Associations between newborn thyroid-stimulating hormone concentration and neurodevelopment and growth of children at 18 months

Abstract

The study aimed to assess the associations between newborn thyroid-stimulating hormone (TSH) concentration, a marker of iodine nutrition in early life, and childhood neurodevelopment and growth using data collected from two pregnancy studies, one in a borderline iodine-deficient setting (DHA to Optimize Mother Infant Outcome (DOMInO) Study) and one in an iodine-sufficient setting (Pregnancy Iodine and Neurodevelopment in Kids (PINK) Study). TSH data were obtained from routine newborn screening. Neurodevelopment was assessed at 18 months using the Bayley Scales of Infant and Toddler Development, third edition (Bayley-III). Weight, height and head circumference were measured at 18 months. In total, 1467 children were included in the analysis. Comparing the highest with the lowest TSH quartile, the mean differences (MD) in the Bayley-III scores ranged from −2·0 (95 % CI −4·7, 0·7) to −2·2 (95 % CI −5·8, 1·3) points in DOMInO and 1·0 (95 % CI −1·6, 3·6) to 2·0 (95 % CI −0·4, 4·4) points in PINK in the cognitive, language and motor scales; the MD in the anthropometric z scores ranged from −0·01 (95 % CI −0·5, 0·5) to −0·5 (95 % CI −0·9, −0·1) in both studies. A 1 mIU/l increase in TSH was associated with −0·3 (95 % CI −0·9, 0·2) point and 0·2 (95 % CI −0·3, 0·7) point changes in the mean cognitive score in the DOMInO and PINK, respectively. A null association between TSH and growth was also observed in both studies. Longitudinal studies that utilise newborn TSH data and examine neurodevelopmental outcomes at later ages are warranted, as neurodevelopmental assessments in older children are more predictive of later achievement.

Iodine is an integral part of thyroid hormone, which in turn is required for normal brain development and physical growth during the early critical periods of life⁽¹⁾. Insufficient maternal iodine intake in pregnancy has been linked to poor growth^(2,3) and impaired neurodevelopment in children^(3–5) and can result in cretinism if iodine deficiency in pregnancy is severe⁽³⁾. Because of the concern of potential adverse effects on offspring, iodine supplementation in pregnancy has been recommended in many countries, including countries with sustained programmes to combat iodine deficiency like Australia with mandatory iodine fortification of bread⁽⁶⁾. However, the WHO technical consultation does not recommend iodine supplementation during pregnancy in iodine-sufficient populations with sustained programmes to combat iodine deficiency⁽⁷⁾. Evidence from randomised controlled trials suggested that iodine supplementation during pregnancy improved child neurodevelopment and growth in severe iodine deficiency settings^(2,8) , but there is lack of benefit in mild iodine deficiency to iodine sufficiency settings based on limited data^(2,9,10) . There is a concern of excess iodine intake in pregnant women from iodine-sufficient settings⁽¹¹⁾. Excessive iodine intake in pregnancy has been linked with subclinical hypothyroidism and isolated hypothyroxinaemia^(11,12) .

Both iodine deficiency^(13–15) and excess^(16–18) in pregnancy were associated with elevated newborn thyroid-stimulating hormone (TSH) concentration. Newborn TSH has been suggested as a marker of iodine nutrition in pregnancy⁽¹⁹⁾, and it may be useful to identify children at risk of neurodevelopmental or growth delay in countries where newborn screening is routinely practiced.

There are several studies that assessed the association between newborn TSH concentration and childhood neurodevelopment. Studies in iodine-deficient populations showed that newborns with higher TSH concentrations of ≥5 mIU/l^(20–22) had an increased risk of developmental delay, but the association between newborn TSH and neurodevelopmental scores was inconsistent in iodine-sufficient populations^(23–26). Most published studies had methodological limitations like inadequate adjustment for potential confounders^(20,22) , small sample sizes of <30 participants⁽²⁷⁾ or subjective assessment of the outcome^(23,24) . Only a few studies have examined the association between newborn TSH concentration and childhood growth^(23,28,29) .

The quality of the studies examining the relationship between newborn TSH concentration and neurodevelopment or growth of children was limited, and the findings were inconsistent. The objective of the current study was to investigate the associations between newborn TSH concentration and neurodevelopment and growth of children at 18 months of age utilising data collected from two large pregnancy studies in Australia, one in a borderline iodine-deficient setting before the implementation of the mandatory iodine fortification of bread and the other in an iodine-sufficient setting after fortification, with adjustment for key confounding variables.

