Impact of Subclinical Hypothyroidism on Cardiometabolic Biomarkers in Women

Paulo H. N. Harada; Julie E. Buring; Nancy R. Cook; Michael E. Cobble; Krishnaji R. Kulkarni; Samia Mora


J Endo Soc. 2017;1(2):113-123. 

In This Article


Study Design, Sample, and Exposure Assessment

The sample comprised apparently healthy middle-aged and older women at study entry of an ongoing prospective cohort, the Women's Health Study.[14] At enrollment, women gave written informed consent and completed questionnaires on demographics, anthropometrics, medical history, and lifestyle factors. Race was self-referred, and body mass index (BMI) was measured as weight in kilograms by squared height in meters. The study was approved by the Institutional Review Board of the Brigham and Women's Hospital (Boston, MA).

As our initial inclusion criteria were similar to the original protocol, that is, women without CVD or cancer, 28,024 participants with available stored frozen plasma at baseline were eligible for our study. From those, 3914 women were randomly selected for thyroid function testing, and after excluding those on lipid-lowering therapy and ineligible thyroid categories amounted to 3321 individuals. Eligible categories based on the Roche Cobas assay recommendations by Atherotech Diagnostics Laboratory (Birmingham, AL) were defined as follows: euthyroid [TSH, 0.27 to 4.2 mIU/L; free thyroxine (FT4), 0.93 to 1.7 ng/dL), SCH (TSH > 4.2 mIU/L; FT4, 0.93 to 1.7 ng/dL), and hypothyroid (TSH > 4.2 mIU/L; FT4 < 0.93 ng/dL). Thyroid category distributions were: euthyrodism, 2571 (77.4%); SCH, 573 (17.3%); and HT, 177 (5.3%). We performed additional analyses of TSH quintiles within normal FT4 individuals (0.93 to 1.7 ng/dL), which includes only euthyroid and SCH, 3144 (94.7%).

Biomarker Assessment

Samples were obtained from stored blood in vapor-phase liquid nitrogen (−170°C) at the time of enrollment into the Women's Health Study and thawed for the following laboratory analyses. TSH and FT4 were measured at Atherotech Diagnostics Laboratory using the Roche Cobas e601 analyzer with coefficients of variation (CVs) of ≤8.7% and ≤6.6%, respectively. In a laboratory certified by the National Heart, Lung, and Blood Institute/Centers for Disease Control and Prevention Lipid Standardization program, standard lipids were measured directly with reagents from Roche Diagnostics (Indianapolis, IN)[15] and apolipoprotein (apo) B and A1 were measured with immunoturbidimetric assays (DiaSorin, Stillwater, MN). All measures had CVs of ≤5% and comprised total cholesterol, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides, apoB, and apoA1. Lipoprotein(a) [Lp(a)] was measured with a well characterized in-house immunoturbidimetric assay using a Technicon Axon autoanalyser (Miles Inc., Tarrytown, NY), rabbit anti-human lipoprotein(a) polyclonal antibodies (Q023, DAKO, Glostrup, Denmark), and a human serum lipoprotein(a) calibrator (DAKO).[16] Nuclear magnetic resonance (NMR) spectroscopy (LabCorp, previously LipoScience, Raleigh, NC) measured the methyl terminal of lipoprotein subclasses and size of LDL, HDL, and very-low-density lipoprotein (VLDL) particles using a targeted metabolomics platform (NMR LipoProfile analysis by the LP3 algorithm).[17] This platform comprises concentrations of large, medium, and small VLDL particles; large and small LDL particles; very large, large, medium, small, and very small HDL particles; and average size of VLDL, LDL, and HDL particles. As the distribution of small LDL particles is bimodal in this population, we analyzed this parameter separately in those with concentrations higher or lower than 164 nmol/L, which is the nadir between modes. The group with the higher small LDL concentration is referred to as pattern B; the group with the lower small LDL concentration is referred to as pattern A. The lipoprotein insulin resistance (LPIR) score is part of the same panel, and is a weighted composite score of lipoprotein subclasses independently related to insulin resistance.[18] It includes 6 parameters (respective maximum points and direction for more insulin resistance): VLDL size (32 points, larger size); large VLDL particles (22 points, increasing concentration); LDL size (6 points, smaller size); small LDL particles (8 points, increasing concentration); HDL size (20 points, smaller size); and large HDL particles (12 points, decreasing concentration). The summed score ranges from 0 to 100, and higher values indicate increasing insulin resistance.[18]

