sign in sign up
<<
Home , Hyperthyroidism, Hyperthyroxinemia, Hypothyroidism

Diabetes Care 1

Impaired Sensitivity to Martin Laclaustra,1,2,3 Belen Moreno-Franco,1,2 Is Associated With Jose Manuel Lou-Bonafonte,1,4,5 Rocio Mateo-Gallego,1,2 and Jose Antonio Casasnovas,1,2 Pilar Guallar-Castillon,6,7,8 Ana Cenarro,1,2 1,2 https://doi.org/10.2337/dc18-1410 and Fernando Civeira

OBJECTIVE Diabetes prevalence and incidence increase among individuals with but also among those with , which seems contradictory. Both high free thyroxine (fT4) and high thyroid-stimulating (TSH) are present in the resistance to thyroid hormone syndrome. A mild acquired resistance to thyroid hormone might occur in the general population and be associated with diabetes. We aimed to analyze the association of resistance to thyroid hormone indices (the Thyroid Feedback Quantile-based Index [TFQI], proposed in this work, and the previously used Thyrotroph T4 Resistance Index and TSH Index) with diabetes.

RESEARCH DESIGN AND METHODS 1Instituto de Investigacion´ Sanitaria de Aragon´ We calculated the aforementioned resistance to thyroid hormone indices based on (IIS Aragon),´ Universidad de Zaragoza, Zaragoza, ‡ Spain a U.S. representative sample of 5,129 individuals 20 years of age participating in 2Translational Research Unit, Hospital Univer- the 2007–2008 National Health and Nutrition Examination Survey (NHANES). Also, sitario Miguel Servet, Zaragoza, Spain and to approximate TFQI, a U.S.-referenced Parametric TFQI (PTFQI) can be calculated CIBERCV, Madrid, Spain 3 RISK METABOLIC AND CARDIOVASCULAR with the spreadsheet formula 5NORM.DIST(fT4_cell_in_pmol_per_L,10.075, Fundacion´ Agencia Aragonesa para la Inves- 1 2 tigacion´ y el Desarrollo (ARAID), Zaragoza, Spain 2.155,TRUE) NORM.DIST(LN(TSH_cell_in_mIU_per_L),0.4654,0.7744,TRUE) 1. 4Instituto Agroalimentario de Aragon,´ CITA- Outcomes of interest were glycohemoglobin ‡6.5%, diabetes , Universidad de Zaragoza, Zaragoza, Spain diabetes-related deaths (diabetes as contributing cause of death), and addi- 5CIBEROBN, Madrid, Spain 6 tionally, in a fasting subsample, diabetes and metabolic syndrome. Logistic Department of Preventive Medicine and , School of Medicine, Universidad Autonoma´ and Poisson regressions were adjusted for sex, age, and race/ethnicity. de Madrid-IdiPaz, and CIBERESP, Madrid, Spain 7IMDEA Food Institute, CEI UAM+CSIC, Madrid, RESULTS Spain OddratiosforthefourthversusthefirstquartileofTFQI were1.73(95% CI 1.32,2.27) 8Welch Center for Prevention, Epidemiology, (P 5 0.002) for positive glycohemoglobin and 1.66 (95% CI 1.31, 2.10) (P 5 and Clinical Research, Johns Hopkins University, trend trend Baltimore, MD 0.001) for medication. Diabetes-related death rate ratio for TFQI being above versus P 5 Corresponding author: Martin Laclaustra, below the median was 4.81 (95% CI 1.01, 22.94) ( trend 0.015). Further adjustment [email protected] for BMI and restriction to normothyroid individuals yielded similar results. Per 1 SD Received 2 July 2018 and accepted 4 November in TFQI, odds increased 1.13 (95% CI 1.02, 1.25) for diabetes and 1.16 (95% CI 1.02, 2018 1.31) for metabolic syndrome. The other resistance to thyroid hormone indices This article contains Supplementary Data online showed similar associations for diabetes-related deaths and metabolic syndrome. at http://care.diabetesjournals.org/lookup/suppl/ doi:10.2337/dc18-1410/-/DC1. CONCLUSIONS © 2018 by the American Diabetes Association. Higher values in resistance to thyroid hormone indices are associated with , Readers may use this article as long as the work metabolic syndrome, diabetes, and diabetes-related mortality. Resistance to is properly cited, the use is educational and not for profit, and the work is not altered. More infor- thyroid hormone may reflect energy balance problems driving . mation is available at http://www.diabetesjournals These indices may facilitate monitoring treatments focused on energy balance. .org/content/license. Diabetes Care Publish Ahead of Print, published online December 14, 2018 2 Resistance to Thyroid Hormone and Diabetes Diabetes Care

