Abstract Objective Thyroid hormones are ubiquitously involved in human metabolism. However, the precise molecular patterns associated with alterations in thyroid hormones levels remain to be explored in detail. A number of recent studies took great advantage of metabolomics profiling to outline the metabolic actions of thyroid hormones in humans. Methods Among 952 participants in the Study of Health in Pomerania, data on serum free thyroxine (FT4) and thyrotropin and comprehensive nontargeted metabolomics data from plasma and urine samples were available. Linear regression analyses were performed to assess the association between FT4 or thyrotropin and metabolite levels. Results and Conclusion After accounting for major confounders, 106 of 613 plasma metabolites were significantly associated with FT4. The associations in urine were minor (12 of 587). Most of the plasma metabolites consisted of lipid species, and subsequent analysis of highly resolved lipoprotein subclasses measured by proton nuclear magnetic resonance spectroscopy revealed a consistent decrease in several of these species (e.g., phospholipids) and large low-density lipoprotein and small high-density lipoprotein particles. The latter was unique to men. Several polyunsaturated and saturated fatty acids displayed an association with FT4 in women only. A random forest-based variable selection approach using phenotypic characteristics revealed higher alcohol intake in men and an adverse thyroid state and menopause in women as the putative mediating factors. In general, our observations have confirmed the lipolytic and lipogenic effect of thyroid hormones even in the physiological range and revealed different phenotypic characteristics (e.g., lifestyle differences) as possible confounders for sex-specific findings. Thyroid hormones (THs), with triiodothyronine (T3) and thyroxine (T4) as the major secretion products of the thyroid gland, play a vital role in the human organism owing to the substantial expression of their corresponding nuclear receptors in almost all tissues (1). They are crucial for the regulation of energy expenditure, carbohydrate and lipid metabolism, and thermogenesis (2). Furthermore, THs exert effects in several other tissues, including the brain, liver, and gut (3). Within the circulation, THs are mainly bound to proteins (3) and appear only in minor free fractions (FT3, FT4), which are considered as the biologically active pool. Synthesis and secretion of THs are under control of the hypothalamic–pituitary–thyroid axis, with thyrotropin (TSH) released from the pituitary as the main stimulus. Release of TSH, in turn, is suppressed by high serum TH levels, establishing a negative feedback loop that is regarded as the reference standard for diagnosis of thyroid disorders in clinical practice (4). In a very recent study (5), we were able to show that a hypothesis-free profiling of the small molecule and protein content of plasma samples (so-called metabolomics and proteomics) in a clearly defined human model of thyrotoxicosis allowed for the derivation of a TSH- and FT4-independent diagnostic signature. The signature comprised molecules likely representing increased resting energy expenditure accompanied by augmented defense against systemic oxidative stress and a prothrombotic and proinflammatory state (5). The results (e.g., elevated acylcarnitine levels) were in agreement with previous population-based findings (6) and recovery of euthyroidism among women with Graves disease (7). Profiling of urine samples using proton nuclear magnetic resonance (NMR) spectroscopy among large cohorts (8, 9) complemented the plasma fingerprint with respect to glucose metabolism and nutritional behavior and, in particular, questioned the proposed diagnostic redundancy of TSH and FT4. In part, these observational results were supported by intervention studies in rodent models (10, 11). In summary, those studies emphasized the particular value of metabolomics techniques to reveal novel modes of action in relation to THs and putatively improve the diagnosis and treatment of thyroid disorders. All the previous work relied on distinct platforms for metabolomics analyses, which might account for the discrepancies and missing replication of findings seen between the studies. To overcome this drawback, the present study has related the alterations in serum TSH and FT4 with metabolomics data using comprehensive profiling of plasma and urine samples measured via mass spectrometry (MS) and proton NMR spectroscopy within a sample from the general population. Materials and Methods Study population The Study of Health in Pomerania (SHIP-TREND) is a population-based study conducted in West Pomerania, a rural region in northeast Germany, and a detailed description of the sampling procedure and the study population has been previously reported (12). In total, 4420 subjects chose to participate (50.1% response). All participants gave written informed consent before participating in the study. The ethics committee of the University of Greifswald approved the study, which conformed to the principles of the Declaration of Helsinki. The SHIP data are publicly available for scientific and quality control purposes by application (available at: www.community-medicine.de). For a subsample of 1000 subjects without self-reported diabetes, plasma and urine metabolomics data based on MS and NMR were available. Subjects with missing values for either TSH or FT4 (n = 36) or confounders considered (n = 3), and subjects reporting intake of medication that influences hormone levels [n = 9; amiodarone (ATC code C01BD01) or oral glucocorticoids (ATC code H02AB)] were excluded, resulting in a study population of 952 subjects. Laboratory measurements and phenotypic characterization Smoking status (current, former, or never smokers), daily alcohol consumption, and physical activity (≥1 hour training a week) were assessed using computer-aided personal interviews. Participants’ self-reported history of thyroid disease was recorded and included hyperthyroidism, hypothyroidism, goiter, thyroid nodules, or other (Supplemental Table 1). The waist circumference was measured midway between the lower rib margin and the iliac crest in the horizontal plane. The mean daily alcohol consumption was calculated using beverage-specific pure ethanol volume proportions. Fasting blood