., Eccles, M. P., & Boynton, J. (2001). Selective serotonin reuptake inhibitors (SSRIs) for depression [Cochrane Review]. Available on the World Wide Web:

Gilbody, S. M., & Song, F. (2000). Publication bias and the integrity of psychiatry research. Psychological Medicine, 30, 253-258.

Gottlieb, A. M., Killen, J. D., Marlatt, G. A., & Taylor, C. B. (1987). Psychological and pharmacological influences in cigarette smoking withdrawal: Effects of nicotine gum and expectancy on smoking withdrawal symptoms and relapse. Journal of Consulting and Clinical Psychology, 55, 606-608.

Gram, L. F. (1994). Fluoxetine. New England Journal of Medicine, 331, 1354-1361.

Greenberg, R. P., & Fisher, S. (1997). Mood-mending medicines: Probing drug, psychotherapy, and placebo solutions. In S. Fisher & R. P. Greenberg (Eds.), From placebo to panacea: Putting psychiatric drugs to the test (pp. 115-172). New York: Wiley.

Hall, L. H., & Roberston, M. H. (1998). Undergraduate ratings of the acceptability of single and combined treatments for depression: A comparative analysis. Professional Psychology: Research and Practice, 29, 269-272.

Hamilton, M. (1960). A rating scale for depression. Journal of Neurology, Neurosurgery, and Psychiatry, 12, 56-62.

Healy, D. (2002). Conflicting interests in Toronto: Anatomy of a controversy at the interface of academia and industry. Perspectives in Biology and Medicine, 45, 250-263.

Hollon, S. D., Shelton, R. C., & Loosen, P. T. (1991). Cognitive therapy and pharmacotherapy for depression. Journal of Consulting and Clinical Psychology, 59, 88-99.

Horton, R. (2001). Lotronex and the FDA: A fatal erosion of integrity. Lancet, 357,1544-1545.

Hughes, J. R., & Krahn, D. (1985). Blindness and the validity of the double-blind procedure. Journal of Clinical Psychopharmacology, 5, 138-142.

Hunsley, J. (in press). Cost-effectiveness and medical cost offset considerations in psychological service provision. Canadian Psychology.

Jacobson, N. S., Roberts, L. J., Berns, S. B., & McGlinchey, J. B. (1999). Methods for defining and determining the clinical significance of treatment effects: Description, application, and alternatives. Journal of Consulting and Clinical Psychology, 67, 300-307.

Jorm, A. F. (2000). Mental health literacy: Public knowledge and beliefs about mental disorders. British Journal of Psychiatry, 177, 396-401.

Khan, A., Khan, S., & Brown, W. A. (in press). Are placebo controls necessary to test new antidepressants and anxiolytics? International Journal of Neuropsychopharmacology.

Khan, A., Warner, H. A., & Brown, W. A. (2000). Symptom reduction and suicide risk in patients treated with placebo in antidepressant clinical trials. Archives of General Psychiatry, 57, 311-330.

Kirsch, I., Moore, T.J., Scoboria, A., & Nicholls, S.S. (2002). The emporer's new drugs: An analysis of antidepressant medication data submitted to the U.S. Food and Drug Administration. Prevention & Treatment, 5, Article 23. Available on the World Wide Web:

Kirsch, I., & Sapirstein, G. (1998). Listening to Prozac, but hearing placebo? A meta-analysis of antidepressant medication. Prevention & Treatment, 1, Article 2. Available on the World Wide Web:

Klein, D. F. (2000). Flawed meta-analyses comparing psychotherapy with pharmacotherapy. American Journal of Psychiatry, 157, 1204-1211.

Krimsky, S., & Rothenberg, L. S. (2001). Conflict of interest policies in science and medical journals: Editorial practices and author disclosures. Science and Engineering Ethics, 7, 205-218.

Krimsky, S., Rothenberg, L. S., Stott, P., & Kyle, G. (1998). Scientific journals and their authors' financial interests: A pilot study. Psychotherapy and Psychosomatics, 67, 194-201.

Kroenke, K., West, S. L., Swindle, R., Gilsenan, A., Eckert, G. J., Dolor, R., et al. (2001). Similar effectiveness of paroxetine, fluoxetine, and sertraline in primary care. Journal of the American Medical Association, 286, 2947-2955.

Lancet. (2001). The tightening grip of big pharma [Editorial]. The Lancet, 357, 1141.

Langer, G. (2000, April 10). Use of antidepressants is a long-term practice. Retrieved December 2001, from

Laughren, T. P. (2001). The scientific and ethical basis for placebo-controlled trials in depression and schizophrenia: An FDA perspective. European Psychiatry, 16, 418-423.

