Thyroid-Stimulating Hormone (TSH) Concentration at Birth in Belgian Neonates and Cognitive Development at Preschool Age

The main objective of the study was to investigate the effect of MID during late pregnancy, assessed by the thyroid-stimulating hormone (TSH) concentration at neonatal screening, on cognitive development of preschool children. A retrospective cohort study including 311 Belgian preschool children of 4–6 years old was conducted. Children were selected at random from the total list of neonates screened in 2008, 2009, and 2010 by the Brussels new-born screening center. Infants with congenital hypothyroidism, low birth weight, and/or prematurity were excluded from the selection. The selected children were stratified by gender and TSH-range (0.45–15 mIU/L). Cognitive abilities were assessed using Wechsler Preschool and Primary Scale of Intelligence—third edition. In addition, several socioeconomic, parental, and child confounding factors were assessed. Neonatal TSH concentration—a surrogate marker for MID—was not associated with Full Scale and Performance IQ scores in children. Lower Verbal IQ scores were found in children with neonatal TSH values comprised between 10–15 mIU/L compared to lower TSH levels in univariate analysis but these results did not hold when adjusting for confounding factors. Current levels of iodine deficiency among pregnant Belgian women may not be severe enough to affect the neurodevelopment of preschool children.


Introduction
Iodine is necessary for the synthesis of the thyroid hormones thyroxin (T4) and tri-iodothyronine (T3) [1], which are essential for the development of the brain during fetal and early postnatal life.
In these critical periods, severe iodine deficiency can induce irreversible brain damage in the fetus and the infant, resulting in retarded cognitive or/and psychomotor development [2].
While severe iodine deficiency is disappearing in developed countries, mild iodine deficiency (MID) in women of child-bearing age and pregnant women is still present in some European countries including Belgium [3][4][5]. There is some evidence that MID during pregnancy may be associated with impaired neurodevelopment of the offspring [6]. An eventual impact of maternal MID on neurodevelopment of the children is a major public health concern and justifies further study given the limited data available.
Thyroid-stimulating hormone (TSH) concentration in whole blood measured at birth has been proposed as an indicator of maternal iodine status during late pregnancy [7,8]. Thyroid hormone level is dependent on iodine nutrition and iodine stocks when thyroid disease is excluded [9]. Liberation of TSH occurs to stimulate production of the thyroid hormone when it is insufficient. When the concentration of thyroid hormone returns to normal, TSH liberation stops. In the case of iodine deficiency, the thyroid hormone level does not adapt. TSH continues to be secreted and its concentration within the blood increases.
Previous studies have shown an association between elevated neonatal TSH (>5 mIU/L) concentration (cord serum TSH or whole blood) and subclinical impairments of cognitive development during childhood [10][11][12]. These studies are scarce and they used very small sample sizes [10][11][12]. In addition, other pregnancy-related factors than MID may explain both the variation of neonatal TSH concentrations in blood and impairments in cognitive development such as premature birth or low birth weight [6]. Additional studies are needed to understand the potential harmful effect of mild neonatal thyroid dysfunction on later cognitive development.
The aim of this study was to assess the association between neonatal TSH concentration and cognitive development at preschool age in Belgium, a country where MID is frequent during pregnancy. This association was studied using a sample size calculation and excluding children born prematurely and/or with a low birth weight and without congenital hypothyroidism, as well as taking into account confounding factors using multivariate analysis. In addition, this is the first study to use a sample stratified by sex and TSH levels in order to study this association.

