Am J Psychiatry 161:9, September 2004
Incidence of Newly Diagnosed Diabetes
Attributable to Atypical Antipsychotic Medications
Douglas L. Leslie, Ph.D.
Robert A. Rosenheck, M.D.
Objective: The purpose of the study was to determine the pro-
portion of patients with schizophrenia with a stable regimen of
antipsychotic monotherapy who developed diabetes or were
hospitalized for ketoacidosis.
Method: Patients with schizophrenia for whom a stable regi-
men of antipsychotic monotherapy was consistently prescribed
during any 3-month period between June 1999 and September
2000 and who had no diabetes were followed through Septem-
ber 2001 by using administrative data from the Department of
Veterans Affairs. Cox proportional hazards models were devel-
oped to identify the characteristics associated with newly diag-
nosed diabetes and ketoacidosis.
Results: Of the 56,849 patients identified, 4,132 (7.3%) devel-
oped diabetes and 88 (0.2%) were hospitalized for ketoacidosis.
Diabetes risk was highest for clozapine (hazard ratio=1.57) and
olanzapine (hazard ratio=1.15); the diabetes risks for quetia-
pine (hazard ratio=1.20) and risperidone (hazard ratio=1.01)
were not significantly different from that for conventional an-
tipsychotics. The attributable risks of diabetes mellitus asso-
ciated with atypical antipsychotics were small, ranging from
0.05% (risperidone) to 2.03% (clozapine).
Conclusions: Although clozapine and olanzapine have greater
diabetes risk, the attributable risk of diabetes mellitus with
atypical antipsychotics is small.
(Am J Psychiatry 2004; 161:1709–1711)
T here is some evidence to suggest that atypical antip-
sychotics can cause weight gain and increased risk of dia-
betes mellitus (1–5). Other studies have suggested that
there might be a link between atypical antipsychotics and
diabetic ketoacidosis (6, 7). However, few published stud-
ies have examined diabetes mellitus prevalence (8) or risk
of new-onset diabetes mellitus or diabetic ketoacidosis for
patients treated with antipsychotics (6, 9), and, to our
knowledge, no studies have reported diabetes mellitus in-
cidence rates in this population.
The goals of this study were to determine the proportion
of patients with schizophrenia with a stable regimen of an-
tipsychotic monotherapy who developed diabetes melli-
tus or were hospitalized for diabetic ketoacidosis and to
identify patient demographic, clinical, and pharmacologi-
cal characteristics associated with these adverse events.
We identified 73,946 patients with schizophrenia in the De-
partment of Veterans Affairs (VA) for whom a stable regimen of
antipsychotic monotherapy was consistently prescribed during
any 3-month period between June 1999 and September 30, 2000.
We defined five groups of antipsychotic medications: clozapine,
risperidone, olanzapine, quetiapine, and all conventional antip-
sychotics. Ziprasidone and aripiprazole were not included in the
study because they had only recently been approved for use.
Patients with any outpatient claims for diabetes mellitus (N=
11,069) or less than two medical primary care visits (N=6,028) in
the previous 6 months were excluded from the sample. Stably
medicated patients with no diabetes mellitus were followed
through September 30, 2001. Patients who received a diagnosis of
diabetes mellitus or who were hospitalized for diabetic keto-
acidosis during that period were identified.
Cox proportional hazards models were used to model the time
to diabetes mellitus diagnosis and time to diabetic ketoacidosis
hospitalization. Independent variables included in the models
were antipsychotic agent prescribed during the stable period,
date of the end of the stable period, age, gender, race, income, co-
morbid mental health diagnoses, levels of service use during the
stable period, and the degree of VA service-connected disability.
We also calculated the attributable risk of diabetes mellitus and
diabetic ketoacidosis associated with each atypical antipsychotic,
which is the estimated proportion of patients taking each drug
who would not have received a diagnosis of diabetes mellitus or
diabetic ketoacidosis if they had been taking a conventional an-
tipsychotic (10, p. 38).
Of the 56,849 patients in the sample, 4,132 patients
(7.3%) received a diagnosis of diabetes mellitus during the
follow-up period, representing an annual incidence rate of
4.4%. Only 88 patients (0.2%) were hospitalized for dia-
betic ketoacidosis. Figure 1 shows the fitted survival func-
tions associated with each medication group from the
model predicting diabetes mellitus diagnosis. The attrib-
utable risk associated with these medications was highest
for clozapine (2.03%), followed by quetiapine (0.80%),
olanzapine (0.63%), and risperidone (0.05%).
In the diabetic ketoacidosis model, hazard ratios associ-
ated with atypical antipsychotics were larger than in the
diabetes mellitus model, but the attributable risks were
much smaller (ranging from 0.004% for risperidone to
0.071% for clozapine). In the entire sample, hazard ratios
for diabetic ketoacidosis were significant only for cloza-
pine (hazard ratio=3.75, 95% confidence interval [CI]=
1.39–10.09) and olanzapine (hazard ratio=1.77, 95% CI=
1.05–2.98). None of the hazard ratios for diabetic ketoaci-