Safety Profile of Lithium: A Disproportionality Analysis Using the FDA Adverse Event Reporting System Open Access

Bipolar disorder is a recurrent episodic psychiatric illness characterized by significant fluctuations in mood and energy over short periods [1]. Although its relatively low lifetime prevalence is approximately 2.4% [2], its considerable negative impact on physical and mental health ranks it among the top five psychiatric disorders in terms of disability-adjusted life years [3], imposing substantial economic and healthcare burdens globally [4].

For over 70 years, lithium has remained the gold standard for mood stabilization in the treatment of bipolar disorder [5, 6]. Despite its therapeutic mechanisms not being fully understood, its effects on cellular function, neurotransmission, and higher-order regulatory systems are well recognized [7, 8]. For instance, lithium competes with sodium channels at the cellular level, reducing elevated sodium levels and neuronal membrane potential in bipolar patients [9, 10]. In neurotransmission, lithium modulates excitatory neurotransmitters such as glutamate [11], dopamine [12], acetylcholine [13], and glycine [14], as well as inhibitory neurotransmitters like gamma-aminobutyric acid [15], thereby alleviating manic behavior while improving depressive mood. Recent studies have also linked lithium’s effects to changes in higher-order regulatory systems, including circadian rhythms [16], the hypothalamic-pituitary-adrenal axis [17], and the gut-brain axis [18]. Moreover, lithium’s anti-impulsive and anti-suicidal properties have demonstrated efficacy in other severe psychiatric disorders [19, 20], making it the most promising mood stabilizer in whole psychiatry [20].

However, despite its widespread recommendation as a first-line treatment in numerous clinical guidelines for bipolar disorder management [21‒23], the prescription rate of lithium among bipolar patients has nearly halved (from 30.4% to 17.6%) over the past two decades, during which outpatient visits for bipolar disorder increased from 12.4% to 51.4% [24]. This significant decline in lithium’s once-trusted status can be partially attributed to the expanded medication choices driven by the second-generation antipsychotics [24] but more fundamentally reflects ongoing concerns among clinicians and patients about its potential adverse events [25, 26]. Lithium has long been plagued by the “old kidney story” [27] and continues to face criticism for its potential to impair renal and thyroid function [28]. Recent studies have further highlighted its associations with several newly identified adverse events, including extrapyramidal symptoms, sexual dysfunction, psoriasis, and acne keloidalis nuchae [29, 30]. The persistent negative reputation is contributing to lithium becoming a “forgotten drug” in clinical practice [31].

However, the disagreement over the side effects of lithium persists. Some research emphasizes the need for vigilance against its wide-ranging adverse effects [32], while another study argues that many side effects are transient and mild for most patients [33]. For instance, several investigations propose that lithium may impair cognitive function by inducing neurotransmission abnormalities [34, 35], whereas other evidence indicates that cognitive function is preserved in lithium-treated patients [36]. Additionally, the well-known thyroid risks associated with lithium have also been challenged, with researchers noting that patients on lithium therapy typically undergo more than twice the number of thyroid tests compared to those receiving other treatments [37]. The substantial underdiagnosis of thyroid dysfunction in non-lithium patients may lead to an over-perceived thyroid risk of lithium [37]. Furthermore, age-specific differences in adverse events, the cumulative impact of polypharmacy on adverse event rates, and unaddressed confounding factors may contribute to an overestimation of lithium’s risks [32, 38].

This discrepancy between evidence-based recommendations and clinical practice underscores the necessity for a thorough assessment of this valuable medication’s safety profile. Therefore, this study aims to perform a pharmacovigilance assessment of lithium using disproportionality analysis based on real-world data from the FAERS. FAERS stores and manages tens of millions of post-marketing adverse event reports (AERs) and has been widely used to explore potential safety risks of medications [39]. Disproportionality analysis can identify drug risk signals by comparing the reporting proportion of a specific drug-adverse event pair to that of all drugs in the database for that adverse event, thereby transcending the limitations of clinical trials that are confined to limited adverse events [40]. The objectives of this study are to discuss the consistency of previously documented adverse events with real-world reporting data and to identify unlisted adverse events in prescribing information, thereby providing a more comprehensive understanding of lithium’s safety profile.

FAERS has been providing quarterly updated AERs since 2004 in compliance with International Conference on Harmonisation (ICH) E2B guideline. Given that lithium was approved for treatment in the USA in 1970 [6], this study extracted AERs in American Standard Code for Information Interchange (ASCII) format for all years and populations from the FAERS database, covering the period from the first quarter of 2004 to the fourth quarter of 2024.

The extracted dataset consists of seven sub-datasets, including drug, demographic (DEMO), reaction (REAC), outcome, report source, therapy, and indication. To enhance reliability and minimize interference, duplicate records were removed, retaining only the latest record for each case with the same identifier. The sub-datasets were then merged using the primary identifier (primaryid), and invalid entries (primaryid = 0) were eliminated.

