Lithium and risk of cardiovascular disease, dementia and venous thromboembolism

To determine if long‐term lithium treatment is associated with protective effects or increased risk of vascular, neurological, and renal disorders.


| INTRODUC TI ON
Lithium is a common and widely used mood-stabilizing agent for bipolar disorder. 1 In addition, lithium is frequently used in severe recurrent depression and schizoaffective disorder. These conditions are all associated with greatly increased morbidity and mortality, both with regard to somatic disease as well as suicide. [2][3][4] The most common cause of death in bipolar disorder is cardiovascular disease. 2,5 Furthermore, bipolar disorder appears to be associated with poorer long-term cognitive outcome and a higher risk of developing dementia. 6 A similar relationship seems to exist between depression and dementia, 7 as well as psychosis and dementia. 8 It is well known that lithium treatment is associated with multiple adverse side effects and somatic complications. However, over the last 20 years, there is growing evidence, from both animal and human studies, that lithium can have neuroprotective effects. For example, lithium treatment has been shown to have protective and antiapoptotic effects on hippocampal neurons after irradiation, and improves cognitive performance in mice. 9,10 In humans, lithium has been shown to be associated with higher brain gray matter volume, 11,12 and appears to have a positive effect on cognitive performance in Alzheimer's disease and mild cognitive impairment in individuals without psychiatric disorders. 13 Lithium has also been studied in both humans and rodents with regard to ischemic stroke with indications of neuroprotective effects. 14,15 A retrospective chart review investigating the effects of lithium treatment on several neurological and cardiovascular disorders, including 1028 adult psychiatric outpatients attending lithium clinics in New York City (USA) found a lower prevalence of dementia, seizures, and amyotrophic lateral sclerosis (ALS) in patients treated with lithium compared to patients of the same clinics who did not receive lithium. 16 A systematic review and meta-analysis by Velosa et al. published in 2020, included six studies on lithium and risk of dementia, of which four indicated that lithium has a protective effect on dementia development in some instances, e.g., continued use in elderly bipolar patients. 17 Lithium has also been proposed to reduce excess mortality from cardiovascular disease in affective disorders (n = 827 patients), 18 and was associated with a lower risk of myocardial infarction among the 1028 patients attending lithium clinics. 16 There is some indication that lithium may have beneficial effects on cardiac remodeling following ischemia. 19 A small study by Kallner et al. (n = 497 patients) published in 2000 suggested excess mortality from pulmonary embolism in lithium-treated individuals. 20 These findings have, to our knowledge, not been investigated further since.
All things considered, the relationship between long-term lithium treatment and health outcomes remains unclear; hence, large population-based studies are needed. In this nationwide registerbased study, we set out to examine if long-term lithium treatment is associated with a higher or lower risk of common disorders with a focus on vascular and neurological disorders, within the Finnish population.

| Study population and registers
This is a register-based cohort study. Cases were defined as all citizens and permanent residents in Finland that had at least one dispensation of prescribed lithium (ATC-code N05AN) for three or more consecutive years from January 1, 1995

| Outcomes and covariates
The cases and controls were followed up in HILMO until December 31, 2016 for somatic diagnosis outcomes. Outcomes were largely cardiovascular, cerebrovascular, and neurodegenerative disorders, as well as well-known lithium-associated disorders (kidney disease) and suicide. The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) diagnosis codes were used throughout the study as they were in routine use since 1996. 'Diagnosis at death' was taken from the Cause-of-Death Register and was included to address survival bias. Information on smoking (ICD-10 code F17) and alcohol abuse (F10), as well as well-known risk factors for cardio-and cerebrovascular disorders, were also obtained from HILMO and RRD: Hypertension (ICD-10 codes I10, I15 and/or ATC C10 dispensation), hypercholesterolemia (ICD-10 codes E78.0, E78.5 and/or ATC C07 dispensation), diabetes mellitus type 2 (ICD-10 codes E11, E14 and/or having special rights for medication reimbursement for the treatment of non-insulin-treated diabetes) and use of antipsychotic medication (ATC N05A excluding lithium). These risk factors were, together with sex, age, and municipality, used as covariates in different models.

