Association Between Lithium Use and Suicide‐Related Behaviors in Bipolar Disorder: A Multicenter Study

BACKGROUND: Lithium has long been associated with reduced suicide risk in bipolar disorder; however, evidence from real-world clinical populations remains susceptible to confounding, and the behavioral correlates related to this association are not well characterized. METHODS: In this multicenter cross-sectional study, 616 individuals aged 12-45 years with bipolar disorder were recruited from 14 psychiatric centers, of whom 585 had complete data for the present analyses. Lithium exposure within the past year was determined from clinical records. The primary outcome was a composite of suicide-related behaviors, including suicidal ideation, suicide attempts, and non-suicidal self-injury. A pre-specified analytical framework was applied, with multivariable logistic regression as the primary analysis and propensity score matching (PSM) as the principal approach to improve covariate balance. Regression-based decomposition analyses were conducted to examine whether aggression was statistically associated with the exposure-outcome relationship, and restricted cubic spline models were used to characterize the dose-response association between aggression and suicide-related behaviors. Robustness analyses included inverse probability of treatment weighting (IPTW), alternative matching specifications, E-value analyses, cluster-robust standard errors, and exploratory age-stratified analyses. RESULTS: Lithium exposure was associated with lower odds of suicide-related behaviors in both the fully adjusted model (OR = 0.55, 95% CI: 0.34-0.89) and the propensity score-matched sample (OR = 0.53, 95% CI: 0.33-0.86). Higher aggression scores were linearly associated with increased suicide-related risk. Exploratory regression-based decomposition analyses indicated that approximately 20.93% of the observed lithium-outcome association was statistically accounted for by aggression. Results from robustness analyses were directionally consistent with the primary findings. CONCLUSIONS: In this multicenter clinical sample, lithium exposure was consistently associated with lower odds of suicide-related behaviors across multiple analytic approaches. Aggression was independently associated with suicide-related risk and statistically accounted for part of the observed association between lithium exposure and suicide-related outcomes. However, causal inference is limited by the cross-sectional design, and residual confounding-particularly confounding by indication-remains a key concern. Prospective longitudinal studies are required to clarify temporal relationships and disentangle treatment effects from selection processes.