Association between mood stabilizers and hypothyroidism in patients with bipolar disorders: a nested, matched case-control study
Susan Shur-Fen Gau, M.D., Ph.D.
Department of Psychiatry
National Taiwan University Hospital
No. 7, Chung-Shan South Road
Taipei 10002, Taiwan
Fax: +886-2-23812408
E-mail: gaushufe@ntu.edu.tw
Abstract
Gau C-S, Chang C-J, Tsai F-J, Chao P-F, Gau SS-F. Association between mood stabilizers and hypothyroidism in patients with bipolar disorders: a nested, matched case-control study.
Bipolar Disord 2010: 12: 253–263. © 2010 The Authors.
Journal compilation © 2010 John Wiley & Sons A/S.
Objectives: This study investigated whether lithium, carbamazepine, and valproate increased the risk for hypothyroidism using Taiwan’s National Health Insurance Dataset.
Methods: The sample included 557 bipolar disorder patients with incident hypothyroidism first diagnosed between 1998 and 2004, and 2,228 sex-, age-, and index date-matched bipolar disorder patients without hypothyroidism from 1996–2004. We compared the use of lithium, carbamazepine, and valproate before the onset of hypothyroidism between the two groups using a conditional logistical regression model.
Results: Compared with patients who had never used any of the three mood stabilizers, patients were more likely to have hypothyroidism if they only used carbamazepine [odds ratio (OR) = 1.68; 95% confidence interval (CI): 1.07–2.65]; or comedication of lithium and valproate (OR = 2.40; 95% CI: 1.70–3.40), lithium and carbamazepine (OR = 1.52; 95% CI: 1.10–2.08), and three mood stabilizers (OR = 2.34; 95% CI: 1.68–3.25). There was a dose-response relationship between the number of mood stabilizers and risk for hypothyroidism (OR = 1.34, 95% CI: 1.21–1.49) and a significant interaction between lithium and valproate on the risk for hypothyroidism (p = 0.020).
Conclusions: Our findings indicate that lithium, carbamazepine, and valproate may increase the risk for hypothyroidism, particularly if combined, and suggest regular monitoring of thyroid function and monotherapy of mood stabilizers for treating patients with bipolar disorders.
Bipolar disorders are common mental disorders with a lifetime prevalence ranging from 0.24% to 1.0% for bipolar I disorder and from 0.3% to 1.1% for bipolar II disorder in the general population (1–3). Intensive treatment is usually needed for patients with bipolar disorders because of functional impairments in several domains. Mood stabilizers, such as lithium, valproate, and carbamazepine, are effective regimens in acute and maintenance treatment (4–11), and lamotrigine prevents both manic and depressive episodes in patients with bipolar disorders (4).
Hypothyroidism is classified as congenital or acquired, primary or secondary, and overt/clinical or mild/subclinical, according to its onset, temporality, etiological region, and severity, respectively (12). The prevalence of hypothyroidism in the general population was reported to range from 2.18% to 2.98% in an earlier study (13) and to be 3.7% (0.3% overt and 3.4% mild) in a recent survey (14). The incidence of hypothyroidism was in the range of 1.35–3.82 per 10,000 person-years in Demark (15–17), in the range of 3.90–4.89 per 1,000 female-years and 0.65–1.01 per 1,000 male-years in the UK (18), and in the range of 0.2–0.5% in China (19).
Numerous studies have provided evidence to support the association between lithium and hypothyroidism (20–22). An annual rate of subclinical hypothyroidism in lithium-treated patients was estimated to be 1.7% from a 10-year longitudinal study (21). The prevalences of overt and subclinical hypothyroidism in lithium-treated patients were reported to be 8-19% and 23%, respectively (20), higher in females (23–26), and particularly high in the presence of thyroid autoimmunity (21). A recent study in the Chinese population also demonstrated a higher rate of hypothyroidism in lithium-treated (28–32%) than in lithium-free (6.3–10.8%) bipolar disorder patients (22).
