Résumé : Objective: Guidelines on the management of thyroid dysfunction during pregnancy have recently been updated and, for the diagnosis of subclinical hypothyroidism (SCH), a thyroid-stimulating hormone (TSH) upper reference limit (cut-off) of 4.0 mIU/L has been proposed when no institutional values are available. It is also suggested that serum TSH and thyroid autoimmunity (TAI) may be different according to the ethnic background of the women. We therefore determined the prevalence of TAI and SCH in pregnant women with different ethnic backgrounds and, to define SCH, we used different first trimester TSH upper reference cut-offs (institutional, ethnicity-specific, 2.5 mIU/L [Endocrine Society] and 4.0 mIU/L [American Thyroid Association]). Design: Cross-sectional data analysis of 1683 pregnant women nested within an ongoing prospective database of pregnant women. Method: The study was performed in a single centre in Brussels, Belgium. During the first antenatal visit, thyroid peroxidase antibodies (TPO-abs), TSH and free T4 (FT4) were measured and baseline characteristics recorded. Data from 481 women with sub-Saharan (SaBg; 28.6%), 754 North African (NaBg; 44.8%) and 448 Caucasian (CaBg; 26.6%) backgrounds were analysed. For the calculation of TSH reference ranges, women with TAI, outliers, twin and assisted pregnancies were excluded. Results: The prevalence of TAI was significantly lower in the SaBg group than in NaBg and CaBg groups (3.3% vs 8.6% and 11.1%; P<.001, respectively). Median TSH was significantly lower in SaBg and NaBg groups as compared with the CaBg group (1.3 and 1.4 vs 1.5 mIU/L; P=.006 and.014, respectively). The prevalence of women with SCH was comparable between all groups when 2.5 mIU/L was used as cut-off, but when 4.0 mIU/L or the institutional cut-off (3.74 mIU/L) was used, it was significantly higher in the CaBg group vs the NaBg group (5.4% vs 2.1% and 7.1% vs 3.3%, P=.008 and.013, respectively). The use of ethnicity-specific cut-offs did not change the prevalence of SCH as compared to the use of institutional cut-offs. However, when these cut-offs were used, the prevalence of SCH reduced by >70% (4.5% instead of 16.7%; P<.001) relative to the 2.5 mIU/L cut-off. Conclusions: Pregnant women with a sub-Saharan African background had a lower prevalence of TAI and TSH levels as compared with women from other backgrounds. The use of ethnicity-specific TSH cut-offs in early pregnancy was not more specific for the diagnosis of SCH as compared to the use of the institutional cut-off.