Objective Congenital hypothyroidism (CHT) is a common cause of preventable mental retardation, and the quantification of intellectual disability due to CHT is needed to assess the public health benefit of newborn screening.
Design Review of published studies conducted among children born prior to the introduction of newborn screening for CHT and reporting cognitive test scores.
Setting Population-based studies.
Patients Children with clinically diagnosed CHT.
Interventions Thyroid hormone substitution.
Main outcome measures Intelligence quotient (IQ) (mean and distribution).
Results The prevalence of recognised CHT rose from one in 6500 prior to screening to approximately one in 3000 with screening. In four population-based studies in high-income countries, among children with clinically diagnosed CHT 8–28% were classified as having intellectual disability (defined as an IQ <70) and the mean IQ was 85 (a leftward shift of 1 SD). Among children with subclinical CHT, the risk of overt intellectual disability was lower (zero in one study), but decreased intellectual potential and increased behavioural abnormalities were documented.
Conclusions Although the prevalence of overt disability among children with CHT in the absence of screening may be less than previously estimated, the preventable burden of intellectual disability due to CHT is substantial and justifies newborn screening. However, changes in existing newborn screening protocols to capture more cases are unlikely to prevent overt cases of disability and should therefore be justified instead by the documentation of other benefits of early detection.
Statistics from Altmetric.com
Newborn screening and treatment for congenital hypothyroidism (CHT), an endocrine disorder that results from the defective development or function of the thyroid gland in utero, constitute an important public health success.1,–,3 Prior to screening, numerous children diagnosed as having CHT in high-income countries were diagnosed as having mental retardation,4 or learning or intellectual disability in current UK or US terminology. Among cohorts of children with CHT diagnosed by screening, the percentage with intellectual disability is generally reported to be 1% or less.5,–,8 Numerous authors have estimated the economic benefit of newborn screening for CHT based on the prevention of intellectual disability.9,–,14 In addition to cognitive impairment, CHT is associated with physical growth retardation, which is readily preventable, although this is not a primary justification for newborn screening.
What is already known on this topic
▶ Congenital hypothyroidism (CHT) is the most common preventable cause of intellectual disability in children.
▶ Early and appropriate treatment following biochemical screening has resulted in the disappearance of intellectual disability (defined as an intelligence quotient <70) from CHT in countries with universal newborn screening.
▶ Newborn screening for CHT is cost-saving because of the prevention of intellectual disability.
What this study adds
▶ The assumption that 40% or more of people with clinically-diagnosed CHT in high-income countries developed intellectual disability in the absence of screening is an overestimate, and the true percentage is probably about 25%.
▶ No evidence is available that children with mild or subclinical CHT who have moderately elevated TSH levels developed overt disability, and it is invalid to extrapolate from outcomes among persons with severe CHT to persons with mild CHT.
▶ Proper assessment of the benefits of newborn screening requires careful, population-based analyses of data on unscreened cohorts that correct for ascertainment bias and consideration of the full range of relevant outcomes.
The relevant policy question in high-income countries is no longer whether to screen for CHT but what criteria to use to screen, diagnose and treat infants with CHT. Screening programmes differ with regard to whether primary screening is done for thyroid stimulating hormone (TSH) or T4, the initial cut-off used to define abnormal values for follow-up testing and the timing of specimen collection. In particular, there has been recent discussion in the UK about the appropriate TSH cut-off.15 Although the UK National Screening Programme Centre recommends that children with TSH <10 mU/l whole blood be classified as normal, the majority of screening laboratories use lower cut-offs, as low as 5 mU/l whole blood.16 Lowering cut-offs or analysing repeat specimens will increase the frequency with which children are followed up, diagnosed and treated for CHT.17 The salient question is whether children who would be detected if screening protocols were modified to refer more children for confirmatory testing would benefit from early detection to the same or a lesser extent than those who meet classical criteria for CHT. There is evidence that children with permanent CHT due to thyroid dyshormonogenesis can have initial TSH values in the 10–15 mU/l range and be missed by screening protocols that set the cut-off at 15 mU/l or higher.18 19 Certainly, treatment of children diagnosed as having mild CHT is routine, but evidence that lack of treatment affects long-term developmental outcomes is lacking.16 20 Indeed, the clinical experience of one of the authors (GVV) is that children with mild CHT missed by newborn screening due to a borderline TSH value appear to have normal school progression.