Increased risk for hypothyroidism in patients with bipolar disorders may be related to the disease itself or to the medications used in these patients. Valle et al. (27) found that 9.2% of patients with bipolar disorders who had never used lithium or carbamazepine had higher rates of subclinical hypothyroidism than the general population. The study of Post et al. (28) did not support an association between hypothyroidism and medication-free rapid-cycling bipolar disorder patients. The possible mechanisms for lithium-induced hypothyroidism include: (i) lithium may suppress thyroid-stimulating hormone (TSH)-induced iodide uptake, and hence reduce iodine organification and de novo thyroid hormone formation (29); (ii) lithium may stabilize thyroid microtubules and subsequently affect thyroid hormone release (30); and (iii) lithium may inhibit thyroxine transform to triodothyronine in the periphery and within neurons (31).
In contrast to lithium, several studies provide only weak evidence of an association between carbamazepine and valproate and hypothyroidism in patients with epilepsy (32–37). Some studies have shown that carbamazepine-treated patients with epilepsy have lower total and free thyroxine levels but that carbamazepine has either no effect on TSH (32–36) or only induces a mild increase in TSH (37). Despite a decrease in thyroxin levels, there are no clinical signs of hypothyroidism in carbamazepine-treated patients, and the changes in serum thyroid hormone are reversible after withdrawal of carbamazepine (32).
The findings regarding valproate are even more controversial. Some studies have suggested no influence by valproate on serum thyroid hormone levels (32, 34, 36) but rather an increase of TSH levels (32, 37) and a reduction in thyroxine level (37). However, two pediatric studies have reported that the combined use of valproate and carbamazepine may increase the risk for hypothyroidism (34, 38).
Although previous studies have consistently reported lithium-induced hypothyroidism in patients with bipolar disorders (20–22), the effects of carbamazepine and valproate on the risk for hypothyroidism have not been well studied (32–37) and have never been studied in patients with bipolar disorders in a non-Western population. Moreover, the majority of previous studies are limited by small sample size and hospital-based samples. In view of this, we conducted this prospective population-based study to investigate the individual and combined use of lithium, carbamazepine, and valproate before the onset of hypothyroidism, and to test whether the number of mood stabilizers used is associated with the risk for hypothyroidism in a dose-response linear relationship among patients with bipolar disorders. Lamotrigine was not included in the main exposures because it had just been approved for treatment for bipolar disorders in Taiwan at the end of 2004, close to the study endpoint.
Patients and methods
Study design
This was a nested, matched, case-control study using the National Health Insurance Research Database (NHIRD).
Data source
In 1996, the reimbursement claims data for the National Health Insurance (NHI) in Taiwan had been transferred to and managed by the National Health Research Institute (NHRI), Taiwan in order to establish a medical claims database, i.e., the NHIRD, for research purposes only. The NHIRD contains medical claim data of about 22.8 million beneficiaries; these data include patients’ demographics, disease diagnosis, contracted medical care institutions, and medical expenditure and prescription claims data. For each medical expenditure reimbursement (both outpatient and inpatient), types of medical services, details of medical orders, and costs are recorded. Information that could be used to identify beneficiaries and medical care providers (institutions and medical professionals) is scrambled by the Bureau of National Health Insurance (BNHI) before being transferred to the NHRI (39). All the research protocols are preapproved by the NHRI and investigators need to sign an agreement that guarantees patient confidentiality before using the NHIRD.
Psychiatric Inpatient Medical Claims (PIMC) dataset
A cohort of patients with at least one psychiatric hospitalization record and one discharge diagnosis of psychiatric disorder coded by International Classification of Diseases, 9th Revision, clinical modification (ICD-9-CM) codes 290–319 or A-codes 210–219 (a simplified coding system used mainly for ambulatory care before 2000) from 1996 to 2001 was created from the NHIRD, which included a total of 91,104 NHI beneficiaries. In 2003, experts in the NHRI retrieved all medical records of these patients from the NHIRD to construct a subset database, the Psychiatric Inpatient Medical Claims (PIMC). The description and validation of the PIMC has been described in details elsewhere (40). The claim data and prescription files of this cohort were initiated in 1997 and extended to 2004 for this study.