Study Population
We used data from a Belgian retrospective cohort study, the PsychoTSH study, including 315 children aged 4, 5, and 6 years old with a neonatal TSH concentration in the range 0.45-15 mIU/L (micro international unit/liter) [13]. A sample size of 315 children was determined based on a detection probability of 95%, a significance level alpha of 5% and a correlation factor between TSH and IQ of 0.2. An anticipated drop-out of 20% was taken into account. Children were randomly selected from the total list of neonates screened in 2008 (N = 29,013), 2009 (N = 29,602) and 2010 (N = 30,126) by the Brussels New-born Screening Centre for Metabolic Disorders (Laboratoire de Pédiatrie, Université Libre de Bruxelles (ULB), Brussels, Belgium). Children with congenital hypothyroidism, prematurity (<37 weeks), and low birth weight (<2500 g) were excluded from the sample frame before the selection procedure. The sampling was stratified by gender and TSH level to ensure that the whole range of neonatal TSH was represented. For each gender and TSH-interval (0.45-1, 1-2, 2-3, 3-4, 4-5, 5-6, 6-7, 7-8, 8-9 and 9-15 mIU/L) 19 new-borns were selected. Within each stratum, for each selected infant, three replacements were selected, in case of refusals or non-contactable individuals. Children with neurological diseases were excluded from the study during the recruitment process.

Approval by Ethical Committee for Human Subjects and the Privacy Commission
A written informed consent was obtained from the parents before the test administration. The protocol and the consent form were approved by the Ethical Committee of the Erasmus hospital (ULB, Brussels, Reference CCB: B40620109191) in accordance with the Code of Ethics of the World Medical Association for experiments involving humans (Declaration of Helsinki). This study was also approved by the Belgian Commission for the Protection of Privacy (Reference: RN 29/2012).

TSH Measurement
TSH levels were measured by a commercial time-resolved fluoroimmunoassay (Autodelfia method) in dried blood spots on filter paper, which were collected by heel stick between 3 to 5 days after birth [14]. The reproducibility of the TSH values in the range 0-15 mIU/L was tested. At 50 different TSH values (e.g., 0.5, 2 mIU/L, 3 mIU/L, 4 mIU/L, 6 mIU/L, 8 mIU/L, 10 mIU/L) TSH was reanalyzed twice in order to determine the coefficient of variation. It was found that for the TSH values ranging from 0.9 to 10 mIU/L the variation coefficient of was below 20%. For values below 0.9 mIU/L a TSH value of 0.45 mIU/L was set in the study.

Cognitive Assessment
Cognitive development was assessed by the French version of the Wechsler Preschool and Primary Scale of Intelligence-third edition (WPPSI-III) [15]. The WPPSI-III is a test administered individually which enables one to measure the intelligence of children aged between 2 years 6 months and 7 years 3 months. The test is composed of 14 subtests: seven verbal tests, five performance tests, and two processing speed tests. These subtests enable the calculation of a Full Scale Intelligence Quotient (FIQ), a Performance IQ (PIQ), a Verbal IQ (VIQ), a Processing Speed Quotient and an optional score of General Language Knowledge. In this study, only the methodology for the age group 4 years to 7 years 3 months was used and only 7 core subtests were administered: Block Design, Information, Matrix Reasoning, Vocabulary, Picture Concepts, Word Reasoning and Coding. Cognitive testing was undertaken by psychologists who were blinded for the neonatal TSH levels of the selected children. Test administration was performed during a home visit. The parents were asked to provide a quiet room and leave the child alone with the psychologist in order to avoid distraction.

Data Collection of Descriptive Variables, Effect Modifiers and Covariates
The body weight, height, and head circumference of the child were measured during the home visit. A sample of urine was collected from the child during the home visit and analyzed using a modification of the Sandell-Kolthoff reaction with spectrophotometric detection [16] at the Clinical Chemistry Laboratory of the Erasmus Hospital (ULB, Brussels).
Date of birth, date of blood sampling, pregnancy duration and body weight at birth were provided by the Brussels new-born screening center for metabolic disorders (Laboratoire de Pédiatrie, ULB, Brussels).
Data concerning body length at birth, head circumference at birth, and Apgar score was collected from the health booklet of the child during the home visit.
Additional information was collected using a self-administered questionnaire filled in by the mother: pre-pregnancy body mass index (BMI) of the mother, thyroid disease, drug intake, chronic disease, alcohol consumption, smoking habits, weight gain and intake of nutritional supplements of the mother during pregnancy, maternal age at birth, reproductive history, parity, gravidity, type of delivery, perinatal asphyxia, health condition of the new-born, child's negative life events, maternal mental health, maternal social support, marital discord, and parent-child interactions. Maternal mental health was assessed using the General Health Questionnaire-12 [17] items and vitality scale questionnaire of the Short Form Health Survey-36 [18].

