Using complete datasets (including unpublished data) and a substantially larger dataset of this type than has been previously reported, we find that the overall effect of new-generation antidepressant medications is below recommended criteria for clinical significance. We also find that efficacy reaches clinical significance only in trials involving the most extremely depressed patients, and that this pattern is due to a decrease in the response to placebo rather than an increase in the response to medication.
Similar to prior reports [3
], this analysis of U.S. FDA data for four new-generation antidepressants suggests an association between initial severity and the benefit of antidepressant medication. Unlike prior studies, we restricted our analysis to complete datasets that included all trials conducted, whether published or not. Thus, simple publication bias cannot underlie the results. We compared drug–placebo differences in improvement to criteria for clinical efficacy, and we used meta-regression procedures [11
] to identify the relation of severity to improvement. Although we were able to replicate previously reported decreases in the placebo response as a function of increasing baseline severity, we found no linear relation between severity and response to medication.
NICE used a three-point difference in HRSD change scores, or a standardized mean difference of 0.50, as criteria of clinical significance [1
]. By that criterion, the differences between drug and placebo were not clinically significant in clinical trials involving either moderately or very severely depressed patients, but did reach the criterion for trials involving patients whose mean initial depression scores were at the upper end of the very severe depression category (mean HRSD baseline ≈ 28; –). Given these data, there seems little evidence to support the prescription of antidepressant medication to any but the most severely depressed patients, unless alternative treatments have failed to provide benefit.
A prior meta-analysis of published data only reported a very small significant difference between the antidepressant effect of fluoxetine and venlafaxine, but did not assess the effect of baseline severity as a moderator [23
]. Our analyses failed to reveal any effect of drug type on efficacy or on the relation between severity and efficacy. It is possible that differences associated with drug type might be found with the inclusion of clinical trials conducted after the approval process, but analyses of head-to-head comparisons suggest that they are not likely to be large enough to be of clinical importance [23
The response to placebo in these trials was exceptionally large, duplicating more than 80% of the improvement observed in the drug groups. In contrast, the effect of placebo on pain is estimated to be about 50% of the response to pain medication [24
]. A substantial response to placebo was seen in moderately depressed groups and in groups with very severe levels of depression. It decreased somewhat, but was still substantial, in groups with the most-severe levels of depression.
Although baseline severity related to degree of improvement in the drug groups, the pattern was not linear. Instead, patients who by APA criteria were moderately depressed and those at the very high end of the severely depressed category (i.e., those with initial HRSD scores greater than 28) showed less improvement than those at the lower end of the severely depressed category. The curvilinear relation depended on only one trial of moderately depressed patients. When that outlier trial is excluded, there is no relation between baseline severity and antidepressant response. However, all of the other trials were with groups with mean initial HRSD scores in the very severe range (i.e., ≥23). What is missing from the FDA data, however, are clinical trials with patients with initial depression scores in the severe range (19–22), and there was only one study with patients in the moderately depressed range. Had groups with a wider array of baseline depression scores been assessed, the curvilinear pattern might have been more obvious; in which case, clinically significant benefits for severely depressed patients might have been obtained. To perform this task in an unbiased way, it would be necessary for data for all approved medications to be available, even those gathered after the medication is approved. Having all the information available would also obviate the need to impute missing standard deviations, a limitation of the current investigation. Public availability of complete data on approved mediations might be made a condition of approval to solve these problems.
Finally, although differences in improvement increased at higher levels of initial depression, there was a negative relation between severity and the placebo response, whereas there was no difference between those with relatively low and relatively high initial depression in their response to drug. Thus, the increased benefit for extremely depressed patients seems attributable to a decrease in responsiveness to placebo, rather than an increase in responsiveness to medication.