Back to 2.6

2.7 Standards for testing the efficacy of SSRIs

The definitive answer to the question "Do antidepressants work?" is the legal one. No one is allowed to market a medicinal drug without a license which, by law, can be given only when the efficacy of a drug has been proved. Most countries require this and have agencies to enforce the law; the "regulators" in the UK include the Medicines Control Agency (MCA), Committee on Safety of Medicines (CSM) and the European Medicines Control Agency (EMEA).

The UK regulators work in strict secrecy which complicates evaluation of their work; (Medawar, 1996) however, it is clear that the scientific basis of some important assessments is slight. For example, given the overlap between "anxiety" and "depression", it seems extraordinary that the regulators should have accepted evidence in which antidepressant efficacy was measured in trials where people took anxiolytics at the same time.

Why has this methodology slipped into the protocols of so many trials? The welfare of patients would have been a factor and convenience and opportunism may also have played some part. But perhaps the driving force was unconscious bias. Its source may have been the conviction that it wouldn’t matter to give both drugs together, because antidepressant and anxiolytic drugs were quite different things. After thirty years of tight market segregation, such assumptions might have been made almost as a matter of course

Much of the bedrock evidence put before the regulators appears suspect for this reason. Publicly available data from the US Food & Drug Administration (FDA) shows that fluoxetine (Prozac) was licensed in spite of, rather than because of, clear-cut evidence of efficacy from controlled trials:

One other standard trial procedure seems capable of wreaking havoc with efficacy evaluations. This is a feature of many protocols which is not only acceptable to the regulators, but also positively advocated by leading authorities. The principle author of the following recommendation is one of the most widely-published experts on the SSRIs; he was also a member of the Committee on Safety of Medicines at the time the main evidence on the SSRIs was being assessed (1987 -1992):

"Studies can be flawed by including too many inappropriate patients. The inclusion of treatment-resistant patients, who are often concentrated in inpatient studies, can reduce the likelihood of finding a positive result. Similarly placebo responders can confuse the picture. The inclusion of a placebo treatment period before the entry severity criteria is applied often helps to reduce this source of error". (Montgomery & Lambert, 1989)

In other words, to demonstrate the efficacy of an antidepressant (a positive result), one should first eliminate from the group of patients studied anyone severely enough depressed to be hospitalised, and anyone whose depression readily responds to placebo. Thus, inpatients were excluded from the pivotal trials on fluoxetine, and all four protocols included a placebo washout period to pre-screen unwanted subjects. The procedure involved measuring the Hamilton (HAM-D) rating for depression to identify possible patients for trials, then putting everyone on placebo for a week and excluding from the trial anyone whose HAM-D rating had dropped to below 80% of the original value, or below a specified HAM-D score.

But how can one use an active drug vs placebo model as the gold standard for efficacy, when permitting the pre-screening of patients to eliminate people who readily respond to placebo ? This would lead (and has led) to gross underestimation of the value of placebo treatment; even in the most severe and obvious cases of depression, it seems like screening a jury to extremes:

"A few years ago, we tried an experimental design in one of our studies which we hoped would eliminate ‘placebo reactors’ and increase our sensitivity in distinguishing between drugs. All depressed patients who entered the hospital and were candidates for the study were first placed on a week of placebo treatment. At the end of the week, the psychiatrist was then asked to make a decision as to whether the patient should be admitted to the study ... We lost 50% of our potential sample, as that number of patients had shown a degree of spontaneous improvement which would have confounded the effects of future treatment. The tendency of depressed patients to improve spontaneously certainly creates difficulties in the clinical evaluation of drugs." (Hollister, 1972)

The increasing tendency (Senn, 1997) to exclude placebo responders might explain the apparent decline in the magnitude of the placebo response reported in clinical trials, over the years. In their review of all properly controlled studies of antidepressants, Smith et al., (1969) reported the median improvement rate on placebo to be 46% (and for active drugs, 61%). Reported response rates on placebo come closer to 33% today.

For licensing purposes, controlled trials to demonstrate the efficacy of SSRIs typically last about six weeks, though the minimum recommended period of treatment in clinical practice is now of the order of six months. Since the 1990s, longer-term studies have been conducted; they appear flawed to about the same extent, but in different ways. See 3.7.


Contents page