Methods

Study design, setting and participants

Participants were children from two pregnancy studies conducted in the same area of Australia: the DHA to Optimize Mother Infant Outcome (DOMInO) trial⁽³⁰⁾ and the Pregnancy Iodine and Neurodevelopment in Kids (PINK) study⁽⁴⁾. Children in the DOMInO study were born before the implementation of the mandatory iodine fortification of bread (born between 2005 and 2008), when the general population was defined as borderline iodine deficient (defined by the median urinary iodine concentration < 100 µg/l), and children in the PINK study were conceived and born after the implementation of the mandatory fortification (born between 2011 and 2013), when the population was defined as iodine sufficient (median urinary iodine concentration ≥ 100 µg/l). Detailed descriptions of DOMInO and PINK are reported elsewhere^(4,30) . Briefly, DOMInO was a multicentre randomised controlled trial that investigated the effect of DHA supplementation during pregnancy on postnatal depression and childhood cognitive and language development. Neurodevelopment and growth were assessed in a subset of children at 18 months of age (n 724), consisting of a random sample of children born at term (n 628) and all children born preterm (n 96) from Adelaide, South Australia. The overall results of the trial showed DHA supplementation during pregnancy had no effect on the neurodevelopmental scores of children at 18 months, though the DHA treatment group had a lower risk of cognitive delay⁽³⁰⁾. PINK was a prospective cohort study investigating the relationship between maternal iodine intake during pregnancy and the neurodevelopment of children at 18 months of age. The results of the PINK study showed that maternal iodine intake during pregnancy in the lowest (<220 μg/d) or highest (≥391µg/d) quartiles was associated with poorer Bayley Scales of Infant and Toddler Development third edition (Bayley-III) scores when compared with maternal iodine intake in the middle quartiles⁽⁴⁾. The inclusion criteria were similar for both the DOMInO and PINK studies: healthy pregnant women <21 weeks of gestation attending antenatal clinics were eligible to take part in the studies. Women with known fetal abnormalities or the following attributes were excluded in both studies: drug or alcohol abuse, English not spoken at home or unable to give informed consent. Mothers with known thyroid disease were also excluded in the PINK study, while the DOMInO study excluded multiple pregnancies. For compatibility, children from multiple pregnancies in the PINK study were excluded in the analysis in the current study. The study protocols were approved by the institutional review boards at each participating centre in both DOMInO (REC1657/12/2007) and PINK (REC2230/12/15 and 076.10), and each participant provided informed consent.

Assessment of outcomes

Neurodevelopment of children

Neurodevelopment of children was assessed at 18 months of age using the Bayley-III in both studies^(4,30) . The Bayley-III Scale consists of cognitive, language, motor, adaptive behaviour and social emotional scales. The cognitive scale assesses sensorimotor development, exploration, manipulation, object relatedness, concept formation, memory and other aspects of cognitive processing. The language scale assesses both receptive and expressive communications. Both fine and gross motor skills were assessed using the motor scale. The adaptive behaviour and social emotional scales assess parent reported child behaviours⁽³¹⁾.

Growth of children

Weight, length and head circumference were measured at the time of Bayley assessment using the standard procedures of the WHO in both studies⁽³²⁾. Length was measured in a laying/supine position using a measuring board and was recorded to the nearest 0·1 cm. Weight was measured by placing the child on a calibrated weighing scale and was recorded to the nearest 10 g. Head circumference was measured using a non-stretching tape positioned above the eyebrows anteriorly and olecranon fossa posteriorly and was recorded to the nearest 0·1 cm⁽³⁰⁾. Weight-for-age z score, weight-for-length z score, length-for-age z score and head circumference-for-age z score were calculated using the WHO’s Child Growth Standard⁽³³⁾.

Assessment of exposure

Newborn TSH concentration data were extracted from the routine newborn screening database in the South Australian Neonatal Screening Centre⁽³⁴⁾. The TSH concentration was measured from newborn whole blood samples taken between 3 and 4 d of infant’s age by heel prick using the dissociation-enhanced fluoroimmunoassay method. The method used to measure newborn TSH was consistent between the two studies.

