The importance of maternal thyroid hormones for fetal central nervous system development is well-established. Lack of dietary iodine, which causes severe maternal and fetal thyroid deficiency, may result in profound neurological impairments among offspring.1,2 In 1999, 2 studies found that pregnant women with mild hypothyroidism were at risk of having children with developmental delays.3,4 Since that time, subclinical hypothyroidism, defined as elevation of thyroid stimulating hormone (TSH) with normal serum free thyroxine, has been associated with increased risk for preterm birth before 35 weeks, potentially leading to long-term sequelae.5 The American Thyroid Association, American Association of Clinical Endocrinologists, and Endocrine Society recently recommended levothyroxine replacement for pregnant women with subclinical hypothyroidism, with the goal of restoring the TSH concentration to the reference range.6 Importantly, it is the position of The American College of Obstetricians and Gynecologists (ACOG) that since there have been no intervention trials demonstrating efficacy of screening or treatment, it would be premature to recommend routine screening for hypothyroidism during pregnancy.7
Subclinical hypothyroidism occurs in 2.3% of pregnancies, or up to 100,000 pregnancies annually in the United States.5 Unfortunately, use of serum TSH thresholds derived from nonpregnant populations may not be reliable once a woman becomes pregnant, because of secretion of placental human chorionic gonadotropin (hCG) throughout gestation. Human chorionic gonadotropin is produced in prodigious amounts early in pregnancy and, by stimulating thyroid hormone secretion, has a characteristic suppressive effect on TSH. Indeed this is a known cause of misdiagnosis of subclinical hyperthyroidism. Glinoer et al8 found that such TSH suppression occurred in 18% of early pregnancies, and effects of hCG on TSH may be even more pronounced in twin gestations.9 Of greater concern is the potential to miss the identification of women with mild hypothyroidism because their TSH concentrations are no longer above the upper limit of normal.
We conducted a prospective observational study of our entire pregnant population to estimate a normal reference range for TSH at each week of gestation. Previously, our group described pregnancy outcomes in 404 women with subclinical hypothyroidism from a cohort of 17,298 singleton pregnancies with TSH screening in the first half of gestation.5 There we identified significant suppression of TSH by hCG, motivating us to further investigate TSH beyond 20 weeks and to construct nomograms that would be valid throughout gestation for twin as well as singleton pregnancies. To evaluate the impact of having such a reference, a second objective was to compare our detection of subclinical thyroid disease with that obtained using the published reference range for the assay, as is currently done in practice across the country. Because TSH results might be specific to our population and even to the assay itself, we reasoned that conversion to multiples of the median (MoM) would facilitate comparisons across different laboratories.
SUBJECTS AND METHODS
All pregnant women who enrolled for prenatal care at our hospital between December 1, 2000, and November 30, 2001, were screened for thyroid disease at their initial visit. Testing was performed with the Immulite 2000 third-generation TSH assay (Diagnostic Products Corporation, Los Angeles, CA), which has a detection limit of 0.002 mU/L. The assay reference range is 0.4–4.0 mU/L, representing the central 95% of values from nonpregnant adults with no known thyroid dysfunction. Thyroid-stimulating hormone evaluation was conducted in a blinded fashion. Results were entered into a computerized obstetrical database that contains de-identified data, using a secure file server that is password protected and encrypted. The study was approved by the Institutional Review Board of the University of Texas Southwestern Medical Center.
We considered that a small number of women would have overt, undiagnosed thyroid disease and would benefit from prompt medical treatment. As a clinical safeguard, preliminary TSH percentile values were derived for the entire group without adjustment for gestational age after the first 2,000 specimens had been processed. A free thyroxine level was obtained if the TSH value was above the 95th percentile or below the 5th percentile. Women with free thyroxine levels outside established laboratory norms were contacted via phone and sent a letter for immediate referral to a special obstetrical complications clinic for evaluation and treatment. The Institutional Review Board of the University of Texas Southwestern Medical Center approved this identification and referral of women with clinical thyroid disease. At the end of the study period, all TSH values were compiled, including these 2,000, and a nomogram specific for gestational age and fetal number was then created.
Gestational age at screening was derived retrospectively from our database, using the obstetric estimate of gestational age recorded at delivery. Outcomes for all women and their infants are entered into this database by perinatal research nurses who assess the data for consistency and completeness before electronic storage. Gestational age was based on the last menstrual period (LMP), with sonography performed if there were discrepancies between fundal height and LMP or if the LMP was uncertain. Sonography was routinely performed in twin pregnancies and was used for dating if discrepant from the LMP. This method of gestational age determination has been found to correlate well with sonographic and pediatric estimates in our population.10 Only women who initially presented for care between 6 and 42 weeks of gestation and were subsequently delivered of live or stillborn singletons or twins weighing at least 500 g were included. Triplets and higher-order multiples were excluded.
