A meta-analysis of patient-level data found that the benefits of antidepressive drugs in comparison with placebo rise with increasing severity of depression, and may be limited or non-existent in mild depression.

Previous analyses have indicated that the therapeutic effect of antidepressive drugs is greater in patients with more severe depression, however these have not included many patients with less severe depression and also had other limitations. The authors of this study aimed to clarify the effect by ensuring inclusion of patients with a range of baseline severity, by excluding studies that excluded some placebo responders with a placebo washout period, and by analysing patient-level data. They carried out a literature search for randomised, placebo-controlled trials of antidepressive drugs in out-patient populations with a range of depression severity, without a placebo washout period, and where original patient-level data were available. Other criteria included at least six weeks active vs. placebo treatment, and HAM-D scores at baseline and end of treatment. The analysis studied the relationship between baseline symptom severity and subsequent symptom change from intake to the end of acute treatment.

There were 2,164 citations identified in the original literature search, of which 281 were retrieved for full assessment; 258 of these had to be rejected, most frequently (118) because the study included a placebo washout. Of the remaining 23 studies, patient-level data were available for six (n=718), which were thus used for the analysis. Three of the included studies used imipramine and three paroxetine; study quality was generally good, and the range of baseline HAM-D scores was 10 to 39 (for HAM-D, a score of 8 to 13 is considered to be mild and 23 and above very severe depression).

Analysis found a significant interaction between baseline symptom severity and the change in symptoms with treatment. Plotting regression lines for the relation between initial severity and change in symptoms showed a positive effect for both drug and placebo, however the slopes of the two lines were different: they converged at the lower end of symptom scores and the size of the difference between the lines increased as symptom score increased. For illustrative purposes, the symptom scores were divided into mild to moderate, severe, and very severe using standard criteria, and numbers-needed-to-treat (NNT) compared to placebo calculated for the three categories. These were 16, 11, and 4 respectively. Tests by drug class found no significant difference between imipramine and paroxetine.

The authors also estimated the point at which the drug-placebo difference crossed the thresholds of clinical significance defined for antidepressive treatment by NICE. These thresholds were met where the baseline HAM-D scores were 25 or above for the more liberal threshold and 27 and above for the tighter threshold.

Based on their analysis, the authors conclude that the efficacy of antidepressive drug treatment varies with baseline symptom severity. For patients with less severe depression, the true drug effect was limited whereas for patients with very severe symptoms drugs were markedly superior to placebo. They note that there are limitations to their analysis, but comment that its results are consistent with previous work; they emphasise that their analysis applies only to acute, not chronic and maintenance therapy.