OBJECTIVES: Insomnia and the medications used to manage it are both associated with suicide risk, but the comparative safety of agents is unkown. The objective of this study is to estimate the comparative risk of death by suicide after first exposure by insomnia drug class. METHODS: Medications with fills totaling =>1% of all Department of Veterans Affairs (VA) insomnia fills (2005-2015) included: hypnotics (zolpidem), serotonin antagonist and reuptake inhibitors (SARIs) (trazodone), first generation antihistamines (hydrodxyzine, diphenhydramine), benzodiazepines (lorazepam, diazepam, temazepam, alprazolam), tetracyclic antidepressants (mirtazapine), tricyclic antidepressants (TCAs) (amitriptyline, doxepin), and non-benzodiazepine anxiolytics (buspirone). More than 2 million patients filled an insomnia monotherapy in 2005-2015 after 12 months of no use. Inpatient, outpatient and medication records were linked at the patient level. Multinomial logistic regression was used to estimate propensity scores. Inverse probability weighting was used to estimate average treatment effects. Extreme values (1^st, 99^th percentiles) were trimmed to generate stabilized weights. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using Cox proportional hazards regression adjusted for length of the first continuous days’ supply and covariates: inpatient days, Gagne comorbidity score, diagnoses (psychiatric, substance use, pain or sleep-related), fills for central nervous system or analgesic drug classes and demographics. RESULTS: Compared to other classes, TCAs had HRs for suicide significantly below 1. The lowest was for TCAs compared to tetracyclic antidepressants (mirtazapine) (HR 0.22 (95% CI 0.12-0.41)). Tetracyclic antidepressants (mirtazapine) compared to SARIs (trazodone) and hypnotics (zolpidem), respectively, were significantly above 1 (HR 1.65 (95% CI 1.20-2.28) and HR 1.62 (95% CI 1.10-2.38)). CONCLUSIONS: In the VA (2005-2015), insomnia medication classes varied widely by risk of death by suicide after adjusting for covariates. TCAs were associated with significantly reduced risk compared to each alternative. Tetracyclic antidepressants were significantly higher risk compared to SARIs and sedative-hypnotics.