samples were taken from the cubital vein with the participants in the supine position between 7:00 am and 12:00 pm. In the same period, spot urine samples were taken. All samples were either analyzed immediately or stored at −80°C in the Integrated Research Biobank (LiCONiC Instruments, Mauren Principality, Liechtenstein). The serum levels of TSH, FT3, and FT4 were measured using an immunoassay (Dimension VISTA; Siemens Healthcare Diagnostics, Eschborn, Germany) with a functional sensitivity of 0.005 mU/L for TSH, 0.77 pmol/L for FT3, and 1.3 pmol/L for FT4. Serum cystatin C, lipids [total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL), triglycerides], and serum activities of alanine aminotransferase were measured using standard methods (Dimension VISTA; Siemens Healthcare Diagnostics). The cystatin C-based estimated glomerular filtration rate was calculated using the CKD-EPI equation (13). Metabolomics measurements A detailed description of all applied measurement techniques is given in the Supplemental Methods. In brief, four different approaches were combined: (1) nontargeted MS-based profiling of plasma and urine samples, as reported previously (14); (2) targeted MS-based profiling of plasma samples using the AbsoluteIDQ p180 Kit (Biocrates Life Sciences AG, Innsbruck, Austria); (3) NMR-based profiling of urine samples, as reported previously (15); and (4) NMR-based profiling of plasma samples to derive lipoprotein particles. After quality control and preprocessing (Supplemental Methods), 613 plasma and 587 urine metabolites were available for statistical analysis. A total of 177 plasma metabolites and 302 urine metabolites could not be unambiguously assigned to a chemical identity and have been referred to hereafter with the notation “X” followed by a unique number. Data on lipoprotein particles comprised 117 measures describing the gradient from very-low-density lipoprotein (VLDL) particles to HDL particles, including their triglycerides, cholesterol, free cholesterol, phospholipid, and apolipoprotein (Apo)B, ApoA1, and ApoA2 content. Statistical analysis Linear regression models were created to assess the association between TSH, FT3, or FT4 (independent) and plasma and urine metabolites (dependent). The concentrations of TSH were log-transformed. Possible confounding was avoided by controlling for age, sex, waist circumference, smoking behavior, physical activity, estimated glomerular filtration rate, and serum alanine aminotransferase activities. The models were separated by sex, if an at least nominally statistically significant (P < 0.05) interaction term between one of the hormones and sex became obvious. Furthermore, all analyses were repeated considering only euthyroid subjects [n = 771; TSH, 0.49 to 3.29 (16); FT4, 9.4 to 18.5 pmol/L; and no intake of thyroid-related medication; ATC code, H03A/B]. Correction for multiple testing was performed using the Benjamini-Hochberg procedure, controlling the false-discovery rate (FDR) at 5%. To elucidate sex-specific findings, we used Boruta feature selection (17) to derive important predictors of serum FT4 levels separately for men and women (Supplemental Methods). Statistical analyses were performed using SAS statistical software, version 9.4 (SAS Institute, Inc., Cary, NC) and R, version 3.1.1 (R Foundation for Statistical Computing, Vienna, Austria). Results The study included subjects aged 20 to 81 years. Men showed a less favorable health status compared with women, including an unfavorable lipid profile and higher alcohol consumption (Table 1). Although women had a greater prevalence of thyroid disease, no difference in TSH values became apparent; however, the FT4 values were slightly lower and FT3 values were slightly higher in men (Table 1). Table 1. General Characteristics of Study Population Stratified by Sex Characteristic Men (n = 419) Women (n = 533) P Value a Age, y 49 (39; 61) 51 (40; 60) 0.71 Smoking, % Never 32.0 50.5 <0.01 Former 44.6 28.0 Current 23.4 21.5 Physical activity, % >1 h/wk 27.7 25.5 0.45 <1 h/wk 72.3 74.5 Alcohol consumption, g/d 8.7 (3.1; 18.4) 1.4 (0.7; 5.5) <0.01 BMI, kg/m2 27.7 (25.0; 30.2) 26.1 (23.1; 29.5) <0.01 WC, cm 94 (86; 102) 81 (74; 90) <0.01 TSH, mU/L 1.10 (0.78; 1.51) 1.21 (0.83; 1.76) 0.02 FT4, pmol/L 13.1 (12.2; 14.2) 13.4 (12.5; 14.6) <0.01 FT3, pmol/L 5.0 (4.6; 5.3) 4.5 (4.2; 4.9) <0.01 eGFR, mL/min/1.73 m2 117 (108; 125) 112 (103; 121) <0.01 LDL cholesterol, mmol/L 3.41 (2.76; 4.00) 3.32 (2.74; 3.98) 0.43 HDL cholesterol, mmol/L 1.27 (1.10; 1.48) 1.58 (1.35; 1.83) <0.01 Total cholesterol, mmol/L 5.3 (4.6; 6.1) 5.5 (4.9; 6.2) <0.01 Triglycerides, mmol/L 1.30 (0.92; 1.89) 1.15 (0.84; 1.61) <0.01 ALT, µkat/L 0.47 (0.35; 0.65) 0.30 (0.24; 0.42) <0.01 History of thyroid disease, % 6.0 29.3 <0.01 Characteristic Men (n = 419) Women (n = 533) P Value a Age, y 49 (39; 61) 51 (40; 60) 0.71 Smoking, % Never 32.0 50.5 <0.01 Former 44.6 28.0 Current 23.4 21.5 Physical activity, % >1 h/wk 27.7 25.5 0.45 <1 h/wk 72.3 74.5 Alcohol consumption, g/d 8.7 (3.1; 18.4) 1.4 (0.7; 5.5) <0.01 BMI, kg/m2 27.7 (25.0; 30.2) 26.1 (23.1; 29.5) <0.01 WC, cm 94 (86; 102) 81 (74; 90) <0.01 TSH, mU/L 1.10 (0.78; 1.51) 1.21 (0.83; 1.76) 0.02 FT4, pmol/L 13.1 (12.2; 14.2) 13.4 (12.5; 14.6) <0.01 FT3, pmol/L 5.0 (4.6; 5.3) 4.5 (4.2; 4.9) <0.01 eGFR, mL/min/1.73 m2 117 (108; 125) 112 (103; 121) <0.01 LDL cholesterol, mmol/L 3.41 (2.76; 4.00) 3.32 (2.74; 3.98) 0.43 HDL cholesterol, mmol/L 1.27 (1.10; 1.48) 1.58 (1.35; 1.83) <0.01 Total cholesterol, mmol/L 5.3 (4.6; 6.1) 5.5 (4.9; 6.2) <0.01 Triglycerides, mmol/L 1.30 (0.92; 1.89) 1.15 (0.84; 1.61) <0.01 ALT, µkat/L 0.47 (0.35; 0.65) 0.30 (0.24; 0.42) <0.01 History of thyroid disease, % 6.0 29.3 <0.01 Data presented as median (25th; 75th percentile). Abbreviations: ALT, alanine aminotransferase; BMI, body mass index; eGFR, estimated glomerular filtration rate; WC, waist circumference. a Mann-Whitney U test for continuous data and χ2 test for categorical data were used for comparisons. View Large Table 1. General Characteristics of Study Population Stratified by Sex Characteristic Men (n = 419) Women (n = 533) P Value a Age, y 49 (39; 61) 51 (40; 60) 0.71 Smoking, % Never 32.0 50.5 <0.01 Former 44.6 28.0 Current 23.4 21.5 Physical activity, % >1 h/wk 27.7 25.5 0.45 <1 h/wk 72.3 74.5 Alcohol consumption, g/d 8.7 (3.1; 18.4) 1.4 (0.7; 5.5) <0.01 BMI, kg/m2 27.7 (25.0; 