The PIMC dataset obtained from the NHRI was further refined to eliminate patients with a scrambled ID not in the correct format or with missing data (n = 1,240; 1.4%). For those under the same scrambled ID but showing more than one birthday in the database, the birthdays of these patients were verified by comparing them with other databases, such as the Registry Files and the Catastrophic Illness Registry Files for Beneficiaries, to make any necessary corrections. Children under five years of age with pervasive developmental disorders were excluded from this study. Finally, the claims data of 89,618 beneficiaries (98.4% of the original PIMC), referred to as the refined PIMC, was used in this study (40). The details of the data refinement procedure is shown in Figure 1 and described elsewhere (40).
Flow chart of the dataset of patients with bipolar disorder derived from the Psychiatric Inpatient Medical Claims (PIMC) dataset.
Study subjects
We selected 12,424 patients with a new diagnosis of ICD-9-CM codes 296.0, 296.1, 296.4, 296.5, 296.6, 296.7, 296.80, 296.81, or 296.89 from the refined-PIMC database (Fig. 1). The medical records for these screened patients during 1996 and 2004 were then retrieved to construct a bipolar disorders subset database for this study (40). Figure 2 presents the enrollment procedure of the case and control groups.
Flow chart of sampling procedure of the case and control groups.
Case group. We identified bipolar disorder patients with incident hypothyroidism from the bipolar disorders subset database. The inclusion criteria were: (i) ICD-9-CM codes 244, 244.3, 244.8, or 244.9 that appeared in the first five diagnoses of patients’ inpatient care data or the first three diagnoses of outpatient care data and (ii) prescriptions of thyroxine, liothyronine, or thyroid hormone at the same time. A total of 717 patients met the inclusion criteria during the years 1998–2004. There were also 160 patients who met these inclusion criteria during the years 1996 and 1997, but they were not included in this study. Accordingly, the case group consisted of 557 patients (425 females, 76.3%; mean age ± SD: 41.3 ± 13.7) with incident hypothyroidism (Fig. 2).
Control group. The control subjects were selected from patients without a record related to hypothyroidism from 1996–2004 in the bipolar disorders dataset (n = 12,424). We selected control subjects randomly by matching sex, age, and the index date of the subjects in the case group at a ratio of one case to four controls. Accordingly, the control group included 2,228 subjects. All the medical records of 2,785 subjects were then retrieved from the bipolar disorder database.
Definition of variables
The index date was the date when the case subjects were diagnosed as having hypothyroidism (ICD-9-CM code 244, 244.3, 244.8, or 244.9) as indicated for the first time in the claim medical records. The index date of the controls was the same as that of the corresponding cases. The age variable was defined by the subject’s age on the index date.
Exposure variables and covariables
Exposure variables and covariables identified from the medical records based on the ICD-9-CM code and medical expenditure and prescription claims data were retrieved only when they existed before the index date. The exposure variables included use of one or more of the following mood stabilizers: lithium, carbamazepine, and valproate. The number of mood stabilizers was also analyzed as a count exposure variable by treating the model as a first-order model, i.e., a linear trend model, after testing for lack of fit of the first-order model. The lack-of-fit test was applied because there were replicate observations among the subjects (i.e., 0, 1, 2, or 3). We found there was no significant difference (p > 0.05) between the one under primary consideration (first-order model) and the more complex one (i.e., model including second order of number of mood stabilizers).
The covariables included: (i) systemic disease related to lithium, carbamazepine, or valproate: ICD-9-CM code for liver disease, renal disease, and/or epilepsy; (ii) medication related to hypothyroidism: amiodarone, phenytoin, interferon, and antithyroid agents (carbimazole, methimazole, propylthiouracil); and (iii) antidepressants: antidepressants available in Taiwan (amitriptyline, bupropion, citalopram, clomipramine, doxepin, fluoxetine, fluvoxamine, imipramine, maprotiline, mirtazapine, moclobemide, paroxetine, sertraline, trazodone, venlafaxine).
Data analysis
Statistical analyses were performed using SAS 9.1 (SAS Institute, Inc., Cary, NC, USA). Descriptive statistics were first used to present the distribution of demographic characteristics, physical illness, and medication related to hypothyroidism in the cases and controls. Since this was a nested, matched case-control study, a conditional logistic regression was employed to assess the risk for hypothyroidism using the PROC PHREG procedure and to calculate the odds ratio (OR) and 95% confidence interval (CI) of incident hypothyroidism for each exposure and covariable with and without controlling for other covariables, which showed significant associations with incident hypothyroidism (case/control status) in the univariate analysis of each covariable. To test whether the effect of mood stabilizers varied across age, sex, and use of the other mood stabilizers, multiple conditional logistic regression analysis was then exercised to examine the significant two- and three-way interactive effects of age and sex with each of the mood stabilizers separately, as well as all possible interactions among the mood stabilizers. Both two-way and three-way interactions were evaluated with the caveat that with the sample size there is not sufficient power to assess such interactions, especially three-way interactions. Preselected alpha level was p < 0.05.