Data Analysis
Statistical analysis was performed using SAS statistical software 9.3 (SAS Institute Inc., Cary, NC, USA) for univariate analysis and Stata version 13 (StataCorp, College Station, TX, USA) for multivariate analysis. Statistical tests were two-sided and tests with p-value <0.05 were considered statistically significant. The missing values were kept as missing and the analyses were restricted to subjects with complete data on the variables involved in the analysis. Neonatal TSH values and UIE were presented as median and IQR and IQ scores were presented as mean and SD. Neonatal TSH values were classified in three groups: below 5 mIU/L, between 5 and 9 and equal or above than 10 mIU/L. Student t-test, anova with Bonferroni correction and simple linear regression were performed to examine univariate associations between IQ and studied parameters. Association between TSH level and IQ was assessed using Pearson correlation.
Multivariate linear and logistic regression models were used to study the predictors of variation of IQ scores in children. All continuous variables and categorical variables associated with IQ scores with a test p < 0.20 were included in the selection procedure [19]. A multiple linear regression model was built, a stepwise backward selection procedure with an entry probability of 0.10 and exit probability of 0.15 was used to define variables to be inserted into the model. Univariate association between these variables were tested with Pearson correlation. Highly correlated variables were not included together in the model. Normal plot of residuals was used to check the normality of the residuals distribution. Linearity and homoscedasticity of residuals were checked by examination of the plot of standardized residuals. Colinearity between predictors was tested using the test of variance inflation factor (VIF) and individuals' VIF for each parameter in the model were around 1.
Logistic regression was used to estimate the risk of having a developmental score below 85 (one standard deviation below the mean of 100). A stepwise selection with an exit probability of 0.15 was used to determine the variable to insert into the final logistic regression model. For the logistic regression, adjusted odds ratios were used and 95% CI was derived from the final model. The adequacy of regression models has been tested by the Hosmer and Lemeshow test.

Demographical Characteristics, Neonatal Thyroid Status and Current Iodine Status of the Study Sample
Descriptive and demographic characteristics of the study population according to gender are shown in Table 1. In total, 311 children (N = 137 girls) aged 4, 5, and 6 years were included in the study. The median (range) TSH level of the study sample was 3.6 mIU/L (1.8-5.8 (IQR), 0.45-13.9 (min-max)). The median iodine concentration of the sample was 138.8 µg/L (87.7-238.7 (IQR)) indicating iodine sufficiency of the study population.
Univariate associations between cognitive scores at preschool age and iodine status are shown in Table 4.
No differences were found in FIQ and PIQ scores in children with TSH levels lower than 5 mIU/L compared to TSH levels between 5 and 9 mIU/L and between 10 and 15 mIUL/L. Lower VIQ scores were found in children with TSH levels between 10-15 mIU/L compared to lower levels (0.45-9 mIU/L) (p = 0.006). After correction for confounding factors-such as monthly income of the household, education level of the mother or child bilingualism-by multiple regression analysis (see below), this difference was no longer significant.
The use of iodized salt was related to higher PIQ scores (p = 0.013) and the intake of dietary supplement by the children was associated with higher FIQ (p = 0.023) and PIQ (p = 0.020).
No associations were found between cognitive scores at preschool age and the following factors: Apgar score at 5 min, health problems at birth, neonatal hospital attendance, previous cognitive assessment, school attendance, child's negative life events, child custody, rural or urban residency, delivery type, previous miscarriage, Graves' disease or Hashimoto's thyroiditis during pregnancy, hypothyroidism during pregnancy, diabetes during pregnancy (data not shown).

