Antidepressants 'don't work'

“Prozac, used by 40m people, does not work say scientists” reads the headline in The Guardian today. This newspaper and others report that a study which pooled all available data comparing Prozac and similar antidepressants with inactive “dummy” pills found that placebo was just as effective as the drugs. The study’s authors say that antidepressants proved more effective among severely depressed patients. However, the scientists add that this could be due to a reduction in the effect of placebo rather than because the drugs worked better, according to The Times .

The scientists told The Independent that “given these results, there seems to be little reason to prescribe antidepressant medication to any but the most severely depressed patients, unless alternative treatments have failed".

The researchers pooled all studies submitted up to 1999 to the US Food and Drug Administration (FDA) for the approval of four antidepressant drugs: fluoxetine (Prozac), venlafaxine (Effexor), nefazodone (Serzone) and paroxetine (Seroxat). The antidepressants produced an overall reduction in depression symptoms compared with placebo. However, the authors of this research suggest that these improvements are not clinically meaningful, except in patients with the most severe depression.

The research did not include trials that were carried out after the drugs were approved. Further studies should include these to see whether they get similar results. This study does not show that antidepressants have no effect. However, it does show that the benefits of drugs might vary for people with different levels of symptoms, and any current debate should centre on how severe symptoms should be before antidepressants are used. Doctors already take into account this severity and try non-drug treatments for depression before prescribing antidepressants. However, for people with very severe symptoms that do not respond to other treatments, antidepressants are an important option.

Professor Irving Kirsch, the study’s main author, emphasises in the newspaper reports that patients should not change their treatment without speaking to their doctor. He says that other forms of treatment, including physical exercise, talking therapies and self-help books, could be considered for less severe cases.

Where did the story come from?

Professor Kirsch of the University of Hull and colleagues from universities in the US and Canada, and the Institute for Safe Medication Practices in the US carried out the research. The study received no specific funding and was published in peer-reviewed medical journal PLoS Medicine.

What kind of scientific study was this?

This systematic review and meta-analysis examined the effects of antidepressants on differing severities of depression.

The researchers asked the FDA for data on all of the double-blind randomised controlled trials (RCTs) that compared six antidepressants (fluoxetine, venlafaxine, nefazodone, paroxetine, sertraline and citalopram) with placebo in people with major depressive disorder. The participants were diagnosed according to standard criteria. These trials were submitted to the FDA as part of the drug licensing process and included all drug company sponsored RCTs that had been published prior to the drugs’ approvals, which were granted between 1987 and 1999. Published and unpublished studies were included.

The researchers supplemented the FDA information with data from the drug companies’ websites and PubMed, an electronic literature database. They used PubMed to search for publications from 1985 to May 2007. The researchers also obtained data from RCTs mentioned in the studies, review publications and the Swedish drug regulatory authority.

The researchers excluded studies that did not report on participants who dropped out and those that were conducted at multiple sites but only reported data from one site.

The remaining RCTs were searched for those that looked at changes in symptoms of depression between the start of the study and the last study visit. Some, but not all, trials had looked at this outcome and the researchers only included data for drugs where all of the RCTs of that drug provided data on this outcome. All studies measured symptoms of depression on the Hamilton Rating Scale of Depression (HAM-D), an accepted scale.

The results from the eligible RCTs were then pooled using meta-analysis. The researchers used statistical techniques to determine whether the severity of the participants’ depression when they began the trial affected these results.

What were the results of the study?

The researchers identified 47 RCTs from the information provided by the FDA; only 35 of these provided results that could be included in the meta-analysis. These trials assessed the drugs fluoxetine (five trials), venlafaxine (six trials), nefazodone (eight trials) and paroxetine (16 trials). In total, the study covered 5,133 people.

Overall, antidepressants improved symptoms more than placebo, and this difference was statistically significant. However, the difference between antidepressants and placebo was relatively small (1.8 points on the HAM-D scale) and the researchers reported that it was not clinically meaningful according to the standard requirements of the National Institute for Clinical Excellence (three points on the HAM-D scale).

The researchers found that the more severe a participant’s depression was initially, the more effect antidepressants had in terms of improving symptoms compared with placebo. However, this improvement was only large enough to make a clinical difference in people with the most severe depression (people with scores of more than 28 on the HAM-D). The researchers found that antidepressants were more effective among severely depressed patients largely because these participants did not respond to placebo as well as those with milder depression.

What interpretations did the researchers draw from these results?

The researchers concluded that the difference between the effectiveness of antidepressants and placebo grew with the severity of depression. However, the differences were comparatively small, even among people with very severe depression. The most severely depressed people are less likely to respond to placebo, which is why antidepressants appear relatively more effective in this group.

What does the NHS Knowledge Service make of this study?

The strength of this research is that it included studies that have not been published. Published studies often report on significant findings that can bias any estimates of the overall effect. However, there are still some limitations to consider:

• The authors looked at all studies submitted to the FDA, which included all industry-sponsored research of these drugs. However, there could be non-industry sponsored trials that were missed.
• This meta-analysis only included studies performed before approval of these drugs (up to 1999). If the researchers had included studies published after approval, the results may have been different. Trials conducted to achieve approval of a drug often use highly selected individuals that are not representative of the patient population in general, and often place strict limitations on how the drug is used and what other treatments can be used at the same time. Trials conducted after a drug has been approved often have less strict inclusion criteria and give a better assessment of how well these drugs work in real life. For example, it is likely that antidepressants are used at the same time as non-drug therapies, but it is not clear whether the RCTs allowed this.
• The researchers point out that the trials mostly involved people with very severe depression. There were no trials involving people in the severe range, and only one that studied people with moderate depression. Therefore, these results cannot be applied to people in the moderate to severe depression range.
• Some of the figures that the researchers needed were missing and they had to use estimates instead. Any inaccurate estimations would affect the validity of the results.
• The authors only looked at depression symptoms as assessed on one scale. There are various ways of looking at recovery from depression, and these might show different results.
• All of the studies included in this meta-analysis were of adults. We cannot assume that the results will apply to children.

This study supports the idea that antidepressants do not work as well in people with milder depression. However, the assessment of severity itself is a skilled task and an individual’s response to treatment may vary. Patients should not, therefore, stop their treatment without consulting a health professional.

Sir Muir Gray adds…

Always look for the systematic review. This method of synthesising research produces the least biased, most accurate results.