Bird et al's review of early psychosis intervention provides a useful meta-analysis of methodologically sound studies. ^{Reference Bird, Premkumar, Kendall, Whittington, Mitchell and Kuipers1} However, there are major problems with it. The authors have stated that their review focuses on the first 3–5 years following onset of illness. Yet they have omitted the 5-year results from the Danish OPUS study ^{Reference Bertelsen, Jeppesen, Petersen, Thorup, Øhlenschlaeger and le Quach2} and the UK Lambeth Early Onset (LEO) study. ^{Reference Gafoor, Nitsch, McCrone, Craig, Garety and Power3} Both of these follow-up studies found that, despite promising early results, ^{Reference Petersen, Jeppesen, Thorup, Abel, Øhlenschlaeger and Christensen4,Reference Craig, Garety, Power, Rahaman, Colbert and Fornells-Ambrojo5} which were included in Bird et al's review, positive effects were not sustained at 5 years.

Bertelsen et al concluded that intensive early intervention improved clinical outcomes in OPUS after 2 years, but the effects were ‘not sustainable up to five years later’. ^{Reference Bertelsen, Jeppesen, Petersen, Thorup, Øhlenschlaeger and le Quach2} This finding was not reported by Bird et al. In fact, the Bertelsen et al study was not cited at all. According to a personal communication from the authors, it was included in their review. However, they used primary references (in this case Petersen et al's analysis of 1- and 2-year outcomes ^{Reference Petersen, Jeppesen, Thorup, Abel, Øhlenschlaeger and Christensen4} ) to refer to all papers for all the trials included in the review. This seems idiosyncratic to say the least. More important, all they reported about the Bertelsen et al study is that ‘Only one trial of an early intervention service provided long-term data (up to 5 years post-randomisation)’.

Gafoor et al similarly found that specialist early intervention did not markedly improve outcomes at 5 years in LEO, ^{Reference Gafoor, Nitsch, McCrone, Craig, Garety and Power3} in accord with the 5-year findings from OPUS. Again this was not reported and the study was not cited. Bird et al's review was initially submitted in January 2009, long before the publication of Gafoor et al's study in this journal, but the final revision occurred after the latter was published. Although it would not have been practical to include Gafoor et al in the meta-analysis, publication of the review could have been delayed, if necessary, to allow a brief discussion of Gafoor et al's findings to be added. They significantly strengthened the evidence that promising early benefits are not sustainable, a very significant finding for a review of the effectiveness of early intervention in psychosis.

Bird et al have concluded that ‘it remains to be determined whether the effects of early intervention services are sustained’, yet they have omitted the best evidence of exactly that.