Exposure to Lithium and Cesium Through Drinking Water and Thyroid Function During Pregnancy: A Prospective Cohort Study

Introduction

The alkali metal lithium (Li) has long been used in the treatment of bipolar disease (1). Impairment of the thyroid function is a well-documented side effect of lithium medication (2,3). The general population may also be exposed to lithium through drinking water, including bottled water (4 –7), and we recently reported that such exposure was associated with increasing thyrotropin (TSH) and decreasing free thyroxine (fT4) levels (8).

An adverse effect of lithium on the thyroid may be even more critical during pregnancy. The thyroid gland is in charge of producing hormones that are essential for pre- and postnatal growth and cognitive development (9). The fetal thyroid starts developing around the fourth week of pregnancy and becomes functional near mid gestation, which is why an appropriate maternal thyroid function is crucial for fetal development (9,10). Hypothyroidism, a condition in which the thyroid gland is incapable of producing enough thyroid hormones, has been associated with increased risk of gestational hypertension, placental abruption, preterm delivery and fetal loss (11,12), as well as lower birth weight, congenital hypothyroidism, and impaired neurological function (13 –15). We recently found inverse associations between maternal exposure to lithium during pregnancy and birth size (16) and hypothesize that an impairment of the thyroid hormonal system might be an underlying mechanism. Therefore, in this study we aimed at elucidating the potential impact of lithium from drinking water on maternal thyroid function during pregnancy.

Materials and Methods

Results

Out of the 176 recruited pregnant women (Supplementary Fig. S1), we excluded five who tested positive for anti-TPO antibodies. All the remaining 171 women had at least one measurement of both exposure and thyroid markers during pregnancy (n = 27 in first, n = 82 in second, and n = 146 in third trimester; in total 255 measurements, Supplementary Fig. S1).

The drinking water lithium concentrations varied between 530 and 830 μg/L (n = 58) in San Antonio de los Cobres and between 5.0 and 324 μg/L (n = 83) in the surrounding villages, except for a small village 20 km east of Cobres (Aguas Blancas), where the concentrations were 958–1660 μg/L (n = 3). Overall, concentrations of both cesium (0.01–717 μg/L) and arsenic (0.82–320 μg/L) in drinking water varied in the different villages. The Spearman's rank correlation coefficients were: blood and urinary lithium, r_s = 0.84 (p < 0.001); blood and water lithium (also collected at the same time), r _s = 0.40 (p < 0.001); and urinary and water lithium, r _s = 0.44 (p < 0.001). Blood and urinary lithium were also positively correlated with blood cesium (r _s = 0.55, p < 0.001, and r _s = 0.47, p < 0.001, respectively); urinary cesium (r _s = 0.53, p < 0.001, and r_s = 0.61, p < 0.001, respectively); and urinary arsenic (r _s = 0.53, p < 0.001, and r _s = 0.67, p < 0.001, respectively) but only weakly correlated with urinary iodine (r _s = 0.23, p < 0.001, and r _s = 0.25, p < 0.001, respectively). Blood cesium was moderately correlated with urinary cesium (r _s = 0.43, p < 0.001) and water cesium (r_s  = 0.38, p < 0.001). To avoid collinearity in the linear quantile regression, models for blood lithium were adjusted for urinary cesium, and models for blood cesium were adjusted for urinary lithium.

Compared with the first tertile of blood lithium [median 11 μg/L (0.0016 mmol/L), range 2.1–19 (0.00030–0.0027)], the concentrations of urinary lithium, blood and urinary cesium, and urinary arsenic and iodine were significantly higher in the third tertile [median 42 μg/L (0.0060 mmol/L), range 34–145 (0.0049–0.021)] (Table 1). There was no difference in the other characteristics across the tertiles of blood lithium concentrations. More than 60% of the pregnant women had urinary iodine concentrations <150 μg/L at some point during pregnancy (Supplementary Table S2), and among the 82 women with 2–3 iodine measurements, 43 (52%) had all measurements <150 μg/L. Only two women had serum selenium concentrations <60 μg/L (Supplementary Table S2).

As shown in Table 2, the serum concentrations of TSH and thyroglobulin increased with advancing pregnancy, while fT4, T4, fT3, and serum selenium decreased. The fraction of women with TSH levels above the reference value (28) during the first (0.1–2.5 mIU/L), second (0.2–3.0 mIU/L), and third (0.3–3.5 mIU/L) trimesters was 19%, 12%, and 13%, respectively (Supplementary Table S2). Out of all the TSH values above the reference range, 83% belonged to the middle and highest tertiles of blood lithium, and only 17% to the lowest tertile. Out of all fT3 values below the reference range, 50%, 28%, and 22% belonged to the highest, middle and lowest tertile of blood lithium, respectively. Blood lithium increased during pregnancy, while blood cesium and serum selenium tended to decrease (Table 2).

