Risk factors and a clinical prediction model for low maternal thyroid function during early pregnancy: two population-based prospective cohort studies

Tim I. M. Korevaar, Daan Nieboer, Peter H. L. T. Bisschop, Mariette Goddijn, Marco Medici, Layal Chaker, Yolanda B. de Rijke, Vincent W. V. Jaddoe, Theo J. Visser, Ewout W. Steyerberg, Henning Tiemeier, Tanja G. Vrijkotte, Robin P. Peeters

Research output: Contribution to journalArticleAcademicpeer-review

23 Citations (Scopus)

Abstract

Low maternal thyroid function during early pregnancy is associated with various adverse outcomes including impaired neurocognitive development of the offspring, premature delivery and abnormal birthweight. To aid doctors in the risk assessment of thyroid dysfunction during pregnancy, we set out to investigate clinical risk factors and derive a prediction model based on easily obtainable clinical variables. In total, 9767 women during early pregnancy (≤18 week) were selected from two population-based prospective cohorts: the Generation R Study (N = 5985) and the ABCD study (N = 3782). We aimed to investigate the association of easily obtainable clinical subject characteristics such as maternal age, BMI, smoking status, ethnicity, parity and gestational age at blood sampling with the risk of low free thyroxine (FT4) and elevated thyroid stimulating hormone (TSH), determined according to the 2·5th-97·5th reference range in TPOAb negative women. BMI, nonsmoking and ethnicity were risk factors for elevated TSH levels; however, the discriminative ability was poor (range c-statistic of 0·57-0·60). Sensitivity analysis showed that addition of TPOAbs to the model yielded a c-statistic of 0·73-0·75. Maternal age, BMI, smoking, parity and gestational age at blood sampling were risk factors for low FT4, which taken together provided adequate discrimination (range c-statistic of 0·72-0·76). Elevated TSH levels depend predominantly on TPOAb levels, and prediction of elevated TSH levels is not possible with clinical characteristics only. In contrast, the validated clinical prediction model for FT4 had high discriminative value to assess the likelihood of low FT4 levels
Original languageEnglish
Pages (from-to)902-909
JournalClinical endocrinology
Volume85
Issue number6
DOIs
Publication statusPublished - 2016

Cite this