Maternal thyroid deficiency and pregnancycomplications: implications for populationscreening

W C Allan, J E Haddow, G E Palomaki, J R Williams, M L Mitchell, R J Hermos,J D Faix, R Z Klein

AbstractObjective—To examine the relation be-tween certain pregnancy complicationsand thyroid stimulating hormone (TSH)measurements in a cohort of pregnantwomen.Methods—TSH was measured in seraobtained from women during the secondtrimester as part of routine prenatal care.Information was then collected aboutvaginal bleeding, premature delivery, lowbirthweight, abruptio placentae, preg-nancy induced hypertension, need forcesarean section, low Apgar scores, andfetal and neonatal death.Results—Among 9403 women with single-ton pregnancies, TSH measurements were6 mU/l or greater in 209 (2.2%). The rate offetal death was significantly higher inthose pregnancies (3.8%) than in thewomen with TSH less than 6 mU/l (0.9%,odds ratio 4.4, 95% confidence interval1.9–9.5). Other pregnancy complicationsdid not occur more frequentlyConclusion—From the second trimesteronward, the major adverse obstetricaloutcome associated with raised TSH inthe general population is an increased rateof fetal death. If thyroid replacementtreatment avoided this problem this wouldbe another reason to consider populationscreening.(J Med Screen 2000;7:127–130)

Keywords: thyroid stimulating hormone; pregnancy;fetal death

Although clinically apparent maternal hypo-thyroidism during pregnancy has long beenknown to be associated with both maternal andfetal complications, most of the studies docu-menting those problems have focused onpatients in specialty or high risk clinics. Thepresent study assesses the impact of maternalhypothyroidism in a cohort of pregnant womenbeing managed in primary care settings. In1991, our group reported thyroid stimulatinghormone (TSH) measurements in a cohort of2000 pregnant women whose sera were beingobtained for á-fetoprotein measurements at15–18 weeks’ gestation as part of routine care.1

TSH measurements were at, or above, 6 mU/lin 49 of the women (2.4%). In six of these 49women the thyroxine (T4) and/or free T4 meas-urements were suYciently abnormal (morethan two standard deviations below the aver-age) for clinical manifestations to be antici-

pated. It was not possible to gather individualinformation about health status in this cohort,because all identifiers were removed from theserum samples before the thyroid measure-ments were performed. Given the frequency ofthyroid deficiency identified in this pregnancypopulation, it was decided that its possibleimpact on late pregnancy and delivery shouldbe evaluated. To accomplish this, it was neces-sary to study a second, larger cohort ofpregnant women.

Materials and methodsThe Foundation for Blood Research oVersprenatal serum screening services for openneural tube defects and Down’s syndrome toall primary care prenatal practices in Maine.The testing is generally performed between 15and 18 weeks’ gestation. Approximately twothirds of the women receiving prenatal care inMaine opt for those services.

Between July 1990 and June 1992 the orderform for the prenatal screening test contained asupplementary consent form, asking women ifthey would agree to have thyroid functionmeasurements performed, in addition to theprenatal screening test being ordered. OurInstitutional Review Board approved the con-sent form and project. The study design calledfor limiting enrollment to women with single-ton pregnancies having prenatal screening forneural tube defects and Down’s syndrome.Approximately 20 900 women were eligible forprenatal screening, and samples and consentwere received from 10 010. Of these, 170 werefrom women not being tested for screeningpurposes and another 369 consents wereprovided by women who had signed up asecond time during the collection of a followup sample. Thus, 9471 women were eligiblefor the study, and pregnancy outcome wasavailable for 9403 (99.2%). Among the re-maining 68 women, about three quarters wereknown to have moved to another state beforedelivery. The women provided selected infor-mation about their pregnancy—for example,gravidity, parity, vaginal bleeding, and smokingstatus, at the time of enrollment. Also includedwas a question about insulin dependent diabe-tes that was independently verified as part ofanother project.2 Information about pregnancyoutcome—for example, viability, length of ges-tation, birthweight, and Apgar score wasobtained via a collaborative agreement with thestate’s Bureau of Vital Records. In the fewinstances where pregnancy outcome was not

J Med Screen 2000;7:127–130 127

Foundation for BloodResearch,Scarborough, Maine,USAW C Allan, director,clinical servicesJ E Haddow, vicepresident/medical directorG E Palomaki, director ofbiometryJ R Williams, data analystmanager

New England NewbornScreening Program,University ofMassachusetts MedicalSchool, Jamaica Plain,Massachusetts, USAM L Mitchell, professor ofmedicineR J Hermos, mph

Beth Israel DeaconessMedical Center,Boston,Massachusetts, USAJ D Faix, director ofclinical immunology

