First-Trimester Thyroid-Stimulating Hormone Levels and Gestational Diabetes Mellitus: A Retrospective Cohort Study

Background: The clinical relevance of a mildly elevated first-trimester thyroid-stimulating hormone (TSH) concentration is uncertain, recent meta-analyses disagree, and few studies have directly contrasted a TSH of 2.5–4.0 mIU/L with a TSH above 4.0 mIU/L. Methods: In this single-centre retrospective cohort study, 218 singleton pregnancies with a first-trimester TSH ≥ 2.5 and ≤10.0 mIU/L and a concurrent free thyroxine (fT4) measurement were classified as high-normal TSH (2.5–4.0 mIU/L; n = 168) or elevated TSH (>4.0–10.0 mIU/L; n = 50; subclinical hypothyroidism in 47 and biochemically overt hypothyroidism in 3). The primary outcome was gestational diabetes mellitus (GDM); secondary outcomes were a composite obstetric complication, individual obstetric and neonatal outcomes, and first- and second-trimester screening markers. Two logistic regression models adjusted for maternal age were fitted; because levothyroxine was started after the index TSH measurement, it was treated as a potential intermediate and examined only in post hoc sensitivity analyses. Results: GDM occurred in 7/50 (14.0%) women with TSH > 4.0 mIU/L versus 6/168 (3.6%) with TSH 2.5–4.0 mIU/L (Fisher’s exact test, p = 0.013; unadjusted odds ratio (OR) 4.40, 95% confidence interval (CI) 1.40–13.76; risk difference 10.4 percentage points, 95% CI 0.4–20.4). After adjustment for maternal age, TSH > 4.0 mIU/L remained associated with GDM (adjusted OR 5.24, 95% CI 1.60–17.16; p = 0.006), an estimate based on only 13 events; additional adjustment for levothyroxine use gave a similar estimate (adjusted OR 4.51, 95% CI 1.33–15.22). The composite obstetric complication did not differ (32.0% vs. 20.2%; χ2(1) = 3.02, p = 0.082; adjusted OR 1.95, 95% CI 0.96–3.98), nor did individual obstetric outcomes, neonatal outcomes or screening markers. Conclusions: First-trimester TSH > 4.0 mIU/L was associated with a higher frequency of GDM. Given the few events, the wide confidence interval, unmeasured body-mass index and thyroid-peroxidase-antibody status and the absence of a TSH < 2.5 mIU/L comparison group, the finding is hypothesis-generating and does not establish 4.0 mIU/L as a clinically discriminatory threshold.