| Statistical analysis
Using Cox proportional hazards modeling, the effect of lithium on All statistical analyses were performed using SAS, version 9.4 (SAS Institute Inc). Statistical significance was set at α = 0.01 in twosided tests.

| Ethical permission
The data protection authority and the authorities providing the data approved the study. Informed consent was not obtained, as this is not required for anonymous register data according to Finnish regulations. A personal identification number, issued to all Finnish citizens and permanent residents, was used to link and combine data from different registries and datasets. This was conducted as set out in the permission from the data-keeping organizations (THL/2218/5.05.00/2018, Kela 67/522/2018, VRK/5862/2018-1, TK-52-1109-19) after consulting the data protection authority.

| RE SULTS
A total of 9698 lithium-treated cases and 96,507 general population controls without lithium prescription were identified for the full cohort analyses. For the subcohort analyses of individuals with a mood or psychotic disorder diagnosis (ICD-10 F20-F29, F31-F33), a total of 8762 lithium-treated cases and 8786 controls were identified. The 936 lithium-treated cases not included in the subcohort were treated in primary care only and therefore had no diagnosis except diabetes in HILMO. The demographic characteristics are presented in Table 1, and the geographic origin of the subjects is represented in eFigure 1A-D. While the subjects were matched for age, sex, and municipality in the full cohort comparison, several differences were observed regarding diagnoses and cardiovascular risk factors, such as higher prevalence of psychiatric disorders overall, antipsychotic medication usage, kidney disorders, diabetes mellitus type 2, smoking and alcohol usage among the lithium-treated cases.
In the subcohort analyses, comparing lithium-treated cases to the group of controls with a diagnosis of mood or psychotic disorder, prevalence of antipsychotic usage (p = 0.001), diabetes mellitus type 2 (p < 0.0001), hypercholesteremia (p = 0.0004), and kidney disorders (p < 0.001) were higher amongst cases, whilst hypertension (p = 0.0889) and alcohol-related disorders (p = 0.246) were not statistically different in frequency compared to in controls. Amongst the lithium-treated cases, diagnoses of bipolar disorder and psychotic disorders were more common, whereas depressive disorders were in majority in the subcohort control group. In these subgroup analyses, the cases and controls were not individually matched, but no major differences were observed for age, sex, or municipality (Table 1, eFigure 1C,D).

| Effects of lithium treatment on disease compared to the general population
In the comparison between lithium users and individuals representing the general population, the lithium-treated cases had a higher somatic disease burden overall (Table 2, Figure 1

