OBJECTIVES: To analyze the comparative safety of prescription medications routinely used to treat sleep problems in VA.

METHODS: VA guidelines and formularies were used to identify prescriptions to treat sleep in 2011-2012: zolpidem (preferred), trazodone, benzodiazepines and sedating antihistamines. Subjects were Veterans with a new prescription who did not have a history of sleep medication use or suicidal behavior. Propensity scoring matched subjects with new prescriptions for the reference agent (zolpidem) to one subject exposed to each of the other agents, controlling for more than 30 variables from the VA Corporate Data Warehouse (CDW) including demographics, disability, diagnoses (sleep disorder, psychiatric, substance use, alcohol use, cardiovascular, oncologic, neurologic and pain conditions) and inpatient days. Cox proportional hazards models estimated exposure effects on suicide attempts first in VA electronic medical records, and then repeated using Suicide Prevention Action Network (SPAN) reports. The VISN 2 Human Subjects Review Board approved this study.

RESULTS: In 2011-2012, 348,449 Veterans met inclusion criteria. Compared to zolpidem, hazard ratios for suicide attempt in electronic medical records were 1.656 (95% CI: 1.107-2.477) for trazodone, 1.367 (95% CI: 0.877-2.129) for benzodiazepines, and 1.545 (1.028-2.322) for sedating antihistamines after adjusting for history of inpatient days, comorbidity burden, demographics and a range of psychiatric, substance use and other diagnoses in the twelve months prior to the first exposure. Yet, SPAN data included more suicide attempts for the cohorts and the same models using these outcomes data did not identify any significant differences in suicide attempt risk by agent.

CONCLUSIONS: Compared to zolpidem, the hazard of suicide attempt was 65.6% higher with trazodone, 36.7% higher on benzodiazepines and 54.5% higher on sedating antihistamines. Yet, the incidence of suicide attempt as recorded in VA SPAN data was similar across all drugs.