Recent evidence suggests that hyper- because they can be quantified just approved by the National Center for thyroxinemia may be cross-sectionally observing and TSH Health Statistics (NCHS) Research Ethics and prospectively associated with type 2 concentrations or indices derived from Review Board (Continuation of Protocol diabetes (1). This is at odds with the them (13,14), although suppression tests no. 2005-06). physiological metabolism-activating have also been described (15). action of thyroid hormones, which Thus, given the thyroid feedback loop, Thyroxine, TSH, and Resistance to is considered capable of ameliorating the aforementioned reports showing Thyroid Hormone Indices obesity-related morbidities (2). It is association of high thyroid hormones Free thyroxine (fT4) was quantified with also in conflict with the increased in- or high TSH with the prevalence and the Access Free T4 assay, a two-step cidence of diabetes reported among incidence of metabolic syndrome and enzyme immunoassay, and TSH was quan- patients with hypothyroidism (3) and diabetes provide apparently contradict- tified with the Access HYPERsensitive with the increased prevalence of met- ing results. In this work, we hypothesized hTSH Assay, a two-site immunoenzymatic abolic syndrome described in the that these conflicting results might be (“sandwich”) assay. Measurements were upper-normal range of the pituitary conciliated if high thyroid hormones and performed at the University of Washing- thyroid-stimulating hormone (TSH) (also high TSH co-occurrence in diabetes re- ton (Seattle, WA). The laboratory normality known as thyrotropin) (4), a range as- flected a central resistance to thyroid reference ranges were 7.74–20.64 pmol/L sociated with lower thyroid activity. Pre- hormone. If that were the case, this for fT4 and 0.34–5.60 mIU/L for TSH. vious evidence is inconsistent (4): some central resistance would probably be Thyrotroph T4 Resistance Index (TT4RI) studies find diabetes or metabolic syn- the expression of a general reduced sen- was calculated as fT4 (pmol/L) z TSH drome to be associated with TSH but not sitivity to thyroid hormones, i.e., not only (mIU/L) (13). TSH index (TSHI) was cal- with thyroid hormones, whereas others central but also peripheral. This resistance culated as ln TSH (mIU/L) + 0.1345 z fT4 find an association only with thyroid hor- to thyroid hormone could be one addi- (pmol/L) (14). mones, occasionally linking opposite tional characteristic of the cardiome- ends of the thyroid hormone concen- tabolic traits predicting diabetes. We TFQI, a New Resistance to Thyroid tration range with diabetes. The effect analyzed continuous National Health Hormone Index of thyroid hormones on sensitiv- and Nutrition Examination Survey Given encouraging preliminary results ity differs by tissue; it enhances glu- (NHANES) (16) data to gain insight from the description of participants cose uptake in the muscle but reduces into this issue and propose the Thyroid who were simultaneously in the extreme it in the liver (5). The overall net effect Feedback Quantile-based Index (TFQI), quartiles of fT4 and TSH (see STATISTICAL in hypothyroidism favors insulin resis- a new resistance to thyroid hormone ANALYSES and RESULTS sections), we developed tance, but in it may index focused on deviations close to a new index to quantify, in a continuous also occasionally favor glucose intoler- normality. manner, deviations from the median pi- ance (5). These metabolic mechanisms tuitary response (inhibition) to thyroid described in clinical thyroid diseases hormones: the TFQI, which stands for may not be enough to explain the asso- RESEARCH DESIGN AND METHODS thyroid feedback quantile-based index. ciations found within the normothyroid Design and Participants Ranks (order position from minimum range. Thyroid hormones and TSH were mea- to maximum value) of fT4 and TSH were Physiologically, thyroid hormones and suredinserumin5,222(N1)adults converted to quantiles between 0 and 1, TSH are inversely correlated owing to a 20 years of age and older in the con- taking into account sampling weights. negative feedback loop. However, high tinuous NHANES (16) performed in Fractional quantiles are a generalization thyroid hormones coexist with high TSH 2007and2008,whichrecruitedarep- that summarizes in a single fraction the in individuals with resistance to thyroid resentative sample of the noninstitu- ordering number of each part in the hormone syndrome, an inherited auto- tionalized U.S. population. Data on sequence and the number of parts, somal recessive trait (6–8). Yet, a possibly mortality up to 2011 from the National e.g., the upper limit of the second tertile reversible acquired resistance to thyroid Death Index are available for all but six is the value of the quantile 0.66 and hormone, due to homeostatic compen- individuals. The current study estimates the upper limit of the third quartile is sation, has been hypothesized (9). Pro- cross-sectional associations between the value of the quantile 0.75. That longed fasting produces drops in TSH and resistance to thyroid hormone indices conversion is achieved by applying the increases in pituitary sensitivity to thyroid and diabetes-related conditions as well population empirical cumulative distri- hormones (10,11). In contrast, morbidly as longitudinal associations between bution function (cdf) to hormone con- obese individuals tend to have higher the indices and the diabetes-related centration. TFQI was calculated as cdf levels of both thyroid hormones and mortality rate. Analyses were performed fT4 2 (1 2 cdf TSH), i.e., the difference TSH (12). Resistance to thyroid hormone using the entire NHANES data set (N1)and between fT4 quantile and the reversed manifestations can be systematized into several subsets (main complete case sam- TSH quantile, because they are inversely central resistance phenomena, which ple, N2 = 5,129; normothyroid sample, correlated in the negative feedback loop. the feedback loop set point in the N3 = 4,750; and fasting normothyroid This index ranges between 21and1. central nervous system, and peripheral sample, N4 = 1,997) (see flowchart in Negative values indicate lower TSH resistance phenomena, which decrease Supplementary Data). NHANES partici- (higher inhibition by fT4) than that hormones’ metabolic effects. The for- pants provided written informed consent expected for the actual fT4 (which mer are easier to evaluate than the latter to participate in the survey, which was means higher sensitivity to fT4). Similarly, care.diabetesjournals.org Laclaustra and Associates 3