FAERS has standardized the collected adverse drug events using Preferred Terms (PTs). In this study, PTs reported three or more times were included and further categorized according to System Organ Classes (SOCs) using the Medical Dictionary for Regulatory Activities (MedDRA) version 27.1. Drug names were standardized using the World Health Organization Drug Dictionary (September 2024), and AERs containing unidentifiable drug names were excluded.

In line with the study objectives, we extracted AERs where lithium was identified as the primary suspect drug, excluding cases where lithium was considered a secondary suspect or a concomitant drug. The detailed flowchart of data processing is presented in online supplementary material S1 (for all online suppl. material, see https://doi.org/10.1159/000546602).

Descriptive analysis was employed to illustrate the clinical characteristics of the AERs, including gender, age, reporter identity, reported country or region, route of administration, main indications, main combination medications, and severe clinical outcomes. Subsequently, four disproportionality analysis methods, including ROR, PRR, BCPNN, and EBGM, were applied for signal detection. ROR and PRR are frequency-based algorithms, whereas BCPNN and EBGM are categorized under Bayesian algorithms. The combined use of these algorithms can reduce false positives and enhance the accuracy of detecting rare adverse events. All analyses were executed based on the 2 × 2 contingency table (online suppl. material S2), with specific formulas and thresholds detailed in online supplementary material S3. In this study, adverse events with a reporting frequency of 3 or higher and meeting the thresholds were considered valid. Additionally, year-by-year signal detection was conducted as a sensitivity analysis to validate the robustness of the identified signals.

All signal analyses were performed using R software (version 4.3.2). The study findings are reported in accordance with the REporting of A Disproportionality analysis for drUg Safety signal detection using individual case safety reports in PharmacoVigilance (READUS-PV) guideline. The READUS-PV checklist is available in online supplementary material S4.

From the first quarter of 2004 to the fourth quarter of 2024, FAERS recorded a total of 18,082,548 AERs, of which 6,909 identified lithium as the primary suspect drug. The annual distribution of these reports is illustrated in online supplementary material S5.

Table 1 summarizes the clinical characteristics of these AERs. The number of AERs involving female patients (3,333, 48.24%) was higher than those involving male patients (2,598, 37.60%). Middle-aged individuals (45–65 years) comprised the largest proportion of the reports (2,185, 31.63%). The majority of AERs were submitted by medical professionals, such as physicians (1,576, 22.81%), pharmacists (1,618, 23.42%), and other health professionals (1,656, 23.97%). The primary source of the reports was the USA (2,579, 37.33%). The most common administration route was oral (2,363, 34.20%). The most frequently reported severe clinical outcomes were hospitalization (3,295, 40.44%), followed by death (535, 6.57%), and life-threatening conditions (413, 5.07%). In addition, lithium was most frequently used to treat bipolar disorder (2,990, 43.28%), followed by schizophrenia (347, 5.02%) and depression (328, 4.75%). The most common combination medications used with lithium included quetiapine (354, 5.12%), levothyroxine (337, 4.88%), clonazepam (245, 3.55%), risperidone (238, 3.44%), and olanzapine (216, 3.13%).

Table 2 presents the number of lithium-related adverse drug events across 24 SOCs, as well as the signal values derived from the four algorithms. Signals meeting the predefined thresholds for all four algorithms were identified in following SOCs: endocrine disorders (ROR = 12.02, PRR = 11.68, IC = 3.54, EBGM = 11.61), renal and urinary disorders (ROR = 4.22, PRR = 3.98, IC = 1.99, EBGM = 3.97), psychiatric disorders (ROR = 2.56, PRR = 2.34, IC = 1.23, EBGM = 2.34), and nervous system disorders (ROR = 2.49, PRR = 2.20, IC = 1.14, EBGM = 2.20).

Figure 1 illustrates the top 50 PTs with the strongest signals, ranked by ROR. These PTs are distributed across 11 SOCs, with 33 adverse events mentioned in the FDA’s lithium prescribing information, such as nephrogenic diabetes insipidus, neonatal toxicity, renal cyst, hyperparathyroidism, visuospatial deficit, and others.

Furthermore, 17 PTs not mentioned in the prescribing information were identified, such as renal oncocytoma, benign parathyroid tumor, cytotoxic lesions of corpus callosum, increased antipsychotic drug levels, and Adams-stokes syndrome. More detailed information is provided in online supplementary material S6.

To validate the robustness of results, year-by-year signal detection was performed on four detected PTs (two well-known in prescribing information and two unrecorded ones). As illustrated in online supplementary material S7, the lower bounds of the confidence intervals for the signal values of the documented nephrogenic diabetes insipidus and hyperparathyroidism remained above zero across all years. Similarly, the lower bounds of confidence intervals for the undocumented increased antipsychotic drug levels and delirium were greater than zero annually, confirming the temporal stability of these signals.

