We agree with the comments of Grover & Kulhara on the lack of information about the specific reasons for the prescription of antiparkinsonian drugs in our observational study. We have stated that such prescribing might have been influenced by factors other than the occurrence of EPS. However, previous naturalistic studies have shown that the use of antiparkinsonian medication was highly correlated with clinical indices of EPS when patients were prescribed antipsychotics (Reference Barak, Shamir and WeizmanBarak et al, 2002; Reference Bobes, Gilbert and CiudadBobes et al, 2003; Reference Montes, Ciudad and GasconMontes et al, 2003). In addition, the sudden change in the incidence of antiparkinsonian drug use following introduction of atypical antipsychotics in the entire population (not just among patients who switched type of antipsychotic therapy) makes it unlikely that physician prescribing habits were a strong alternative explanation for our findings.

Since we observed the same patients over time in the analysis of drug switching, changes in antiparkinsonian drug prescribing following the switch could be explained by the differential effects of antipsychotics on EPS.

Nevertheless, antiparkinsonian drug prescribing is only a marker of EPS and cannot perfectly reflect the incidence of EPS. Owing to the limitation of our data-set (which did not include indications for prescriptions), we cannot exclude the possibility that some patients may have been prescribed antiparkinsonian medication because they had Parkinson's disease, not because they had EPS caused by antipsychotics.

Grover & Kulhara question why we included only 266 GPs in this study. We selected from the GPRD only those patients who had been diagnosed with schizophrenia and prescribed antipsychotics between 1992 and 2000. Therefore 6356 patients who met those requirements and their 266 general practices were included in the study.

Grover & Kulhara raise the possibility that patients might have taken both classes of antipsychotics simultaneously. We examined the effects of switching antipsychotics on antiparkinsonian drug prescribing by classifying patients into two groups. We defined the TA group as patients who had been prescribed typical antipsychotics with no atypical antipsychotic use before the switch, completely stopped typical antipsychotics and subsequently switched to atypical antipsychotics, with no typical antipsychotic use for at least 2 years after the switch. The TT group included patients who were prescribed one typical antipsychotic (e.g. chlorpromazine) then switched to a different typical antipsychotic (e.g. haloperidol), and who never received an atypical antipsychotic during the study period. Therefore, by definition, no patients in our study were receiving a combination of both classes of antipsychotics.