Document resource
Objectives Lithium is a first-line treatment option for preventing relapse in bipolar disorder but may be discontinued periconceptionally due to safety concerns. We aimed to examine the prevalence of lithium discontinuation among pregnant women with bipolar disorder in Sweden, using a novel method to identify daily dose, and to describe the characteristics of the discontinuers and continuers.Design Cohort study.Setting Swedish population-based registers (2006–2019).Participants Women with bipolar disorder and an ongoing lithium treatment episode 6 months prior to pregnancy were identified. Natural language processing (NLP) of the prescription free-text information was used to define the daily dose, and to construct treatment episodes.Interventions Not applicable.Main outcome measures Lithium discontinuation was defined as the end of a continuous treatment episode between 6 months before the start of pregnancy and delivery. The performance of the NLP method was compared with manual review of the prescription texts. The characteristics of the discontinuers and the continuers were described, including psychiatric comorbidities and other medications used.Findings In 662 pregnancies, 438 women (70%) discontinued lithium treatment, the majority in prepregnancy or first trimester. The NLP method showed high internal validity compared with manual review of prescription texts; however, the estimated date of discontinuation could not be validated against measures of actual drug use. Discontinuers, compared with continuers, had a higher prevalence of comorbid substance use disorder (7.8% vs 2.7%), personality disorder (11.4% vs 6.0%) and attention-deficit/hyperactivity disorder (13.5% vs 7.6%). After discontinuation, there was no increased use of other mood stabilisers, and 46% had a new dispensation of lithium within 6–12 months.Conclusions The majority of lithium-treated women with bipolar disorder discontinued treatment around the start of pregnancy; however, lithium discontinuation is also common outside pregnancy, and the reason for discontinuation in our cohort remains unknown. Discontinuers had a higher rate of psychiatric comorbidity, and switches to other mood stabilising treatments did not appear to be common. Given that both the peripartum period and withdrawal of mood stabilising treatment may contribute to increased risk of relapse, the consequences of lithium discontinuation for maternal psychiatric health should be studied further.