| Effects of lithium treatment on disease compared to the psychiatric control group
Comparing lithium users to individuals with a diagnosis of mood or psychotic disorder without lithium treatment, we found that lithium users had a lower risk for cardiovascular and cerebrovascular disor-  Figure 1, right panel). The hazard ratio for suicide showed no significant difference between lithium users and controls in this comparison (Figure 1, right panel). TA B L E 1 Demographic description of lithium-treated cases and population-based controls in full cohort matched for age, sex, and municipality, and in subcohort of individuals with a diagnosis of a mood or psychotic disorder (ICD-10 F20-F29, F31-F33) Note: Differences in the between cases and controls subcohort of individuals with a diagnosis of a mood or psychotic disorder were tested using twosided chi-squared tests. a Matching criteria: age, sex, and municipality, for geographical origin, see eFigure 1A-D.  It is well known that patients with severe mental illnesses such as bipolar disorder, major depression, and schizophrenia have a higher disease burden from cardiovascular disorders than the general population. 23 The results from our study support that this is true also for lithium-treated individuals in comparison with the general population.
However, when comparing lithium users to individuals having received a diagnosis of a mood or a psychotic disorder with no lithium use, lithium treatment was associated with lower cardio-and cerebrovascular disease risk, particularly with regards to chronic ischemic heart disease and cerebral infarction, pointing to a possible cerebrovascularand cardioprotective effect of lithium. These results remained largely unchanged in our different analysis models, for example, after correcting for cardiovascular risk factors and the use of antipsychotic medication. However, we cannot conclude whether this is a direct effect of lithium, or if long-term lithium use is a proxy-marker for a stable treatment regimen with high compliance, regular medical follow-up, or in other ways a proxy for an altogether different health behavior compared to other mood or psychotic disorder patients.
Contrary to the cardioprotective effects, we could not replicate the neuroprotective effect of lithium on dementia and ALS reported in earlier studies. 16 In contrast to the conclusion of the systemic review and meta-analysis by Velosa et al., 17 we found no indication that lithium treatment significantly reduces the risk of dementia. In our study, comparing lithium users to individuals with a diagnosis of mood or psychotic disorder without lithium treatment, the risk for dementia among lithium users showed no significant difference. However, the discrepancy of the diagnostics groups within the psychiatric subcohort may influence the results, as the risk of dementia in mood disorders seems to be associated with severe affective episodes 24 which are likely more frequent in the lithium group, that consists mainly of individuals with bipolar disorder. A recently published study by Chen et al. 25 including 548 cases exposed to lithium, concludes that lithium is associated with lower risk of dementia and its subtypes in older adults. Our results do not contradict this as we have not investigated different age groups, nor subtypes of dementia.
We also assessed the risk for thromboembolic events as psychiatric illness, psychotic, and mood disorders in particular, appear to be associated with a higher risk of venous thromboembolism (VTE). 26 Although psychotropic medication may play a role in this correlation, 27,28 to our knowledge, lithium has not been comprehensively studied in this context. In our study, the risk for VTE was significantly higher in lithium-treated individuals, in comparison with the general population, as well as in comparison with individuals having received a diagnosis of a mood or psychotic disorder without lithium medication. Previous research has linked major depression, bipolar disorder and schizophrenia with VTE. [29][30][31] The use of antidepressants has been suggested to be causal for venous thrombosis in this context, 32 and antipsychotics have also been linked to higher risk of VTE. 27,31 There are a few case reports on VTE in lithium-treated patients associated with lithium toxicity, 33,34 and case reports of lithium-induced nephrogenic diabetes insipidus contributing to the development of dural sinus 35 and superior sagittal sinus thrombosis, 36 respectively. Excess mortality from pulmonary embolism in lithium-treated patients has been previously described. 20 This possible association of lithium with VTE is not well understood, nor is it routinely considered in clinical practice. Overlapping risk factors for thrombosis with disease manifestations of bipolar disorder (e.g., immobility due to depression), consequences of acute management (e.g., use of physical restraints in mania 37 and/or use of antipsychotics 27 ), or properties of lithium specifically such as hematological effects of lithium, including effects of bone marrow hemopoietic stems cells, neutrophil, and platelet count, might all contribute. [38][39][40][41] Although this question remains complex, the findings in this study do further strengthen the evidence of an association of venous thromboembolic disease with lithium. Further investigation on this clinically relevant question is however needed.