Results
Thyroid disease, systemic disease, and comedications
There were 557 incident cases of hypothyroidism in years 1998∼2004 among 12,264 subjects at risk for hypothyroidism. The incidence of hypothyroidism among bipolar disorder patients was 65 per 10,000 person-years. Table 1 presents the distribution of systemic disease and comedications for both groups before the onset of hypothyroidism. The bipolar disorder patients with incident hypothyroidism were more likely than the controls to suffer from liver and renal disease and to have taken comedications comprising antithyroid agents and antidepressants before the onset of hypothyroidism. There were no differences in the rates of epilepsy and comedications with amiodarone and phenytoin between the two groups.
| Case group (n = 557) | Control group (n = 2,228) | Odds ratio (95% CI) | |
|---|---|---|---|
| Age (mean ± SD) | 41.3 ± 13.7 | 41.2 ± 13.7 | |
| Gender, n (%) | |||
| Female | 425 (76.3) | 1700 (76.3) | |
| Male | 132 (23.7) | 528 (23.7) | |
| Systemic disease, n (%) | |||
| Liver disease | 184 (33.0) | 599 (26.9) | 1.36 (1.11–1.67)a |
| Renal disease | 97 (17.4) | 243 (10.9) | 1.74 (1.34–2.26)b |
| Epilepsy | 25 (4.5) | 136 (6.1) | 0.72 (0.46–1.12) |
| Comedications, n (%) | |||
| Amiodarone | 4 (0.7) | 10 (0.5) | 1.60 (0.50–5.12) |
| Phenytoin | 16 (2.9) | 55 (2.5) | 1.17 (0.66–2.07) |
| Antithyroid agents | 93 (16.7) | 84 (3.8) | 5.08 (3.70–6.97)b |
| Antidepressants | 401 (72.0) | 1280 (57.5) | 1.98 (1.61–2.44)b |
- CI = confidence interval.
- ap < 0.01; bp < 0.001.
Effect of individual mood stabilizers
Table 2 summarizes the rates of mood stabilizer use for the two groups according to the following three usage definitions: had ever been prescribed at least once, seven or more defined daily doses (DDDs) in one prescription, and seven or more DDDs in cumulated prescriptions of lithium, carbamazepine, and valproate. Lithium was used the most, followed by carbamazepine and valproate in both cases (bipolar disorder patients with incident hypothyroidism) and controls (bipolar disorder patients without hypothyroidism). Regardless of which definition of medication use was employed in the analysis, the cases were more likely than the controls to use lithium, carbamazepine, and valproate before the onset of hypothyroidism. We further examined the interaction between age and gender and the use of three mood stabilizers on the risk for hypothyroidism by only presenting the results using the definition ‘use of mood stabilizer’ as having seven or more DDDs of any mood stabilizers in one prescription (Table 2). We found a significant gender effect with greater magnitude of association between valproate and hypothyroidism in males than females (β = 0.82, chi-square = 10.50, p = 0.001). There were no other modifying effects from age or sex on the associations between the three mood stabilizers and hypothyroidism.