positive values indicate higher TSH (lower i.e., diagnosis was issued when meeting indices, this outcome was compared inhibition by fT4) than that expected for three of the following criteria: 1)waist between values above and below the the actual fT4 (lower sensitivity to fT4). circumference $102 cm in men or $88 median, and trend was tested with each Median TSH response to fT4 is represented cm in women; 2) triglycerides $150 mg/dL; index as a continuous variable. by the value 0. This index is more stable 3) HDL ,40 mg/dL in men In the main sample and the normo- than TT4RI and TSHI at abnormal ranges or ,50 mg/dL in women; 4) systolic or thyroid sample, interquartile compari- of the thyroid gland responsiveness to diastolic blood pressure $130 mmHg sons were performed for BMI, obesity, TSH. That is, TFQI does not reach extreme or $85 mmHg, respectively (or receiv- HbA1c $6.5%, diabetes medication use, values in individuals with thyroid gland– ing medication for ); and 5) and diabetes-related mortality. BMI was caused (primary) clinical hyper- or hypo- glucose $100 mg/dL (or receiving med- modeled with linear regression; obesity, thyroidism (see data points in the top-left ication for diabetes). HbA1c $6.5%, and diabetes medication and bottom-right areas of TSH versus were modeled with logistic regression, fT4 scatterplots in Supplementary Fig. 1). Mortality and death was modeled with Poisson In order to provide an index that can be Survey data were linked with the Na- regression. calculated for any new value or adapted tional Death Index by the NCHS. Besides In the fasting subsample, ORs for di- to other populations, an approximation the leading cause of death, records also abetes, metabolic syndrome, and meta- with the same range and interpretation, include contributing causes of death in a bolic syndrome criteria were calculated the Parametric Thyroid Feedback Quantile- 20 entity axis set of codes for multiple for1-SDincreaseofresistancetothyroid based Index (PTFQI) can be obtained causes of death. The public use linked hormone indices except for TT4RI. This from fT4 in pmol/L and TSH in mIU/L mortality data set provided by the NCHS index was analyzed using the logarithmic using the standard normal cumulative includes a flag variable for diabetes when scale, and its OR was calculated for a distribution function as follows: ICD-10 codes E10–E14 appear as one of TT4RI multiplicative increase by an SD log2 TT4RI F((fT4 2 mfT4)/sfT4) 2 (1 2 F((ln TSH 2 the 20 contributing or multiple causes of SD equivalent (3 2 ). Diabetes, mln TSH)/sln TSH)), where mfT4 = 10.075, death. We used this flag, i.e., diabetes- metabolic syndrome, and metabolic sfT4 = 2.155, mln TSH = 0.4654, and sln TSH = related death, as our end point. syndrome criteria were modeled with 0.7744 for the U.S. population. This logistic regression. can be easily achieved with the Excel Statistical Analyses As part of sensitivity analyses, we or LibreOffice spreadsheet formula = Mean and SD of thyroid parameters were further adjusted models for education NORM.DIST(fT4_cell,10.075,2.155,TRUE)+ calculated for population sex, age, and level, physical activity performed at work NORM.DIST(LN(TSH_cell),0.4654,0.7744, race strata. TSH and TT4RI were processed or during leisure time, and daily seden- TRUE)21. in the logarithmic scale given their skewed tary time. All analyses were performed distributions. We created scatterplots of with statistical computing software R Other Variables TSH versus fT4 and plotted population- (18), version 3.4.4, with the package Demographic (including race/ethnicity based percentiles of the indices. “survey” (19) to account for the NHANES as nonhispanic white, nonhispanic black, The interesting results observed when complex survey design. and other), clinical interview, physical we compared participants who were examination, and other laboratory var- either at the lowest or at the highest iables were obtained following protocols quartiles of fT4 and TSH simultaneously RESULTS available in NHANES operation manuals (see Supplementary Material and RESULTS) The 5,222 participants in the sample (16). Briefly, glucose was measured with led us to develop TFQI and PTFQI (de- represented the U.S. population 20 years the glucose oxidase method and glyco- scribed above). These indices enabled of age and older, and they had a mean hemoglobin (HbA1c) on the A1c G7 HPLC us to study pituitary feedback response (SD) age of 46.9 (16.6) years (Table 1). Glycohemoglobin Analyzer (Tosoh Medics, with a continuous variable. Indices of resistance to thyroid hormone Inc., San Francisco, CA). Blood was We created population-based quar- were higher among older people and drawn from participants who were vis- tiles for each resistance to thyroid hor- lower among black individuals (Table 1). ited at any time throughout the day, mone index (TFQI, PTFQI, TT4RI, and In a preliminary analysis, obesity prev- except for a subsample of participants TSHI). Differences, odds ratios (ORs), alence and, especially, HbA1c $6.5% and appointed for a morning visit and re- and rate ratios (RRs) across quartiles, use of diabetes medication showed a quested to fast. In the main sample data using the lowest as reference, were tendency to be higher among partici- set(which includes nonfasting data), high estimated with generalized linear, logis- pants whose fT4 and TSH were simulta- HbA1c ($6.5%, i.e., $48 mmol/mol) and , and Poisson regression models, re- neously above their 75th percentiles use of diabetes medication were ana- spectively. Model 1 adjusted for sex, age, (Supplementary Data). lyzed, as they can be interpreted as and race. Model 2 further adjusted for A scatterplot of TSH versus fT4 shows proxies for diabetes. In the fasting sub- BMI. Model 3 further adjusted for initially the negative correlation of these varia- sample, diabetes was defined as fasting presenting HbA1c $6.5% and diabetes bles, while the contour lines show se- ($8 h) glycemia $126 mg/dL or present- medication use. For each trend, signifi- lected percentiles for each one of the ing the previously mentioned proxies. cance was tested with each index quartile resistance to thyroid hormone indices Obesity was defined as a BMI $30 kg/m2. ordinal as a continuous variable. Given (Fig. 1 and Supplementary Fig. 1). Metabolic syndrome diagnosis was estab- that there were no diabetes-related TT4RI and TSHI (Supplementary Fig. 1) lished using the harmonized criteria (17); deaths within the first quartile of three take extreme values (.95th and ,5th 4 Resistance to Thyroid Hormone and Diabetes Diabetes Care

Table 1—Characteristics and thyroid parameters of the U.S. population represented in the NHANES sample (N1 = 5,222) % fT4 P TSH* P TFQI P TT4RI* P TSHI P All 10.08 (2.15) 1.59 0.00 (0.37) 15.73 1.82 (0.74)