Assessment of confounders

The potential confounders of child neurodevelopment and growth were identified based on the literature and are presented in Supplementary Figs S1 and S2. Socio-economic status data, including parental education and occupation, were collected using a structured self-report questionnaire. The Home Screening Questionnaire (HSQ) was completed when the parent attended the 18-month Bayley-III appointment to assess the level of cognitive, social and emotional support available to the child in the home environment⁽³⁵⁾. The HSQ score was also used as a proxy for overall maternal care and support during pregnancy⁽³⁶⁾. Pregnancy-related background data on smoking, alcohol consumption, multivitamin intake and history of depression were collected during pregnancy by self-report. Mother’s BMI was calculated using maternal weight and height measured at enrolment. Information on caesarean-section delivery, gestational age at birth, 5-min Apgar score (a proxy for perinatal stress) and birth weight were retrieved from a case note audit at birth. Birth weight-for-gestational age z score was calculated using the Australian national birth weight percentiles by gestational age data⁽³⁷⁾.

Statistical analysis

The data were analysed using Stata 15 (Statacorp LP). Multiple imputation using chained equations was performed to handle missing information. The proportion of missing data in the variables included in the analysis ranged from 0·2 to 18·5 %. The pattern of missingness was non-monotonic, and the data were assumed to be missing at random⁽³⁸⁾. A total of twenty imputed data sets were generated. The multiple imputation model included the exposure variable, outcome variables, auxiliary variables that affect missingness of the imputed variables or correlate with the imputed variables (age at TSH assessment, hospital of enrolment, maternal height and weight) and potential confounders (parity, ethnicity, occupation, education, smoking, alcohol consumption, supplement intake, depression, mode of delivery, mother’s age, gestational age, sex of the child, Apgar score, HSQ score and birth weight-for-gestational age z score). The newborn TSH in quartiles and the binary variables (developmental delay, growth delay and dichotomised TSH as ≤5 mIU/l v. >5 mIU/l) were created before the imputation. We compared these variables descriptively between the observed data set and imputed data set in the DOMInO and PINK studies, and they were similar in both studies (see online Supplementary material, Supplementary Table S1). We analysed trend across categorical variables using a non-parametric test that is an extension of the Wilcoxon rank-sum test. All analyses were performed on the imputed data sets.

The primary analyses were performed on the DOMInO and PINK data separately for several reasons. First, the two studies were conducted over different time periods where the iodine status of the population differed. DOMInO was conducted in the borderline iodine-deficient setting before the implementation of mandatory iodine fortification of bread and recommendation of routine iodine supplementation for pregnant women, while PINK was conducted in the iodine-sufficient setting after those national iodine interventions. Second, the two studies may not be directly compatible due to possible changes in the demographic and economic characteristics of the general population or the type of antenatal care over time. Third, while the participants were from the same geographic area in Australia, the design of the studies differed (DOMInO was a randomised controlled trial, while PINK was a cohort study). Finally, the direction of the association between TSH and neurodevelopment differed in the two studies. Meta-analysis was performed to examine the overall effect combining both studies.

To assess the association between newborn TSH and Bayley-III and growth outcomes, the exposure variable newborn TSH was treated as a continuous variable and categorised into quartiles in separate analyses. The newborn TSH distribution differed between DOMInO and PINK studies, and hence different TSH quartile cut-offs were applied in each study. Sensitivity analyses were conducted to examine the robustness of the main findings by examining newborn TSH in tertiles, or quintiles, or applying the DOMInO TSH quartile cut points to PINK and vice versa, and using the TSH corrected for infant’s age at blood sampling (TSH∼age) because infants in the highest TSH quartile were sampled earlier on average than infants in the lowest TSH quartile. It is well recognised that the age at sampling affects newborn TSH concentration due to the physiological surge in newborn TSH concentration for early infant blood sampling⁽³⁹⁾. The TSH∼age variable was generated by adding the mean TSH to the residuals from the regression model of TSH on age at blood sampling. TSH and age at blood sampling were log-transformed for the analysis and back-transformed to the original scales^(40,41) . TSH was also dichotomised based on the commonly reported categories in literature (>5 mIU/l v. ≤5 mIU/l) to facilitate comparison with previous studies⁽⁴²⁾.

