Anselmo J, Cao D, Karrison T, Weiss RE, Refetoff S. Fetal Loss Associated With Excess Thyroid Hormone Exposure. JAMA. 2004;292(6):691–695. doi:10.1001/jama.292.6.691
Author Affiliations: Endocrinology Unit, Hospital Divino Espírito Santo, Ponta Delgada, Azores-Portugal (Dr Anselmo); Department of Health Studies (Drs Cao and Karrison), Department of Medicine and Committees of Molecular Medicine and Nutritional Biology (Dr Weiss), and Departments of Medicine and Pediatrics and Committees on Genetics and Molecular Medicine, University of Chicago, Chicago, Ill (Dr Refetoff).
Context Maternal hypothyroidism and hyperthyroidism have deleterious effects
on the outcome of pregnancy. While the effects of thyroid hormone (TH) deprivation
on the fetus, independently from that on the mother, can be studied in infants
with congenital hypothyroidism, this is not the case in those with fetal thyrotoxicosis.
Objective To study the effects of TH excess on fetuses carried by mothers with
resistance to TH (RTH) who are euthyroid despite high TH levels but who may
carry normal fetuses that are exposed to high maternal hormone levels.
Design, Setting, and Participants Retrospective study of 167 members of an Azorean family with RTH. Affected
individuals had the RTH phenotype (high serum concentration of free thyroxine
and triiodothyronine without suppressed thyrotropin) confirmed by genotyping
to identify the Arg243→Gln mutation in the TH receptor β gene.
Main Outcome Measures Pregnancy outcome of affected mothers vs that of unaffected mothers
carrying fetuses conceived by affected fathers, as well as that of unaffected
first-degree relatives and outcomes from the general island population. Comparison
of birth weights and blood concentrations of thyrotropin (TSH) obtained during
routine neonatal screening of infants born to these 3 groups.
Results Thirty-six couples with complete information belonged to 1 of 3 groups:
affected mothers (n = 9), affected fathers (n = 9), and unaffected relatives
(n = 18). Mean miscarriage rates were 22.9%, 2.0%, and 4.4%, respectively
(χ2 = 8.66, P = .01). Affected mothers
had an increased rate of miscarriage (z = 3.10, P = .002, by Wilcoxon rank-sum test). They had marginally
higher than expected numbers of affected offspring, ie, 20 affected and 11
unaffected children (P = .07), while affected fathers
had 15 affected and 12 unaffected children (P = .35).
Unaffected infants born to affected mothers were significantly smaller than
affected infants, having a mean SD score for gestational age of –1.79
(SD, 0.86) vs −0.06 (SD, 1.11) to –0.22 (SD, 0.70) for all other
groups (P<.001). Only unaffected infants born
to affected mothers had undetectable blood levels of TSH.
Conclusion There was a higher rate of miscarriage in mothers affected by RTH that
may have involved predominantly unaffected fetuses. The lower birth weight
and suppressed levels of TSH in unaffected infants born to affected mothers
indicates that the high maternal TH levels produce fetal thyrotoxicosis. These
data indicate a direct toxic effect of TH excess on the fetus.
Thyroid hormone (TH) plays an important role in embryogenesis and fetal
maturation.1 In rat embryos cultured in vitro,
addition of serum from animals with either hypothyroidism or hyperthyroidism
induced malformations.2 Evidence of a similar
effect in humans, including higher miscarriage rates, intrauterine growth
retardation, and congenital malformations, however, is indirect since it comes
from studies of women with autoimmune thyroid disease.3- 5 Since
the presence of thyroid autoantibodies is associated with an increased rate
of miscarriages,6,7 it remains
unclear whether maternal TH has a direct toxic effect on the growing fetus.
In the present study we took advantage of a disorder that circumvents
the confounding factors that prevent assessment of the direct impact of maternal
TH on the fetus. Indeed, individuals with resistance to TH (RTH) achieve euthyroidism
by maintaining high serum levels of free TH. Furthermore, their increased
thyroid gland activity is not mediated through thyroid-stimulating antibodies.8 The disorder is caused by mutations of the TH receptor β
gene (TRβ). Mutant receptors not only have reduced
function, but also interfere with the function of the normal receptor, explaining
the dominant mode of inheritance.