30.2) 26.1 (23.1; 29.5) <0.01 WC, cm 94 (86; 102) 81 (74; 90) <0.01 TSH, mU/L 1.10 (0.78; 1.51) 1.21 (0.83; 1.76) 0.02 FT4, pmol/L 13.1 (12.2; 14.2) 13.4 (12.5; 14.6) <0.01 FT3, pmol/L 5.0 (4.6; 5.3) 4.5 (4.2; 4.9) <0.01 eGFR, mL/min/1.73 m2 117 (108; 125) 112 (103; 121) <0.01 LDL cholesterol, mmol/L 3.41 (2.76; 4.00) 3.32 (2.74; 3.98) 0.43 HDL cholesterol, mmol/L 1.27 (1.10; 1.48) 1.58 (1.35; 1.83) <0.01 Total cholesterol, mmol/L 5.3 (4.6; 6.1) 5.5 (4.9; 6.2) <0.01 Triglycerides, mmol/L 1.30 (0.92; 1.89) 1.15 (0.84; 1.61) <0.01 ALT, µkat/L 0.47 (0.35; 0.65) 0.30 (0.24; 0.42) <0.01 History of thyroid disease, % 6.0 29.3 <0.01 Characteristic Men (n = 419) Women (n = 533) P Value a Age, y 49 (39; 61) 51 (40; 60) 0.71 Smoking, % Never 32.0 50.5 <0.01 Former 44.6 28.0 Current 23.4 21.5 Physical activity, % >1 h/wk 27.7 25.5 0.45 <1 h/wk 72.3 74.5 Alcohol consumption, g/d 8.7 (3.1; 18.4) 1.4 (0.7; 5.5) <0.01 BMI, kg/m2 27.7 (25.0; 30.2) 26.1 (23.1; 29.5) <0.01 WC, cm 94 (86; 102) 81 (74; 90) <0.01 TSH, mU/L 1.10 (0.78; 1.51) 1.21 (0.83; 1.76) 0.02 FT4, pmol/L 13.1 (12.2; 14.2) 13.4 (12.5; 14.6) <0.01 FT3, pmol/L 5.0 (4.6; 5.3) 4.5 (4.2; 4.9) <0.01 eGFR, mL/min/1.73 m2 117 (108; 125) 112 (103; 121) <0.01 LDL cholesterol, mmol/L 3.41 (2.76; 4.00) 3.32 (2.74; 3.98) 0.43 HDL cholesterol, mmol/L 1.27 (1.10; 1.48) 1.58 (1.35; 1.83) <0.01 Total cholesterol, mmol/L 5.3 (4.6; 6.1) 5.5 (4.9; 6.2) <0.01 Triglycerides, mmol/L 1.30 (0.92; 1.89) 1.15 (0.84; 1.61) <0.01 ALT, µkat/L 0.47 (0.35; 0.65) 0.30 (0.24; 0.42) <0.01 History of thyroid disease, % 6.0 29.3 <0.01 Data presented as median (25th; 75th percentile). Abbreviations: ALT, alanine aminotransferase; BMI, body mass index; eGFR, estimated glomerular filtration rate; WC, waist circumference. a Mann-Whitney U test for continuous data and χ2 test for categorical data were used for comparisons. View Large Association between hormone measurements and plasma metabolome The FT4 levels showed substantial associations with 106 of 613 plasma metabolites (Fig. 1), mainly lipids (n = 84) and, to a far lesser extent, amino acids (n = 6) and other compounds. The lipid profile comprised nonesterified fatty acids (NEFAs), including monounsaturated (e.g., oleate) and polyunsaturated (e.g., docosahexaenoate and linoleate) species of medium- to long-chain length. All of them were positively associated. Further positive associations with FT4 included numerous acylcarnitine species, 3-hydroybutyrate, 1-monopalmitin, and 1-monoolein. Consistent inverse associations with FT4 were noted for complex lipids [i.e., phosphatidylcholines and sphingomyelins with diverse fatty acid (FA) residues], lysophosphatidylcholines (LPCs) with FA residues of medium- to long-chain lengths, and lysolipids containing ethanolamine as the head group. In contrast, FT4 showed positive associations with lysolipids with polyunsaturated FA residues. Further positive associations with FT4 included two androstenedione derivatives, S-methylcysteine and (E,E)-bilirubin, and the complement C3 fragment HWESALLR. Not least, FT4 and uridine were inversely associated. Associations with FT3 were less apparent in plasma (n = 55; Supplemental Fig. 1) compared with FT4 and mirrored in part the positive associations with lipid species (e.g., acylcarnitines and FAs). Additionally, FT3 showed positive associations with arginine, tryptophan, and tyrosine in plasma, which were not seen with respect to FT4. Figure 1. View largeDownload slide Standardized β-estimates from linear regression analysis with FT4, with exposure and plasma metabolites as outcomes, including the whole population (squares), only men (circles), or only women (diamonds). Displayed are only metabolites that were statistically significant (controlling for the FDR at 5%) in at least one of the subsets (indicated by colors). Metabolites printed in bold showed a nominal statistically significant (P < 0.05) interaction term between FT4 and sex. Figure 1. View largeDownload slide Standardized β-estimates from linear regression analysis with FT4, with exposure and plasma metabolites as outcomes, including the whole population (squares), only men (circles), or only women (diamonds). Displayed are only metabolites that were statistically significant (controlling for the FDR at 5%) in at least one of the subsets (indicated by colors). Metabolites printed in bold showed a nominal statistically significant (P < 0.05) interaction term between FT4 and sex. Unique positive associations with FT4 among women became apparent with respect to medium-chain NEFAs, long-chain polyunsaturated fatty acids, and lysophospholipids carrying such FA residues (Fig. 1). Furthermore, the inverse association between FT4 and kynurenine, citrulline, histidine, and dihydroxyphenylalanine and the positive associations with 3-hydroyisobutyrate and L-urobilin were unique to women. The inverse associations between serum FT4 and the phosphatidylcholines (PCs) PC aa C30:0, PC aa C32:1, PC aa C34:1, and PC aa C34:4 were stronger in men than in women. Based on a data-dependent metabolic network construction (Supplemental Methods), we identified a subset of unknown compounds strongly related to bile acids (Fig. 2), with positive associations between FT4 and, for example, glycolithocholate sulfate. These associations were more apparent among women (Fig. 2). After correcting for multiple testing, no associations with TSH were apparent. Figure 2. View largeDownload slide Subnetwork of the derived Gaussian graphical model with emphasize on bile acids and related compounds. On each node, the results from linear regression analysis for visceral fat were mapped for the whole population (dark gray), only women (light gray), or only men (gray) as a portion of the association strength given as –log10(FDR − value). Statistically results in at least one population, with an FDR <5%, are highlighted by colors. Node sizes were chosen to represent the maximum association strength. Edges represent statistically significant partial correlations (par. cor.) between metabolites. Type and color represent metabolite and fluid dependencies. diff., different; Met., metabolite; non-sign., not statistically significant; P, plasma metabolites; sign., statistically significant; U, urine metabolites. Figure 2. View largeDownload slide Subnetwork of the derived Gaussian graphical model with emphasize on bile acids and related