There are no well-designed studies on which to base power calculations for the assessment of childhood neurodevelopment and growth outcomes between children in the lowest and highest TSH quartiles. However, a five-point difference on average in the cognitive and psychomotor development between children with adequate v. deficient iodine or iron intake was considered clinically significant^(4,43) . With the sample sizes available in DOMInO (n 724) and PINK (n 743), a mean Bayley-III score of 100 and a sd of 15, the studies have 88 % and 90 % power to detect a difference of five points in the Bayley-III scale between children in the highest and lowest TSH quartiles.

The neurodevelopment and growth outcomes of children were analysed as both continuous and categorical outcomes. For the categorical outcomes, developmental delay was defined as Bayley-III scores below 85 and growth delay was defined as anthropometric z scores below −1 (i.e. more than 1 sd below the mean).

A linear regression model was used to analyse continuous outcomes, and a log Poisson regression model with robust variance estimation was used to analyse binary outcomes⁽⁴⁴⁾. Both unadjusted and adjusted analyses were performed, with adjusted analyses considered to be the primary analysis. In DOMInO, we adjusted for DHA treatment group to investigate potential confounding as DHA may influence child development⁽⁴⁵⁾, though there is no known biological mechanism for an effect of DHA supplementation on iodine metabolism and newborn TSH concentration. We also stratified the analysis by the treatment group in the DOMInO study. The following confounders were controlled in the adjusted analyses for neurodevelopmental and growth outcomes in both DOMInO and PINK: parity, ethnicity, parents’ occupation, parents’ education, maternal smoking and alcohol consumption during and before pregnancy (yes or no), maternal multivitamin intake (yes or no) during pregnancy, maternal depression during and before pregnancy (yes or no), mother’s BMI, HSQ score, delivery by caesarean section (yes or no), mother’s age, gestational age at birth, 5-min Apgar score, sex and birth weight-for-gestational age z score. The mean differences in Bayley-III scores with the 95 % CI estimated from the linear regression model and the relative risks (RR) and 95 % CI estimated from the log Poisson regression model were reported to indicate the magnitude and direction of the associations. In line with current best practice^(46–49), we have not used P-values to interpret the associations. The STATA ‘metan’ command was used to perform the meta-analysis⁽⁵⁰⁾. I ² was used to evaluate heterogeneity between the two studies. Heterogeneity between the results was classified as low (I ²: <25 %), moderate (I ²: 25–75 %) and high (I ²: >75 %)^(51,52) . A random effects model was applied due to the heterogeneity between the two studies. The pooled effect sizes were calculated to see the magnitude and direction of the associations between newborn TSH and neurodevelopment and growth. The pooled effect sizes were reported using forest plots.

Results

Maternal and infant characteristics

A total of 724 children in DOMInO and 743 children in PINK were included in the analysis (Fig. 1). Table 1 shows the characteristics of the DOMInO and PINK participants. The two studies differed in some sociodemographic characteristics. There were more Caucasian women, smokers and alcohol consumers in DOMInO compared with PINK. As a result of the study design that included a subset of randomly selected term infants and all preterm infants⁽³⁰⁾, there was a higher percentage of preterm infants in DOMInO. More PINK mothers completed secondary education and took multivitamin supplements during pregnancy compared with DOMInO mothers. The two studies had similar mean HSQ scores and birth weight-for-gestational age z scores.

Fig. 1. Selection of participants in the DHA to Optimize Mother Infant Outcome (DOMInO) and Pregnancy Iodine and Neurodevelopment in Kids (PINK) studies. Bayley-III, Bayley Scales of Infant and Toddler Development, third edition; TSH, thyroid-stimulating hormone.

Table 1. Characteristics of the participants

(Mean values and standard deviations; numbers and percentages; medians and interquartile ranges)

DOMInO, DHA to Optimize Mother Infant Outcome; PINK, Pregnancy Iodine and Neurodevelopment in Kids; HSQ, home screening questionnaire; GA, gestational age; TSH, thyroid-stimulating hormone; Bayley-III, Bayley Scales of Infant and Toddler Development, third edition.

As shown in Table 1, the median newborn TSH concentration was 1·7 (interquartile range 1·0, 2·9) mIU/l in DOMInO and 2·2 (interquartile range 1·4, 3·3) mIU/l in PINK. The percentage of newborns with TSH concentration >5 mIU/l was 4·6 % in the DOMInO and 5·5 % in the PINK. Newborn heel prick blood samples were collected between 48 and 96 h of age in both studies, and the median age at the TSH assessment was 57·7 h in PINK, which was 1·1 h earlier than in DOMInO. In DOMInO, the median age at sampling was 1 h earlier in preterm than term infants, but the median newborn TSH was comparable between these subgroups (1·8 mIU/l in preterm v. 1·7 mIU/l in term infants).