We retrospectively examined the pregnancy outcome in a large family
harboring the TRβ mutation, Arg243→Gln
(R243Q),9 which has been previously characterized
in detail.10,11 The estimated
miscarriage rate in affected and unaffected women, together with determination
of the genotype of all live births, their weights, and neonatal blood thyrotropin
(TSH) concentrations, allowed us to evaluate the influence of high maternal
levels of TH on the fetuses.
More than 200 members of an Azorean family of Portuguese ancestry with
RTH were identified. They were traced to a late 19th-century founder couple,
the progenitor of 61 nuclear families spanning 4 generations (generation 1,
6 families; 2, 19 families; 3, 35 families; and 4, 1 family). Of the 167 individuals
available for examination, 44 were affected (23 women, 21 men) and 123 were
unaffected (35 first-degree relatives [18 women, 17 men], 54 distant relatives
[21 women, 33 men], and 34 relatives by marriage [14 women, 20 men]). All
couples bore children and/or had pregnancies. In 23 couples, one of the spouses
was affected; in 19 unaffected couples, one of the spouses was first-degree
relative to an affected individual (ie, an unaffected child of an affected
parent). The remaining 19 couples were distant (second-generation) relatives.
Two of the affected couples were excluded because of incomplete information
and another 3 because of thyroidectomy-induced hypothyroidism (TSH values
of 15-75 mU/L) during all pregnancies. Information from 1 unaffected couple
A full clinical history with particular emphasis on pregnancies and
miscarriages was obtained by interviewing all mothers, available spouses,
and family members. The presence of pregnancy and miscarriage was made by
clinical criteria (miscarriages were scored when vaginal bleeding had occurred
with or without cramping and passing of clots after pregnancy had been diagnosed
by a physician and confirmed by conventional pregnancy test). Records from
hospital admissions, family physicians, and other relevant sources of clinical
information were also obtained. The presence of RTH was not known to the participants
or to the record keepers. The study was approved by the institutional review
boards of the Hospital Divino Espérito Santo, Azores-Portugal, and
the University of Chicago, Chicago, Ill. Informed consent (oral consent in
Azores, written consent in Chicago) was obtained from all participants.
The RTH phenotype was identified by the presence of elevated serum concentrations
of free thyroxine (T4) and triiodothyronine (T3) in
the presence of nonsuppressed TSH levels. The diagnosis was confirmed in all
participants by the presence of a single nucleotide substitution in 1 allele
of the TRβ gene. This mutation, R243Q, was described
in several unrelated families.10- 12
Blood TSH values were obtained from the registry of the National Center
for Neonatal Diagnosis in mainland Portugal. Of the 123 participants included
in the 3 groups under analysis, 50 (41%) were born after the institution of
routine screening for neonatal hypothyroidism; TSH values were recovered from
44 of these 50 (88%). Blood was collected on filter paper between the 6th
and 10th day of life.
Thyroid function test results and serum TSH levels in different groups
were compared using analysis of variance.
Miscarriage rates for the groups of affected mothers, affected fathers,
and unaffected first-degree relatives were first calculated as the ratio of
total number of miscarriages to the total number of pregnancies (the "per
pregnancy" rate). Since outcomes among multiple pregnancies in the same couple
are potentially correlated, we also calculated the miscarriage rate for each
couple and averaged the rates over the couples in the group (the "per couple"
rate). The distribution of affected children for each couple was similarly
The per-couple miscarriage rates were analyzed using the Kruskal-Wallis
test and the Wilcoxon rank-sum test. We used a 1-sided binomial test to analyze
whether the observed ratio of affected to unaffected progeny in couples with
an affected mother or father deviates significantly from 1. The effect of
parental and infant genotype on birth weight was assessed by repeated-measures
analysis of variance to account for the correlation among infants' birth weights
within the same couple. Available data from the general population of San
Miguel Island was used for comparison. For birth weights and neonatal TSH
values, groups were compared using analysis of variance with the Tukey adjustment
for multiple pairwise comparisons. Analyses were performed using SAS version
8.2 (SAS Institute Inc, Cary, NC); P<.05 was used
to determine statistical significance.