Male 48.3 10.04 0.431 1.62 0.220 0.00 0.403 16.02 0.147 1.83 0.299 Female 51.7 10.11 1.57 20.01 15.47 1.81

$20 years and ,40 years 37.0 10.00 ,0.001 1.47 ,0.001 20.05 ,0.001 14.41 ,0.001 1.73 ,0.001 $40 years and ,60 years 39.3 9.80 1.64 20.03 15.79 1.81 $60 years 23.7 10.67 1.72 0.13 17.96 1.98

White 70.7 10.06 0.234 1.69 ,0.001 0.02 ,0.001 16.68 ,0.001 1.88 ,0.001 Black 10.0 9.91 1.19 20.17 11.51 1.51 Other 19.3 10.22 1.49 20.01 14.95 1.77 Data are percentages or means (SD) unless otherwise indicated. P values are for differences between groups and were calculated from linear regression models. *Calculated as geometric means.

percentile) for almost all patients diag- mentioned associations and provided progressively higher across TFQI quar- nosed with clinical hypo- and hyperthy- more moderate values among patients tiles after adjustment for sex, age, and roidism. This seemingly inappropriate with clinical thyroid gland–caused (primary) race (all P , 0.01). ORs of the fourth result together with our observation disease is what led us to develop TFQI. versus the first TFQI quartile for that a quantile-based approach (Supple- BMI, obesity prevalence, HbA1c $ 6.5%, HbA1c $6.5% and use of diabetes med- mentary Data) uncovered the above- and use of diabetes medication were ication were 1.73 (95% CI 1.32, 2.27) and

TFQI = cdf fT4 − ( 1 − cdf TSH )

P95 P25

P75 P95 P50 P5 P25 P75 TFQI interval P5 300 −P5 P5−P25 P25−P50 P50−P75 P75−P95 P95− 250 Diabetes−related mortality (crude rate) 200 (white dots) 150 0.1 1 10 100 TSH (mIU/L) 100 50 0 0 0 38 108 205 285 deaths per 105 person−years

P50 P5 P25 P75 P95 P50 0.001 0.01 0 102030405060

fT4 (pmol/L)

Figure 1—Scatterplot of TSH versus fT4 and bar plot of crude diabetes-related death rates across intervals of TFQI in the NHANES sample (N1 = 5,222). Scatterplot: dots represent observations. White dots represent participants who died during follow-up and whose deaths were diabetes related. Gray dashed vertical and horizontal straight lines mark fT4 and TSH normality ranges. Inner marks in the axes represent the 5th, 25th, 50th, 75th, and 95th percentiles of fT4 and TSH. TFQI was calculated for each observation, and its value grows gradually across the plot area, from the lower-left corner to the upper-right corner. Black dashed curves represent the 5th, 25th, 50th, 75th, and 95th percentiles of TFQI. Inner plot: the bars and the number at the base show crude death rates for each TFQI interval (percentile ,5, 5–25, 25–50, 50–75, 75–95, and $95) in deaths per 105 person-years. The widths of the bars are proportional to the fraction of the population at risk in each bar so that the area of the bar is proportional to the absolute number of deaths (i.e., the count of white dots in the scatterplot). care.diabetesjournals.org Laclaustra and Associates 5

Table 2—Association of TFQI with BMI, obesity, high HbA1c, use of diabetes medication, and diabetes-related deaths in the main sample (N2 = 5,129) TFQI Quartile 1 Quartile 2 Quartile 3 Quartile 4

[$21, and ,20.25) [$20.25 and ,0) [$0 and ,0.25) [$0.25 and #1) Ptrend N 1,319 1,259 1,270 1,281

BMI (kg/m2) 28.1 28.0 29.0 28.8 Model 1, difference (95% CI) 0.0 (Reference) 0.1 (20.6, 0.7) 1.1 (0.7, 1.5) 0.9 (0.1, 1.6) 0.006

Obesity (%) 31.0 29.0 36.7 35.8 Model 1, OR (95% CI) 1.00 (Reference) 0.93 (0.76, 1.14) 1.34 (1.17, 1.54) 1.29 (1.11, 1.50) 0.001

HbA1c $6.5% (%) 5.0 6.4 8.9 10.0 Model 1, OR (95% CI) 1.00 (Reference) 1.28 (0.96, 1.71) 1.70 (1.37, 2.12) 1.73 (1.32, 2.27) 0.002 Model 2, OR (95% CI) 1.00 (Reference) 1.19 (0.91, 1.54) 1.38 (1.06, 1.78) 1.43 (1.07, 1.92) 0.034

Diabetes medication (%) 4.8 7.2 9.2 9.9 Model 1, OR (95% CI) 1.00 (Reference) 1.46 (1.00, 2.13) 1.72 (1.21, 2.44) 1.66 (1.31, 2.10) 0.001 Model 2, OR (95% CI) 1.00 (Reference) 1.36 (0.96, 1.94) 1.41 (1.00, 1.98) 1.39 (1.10, 1.75) 0.028

Diabetes-related deaths (n)04716 Crude rate (per 100,000 person-years) 0.0 38.9 108.9 211.3 Crude rate (per 100,000 person-years) 19.5 159.2 Model 1, RR (95% CI) 1.00 (Reference) 4.81 (1.01, 22.94) 0.015# Model 2, RR (95% CI) 1.00 (Reference) 3.92 (0.88, 17.51) 0.035# Model 3, RR (95% CI) 1.00 (Reference) 3.51 (0.78, 15.85) 0.099# Differences, ORs, and RRs are estimated with generalized linear, logistic, and Poisson regression models, respectively. Model 1 is adjusted for sex, age, and race. Model 2 is further adjusted for BMI. Model 3 is further adjusted for initially presenting HbA1c $6.5% and using diabetes medication. Ptrend is calculated with the TFQI quartile ordinal as a continuous variable except in the models with only two categories in which it is calculated with TFQI as a continuous variable. #Ptrend for TFQI as a continuous variable.