compounds. On each node, the results from linear regression analysis for visceral fat were mapped for the whole population (dark gray), only women (light gray), or only men (gray) as a portion of the association strength given as –log10(FDR − value). Statistically results in at least one population, with an FDR <5%, are highlighted by colors. Node sizes were chosen to represent the maximum association strength. Edges represent statistically significant partial correlations (par. cor.) between metabolites. Type and color represent metabolite and fluid dependencies. diff., different; Met., metabolite; non-sign., not statistically significant; P, plasma metabolites; sign., statistically significant; U, urine metabolites. Association between hormone measurements and urine metabolome In general, substantial associations between FT4 and urinary metabolites (n = 12) were sparse (Supplemental Fig. 2). For example, FT4 was positively associated with 3-hydroxysebacate and inversely with isovalerylglycine. Similar findings held true for FT3 (n = 13 urine-associated metabolites; Supplemental Fig. 1). In brief, serum FT3 was inversely associated with several glycine conjugates. No substantial associations with TSH were seen. Lipoprotein particles FT4 levels were significantly associated with more than one half of the derived lipoprotein measures (Fig. 3). Those included classic parameters (e.g., total cholesterol or LDL cholesterol and HDL cholesterol) and more specific parameters (e.g., total ApoA1, ApoA2, and ApoB levels). FT4 was inversely associated with large and small dense LDL and with HDL particles and related measures. In part, these associations were mirrored by TSH but with the association direction inverted (all positively associated with TSH). Serum FT3 was associated positively with large VLDL particles and small dense LDL and HDL particles (Supplemental Fig. 3). Figure 3. View largeDownload slide Standardized β-estimates from linear regression analysis with either (left) FT4 or (right) TSH, with exposure and lipoprotein particles as outcomes, including the whole population (squares), only men (circles), only or women (diamonds). Displayed are only the lipoprotein subfractions that were statistically significant (controlling for the FDR at 5%) in at least one of the subsets (indicated by colors). Lipoproteins printed in bold showed a nominal statistically significant (P < 0.05) interaction term between either FT4 or TSH and sex. IDL, intermediate-density lipoprotein. Figure 3. View largeDownload slide Standardized β-estimates from linear regression analysis with either (left) FT4 or (right) TSH, with exposure and lipoprotein particles as outcomes, including the whole population (squares), only men (circles), only or women (diamonds). Displayed are only the lipoprotein subfractions that were statistically significant (controlling for the FDR at 5%) in at least one of the subsets (indicated by colors). Lipoproteins printed in bold showed a nominal statistically significant (P < 0.05) interaction term between either FT4 or TSH and sex. IDL, intermediate-density lipoprotein. A sex-specific effect became apparent with respect to medium-dense HDL particles (HDL3) and related measures, which were inversely associated with FT4 among men only (Fig. 3). Restriction to a euthyroid subsample The vast majority of associations apparent in the whole population remained substantial when restricted to a euthyroid subsample (78 of 106; Supplemental Fig. 4). However, several of these interactions between men and women were no longer statistically significant (Supplemental Fig. 5). Persisting plasma metabolites with substantial FT4–sex interactions included citrulline, docosapentaenoate, 2-palmitoleoylglycerophosphocholine, 2-palmitoylglycerophosphocholine, PC aa C32:1, and the unknown compound X-11315. Sex-specific feature selection To test whether differences in phenotypic characteristics rather than sex per se accounted for the observed sex differences, we used a Boruta feature selection approach, which provides an inherent measure of variable importance. From a list of >50 phenotypic attributes (Supplemental Methods), including lifestyle, thyroid state, and anthropometric and laboratory parameters, 15 and 13 were identified as important factors for FT4 levels in women and men, respectively (Fig. 4). Strong discrepancies between the sexes were noted with respect to measures of the thyroid state (more relevant in women) and blood lipids and alcohol intake (more relevant in men). Figure 4. View largeDownload slide Final important distributions from Boruta feature selection of phenotypic traits associated with serum FT4 levels in (left) women and (right) men. Boxplots in dark gray indicate statistically significant (P < 0.01) attributes across 1000 permutations. Values of the shadow variables are printed in light gray for comparison purposes. eGFR, estimated glomerular filtration rate. Figure 4. View largeDownload slide Final important distributions from Boruta feature selection of phenotypic traits associated with serum FT4 levels in (left) women and (right) men. Boxplots in dark gray indicate statistically significant (P < 0.01) attributes across 1000 permutations. Values of the shadow variables are printed in light gray for comparison purposes. eGFR, estimated glomerular filtration rate. Discussion The present study used comprehensive metabolic profiling in a large cohort from the general population to characterize variation in the two most important surrogates of thyroid function, namely TSH and FT4. The results with respect to lipid metabolism confirmed previous observations, and the combination with comprehensive lipoprotein profiling allowed us to provide a reasonable mechanism for alterations in phospholipids. Furthermore, sex-specific findings were most likely attributable to an adverse thyroid state in women and higher alcohol intake in men. In general, the present study has provided further evidence for the numerous metabolic changes associated with serum FT4 and the missing capacity of TSH to capture those. This is of particular interest for clinical practice, because TSH might not be suitable for monitoring an adequate TH supply to the periphery. Signature of increased usage of NEFAs The metabolic fingerprint of FT4 in plasma showed a strong enrichment of lipid species. In particular, positive