Dartmouth MedicalSchool, Hanover, NewHampshire, USAR Z Klein, consultingendocrinologist

Correspondence to:W C Allan, 69 US RouteOne, Scarborough, ME04070-0190, USAallan@fbr.org

www.jmedscreen.com

found in either the birth or death records, theinformation was collected from the health careprovider who cared for the woman. Earlierpublications by us have verified the reliabilityof this method of collecting both secondtrimester and pregnancy outcome data.3 4

The serum TSH measurements were per-formed at the New England Newborn Screen-ing Program in Boston on sera from allenrolled pregnancies within one month ofreceipt. The women’s physicians were notifiedif the TSH measurement was greater than10 mU/l. Additional thyroid function measure-ments were performed at the Beth IsraelDeaconess Medical Center at a later date onall serum samples with TSH measurements> 6 mU/l (the definition of thyroid deficiencyfor the current study) and on sera from thepregnancies enrolled immediately before andafter each of those samples. Details of theassays used for the various measurements havebeen published elsewhere.1

Statistical significance was taken at the 0.05value. Categorical variables were comparedusing the ÷2 test or, when there were few obser-vations, Fisher’s exact test. Exact confidenceintervals were estimated using Cornfield’sapproximation. All statistical analyses wereperformed using software from either BMDP5

or True Epistat.6

ResultsFigure 1 displays TSH measurements in the209 women (2.2%) whose values were> 6 mU/l. TSH measurements were 10 mU/lor higher in 37 of the women (0.4%). Table 1shows that women with TSH measurements of10 mU/L or higher were, on average, 2.8 yearsolder and 4.08 kg heavier than women whoseTSH measurements were not raised; fewer ofthem smoked cigarettes. Two of the 209women had insulin dependent diabetes. Itshould be noted that the women declininginvolvement in the study were quite similar towomen with TSH measurements less than6 mU/l. Table 2 shows detailed thyroid func-tion studies which were performed in all of thepregnancies with elevated TSH measurementsand in a subset of the pregnancies with valuesless than 6 mU/l (see Methods). Mean T4, freeT4, and T4/thyroxine binding globulin (TBG)ratios all become progressively lower, as TSHmeasurements increase. In addition, the per-cent of women with thyroid antibodies in-creases.

Table 3 shows that most of the complicationsof pregnancy and delivery occur at similar ratesamong women with raised TSH measurementsand the rest of the study population; averagegestational ages at delivery and birthweightsare also similar. Among the 9194 womenwhose TSH measurements were not raised,there were 83 fetal deaths (0.9%) as opposedto eight fetal deaths among the 209 womenwith TSH measurements at or above 6 mU/l(3.8%, odds ratio 4.4, 95% confidence interval1.9 to 9.5). The rate of fetal death is furtherstratified by degree of TSH increase in table 3.None of the fetal or neonatal deaths amongwomen with raised TSH measurements oc-curred in the women who also had insulindependent diabetes.

Information could be obtained regarding thethyroid status of the 16 women with the highestTSH measurements (at or above 20 mU/l).Two had undergone thyroidectomies previ-ously for “cancer” and four for Graves disease,

Figure 1 Individual TSH measurements for the 209 women with TSH measurements ator above 6.0 mU/l (approximately the 98th centile) are shown on a logarithmic scale.

300

100

200

50

20

30

10

6

TS

H (

mU

/l)

Table 1 Demographic characteristics of pregnant women with raised TSH measurements,in comparison to the general pregnancy population

Women notconsenting

Women giving consent TSH category(mU/l)

<6 6–9.99 >10

Number of pregnancies 10857 9194 172 37Mean maternal age (years) 26.5 26.9 28.0 29.7 (p<0.001)Mean gestational age at screening (days) 119 121 121 121Mean maternal weight (kg) 68.49 68.49 70.3 72.57 (p<0.05)Mean maternal weight gain (kg) 14.06 14.06 14.51 12.24More than high school education (%) 42 45 50 55Smoke cigarettes (%) 26 23 17 15 (p<0.03)Insulin dependent diabetes (%) 0.3 0.4 0.6 2.7

Table 2 Measurements of thyroid function in maternal sera during the second trimester, stratified by TSH measurement

TSH category (mU/l)

<6* 6–9.99 >10 p-value†

Number of pregnancies 418 172 37Mean T4 (nmol/l) 145 (108–192) 138 120 <0.001Mean free T4 (pmol/l) 11.8 (7.6–15.3) 11.0 9.6 <0.001Thyroxine binding globulin (nmol/l) 770 (530–1060) 760 800 NST4/TBG (molar ratio) 0.19 (0.13–0.25) 0.18 0.15 <0.001Thyroid antibodies (% positive) 9 (28/304) 55 (66/119) 80 (24/30) <0.001‡

*Two pregnancies with TSH measurements <6 mU/l were randomly selected for each sample with an raised TSH measurement.Numbers in parenthesis are the observed 5th and 95th centiles.†P value for comparison between groups with TSH <6 and >10 mU/l.‡P value for comparison between group with TSH <6 and both TSH 6–9.99 and TSH >10 mU/l.