1.66 (1.31, 2.10), respectively, and the 3 years and 11 months, during which and race-adjusted RR for diabetes- association was independent of BMI, there were 27 deaths with diabetes related death was 4.81 (95% CI 1.01, given that after adjustment, OR estima- listed within the entity axis codes for mul- 22.94). TFQI was linearly associated with tions were only partially reduced to tiple causes of death. In nine of these the rate for diabetes-related death (P = 1.43 (1.07, 1.92) and 1.39 (1.10, 1.75), cases, diabetes was the leading cause of 0.015), and the association remained respectively (Ptrend ,0.05) (Table 2). death. For participants with TFQI above statistically significant even after adjust- Participants had a median follow-up of the median (versus below it), sex-, age-, ment for BMI (P = 0.035). One-fourth of

Table 3—ORs (95% CI) of diabetes, metabolic syndrome, and metabolic syndrome criteria for 1-SD increase of TFQI,

PTFQI, and TSHI or multiplication by an SD equivalent of TT4RI in the fasting normothyroid sample (N4 = 1,997) TFQI (+1 SD) PTFQI (+1 SD) TT4RI (3 2SD log2 TT4RI) TSHI (+1 SD) Diabetes 1.13 (1.02, 1.25) 1.13 (1.02, 1.25) 1.09 (0.89, 1.33) 1.10 (0.93, 1.31) P 0.041 0.036 0.420 0.299 Metabolic syndrome 1.16 (1.02, 1.31) 1.16 (1.03, 1.31) 1.30 (1.09, 1.56) 1.28 (1.09, 1.50) P 0.040 0.031 0.014 0.013 Waist criterion 0.99 (0.88, 1.11) 0.99 (0.89, 1.11) 1.06 (0.91, 1.24) 1.05 (0.91, 1.21) P 0.814 0.870 0.439 0.536 Triglyceride criterion 1.07 (0.88, 1.30) 1.07 (0.88, 1.29) 1.40 (1.16, 1.69) 1.33 (1.10, 1.61) P 0.510 0.506 0.005 0.013 HDL criterion 1.21 (1.05, 1.39) 1.21 (1.06, 1.38) 1.21 (1.04, 1.40) 1.22 (1.06, 1.41) P 0.024 0.018 0.034 0.021 Blood pressure criterion 1.18 (1.04, 1.34) 1.19 (1.06, 1.35) 1.22 (1.07, 1.38) 1.21 (1.08, 1.37) P 0.023 0.016 0.013 0.008 Glycemia criterion 1.06 (0.94, 1.20) 1.06 (0.94, 1.19) 1.08 (0.88, 1.32) 1.07 (0.90, 1.28) P 0.364 0.371 0.472 0.461 ORs are estimated with generalized logistic regression models adjusted for sex, age, and race (model 1) for the increase of 0.37 units TFQI, 0.33 units PTFQI, and 0.74 units TSHI (1 SD) and for the multiplication of TT4RI by 2.1 (1 SD equivalent). Indices enter the models as continuous variables. TT4RI enters the model as base 2 log so that the regression coefficient is for each duplication of its value. ORs are comparable across columns in terms of influence of the index variation on the outcome prevalences. 6 Resistance to Thyroid Hormone and Diabetes Diabetes Care