associations with NEFAs were observed, suggesting increased lipolysis from white adipose tissue with increasing FT4 levels. A T3-dependent increase in catecholamine sensitivity due to enhanced expression of β2-adrenergic receptors on adipocytes is a described mechanism (18, 19). Even the mitochondrial oxidation of NEFAs was shown to be accelerated by T3, mainly owing to increased conjugation of NEFAs with carnitine to facilitate their entrance to mitochondria (20). The resulting intracellular accumulation of acylcarnitine species likely induces their spill into the blood stream and hence explains the pronounced acylcarnitine signature seen in the present and previous studies (5–7). Ubiquitous decline in lipoprotein particles FT4 exhibited pronounced inverse associations with different characteristics of LDL and HDL particles. Previous work (21) has already described a reduction in large LDL and HDL particles in hypothyroid women after levothyroxine treatment, an observation confirmed in the Framingham Heart Study (21). With respect to LDL particles, THs accelerate their removal from blood through transcriptional upregulation of hepatic LDL receptor expression (22, 23). In addition, THs regulate the expression of key genes involved in lipoprotein metabolism (24), including those encoding for cholesteryl-ester transport protein, hepatic lipase, and lipoprotein lipase (25, 26). Increasing cholesteryl-ester transport protein activity would enhance the transfer of core lipids, triglycerides, and cholesteryl esters from HDL2 toward VLDL and intermediate-density lipoprotein and remnants, forcing depletion of the triglyceride content of HDL particles, which aligns with our observations. Through hydrolysis of atherogenic HDL2 and their subsequent conversion to HDL3, even a TH-dependent increase in hepatic lipase activity could additionally explain our observations (26). In addition to the modification of HDL particles within the circulation, THs also increase their hepatic uptake by inducing the multiple-ligand binding scavenger receptor class B member 1, which is the primary receptor for HDL. The TH-mediated decrease in lipoprotein particles likely accounts for the pronounced inverse associations between several phospholipid species (PCs) and FT4. This assumption is supported by the observation of strong positive correlations between lipoprotein particles and PCs (Supplemental Fig. 6). Furthermore, the single-layer surrounding a lipoprotein is composed of different PC species. Even if the PC content of lipoproteins differs during their turnover, the ubiquitous effect of THs in the present study included lipoproteins of different density and function and hence different PC compositions. FT4 and bile acid metabolism Excess hepatic cholesterol stimulates bile acid production. The limiting step in this process, namely the addition of a hydroxyl group, is facilitated by cholesterol 7-hydroxylase and is moderately stimulated by THs (27, 28). In line with this, FT4 was positively associated with the downstream bile acid derivative glycocholate sulfate and among women with glycolithocholate sulfate. Even an opposing effect direction might be reasonable because bile acids are known to stimulate hepatic deiodinase 2 activity and hence conversion of T4 to T3 (29). However, no relation between FT3 and these bile acids became apparent. Greater general clinical relevance might arise because both named bile acid species have been suggested to predict the onset of atrial fibrillation (30), and subclinical hyperthyroidism constitutes a well-established risk factor for atrial fibrillation (31), remembering that the association proposed between bile acids and atrial fibrillation might have resulted from not controlling for thyroid function in the study (30) rather than an effect of bile acids per se. FA residue-dependent associations with LPCs We observed an inverse association between FT4 and various LPCs and lysophosphatidylethanolamines. LPCs are generated, among others, through hydrolysis of membrane phospholipids by phospholipase A2 activity (32). A decrease of hepatic phospholipase A2 messenger RNA was observed in T3-treated rats (33). Decreased phospholipase A2 activity would result in a strong decline of LPC generation and might explain our observation. Further confirming observations were noted in previous rodent studies of thyroid disease (10, 11). However, LPC species with polyunsaturated FA residues showed positive associations with FT4, and we made similar observations previously (5). We supposed that adaptions in mitochondrial membrane composition, replacing polyunsaturated with saturated FAs, to avoid lipid peroxidation due to an increased presence of reactive oxygen species from enhanced respiration might be a reasonable explanation. In addition, LPCs are known to decrease under inflammatory conditions (e.g., inversely associated with plasma fibrinogen) (34). Fibrinogen levels, in turn, were a common predictor of FT4 levels in both sexes in the present study. However, the link between an inflammatory state and thyroid function has been rather elusive, apart from autoimmunity, and further studies are needed to explore involved mechanisms more in detail. Observed sex differences resulted in part from an adverse thyroid state in women We observed different association patterns between women and men for a number of metabolites in relation to FT4, with a particular enrichment in women. Restriction to a euthyroid subsample led to the disappearance of several former statistically significant interaction terms, indicating an influence of the adverse thyroid state among women (∼30% with previous thyroid disease). Nonetheless, the positive association between docosapentaenoate and FT4 in women persisted. In previous work (35), we observed a positive association between dehydroepiandrosterone sulfate and docosapentaenoate, which was also unique to women. In addition to its production in the adrenal glands, dehydroepiandrosterone sulfate is released from the ovaries and might therefore suggest a role for estrogen in this context. Even the hormonal shift during menopause might have contributed to our exclusive finding in women, because postmenopausal women exhibited higher FT4 levels in our population, and a previous metabolomics study (36) revealed an increase in polyunsaturated FA species during menopause similar