Kessing et al ^{Reference Kessing, Thomsen, Mogensen and Andersen1} describe the risks of diabetes in clinical practice from a large-cohort, observational study of Danish patients requiring antipsychotics. We believe that the relative risks of subsequent incident diabetes that they report for individual antipsychotics are at odds with established literature. The preponderance of evidence has demonstrated that ziprasidone has limited effect on metabolic indices associated with the development of diabetes. We present some of that evidence below.

In the CATIE study of 1493 patients with schizophrenia receiving olanzapine, quetiapine, risperidone, ziprasidone or perphenazine for up to 18 months, ziprasidone was the only drug associated with improvement in glycosylated haemoglobin, total cholesterol and triglycerides. Meyer and colleagues ^{Reference Meyer, Davis, Goff, McEvoy, Nasrallah and Davis2} reported that, in the CATIE trial, the prevalence of metabolic syndrome increased for olanzapine (from 34.8% to 43.9%) but decreased for ziprasidone (from 37.7% to 29.9%), and that the comparison between ziprasidone and olanzapine was statistically significant (P = 0.001).

In the EUFEST study of 498 patients with first-episode schizophrenia assigned to haloperidol, amisulpride, olanzapine, quetiapine or ziprasidone for 12 months, weight change was greatest on olanzapine and among the lowest on ziprasidone. ^{Reference Kahn, Fleischhacker, Karayal, Siu and Pappadopulos3} Ziprasidone was not associated with untoward effects on any other metabolic risk factors.

A pooled analysis in 2009 of over 100 Pfizer-sponsored randomised controlled trials found no significant differences between 1605 individuals given ziprasidone and 677 given placebo in total cholesterol, fasting glucose or fasting triglycerides (details available on request).

Yood et al, ^{Reference Yood, DeLorenze, Quesenberry, Oliveria, Tsai and Willey4} in a 55 287-member inception cohort of antipsychotic users, found 357 cases of newly treated diabetes. Ziprasidone was among the group of agents with the lowest risk of diabetes. Patients exposed to olanzapine and clozapine had an increased risk of the illness.

A consensus statement on antipsychotic drugs and obesity published by the American Diabetes Association et al in 2004 concluded that increased risks of obesity, dyslipidaemia and diabetes are most associated with clozapine and olanzapine; little or no significant weight gain, diabetes and dyslipidaemia was associated with aripiprazole and ziprasidone, although it should be noted that these agents had not yet been used extensively at that time. ⁵ Further, the panel suggested switching patients who develop worsening glycaemia or dyslipidaemia to a second-generation antipsychotic not associated with significant weight gain or diabetes (i.e. ziprasidone or aripiprazole). Standards of practice that promote agents with lower metabolic risks may be a confounding factor in naturalistic studies.

Kessing et al acknowledge that ‘individuals at higher risk of diabetes because of a personal history of obesity or inactivity, a family history of diabetes or other risk factors may have been prescribed agents perceived to confer a lower risk of diabetes’. This channelling bias affects the generalisability of their results.

They report a low risk of diabetes for aripiprazole, but the drug did not becoming commercially available in Denmark until 2004, only 1.5 years before the end of this 10-year study. A small number of patients were exposed for a limited period of time, making the direct comparison with ziprasidone not meaningful.

With regard to Table 3, given the widely differing times of drug exposure and the ultimate position of any individual agent within a single patient's treatment regimen, conclusions about the risk of an individual agent v. a drug class may be inappropriate based on this study design.

We are concerned about how clinicians will interpret Kessing et al's findings for ziprasidone, as the results stand in contrast to the relative risks for diabetes reported in the established literature.