deaths occurred among participants hyperthyroxinemia (1) but also call for diabetes onset in a previously meta- without initial presence of proxies for a new interpretation of thyroid hor- bolically normal individual. diabetes. However, in spite of a similar mone changes associated with diabetes. Besides secretion regulation by the magnitude of association for being above As far as we are aware, this is the first hypothalamic-pituitary-thyroid axis, thy- the median, RR 3.51 (0.78, 15.85), such a time that indices of resistance to thy- roid hormone action is modulated (6,26) small number of events made it impos- roid hormone have been evaluated in at transport, enzymatic sible to detect a statistically significant the general population and associated modification into active and inactive association after adjustment for the ini- with cardiometabolic health charac- species by deiodinases, and nuclear re- tial condition. All results were similar in teristics. Our analyses showed that ceptors. Nutritional signals can modulate examination of PTFQI instead of TFQI TT4RI and TSHI indices seemed to sys- the thyroid system during some of the (Supplementary Data). Adjustment for tematically classify thyroid gland–caused regulation, secretion, and action steps. education, physical activity, and sed- (primary) hypothyroidism and hyper- Hypothalamic neurons inhibit secretion entary time did not substantially thyroidism (clinical and subclinical) as when leptin levels fall, providing a system change these results (Supplementary resistant or very sensitive to thyroid to protect against starvation (27). Mean- Data). hormones, respectively. Thus, we pro- while, an increase in leptin derived from TT4RI and TSHI were cross-sectionally posed a new index, TFQI. TFQI is based adipose tissue accumulation may indi- associated with obesity, although not on the empirical joint distribution of cate availability of energetic substrates, with diabetes proxies. Nonetheless, fT4 and TSH with the advantage of not and, thus, the subsequent activation of both were prospectively associated yielding extreme values in cases of thy- the thyroid axis could also be interpreted with diabetes-related mortality inde- roid gland dysfunction. Its parametric as a homeostatic regulation. Accordingly, pendently of initial presence of diabe- version (PTFQI) can readily be used high TSHdwithin normal levelsdis as- tes proxies (Supplementary Data). All of as a formula to calculate the index for sociated with increased adiposity (28) these results persisted in the normothy- any particular individual with reference and consistent with near-hypothyroidism roid subsample (Supplementary Data) to the U.S. population, or it can be hormonal end effects. This can be inter- with the exception of RR for mortality. adapted to other reference populations. preted as a deficient thyroid hormone Although the magnitudes for its indica- All these resistance to thyroid hor- production in relative terms, unable to tors were very similar, they failed to mone indices measure central sensitiv- meet the metabolic increase required reach statistical significance, probably ity/resistance, i.e., the grade of pituitary to compensate and avoid more fat ac- due to the statistical loss of power gland inhibition by fT4 levels. Thus, they cumulation.Interestingly,inmorbidly caused by excluding two deceased par- evaluate the set point of the central obese individuals thyroid hormones are ticipants who had marginally subclinical regulation of thyroid hormone concen- also elevated (12), hinting at a down- hypothyroidism (Fig. 1 and Supplemen- tration. Among participants with high stream signal problem. Activity of de- tary Data). values, peripheral resistance could also iodinases (29), epigenetic modification We tested cross-sectional associations be present because, despite higher fT4, of histones, and interaction with signal- of the indices with diabetes and meta- we find in this group higher prevalence ing of other factors (30) bolic syndrome in the fasting subsam- of obesity, diabetes, and metabolic syn- also regulate, at the cellular level, thy- ple of normothyroid participants. Only drome, phenotypes usually associated roid hormone actions on metabolism. In TFQI (and its approximation, PTFQI) with hypothyroidism. addition, in obesity a steady consump- was associated with diabetes based on Thyroid hormones increase energy tion of high-calorie foods may even glycemia, HbA1c, and medication. All expenditure and thermogenesis (20), in- overpower the thyroid stimulus that resistance to thyroid hormone indices crease glucose and fatty acid oxidation in increases resting metabolic rate (31), were associated with metabolic syndrome muscle (21) and the liver (22), increase leading to a biochemical phenotype in general and with HDL and blood pres- adipose tissue lipolysis (23), and promote compatible with an apparent resistance sure metabolic syndrome criteria. Lastly, lower body weight (24). In sum, they have to thyroid hormone. We previously pro- TT4RI and TSHI were also associated an overall effect on metabolism that posed a similar interpretation for the with triglyceride criterion (Table 3). would prevent diabetes development. origin of insulin resistance in obesity (32). Interestingly, thyroid hormones also sen- From a clinical point of view, there CONCLUSIONS sitize to catecholamines, which increase seems to be a gradient of increasing Based on a sample representative of the liver hepatic glycogenolysis, neogluco- thyroid hormones levelsdwithin the U.S. population, we provide evidence genesis, and glucose output (25); reduce normal rangedacross early stages of of the association between indices insulin anabolic actions in the liver; and diabetes (33). Furthermore, prospec- measuring resistance to thyroid hor- increase glucose intestinal absortion (5), tively, higher levels of thyroid hormones mone and prevalence of obesity, diabe- though the latter effect plays a minor role are associated with incident diabetes (1) tes, and metabolic syndrome as well as in glucose level derangement. In overt in spite of the fact that hypothyroidism between these indices and the incidence hyperthyroidism, these changes could is the disorder that clearly increases of diabetes-related deaths. These associ- be responsible for decompensation of diabetes risk (3). As resistance to thyroid ations were also identified within the preexisting diabetes and even trigger hormone is one of the differential di- normal fT4 and TSH ranges. These results ketoacidosis (25), but they may be less agnoses when both fT4 and TSH are not only reconcile previous reports linking important within the normal range of elevated (34), our results offer an expla- diabetes to hypothyroidism (3) and to thyroid hormones and unlikely to cause nation for thyroid profiles commonly care.diabetesjournals.org Laclaustra and Associates 7