| Mood stabilizer | Case group (n = 557) | Control group (n = 2,228) | Odds ratio (95% CI)a |
|---|---|---|---|
| Prescribed at least once, n (%) | |||
| Lithium | 401 (72.0) | 1448 (65.0) | 1.41 (1.14–1.74)c |
| Carbamazepine | 271 (48.7) | 922 (41.4) | 1.37 (1.13–1.65)c |
| Valproate | 244 (43.8) | 730 (32.8) | 1.72 (1.40–2.11)d |
| Prescribed for at least 7 DDDs in one prescription, n (%) | |||
| Lithium | 384 (68.9) | 1382 (62.0) | 1.38 (1.13–1.69)c |
| Carbamazepineb | 197 (35.4) | 664 (29.8) | 1.30 (1.07–1.59)c |
| Valproate | 201 (36.1) | 571 (25.6) | 1.75 (1.41–2.16)d |
| Prescribed for at least 7 DDDs in combined prescriptions, n (%) | |||
| Lithium | 390 (70.0) | 1406 (63.1) | 1.39 (1.13–1.70)c |
| Carbamazepine | 229 (41.1) | 761 (34.2) | 1.37 (1.12–1.66)c |
| Valproate | 226 (40.6) | 646 (29.0) | 1.79 (1.45–2.20)d |
- DDD = defined daily dose; CI = confidence interval.
- aAdjusted for covariables associated with hypothyroidism and/or with use of mood stabilizers (i.e., thyroiditis, thyroid surgery, goiter, thyrotoxicosis, renal disease, liver disease, epilepsy, phenytoin, antithyroid agents, and other thyroid disease) using multivariate conditional logistical regression.
- bInteraction between sex (male versus female) and valproate on the risk for hypothyroidism: regression coefficient = 0.82, chi-square = 10.496, p = 0.0012.
- cp < 0.01.
- dp < 0.001.
Different combinations of mood stabilizer usage
Table 3 summarizes the ORs and 95% CIs of monotherapy and combined mood stabilizer usage relative to not using mood stabilizers (as a reference group) with and without further controlling for the covariables, which showed significant differences between the cases and controls on association with use of mood stabilizers in univariate covariable analyses (i.e., thyroiditis, thyroid surgery, goiter, thyrotoxicosis, renal disease, liver disease, epilepsy, phenytoin, antithyroid agents, and other thyroid disease). Univariate analysis revealed greater likelihood of combined use of lithium and valproate or all three mood stabilizers in the cases than the controls. The statistical significance of these findings remained after controlling for physical illness and comedications.
| Usage pattern | Case group (n = 557) | Control group (n = 2,228) | Odds ratio (95%CI) | |
|---|---|---|---|---|
| Crude | Adjusteda | |||
| No mood stabilizers | 92 (16.5) | 517 (23.2) | 1.00 | |
| Use of one mood stabilizer only | 135 (24.2) | 661 (29.7) | 1.21 (0.90–1.63) | 1.11 (0.80–1.54) |
| Lithium | 86 (15.4) | 453 (20.3) | 1.03 (0.70–1.52) | 1.26 (0.76–2.10) |
| Valproate | 17 (3.1) | 94 (4.2) | 1.26 (0.53–2.98) | 1.64 (0.48–5.59) |
| Carbamazepine | 32 (5.7) | 114 (5.1) | 1.62 (0.85–3.09) | 1.39 (0.58–3.34) |
| Comedication of two mood stabilizers | 209 (37.5) | 711 (31.9) | 1.81 (1.36–2.40)c | 1.66 (1.21–2.27)b |
| Lithium and valproate | 91 (16.3) | 242 (10.9) | 2.28 (1.33–3.90)b | 3.38 (1.55–7.38)b |
| Lithium and carbamazepine | 103 (18.5) | 414 (18.6) | 1.25 (0.84–1.86) | 1.10 (0.69–1.77) |
| Valproate and carbamazepine | 15 (2.7) | 55 (2.5) | 1.53 (0.54–4.34) | 0.74 (0.13–4.22) |
| Comedication of three mood stabilizers | 121 (21.7) | 339 (15.2) | 2.46 (1.57–3.86)c | 2.65 (1.58–4.45)c |
- Values reported as n (%). CI = confidence interval.
- aAdjusted for covariables associated with hypothyroidism and/or with use of mood stabilizers (i.e., thyroiditis, thyroid surgery, goiter, thyrotoxicosis, renal disease, liver disease, epilepsy, phenytoin, antithyroid agents, and other thyroid disease) using multivariate conditional logistical regression.
- bp < 0.01.
- cp < 0.001.