128 Allan, Haddow, Palomaki, et al

www.jmedscreen.com

one woman had congenital hypothyroidism,five were reported to have idiopathic acquiredhypothyroidism. The remaining four werediagnosed as a result of the TSH measurementobtained for this study. Of the 12 withpreviously recognised disease, one admitted tonon-compliance, and their physicians sus-pected the others of non-compliance.

DiscussionThis is the first large, population based study ofcomplications in pregnancies with raised TSHmeasurements. Our finding that 2.2% of 9403unselected pregnant women have a TSHmeasurement at, or above, 6 mU/l is similar tothe frequency in our previous study (2.4% of2000 pregnancies) and that of Glinoer (2.2%of 1900 pregnancies).1 7 Neither of these earlierstudies linked TSH measurements with preg-nancy complications or outcome in the entirecohort. Other studies suggested that adverseevents such as fetal death, premature birth, lowbirthweight, placental abruption, and preg-nancy induced hypertension8–11 occur moreoften in women with clinically diagnosedhypothyroidism. These observations, however,have been limited to women attending high riskor specialty clinics and may not reflect findingsin the general population. Of the adverseevents that were examined in the presentcohort, only fetal death occurred more often inthe women with raised TSH measurements. Inprevious studies, the frequency of fetal death inthe presence of maternal hypothyroidismranged from 1.5 to 10%.7 9–13 However, none ofthese was a cohort study. Our study onlyincluded fetal deaths after 16–18 weeks’ gesta-tion, and this might explain why our rateamong women with raised TSH measurementswas towards the lower end of the reportedrange (3.8%).

Although the proportion of TSH associatedlate fetal deaths appears not to be great (3.8%of 2.2% of all pregnancies in the general popu-lation), it may be avoidable. Glinoer7 reportedfour miscarriages among the 41 women withraised TSH measurements in his study of 1900consecutive pregnant women. In that study,follow up was limited to women with elevatedTSH measurements. He performed screeningat the first prenatal visit, and all of the fetaldeaths occurred before he could introduce

treatment. In 1990, Stagnaro-Green and asso-ciates13 examined the relation between thyroidantibodies and fetal death in a cohort of 522women enrolled before 13 weeks’ gestation;108 of these women were classified as havingpositive antibody studies (21%). Among thisgroup, the fetal death rate was 17%, as opposedto 8% in the antibody negative women. Five ofthe 17 deaths among the antibody positivewomen occurred in association with TSHincreases. The authors speculated that thyroidantibodies might serve as a marker for otherautoimmune conditions that might, in turn, beresponsible for the fetal death. In the presentstudy, TSH measurements alone were used tostratify the rate of fetal death. It was alsoshown, however, that thyroid antibodies werepresent in a much greater proportion ofwomen with raised TSH measurements. Forthat reason, it is not possible to attributecausality to either marker with confidence.Although it is not certain that adequatetreatment of hypothyroidism can avoid thiscomplication, it is apparent that there is roomfor improvement in monitoring and treatingmaternal hypothyroidism. One possibility isthat adequate treatment may, if introducedearly enough, reduce fetal deaths in womenwith raised TSH measurements.

In a previous report, we documented anassociation between untreated hypothyroidismduring pregnancy and lower IQ in the oV-spring.14 That study also found that an averageof five years elapsed before a clinical diagnosisof hypothyroidism could be made. In a fewinstances, the diagnosis was not made until 10years later.15 The current finding that raisedTSH measurements are associated with anincreased rate of fetal death adds anotherdimension to the argument that routine TSHscreening should be evaluated as a way topotentially improve pregnancy outcome andmaternal well being.

Funding for this study was provided by general research fundsat the Foundation for Blood Research, Scarborough, Maine,USA and by a grant from the Hitchcock Foundation of Hano-ver, New Hampshire, USA.

1 Klein RZ, Haddow JE, Faix JD, et al. Prevalence of thyroiddeficiency in pregnant women. Clin Endocrinol 1991;35:41–6.

2 Willhoite MB, Bennert HW, Palomaki GE, et al. The impactof preconception counseling on pregnancy outcomes. Dia-betes Care 1993;16:450–5.

3 Haddow JE, Knight GJ, Kloza EM, et al. Cotinine-assistedintervention in pregnancy to reduce smoking and lowbirthweight delivery. Br J Obstet Gynaecol 1991;98:859–65.

