The effect size of antidepressant may have been overestimated by a third, an extensive review of trial data has concluded.
The date authors obtained reviews for all antidepressant trial registered with the FDA between 1987 and 2004. A systematic literature search was carried out to determine whether the results has been published.
There were 74 trials, involving a total of 12,564 patients, investigating 12 different antidepressant drugs. The study found that 23 trials (31%), accounting for 3449 participants, had not been published.
The FDA judged 38 of the trials to have a positive finding, of which all but one were published. The remaining trials were considered to be negative(24) or questionable(12). Only three of these trials were published as a negative outcome; 22 were not published and the remaining 11 were published in a way that conveyed a positive outcome.
Of the published trials, 94% claimed a positive outcome, whereas the FDA analysis showed a positive result in only 51%. The overall effect size for the published studies was 0.41. However, when the unpublished studies were included the overall effect size was reduced to 0.31. The reduction in the individual effect size for the 12 drugs used in these trials ranged from 11% to 69%.
Kirsch et al calculated an almost identical overall effect size of 0.32 from FDA data for four of the 12 drugs. They suggest that this effect is too small to have any clinical significance; indeed, an overall effect size of 0.41 still remains below their cutoff of 0.5.
Other studies do not agree with the use of such an arbitrary demarcation between significance and non-significance.
However, compared with baseline the effect sizes for antidepressants(1.24) and placebo(0.92) are both extremely large; it is only the difference between them that is small(0.32). The fact that most of the improvement is because of the placebo effect has long been known; the Quality Assurance Project in the mid-1980s attributed 60% of the response to antidepressants to the placebo effect.
There has been a significant increase in placebo response rates in antidepressant trial over the past 20 years. It has been suggested that this is, in part, caused by a larger proportion of patients with milder self-limiting depression; the "effect" is caused by spontaneous remission rather than a therapeutic effect. This is not consistent, however, with the evidence that active placebos may be more effective.
In my opinion, the key question is whether the placebo effect is mainly caused by spontaneous remission or a genuine therapeutic effect. If the former, we should perhaps adopt a policy of watchful waiting, only offering antidepressant treatment if their is no improvement with in a specified time period. If the later, unless we have access to an effective alternative (such as CBT), we need to consider whether we can justify withholding such a highly effective treatment from our patients.
The date authors obtained reviews for all antidepressant trial registered with the FDA between 1987 and 2004. A systematic literature search was carried out to determine whether the results has been published.
There were 74 trials, involving a total of 12,564 patients, investigating 12 different antidepressant drugs. The study found that 23 trials (31%), accounting for 3449 participants, had not been published.
The FDA judged 38 of the trials to have a positive finding, of which all but one were published. The remaining trials were considered to be negative(24) or questionable(12). Only three of these trials were published as a negative outcome; 22 were not published and the remaining 11 were published in a way that conveyed a positive outcome.
Of the published trials, 94% claimed a positive outcome, whereas the FDA analysis showed a positive result in only 51%. The overall effect size for the published studies was 0.41. However, when the unpublished studies were included the overall effect size was reduced to 0.31. The reduction in the individual effect size for the 12 drugs used in these trials ranged from 11% to 69%.
Kirsch et al calculated an almost identical overall effect size of 0.32 from FDA data for four of the 12 drugs. They suggest that this effect is too small to have any clinical significance; indeed, an overall effect size of 0.41 still remains below their cutoff of 0.5.
Other studies do not agree with the use of such an arbitrary demarcation between significance and non-significance.
However, compared with baseline the effect sizes for antidepressants(1.24) and placebo(0.92) are both extremely large; it is only the difference between them that is small(0.32). The fact that most of the improvement is because of the placebo effect has long been known; the Quality Assurance Project in the mid-1980s attributed 60% of the response to antidepressants to the placebo effect.
There has been a significant increase in placebo response rates in antidepressant trial over the past 20 years. It has been suggested that this is, in part, caused by a larger proportion of patients with milder self-limiting depression; the "effect" is caused by spontaneous remission rather than a therapeutic effect. This is not consistent, however, with the evidence that active placebos may be more effective.
In my opinion, the key question is whether the placebo effect is mainly caused by spontaneous remission or a genuine therapeutic effect. If the former, we should perhaps adopt a policy of watchful waiting, only offering antidepressant treatment if their is no improvement with in a specified time period. If the later, unless we have access to an effective alternative (such as CBT), we need to consider whether we can justify withholding such a highly effective treatment from our patients.
