Prevalence of Thyroid Dysfunction in Postpartum Women With Suboptimal Iodine and Selenium and Adequate Iron Status

Ying Jin; Jane Coad; Shao J. Zhou; Sheila Skeaff; Thiagarajah Ramilan; Louise Brough


Clin Endocrinol. 2021;95(6):873-881. 

In This Article

Materials and Methods

The Mother and Infant Nutrition Investigation (MINI) study was approved by the Health and Disability Ethics Committee, New Zealand (15/NTA/172), in December 2015 (ACTRN12615001028594). Prior written consent was obtained from all participants. Infants' participation was consented to by their mothers.

Study Design and Participants

The MINI study was an observational longitudinal cohort study spanning the first postpartum year in Palmerston North within the North Island of New Zealand. Data from six months postpartum are reported here, since biomarkers for iodine, selenium and iron status, and thyroid hormone concentrations were only determined at this time point. Women aged 16 years and older, who had given birth to a healthy term singleton infant aged less than three months of age, and who were breastfeeding, were invited to join the study. Women were excluded: 1) if they had pre-existing or developed significant health problems, such as metabolic disease and cancer; and 2) if they had been diagnosed or treated at any time for hyperthyroidism or hypothyroidism. The full study protocol is published.[10]

Data Collection

Assessment of Thyroid Hormones and Thyroid Volume. Thyroid hormone biomarkers (serum-free T3, serum-free T4, and thyroid-stimulating hormone [TSH]) were measured using the chemiluminescent microparticle immunoassay (CMIA) method. The adult reference ranges of these biomarkers were used due to the absence of reference ranges specifically for lactating women. Thyroid peroxidase antibody (TPOAb) concentration above 10 IU/mL was regarded as indicative of a potential autoimmune disorder. Reference ranges for euthyroid are TSH, 0.40–4.00 mIU/L; free T4, 10–24 pmol/L; and free T3, 2.5–6.0 pmol/L. These reference ranges were used to calculate the prevalence of thyroid dysfunction including subclinical hypothyroidism (TSH > 4.00 mIU/L and normal free T4), overt hypothyroidism (TSH > 4.00 mIU/L and free T4 < 10 pmol/L), subclinical hyperthyroidism (TSH < 0.40 mIU/L and normal free T4 and free T3) and overt hyperthyroidism (TSH < 0.40 mIU/L and free T4 > 24 pmol/L, and free T3 > 6.0 pmol/L).[11]

Thyroid volume was measured by a portable ultrasound (Terason uSmart3200T™, USA) equipped with a linear transducer (7–15 MHz). The total volume of thyroid gland was calculated for each lobe as: Anteroposterior (AP) diameter × Width × Length × 0.479.[12] A total volume greater than 18 ml was defined as thyroid enlargement, based on the normative thyroid volume in an iodine-sufficient female population.[12] Any participant with observed abnormalities was referred without delay to clinical health professionals for further assessment.

Assessment of Iodine, Selenium and Iron Status. Maternal non-fasting spot urine samples (approximately 120 mL) were collected to measure maternal urinary iodine concentration (UIC, μg/L) and creatinine concentration to determine maternal urinary iodine-to-creatinine ratio (μg/g). Women were asked to provide a breast milk sample (approximately 30–50 ml) using an electric breast pump, if needed, and breast milk iodine concentration (BMIC, μg/L) was determined. The timing of breast milk collection was not standardised, although all samples were collected before 12 noon on the study visit day. All samples were stored without preservative at −20°C prior to analysis. Non-fasting maternal venous blood samples (22 ml) were collected by a phlebotomist and separated into plasma and serum samples before storage at −80°C.