Thyroid-stimulating hormone concentration and Bayley Scales of Infant and Toddler Development third edition scores at 18 months of age

Table 2 reports the Bayley-III scores of children in the DOMInO and PINK studies. The mean cognitive score was 102 (sd 12) points in DOMInO compared with 97 (sd 13) points in PINK, and the percentage of children with cognitive delay was 4·9 % in DOMInO compared with 9·2 % in PINK. Children in the highest TSH quartile had the lowest mean Bayley-III scores in the DOMInO study but the highest mean scores in PINK study when compared with the other TSH quartiles.

Table 2. Bayley Scales of Infant and Toddler Development, third edition (Bayley-III) outcomes at 18 months of age by newborn thyroid-stimulating hormone (TSH) concentration in quartiles

(Mean values and standard deviations; numbers and percentages)

DOMInO, DHA to Optimize Mother Infant Outcome; PINK, Pregnancy Iodine and Neurodevelopment in Kids.

* TSH was categorised into quartiles: quartile 1: <1·1 mIU/l; quartile 2: 1·1–1·7 mIU/l; quartile 3: 1·8–2·9 mIU/l; quartile 4: ≥3·0 mIU/l.

† TSH was categorised into quartiles: quartile 1: < 1·5 mIU/l; quartile 2: 1·5–2·2 mIU/l; quartile 3: 2·3–3·3 mIU/l; quartile 4: ≥3·4 mIU/l.

‡ Developmental delay was defined as Bayley-III scores below 85.

§ P values from the trend test using the extension of the Wilcoxon rank-sum test.

There was a null association between newborn TSH and neurodevelopment. For example, when TSH was treated as a continuous exposure (Table 3 and Supplementary Fig. S3). A 1 mIU/l increase in TSH was associated with a –0·3 (95 % CI –0·9, 0·2) point change in the mean cognitive score in DOMInO, compared with a 0·2 (95 % CI –0·3, 0·7) point change in PINK, and a RR of 1·1 (95 % CI 0·9, 1·3) for cognitive delay in DOMInO compared with a RR of 0·9 (95 % CI 0·7, 1·0) in PINK. When TSH was categorised into quartiles, the mean difference for children in the highest TSH quartile compared with the lowest quartile was –2·1 points (95 % CI –4·8, 0·6) for the cognitive score in DOMInO compared with 1·0 points (95 % CI –1·6, 3·6) in PINK, and the RR was 1·7 (95 % CI 0·5, 5·6) for cognitive delay in DOMInO compared with 0·5 (95 % CI 0·2, 1·4) in PINK (Table 3).

Table 3. Adjusted associations between newborn thyroid-stimulating hormone (TSH) concentration and Bayley Scales of Infant and Toddler Development, third edition (Bayley-III) outcomes at 18 months of age

(Mean differences (MD), relative risks (RR) and 95 % confidence intervals)

DOMInO, DHA to Optimize Mother Infant Outcome; PINK, Pregnancy Iodine and Neurodevelopment in Kids; Ref, reference category.

* Developmental delay was defined as Bayley-III scores below 85.

† The MD and 95 % CI were estimated with multivariable linear regression model.

‡ The RR and 95 % CI were estimated with multivariable log Poisson regression model with robust variance estimation.

§ The regression models were adjusted for sex, parity, ethnicity, occupation, education, smoking during second trimester, previous smoking, alcohol consumption during second trimester, previous alcohol consumption, supplement intake during second trimester, previous depression, depression during second trimester, mode of delivery, mother’s age, gestational age at birth, 5-min Apgar score, mother’s BMI, home screening questionnaire score and birth weight-for-gestational age z score in both DOMInO and PINK studies. In addition, treatment group in the DOMInO study was also added to the adjusted models when analysing the DOMInO data.

|| TSH is modelled continuously.