The TSH concentration was considered elevated if more than 2 standard deviations above the population mean for week of gestation, ie, above the 97.5th percentile on our nomogram. Thyroid-stimulating hormone was considered suppressed if more than 2 standard deviations below the mean for week of gestation. Thresholds were derived for singleton and twin pregnancies. To determine the impact of using these thresholds, we calculated the percentage of women with TSH elevation based on our gestational age-specific nomograms who would not have been identified had a standard reference range (0.4–4.0 mU/L) been used.
Thyroid-stimulating hormone reference ranges were constructed using quantile regression analysis, a form of regression in which the specified quantile of TSH is expressed as a function of gestational age and, when evaluating the twin pregnancies, fetal number.11 The addition of the twin nomogram was accomplished by using multivariable regression, including a twin/singleton factor, a gestational age factor, and their interaction. Because the relationship between TSH and gestational age may not be linear, a cubic smoothing spline was used, with knots fixed at 3-week intervals. The model was estimated for each quantile independently. Statistical analyses were performed by using χ2 for categorical variables and analysis of variance (ANOVA). P values less than .05 were considered significant. SAS 9.0 (SAS Institute, Cary, NC) was used for statistical computation.
During the 1-year study period, 13,737 pregnant women were screened for thyroid disease and delivered infants weighing at least 500 g. Excluded from analysis were 5 sets of triplets and 1 set of quadruplets; the remaining 13,599 singleton and 132 twin pregnancies composed the study group. None of the pregnancies in this series resulted from assisted reproductive technology or was reduced from a higher-order multiple gestation. Most women presented for care and underwent screening before midpregnancy—42% by 12 weeks and 69% by 20 weeks. Overall, TSH was evaluated in 95% before 37 weeks.
Selected demographic characteristics for these singleton and twin pregnancies are presented in Table 1. There were no significant differences with respect to maternal age or nulliparity. Hispanic women made up a slightly smaller percentage of the twin cohort, with a correspondingly larger percentage of twins born to African-American women, P < .05.
Empirical TSH percentile values for the singleton pregnancies are presented in Table 2, shown according to gestational age. The median TSH concentration decreased in the first trimester, with nadir at 10 weeks of gestation, returning to baseline at 16–18 weeks. In both singletons and twins, the first-trimester decrease in TSH was statistically significant (P < .001), and the degree of suppression was greater in twin pregnancies (P < .001). A quantile regression nomogram is shown in Figure 1, with singleton and twin pregnancies shown separately. Represented in the figure are the 97.5th and 2.5th percentiles, the threshold values for diagnosing significant TSH elevation and suppression, respectively. In the first half of pregnancy, the threshold for identifying TSH elevation was approximately 0.4 mU/L lower in women with twin pregnancies than in those with singletons.
There were 342 women with singleton pregnancies whose TSH values were above the 97.5th percentile, ie, more than 2 standard deviations above the mean for week of gestation. Of note, 95 of these women (28%) would not have been identified with TSH elevation if the manufacturer's recommended threshold of 4.0 mU/L had been used (Fig. 1). Similarly, 340 women with singleton pregnancies had TSH values below the 2.5th percentile. If the recommended threshold of 0.4 mU/L had been used, 1,448 euthyroid women (11%) would have been incorrectly characterized as abnormal. None of the 132 twin pregnancies had TSH values above 4.0 mU/L, and 15 (11%) had TSH values below 0.4 mU/L.
To facilitate use of our gestational age–specific data in other populations, TSH values were converted to multiples of the median (MOM). Shown in Figure 2 are the 97.5th percentile values for TSH in singleton and twin pregnancies, presented as multiples of the singleton median for each week of gestation. The 97.5th percentile is approximately 4.0 MoM in first-trimester singletons and 3.5 MoM in first-trimester twins. Thereafter, the approximate upper limit of normal was 2.5 MoM for both singleton and twin pregnancies.
This is a population-based study of the normal range of maternal serum thyroid-stimulating hormone (TSH) throughout gestation in singleton and twin pregnancies. Our results demonstrate that physiologic suppression of TSH during pregnancy has a predictable and substantial impact on the reference range. Indeed, more than a fourth of women with abnormally elevated TSH levels would not have been detected by using the assay recommended upper limit of normal for nonpregnant individuals, 4.0 mU/L. The American Thyroid Association, American Association of Clinical Endocrinologists, and Endocrine Society recently redefined the upper limit of serum TSH concentration to be 4.5 mU/L.6 Had we used this threshold in our current study, we would have missed more than a third with subclinical hypothyroidism. Such a possibility had been previously suggested by other investigators.12
We also found that TSH suppression is more frequent and more pronounced in women with twin pregnancies. In fact, using the above cutoffs, none of the twin pregnancies in this series would have been diagnosed with TSH elevation (Fig. 1). In 2 previous reports in which TSH concentrations were compared, the TSH differences between singleton and twin pregnancies were not significantly different, likely due to the small sample size.9,13 Our data indicate that if TSH screening is performed in twin pregnancies, a separate reference range is needed.