To our knowledge, this is the first study to perform a pharmacovigilance analysis of lithium using real-world data. Our findings indicate that despite a notable decrease in lithium prescription rates [24], the annual number of AERs recorded in FAERS has remained relatively stable at around 329 cases, suggesting a potential rise in the spontaneous reporting rate of adverse events. In terms of the clinical characteristics of the reports, the majority of AERs involve female and middle-aged patients, which may be linked with the higher sensitivity to lithium and decreased lithium clearance in these populations [38, 41].

Signal detection revealed the strongest signals in the endocrine system, with adverse events concentrated in the thyroid and parathyroid. Beyond the well-documented induction of hyperthyroidism, hypothyroidism, and hyperparathyroidism by lithium through interference with thyroid iodine uptake and parathyroid calcium-sensing receptors [32], this study also identified significant signals for thyrotoxic crisis and benign parathyroid tumor. These findings suggest that clinicians should not only monitor common functional abnormalities during lithium use but also remain vigilant for such severe complications. Caution is advised when prescribing lithium to patients with pre-existing thyroid disease or genetic predispositions [42]. For individuals with biochemical evidence of parathyroid dysfunction, parathyroid ultrasound examinations may be considered to evaluate for structural abnormalities [43]. Similarly, in the renal and urinary systems, we identified signals for common adverse events, such as nephrogenic diabetes insipidus, renal cysts, chronic glomerulonephritis, and less common events like renal oncocytoma. This is consistent with previous research indicating lithium exposure increases the risk of solid renal tumors [44]. Given that lithium has been found to increase the number of intercalated cells in the collecting ducts and activate proliferative signaling pathways [45], clinicians need to balance the benefits of mental illness control against the potential tumor risk in high-risk populations prone to lithium nephrotoxicity.

Furthermore, this study supports the association of lithium with up to 21 types of neurological and psychiatric disorders. Prior research has documented that lithium may lead to widespread demyelination in various parts of the peripheral and central nervous systems, as well as excessive gliosis [46]. For instance, demyelination in the cerebellum may result in the loss of Purkinje cells, which are crucial for motor coordination and balance control [47], while excessive gliosis can disrupt the normal structure and function of the brain [48]. These organic changes can lead to irreversible neurological and psychiatric disorders, warranting significant clinical attention [46].

Finally, this study identified risk signals for several cardiovascular disorders, including Adams-Stokes syndrome, acute cardiomyopathy, sinoatrial block, and sinus bradycardia. This may be attributed to lithium’s property of blocking cardiac sodium channels [49]. Lithium has been found to interfere with the normal transport of sodium ions in myocardial cells, affecting the depolarization and repolarization processes, thereby leading to sinoatrial node dysfunction and cardiac arrhythmias [49]. Existing evidence indicates that lithium can induce unstable electrocardiogram changes at both therapeutic and toxic levels [50]. In severe cases, this mechanism may induce critical conditions identified in this study, such as Adams-Stokes syndrome and acute cardiomyopathy. These findings highlight the need for regular electrocardiogram monitoring in patients on lithium therapy to closely track changes in heart rhythm.

The findings of this study should be interpreted with the following limitations. First, reliance on spontaneous reports in FAERS may introduce underreporting bias, especially if patients or healthcare providers perceive certain events as expected or trivial. This could potentially explain the absence of signals for minor gastrointestinal events such as nausea and vomiting. Additionally, disproportionality analysis can only provide associational inferences rather than causal ones. Further research should be designed to explore these associations in greater depth.

This study systematically evaluated the safety profile of lithium and identified several previously overlooked adverse events using real-world data from FAERS, thereby contributing to a more comprehensive understanding of its clinical risks.

The authors would like to express their gratitude to everyone who has submitted reports to the FAERS as well as to the staff members who are responsible for maintaining the system.

An ethics statement was not required for this study type as it is based exclusively on data extracted from FAERS database.

The authors have no conflicts of interest to declare.

The authors declare that no funds were received to assist with this study.

Conceptualization: Hao Zhu, Jack Guo, Hannah Lui, and Patrick Ip; methodology, formal analysis, and investigation: Hao Zhu and Hannah Lui; writing – original draft preparation: Hao Zhu; writing – review and editing: Hao Zhu, Jack Guo, and Patrick Ip; and supervision: Patrick Ip.

The data that supports the findings of this study are openly available in FAERS Website (https://fis.fda.gov/extensions/FPD-QDE-FAERS/FPD-QDE-FAERS.html). The sub-dataset generated and analyzed during the current study is available from the corresponding author upon reasonable request.