to those observed in the present study. Higher alcohol intake mediates effects seen only in men FT4 levels were inversely associated with medium dense HDL particles (HDL3) among men only, an effect not previously described. Moreover, HDL3 particles were strongly and positively correlated (Supplemental Fig. 6) with those diacyl PCs evenly and uniquely associated with FT4 in men. A recent Mendelian randomization approach (37) provided convincing evidence that HDL3 particles are causally altered by alcohol intake. In line with that, these diacyl PCs were associated with alcohol intake in men only in a population-based study (38). It is known that chronic high alcohol consumption impairs TH action at multiple levels (e.g., diminishes thyroid volume or impairs hepatic function, resulting in less peripheral conversion of THs) (39, 40). Together, these processes result in less bioactive THs and might therefore diminish further the downstream metabolic effects of THs or the observed associations are purely mediated by alcohol consumption, decreasing FT4 levels and increasing HDL3 particles. Inconsistent metabolic profile of FT3 In general, T4 is considered a prohormone and the metabolic effects of TH are mediated in most by T3. One might consider FT3 measurements as a more suitable surrogate marker of metabolic TH action. At the least, in our study, serum FT3 was associated with fewer metabolites compared with FT4, merely mirroring the lipolytic effect discussed. However, distinct associations did become apparent; in particular, the positive associations with small dense LDL particles contrasted with the associations seen with FT4. A possible mechanism to understand this discrepancy might arise from the origin of FT3. Most circulating T3 originates from peripheral conversion of T4, with only minor amounts released by the thyroid gland (1). Hence, serum FT3 relies strongly on tissue-dependent deiodinase activity, in particular, in the liver. Hepatic steatosis has been shown to increase FT3 even in euthyroid subjects (41) and leads to diminished LDL uptake (42). A prolonged dwelling time of LDL particles in the circulation renders them susceptible to hydrolysis, generating small dense LDL particles. In summary, the inconsistent metabolic profile of FT3 is at least partially explainable by the higher dependency of serum FT3 on nonthyroidal circumstances compared with FT4. Study strengths and limitations The strengths of our study were our large sample size of a well-characterized cohort, the availability of detailed information on potential confounders, and the comprehensive metabolic profiling using MS and NMR technologies. The study limitations arose from the cross-sectional design, which does not permit drawing causal conclusions. The replication of our findings in independent cohorts is a further important issue. As our study population did not comprise those with untreated thyroid disorders, it would be of particular interest if the molecular signatures we have observed are evenly present in such patients and, moreover, might be able to guide adequate levothyroxine treatment, given that about ~5% to 10% of patients report impaired psychological well-being, depression, or anxiety despite normal TSH levels with standard levothyroxine substitution (43). Finally, further sex-specific effects of thyroid function (e.g., on the blood lipid profile) might not have been detected within the present study owing to missing enrichment of subjects with subclinical thyroid disorders, in particular, hypothyroidism (44). Conclusions The present study used state-of-the-art metabolomics techniques to address the metabolic implications of TSH and FT4 among a sample from the general population. Manifold associations with respect to FT4 and lipid metabolism were partially expected; however, the present work is an important generalization of previous experimental studies. Moreover, the combination of in-depth phenotyping using classic clinical tools and state-of-the-art metabolic profiling allowed us to clarify the sexual dimorphic effects observed and underscore the need for comprehensive data analyses using advanced tools. In general, our results argue for the consideration of FT4 as a relevant marker to monitor an adequate supply of THs to the peripheral tissues in the diagnosis and monitoring of thyroid disorders, because TSH measurements will not reflect these effects. Abbreviations: Abbreviations: Apo apolipoprotein FA fatty acid FDR false-discovery rate FT3 free triiodothyronine FT4 free thyroxine HDL high-density lipoprotein LDL low-density lipoprotein LPC lysophosphatidylcholine MS mass spectrometry NEFA nonesterified fatty acid NMR nuclear magnetic resonance T3 triiodothyronine T4 thyroxine TH thyroid hormone TSH thyrotropin VLDL very-low-density lipoprotein Acknowledgments We thank Bianca Schmick from the Genome Analysis Center for expert technical assistance. Financial Support: This work was funded by the German Federal Ministry of Education and Research (Grants 01ZZ0403, 01ZZ0103, 01GI0883, and AtheroSysMed 03IS2061B), the Ministry for Education, Research and Cultural Affairs, and the Ministry of Social Affairs of the Federal State of Mecklenburg-West Pomerania. The present study was also part of the research project Greifswald Approach to Individualized Medicine (GANI_MED). The GANI_MED consortium is funded by the Federal Ministry of Education and Research and the Ministry of Cultural Affairs of the Federal State of Mecklenburg-West Pomerania (Grant 03IS2061A). A part of our study was also supported by the German Center Diabetes Research (grant to J.A.). The analyses were supported by grants from the German Research Foundation as a part of the priority program “Thyroid Trans Act” (Grants DFG FR 3055/4-1 and VO 1444/9-1). Disclosure Summary: The authors have nothing to disclose. References 1. Yen PM . Physiological and molecular basis of thyroid hormone action . Physiol Rev . 2001 ; 81 ( 3 ): 1097 – 1142 . Google Scholar CrossRef Search ADS PubMed 2. Mullur R , Liu YY , Brent GA . Thyroid hormone regulation of metabolism . Physiol Rev . 2014 ; 94 ( 2 ): 355 – 382 . Google Scholar CrossRef Search ADS PubMed 3. Boelaert K , Franklyn JA . Thyroid hormone in health and disease . J Endocrinol . 