¶ TSH was categorised into quartiles separately in DOMInO and PINK. DOMInO quartile 1: <1·1 mIU/l; quartile 2: 1·1–1·7 mIU/l; quartile 3: 1·8–2·9 mIU/l; quartile 4: ≥ 3·0 mIU/l. PINK quartile 1: <1·5 mIU/l; quartile 2: 1·5–2·2 mIU/l; quartile 3: 2·3–3·3 mIU/l; quartile 4: ≥3·4 mIU/l.

A similar pattern of null association was observed between newborn TSH and neurodevelopment in both studies regardless of whether TSH was categorised into tertiles (see online Supplementary material, Supplementary Table S2), quintiles (see online Supplementary material, Supplementary Table S3) or dichotomised at 5 mIU/l (see online Supplementary material, Supplementary Table S4), or when the quartile cut points from the PINK study were applied to DOMInO and vice versa (see online Supplementary material, Supplementary Table S5), or when TSH was corrected for age at sampling (see online Supplementary material, Supplementary Table S6), or the analyses stratified by the treatment group in the DOMInO study (see online Supplementary material, Supplementary Table S7). Null association was also observed in the unadjusted analyses (see online Supplementary material, Supplementary Table S8).

Thyroid-stimulating hormone concentration and growth at 18 months of age

The mean weight-for-length z score, length-for-age z score, weight-for-age z score and head circumference-for-age z score scores were mostly greater than zero for both DOMInO and PINK, indicating children were larger on average than the populations used to define the WHO growth standards (see online Supplementary material, Supplementary Table S9). The percentage of children with z scores < –1 ranged from 3·5 % to 15·5 % in the DOMInO and 3·8 % to 15·2 % in the PINK studies (see online Supplementary material, Supplementary Table S9).

The associations between newborn TSH and growth outcomes were similar in both PINK and DOMInO. Null associations were observed between newborn TSH and growth outcomes in both studies regardless of whether TSH was treated as a continuous variable or categorised into quartiles (Table 4), tertiles or quintiles (data not shown) or when TSH was dichotomised at 5 mIU/l (see online Supplementary material, Supplementary Table S10), or when TSH was corrected for age at sampling (see online Supplementary material, Supplementary Table S11). A similar pattern of null association was also observed in the unadjusted analyses (see online Supplementary material, Supplementary Table S12).

Table 4. Adjusted associations between newborn thyroid-stimulating hormone (TSH) and growth outcomes at 18 months of age

(Mean differences (MD), relative risks (RR) and 95 % confidence intervals)

DOMInO, DHA to Optimize Mother Infant Outcome; PINK, Pregnancy Iodine and Neurodevelopment in Kids; WLZ, weight-for-length z score; LAZ, length-for-age z score; Ref, reference category; WAZ, weight-for-age z score; HCZ, head circumference-for-age z score.

* Growth delay was defined as z scores below 1 sd.

† The MD and 95 % CI were estimated with multivariable linear regression model.

‡ The RR and 95 % CI were estimated with multivariable log Poisson regression model with robust variance estimation.

§ The regression models were adjusted for sex, parity, ethnicity, occupation, education, smoking during second trimester, previous smoking, alcohol consumption during second trimester, previous alcohol consumption, supplement intake during second trimester, previous depression, depression during second trimester, mode of delivery, mother’s age, gestational age at birth, 5-min Apgar score, mother’s BMI, home screening questionnaire score and birth weight-for-gestational age z score in both DOMInO and PINK studies. In addition, treatment group in the DOMInO study was also added to the adjusted models when analysing the DOMInO data.

|| TSH is modelled continuously.

¶ TSH was categorised into quartiles in both DOMInO and PINK. DOMInO quartile 1 (lowest): <1·1 mIU/l; quartile 2: 1·1–1·7 mIU/l; quartile 3: 1·8–2·9 mIU/l; quartile 4 (highest): ≥3·0 mIU/l. PINK quartile 1 (lowest): <1·5 mIU/l; quartile 2: 1·5–2·2 mIU/l; quartile 3: 2·3–3·3 mIU/l; quartile 4 (highest): ≥3·4 mIU/l.

Meta-analysis combined the DHA to Optimize Mother Infant Outcome and Pregnancy Iodine and Neurodevelopment in Kids studies

As shown in online Supplementary Figs. S4 and S5, there was high heterogeneity between the DOMInO and PINK studies in the developmental outcomes, but low heterogeneity in the growth outcomes (see online Supplementary material, Supplementary Figs. S6 and S7) with null associations in all meta-analyses.

