Lithium treatment extends human lifespan: findings from the UK Biobank

Introduction

The trace element lithium is a naturally occurring alkali metal. Owing to its mood-stabilizing properties [1, 2], lithium has been prescribed for more than 50 years to individuals affected by mood and affective disorders.

Lithium is the most effective maintenance treatment for bipolar disorders [3, 4] and alone or in conjunction with other psychotropic drugs, it can be used as an acute therapy for other mood and affective disorders (ICD10 F30-39), such as acute mania and bipolar depression [5, 6]. Since lithium requires constant monitoring to ensure that its plasma concentration is maintained between target clinical concentrations of 0.4-1mmol, in the current clinical UK guidelines for affective disorders [3], lithium is not recommended as the first-line acute treatment to mania or hypomania, while other anti-psychotic medications are preferred. While lithium efficacy in relieving affective disorders long-term and preventing suicides is well established [7], recent studies both in animal models and in humans have uncovered that lithium may have life-extending properties.

Lithium increases longevity in the nematode Caenorhabditis elegans, both at clinical concentrations [8] and in trace amounts [9], in the fruit fly Drosophila melanogaster [10], as well as in yeast (Schizosaccharomyces pombe) [11]. In Drosophila, lithium increases lifespan by inhibiting glycogen synthase kinase-3 (GSK-3) and activating nuclear factor erythroid 2-related factor (NRF-2) [10], suggesting that lithium regulates metabolism and stress resistance. When administered to mice, lithium given at concentrations comparable to the therapeutic ranges for the treatment of human bipolar disorder improves metabolic parameters without significantly extending lifespan [12, 13]. The broad effects of lithium on lifespan in several model organisms suggest that it exerts evolutionary conserved mechanisms to promote healthy aging.

Nutritionally ingested lithium, as present in plant-derived foods and drinking water, is readily bioavailable and evenly disseminated in every tissue, but its distribution in food and water is variable and region-dependent [14, 15]. Using geographic differences in lithium concentration as a natural experiment, studies in independent populations worldwide have demonstrated a positive correlation between trace amounts of lithium in drinking water and longevity [9, 16, 17]. The protective effects of higher dose of lithium intake on all-cause mortality have indirectly been evaluated in prospective intervention studies aimed at comparing the clinical efficacy and suicide rates of lithium on patients suffering from bipolar and related disorders [18, 19]. While highly informative, these studies include mostly middle-aged individuals (43.78 ± 12.94 years) and, more importantly, have a rather short intervention/observation period (2.52 ± 1.05 years). The demographics and observation time of these studies are therefore suboptimal to capture the life-extending properties of lithium in the aging population. In addition, accidental causes of deaths were not excluded from most of these studies potentially confounding the findings. Addressing these shortcomings, we leveraged data from the UK Biobank [20], a large well-characterized and long-term observational population cohort where age of recruitment is 56.41 + 8.09 years and follow-up time is 11.90 ± 1.24 years, to provide interventional evidence that lithium increases lifespan in humans.

Results

Mood disorders, which are classified in the International Statistical Classification of Diseases and Related Health Problems (10th Edition) (ICD10) with diagnosis codes F30-F39 (onwards globally referred to as F3X) [21], are disturbances of affect or mood and often lead to depression or elation. In the highly characterized and healthcare-linked UK Biobank cohort, we have analyzed National Health Services (NHS) clinical records of participants to identify individuals with at least one ICD10-defined F3X mood affective disorder diagnosis since 1990. F3X-diagnosed individuals constitute a demographics that is significantly different from the rest of the UK Biobank cohort: they are older, have a higher suicide rate, are more frequently affected by comorbidities such as diabetes, cardiovascular diseases and cancer (Table 1 [excluding suicides], Supplementary Table 1 [including suicides]), and, importantly, have a significantly lower lifespan than controls (Figure 1A, Supplementary Figure 1A). A Cox proportional hazard model shows that F3X diagnoses significantly contribute to shortened lifespan (Figure 1B, Supplementary Figure 1B). Analysis of the causes of death indicates that F3X-diagnosed individuals are more likely to die due to suicides, respiratory diseases, mental disorders, diseases of the nervous system, as well as COVID-19, while they are less likely to succumb from cancer (Supplementary Table 2).

Due to the detrimental and diagnosis-specific effects on lifespan of individuals with mood and affective disorders, and considering that all the patients that take lithium must have at least one F3X diagnosis, we evaluated differences in survival in individuals with F3X diagnoses that are either taking lithium or other psychotropic medications (Figure 2A). By analyzing NHS prescription data, we selected individuals that had taken either lithium, in the therapeutic formulation of lithium carbonate (including brand name equivalent compounds Priadel™, Camcolit™, and Liskorum™), or other anti-psychotic drugs (Supplementary Figure 2) for at least three consecutive months and within 6 months of the time of last recorded event (either death, or at the date of last censoring) (Figure 2A). Patients prescribed with lithium took the drug on average longer than individuals prescribed other anti-psychotic medications, had a longer interval between the first and last F3X diagnoses, and had more cumulative F3X episodes (Supplementary Figure 3B, 3C, 3E, 3F), while the age at onset of the diagnosis and episodes per year were unchanged (Supplementary Figure 3A–3D, 3F). These data support the current UK indications for lithium use in mood affective disorders [3]. Among the 348 lithium users, there were no suicide deaths, while in 1717 users of other psychotic drugs there were 5 deaths by suicides (Supplementary Table 3). Since deaths by suicides were not present among lithium treated individuals, only deaths not caused by accident nor self-inflicted causes were considered in subsequent analyses.