A four-fold higher risk of suicide was seen in our cases compared to the general population. This is expected as this represents the over-risk associated with psychiatric illness. In comparison with individuals diagnosed with mood or psychotic disorders without lithium treatment, there was no over-risk for suicide. This is particularly remarkable as more severe diagnoses like bipolar disorder and schizophrenia, which both carry a substantially higher risk of suicide, [42][43][44] were over-represented among the cases in the subcohort (Table 1).
This result may therefore be attributed to the probable antisuicidal effect of lithium 45,46 compensating for the difference in diagnoses.
As expected, the HR for kidney disease was significantly higher in the lithium-treated group both in comparison to the general population and to individuals having received a diagnosis of a mood or psychotic without lithium treatment. It is a well-known fact is that lithium can cause nephrogenic diabetes insipidus, reduced glomerular filtration, and chronic kidney disease, but the issue of lithium's ability to cause end-stage kidney disease is a more complex question. 47 This study supports the notion that lithium is associated with and likely causal of different types of kidney disease to a high degree. However, due to the limited sample size of chronic kidney disease diagnoses in this study, a sub-analysis of different stages of chronic kidney disease, including end-stage kidney disease, was not possible. Important to note, however, is that due to close monitoring in lithium patients, there is an obvious risk of surveillance bias in terms of kidney disease.
Bipolar disorder has been shown to be associated with an increased risk of developing Parkinson's disease. 48 previous studies, 50 lithium maintenance treatment was defined as at least one lithium prescription/year continuously for three or more years, which represents a significant treatment time.
There are, however, limitations to this study. There is a lack of detailed information regarding the time of receiving lithium treatment before the follow-up period; therefore, it was not possible in this study to analyze the impact of time (on lithium treatment). Further, in the comparison to the general population, there is an obvious and nearly complete overlap between psychiatric diagnosis and lithium treatment, making it difficult to distinguish between the effects of psychiatric illness and that of lithium treatment. However, these results are important as they illustrate that the potential protective effects of lithium cannot compensate for the heavy disease burden associated with psychiatric illness in general, and bipolar disorder in particular.
The subcohort analyses allowed us to more specifically look at lithium effects within populations with a hypothesized similar baseline health and somatic disease risk. It is important to note though, the discrepancy in psychiatric diagnoses between cases and controls; within the lithium-treated group, patients with bipolar disorder were in majority, followed by individuals with a diagnosis of a psychotic disorder, whereas in the subcohort control group, diagnoses of depressive episode or depressive disorder were by far the most common. It was not possible to match the psychiatric diagnoses entirely because the absolute majority of bipolar patients were prescribed lithium. While this dissimilarity might influence the results, it would be expected that patients with bipolar or psychotic disorder have more severe comorbidities. Therefore, our results, support the idea that lithium treatment is associated with reduced somatic morbidity. However, no conclusion on causality can be drawn, nor can we say whether these effects are exerted by direct biological effects of lithium or if long-term lithium treatment is associated with, or a proxy-marker for, a stable treatment regimen with high compliance and regular medical follow-ups.
There is a risk of multiple testing in this study due to investigating association with several disorders. Therefore, we chose a more stringent significance threshold and report 99% CIs to reduce the risk of false positives. 16 Alcohol use and smoking are common and well-known risk factors associated with somatic disorders as well as mood and psychotic disorders. However, these conditions are likely to be underreported in clinical settings and probably even less often coded with their ICD-codes, especially in the large control group from the general population and therefore these variables were not used in this study. Furthermore, there are more unmeasured variables known to be associated with psychiatric illnesses influencing the incidence of somatic disorders, such as food behavior patterns, obesity, physical activity as well as socioeconomic factors which may therefore act as confounders.
In conclusion, lithium use was associated with a lower risk of cardiovascular and cerebrovascular disease, but shows no marked effect on dementia in individuals with a mood or psychotic disorder diagnosis. Venous thromboembolism, Parkinson's disease, and kidney disease were significantly more prevalent in individuals prescribed lithium, in comparison with both the general population and individuals with a similar psychiatric background. The association of VTE and lithium treatment constitutes an important area of further research. Overall, while supporting the potential beneficial effects of lithium treatment on cardiovascular and cerebrovascular health in individuals with a mood or psychotic disorder diagnosis, we also provide evidence that this effect does not compensate for the higher somatic disease burden that is associated with psychiatric illness.

CO N FLI C T O F I NTE R E S T
The authors of this paper do not have any commercial or other association that constitutes a conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from