Leuchter, A. F., Cook, I. A., Witte, E. A., Morgan, M., & Abrams, M. (2002). Change in brain function of depressed subjects during treatment with placebo. American Journal of Psychiatry, 159, 122-129.

Lo, B., Wolf, L. E., & Berkeley, A. (2000). Conflict-of-interest policies for investigators in clinical trials. New England Journal of Medicine, 343, 1643-1645.

Louie, L. (2001). A prescription for profits. Upside. Available on the World Wide Web:

Marks, I. M., Swinson, R. P., Basoglu, M., Kuch, K., Noshirvani, H., O'Sullivan, G., et al. (1993). Alprazolam and exposure alone and combined in panic disorder with agoraphobia: A controlled study in London and Toronto. British Journal of Psychiatry, 162, 776-787.

Martin, S. D., Martin, E., Rai, S. S., Richardson, M. A., & Royall, R. (2001). Brain blood flow changes in depressed patients treated with interpersonal psychotherapy or venlafaxine hydrochloride: Preliminary findings. Archives of General Psychiatry, 58, 641-648.

Mayberg, H. S., Silva, J. A., Brannan, S. K., Tekell, J. L, Mahurin, R. K., McGinnis, S., & Jerabek, P. A. (2002). The functional neuroanatomy of the placebo effect. American Journal of Psychiatry, 159, 728-737.

Miracle drugs or media drugs? (1992, March). Consumer Reports, 142-146.

Misakian, A. L., & Bero, L. A. (1998). Publication bias and research on passive smoking: Comparison of published and unpublished studies. Journal of the American Medical Association, 280, 250-253.

Monbiot, G. (2002). The fake persuaders: Corporations are inventing people to rubbish their opponents on the internet. Guardian. Retrieved May 14, 2002, from,9236,715160,00.html

Moncrieff, J. (2001). Are antidepressants overrated? A review of methodological problems in antidepressant trials. Journal of Nervous and Mental Disorders, 189, 288-295.

Moncrieff, J. (2002). The antidepressant debate. British Journal of Psychiatry, 180, 193-194.

Moncrieff, J., Wessely, S., & Hardy, R. (2001). Antidepressants using active placebos [Cochrane Review]. Cochrane Database Systematic Review, 2, CD003012.

Moses, H., & Martin, J. B. (2001). Academic relationships with industry: A new model for biomedical research. Journal of the American Medical Association, 285, 933-935.

Moynihan, R., Bero, L., Ross-Degnan, D., Henry, D., Lee, K., Watkins, J., et al. (2000). Coverage by the news media of the benefits and risks of medications. New England Journal of Medicine, 342, 1645-1650.

Muņoz, R. A. (2000, August). Rumble in Reno: Psychopharmacologic Nihilism. Psychiatric Times, 17(8). Retrieved August 2000, from

Olfson, M., Marcus, S. C., Druss, B., Elinson, L., Tanielian, T., & Pincus, H. A. (2002). National trends in the outpatient treatment of depression. Journal of the American Medical Association, 287, 203-209.

Otto, M. W., & Nierenberg, A. A. (in press). Assay sensitivity, failed clinical trials, and the conduct of science. Psychotherapy and Psychosomatics.

Paykel, E. S., Hart, D., & Priest, R. G. (1998). Changes in public attitudes to depression during the Defeat Depression Campaign. British Journal of Psychiatry, 173, 519-522.

Pellegrino, E. D., & Relman, A. S. (1999). Professional medical associations: Ethical and practical guidelines. Journal of the American Medical Association, 282, 984-986.

Petkova, E., Quitkin, F. M., McGrath, P. J., Stewart, J. W., & Klein, D. F. (2000). A method to quantify rater bias in antidepressant trials. Neuropsychopharmacology, 22, 559-565.

Piasecki, M., Antonuccio, D. O., Steinagel, G., & Kohlenberg, B. S. (in press). Penetration of the blind in a controlled study of Paxil used to treat cocaine addiction. Journal of Behavior Therapy and Experimental Psychiatry.

Preda, A., MacLean, R. W., Mazure, C. M., & Bowers, M. B. (2001). Antidepressant-associated mania and psychosis resulting in psychiatric admissions. Journal of Clinical Psychiatry, 62, 30-33.

Preskorn, S. H. (1997). Clinically relevant pharmacology of selective serotonin reuptake inhibitors. Clinical Pharmacokinetics, 32 (Suppl. 1), 1-21.

