Page 1 of 21
Accepted Preprint first posted on 9 March 2009 as Manuscript EJE-08-0953
N.Benhadi 1
Higher maternal TSH levels in pregnancy are associated with
2
increased risk for miscarriage, fetal or neonatal death
3
N. Benhadi1, W.M. Wiersinga1, J.B. Reitsma2, T.G.M. Vrijkotte3, G.J. Bonsel3.
4 5 6
1
7
of Amsterdam, Meibergdreef 9, 1105 AZ Amsterdam.
Departement of Endocrinology and Metabolism, Academic Medical centre of the University
8 9 10
2
Departement of Clinical Epidemiology, Biostatistics and Bioinformatics, Academic Medical
Centre of the University of Amsterdam, Meibergdreef 9, 1105 AZ, Amsterdam
11 12
3
13
Meibergdreef 9, 1105 AZ, Amsterdam.
Departement of Social Science, Academic Medical Centre of the University of Amsterdam,
14 15
Corresponding author:
16 17
N. Benhadi, Departement of Endocrinology and Metabolism, Academic Medical Centre of the
18
University of Amsterdam, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands.
19
Telephone: +31 (0)20 5666964, Fax: +31 (0)20 6917682, Email:
[email protected].
20 21 22
Running title: Increasing TSH associated with higher risk for child loss.
23 24
Word count: 2,675
25
1 Copyright © 2009 European Society of Endocrinology.
Page 2 of 21
N.Benhadi 1
Abstract
2
Background: To examine the relationship between maternal Thyroid Stimulating Hormones
3
(TSH) and free Thyroxine (T4) concentrations in early pregnancy and the risk of miscarriage,
4
fetal or neonatal death.
5
Method: Cohort study of 2497 Dutch women. TSH, free T4 and Thyroid Peroxidase
6
Antibodies (TPO-Ab) concentrations were determined at first booking. Child Loss was
7
operationalised as miscarriage, fetal or neonatal death. Women with overt thyroid dysfunction
8
were excluded.
9
Results: 27 cases of child loss were observed. The mean TSH and free T4 level in the women
10
with child loss was 1.48 mU/L and 9.82 pmol/L compared to 1.11 mU/L and 9.58 pmol/l in
11
women without child loss. The incidence of child loss increased by 60% (OR= 1.60
12
(95%CI:1.04-2.47)) for every doubling in TSH concentration. This association remained after
13
adjustment for smoking, age, parity, DM, Hypertension, previous preterm deliveries and
14
previous preterm stillbirth/miscarriage (AOR= 1.80 (95%CI: 1.07- 3.03)). This was not true
15
for FT4 concentrations (OR= 1.41 (95%CI: 0.21- 9.40); P=0.724).
16
Conclusion: In a cohort of pregnant women without overt thyroid dysfunction, the risk of
17
child loss increased with higher levels of maternal TSH, Maternal free T4 concentrations and
18
child loss was not associated.
19 20 21
Key Words: TSH, pregnancy outcome, thyroid function, risk.
22 23 24 25
2
Page 3 of 21
N.Benhadi 1 2
Introduction Thyroid hormones are vital for the development of the brain both during fetal and
3
early postnatal life. Impaired maternal thyroid hormone availability may induce irreversible
4
brain damage with consequent neurological abnormalities (1-8). So far, only the effects of
5
subclinical and overt hypo- and hyperthyroidism on pregnancy outcome have been studied.
6
Overt maternal hypothyroidism in the first trimester is associated with preterm delivery and
7
fetal death (9). Haddow JE and colleagues (1999) reported that children of women whose
8
thyroid stimulating hormone (TSH) levels were increased during midterm of pregnancy had a
9
slight but significant reduction in intelligence quotient score between 7 and 9 years of age if
10
compared with infants of euthyroid women (10). Allan WC, Haddwon JE et al (2002) showed
11
that TSH levels above 6 mU/L are significantly associated with a higher frequency of still
12
birth (11). Women with resistance to thyrod hormone (RTH), (high free T4 thyroxine (T4)
13
levels accompanied with normal or slightly elevated TSH levels) have a higher rate of
14
miscarriage (12). In a study of Casey BM, Dashe JS et al. (2007) there was no significant
15
adverse effects on perinatal outcome in women with maternal hypothyroxinemia (low free T4
16
levels accompanied with normal TSH levels) in the first half of their pregnancy (13).