There are other considerations that may affect application of our findings to different cohorts. For example, population demographics differ considerably across the United States, and it is well known that the various commercially available TSH assays have slightly different, albeit overlapping, reference ranges.14 To obviate this, we converted the 97.5th percentiles for singleton and twin pregnancies into multiples of the singleton median, mirroring the alpha-fetoprotein screening system.15 This would allow our findings to be used by other laboratories, after they have established their own gestational age–specific medians. Additional research to validate the extrapolation of our specific MoM thresholds to other centers would be an important next step.
The normative curves now presented will be useful for more accurate identification of women with thyroid deficiency that requires thyroxine replacement. Regarding subclinical hypothyroidism, we acknowledge the current controversy, and even contention, surrounding the issues of routine screening and treatment during pregnancy. At this juncture, we do not endorse universal prenatal screening for subclinical hypothyroidism, which was recently recommended in a joint management statement by American Association of Clinical Endocrinologists, the American Thyroid Association, and the Endocrine Society.16 We believe there are currently no data to justify identification of women with subclinical hypothyroidism, those who by definition have normal free thyroxine concentrations. Our view is also shared by ACOG.7 Further studies are necessary to determine whether hormonal interdiction in such women might improve perinatal outcome. The normative curves now presented may also be used for these studies.
1. Delange F. The disorders induced by iodine deficiency. Thyroid 1994;4:107–28.
2. Utiger RD. Maternal hypothyroidism and fetal development. N Engl J Med 1999;341:601–2.
3. Pop VJ, Kuljpens JL, van Baar, AL, Verkerk G, van Son, MM, de Vijlder, JJ, et al. Low maternal free thyroxine concentrations during early pregnancy are associated with impaired psychomotor development in infancy. Clin Endocrinol 1999;50:149–55.
4. Haddow JE, Palomaki GE, Allan WC, Williams JR, Knight GJ, Gagnon J, et al. Maternal thyroid deficiency during pregnancy and subsequent neuropsychological development of the child. N Engl J Med 1999;341:549–55.
5. Casey BM, Dashe JS, Wells CE, McIntire DD, Byrd W, Leveno KJ, et al. Subclinical hypothyroidism and pregnancy outcomes. Obstet Gynecol 2005;105:239–45.
6. Surks MI, Ortiz E, Daniels GH, Sawin CT, Col NF, Cobin RH, et al. Subclinical thyroid disease: scientific review and guidelines for diagnosis and management. JAMA 2004;291:228–38.
7. Thyroid disease in pregnancy. ACOG Practice Bulletin No. 32. American College of Obstetricians and Gynecologists. Obstet Gynecol 2001;98:879–88.
8. Glinoer D, De Nayer, P, Robyn C, Lejeune B, Kinthaert J, Meuris S. Serum levels of intact human chorionic gonadotropin (HCG) and its free α and \S subunits in relation to maternal thyroid stimulation during normal pregnancy. J Endocrinol Invest 1993;16:881–8.
9. Grun JP, Meuris S, De Nayer, P, Glinoer D. The thyrotropic role of human chorionic gonadotropin (hCG) in the early stages of twin (versus single) pregnancies. Clin Endocrinol 1997;46:719–25.
10. McIntire DD, Bloom SL, Casey BM, Leveno KJ. Birthweight in relation to morbidity and mortality among newborn infants. N Engl J Med 1999;340:1234–8.
11. Austin PC, Tu JV, Daly PA, Alter DA. The use of quantile regression in health care research: a case study examining gender differences in the timeliness of thrombolytic therapy. Stat Med 2005;24:791–816.
12. Smallridge RC, Ladenson PW. Hypothyroidism in pregnancy: consequences to neonatal health. J Clin Endocrinol Metab 2001;86:2349–53.
13. Ogueh O, Hawkins AP, Abbas A, Carter GD, Nicolaides KH, Johnson MR. Maternal thyroid function in multifetal pregnancies before and after fetal reduction. J Endocrinol 2000;164:7–11.
14. College of American Pathologists. Survey 2000 K/KN-C Ligand (General): Participant Summary Report. Northfield (IL): College of American Pathologists; 2000. p. 29–31.
15. Wald NJ, Cuckle H, Brock JH, Peto R, Polani PE, Woodford FP. Maternal serum alpha-fetoprotein measurement in antenatal screening for anencephaly and spina bifida in early pregnancy. Report of the U.K. Collaborative Study on alpha-fetoprotein in relation to neural tube defects. Lancet 1977;1:1323–32
© 2005 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
16. Gharib H, Tuttle RM, Baskin J, Fish LH, Singer PA, McDermott MT. Consensus statement #1: Subclinical thyroid dysfunction: a joint statement on management from the American Association of Clinical Endocrinologists, the American Thyroid Association, and the Endocrine Society. Thyroid 2005;15:24–8.