Publication Date


Document Type

Doctoral Dissertation

Academic Program

Millennium PhD


Quantitative Health Sciences

First Thesis Advisor

Jeroan Allison, MD, MPH


Suicide, Bipolar Disorder, Lithium, Valproate, Veterans Health


Dissertations, UMMS; Suicide; Bipolar Disorder; Lithium; Valproate; Veterans Health


Background: The mood stabilizer lithium has long been reported to be associated with reduced suicide risks, but many studies reporting associations between lithium and reduced suicide risks also have been nonrandomized and lacked adjustment for many potential confounders, active controls, uniform follow-up, or intent-to-treat samples. Concerns also have been raised that medications being considered as potential suicide preventative might increase risks of nonsuicide mortality while reducing risks of suicide.

Methods: Three studies of Veterans Health Administration (VHA) patients were conducted combining high-dimensional propensity score matching with intent-to-treat analyses to examine the associations between lithium and valproate and one-year suicide and nonsuicide mortality outcomes.

Results: In intention-to-treat analyses, initiation of lithium, compared to valproate, was associated with increased suicide mortality over 0-365 days among patients with bipolar disorder (Hazard Ratio (HR) 1.50 [95% Confidence Interval 1.05, 2.15]) Nonsuicide mortality among VHA patients with or without bipolar disorder was not significantly associated with the initiation of lithium compared to valproate ( HR 0.92 [0.82-1.04]). Rates of treatment discontinuation, however, were very high (≈ 92%). Longitudinal analyses revealed that the increased suicide risks associated with initiating lithium among patients with bipolar disorder occurred exclusively after discontinuation of lithium vii treatment. In secondary analyses restricted to patients still receiving their initial treatment, there was no difference in suicide risk between the initiation of lithium or valproate.

Conclusions: Significantly increased risks of suicide were observed at one year among VHA patients with bipolar disorder initiating lithium compared to valproate, related to risks observed after the discontinuation of lithium treatment Since these studies are nonrandomized, confounding may account for some or all of our findings, including the risks observed after lithium discontinuation. Nevertheless, these results suggest that health systems and providers consider steps to minimize any potential lithium discontinuation-associated risk. Approaches might include educating patients about possible risks associated with discontinuation and closely monitoring patients after discontinuation if feasible. Given the obvious importance of any substantive difference between lithium and valproate in suicide or nonsuicide mortality risk, our studies also suggest that further research is needed, especially research that can further minimize the potential for confounding.



Rights and Permissions

Copyright is held by the author, with all rights reserved.