Iodine concentrations in urine and breast milk samples were determined by Hill Laboratories, Hamilton, New Zealand, using inductively coupled plasma mass spectrometry. Each batch (n = 25) of urinary samples was analysed together with an external reference standard (Seronorm Trace Elements Urine, L-2, Norway) giving a mean ± SD iodine concentration of 286 ± 12 μg/L, (published value: 297 μg/L) with a coefficient of variance (CV) of 4.2% (n = 14). Each batch (n = 25) of breast milk samples was analysed together with an external reference standard (Skimmed milk powder, Elements in organic matrix, European Reference Material) giving a mean ± SD iodine concentration of 1.603 ± 0.029 mg/L (published value: 1.78 mg/L), with a coefficient of variance (CV) of 4.9% (n = 6). Serum thyroglobulin (Tg) was measured by the Beckman Coulter Access method, and anti-thyroglobulin antibodies were determined by the CMIA method, at Canterbury Health Laboratories, Christchurch, New Zealand. Serum Tg has been suggested as a biomarker to assess iodine status reflecting a period of weeks or months;[13] if anti-thyroglobulin antibodies were detected positive (≥10 IU/ml), Tg concentrations were disregarded.

Plasma selenium concentration was assessed by the inductively coupled plasma spectrometry method at Canterbury Health Laboratories, New Zealand, which is accredited with International Accreditation New Zealand (IANZ). Accuracy was assessed by using external certified reference material (mean ± SD) Seronorm Trace Elements serum 1 (86.65 ± 2.67 μg/L) and serum 2 (138.23 ± 4.03 μg/L). The two Seronorm plasma quality controls were run with each batch of selenium analysis to determine the interassay CV which was 3.1% (n = 142) at 88 μg/L and 2.9% (n = 142) at 139 μg/L, respectively. A plasma selenium concentration of 95 μg/L has been suggested to saturate GPX activity;[14] this was used as a cut-off for the current study.

Haemoglobulin (Hb), serum ferritin (SF) and soluble transferrin receptor (sTfR) concentrations were determined to evaluate maternal iron status. Hb concentrations were measured using a handheld Hemocue Hb 201+ device (HemoCue® Hb 201+, Sweden).[15] sTfR (using the Nephelometry method) was measured to determine iron demand versus iron supply. Serum ferritin (using the CMIA method) and C-reactive protein (CRP) were determined (using an immunoturbidimetric method analysed on an Abbott c series analyser), and, if CRP ≥ 8 mg/L (indicating inflammation), the serum ferritin concentration was disregarded. The iron status of participants was defined using the following definitions: sufficient iron stores: SF ≥ 12 μg/L and Hb ≥ 120 g/L; anaemia without iron deficiency: SF ≥ 12 μg/L but Hb <120 g/L; iron deficiency (ID): SF < 12 μg/L and Hb ≥ 120 g/L; iron-deficiency anaemia (IDA): SF < 12 μg/L and Hb < 120 g/L.[16]

Data Analysis

All data were analysed using IBM SPSS (Statistics Package for the Social Sciences, IBM) version 20 and R statistical programme (

Data were tested for normality using the Shapiro-Wilk test. Nonparametric data were expressed as median (25th, 75th percentile), and parametric were data expressed as mean (± standard deviation; SD). Bivariate correlations were tested using either the parametric Pearson correlations or the nonparametric Spearman's rho correlation coefficient, as appropriate. Seventy-four blood samples were obtained, and three samples from women who ceased to breastfeed were excluded from further analysis. Differences in categorical variables were tested by Fisher's exact t test. Plasma selenium concentrations were split into two categories (≥ 95 μg/L or < 95 μg/L) for comparison with the biomarkers; independent t test was used for parametric data after natural log transformation (including UIC, urinary iodine-to-creatinine ratio, BMIC, Hb, SF, free T3:T4), and biomarkers which were unable to be transformed into parametric data (including serum Tg, sTfR, and TSH) were tested by Mann-Whitney U test. A logistic regression model was employed to consider factors likely to influence abnormal TSH concentrations (< 0.4 mIU/L or > 4.0 mIU/L). The dependent variable was binary, with the value of one being abnormal TSH concentration and the value of zero being normal TSH concentration. Age, parity and iodine status have been suggested as risk factors for maternal thyroid disease.[17] Selenium is involved in thyroid autoimmunity,[7] and iron deficiency impairs the synthesis of thyroid hormones.[18] Therefore, age of participants; parity (categorical variable as primiparity and multiparity); UIC; plasma selenium; and sTfR were selected as covariates to enter the logistic regression model simultaneously (Table 1). Average marginal effects are less sensitive to changes in the specification of the logistic regression model when compared to odds ratios.[19] Therefore, average marginal effects were calculated to illustrate the effect of small changes in covariates on the binary dependent variable (TSH concentration).