The distribution of all unadjusted thyroid markers across tertiles of blood lithium is shown in Fig. 1. TSH was higher while fT3 and total T3 were lower in the highest tertile of blood lithium compared with the lowest (Fig. 1). The multivariable adjustment of the linear quantile regression models showed only minor changes in the estimates. The regression models revealed positive associations between log2 blood lithium concentrations and TSH in serum, particularly in the lower tail of the TSH distribution (Table 3, Supplementary Fig. S2A). Each doubling of blood lithium concentrations was associated with an increase of about 0.2 mIU/L at both the fifth and twenty-fifth percentiles of TSH (Table 3). Also, the blood lithium concentrations were associated with a decrease in transthyretin, particularly in the highest percentiles (Table 3, Supplementary Fig. S2B), and with a decrease in fT3 and total T3 with less obvious variation in the estimates across outcome percentiles (Table 3, Supplementary Fig. S2C, D). Associations with fT4, total T4, and thyroglobulin were inverse, but not statistically significant (Table 3) and therefore not included in Supplementary Figure S2. Of note, gestational age was significantly associated with TSH, fT4, total T4, fT3, and transthyretin in the multivariable adjusted analyses (data not shown).

OPEN IN VIEWER

FIG. 1.

Box plots of maternal blood lithium concentrations and thyrotropin (TSH) (A); free thyroxine (fT4) (B); total T4 (C); free triiodothyronine (fT3) (D); total T3 (E); thyroglobulin (F); and transthyretin (G). Box layers describe the 75th, 50th, and 25th percentiles. Extreme outliers are not shown.

Cesium concentrations were inversely associated with fT3 and T3, particularly at the higher percentiles (Table 3, Supplementary Fig. S2E, F). For each doubling of blood cesium, the 95th percentile of fT3 and total T3 decreased by 0.13 pmol/L, and 0.17 nmol/L, respectively (Table 3).

In a sensitivity analysis, we tested if the observed associations of blood lithium and cesium with T3 and fT3 could be influenced by illness during pregnancy. Different types of infections (i.e., cold, vaginal infection, urinary tract infection, and fever) were the most frequently reported illnesses. When we grouped all infections, neither fT3 nor T3 was different. In those having infection, mean fT3 was 4.46 pmol/L (range 4.31–4.60), and in women without any infection 4.43 (4.34–4.5; p = 0.67); while mean T3 was 2.77 nmol/L (range 2.65–2.87) in women with infection and 2.69 (range 2.62–2.76) in women without infection (p = 0.24). When we adjusted the statistical models of fT3 and T3 in relation to blood lithium or cesium for this variable (infections yes/no), the estimates remained essentially the same and the variable “infections” was not statistically significant in any of the models. The frequency of most other illnesses encountered was generally low—for example, nausea/abdominal pain (first trimester n = 2, second trimester n = 2, third trimester n = 5)—and for this reason we could not test this statistically. However, women having any illnesses other than infection had similar fT3 and T3 as women without such illnesses.

Discussion

This study provides the first indication that exposure to lithium and cesium through drinking water may adversely affect thyroid function during pregnancy. In particular, increasing blood lithium concentrations were associated with higher serum TSH, but also lower fT3, total T3 and transthyretin levels. The quantile regression models indicate that the women in the lowest percentiles of TSH and highest percentiles of T3 are more susceptible to lithium. An unexpected secondary finding is the inverse association between cesium, which also occurred at highly varying concentrations in the drinking water, and both fT3 and T3.

One previous study from Buenos Aires, on the East coast of Argentina, presumably with a very low water lithium concentration, reported median TSH concentrations of 0.95, 1.5 and 1.6 mIU/L in the first, second, and third trimesters, respectively, (27) which is similar to the observed concentrations in our study, particularly in the first lithium tertile, (1.3, 1.7, and 1.8 mIU/L, respectively). Our results on lithium exposure and thyroid function are in agreement with previous studies, although most of those concerned lithium medication, which results in much higher blood lithium concentrations (therapeutic range 0.8–1.2 mmol/L). A recent meta-analysis and a retrospective cohort study of clinical studies on lithium therapy showed that lithium therapy is associated with an increased risk of hypothyroidism (2,3). In support of a low-dose effect at the population level, we previously observed a positive association between lithium exposure through drinking water and TSH in nonpregnant women in the same Andean area (8). The potential health consequences of the indicated effects of lithium on thyroid markers, including the potential relation to the previously found lithium-related shorter birth length in the present cohort (16), need to be elucidated. In particular, the children need to be followed-up, as lithium readily crosses the placenta (28) and thyroid hormones are known to play a critical role in fetal growth and development (9,10), including cognitive function (14,15).