17
From the evidence in cases with overt hypo- or hyperthyroidism we would expect both
18
low and high TSH and free T4 to be associated with miscarriage, fetal or neonatal death (child
19
loss). The purpose of the present study was to examine the association between maternal
20
levels of TSH, free T4, thyroid peroxidase antibodies (TPO-Ab) measured in pregnancy and
21
subsequent child loss, excluding women with known thyroid disease or overt hypo- or
22
hyperthyroidism.
23
The study was carried out using data from native Dutch pregnant women without
24
known or clinical thyroid disease, who participated in a large population-based cohort study
25
(ABCD-study, Amsterdam 2003-2004). Other ethnic groups participating in this study were
3
Page 4 of 21
N.Benhadi 1
excluded from the present analysis, because of significant difference in serum TSH and in
2
pregnancy outcome that exist between the various ethnic groups (14).
3 4
Methods
5
Subjects
6
Our study was nested within a prospective cohort study of pregnant women from the
7
Amsterdam Born Children and their Development (ABCD) study (14). The main objective of
8
the ABCD study is to examine differences in pregnancy outcome, focussing on ethnic
9
background, maternal lifestyle factors and psychosocial conditions on the outcome of the
10
pregnancy and the baby’s health. The ABCD study is a collaborative effort of the Municipal
11
Health Services (GGD) and all hospitals and midwife practices in Amsterdam, the
12
Netherlands. All pregnant women living in the city of Amsterdam were invited to participate
13
at their first visit to an obstetric caregiver between January 2003 and March 2004. Of 12,377
14
pregnant women invited, 8,266 women agreed to participate (response rate 67%). These
15
women filled out a questionnaire including questions about social demographic characteristics
16
and ethnic background. 4,267 women gave additional informed consent for blood collection
17
during their first visit which took place on average in the 13th week of gestation. The study
18
protocol was approved by the medical ethnical committees of all Amsterdam hospitals and the
19
Registration Committee of Amsterdam, and participants gave their written informed consent.
20 21 22 23 24 25
4
Page 5 of 21
N.Benhadi 1 2
Baseline characteristics All pregnant women received a pregnancy questionnaire at their home address two
3
weeks after their first antenatal visit. The questionnaire contained demographic, health
4
history, medication and lifestyle questions, all from existing validated sources. All
5
approached women were asked to return the pregnancy questionnaire by prepaid mail. A
6
written reminder was sent two weeks after the initial mailing.
7
Smoking status during pregnancy, parity, age and BMI were determined from the self
8
reported information. In this study we recoded smoking as a dichotomous question for current
9
smoking (yes/no), regardless the amount of cigarettes smoked daily. Parity was recoded in a
10
categorical variable 90, 1 and 2+). BMI was based on the length and weight of the mother
11
before pregnancy. Alcohol was not included in the analysis of our study in view of the very
12
low levels of alcohol use among the Dutch pregnant women in our study population.
13
Hypertension (no, pre-existent and gestational hypertension) and diabetes mellitus (nu,
14
pre-existent and gestational diabetes) were based on self-reported information from the
15
questionnaire and completed by information from the national obstetric registry (Perinatal
16
Registration centre of the Netherlands) (14). This registry was linked to the ABCD data by
17
probabilistic record linkage (15). These data were gathered by a trained health care provider.
18
We also recorded women who had a preterm delivery in previous pregnancies and
19
previous miscarriage or stillbirth. In our stillbirth/miscarriage variable we included intra-
20
uterine deaths as well. All these data were based on the information from trained health care
21
providers from the national obstetric registry (Perinatal Registration centre of the
22
Netherlands) (14).Validation has shown close to perfect results for these variables.
23 24 25
5
Page 6 of 21
N.Benhadi 1
Assays
2
For our study serum TSH, free Thyroxine (FT4) and antibodies against thyroid
3
peroxidase (TPO-Ab) were assayed. TSH (reference range, 0.34-5.60 mU/L) and FT4
4
concentration (reference range, 7.5- 21.1 pmol/L) were measured in serum by means of
5
Access immunoanalyzer of Beckman Coultier, Inc. The inter-assay variation for TSH was
6
5.0% and for FT4 the inter-assay variation was 3.1-5.0%. Antibodies against Thyroid
7
Peroxidase (TPO-Ab) were determined by Elisa ELIZEN TG Ab (E-CK-96), Zentech, Luik,
8
Belgium. A TPO-Ab concentration above 80 kU/L was considered as positive. The inter-
9
assay variation was 13.4%.