2005 ; 187 ( 1 ): 1 – 15 . Google Scholar CrossRef Search ADS PubMed 4. Benhadi N , Fliers E , Visser TJ , Reitsma JB , Wiersinga WM . Pilot study on the assessment of the setpoint of the hypothalamus-pituitary-thyroid axis in healthy volunteers . Eur J Endocrinol . 2010 ; 162 ( 2 ): 323 – 329 . Google Scholar CrossRef Search ADS PubMed 5. Pietzner M , Engelmann B , Kacprowski T , Golchert J , Dirk AL , Hammer E , Iwen KA , Nauck M , Wallaschofski H , Führer D , Münte TF , Friedrich N , Völker U , Homuth G , Brabant G . Plasma proteome and metabolome characterization of an experimental human thyrotoxicosis model . BMC Med . 2017 ; 15 ( 1 ): 6 . Google Scholar CrossRef Search ADS PubMed 6. Jourdan C , Linseisen J , Meisinger C , Petersen AK , Gieger C , Rawal R , Illig T , Heier M , Peters A , Wallaschofski H , Nauck M , Kastenmüller G , Suhre K , Prehn C , Adamski J , Koenig W , Roden M , Wichmann HE , Völzke H . Associations between thyroid hormones and serum metabolite profiles in an euthyroid population . Metabolomics . 2014 ; 10 ( 1 ): 152 – 164 . Google Scholar CrossRef Search ADS PubMed 7. Chng CL , Lim AY , Tan HC , Kovalik JP , Tham KW , Bee YM , Lim W , Acharyya S , Lai OF , Chong MF , Yen PM . Physiological and metabolic changes during the transition from hyperthyroidism to euthyroidism in Graves’ disease . Thyroid . 2016 ; 26 ( 10 ): 1422 – 1430 . Google Scholar CrossRef Search ADS PubMed 8. Friedrich N , Pietzner M , Cannet C , Thuesen BH , Hansen T , Wallaschofski H , Grarup N , Skaaby T , Budde K , Pedersen O , Nauck M , Linneberg A . Urinary metabolomics reveals glycemic and coffee associated signatures of thyroid function in two population-based cohorts . PLoS One . 2017 ; 12 ( 3 ): e0173078 . Google Scholar CrossRef Search ADS PubMed 9. Pietzner M , Budde K , Homuth G , Völker U , Völzke H , Nauck M , Wallaschofski H , Friedrich N . Distinct urinary metabolic profiles associated with serum TSH and FT4 concentrations . Metabolomics . 2015 ; 11 ( 5 ): 1316 – 1326 . Google Scholar CrossRef Search ADS 10. Montoya GA , Strauss V , Fabian E , Kamp H , Mellert W , Walk T , Looser R , Herold M , Krennrich G , Peter E , van Ravenzwaay B . Mechanistic analysis of metabolomics patterns in rat plasma during administration of direct thyroid hormone synthesis inhibitors or compounds increasing thyroid hormone clearance . Toxicol Lett . 2014 ; 225 ( 2 ): 240 – 251 . Google Scholar CrossRef Search ADS PubMed 11. Wu S , Tan G , Dong X , Zhu Z , Li W , Lou Z , Chai Y . Metabolic profiling provides a system understanding of hypothyroidism in rats and its application . PLoS One . 2013 ; 8 ( 2 ): e55599 . Google Scholar CrossRef Search ADS PubMed 12. Völzke H , Alte D , Schmidt CO , Radke D , Lorbeer R , Friedrich N , Aumann N , Lau K , Piontek M , Born G , Havemann C , Ittermann T , Schipf S , Haring R , Baumeister SE , Wallaschofski H , Nauck M , Frick S , Arnold A , Jünger M , Mayerle J , Kraft M , Lerch MM , Dörr M , Reffelmann T , Empen K , Felix SB , Obst A , Koch B , Gläser S , Ewert R , Fietze I , Penzel T , Dören M , Rathmann W , Haerting J , Hannemann M , Röpcke J , Schminke U , Jürgens C , Tost F , Rettig R , Kors JA , Ungerer S , Hegenscheid K , Kühn JP , Kühn J , Hosten N , Puls R , Henke J , Gloger O , Teumer A , Homuth G , Völker U , Schwahn C , Holtfreter B , Polzer I , Kohlmann T , Grabe HJ , Rosskopf D , Kroemer HK , Kocher T , Biffar R , John U , Hoffmann W . Cohort profile: the study of health in Pomerania . Int J Epidemiol . 2011 ; 40 ( 2 ): 294 – 307 . Google Scholar CrossRef Search ADS PubMed 13. Inker LA , Schmid CH , Tighiouart H , Eckfeldt JH , Feldman HI , Greene T , Kusek JW , Manzi J , Van Lente F , Zhang YL , Coresh J , Levey AS ; CKD-EPI Investigators . Estimating glomerular filtration rate from serum creatinine and cystatin C . N Engl J Med . 2012 ; 367 ( 1 ): 20 – 29 . Google Scholar CrossRef Search ADS PubMed 14. Knacke H , Pietzner M , Do KT , Römisch-Margl W , Kastenmüller G , Völker U , Völzke H , Krumsiek J , Artati A , Wallaschofski H , Nauck M , Suhre K , Adamski J , Friedrich N . Metabolic fingerprints of circulating IGF-1 and the IGF-1/IGFBP-3 ratio: a multifluid metabolomics study . J Clin Endocrinol Metab . 2016 ; 101 ( 12 ): 4730 – 4742 . Google Scholar CrossRef Search ADS PubMed 15. Pietzner M , Homuth G , Budde K , Lehmphul I , Völker U , Völzke H , Nauck M , Köhrle J , Friedrich N . Urine metabolomics by (1)H-NMR spectroscopy indicates associations between serum 3,5-T2 concentrations and intermediary metabolism in euthyroid humans . Eur Thyroid J . 2015 ; 4 ( Suppl 1 ): 92 – 100 . Google Scholar CrossRef Search ADS PubMed 16. Ittermann T , Khattak RM , Nauck M , Cordova CM , Völzke H . Shift of the TSH reference range with improved iodine supply in Northeast Germany . Eur J Endocrinol . 2015 ; 172 ( 3 ): 261 – 267 . Google Scholar CrossRef Search ADS PubMed 17. Kursa MB , Rudnicki WR . Feature selection with the Boruta package . J Stat Softw . 2010 ; 36 ( 11 ): 1 – 13 . Google Scholar CrossRef Search ADS 18. Nedvidkova J , Haluzik M , Bartak V , Dostalova I , Vlcek P , Racek P , Taus M , Behanova M , Svacina S , Alesci S , Pacak K . Changes of noradrenergic activity and lipolysis in the subcutaneous abdominal adipose tissue of hypo- and hyperthyroid patients: an in vivo microdialysis study . Ann N Y Acad Sci . 2004 ; 1018 ( 1 ): 541 – 549 . Google Scholar CrossRef Search ADS PubMed 19. Viguerie N , Langin D . Effect of thyroid hormone on gene expression . Curr Opin Clin Nutr Metab Care . 2003 ; 6 ( 4 ): 377 – 381 . Google Scholar PubMed 20. Jackson-Hayes L , Song S , Lavrentyev EN , Jansen MS , Hillgartner FB , Tian L , Wood PA , Cook GA , Park EA . A thyroid hormone response unit formed between the promoter and first intron of the carnitine palmitoyltransferase-Ialpha gene mediates the liver-specific induction by thyroid hormone . J Biol Chem . 2003 ; 278 ( 10 ): 7964 – 7972 . Google Scholar CrossRef Search ADS PubMed 21. Pearce EN , Wilson PW , Yang Q , Vasan RS , Braverman LE . Thyroid function and lipid subparticle sizes in patients with short-term hypothyroidism and a population-based cohort . J Clin Endocrinol Metab . 2008 ; 93 ( 3 ): 888 – 894 . Google Scholar CrossRef Search ADS PubMed 22. Ness GC , Pendleton LC , Li YC , Chiang JY . Effect of thyroid hormone on hepatic cholesterol 7 alpha hydroxylase, LDL receptor, HMG-CoA reductase, farnesyl pyrophosphate synthetase and apolipoprotein A-I mRNA levels in hypophysectomized rats . Biochem Biophys Res Commun . 