All participants with RTH, except for the 3 who underwent thyroid surgery,
had serum free T4 and T3 values above the upper limit
of normal without suppressed concentrations of TSH. The latter ranged from
0.6 to 5.8 mU/L (reference range, 0.4-3.8 mU/L). Serum thyroglobulin concentrations
were also high in the majority of affected participants, in agreement with
the increased activity of their thyroid glands (Table 1). Only individuals expressing the RTH phenotype harbored
the R243Q mutation, but none manifested typical symptoms and signs of thyrotoxicosis.
The 18 affected and 18 unaffected couples used in the analysis of miscarriages
had 89 and 68 pregnancies, respectively. The per pregnancy rates of miscarriage
in the affected mothers, affected fathers, and unaffected first-degree relatives
were 23.7%, 6.7%, and 8.8%, respectively, while the overall miscarriage rate
in the general population was 8.1% (Table
2). The per couple mean miscarriage rates, on the other hand, were
22.9%, 2.0%, and 4.4% for the 3 groups, respectively (χ2 =
8.66, P = .01). A 2-sample Wilcoxon rank-sum test
showed that the miscarriage rate in the affected mothers group was significantly
higher than in the combined group without affected mothers (z = 3.10, P = .002).
Affected mothers delivered 20 affected and 11 unaffected children, while
spouses of affected fathers delivered 15 affected and 12 unaffected children.
The 56% rate of affected children in the latter group is in accordance with
the expected mendelian distribution of a dominantly inherited trait. The test
that the observed genotype distribution ratio of children born to affected
mothers is different than 1 has a P value of .07.
The per pregnancy prevalences of affected children were 64.5% and 55.6% for
those born to affected mothers and affected fathers, respectively; the per
couple prevalences were 77.9% and 48.1%, respectively.
Birth weight records were found for 67% to 92% of the newborns of the
different groups. Unaffected infants born to affected mothers had mean SD
scores for gestational age of –1.79 (SD, 0.86) vs −0.06 (SD, 1.11)
for affected infants born to the same mothers (P<.001)
(Figure 1). This was not true for
infants born to affected fathers. Except for unaffected infants born to affected
mothers, the mean birth weights of infants from all other groups (3.14-3.16
kg) were not different from the mean birth weight for infants born on San
Miguel island (3.18 kg).
As shown in Figure 2, TSH
values of infants born to the various groups were not significantly different
except for the unaffected infants born to affected mothers, in whom TSH was
undetectable (<0.1 mU/L). One of these infants had a low birth weight of
2.1 kg (−2.7 SDs). Thyroid function tests were assessed prospectively
in the last born. Levels of TSH were not detectable at 1 and 7 days, and levels
of free T4 and T3 were above the reference ranges. On
day 14, levels of free T4 had normalized and TSH level was just
below reference range (0.4 mU/L); however, by day 28, TSH level had increased
to 4.8 mU/L.
It is of interest to note that 1 of the affected mothers, excluded from
the study because she had postablative hypothyroidism with TSH values above
15 mU/L but free T4 levels still above reference range, delivered
2 infants with low but detectable blood TSH values (1.9 and 2.6 mU/L).
Since the first description of RTH,13 more
than 200 families have been identified.14 Affected
patients are usually heterozygous for mutations in the TRβ gene, as was also the case in the large Azorean family described
herein. Screening was extended to 74 descendants of siblings of the founding
mother and 108 descendants of the founding father and no RTH was identified,
indicating that the R243Q mutation occurred de novo in 1 of the founders.
This is not surprising considering that the mutation involves a CpG mutational
The study was undertaken to assess the effect of TH excess on the fetus
without the concomitant effect of hormone excess on the mother and without
the possible effect of antibodies associated with autoimmune thyrotoxicosis.