Predictors of IQ Scores Assessed by Multiple Linear Regression Analysis
The results of multiple linear regressions of factors explaining variation of IQ scores at preschool age are shown in Table 5. TSH concentration at birth was not associated with any cognitive scores. The predictors of higher IQ scores were higher household income (FIQ, p = 0.007; PIQ, p = 0.014; VIQ, p = 0.022), higher mother's education level (FIQ, p = 0.017; PIQ, p = 0.046; VIQ, p ď 0.0001), and mother's optimal vitality (FIQ, p = 0.04, PIQ, p = 0.013 VIQ, p = 0.012). Female gender was associated with higher FIQ (p = 0.019). A positive association was found between PIQ and child dietary supplement intake (p = 0.037), use of iodized household salt (p = 0.007), and stronger maternal social support (p = 0.003). Bilingualism was associated with lower VIQ (p = 0.001) and being the first born was associated with higher VIQ (p = 0.012).

Discussion
The main objective of the present study was to investigate the association between neonatal TSH concentration and cognitive development at preschool age in Belgium, where MID during pregnancy is frequent, taking into account confounding factors. Children with a neonatal TSH above 5 mIU/L did not have a higher risk of an impaired cognition compared to those above this cut-off point. Major factors influencing IQ status in preschool children were gender, parity, child dietary supplement intake, use of iodized salt within the household, household income, mother's education level, mother's social support, mother's vitality, and child bilingualism.
Data on the association between neonatal TSH concentration, either measured in cord blood or between the 3rd and 5th day of life, and cognitive development [10][11][12] are limited and contradictory. An Italian case-controlled study has found lower Full Scale and Performance cognitive scores in a small cohort (N = 18) of 6 to 9 year old children with transient congenital hypothyroidism or hyperthyrotrophinaemia measured in cord blood at birth (N = 9) in comparison with control group (N = 9) [10]. In Spain, a retrospective cohort study (N = 61) showed that children aged 3 years with an elevated TSH level measured 3 days after birth (between 5-10 mIU/L) had lower scores in Perceptual Performance Skills, Memory Development, and General Cognitive Index (N = 13) than those with TSH level below 5 mIU/L (N = 48) [11]. These two studies were performed with a rather small sample in which the confounding factors were not assessed with statistical analysis. A more recent Spanish retrospective cohort study which evaluated 178 boys aged 3 years old born between 2000-2002, recruited from a general population, found that the TSH in cord blood (TSH range 0.24-17 mIU/L) was significantly negatively correlated with general cognitive scores and quantitative scores [12]. This study, although including a larger sample size and assessing confounding factors, still needs to be criticized because it only presented results for boys without explaining the rationale behind this.
Our study corroborates with another study conducted in Scotland reporting no evidence for an association between TSH cord blood level and child neurodevelopmental scores at 5 years old (N = 97) [20]. Another study including 500 children found no association between variation in newborn T4 concentration and cognitive scores at 6 months and 3 years old [21].
In the present study, the sampling was done according to a sample size calculation, stratified by TSH levels and by gender to ensure that the whole range of TSH values were included in the study. In addition, cognitive outcomes of the children were tested by blinded psychologists using a validated test and before entering primary school, thereby reducing the time between TSH measurement and cognitive testing. However, the use value of neonatal TSH as a parameter for MID during late pregnancy has recently been questioned. Indeed, the actual cut-off value of 5 mIU/L was not sensitive enough to detect MID in several studies [22][23][24]. On the other hand, several studies have investigated the effect of MID during pregnancy using maternal urinary iodine excretion (UIE) concentration as an indicator. A UK study showed that lower UIE is related to a higher risk of impaired verbal abilities [25]. The Dutch Generation R Study found an association between low urinary iodine during pregnancy and impaired executive functioning in children aged 4 years old [26]. In contrast, the Dutch Generation R study, including 1525 children aged 6 years old, found no relationship between maternal low urinary iodine concentrations before 18 weeks of pregnancy and cognitive scores [27].