foundinpatientswithdiabetes.Thatis,at results between cross-sectional associa- diabetes morbidity and mortality were the population level, measurements of tions and longitudinal mortality analyses independent of obesity. We propose the resistance to thyroid hormone are cross- support our conclusions. However, some new TFQI to identify mild levels of ac- sectionally associated with diabetes, in- limitations exist. Indices of resistance to quired resistance in the general popula- dependently of the degree of obesity as thyroid hormone increase markedly with tion. Type 2 diabetes is in part driven by measured with BMI, suggesting that aging. Whereas all analyses are adjusted an energy balance problem, and the there might be other underlying mech- for age and other potential confound- associated central resistance measured anisms linking diabetes and resistance to ers, some residual confounding may by indices of resistance to thyroid hor- thyroid hormone. Furthermore, prospec- remain. Given that many factors may mone may be the result of the physiologic tively, these measurements were also be associated with diabetes and meta- contraregulation against it. These indices associated with diabetes-related deaths, bolic syndrome, it is not possible to may facilitate the monitoring and eval- even independently of the initial diabetes adjust for unavailable variables and un- uation of new therapies that focus on status. known factors, and the potential re- energy expenditure. Measurements of resistance to thy- sidual confounding must be considered roid hormone were also cross-sectionally when interpreting the study results. associated with metabolic syndrome on One antidiabetes oral agent, metformin, Funding. The research activity of M.L. is funded a consistent basis. Previous attempts to lowers TSH (leaving fT4 unchanged) by Agencia Aragonesa para la Investigacion´ y el relate metabolic syndrome with the thy- among patients with hypothyroidism Desarrollo (ARAID). This work was supported by grants from Instituto de Salud Carlos III (nos. roid axis either showed an association (37), and there is some evidence hinting PI14/00009 and PI17/01709) (State Secretary of with TSH but not with fT4 (35) or yielded that it sensitizes the and R+D+I and cofunded by European Regional De- inconsistent results. We hypothesized peripheral tissues to thyroid hormone velopment Fund/European Social Fund “Invest- that whereas a metabolic status with action (40). Thus, our results may un- ing in your future”). excess energy may stimulate the thyroid derestimate the true association be- None of the funding sources had any role in the writing of the manuscript or the decision to axis, the thyroid boost to energy expen- tween resistance to thyroid hormone submit it for publication. diture fails to compensate. This results in and diabetes; i.e., the association may Duality of Interest. No potential conflicts of an expression compatible with resistance be even stronger than the statistically interest relevant to this article were reported. to thyroid hormone. Thyroid stimuli significant association reported here. Author Contributions. M.L. conceptualized fi the research question and designed the study. still increase, although insuf ciently, the Because our analysis focused on resis- M.L. and B.M.-F. performed the statistical data uncoupled thermogenic oxidative pro- tance to thyroid hormone indices, which analysis. M.L., B.M.-F., J.M.L.-B., and F.C. inter- cesses, which can be observed as an are all fT4 based, fT3 was intentionally preted the results. M.L. and B.M.-F. wrote the increase of oxidative stress in metabolic excluded from the models, since it was manuscript. R.M.-G., J.A.C., P.G.-C., A.C., and F.C. syndrome (36). A sustained compensa- considered to be downstream in the revised the manuscript for important intellec- tual content. M.L., B.M.-F., J.M.L.-B., R.M.-G., tory TSH stimulus could also explain, in action pathway. Actually, deiodinases J.A.C., P.G.-C., A.C., and F.C. approved the final part, the recently reported association of balance may play a role in the modulation version of the manuscript for publication. M.L. metabolic syndrome, insulin resistance, of sensitivity to thyroid hormones (29), is the guarantor of this work and, as such, had and diabetes with thyroid enlargement, and future projects should be devoted to full access to all the data in the study and takes thyroid nodules, and (37). specifically study their role in resistance responsibility for the integrity of the data and the accuracy of the data analysis. We cannot exclude that other condi- to thyroid hormone in obesity, metabolic tions, like immunity disorders, could also syndrome, and diabetes. Diabetes and References explain a coincidence of elevated thyroid related outcomes were based on one 1. Ittermann T, Schipf S, Dorr¨ M, et al. Hyper- hormones, due to autoimmune thyroid single measurement of HbA1c or fast- thyroxinemia is positively associated with prev- disease, and latent autoimmune diabe- ing plasma glucose, which is a common alent and incident type 2 diabetes mellitus in tes of adults. However, from a pop- limitation in population-based studies. two population-based samples from Northeast ulation perspective, energy imbalance is Finally, the small number of observed Germany and Denmark. Nutr Metab Cardiovasc much more prevalent. diabetes-related deaths limits the statis- Dis 2018;28:173–179 Type 2 diabetes requires a broad tical power needed to find subtle effects 2. Vargas-Castillo A, Fuentes-Romero R, Rodriguez- clinical management beyond glucose as well as to perform subgroup analyses. Lopez LA, Torres N, Tovar AR. Understanding the biology of thermogenic fat: is browning a new control,focusedonenergybalance. Notwithstanding, the association was approach to the treatment of obesity? Arch Med Monitoring resistance to thyroid hor- strong enough to bear statistically sig- Res 2017;48:401–413 mone could be used to evaluate the nificant results. 3. Gronich N, Deftereos SN, Lavi I, Persidis AS, intended reversal of the abnormal en- Abernethy DR, Rennert G. Hypothyroidism is a ergy balance. Interestingly, where tradi- Conclusion risk factor for new-onset diabetes: a cohort tional interventions of food intake Higher values in indices of resistance study. Diabetes Care 2015;38:1657–1664 reduction and physical exercise increase to thyroid hormone are associated with 4. Laclaustra M, Hurtado-Roca Y, Sendin M, et al. seem to fail, new treatments trying to obesity, metabolic syndrome, and diabe- Lower-normal TSH is associated with better metabolic risk factors: a cross-sectional study modify the activity of uncoupling pro- tes for the entire population in general on Spanish men. Nutr Metab Cardiovasc Dis teins have been proposed (2,38,39). and also for normothyroid individuals 2015;25:1095–1103 This study is based on a representative in particular. Higher values are also 5. Dimitriadis GD, Raptis SA. Thyroid hormone sample of the U.S. population with strict prospectively associated with diabetes- excess and glucose intolerance. Exp Clin Endo- data collection protocols. Consistency in related mortality. These associations with crinol Diabetes 2001;109(Suppl. 2):S225–S239 8 Resistance to Thyroid Hormone and Diabetes Diabetes Care

6. Dumitrescu AM, Refetoff S. The syndromes Lung, and Blood Institute; American Heart As- 28. de Moura Souza A, Sichieri R. Association of reduced sensitivity to thyroid hormone. sociation; World Heart Federation; International between serum TSH concentration within the Biochim Biophys Acta 2013;1830:3987–4003 Society; International Associa- normal range and adiposity. Eur J Endocrinol 7. Onigata K, Szinnai G. Resistance to thyroid tion for the Study of Obesity. Harmonizing 2011;165:11–15 hormone. Endocr Dev 2014;26:118–129 the metabolic syndrome: a joint interim state- 29. Gereben B, Zavacki AM, Ribich S, et al. 8. Ortiga-Carvalho TM, Sidhaye AR, Wondisford ment of the International Diabetes Federation Cellular and molecular basis of deiodinase- FE. Thyroid hormone receptors and resistance Task Force on Epidemiology and Prevention; regulated thyroid hormone signaling. Endocr Rev to thyroid hormone disorders. Nat Rev Endo- National Heart, Lung, and Blood Institute; Amer- 2008;29:898–938 – crinol 2014;10:582 591 ican Heart Association; World Heart Federation; 30. Mullur R, Liu Y-Y, Brent GA. Thyroid hormone 9. Tjørve E, Tjørve KMC, Olsen JO, Senum R, International Atherosclerosis Society; and Inter- regulation of metabolism. Physiol Rev 2014;94: Oftebro H. On commonness and rarity of thy- national Association for the Study of Obesity. 355–382 roid hormone resistance: a discussion based on Circulation 2009;120:1640–1645 31. Carneiro IP, Elliott SA, Siervo M, et al. Is mechanisms of reduced sensitivity in peripheral 18. R Core Team. R: A Language and Environ- obesity associated with altered energy expen- tissues. Med Hypotheses 2007;69:913–921 ment for Statistical Computing. Vienna, Austria, diture? Adv Nutr 2016;7:476–487 10. Azizi F. Effect of dietary composition on R Foundation for Statistical Computing, 2018 fasting-induced changes in serum thyroid hor- 32. Laclaustra M, Corella D, Ordovas JM. Met- 19. Lumley T. Analysis of complex survey sam- abolic syndrome pathophysiology: the role of mones and thyrotropin. Metabolism 1978;27: ples. J Stat Softw 2004;9:1–19 935–942 adipose tissue. Nutr Metab Cardiovasc Dis 2007; 20. Kim B. Thyroid hormone as a determinant 17:125–139 11. Warren MP. Endocrine manifestations of of energy expenditure and the basal metabolic eating disorders. J Clin Endocrinol Metab 2011; 33. Lambadiari V, Mitrou P, Maratou E, et al. rate. Thyroid 2008;18:141–144 96:333–343 Thyroid hormones are positively associated 21. Salvatore D, Simonides WS, Dentice M, 12. Laurberg P, Knudsen N, Andersen S, Carle´ with insulin resistance early in the development Zavacki AM, Larsen PR. Thyroid hormones A, Pedersen IB, Karmisholt J. Thyroid function of type 2 diabetes. Endocrine 2011;39:28–32 and skeletal muscle–new insights and potential and obesity. Eur Thyroid J 2012;1:159–167 34. Koulouri O, Moran C, Halsall D, Chatterjee K, implications. Nat Rev Endocrinol 2014;10: 13. Yagi H, Pohlenz J, Hayashi Y, Sakurai A, Gurnell M. Pitfalls in the measurement and in- 206–214 Refetoff S. Resistance to thyroid hormone caused terpretation of . Best Pract 22. Sinha RA, Singh BK, Yen PM. Direct effects of by two mutant thyroid hormone receptors beta, Res Clin Endocrinol Metab 2013;27:745–762 thyroid hormones on hepatic lipid metabolism. R243Q and R243W, with marked impairment 35. Iwen KA, Schroder¨ E, Brabant G. Thyroid Nat Rev Endocrinol 2018;14:259–269 of function that cannot be explained by altered hormones and the metabolic syndrome. Eur 9 fi 23. Carmean CM, Cohen RN, Brady MJ. Systemic in vitro 3,5,3 -triiodothyroinine binding af nity. Thyroid J 2013;2:83–92 J Clin Endocrinol Metab 1997;82:1608–1614 regulation of adipose metabolism. Biochim Bio- – 36. Hurtado-Roca Y, Bueno H, Fernandez-Ortiz 14. Jostel A, Ryder WDJ, Shalet SM. The use of phys Acta 2014;1842:424 430 24. Fox CS, Pencina MJ, D’Agostino RB, et al. A, et al. Oxidized LDL is associated with metabolic thyroid function tests in the diagnosis of hypo- Syndrome traits independently of central obesity pituitarism: definition and evaluation of the TSH Relations of thyroid function to body weight: and insulin resistance. Diabetes 2017;66:474–482 Index. Clin Endocrinol (Oxf) 2009;71:529–534 cross-sectional and longitudinal observations in 37. Meng X, Xu S, Chen G, Derwahl M, Liu C. 15. Ercan-Fang S, Schwartz HL, Mariash CN, a community-based sample. Arch Intern Med Metformin and . J Endocrinol Oppenheimer JH. Quantitative assessment of 2008;168:587–592 2017;233:R43–R51 pituitary resistance to thyroid hormone from 25. Potenza M, Via MA, Yanagisawa RT. Excess 38. Song N-J, Chang S-H, Li DY, Villanueva CJ, plots of the logarithm of thyrotropin versus thyroid hormone and carbohydrate metabolism. serum free thyroxine index. J Clin Endocrinol Endocr Pract 2009;15:254–262 Park KW. Induction of thermogenic adipocytes: Metab 2000;85:2299–2303 26. Refetoff S, Bassett JHD, Beck-Peccoz P, et al. molecular targets and thermogenic small mol- 16. Zipf G, Chiappa M, Porter KS, Ostchega Y, Classification and proposed nomenclature for ecules. Exp Mol Med 2017;49:e353 Lewis BG, Dostal J. National health and nutrition inherited defects of thyroid hormone action, 39. Tamucci KA, Namwanje M, Fan L, Qiang L. The examination survey: plan and operations, 1999- cell transport, and metabolism. Thyroid 2014; dark side of browning. Protein Cell 2018;9:152–163 2010. Vital Health Stat 1 2013:1–37 24:407–409 40. Krysiak R, Okopien B. Thyrotropin-lowering 17. Alberti KGMM, Eckel RH, Grundy SM, et al.; 27. Hollenberg AN. The role of the thyrotropin- effect of metformin in a patient with resistance International Diabetes Federation Task Force on releasing hormone (TRH) neuron as a metabolic to thyroid hormone. Clin Endocrinol (Oxf) 2011; Epidemiology and Prevention; Hational Heart, sensor. Thyroid 2008;18:131–139 75:404–406