Stimpson et al (Reference Stimpson, Agrawal and Lewis2002) have taken an ‘all or nothing’ approach to evaluating randomised controlled trials (RCTs) for their systematic review. Their rigorous procedures eliminated over 98% of the 919 RCTs considered (although we note that the flow chart in Fig. 1 appears to ‘lose’ 166 of them without explanation). As a consequence, they have provided a matchless summary of the very best evidence about intervention for treatment-refractory unipolar depression but have left undescribed the very large quantity of remaining levels of evidence.

In 1999 Bauer and Dopfmer identified 11 placebo-controlled studies of lithium augmentation. As always, the trials were of varying quality; nevertheless, they concluded (using the three studies of highest quality, two of which were used by Stimpson et al) that there is ‘firm evidence’ in favour of lithium as an augmentation strategy for treatment-refractory unipolar depression, with a number needed to treat of 3.7. They supported their conclusion by performing a separate analysis adding a further six studies (that used either lower doses or shorter duration of lithium augmentation) and found a similar, indeed slightly stronger, effect size (Reference Bauer and DopfmerBauer & Dopfmer, 1999).

We note that there have been no studies of lithium augmentation against placebo for treatment-resistant unipolar depression that are of a suitable quality for a systematic review in the approximately 3-year period between the acceptance dates of the two papers cited above. We suggest that many clinicians now consider the weight of evidence (at many levels) supporting the use of lithium as an augmentation strategy for treatment-refractory unipolar depression sufficiently compelling. Thus, it is unusual for our service dedicated to treatment-resistant depression to receive referrals of patients not yet tried on lithium. Although further and better RCTs of lithium augmentation would be welcome (even Bauer & Dopfmer identified only 234 subjects studied), many would feel that other questions now have more clinical salience. Pressing examples might include whether psychological treatments are effective in these patients, how they compare with lithium augmentation, and how olanzapine augmentation (for which a large body of evidence is emerging; see Reference Dube, Anderson and PaulDube et al, 2002) compares with both.