Priest, R. G., Vize, C., Roberts, A., Roberts, M., & Tylee, A. (1996). Lay people's attitude to treatment of depression: Results of opinion poll for Defeat Depression Campaign just before its launch. British Medical Journal, 313, 858-859.

Pushing drugs to doctors. (1992, February). Consumer Reports, 87-94.

Quick, J. (2001, December). Maintaining the integrity of the clinical evidence base (Ref. No. 01-1602). Bulletin of the World Health Organization, 79(12), 1093.

Quitkin, F. M. (1999). Placebos, drug effects, and study design: A clinician's guide. American Journal of Psychiatry, 156, 829-836.

Rennie, D. (1997). Thyroid storm. Journal of the American Medical Association, 277, 1238-1243.

Rennie, D. (1999). Fair conduct and fair reporting of clinical trials. Journal of the American Medical Association, 282, 1766-1768.

Romano, M. (2001). Ethical quandary: Smaller AMA moves slowly on rules for drug firms' influence over prescribing. Modern Healthcare, 31, 4-14.

Rosenbaum, J. F., Fava, M., Hood, S. L., Ashcroft, R. C., & Krebs, W. B. (1998). Selective serotonin reuptake inhibitor discontinuation syndrome: A randomized clinical trial. Biological Psychiatry, 44, 77-87.

Ross, J., Lurie, P., Wolfe, S. M., & Public Citizen's Health Research Group. (2000). Letter to the accreditation council for graduate medical education regarding a public citizen study describing medical education services suppliers (HRG Publication No. 1530). Retrieved July 2000, from

Shea, M. T., Elkin, I., Imber, S. D., Sotsky, S. M., Watkins, J. T., Collins, J. F., et al. (1992). Course of depressive symptoms over follow-up: Findings from the National Institute of Mental Health Treatment of Depression Collaborative Research Program. Archives of General Psychiatry, 49, 782-787.

Shean, G. (2001). A critical look at some assumptions of biopsychiatry. Ethical Human Sciences and Services, 3, 77-96.

Silverstein, K. (1999, November/December). An influential mental health nonprofit finds its "grassroots" watered by pharmaceutical millions. Mother Jones. Retrieved December 1999, from

Steinbrook, R. (2000). Medical journals and medical reporting. New England Journal of Medicine, 342, 1668-1671.

Storosum, J. G., van Zwieten, B. J., van den Brink, W., Gersons, B. P. R., & Broekmans, A. W. (2001). Suicide risk in placebo-controlled studies of major depression. American Journal of Psychiatry, 158, 1271-1275.

Thase, M. E. (1999). How should efficacy be evaluated in randomized clinical trials of treatments for depression? Journal of Clinical Psychiatry, 60(Suppl. 4), 23-31.

Thase, M. E., Entsuah, A. R., & Rudolph, R. L. (2001). Remission rates during treatment with venlafaxine or selective serotonin reuptake inhibitors. British Journal of Psychiatry, 178, 234-241.

Viguera, A. D., Baldessarini, R. J., & Friedberg, J. (1998). Discontinuing antidepressant treatment in major depression. Harvard Review of Psychiatry, 5, 293-306.

Walsh, B.T., Seidman, S.N. Sysko, R., & Gould, M. (2002). Placebo response in studies of major depression: Variable, substantial, and growing. Journal of the American Medical Association, 287, 1840-1847.

Wayne, L., & Peterson, M. (2001, November 4). Drug industry: A muscular lobby tries to shape nation's bioterror plan. New York Times, pp. 3.1. Abstract available on the World Wide Web:

Wazana, A. (2000). Physicians and the pharmaceutical industry: Is a gift ever just a gift? Journal of the American Medical Association, 283, 373-380.

White, K., Kando, J., Park, T., Waternaux, C., & Brown, W. A. (1992). Side effects and the "blindability" of clinical drug trials. American Journal of Psychiatry, 149, 1730-1731.

Wise, P., & Drury, M. (1996). Pharmaceutical trials in general practice: The first 100 protocols. An audit by the clinical research ethics committee of the Royal College of General Practitioners. British Medical Journal, 313, 1245-1248.

Wolfe, S. (2002). Direct-to-consumer advertising-Education or emotion promotion? New England Journal of Medicine, 346, 524-526.

Woloshin, S., & Schwartz, L. M. (2002). Press releases: Translating research into news. Journal of the American Medical Association, 287, 2856-2858.

Zimmerman, M., Mattia, J. I., & Posternak, M. A. (2002). Are subjects in pharmaceutical treatment trials of depression representative of patients in routine clinical practice? American Journal of Psychiatry, 159, 469-473.