Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Equal on average does not mean equal for everyone


This fascinating title comes from a thoughtful commentary [1] on an even more thoughtful and useful trial [2] examining the effectiveness of SSRIs in a naturalistic trial in primary care. Together they make a useful contribution to thinking about effects of classes of drugs, of designs of trials to determine differences in practice, and how to use formularies.


The trial [2] was conducted in two primary care practice networks in the USA. The networks were of primary care physicians interested in optimising the care they provide through education and practice-based research.

The trial was designed to resemble real world practice. The decision to start an antidepressant was based strictly on physician judgement that there was clinical depression, rather than specific criteria for diagnosis. Patients and physicians knew what was being taken because blinding would obscure typical patient management. All decisions about discontinuation, switching, or dose change were made by physician and patient, with no pre-set criteria. There was no interaction between information gathering and physicians, who were uninfluenced by any results.

Apart from very obvious exclusions, adult patients with clinical depression were randomised (with adequate concealment) to initial doses of paroxetine (20 mg), fluoxetine (20 mg) or sertraline (50 mg). Drugs were free to patients. After enrolment, they were interviewed by telephone and questionnaire at several times over nine months. The primary outcome was SF-36 mental component summary, but a battery of other measures was used. Compliance (switching, discontinuation) was assessed.


There were no differences between the three groups at baseline. Patients were predominantly white women with major depression, with an average age of 46 years.

All three SSRIs showed substantial improvements in depression and other outcomes. There were no differences between treatments for any outcome, using any definition of intention to treat or per protocol analysis, or with any subgroup analysis. In the whole sample, the proportion meeting criteria for major depression fell from 74% at baseline to 26% at nine months (Figure 1). Rapid initial benefit was followed by slow continued improvement.

Figure 1: Major depression with treatment

Patients initially randomised to one treatment frequently changed treatment. By nine months only 44% were still taking the treatment to which they had been randomised. Some (about 15%) were lost to follow up after baseline or when on randomised treatment. Others either switched to another antidepressant or stopped treatment because of adverse effects or lack of efficacy, again without any difference between the three SSRIs (Figure 2).

Figure 2: Stopping or switching


The study was not a formal equivalence study, but rather one that sought to mimic clinical practice, and was powered to detect a moderate change in outcome. It found no practical difference between the three SSRIs. It also found that clinical practice, in which dose or drug changes were done on the basis of patient-physician interaction as they would be in real life, resulted in many patients stopping or switching drugs.

As the accompanying commentary says, all other considerations being equal, an initial choice cased on prescription costs is prudent, ethical, and clinically reasonable. The study is important because it provides evidence for formulary choices for SSRIs in this indication, and indicates how evidence for other choices could be determined.

However, it also provides evidence against formularies that restrict use of SSRIs for subsequent choices when an initial choice is unsatisfactory. Because these three SSRIs are equal on average does not mean that they equal for every individual. There may be a class effect, but class effects are average effects, and the thinking here supports taking individuals into account.


  1. G Simon. Choosing a first line antidepressant. Equal on average does not mean equal for everyone. JAMA 2001 286: 3003-3004.
  2. K Kroenke et al. Similar effectiveness of paroxetine, fluoxetine and sertraline in primary care. JAMA 2001 286: 2947-2995.
previous or next story in this issue