4 Haddow JE, Palomaki GE, Knight GJ. EVect of parity onhuman chorionic gonadotrophin measurements andDown’s syndrome screening. J Med Screen 1995;2:28–30.

5 Dixon, WJ. BMDP Statistical Software Manual. WestSussex, England: University Press of California c/o JohnWiley & Sons, 1992.

6 Gustafson TL. True Epistat Reference Manual, Richardson.Texas: Epistat Services, 1995.

7 Glinoer D. The thyroid gland in pregnancy: a Europeanperspective. Thyroid Today 1995;18:1–11.

8 Greenman GW, Gabrielson MO, Howard-Flanders J, et al.Thyroid dysfunction in pregnancy: fetal loss and follow-upevaluation of surviving infants. N Engl J Med 1962;267:426–31.

9 Montoro M, Collea JV, Frasier SD, et al. Successfuloutcome of pregnancy in women with hypothyroidism. AnnIntern Med 1981;94:31–4.

10 Davis LE, Leveno KJ, Cunningham FG. Hypothyroidismcomplicating pregnancy. Obstet Gynecol 1988;72:108–12.

11 Leung AS, Millar LK, Koonings PP, et al. Perinatal outcomein hypothyroid pregnancies. Obstet Gynecol 1993;81:349–53.

12 Jones WS, Man EB. Thyroid function in human pregnancy.VI. Premature deliveries and reproductive failures of preg-

Table 3 Selected demographic data and rates of complications involving pregnancy anddelivery, according to TSH measurement

TSH measurement (mU/l)

<6 6–9.99 >10

Number of pregnancies 9194 172 37Vaginal bleeding (%) 11 10 11Mean gestational age at delivery (weeks) 40.2 40.3 40.2Mean birthweight (grams) 3448 3451 3498Abruptio placentae (%) 0.7 1.2 0Pregnancy induced hypertension (%) 3.9 4.7 2.7Cesarean section (%) 22 24 16Apgar <3 @ 5 minutes (%) 0.4 0 2.7Apgar <6 @ 5 minutes (%) 1.5 2.9 2.7Fetal deaths† (%) 0.9 2.9 8.1 (p<0.001)*Neonatal deaths‡ (%) 0.4 0.0 2.7

*Only comparison to reach statistical significance compared to value for group <6 mU/l.†Includes all in utero deaths between the time of enrollment and term.‡Includes all deaths occurring after a live birth, up to 1 month of age.

Maternal thyroid deficiency 129

www.jmedscreen.com

nant women with low serum butanol-extractable iodines.Maternal serum TBG and TBPA capacities. Am J ObstetGynecol 1969;104:909–14.

13 Stagnaro-Green A, Roman SH, Cobin RH, et al. Detectionof at-risk pregnancy by means of highly sensitive assaysfor thyroid autoantibodies. JAMA 1990;264:1422–5.

14 Haddow JE, Palomaki GE, Allan WC, et al. Maternalthyroid deficiency during pregnancy and subsequentpsychological development in the child. N Engl J Med1999;341:549–55.

15 Haddow JE. Screening for hypothyroidism in adults:supporting data from two population studies. J Med Screen2000;7:1–3.

Screening briefs published in JMSAIDS from maternally transmitted HIV infection 1997;4:177

Antenatal screening for neural tube defects 1998;5:167

Antenatal and neonatal screening for congenital toxoplasmosis infection 1999;6:217

Antenatal screening for Down’s syndrome 1995;2:56

Antenatal screening for â thalassaemia (and its variants) 1998;5:215

Atrial fibrillation and screening for the prevention of stroke 1998;5:111

Bone density screening for hip fracture 1994;1:203

Breast cancer 1994;1:73

Cervical cancer 1994;1:255

Colorectal cancer 1997;4:54

Congenital dislocation of the hip 1995;2:117

Cystic fibrosis 1995;3:55

Diabetic retinopathy 1995;2:126

Haemolytic disease of the newborn 1995;2:232

Helicobacter pylori testing in screening for the prevention of stomach cancer 1999;6:165

Hepatoma and chronic liver disease from maternally transmitted hepatitis B infection 1998;5:54

Hereditary haemochromatosis 1997;4:248

Lung cancer 1996;3:110

Malignant melanoma 1996;3:216

Ovarian cancer 1994;1:135

Phenylketonuria 1999;6:113

Prostate cancer 1996;3:164

Radiographic screening for stomach cancer 2000;7:120

Ruptured abdominal aortic aneurysm 1997;4:112

Screening to prevent pelvic inflammatory disease from Chlamydia trachomatis genital infection 2000;7:55

130 Allan, Haddow, Palomaki, et al

www.jmedscreen.com