Dr Moncrieff (Reference Moncrieff2003) re-analysed the data of a Cochrane meta-analysis by Wahlbeck et al (Reference Wahlbeck, Cheine and Essali2000) on the comparison between clozapine and conventional antipsychotic drugs for treatment-resistant schizophrenia. After selecting nine randomised controlled trials and analysis she concluded that the Cochrane review might have overestimated the effects of clozapine as she found a lower overall effect. This was explained by the use of data from intention-to-treat analysis in the largest included study by Rosenheck et al (Reference Rosenheck, Cramer and Xu1997) and inclusion of the large study by Essock et al (Reference Essock, Hargreaves and Covell1996), which was excluded in the Cochrane review.

There are good reasons for reporting the results from the studies by Rosenheck et al (Reference Rosenheck, Cramer and Xu1997) and Essock et al (Reference Essock, Hargreaves and Covell1996) separately from the other seven studies rather than giving the overall results. These two studies are long-term studies with durations of 1 and 2 years, respectively. The study populations were much larger than most of the other studies, which were short-term studies lasting 6–29 weeks. The two long-term studies found a small to no difference in treatment effect between clozapine and the conventional antipsychotic. These results have a large negative impact on the overall effect because of the large study populations. However, the use of intention-to-treat analysis will result in smaller differences between the clozapine and control group the longer the study lasts, because drop-outs are classified as relapses irrespective of the reason for discontinuation. Longer studies tend to have larger drop-out rates, as is also apparent in this meta-analysis, resulting in smaller differences between study groups.

Reporting the results from the short-term and long-term studies separately will probably show that clozapine has a higher treatment effect than that reported by Moncrieff. Short-term studies explore the pharmacological efficacy of a medicine whereas long-term studies explore the treatment effect in daily practice and can be influenced by the patient's willingness to continue treatment. These results should be reported separately.