This occurs rarely in congenital thyrotoxicosis due to gain-of-function mutations
in the TSH receptor gene.16 However, even in
these rare occurrences, thyrotoxicosis does not occur before the full development
of the fetal thyroid gland. The study of women with RTH provided a unique
opportunity to evaluate the direct effect of maternal TH excess on the fetus
without producing hypermetabolism in the mother. In these women that carry
both unaffected and affected fetuses, only the former should experience the
metabolic effects of excess maternal hormone. Fetuses harboring the same mutation
as the mother would be exposed to appropriate levels of the hormone. Furthermore,
couples made up of affected fathers and unaffected mothers carrying unaffected
and affected fetuses serve as invaluable controls for any other unknown effect
of the mutant allele harbored by the fetus.
Our data show a 3- to 4-fold increase in the rate of miscarriage in
affected mothers compared with that of spouses of affected fathers or unaffected
first-degree relatives, as well as with the overall miscarriage rate in the
population of San Miguel Island. The latter was similar to that found in other
regions of the world.17 This difference held
true to an even greater extent when rates were calculated per couple.
Fertility was not impaired in affected couples, regardless of whether
women or men harbored the mutant TRβ gene. The
greater number of pregnancies among affected mothers compared with those among
spouses of affected fathers was not significant (z =
1.39, P = .16, by Wilcoxon rank-sum test). Furthermore,
the higher rate of miscarriage was not related to increased maternal age and/or
greater parity, both of which are known risk factors for spontaneous abortion.18 Indeed, the age span of parity in women from both
groups was not different and there was no correlation between parity and rate
Although of borderline significance, the difference in genotype frequency
among descendants of affected mothers (20 affected vs 11 unaffected offspring),
when combined with a significantly higher miscarriage rate, suggests that
these women tend to miscarry more unaffected than affected fetuses. The difference
in genotype frequency was not found in the progeny of affected fathers who
had, as expected, almost equal numbers of affected and unaffected offspring.
Since the mothers were not thyrotoxic and had no thyroid autoantibodies, it
may be concluded that miscarriages were the consequence of the fetal exposure
to the high levels of TH. This is further supported by the improved survival
of the affected fetuses, for whom high TH levels were physiological, just
as they were for their affected mothers. Contrary to findings in women with
uncontrolled hyperthyroidism,19 women with
RTH had no increased frequency of premature labor, preeclampsia, stillbirths,
and perinatal loss, further indicating that mothers with RTH are not thyrotoxic
despite their high circulating levels of TH.
Unaffected infants born to affected mothers showed a significantly lower
birth weight than their affected siblings. This suggests that the high maternal
TH level was able to induce a catabolic state during fetal life, similar to
what happens in children and adults with uncontrolled hyperthyroidism.20 That these infants were thyrotoxic is supported by
their suppressed blood levels of TSH at birth.
Taken together, these findings represent the first evidence in humans
that TH excess can, by itself, impair embryogenesis of growing fetuses through
transplacental passage of maternal TH. Until now, greater attention has been
given to the deleterious effect of insufficient TH passage from mother to
fetus.21,22 Yet, indirect evidence
for the potentially toxic effect of TH excess is contained in the recent finding
of high levels of iodothyronine deiodinase 3 in the uterine lumen surrounding
the fetal cavity.23 This enzyme inactivates
the potential biological effect of T4 and T3 by converting
them into the inactive forms 3,3′,5-triiodothyronine (reverse T3) and 3,3′-diiodothyronine, respectively. In this respect, it
is interesting to note that the relative deficiency in TH levels occurring
in affected fetuses carried by unaffected mothers (spouses of affected fathers)
did not increase fetal attrition.
The data presented herein show, for the first time in humans, that high
levels of TH can exert a direct toxic effect on fetal development. This is
manifested by an increased rate of miscarriages and a lower birth weight of
unaffected infants born to euthyroid mothers with high levels of TH. As expected,
fetuses harboring a mutation that reduces the sensitivity to TH are protected
from this toxic effect of TH excess. Given the established importance of providing
TH replacement to even mildly hypothyroid pregnant women, it is important
to recognize that overreplacement appears to be equally detrimental.