Our review in Diabetes Care (1) described the understanding of schizophrenia and diabetes and antipsychotic medication in relation to how it emerged from the published literature when we wrote the article. New results will change the emphasis in a balanced review, and delays between literature search, writing, and publication may exacerbate this phenomenon. We did not seek to prove causation or to quantify the links between specific atypical antipsychotics and the onset of type 2 diabetes. However, we presented evidence that patients with schizophrenia receiving antipsychotic drug therapy are at elevated risk of type 2 diabetes. Hardy and Breier (2), in their letter in this issue of Diabetes Care, make several points that we would like to address.
First, they question the quality of the studies that we review, stating that they “are not suited to address causation.” Our review of over 50 years of literature included studies of variable quality but this richness would not satisfy modern criteria for systematic reviews. Most of that literature was not specifically designed to evaluate risks of diabetes. Epidemiologic methods can assess the comparative strength of association between a disease (in this case, diabetes) and potential factors that may influence the disease (in this case, the use of certain antipsychotics). These types of studies do not claim to demonstrate causation but are primarily for hypothesis generation to highlight an emerging issue to address in further research.
Second, Hardy and Breier state that we “did not cite studies by Koller and Hedenmalm” (3,4). The study by Koller et al. (3) was published after the cutoff date for our review article. In addition, both of these studies rely on databases (Food and Drug Administration’s MedWatch and World Health Organization, respectively) of spontaneously reported adverse events. Hardy and Breier correctly indicate in the following sentence that spontaneous reports are poorly suited to establishing disease incidence. Our review article only mentions case reports and spontaneously reported events as beacons that support the more rigorous epidemiologic and clinical studies.
Third, Hardy and Breier question why we did not discuss three other studies in which risperidone use was associated with a higher risk (5–7). Buse et al. (5) and Lee et al. (6) were published after the review article was complete, and Lee et al. (6) was published in a non-peer-reviewed journal. There are two concerns with the methodologies used in Lage et al. (7) and Lee et al. (6). Because concomitant medications were not controlled for, differences in the use of other diabetogenic agents between the risperidone and olanzapine groups were unknown and could have biased the results. Comparisons to an untreated control population were not completed, so the studied populations may have had a higher or lower propensity to develop diabetes. The study by Buse et al. (5) was only based on a prescription database, without access to medical information on the underlying disease states. Antipsychotics are used for diagnoses other than schizophrenia, so there are numerous potential confounding factors that cannot be controlled in the analyses.
Fourth, Hardy and Breier state that “none of the studies… reported statistically significant differences in rates of new diabetes.” In fact, in five more recently published studies (8–12), risperidone was significantly less likely to be associated with new-onset diabetes than olanzapine. However, four of these studies were published after the cutoff date for the review article (the study by Koro et al. [10] was included in our review).
Finally, they (reasonably) comment that the conclusions we drew from two studies (13,14) were “weakened by the fact that both studies were cross-sectional and without randomization to treatment.” Instead, they refer to studies by Beasley et al. (15) and Newcomer et al. (16) using the hyperinsulinemic-euglycemic clamp method. However, these studies were performed in normal volunteers, not schizophrenics, and the duration of treatment with the antipsychotic (2–3 weeks) may have been inadequate to induce glucoregulatory changes. Although a regression analysis was done for age, this did not include weight, concomitant drugs, and family history of diabetes. It is not clear that the hyperinsulinemic-euglycemic clamp performed in the two-step manner used in these studies would be able to assess hepatic insulin insensitivity as we noted in dogs because tagged glucose was not given. In Ader’s dog experiments (17), peripheral glucose uptake was also normal.
We stand by the conclusions that we drew in our review article at the time of its writing. Patients with schizophrenia and on antipsychotic drugs should be considered a high-risk group for type 2 diabetes. There are almost certainly differences in the relative hazards (including weight gain) with different antipsychotic drugs, which need to be evaluated in studies specifically designed for that purpose.
References
M.E.J.L. has received honoraria from Roche Products, Janssen-Cilag, Abbott Laboratories and Merck and has received grants from Sanofi-Synthelabo, Roche Products, Janssen-Cilag, GlaxoSmithKline, and Alizyme Therapeutics.