Experts frequently assume that all children with CHT benefit equally from early detection and treatment, even though there is well-documented evidence that outcomes depend, among other factors, on the initial severity of CHT.21 Observers often equate the number of children detected with CHT or other congenital disorders as a result of newborn screening with the number of lives saved from serious disability or death.22 Such claims do not provide an evidence base for newborn screening policy decisions. The evidence from long-term follow-up studies of screened cohorts suggests that it is relatively severely affected children, with low T4 and/or delayed bone maturation at diagnosis, who are at risk for lower cognitive test scores.23,–,25
The primary justification for newborn screening for CHT is the prevention of cognitive impairment in affected children. In order to assess the benefit of early detection and prompt treatment, it is essential to understand the natural history of CHT or, more precisely, the outcomes that can be expected to occur in the absence of screening. Even in the absence of screening, most children with severe CHT will be recognised clinically and treated beginning in infancy, so one is not really considering the ‘natural’ history of the condition.
The evidence that late initiation of treatment of CHT is harmful for cognitive outcome derives from observations of children who were clinically diagnosed in the absence of screening programmes. However, the prevalence of diagnosed CHT following the introduction of screening has more than doubled, and the spectrum of CHT now includes many children with relatively mild, subclinical CHT who would not have been treated in the absence of screening.
The purpose of this paper is to review the published literature to address the following questions. First, how does the reported frequency of CHT vary depending on the mode of identification? Second, what was the magnitude of intellectual disability among children who were clinically recognised with CHT born prior to newborn screening? Third, among children with subclinical CHT who would not be diagnosed in the absence of screening, what is the evidence regarding cognitive or neurological impairments and disability? Finally, what can be said about the expected health impacts from expanding the numbers of children detected with CHT by modifying screening protocols or the operational case definition of CHT?
Using PubMed and a hand search of references, one of us (SDG) identified population-based surveillance studies of CHT published in English or French and conducted in high-income countries among children born prior to newborn screening for CHT. We identified eight such reports, published between 1976 and 1986, all from northern Europe.26,–,33 A search of publications in EMBASE from 1988 to 2009 did not identify any additional studies from high-income countries.
Among the eight population-based surveillance reports involving unscreened cohorts, five, representing four distinct studies, reported quantitative information on cognitive test scores among unscreened cohorts of children with CHT whose characteristics were described.26,–,30 The primary end points considered in this review are the percentage of children with CHT who have a standardised intelligence quotient (IQ) score <70, the percentage who receive special education due to intellectual disability, and the mean and range of the IQ scores. For epidemiological purposes, an IQ cut-off of approximately 70 is typically used to define intellectual or learning disability (formerly ‘mental retardation’), with a cut-off of 50 to define moderate to severe intellectual disability.34 An individual diagnosis of intellectual disability should entail a functional assessment, but the latter is difficult to standardise. Although an IQ score cut-off is a crude indicator to classify a child as having a disability, confirmation by the receipt of disability services can demonstrate functional impairment. In particular, an observable measure of educational disability is whether children are assigned to special classrooms or schools due to inability to make satisfactory progress in regular classes.
Because IQ scores cannot be reliably assessed in children under 5 years of age, most investigators assess IQ in school-age children. Of the four studies included in the review, three assessed scores in children aged 7 or above, and one, which used the developmental quotient (DQ) rather than IQ, assessed 5-year-old subjects. An additional study was found which reported mean IQ scores for 22 Japanese children with CHT born prior to the introduction of screening, aged 36–66 months, but no information was provided as to how these children were identified.35
Consideration of the entire distribution of IQ scores is likewise important because intellectual disability only represents the tip of the distribution. An environmental insult, such as prenatal or postnatal exposure to a neurotoxin like lead, that shifts the entire distribution of cognitive scores to the left can be considered to cause cognitive impairment to everyone, not just the subset below an arbitrary cut-off. IQ tests are periodically renormed to ensure a mean IQ score of 100 and an SD of approximately 15. A child with a normal IQ score of 100 might have had an IQ score of 110 or even 120 in the absence of the insult.
The populations of the four studies with cognitive assessments are summarised here (see table 2). First, investigators in Sweden conducted a retrospective epidemiological study to identify all children with diagnosed CHT who were born during 1969–1975.26 Alm and colleagues prospectively studied 41 children and obtained consent to evaluate IQ in 39 using the Wechsler Intelligence Scale for Children (WISC) IQ test. Second, a retrospective screening study by Alm and colleagues of 100 239 stored blood spots collected in Sweden during 1977 and 1978, prior to the initiation of screening for CHT, found that 32 specimens tested positive for CHT by TSH screening with a cut-off of 40 mU/l plasma,27 approximately equivalent to 20 mU/l whole blood. That finding yields a CHT birth prevalence of 1 in 3100, consistent with prospectively gathered Swedish newborn screening data. All 32 positive specimens were linked to the records of school-aged children, and 31 of the 32 children could be tracked (one family had emigrated). A review of medical records revealed that 15 had been diagnosed clinically with CHT, for a prevalence of clinical CHT of 1 in 6700. Biochemical testing revealed that an additional seven children had previously undiagnosed hypothyroidism, for a current hypothyroidism prevalence of 1 in 4600. The remaining nine (29%) children had normal thyroid hormone and TSH levels when assessed at age of school entry and, therefore, were likely to have had transient TSH elevation as newborns. The investigators administered the Griffiths test of DQ to 26 of 31 children at 5 years of age.
Two epidemiological studies were conducted in different parts of England among cohorts of children born before newborn screening was introduced. First, in four Thames health authorities in southeastern England, a total of 141 children with hypothyroidism were identified by contacting paediatricians, area health authority handicapped registers and long-stay hospitals, with follow-up. An analysis by Hulse,30 published in 1984, reported data for a total of 122 school-age children of which 99 children ages 7 and above completed the WISC-R IQ test, five of whom had acquired hypothyroidism. Second, a similar epidemiological study was conducted in northern England, which identified a total of 90 children by contacting paediatricians, endocrinologists and general practitioners, and by scrutinising handicap registers and regional statistics records.28 A total of 67 children were available for assessment, of which 45 of suitable age (7–15, median 10) completed the WISC-R IQ test.29
Table 1 provides information from five studies of case finding (surveillance) for clinically detected CHT among geographically defined European populations of children not screened for CHT through blood specimens collected during the first week after birth and from the Swedish ‘retrospective screening’ study. The prevalence of clinically detected CHT reported in those studies was approximately half as high as the prevalence of CHT reported following the introduction of screening programmes in the same populations.26 27 36
At the national level, active case finding studies conducted in Denmark,31 Sweden32 and The Netherlands33 yielded CHT prevalence estimates in a range from 1 in 6100 to 1 in 6900 births, rounded to the nearest 100. Two comprehensive CHT surveillance studies conducted in southern and northern England prior to the introduction of screening reported prevalence estimates of clinically detected CHT of 1 in 6700.28,–,30 One of the two UK studies collected complete data from healthcare providers in Kent and East Sussex in southeastern England and reported a prevalence of 1 in 6700,30 the same figure as in the Swedish retrospective screening study.27
Table 2 summarises population-based studies of unscreened cohorts of children with clinically detected CHT with cognitive assessments that met the inclusion criteria. The most favourable estimates came from a study of 41 Swedish children with clinically detected CHT born during the period 1969–1971 and evaluated at 7–9 years of age.26 The tests revealed that three (8%) of 39 children tested had IQ scores <70. An additional three children had scores in the borderline range from 70 to 74. Three children (8%) received special education for intellectual disability, and an additional six children required services for milder learning disabilities.
In the Swedish retrospective screening study, the Griffiths test of DQ was administered to 26 of 31 Swedish children at 5 years of age who had screened positive for CHT in dried blood spot specimens that had been collected during the first week after birth and placed in storage after testing for phenylketonuria was completed.27 Among the 14 children with CHT whose condition had been clinically detected previously, the mean DQ was 87, and two (14%) children had a DQ <70. None of the six children with undiagnosed CHT had DQ <70, although they had significantly lower mean scores on the performance and motor Griffiths scales (by 20 and 14 points, respectively) in comparison with the six euthyroid children.
The two population-based assessments of children with clinically detected CHT in England reported mean IQ scores among school-age children of 82, assessed with the WISC-R test (table 2). The percentage with IQ <70 was 26.6% in one study30 and not reported in the other.28 29 Of 112 school-aged children with CHT in the first study, 32 (28.6%) attended special schools for learning disability, including three in residential schools.30 In the second study, 13% of children with CHT were reported to attend special schools; an additional 30% attended remedial classes within regular schools.29
The population-based surveillance studies reported relatively small numbers of children with CHT who had an IQ <50. Indeed, a Swedish study reported no IQ scores in that range among 39 children tested.26 The second Swedish study reported one of 14 children with clinically diagnosed CHT with a Griffiths development quotient of 45; all others had scores of 58 or greater.27 In England, seven of 94 children with CHT who were tested had IQ <50.30 The same study reported that two of 132 children of all ages with probable CHT resided in long-term care facilities. In the second English study, one of 45 children with CHT had an IQ score of 40; apparently all others had an IQ >50.29 If one pools the data from the three studies that reported the distribution of IQ scores, 4.5% of children with clinically diagnosed CHT had IQ <50. Given a prevalence of CHT of roughly 1 in 6700 in both countries, this implies that approximately 1 in 150 000 children had moderate-to-profound disability (IQ <50) due to CHT.
Perhaps more important than the frequency of test scores indicative of overt disability, the distribution of cognitive ability was shifted to the left by more than 1 SD among cohorts with clinical CHT born prior to newborn screening. The mean IQ scores in the studies of clinically diagnosed children reported in table 2 ranged from 82 to 88. Because IQ scores are reported to increase over time between periodic renorming of tests,37 the average IQ score can be greater than 100, with mean scores of 105–110 reported in control groups.38 39 Therefore, even children who did not experience overt intellectual disability were seriously affected, with a loss of 17–25 IQ points on average.
Also, even children with subclinical CHT appear likely to experience cognitive impairment, albeit to a lesser extent. Although none of the six children with undiagnosed CH in the Swedish retrospective study had test scores less than 70 or needed special schools, they had significantly lower mean scores in comparison with the six euthyroid children who also had elevated TSH values at birth. Based on those data, children with subclinical CHT experienced an average decrement of about 7 IQ points, roughly one-third the decrement among those with clinically diagnosed CHT.
Long-term follow-up studies have documented excellent outcomes among children with CHT for whom an appropriate dosage of thyroid hormones is established soon after birth.5,–,8,38,–,42 The accurate calculation of the magnitude of benefit resulting from newborn screening requires data on outcomes prior to the introduction of screening. Prior to screening, about 1 in 6500 to 1 in 6900 European children were diagnosed and treated for CHT. Between 8% and 29% of those children developed intellectual or learning disability as indicated by either low IQ scores or requirement for special schooling. The highest percentages come from UK studies with relatively large samples of affected children, and it seems reasonable to assume that approximately 25% of children with permanent, clinically diagnosed CHT experienced overt disability prior to the introduction of newborn screening. This compares with published estimates of 35–40% prevalence of intellectual disability from clinical case series.43 44
In order to calculate the aggregate impact of newborn screening for CHT on disability, it is essential to include in calculations those children with CHT who would not have been diagnosed or treated in the absence of screening. CHT is a spectrum, and in the absence of screening only those who are most severely affected will be detected. The doubling of the prevalence of CHT following newborn screening reflects the identification of children with milder or subclinical cases of permanent CHT as well as transient hypothyroidism, which accounts for 10–30% of children diagnosed as having CHT following screening.45,–,47 The Swedish retrospective screening study found that no children who had elevated TSH values at birth and were either euthyroid or had subclinical CHT at age 5 years had an IQ <70 despite never having been treated for CHT.27 Regrettably, little information exists on cognitive outcomes in subclinical CHT or transient TSH elevation. Nonetheless, based on available data, it seems more reasonable to conclude that roughly one-quarter of children with clinically diagnosed CHT, or 1 in 25 000 births, experienced overt intellectual disability prior to the introduction of newborn screening.
The prevention of overt intellectual disability through newborn screening for CHT, although important, is just the proverbial tip of the iceberg. Equally importantly, even children with unscreened clinical CHT who have IQ scores in excess of the 70–75 range used to classify a child as having disability incur serious cognitive impairment. The mean IQ score reported for children with clinical CHT was 82 in the two UK studies,29 30 and 87–88 in the two Swedish studies.26 27 A Japanese study, which was excluded from table 2 because of a lack of detailed information about the sample composition, reported a mean IQ of 87 for 22 children aged 36–66 months who had been tested using various instruments.35 Whether the mean IQ was 82 or 87 is unimportant. A cohort with an average IQ score of 85 has lost an average of 20–25 IQ points relative to their potential, with an expected IQ of 105–110. This cognitive impairment has large negative effects on their economic productivity, with a reduction of at least 1% in lifetime earnings for each one-point drop in IQ scores.48 Also, although none of the children with subclinical CHT in the Swedish retrospective screening study had overt disability, their cognitive test scores were still depressed relative to their peers, by an average of 7 points.
Permanent CHT is associated with other outcomes that are preventable through prompt treatment. These include behavioural abnormalities and physical growth retardation, both of which may occur in relatively mild cases of CHT. For example, the Swedish retrospective screening study found that children with subclinical permanent CHT who had never been treated had a number of behavioural abnormalities despite having cognitive test scores in the normal range.27 The clearest data come from one of the two UK studies. Frost and Parkin29 obtained behavioural assessments for 60 children from parents and teachers and for 56 age- and sex-matched classmates. Twelve (20%) of the CHT sample had marked behavioural disturbances, and 47% had elevated behavioural risk scores compared with 14% of their peers.
Growth retardation is a well-known complication of CHT. Among unscreened cohorts under long-term treatment with thyroxine, the percentage with height below the 10th centile has been reported to range from 19%49 to 31%.30 A child can have a normal initial screen, particularly if a TSH cut-off of 15 mU/l or higher is used, and then be diagnosed as having permanent hypothyroidism on the basis of growth retardation.19
We suggest that extending newborn screening protocols for CHT to pick up additional children with relatively mild CHT who are missed by current protocols is unlikely to lead to prevention of more cases of intellectual disability. Whether changing screening cut-offs will result in preventing other outcomes such as milder cognitive, behavioural or physical sequelae remains to be proven.
Newborn screening for CHT is valuable, but it is essential that screening form part of a comprehensive system of follow-up, diagnosis, treatment, monitoring and evaluation.50 If treatment and healthcare providers are not readily available, and parents are not appropriately educated about the importance of adherence to treatment, the benefits of screening will not be fully realised.51 A recent study from the USA found that both publicly and privately insured children with diagnosed CHT had high rates of discontinuation of treatment in the first 3 years after beginning treatment, contrary to guidelines, although many of these children likely had mild CHT or transient TSH elevations, and it is unknown whether they experienced adverse outcomes.52
In conclusion, CHT is a model disorder for newborn screening because of its simple and inexpensive treatment and improvements in outcomes if treatment is initiated promptly. Even though the benefits of screening for CHT have often been overstated in terms of the numbers of cases of overt intellectual disability prevented, newborn screening results in important public health benefits, including an improvement in the distribution of cognitive ability and expected economic productivity. It is not clear, though, if the modification of screening protocols, including lowering screening cut-offs or requiring repeat screens, will have comparable effects on cognitive outcomes. There is good reason to believe that it would not, based on the clear association of disease severity with cognitive outcomes. Direct evidence that the additional children who are picked up through the use of lower cut-offs or repeat screens would have otherwise experienced cognitive impairment is lacking. Policy analyses that address the expansion of newborn screening protocols should focus on outcomes for which there is evidence of association with mild CHT and not presume that each child treated for CHT would otherwise have been at risk of intellectual disability.
The authors thank C Boyle for contributions to an earlier version of this manuscript. The authors thank AM Comeau, J Haddow, C Hinton, G Krahn, S LaFranchi, M Mitchell and M Yeargin-Allsopp for helpful comments on earlier versions as well as the peer reviewers. The findings in this paper were presented in a Symposium on New Trends in Thyroid Screening held at the 47th Annual Meeting of the European Society for Paediatric Endocrinology in Istanbul, Turkey, 20 September 2008.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.