IMPORTANCE There is interest in reducing long-term benzodiazepine prescribing given harms associated with use, but the cumulative risks or benefits of discontinuation are unknown.

OBJECTIVE To identify the association of benzodiazepine discontinuation with mortality and other adverse events among patients prescribed stable long-term benzodiazepine therapy, stratified by baseline opioid exposure.

DESIGN, SETTING, AND PARTICIPANTS This comparative effectiveness study with a trial emulation approach included data from a US commercial insurance database between January 1, 2013, and December 31, 2017. Eligible participants were adults with stable long-term benzodiazepine prescription treatment. Data were analyzed between December 2022 and November 2023.

EXPOSURE Benzodiazepine discontinuation, defined as no benzodiazepine prescription coverage for 31 consecutive days identified during a 6-month grace period after baseline.

MAIN OUTCOME AND MEASURES Mortality during 12 months of follow-up; secondary outcomes included nonfatal overdose, suicide attempt or self-inflicted injury, suicidal ideation, and emergency department use, identified in medical claims. Inverse probability weighting was used to adjust for baseline confounders that potentially affected treatment assignment and censoring due to death or disenrollment. Primary analysis used an intention-to-treat approach; a secondary per-protocol analysis estimated associations after accounting for nonadherence. Analyses were stratified by opioid use.

RESULTS The study included 213 011 (mean [SD] age, 61.1 [13.2] years) patients with stable long-term benzodiazepine use without and with opioid exposure, respectively. Among the nonopioid exposed, the adjusted cumulative incidence of death after 1 year was 5.5% for discontinuers, an absolute risk difference of 2.1 percentage points higher than for nondiscontinuers. The mortality risk was 1.6 times that of nondiscontinuers. Among those with opioid exposure, the adjusted cumulative incidence of death was 6.3% for discontinuers, an absolute risk difference of 2.4 percentage points (95% CI, 2.2-2.7 percentage points) higher than for nondiscontinuers and a mortality risk 1.6 times that of nondiscontinuers. Cumulative incidence of secondary outcomes was also higher among discontinuers.

CONCLUSIONS AND RELEVANCE This study identifies small absolute increases in risk of harms among patients with stable long-term prescription benzodiazepine treatment who appear to discontinue relative to continuing treatment, including those with and without recent prescription opioid exposure. Policy broadly promoting benzodiazepine discontinuation may have unintended risks.