We thank Dr Temmingh for his interest in our paper. ^{Reference Peluso, Lewis, Barnes and Jones1} The use of sulpiride in CUtLASS-1 was discussed in the original report ^{Reference Jones, Barnes, Davies, Dunn, Lloyd and Hayhurst2} and subsequent correspondence. ^{Reference Jones, Barnes, Elton, Davies, Dunn and Lloyd3} The Cochrane review of sulpiride in schizophrenia ^{Reference Soares, Fenton and Chue4} concluded that extrapyramidal side-effects (EPS) may be less frequent for individuals taking sulpiride but that no result regarding either direct or proxy measures of EPS reached statistical significance. Moreover, the review includes a report that sulpiride seemed to cause problems with increased prolactin levels and galactorrhoea. ^{Reference Weizman, Maoz, Treves, Asher and Ben David5} Claims that the drug shows particular efficacy against negative symptoms were not supported by trial data. Thus, any evidence that sulpiride is a particularly atypical typical antipsychotic is, at best, not strong. It is similar to amisulpride in its chemical structure and receptor pharmacology, with highly selective affinity for pre- and post-synaptic D₂ and D₃ receptors, ^{Reference Schoemaker, Claustre, Fage, Rouquier, Chergui and Curet6} characteristics of both drugs that question the validity of the typical v. atypical classification.

We acknowledge in the paper ^{Reference Peluso, Lewis, Barnes and Jones1} that a cautious approach is needed when undertaking a secondary analysis of any trial data because sample size will have been predicated on the primary, not secondary, hypothesis, and because many hypothesis tests may be undertaken; type 1 and 2 statistical errors lie in wait even for a Cochrane review. That is why we defined a doubling or halving of EPS as a clinically meaningful effect size to use in conjunction with significance testing. This was a matter of clinical judgement rather than being completely arbitrary. Like the conventional 5% cut-off used in significance testing, we hope it has some value while acknowledging that all these decisions are subject to controversy. ^{Reference Glaser7} In deciding to dichotomise EPS in this way, we were aiming to keep things simple and avoid erroneous conclusions from multiple secondary analyses.

We agree that the findings raise important points for the design of superiority (and non-inferiority) trials, and for crucial policy decisions based on health economic evidence. However, we hope that the findings may also remind clinicians that older antipsychotic drugs may be worth a thought when trying to find the right medicine for a particular patient.