Since Kuhn’s paper on imipramine, it has been believed that antidepressants are most effective in severe depression. Many people, who are sceptical of their widespread use for milder cases, maintain that antidepressants are nevertheless effective and necessary in severe depression. It has also been suggested that the reason some studies find little difference between antidepressants and placebo is because they are conducted with people with mild depression, who dilute the antidepressant effect (National Institute for Clinical Excellence 2004). The National Institute for Clinical Excellence guidelines on the treatment of depression have gone as far as to suggest recently that antidepressants should not be prescribed to people with mild depression, and should only be used in cases of moderate and severe depression. However there is actually little evidence for the presumption that antidepressants are effective in severe depression.
An early review of the relation between the type of depression and antidepressant response found that evidence for predictors of antidepressant response was sparse and included few controlled studies. Although they suggested that there was evidence for symptoms associated with an endogenous profile, such as anorexia, retardation and late sleep disturbance being associated with better response, they also noted the inconsistency of studies. They concluded that ‘the relationship between severity of illness and tricyclic response is unclear’. A later and larger review by Joyce & Paykel (1989) did not find enough evidence to suggest that sleep and appetite disturbance predicted antidepressant response, and found that studies disagreed about whether endogenous depression responded better or worse to antidepressants than other types of depression. They concluded by suggesting that tricyclic antidepressants might be most useful in the middle range of severity and ‘endogenicity’.
More recently a few studies and meta-analyses have examined this issue. A meta-analysis by Angst et al. claimed to show evidence that the efficacy of antidepressants relative to placebo was greater for people who were more severely depressed initially. However the effects were weak and mostly not statistically significant. Another meta-analysis found more impressive gradients of effects, but full data were only provided for ‘investigational’ antidepressants and not for ‘established’ antidepressants, whose relationship with severity appeared to be weaker. A recent analysis found that people on antidepressants showed a greater response with increasing initial severity of depression in contrast to the placebo group whose response rates tailed off at higher severity levels. However even for the most severely depressed subgroup, the benefits of antidepressants over placebo were only around 4 points on the Hamilton rating scale, a difference that is of doubtful clinical relevance and can easily be explained by drug-induced effects. The pattern of response in the antidepressant group may also reflect ‘regression to the mean’. This is the phenomenon by which observations naturally tend to gravitate towards the mean value, and hence those people with most severe depression to begin with will tend to show the most improvement. On the other hand, one recent meta-analysis found no effect of initial severity on treatment response. The NICE meta-analysis also failed to find a consistent gradient between severity and antidepressant efficacy. However the review still concluded that a relationship had been shown. In fact the middle severity group tended to show the greatest drug-placebo differences in this analysis, but the number of studies in each group was small (National Institute for Clinical Excellence 2004).
Some individual trials of antidepressants that found no overall effects, found a relatively stronger effect in some of the most severely depressed subjects in post hoc analysis. However post hoc analysis is where the authors look for significant results without predetermining what particular tests are of interest. This sort of analysis is commonly referred to as a ‘fishing expedition’ and it is well known that it can highlight results that are positive just by chance. Another similar trial conducted in primary care found no association between ‘melancholic’ depression, that is the most severe depression, and antidepressant efficacy. These trials were conducted exclusively with people with relatively mild depression and so could not assess the relation over the whole severity range. The meta-analyses too were based mostly on outpatient studies. On the other hand, my own meta-analysis of older trials found that effects in inpatients were small and not statistically significant compared with outpatients, where effects were somewhat larger. In addition, it has long been believed that antidepressants are relatively ineffective in severe depression accompanied by psychotic delusions. A study of inpatients found that it was actually greater severity of depression that predicted a worse response to antidepressants and not the presence of delusions. However the fact that the two are associated has led to the impression that it is psychiatric features that mark a lesser response to antidepressants. Recently a large trial of antidepressants versus placebo showed that people experiencing the depressive phase of manic depression, or bipolar disorder, who were also on ‘mood stabilisers’, showed no better response to antidepressants than they did to placebo. Since the mood swings in manic depression are often more severe than other episodes of depression, this trial provides further evidence that, contrary to current opinion, antidepressants are not superior to placebo even in the most severe forms of depression.
Some of the clinical trial evidence suggests that there may be a group of people in the mid range of severity who benefit most from having an active antidepressant compared with a placebo. This would not be expected from a simple biological effect. It is more likely that people within the middle of the severity spectrum have the highest commitment to the idea of the effectiveness of drug treatment and would therefore be most susceptible to non-specific pharmacological and psychological effects. Many people with milder depression probably do not even consider themselves depressed and many do not want to take drug treatment. People with more severe depression, including the sort of people who end up in hospital, may have little faith in any intervention.