For the model including all combinations (i.e., lithium only, valproate only, carbamazepine only, combination of lithium and valproate, combination of lithium and carbamazepine, combination of valproate and carbamazepine, and combination of all three mood stabilizers), we found that the odds of developing hypothyroidism were significantly greater in patients exposed to carbamazepine only (OR = 1.68, 95% CI: 1.07–2.65) or combined use of lithium and valproate (OR = 2.40, 95% CI: 1.70–3.40), lithium and carbamazepine (OR = 1.52, 95% CI: 1.10–2.08), and all three mood stabilizers (OR = 2.34, 95% CI: 1.68–3.25) compared with patients who had never used any of these three mood stabilizers.
Number of mood stabilizers
When the risk of hypothyroidism was examined for the use of one, two, and three mood stabilizers, the risks for incident hypothyroidism among patients who used a combination of two or all three mood stabilizers were 1.81 times (adjusted OR = 1.66) and 2.46 times (adjusted OR = 2.65) greater, respectively, than the risk among patients who did not use any of the three mood stabilizers (Table 3). When the effect of the combined medications was treated as a linear variable in the order of no use, single use, combination of two, and combination of three mood stabilizers, the result showed no departure from the linear trend (goodness of fit test statistics = 0.867, df = 2, p = 0.648). Therefore, the use of one more mood stabilizer was associated with a 1.34 times (95% CI: 1.21–1.49, p < 0.001) higher likelihood of risk for hypothyroidism.
Interactions among mood stabilizers, age, and sex
We further examined the interactions among the three mood stabilizers, age, and sex starting from the three-way interactions (two mood stabilizers and sex or age) based on the use of a prescription of seven or more DDDs in one prescription. We found that there were neither any significant three-way interactions nor any significant two-way interactions except a significant interaction between lithium and valproate (regression coefficient = 0.605, chi-square = 5.41, p = 0.02) on the risk for hypothyroidism. Lithium-treated patients who also used valproate had 2.03 times the risk of developing hypothyroidism as compared with those who were not treated with any of the three mood stabilizers (Table 4). However, neither valproate-treated patients who did not use lithium nor lithium-treated patients who did not use valproate had a significantly higher risk of developing hypothyroidism than those who were not treated with any of the three mood stabilizers.
| Model | Regression coefficients | |
|---|---|---|
| β value (SE) | (95% CI) | |
| Lithium | 0.077 (0.123) | (–0.164–0.318) |
| Valproate | 0.027 (0.236) | (–0.436–0.490) |
| Carbamazepine | 0.185 (0.105) | (–0.021–0.391) |
| Lithium × valproate | 0.605 (0.260) | (0.095–1.115) |
| Odds ratio | (95%CI) | |
| No carbamazepine | 1.00 | |
| Carbamazepine | 1.20 | (0.98–1.48) |
| Lithium and valproate interaction | ||
| No lithium, no valproate | 1.00 | |
| No lithium, valproate | 1.03 | (0.65–1.63) |
| Lithium, no valproate | 1.08 | (0.85–1.38) |
| Lithium, valproate | 2.03 | (1.54–2.68) |
- β = estimate of regression coefficient; SE = standard error; CI = confidence interval.
Discussion
The current study is among the first few population-based studies to examine hypothyroidism as an adverse outcome of lithium, carbamazepine, and valproate simultaneously using a prospective nested case-control study design. The major findings of the current study are that all three mood stabilizers (lithium, carbamazepine, and valproate) assessed herein increased the risk for hypothyroidism among patients with bipolar disorders, even after controlling for other physical conditions and hypothyroidism-related medications. Moreover, there was a significant interaction between the use of lithium and valproate: a significant positive dose-related linear relationship between the numbers of mood stabilizers used and risk for incident hypothyroidism. Lastly, the incidence of hypothyroidism (65 per 10,000 person-years) among patients with bipolar disorders estimated in this study is higher than those reported in the general population in Western countries (15–18) and China (19).
Consistent with the findings from epidemiological studies (13, 14) and studies of patients treated with lithium (23–26), our findings also demonstrated that females with bipolar disorders are more likely to develop hypothyroidism than their male counterparts. Investigation of the gender difference in the influence of valproate and carbamazepine on thyroid function in future studies is warranted.