Suggested mechanisms for the effects of lithium medication on thyroid function include inhibition of iodine uptake, inhibition of iodotyrosine coupling, alteration in thyroglobulin structure, and inhibition of thyroxine secretion (29,30), most of which would fit with the present findings. The results relating to T3 and fT3 could conceivably be due to an inhibitory action of lithium on the conversion of T4 to T3 (31). Lithium is known to accumulate in the thyroid gland and also to cross the blood–brain barrier and accumulate in the hypothalamus (32,33). Therefore, lithium may impair the thyroid homeostasis both at the pituitary level, directly in the thyroid follicles, and via disruption of the thyroid hormone transport or conversion in the bloodstream. If confirmed in other studies, the found inverse associations between blood lithium and transthyretin might be a novel mechanism of low-level lithium exposure. Transthyretin is one of the proteins responsible for transporting T4 in the body and the most abundant T4 transporter in the cerebrospinal fluid (34). Importantly, transthyretin is produced in placental trophoblastic cells and transports thyroid hormones to the fetus (35). Thus, a decrease in transthyretin might lead to lower fetal thyroid hormone levels (36).

Cesium is also known to accumulate in the thyroid gland, as shown for the radioactive form (21). Still, we found only one previous study evaluating exposure to stable cesium (and multiple other elements) and thyroid function, based on the NHANES 2007–2008 data. It showed an inverse association between urinary cesium concentrations and TSH, but no associations with fT3 or T3 (37). However, the cesium concentrations were extremely low (median 4.6 μg/L, interquartile range 3.4–6.1 μg/L), and the results cannot be compared with those of the present study with two orders of magnitude higher urinary concentrations.

About half of the studied women were found to be iodine deficient, which might increase susceptibility to lithium toxicity. Due to the lack of power, this was not tested in the present study. Although iodine deficiency–related goiter is prevalent in the province of Salta (38,39), the prevalence of goiter among school-aged children decreased after the introduction of iodinated table salt; from 68% in 1975, to 32% in 1981 (38), and to 6.3% by 2008 (40). Our present findings of low iodine concentrations in the Andean part of Salta province raise the need for a thorough follow-up. We found very low iodine concentrations in both the drinking water (5–30 μg/L) and the water from a couple of the salt lakes (salars; 100–400 μg/L). Thus, the use of the salt from the many salars as table salt would provide little of the required iodine.

The exposure to both lithium and cesium had a marked spatial variation in the study area. Similarly high drinking water concentrations of lithium have been found in a few other countries (7,41), but obviously, more intense water monitoring is needed worldwide. A number of brands of bottled water with lithium concentrations above 1 mg/L have been reported (7), but a large number of brands remain to be analyzed for lithium and cesium. Although the drinking water most likely was the main source of exposure in the present study, the correlation with lithium and cesium in blood or urine was only moderate. This suggests a variation in the water consumption, both directly and through food and drinks, e.g. the commonly eaten soups (∼1100 μg lithium/L), especially during the cold winter period. Probably, the food as such (mainly cereals, vegetables, and some meat) contributed little to the exposure as very few crops are grown in the arid Puna region and samples of meat from lama animals, grazing in the surrounding Puna, were found to contain only about 40 μg/kg of lithium. Inhalation of dust might however have contributed some because of the prevailing windy climate. Nevertheless, the variations emphasize the need for repeated measurements of individual exposure biomarkers for reliable assessment of the internal dose.

The strengths of our study include the population-based prospective design, covering most of the Andean part of Salta province, with wide ranges of exposure to lithium and cesium. Another strength is the use of multiple biomarkers of thyroid function, collected longitudinally across pregnancy, along with the measurements of the exposure biomarkers. One of the limitations of our study is the lack of samples for all women at all time points. Also, the potential impact of the low atmospheric oxygen concentration at the high altitudes on thyroid function (42) was not assessed. However, all women in the study were living at high altitude (all at more than 3,000 meters above sea level) and 56% of the women had lived in the area since birth. Moreover, we did not have the possibility to measure thyroxine-binding globulin (TBG), which might have had an impact on the T3 levels due to its increase in early pregnancy (43). However, in any case, an impact of elevated TBG on T3 would rather decrease the possibilities to detect associations with lithium and cesium. Another limitation is the fairly small cohort. Indeed, the findings emphasize the need for larger studies also in other populations.

In conclusion, this study indicates that elevated environmental lithium and cesium exposure may adversely affect the maternal thyroid function during pregnancy. Due to the important role of thyroid hormones in child development, in particular neurodevelopment, a follow up of early-life lithium-exposed children is warranted. Equally important is a follow-up of the iodine status and to make sure that the pregnant women and children, in particular, have adequate iodine intake. Our findings are of public health relevance and they reinforce the need for more drinking water measurements worldwide and also guideline values for drinking water, including bottled water.