Background: Long-term benzodiazepine use is associated with suicidal behavior, yet within the class comparative safety is largely unknown. A new statistical post-market surveillance system found an increased risk signal for alprazolam in commercial claims that required testing. Commercial insurance covers most veterans.
Objectives: To test the previous signal in a national sample of patients and compare effects to other benzodiazepines and buspirone.
Methods:
Design: Exposure-only, within-person dynamic analyses in US adults’ commercial claims over 2 years. We compared results from the alprazolam cohort (N=2,620,554 unique patients) to parallel analyses in (long-acting) lorazepam, (short-acting) diazepam, and buspirone cohorts, respectively. Setting: MarketScan (IBM Watson) (est. 164 million unique patients) with an alprazolam fill in 2012-2019 and without an alprazolam fill in 2010-2011. Exposure: The index prescription was the first fill after >=12 months without a fill. Main Outcome Measure: Suicide attempt or self-harm ICD-9/ICD-10 codes in outpatient, inpatient or emergency department settings. Analyses: A discrete-time survival model based on a logistic regression with complementary log-log link function was used such that the exponential of the estimated regression coefficient is a hazard ratio (HR). Month was the unit of analysis and a categorical variable for the 12-month follow-up. Within each subject we analyzed suicidal attempts and self-harm in months with and without the focal drug. Covariates were age, sex, Charlson Comorbidity Index, suicide attempt history, and baseline diagnoses and medications. A duration-response analysis in patients on the modal daily dose (0.5mg) characterized the cumulative number of treatment months as a time-varying covariate.
Results: In total, 1,413 suicidal events occurred in 11,110,422 alprazolam months (12.72/100,000 months), compared to 1,679 in 30,869,706 months without alprazolam (5.44/100,000 months). Adjusting for age, gender, history of suicidal events, and suicide-related diagnoses and medications, alprazolam had a HR of 2.30 (95% CI = 2.14, 2.47). Duration-response analysis found a 9% increase in suicidal events per additional month of treatment (HR=1.09, 95% CI = 1.07, 1.11). Parallel analyses in diazepam and lorazepam found comparably increased risks to alprazolam, whereas buspirone was associated with significant risk but of a magnitude less than the benzodiazepines. Buspirone demonstrated no increased risk in patients with a history of suicidal behavior.
Conclusions: The previous signal for alprazolam and increased suicidal event risk was confirmed and appears to extend to the benzodiazepine class. The anxiolytic buspirone was comparatively safer.