Future studies aiming to assess the effect of MID during pregnancy on child neurodevelopment should adopt a prospective design and use a more reliable indicator of maternal iodine deficiency during pregnancy than elevated TSH at birth; interesting alternatives could be maternal thyroid parameter or UIE from the first trimester of pregnancy. Indeed, neonatal TSH only indicates iodine status during late pregnancy and many studies showed an impact of MID on child cognitive development when appearing before the first half of pregnancy (12 or 18 weeks) [28][29][30][31][32], before the fetal thyroid gland reaches its functional maturity, and at a time when the fetus is still entirely dependent on the thyroid hormones provided by the mother.
Importantly, more studies are needed to understand the potential beneficial/harmful effect of iodine supplementation for iodine deficient pregnant women on child neurodevelopment. Indeed, relatively little is known about the positive effect of iodine supplementation in the area of MID, only two studies have found a positive effect of potassium iodine intake during pregnancy on children's cognitive [33] and psychomotor [34] development while one study showed a deleterious effect of iodine supplementation on psychomotor development [35]. Finally, a recent randomized controlled trial found no beneficial effect of levothyroxine supplementation for pregnant women on cognitive development of the 3 year old offspring [36].
The actual cut-off used in Belgium by newborn screening for congenital hypothyroidism is 15 mU/L. The benefits of lowering this cut-off for preventing future intellectual disabilities of children are controversial [37]. Since we did not find a significant association between the highest TSH levels groups and lower IQ scores when adjusted for covariates, our results do not support a lowering of the actual TSH cut-off used in Belgium.
In our study, we found that iodized salt consumption has a protective effect on IQ performance scores of children thereby showing the importance of adequate nutritional iodine intake for child neurodevelopment. Our study showed that a various range of factors influence IQ scores such as maternal education, household income, maternal mental health, and social support or child bilingualism. These findings emphasize the importance of taking potential confounding factors into account when trying to assess the role of perinatal factors on later cognitive development because the latter may play a bigger role than the former.
The study limitations include that we used a retrospective design and the fact that even if our sample size was large with respect to one of the previous studies, we calculated that for a sample size of 311, with an error alpha of 0.05 and a power of 0.95, the minimum IQ score difference that we were able to detect was a difference of 4.3 points. Another limitation is that we studied a highly educated population (73% of the mothers had a university degree) that could have biased the results. Additionally, we could have also measured iodine status and thyroid newborn T4 and T3 to obtain all the information about newborn thyroid functioning.
The lack of association between neonatal TSH within the range of 0 to 15 mIU/L, used as a surrogate marker for MID during pregnancy, and cognitive impairment at preschool age suggests that the current level of iodine deficiency in Belgium is probably not severe enough to affect the neurodevelopment of children.
Acknowledgments: This work was supported by the Belgian Federal Science Policy Office and the Fonds de la Recherche Scientifique Medicale (Grant number: 3.4572.11). The funding sources did not have any role in the design of the study, data analysis of results, or preparation of the manuscript. The authors acknowledge the Brussels New-born Screening Centre for Metabolic Disorders of the ULB for providing the screening data. The authors acknowledge all the children and mothers who took part in the study. All authors helped in the interpretation of the results and writing of the manuscript. All authors have approved the final version of the manuscript.
Author Contributions: All authors contributed to the conception and the design of the study, the data interpretation, and the revision of the article. Caroline Trumpff wrote the manuscript, performed the data collection and the data analysis. All the authors approved the final version of the article.

Conflicts of Interest:
The authors declare that they have no conflict of interest that could prejudice the impartiality of the research reported.