In line with other studies, we found that use of lithium (20–22, 41), carbamazepine (32–37), and valproate (38, 42, 43) can lead to the development of hypothyroidism. Moreover, the prediction of lithium use to the risk for hypothyroidism was not only demonstrated in the Western populations (20, 21, 23) but also in ethnic Chinese populations (22). However, the findings of an increased risk for hypothyroidism among valproate-treated patients are not supported by some studies in patients with epilepsy (34, 36, 44). Although lithium has been studied the most in bipolar disorder patients (41, 45, 46), and some studies have been done regarding the use of carbamazepine (32, 34, 36) and valproate (38, 42, 43) in patients with epilepsy, there is a lack of information about whether carbamazepine and valproate have a similar adverse effect in patients with bipolar disorders. Our finding of association between the combined use of carbamazepine and valproate and increased risks for hypothyroidism in bipolar disorder patients is a new contribution of the current study. This novel finding indicates that exploring the adverse effect of these two mood stabilizers on the thyroid function, particularly in bipolar disorder patients, is necessary.
Although, like several studies, we found that use of lithium was associated with higher risk for hypothyroidism (20–22) and combined use of lithium and valproate further increased this risk (47), we did not find a higher risk for hypothyroidism in bipolar disorder patients who were only treated with lithium when compared with patients who were prescribed carbamazepine or valproate only. Taken together, due to the well-established association between lithium use and hypothyroidism, it is recommended that physicians pay closer attention to hypothyroidism in lithium-treated patients, particularly when combining the use of valproate. Once there is any symptom of developing hypothyroidism in lithium users, the medication may be switched to other mood stabilizers. Consequently, close monitoring the use of lithium and thyroid function, especially in high-risk groups (the elderly, women, patients with thyroid autoimmune disorder or thyroid surgery), may further reduce the risk for hypothyroidism.
Among the three mood stabilizers, valproate has been studied the least in regards to hypothyroidism. Despite different study populations, which include patients with epilepsy as reported in the literature (32, 37, 42, 43) and patients with bipolar disorders in this study, these findings provide some evidence to support an association between valproate and hypothyroidism, which contradicts the findings in some studies (34, 36). Surprisingly, the association between valproate and hypothyroidism is found to be even stronger than that reported between carbamazepine and hypothyroidism, suggesting that valproate-treated patients are equally at risk, if not higher, for developing hypothyroidism as carbamazepine-treated patients. Although Shulman et al. (48) reported a higher rate of thyroxine prescriptions in new lithium users than in new valproate users, our study did not support such distinct risks for hypothyroidism between the two mood stabilizers. The current study’s failure to find a higher risk for hypothyroidism in lithium-treated only as compared to valproate-treated only patients may be due to the difference in the age groups, i.e., patients aged older than 65 years in Shulman’s study and patients 15–65 years of age in this study. However, we were not able to show any age effect on the risk for hypothyroidism, which is consistent with the reports in the literature (23, 49).
Similar to the findings of Gracious et al. (47), this study demonstrated an increased risk of hypothyroidism in patients who used a combination of lithium and valproate. Although the current study did not find differential risk for hypothyroidism among the three mood stabilizers, the combined use of lithium and valproate significantly increased the risk for hypothyroidism, but the other two combinations did not.
The finding of greater risk of hypothyroidism in a combination regimen of lithium and valproate may be explained on the basis of pharmacokinetics and drug-drug interactions. Literature did not document significant pharmacokinetic interactions between lithium and valproate (50). Lithium and valproate may synergetically suppress thyroid hormone secretion, a suggestion indirectly supported by a recent study (51), which showed that combination therapy with lithium chloride and histone deacetylase (HDAC) inhibitors valproate and suberoyl bis-hydroxamic acid (SBHA) decreases growth and suppresses hormone secretion of medullary thyroid cancer. The mechanism is that the HDAC inhibitors activate the Notch1 signaling pathway while lithium chloride inhibits the glycogen synthase kinase-3ss (GSK-3ss) pathway. The HDAC inhibitors, including valproate and lithium, limit cell growth and suppress thyroid hormonal secretion (51). Combined lithium and valproate therapy leading to an elation of TSH in bipolar disorder patients is associated with a higher baseline TSH level and a higher lithium and valproate level (47). Since the use of valproate to treat bipolar disorder patients is on the rise in Taiwan, and because the likelihood of its combined use with lithium will increase, our findings have clinical implications; however, further studies are needed to confirm the interaction between lithium and valproate and to explore the underlying mechanism of this interaction.