To ensure that the outcomes of lithium use on survival were not affected by confounding variables, such as baseline population characteristics, we used propensity score matching with nearest neighbor distance algorithm (Figure 2A) [22]. Propensity score matching (PSM) is a statistical matching method that allows to analyze an observational non-randomized study so that it mimics a randomized controlled trial in a post-hoc fashion [23, 24]. While in randomized control trials the propensity score, or probability to assign a treatment based on observed baseline covariates, is defined by the study design, in observational studies PSM allows to selected a posteriori individuals to include in control groups based on a distance algorithm that equalizes the values of a defined set of covariates in the treatment group [22]. In the current study, participants using lithium were outnumbered by users of other anti-psychotic drugs, therefore each lithium user was matched to two users of other anti-psychotic drugs for baseline traits including sex, age at recruitment, most common F3X diagnoses, other comorbidities and age at first treatment as a proxy of the length of treatment (Figure 2A). The resulting matched cohorts, composed of 276 lithium users and 552 users of other psychotropic medications, had identical distribution of the 5 most frequent F3X diagnoses (Figure 2B) and baseline characteristics (Table 2). By applying this widely used approach, we ensured that individuals in both groups were comparable in terms of comorbidities, F3X diagnoses, and length of drug use, and therefore the effects on lifespan mediated by the two treatment regimens could be compared excluding confounding effects [23, 24].

Lithium significantly extends the lifespan of its users compared to individuals taking other anti-psychotic medications (Figure 3A). Cox proportional hazard model shows that taking lithium is the covariate that most significantly contributes to increased longevity in the study (Figure 3B), with a logHR of −1.292 (−2.129067 to −0.455455 CI 95%), or hazard ratio of 0.274 (0.119–0.634 CI 95%) corresponding to 3.641 times lower (95% CI 1.577–8.407) chances of dying at a given age for lithium users compared to other anti-psychotic drug users. Characteristics such as male gender, having been diagnosed with diabetes, blood clots or cancer increase the hazard ratio, as previously shown [25, 26]. Interestingly, lithium significantly increases longevity both in males and female, therefore lithium exerts its geroprotective effect independently of sex (logrank female p-value: 0.0253; logrank male p-value: 0.0315) (Supplementary Figure 4A, 4B). Similar results were obtained when including also the deaths by suicides in the cohort taking other anti-psychotic drugs (Supplementary Figure 5A, 5B, Supplementary Table 4). Overall our data demonstrates in a large population-based aging observational cohort that lithium has life-extending properties. The causes of death between lithium users and other psychotropic drug users were not significantly different (Supplementary Table 5).

Discussion

This study is the first to evaluate the effects of lithium on lifespan in a large aging observational cohort. We have employed a state-of-the-art statistical algorithm, namely propensity score matching, to take advantage of the large amount of observational data in the UK Biobank and simulate a randomized control trial setting to evaluate whether lithium is a geroprotective agent in humans. We have unveiled that lithium at therapeutic concentrations is not only effective in ameliorating mental health, but it is also able to decrease mortality in individuals affected by F3X diagnoses, a demographics that would normally be more fragile and shorter-lived compared to healthy individuals, as also shown in this study. An alternative interpretation for these findings is that lithium in F3X-diagnosed patients has fewer lifespan-decreasing effects than other medications used to treat F3X diagnoses, and therefore it protects its users from side effects which may lead to premature death.

It needs to be noted that lithium is not prescribed in patients unless they suffer from a F3X diagnosis, i.e., there is an indication bias in the selection of individuals included in the analysis. This bias is unavoidable in observational studies, where drugs are administered to participants that have an underlying condition, and only preventable if designing an ad hoc clinical trial where effects of lithium on longevity would be evaluated exclusively in healthy individuals. Even considering these limitations, this study unambiguously revealed that lithium is the most significant factor in reducing mortality in patients with affective disorders, and encourages further investigations in healthy individuals.

Our findings are in line with previous works demonstrating that trace amounts of lithium in drinking water are positively correlated with longer lifespan [9, 17] and cognitive improvement [27], and support the idea that lithium could be used as a nutritional longevity supplement [16]. Lithium use in humans has been linked with molecular changes associated with longevity. Long-term lithium-treated bipolar disorder patients are protected from leukocyte telomere shortening [28], and lithium carbonate therapy has been associated with improved kidney function through inhibition of GSK3beta activity and decreased senescence signaling in urinary exfoliated podocytes [29]. In our analyses, due to the relatively limited size of the lithium cohort, we cannot evaluate if specific pathologies are delayed thanks to lithium, and larger cohorts should be used to pinpoint which biological processes mostly contribute to the lithium-mediated lengthening of lifespan in human cohorts. Therefore, the full extent of the molecular mechanisms of lithium-induced longevity in humans still remains elusive and should be further investigated.

Materials and Methods

Supplementary Materials

Supplementary Figures

Supplementary Tables