Discussion

Our study showed a null association between newborn TSH and both childhood neurodevelopment and growth in the DOMInO and PINK studies. However, children in the highest TSH quartile had the lowest mean Bayley-III score in the DOMInO study, whereas the opposite was observed in the PINK study where children in the highest TSH quartile had the highest mean score. We cannot rule out a different pattern of association between newborn TSH and neurodevelopmental outcomes in the two studies because of the wide CI around the effect estimates, as well as possible residual confounding. Similarly, we cannot rule out a poorer growth outcome of children in the highest TSH quartile compared with children in the lowest quartile in both studies.

Newborn TSH as an indicator of iodine nutrition during pregnancy may depend on maternal iodine intake, as either too little^(13–15) or too much^(16,17) iodine during pregnancy may lead to high TSH. Excess iodine intake may also lead to lower TSH in iodine-induced hyperthyroidism during pregnancy⁽⁵³⁾. DOMInO was conducted in a borderline iodine-deficient population before the mandatory iodine fortification of bread in Australia, and newborns with TSH in the highest quartile were likely be due to suboptimal maternal iodine nutrition in pregnancy. In studies from iodine-deficient populations, like DOMInO, high newborn TSH may be due to inadequate maternal iodine intake during pregnancy. Iodine deficiency during pregnancy can lead to lower maternal-free thyroxine concentration⁽⁵⁴⁾ and shifts the intra-thyroidal formation of thyroxine to the more active metabolite tri-iodothyronine^(16,55) . This leads to a lower transfer of thyroid hormones, primarily thyroxine, to the fetus and may cause impaired neurodevelopment in early childhood. Although a majority of studies conducted in iodine-deficient populations showed a negative association between newborn TSH and childhood neurodevelopment^{(20,22,56,57)} , all except one study⁽²⁰⁾ had a small sample size (ranged from 61 to 250 participants) and did not account for key confounders including maternal socio-economic status, gestational age at birth and home environment. Our finding of a null association in DOMInO was consistent with a large study (n 691) in a mildly iodine-deficient population that controlled for key confounders including maternal socio-economic status and infants age at blood sampling for TSH assessment⁽⁵⁸⁾. Similar to DOMInO, Murcia et al. ⁽⁵⁸⁾ used Bayley scales to assess neurodevelopment of offspring under 2 years of age. Participants in DOMInO and Murcia et al. ⁽⁵⁸⁾ were healthy children from population-based cohorts unlike a study in Italy from hospitalised children⁽⁵⁶⁾ where the cause for a higher TSH may differ. Furthermore, outcome assessors were not blinded to the exposure variable and maternal characteristics in a prospective cohort study by Costeira et al. ⁽⁵⁷⁾, which may lead to bias. All of the studies in iodine-deficient populations assessed neurodevelopment in toddlers and pre-schoolers except a large Australian study⁽²⁰⁾ that assessed educational performance at a median age of 10 years. In that Australian study⁽²⁰⁾, infants without congenital hypothyroidism whose TSH concentration was above the 99th percentile (TSH 12·4–20 mIU/l) had a higher risk of poorer school achievement than infants with TSH < 75th percentile (TSH concentration ≤ 4 mIU/l). Neurodevelopmental assessments at later ages may be more stable and a better indication of cognitive function at later ages⁽⁵⁹⁾.