The inflammation and coagulation biomarkers were: high-sensitivity C-reactive protein (hs-CRP) measured by a high-sensitivity immunoturbidimetric assay on a Hitachi 917 autoanalyzer (Roche Diagnostics), with reagents and calibrators from Denka Seiken;[19] fibrinogen measured by an immunoturbidimetric assay (Kamiya Biomedical, Seattle, WA); homocysteine measured by an enzymatic assay (Catch Inc., Seattle, WA); and soluble intercellular adhesion molecule 1 (sICAM-1) measured by an enzyme-linked immunosorbent assay (R&D Systems, Minneapolis, MN). Additionally, glycan N-acetylglucosamines (GlycA) is a novel composite inflammatory biomarker of acute phase glycoproteins, mostly α1-acid glycoprotein, haptoglobin, α1-antitrypsin, α1-antichymotrypsin, and transferring.[20] Their N-acetyl methyl group protons of the N-acetylglucosamine moieties located on their biantennary, triantennary, and tetraantennary branches were measured by magnetic resonance. This signal only arises from the N-acetylglucosamine units with β1→2 and β1→6 linkages on preceding mannose residue centered at 2.00 ± 0.01 ppm. GlycA signal was successfully deconvoluted from neighboring lipoprotein, majorly triglyceride concentration in VLDL, with a CV of 4.3%.[20] Hemoglobin A1c (HbA1c) was measured by turbidimetric immunoinhibition on hemolyzed whole blood or packed red cells (Roche Diagnostics). Metabolic syndrome was classified by the presence of at least of 3 of the following: HbA1c ≥5.7 or diabetes history; blood pressure ≥130/85, hypertension diagnosis or treatment; BMI >26.7 kg/m2; triglycerides ≥150 mg/dL; or HDL cholesterol <50 mg/dL. As waist circumference was not measured until year 6 of follow-up, baseline BMI was used instead. BMI baseline cut point was chosen according to the same percentile for BMI at year 6 as for 88-cm waist circumference at that moment.

Statistical Analysis

Descriptive demographics of functional thyroid categories were displayed by percentages for categorical variables, and by mean (standard deviation) continuous variables. For univariable statistical comparisons, we used Cochran-Mantel-Haenszel for categorical variables and linear regression for trend for continuous variables.

We addressed the relationship of thyroid categories with each biomarker by adjusted linear regression analysis, with tests for linear trend, and pairwise comparisons adjusted by Bonferroni in the thyroid categories and TSH quintiles. For TSH quintiles, we tested linear trend across median TSH values within each quintile. For pairwise comparisons, the respective references were euthyroid or the bottom TSH quintiles. The covariables were age, race, household income, current smoking, systolic blood pressure, antihypertensive treatment, menopause status, and hormone replacement therapy. Skewed distributed variables were transformed by the natural logarithm [HDL cholesterol, triglycerides, Lp(a), large and small VLDL particles, small LDL particles (pattern B), hs-CRP, fibrinogen, homocysteine, and HbA1c] or by the square root (very large, large, medium, and small HDL particles) for better model fit, and then back transformed. We also conducted sensitivity analyses adjusted for BMI to address possible mediation of adiposity on thyroid function and metabolic abnormalities.

Finally, to assess the risk of having clinical metabolic syndrome across the thyroid categories or TSH quintiles, we applied logistic regression adjusted for age, race, smoking status, income, menopause status, and hormone replacement therapy. Analyses were performed using SAS version 9.3 (SAS Institute, Cary, NC).