1990 ; 172 ( 3 ): 1150 – 1156 . Google Scholar CrossRef Search ADS PubMed 23. Lopez D , Abisambra Socarrás JF , Bedi M , Ness GC . Activation of the hepatic LDL receptor promoter by thyroid hormone . Biochim Biophys Acta . 2007 ; 1771 ( 9 ): 1216 – 1225 . Google Scholar CrossRef Search ADS PubMed 24. Angelin B , Rudling M . Lipid lowering with thyroid hormone and thyromimetics . Curr Opin Lipidol . 2010 ; 21 ( 6 ): 499 – 506 . Google Scholar CrossRef Search ADS PubMed 25. Lam KS , Chan MK , Yeung RT . High-density lipoprotein cholesterol, hepatic lipase and lipoprotein lipase activities in thyroid dysfunction—effects of treatment . Q J Med . 1986 ; 59 ( 229 ): 513 – 521 . Google Scholar PubMed 26. Tan KC , Shiu SW , Kung AW . Effect of thyroid dysfunction on high-density lipoprotein subfraction metabolism: roles of hepatic lipase and cholesteryl ester transfer protein . J Clin Endocrinol Metab . 1998 ; 83 ( 8 ): 2921 – 2924 . Google Scholar PubMed 27. Hashimoto K , Mori M . Crosstalk of thyroid hormone receptor and liver X receptor in lipid metabolism and beyond [review] . Endocr J . 2011 ; 58 ( 11 ): 921 – 930 . Google Scholar CrossRef Search ADS PubMed 28. Bonde Y , Breuer O , Lütjohann D , Sjöberg S , Angelin B , Rudling M . Thyroid hormone reduces PCSK9 and stimulates bile acid synthesis in humans . J Lipid Res . 2014 ; 55 ( 11 ): 2408 – 2415 . Google Scholar CrossRef Search ADS PubMed 29. Watanabe M , Houten SM , Mataki C , Christoffolete MA , Kim BW , Sato H , Messaddeq N , Harney JW , Ezaki O , Kodama T , Schoonjans K , Bianco AC , Auwerx J . Bile acids induce energy expenditure by promoting intracellular thyroid hormone activation . Nature . 2006 ; 439 ( 7075 ): 484 – 489 . Google Scholar CrossRef Search ADS PubMed 30. Alonso A , Yu B , Qureshi WT , Grams ME , Selvin E , Soliman EZ , Loehr LR , Chen LY , Agarwal SK , Alexander D , Boerwinkle E . Metabolomics and incidence of atrial fibrillation in African Americans: the Atherosclerosis Risk in Communities (ARIC) Study . PLoS One . 2015 ; 10 ( 11 ): e0142610 . Google Scholar CrossRef Search ADS PubMed 31. Baumgartner C , da Costa BR , Collet TH , Feller M , Floriani C , Bauer DC , Cappola AR , Heckbert SR , Ceresini G , Gussekloo J , den Elzen WPJ , Peeters RP , Luben R , Völzke H , Dörr M , Walsh JP , Bremner A , Iacoviello M , Macfarlane P , Heeringa J , Stott DJ , Westendorp RGJ , Khaw KT , Magnani JW , Aujesky D , Rodondi N ; Thyroid Studies Collaboration . Thyroid function within the normal range, subclinical hypothyroidism, and the risk of atrial fibrillation . Circulation . 2017 ; 136 ( 22 ): 2100 – 2116 . Google Scholar CrossRef Search ADS PubMed 32. Dennis EA . The growing phospholipase A2 superfamily of signal transduction enzymes . Trends Biochem Sci . 1997 ; 22 ( 1 ): 1 – 2 . Google Scholar CrossRef Search ADS PubMed 33. Boone LR , Lagor WR , Moya ML , Niesen MI , Rothblat GH , Ness GC . Thyroid hormone enhances the ability of serum to accept cellular cholesterol via the ABCA1 transporter . Atherosclerosis . 2011 ; 218 ( 1 ): 77 – 82 . Google Scholar CrossRef Search ADS PubMed 34. Pietzner M , Kaul A , Henning AK , Kastenmüller G , Artati A , Lerch MM , Adamski J , Nauck M , Friedrich N . Comprehensive metabolic profiling of chronic low-grade inflammation among generally healthy individuals . BMC Med . 2017 ; 15 ( 1 ): 210 . Google Scholar CrossRef Search ADS PubMed 35. Piontek U , Wallaschofski H , Kastenmüller G , Suhre K , Völzke H , Do KT , Artati A , Nauck M , Adamski J , Friedrich N , Pietzner M . Sex-specific metabolic profiles of androgens and its main binding protein SHBG in a middle aged population without diabetes . Sci Rep . 2017 ; 7 ( 1 ): 2235 . Google Scholar CrossRef Search ADS PubMed 36. Ke C , Hou Y , Zhang H , Yang K , Wang J , Guo B , Zhang F , Li H , Zhou X , Li Y , Li K . Plasma metabolic profiles in women are menopause dependent . PLoS One . 2015 ; 10 ( 11 ): e0141743 . Google Scholar CrossRef Search ADS PubMed 37. Tabara Y , Arai H , Hirao Y , Takahashi Y , Setoh K , Kawaguchi T , Kosugi S , Ito Y , Nakayama T , Matsuda F ; Nagahama Study Group . The causal effects of alcohol on lipoprotein subfraction and triglyceride levels using a Mendelian randomization analysis: the Nagahama study . Atherosclerosis . 2017 ; 257 : 22 – 28 . Google Scholar CrossRef Search ADS PubMed 38. Jaremek M , Yu Z , Mangino M , Mittelstrass K , Prehn C , Singmann P , Xu T , Dahmen N , Weinberger KM , Suhre K , Peters A , Döring A , Hauner H , Adamski J , Illig T , Spector TD , Wang-Sattler R . Alcohol-induced metabolomic differences in humans . Transl Psychiatry . 2013 ; 3 ( 7 ): e276 . Google Scholar CrossRef Search ADS PubMed 39. Hermann D , Heinz A , Mann K . Dysregulation of the hypothalamic-pituitary-thyroid axis in alcoholism . Addiction . 2002 ; 97 ( 11 ): 1369 – 1381 . Google Scholar CrossRef Search ADS PubMed 40. Rachdaoui N , Sarkar DK . Pathophysiology of the effects of alcohol abuse on the endocrine system . Alcohol Res . 2017 ; 38 ( 2 ): 255 – 276 . Google Scholar PubMed 41. van den Berg EH , van Tienhoven-Wind LJ , Amini M , Schreuder TC , Faber KN , Blokzijl H , Dullaart RP . Higher free triiodothyronine is associated with non-alcoholic fatty liver disease in euthyroid subjects: the Lifelines Cohort Study . Metabolism . 2017 ; 67 : 62 – 71 . Google Scholar CrossRef Search ADS PubMed 42. Min HK , Kapoor A , Fuchs M , Mirshahi F , Zhou H , Maher J , Kellum J , Warnick R , Contos MJ , Sanyal AJ . Increased hepatic synthesis and dysregulation of cholesterol metabolism is associated with the severity of nonalcoholic fatty liver disease . Cell Metab . 2012 ; 15 ( 5 ): 665 – 674 . Google Scholar CrossRef Search ADS PubMed 43. Wiersinga WM . Paradigm shifts in thyroid hormone replacement therapies for hypothyroidism . Nat Rev Endocrinol . 2014 ; 10 ( 3 ): 164 – 174 . Google Scholar CrossRef Search ADS PubMed 44. Tognini S , Polini A , Pasqualetti G , Ursino S , Caraccio N , Ferdeghini M , Monzani F . Age and gender substantially influence the relationship between thyroid status and the lipoprotein profile: results from a large cross-sectional study . Thyroid . 2012 ; 22 ( 11 ): 1096 – 1103 . Google Scholar CrossRef Search ADS PubMed Copyright © 2018 Endocrine Society
Journal of Clinical Endocrinology and Metabolism – Oxford University Press
Published: Mar 12, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera