Main

Hypothyroxinemia is a common finding in preterm infants. It is strongly related to gestational age and normalizes spontaneously in 6-8 wk(1, 2). Whether this transient hypothyroxinemia is harmful in preterm infants remains a matter of debate. A relationship between low thyroxine level after the first day of life and respiratory distress has been found, but it is not clear whether this is a causal relationship(3). A relationship between low thyroxine level and neonatal mortality has also been reported(4, 5).

Normal thyroxine levels are necessary for postnatal development of the brain. From the history in children treated for congenital hypothyroidism we know that early therapy can prevent mental retardation. However, depending on the duration and severity of hypothyroidism, some aspects of mental and neurologic development may remain impaired(6–10). In preterm infants the evidence is contradictory. Kohelet et al.(11) found no statistically significant difference in reaction on brainstem-evoked potentials between hypothyroid and euthyroid preterm infants, whereas de Vrieset al.(12) did find a delay in myelination in preterm infants with hypothyroxinemia. Hadeed et al.(13) and Choundry et al.(14) found no relation between low thyroxine levels and developmental outcome in preterm infants, whereas preterm infants with low triiodothyroxine levels were found in the study of Lucas et al.(15) to have significantly lower values on the Bayley scales at 18 mo of age.

In the nationwide cohort of preterm and/or very low birth weight infants born alive in 1983 in The Netherlands, we previously found a relation between thyroxine levels at the end of the first week of life and a delay in one or more of three developmental milestones at the age of 2 y(16). In the present study we report the outcome at 5 y of age as well as the achievement at school at age 9 in relation to thyroxine levels in the first week of life.

METHODS

Patients. The study population was part of the Project On Preterm and Small for gestational age infants (POPS) which included virtually all infants (n = 1338) born alive in 1983 in The Netherlands, with a gestational age of less than 32 wk and/or a birth weight of less than 1500 g(17). For infants born between April and December in 1983 (n = 1052), the results of thyroxine measurements, obtained in the national thyroid screening program on d 6-9, could be linked to the perinatal data in the POPS study. Not included were the data of 266 infants who died in the 1st wk of life before thyroxine measurements were collected and of 56 infants in whom thyroxine samples were collected either before d 5 or after d 17. Infants with disabling congenital malformation (n = 7), primary congenital hypothyroidism (n = 1), and temporary thyroid hormone replacement (n = 5) were excluded. Thus data on 717 infants were available for the present study.

Screening method. Thyroxine concentrations from filter paper eluates were determined in duplicate by RIA in five laboratories with permanent laboratory control. Thyroxine levels in the eluates were expressed as standard deviations about the mean calculated on a daily basis. The intraassay and interassay coefficients of variation in the eluates were 8 and 10%, respectively(18).

TSH measurements in the national screening program were performed in 20% of the samples with the lowest thyroxine level of each day only. Thyroxine levels were low in virtually all study infants, because in 94% of these infants TSH measurement was necessary, and in only 6% of the infants thyroxine levels were in the normal range.

Follow-up. A follow-up examination was conducted by three specially trained pediatricians during a home visit shortly after the 5th birthday. The examiners were unaware of the perinatal history of the examined children and had no access to neonatal thyroxine data. Interobserver variation was small and varied from 0 to 4% on the different outcome measurements(19). The assessment included neurologic examination according to Touwen(20), developmental assessment based on the Denver Developmental Screening Test(21), and a speech and language assessment validated for Dutch children(22). Neurologic outcome was dichotomized as normal or abnormal, including minor neurologic dysfunction and cerebral palsy. Mental development was assessed as either normal or delayed including both mild (less than 1 y delay) and severe delay. Speech and language assessment was graded as normal or delayed (less than the 10th centile of a standard Dutch population). Overall outcome based on these assessments was given as handicap, disability without handicap, impairment without disability, or normal according to the World Health Organization (WHO) classification(19).

At 9 y of age, school performance was assessed on the basis of a parental questionnaire. Results at school age were assessed as normal when the child was in mainstream education in an age appropriate level and as abnormal when the child was in special education or in mainstream education with grade retention(23).

Statistical analysis. Multivariate logistic regression analyses with unconditional maximum likelihood estimation were used to study the relationship between thyroxine levels and outcome in the various developmental fields at ages 5 and 9 y. Possible confounding factors included were premature rupture of membranes, infants' sex, gestational age in weeks, weight for gestational age (<10th percentile versus ≥10th percentile for the Dutch population), Apgar score at 5 min (<7 versus ≥7), duration of mechanical ventilation, ICH and (or) seizures, septicemia, and the highest observed neonatal serum bilirubin level (μmol/L).

The effect of grading of ICH, a possibly important confounder, could be analyzed only in a subset of 228 infants born in or referred to a neonatal intensive care unit, where routine ultrasound examinations were available in 1983.

RESULTS

Mean thyroxine levels in the study infants were considerably below the mean of the total screened population: 661 (92%) of the 717 studied infants had a thyroxine level of more than 1 SD below the mean, and 32% had a thyroxine level of more than 3 SD below the mean (approximately <60 nmol/L), often considered as hypothyroidism. The percentage of infants with a thyroxine level of more than 3 SD below the mean decreased from 83% in infants less than 26-wk gestation to 15% in infants with a gestational age of 32 wk or more(Table 1). There were no elevated TSH values in those infants for whom TSH was ascertained.

Table 1 Mean thyroxine level and percentage of hypothyroxinemia [thyroxine (T4) ≤3 SD] by gestational age
Full size table

At 5 y of age, 50 infants had died and 27 were lost to follow-up. Of the 640 children available for follow-up (96% of survivors), 85 (13.3%) had a disability, and another 92 (14.3%) were handicapped. Infants who had died and children with a handicap at 5 y of age had significantly lower thyroxine levels than children with a normal outcome even after correction for gestational age (Table 2). At 9 y of age, 552 children were available for follow-up (83% of survivors), and of these 151 (27.4%) had grade retention and 101 (18.3%) were in special education.

Table 2 Mean thyroxine level (SD) by gestational age and outcome at 5 y of age
Full size table

The OR for thyroxine on the various outcome measures are given inTable 3. The unadjusted OR on handicap and on neurologic dysfunction were significantly higher than 1 and slightly increased after adjustment for gestational age. After adjustment for all other perinatal factors the significance for handicap no longer existed, although the OR was similar. The adjusted OR on neurologic dysfunction was 1.3, implying a 30% increase in the odds on neurologic dysfunction for each decrease in thyroxine level with 1 SD. School failure at the age of 9 also appeared to be significantly related to low thyroxine levels in the 1st wk of life (OR unadjusted and adjusted 1.3).

Table 3 Unadjusted and adjusted OR (95% confidence interval) for thyroxine (SD) on different outcome measures at ages 5*and 9†
Full size table

In the subpopulation of 228 children with known grading of ICH, severity of ICH was found not to influence this association when added to the set of confounders.

DISCUSSION

More than 90% of the infants in our study had thyroxine levels more than 1 SD below the mean of the normal population, but the incidence of hypothyroxinemia (thyroxine < 60 nmol/L) was 32% in the total study population and 39% in the infants with less than 32-wk gestation. This is considerably less than the 84% found by Mercado et al.(1). This difference may be due to underestimation in our study, caused by selection bias. It is known that a relationship between neonatal mortality and low thyroxine levels exists(4). The majority of neonatal deaths in our study occurred in the 1st wk of life, before thyroxine measurements were performed, which means that many infants who probably had low thyroxine levels were not included in this study. The relationship between gestational age and hypothyroxinemia is similar to that found by others(1, 2).

Both neurologic dysfunction at age 5 y and school failure at age 9 y appeared significantly more frequent in children with low neonatal thyroxine levels. This is in accordance with specific cerebellar dysfunction as described in relation to low thyroxine by Porterfield and Hendrich(24). The association between low triiodothyroxine levels with a delay in both mental and motor performance in preterm infants, as found by Lucas et al.(15), may have the same origin.

The origin of neurodevelopmental dysfunction may be similar to that of the(minor) neurodevelopmental dysfunction found in children with severe congenital hypothyroidism, believed to be caused by intrauterine hypothyroidism(6–8). Probably, transient hypothyroxinemia at an early stage of development, either intrauterine in full-term infants with early treated congenital hypothyroidism or extrauterine in very preterm infants, can have a subtle but lasting effect on brain maturation.

Because these differences in development are subtle in children with treated congenital hypothyroidism(25), it is not likely that such differences will be found in preterm infants in studies done in a small number of infants or in very young ones, such as the study of Hadeedet al.(13) or Chowdhry et al.(14). Similarly, such subtle differences cannot be detected by the global screening methods for mental development and speech language development at 5 y of age that we used in our study. The significantly higher frequency of school failure we found, however, suggests that mental development is also different in preterm infants.

It is still not clear whether the relationship between low thyroxine in the neonatal period and later developmental disturbances is causal. Although adjustment was made for possible confounders, there might still be a common cause for both low thyroxine levels in the neonatal period and later neurodevelopmental dysfunction, not or not fully taken in account using these confounders. On the other hand, if low thyroxine levels increase the severity of neonatal disease (for instance respiratory distress syndrome), overadjustment may have led to an underestimation of the relationship between low thyroxine and neurodevelopmental dysfunctions. Randomized clinical trials with thyroxine administration in preterm infants are indispensable for the study of causality and the preventive effect. Because the neurodevelopmental disorder to be prevented is subtle, a sufficient number of children should be included, and follow-up should be long and extensive enough to be conclusive. Such trials are urgently needed to establish whether administration of thyroxine can at least partially prevent (minor) neurologic dysfunction and learning disabilities, both common sequelae of very preterm birth(23, 26). Furthermore the postulated necessity for thyroxine administration after birth should be established before new treatments such as prenatal TSH-releasing hormone administration are introduced on a wider scale. This treatment, which is very promising to reduce chronic lung disease(27, 28), may at the same time increase the frequency of (minor) neurologic dysfunction and learning disabilities in very preterm infants, by causing a further decrease of thyroid hormone levels in the neonatal period(29).