To the Editor: In their survey of diabetes mellitus in patients receiving neuroleptics, Michael J. Sernyak, M.D., et al. (1) reported that the prevalence was 9% higher in those treated with atypical neuroleptics than in those treated with typical neuroleptics. Dr. Sernyak et al. acknowledged some of the limitations of their study, including the fact that it was retrospective, there was no attempt to determine diabetes status at baseline, and the screening period was only 4 months long. They acknowledged that this narrow time frame yielded a virtual cross-sectional group, precluding determination of the temporal relationship between neuroleptics and the development of diabetes mellitus—a basic requirement when assessing causality. Thus, they showed an association between atypical neuroleptic treatment and diabetes, but they did not establish causality.

Some other limitations are worth noting. No matched comparison subjects were used in this retrospective study. A history of alcoholism was significantly more common in the atypical neuroleptic group than in the typical neuroleptic group, and alcohol-induced pancreatitis may have accounted for at least part of the higher prevalence of diabetes in the former group.

Finally, the numbers of patients taking each agent varied widely, making statistical analyses difficult. For example, the percentage of patients taking quetiapine was so small that the odds ratio for diabetes in this group was higher than normal in the 40–49-year age range but lower than normal in the 60–69-year age range—a finding that is the opposite of what one would expect to see clinically (2).

The possibility that patients taking neuroleptics may develop diabetes is a valid question, but because of design limitations, the current study did not yield any answers regarding causality.

The authors are employees of AstraZeneca Pharmaceuticals.