10
Overt hyperthyroidism was defined as having a TSH concentration below 0.4 mU/L in
11
combination of FT4 concentration above the upper limit of 21.1 pmol/L. Overt
12
hypothyroidism was defined as a TSH above the 4.0 mU/L upper limite in combination with a
13
FT4 below the lower limit of 7.5pmol/L.
14 15 16
Outcomes The occurrence of miscarriage, fetal death or neonatal death (child loss( was
17
determined from3 overlapping sources; 1) the National Midwife Registry, 2) the National
18
Obstetricians Registry and 3) the National Neonatal Registry. Miscarriage was defined as
19
death of the fetus occurring before 22 weeks of gestation. Fetal death and neonatal death were
20
defined according to existing standards (fetal death: death occurring from 22 weeks of
21
gestation until delivery; neonatal death: death from 0-7 days after delivery). Fetal death
22
includes stillbirth and intra-uterine deaths in our cohort.
23 24 25
6
Page 7 of 21
N.Benhadi 1 2
Statistical Analysis Differences in demographic and clinical characteristics were analyzed in a descriptive
3
way. Logistic regression models were built to explore the relationship between TSH, FT4
4
concentrations determinants, and the occurrence of miscarriage, fetal death and neonatal death
5
in more detail. Restricted also known as natural cubic splines (4knots) were used to examine
6
the functional relationship between TSH and free T4 levels in relation to the outcome (17,
7
18). Based on these graphical analyses, we used a log transformation (to the power of 2) as
8
this clearly improved the linear relationship between TSH and FT4. In addition to the
9
univariate analysis we also build a multivariate model containing the following potential
10
confounders: current smoker (yes/no), mother’s age (continuous), parity (categorical: 0,1,2+),
11
hypertension (no, pre-existent and gestational hypertension), diabetes mellitus (no, pre-
12
existent, gestational diabetes), previous preterm deliveries (yes/no), previous
13
stillbirth/miscarriages (yes/no) and TPO-Ab presence (yes/no) with TPO-Ab concentration
14
below 80kU/L or more as positive. The modelling strategy was the same whether examining
15
the relationship between levels of TSH of FT4 with child loss. P-values of less than 0.05 were
16
considered statistically significant.
17 18
All statistical analysis was performed using the Statistical package of Social Sciences and Problem Solutions (SPSS version 15.0).
19 20 21 22 23 24 25
7
Page 8 of 21
N.Benhadi 1
Results
2
The initial number of native Dutch women within the ABCD cohort giving informed
3
consent for blood sampling was 2,684. From this starting cohort we excluded 116 women in
4
whom blood sampling occurred after the 27th week of gestation (third trimester). Of the
5
remaining, we excluded 29 women who gave birth to twins. We also excluded 18 women
6
with over hypo- or hyperthyroidism defined as TSH >5.6 mU/L in combination with FT4< 7.5
7
pmol/L and TSH 21.1 pmol/L respectively. Of the
8
2,521 women who remained in the study we excluded another 24 with missing TSH values,
9
leaving 2,497 women in our analysis (Figure 1).
10
Our study population of 2,497 Dutch pregnant women had a mean age of 32.0
11
(SD=4.0) years. In 60% of the women this was their first pregnancy and 32% their second.
12
9.2% of all women smoked during their pregnancy. There were 129 (5%) women with a TSH
13
concentration below 0.34 mU/L and 11 (0.5%) with a TSH value above 5.6 mU/L. A FT4
14
concentration below 7.5 pmol/L was seen in 94 women (4%) and there were no women with a
15
FT4 value above the 21.1 pmol/L in our study cohort. The overall prevalence of elevated
16
TPO-Ab (concentration above > 80kU/L) was 5.8% (146/2497), (see Table I).
17
TSH and FT4 levels were negatively correlated (Pearsons’r= -0.375, p