The current study provides strong evidence of a dose-response linear relationship between the number of mood stabilizers used and the risk for hypothyroidism. Despite the absence of information regarding the underlying mechanism of possible drug-drug interactions resulting from combined use, we should nonetheless be cautious about the combined use of mood stabilizers since some studies have reported an association between such a combination and increased risk of hypothyroidism (21, 34). For example, Bocchetta et al. (21) reported that combined use of carbamazepine and lithium may decrease TSH levels, thereby affecting the detection of hypothyroidism in lithium-treated patients. Verrotti et al. (34) found that epileptic children treated with carbamazepine may have subclinical signs of hypothyroidism and that these changes are more evident if carbamazepine is given in association with valproic acid.
Since these mood stabilizers are not only indicated for the treatment of affective symptoms, and because their other adverse effects may have influenced prescription patterns, we assessed and controlled for the following conditions in our data analyses. It has been reported that lithium may induce nephrotoxicity, such as chronic tubulointerstitial nephropathy, and end-stage renal disease (52, 53), and that valproate may cause drug-induced hepatopathy (54). Accordingly, medical records of abnormal renal and liver functions and/or renal and liver diseases may hinder psychiatrists from prescribing lithium or valproate to these patients, respectively (55). In addition, carbamazepine and valproate are more likely than lithium to be prescribed to bipolar disorder patients affected with epilepsy to control their affective symptoms (56). Studies have found that reduced glomerular filtration rate is associated with a higher prevalence of hypothyroidism (57); our finding that the prevalence of renal disease was higher before the onset of hypothyroidism may indirectly support this relationship. Although the physical problems/diseases, including renal disease, assessed herein may not have any cause-effect relationship with hypothyroidism, these disease variables were controlled in the data analyses.
Limitations of the study
This study is one of few prospective studies that have examined hypothyroidism as an adverse outcome of lithium, carbamazepine, and valproate within a single study using a national sample of patients with bipolar disorders in Taiwan. A nested, matched case-control study design using a cohort of patients with bipolar disorders and adequate controls for comorbid physical conditions and comedications constitute the strengths of this study.
However, the major limitation of the present study is that, as it is based on a claims dataset with scrambled patient identification for privacy protection, there is a lack of detailed patient information such as drug adherence and symptom severity, and it is impossible to validate the data in the medical records to confirm the diagnosis of bipolar disorders, hypothyroidism, and other disease. Moreover, there were no data regarding the serum level of thyroid hormone to validate subjects’ thyroid function, which would have been useful to evaluate the severity of hypothyroidism. Instead, the diagnosis of hypothyroidism was based on either the ICD-9 code or prescription of thyroxine, liothyronine, or thyroid hormone. In addition, this study cannot determine whether hypothyroidism is associated with bipolar disorder. Another limitation is the time lag of the claim database. NHRI takes almost one year to update the claim database, so the findings provided here may not be up to date. However, the claim data of Taiwan’s NHI provide plentiful prospective patient data, cover 98% of the national population, have a specific diagnostic algorithm for psychiatric disorders, and allow for psychiatric assessments to be performed quickly and cost effectively. Lastly, the generalizability of the findings is questionable. It is also not possible to generalize our results to a broad ethnic Chinese population because Chinese in Taiwan may have different cultural, psychosocial, and economical backgrounds from the Chinese in China and other countries.
Conclusions
Findings from this study suggest that, in addition to lithium, carbamazepine and valproate also increase the risk for hypothyroidism. The combined use of these mood stabilizers further increases the risk for hypothyroidism, particularly the combined use of lithium and valproate. Our finding of a linear dose-response relationship between number of mood stabilizers and risk for hypothyroidism implies that regular monitoring of thyroid function is necessary for patients taking mood stabilizers. We also recommend that monotherapy of mood stabilizers should be recognized as the guideline for treating affective symptoms.
Acknowledgements
This work was supported by a grant from the Department of Health, Executive Yuan, Taiwan [DOH 93-TD-D-113-020-(2)]. This study is based in part on data from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health, and managed by National Health Research Institutes. The interpretation and conclusions contained herein do not represent those of the Bureau of National Health Insurance, Department of Health, or National Health Research Institutes.