PINK was conducted in an iodine-sufficient setting after mandatory iodine fortification in a previously iodine-deficient population⁽⁶⁰⁾. In contrast to the DOMInO study, the higher TSH concentration in PINK may be due to a high maternal iodine intake as a result of concurrent mandatory iodine fortification and routine iodine supplementation. This is partially explained by the acute Wolff–Chaikoff effect that excess iodine intake in a previously iodine-deficient population temporarily inhibits synthesis of thyroid hormones and resulted in elevated TSH⁽¹⁶⁾. Our finding of a null association between newborn TSH and Bayley-III scores in PINK study is consistent with the PsychoTSH study (n 315)^(24–26) in an iodine-sufficient population in Belgium, which is a prospective cohort study designed to assess the association between newborn TSH and neurodevelopmental outcomes in pre-schoolers. Similar to PINK, children who participated in the PsychoTSH study were born during a transitional period when Belgium introduced a voluntary programme to use iodised salt in bread making in 2009⁽⁶¹⁾, and the higher TSH may be due to the transient excess in iodine intake. Although the PsychoTSH study reported no association between TSH level and most of the neurodevelopmental scales^(24–26), it showed higher mean verbal IQ scores in children with TSH concentration in the range 5–9 mIU/l compared with children whose TSH concentration < 5 mIU/l⁽²⁶⁾. We cannot rule out a higher neurodevelopmental score in children with TSH in the highest quartile (median TSH of 4·5 mIU/l) in the PINK study due to the wide CI around the estimates. In contrast, though not directly compatible to our study, a smaller (n 250) prospective cohort study by Cuestas et al. ⁽²³⁾ in an iodine-sufficient population in Argentina showed a higher odds of developmental delay in school-age children with newborn TSH ≥ 10 mIU/l compared with TSH < 10 mIU/l at birth; however, the developmental outcome assessment was by parents’ evaluation of developmental status, and TSH was sampled earlier at 2–3 d of age outside the WHO recommendation⁽⁴²⁾. Prospective cohorts or population data linkage studies using exposure data from newborn screening and cognitive function at later ages may help to evaluate long-term effect of the newborn TSH concentration on cognition.

The newborn TSH level is influenced by the timing of blood sampling due to physiological surge within 48 h after birth. Higher TSH level may also be a reflection of earlier blood sampling in newborn screening⁽³⁹⁾. Although the median age at blood sampling was earlier in PINK than DOMINO overall, children in the highest TSH quartile had an earlier blood sampling compared with the other quartiles in both the DOMInO and PINK studies. Therefore, age at the newborn TSH assessment is unlikely to explain the different patterns of associations observed between newborn TSH and neurodevelopmental outcomes in the two studies.

There is only one study (n 250)⁽²³⁾, which examined the association between newborn TSH in full-term infants and childhood growth, which was conducted in Argentina, an iodine-sufficient population. It showed no difference in the weight and height at 6 years of age between children with newborn TSH ≥ 10 mIU/l and those with newborn TSH < 10 mIU/l at birth though no adjustment was made for potential confounding variables. While the Argentina study is not directly compatible to our study, both suggested no difference in the growth of children between newborns with high or low TSH in populations of mild iodine deficiency or iodine sufficiency.

The WHO defines iodine deficiency in populations when >3 % of newborns have TSH concentration >5 mIU/l⁽⁴²⁾. Several studies examining the association between newborn TSH and developmental outcomes dichotomised TSH into ≤5 mIU/l or >5 mIU/l^(22,24) , but it is important to note that TSH > 5 mIU/l alone is not a criterion in classifying iodine deficiency either in populations or individuals. Further research is warranted to establish a newborn TSH cut-off associated with impaired neurodevelopment to identify children at risk of developmental delay for monitoring and early intervention as appropriate.

Our results should be interpreted with caution because our study is a secondary analysis utilising data from the DOMInO and PINK studies that differed in the distribution of several confounders. In the DOMInO study, for instance, there was an over-representation of preterm infants due to the study design. Although preterm birth was adjusted for in all models, and Bayley-III scores were assessed based on the child’s corrected age, preterm birth may increase TSH concentration and confound the association. The association did not differ when the analysis was restricted to full-term infants (data not shown). The timing of newborn screening was earlier in PINK, as there was a shift towards earlier discharge from hospitals after birth. We attempted to address this through correcting TSH for infant’s age at blood sampling but residual confounding may still present. Furthermore, iodine status of the DOMInO cohort was inferred from the general population as maternal UIC in pregnancy was not assessed in the DOMInO study. Despite these limitations, our study had adequate statistical power, the outcome assessments were blinded and confounders were adjusted in analyses. The possible different patterns of the association between newborn TSH and developmental outcomes in an iodine-deficient v. iodine-sufficient setting highlights the importance of considering iodine status of the study population when examining and interpreting the relationship.

Conclusion

Our study shows a null association between newborn TSH and both childhood neurodevelopment and growth. However, we cannot exclude the possibility of poorer neurodevelopment or growth in infants with high TSH in a borderline iodine-deficient setting, and better neurodevelopment in infants with high TSH up to 5 mIU/l in an iodine-sufficient setting, due to the wide CI around the estimated effects. Follow-up studies that utilise newborn TSH data obtained from routine newborn screening and examine neurodevelopmental outcomes at later age are warranted to evaluate whether newborn TSH predicts long-term neurodevelopmental outcome.