Association of Thyroid Function Genetic Predictors With Atrial Fibrillation: A Phenome-Wide Association Study and Inverse-Variance Weighted Average Meta-analysis | Atrial Fibrillation | JAMA Cardiology | JAMA Network
[Skip to Navigation]
Sign In
Figure.  Association of AF Risk With Thyrotropin Levels
Association of AF Risk With Thyrotropin Levels

A, Scatterplot summarizing associations between the thyrotropin predictor and 11 diagnoses in 37 154 individuals in the BioVU eMERGE (Electronic Medical Records and Genetics) cohort. Each dot represents a disease association. Significant (P ≤ 1.89 × 10−5) associations are shown in orange, and selected diagnoses are labeled. B, Scatterplot comparing the associations with thyrotropin levels and atrial fibrillation (AF) risk for 24 single-nucleotide variants (SNVs) significantly associated with thyrotropin by prior genome-wide association studies in 17 931 individuals with AF and 115 142 controls in the AF Genetics cohort.4 The x-axis is the SD change in thyrotropin levels per copy of the reference allele and the y-axis is the change in the log(odds-ratio) of AF risk. The horizontal and vertical whiskers around each point represent the 95% CI for the thyrotropin and AF effect sizes, respectively. The line displays the results of an inverse-variance weighted average meta-analysis summarizing the association between thyrotropin levels and AF risk. NOS indicates not otherwise specified.

Table 1.  Association Between AF and Individual SNVs Used to Build the Thyrotropin Polygenic Risk Scores
Association Between AF and Individual SNVs Used to Build the Thyrotropin Polygenic Risk Scores
Table 2.  Demographic and Clinical Characteristics of the PheWAS Analytic Data Set by Tertile of International Thyrotropin Polygenic Risk Score
Demographic and Clinical Characteristics of the PheWAS Analytic Data Set by Tertile of International Thyrotropin Polygenic Risk Score
1.
De Leo  S, Lee  SY, Braverman  LE.  Hyperthyroidism.  Lancet. 2016;388(10047):906-918. doi:10.1016/S0140-6736(16)00278-6PubMedGoogle ScholarCrossref
2.
Ryödi  E, Salmi  J, Jaatinen  P,  et al.  Cardiovascular morbidity and mortality in surgically treated hyperthyroidism: a nation-wide cohort study with a long-term follow-up.  Clin Endocrinol (Oxf). 2014;80(5):743-750. doi:10.1111/cen.12359PubMedGoogle ScholarCrossref
3.
Porcu  E, Medici  M, Pistis  G,  et al.  A meta-analysis of thyroid-related traits reveals novel loci and gender-specific differences in the regulation of thyroid function.  PLoS Genet. 2013;9(2):e1003266. doi:10.1371/journal.pgen.1003266PubMedGoogle ScholarCrossref
4.
Gudmundsson  J, Sulem  P, Gudbjartsson  DF,  et al.  Discovery of common variants associated with low TSH levels and thyroid cancer risk.  Nat Genet. 2012;44(3):319-322. doi:10.1038/ng.1046PubMedGoogle ScholarCrossref
5.
Denny  JC, Ritchie  MD, Basford  MA,  et al.  PheWAS: demonstrating the feasibility of a phenome-wide scan to discover gene-disease associations.  Bioinformatics. 2010;26(9):1205-1210. doi:10.1093/bioinformatics/btq126PubMedGoogle ScholarCrossref
6.
Denny  JC, Bastarache  L, Ritchie  MD,  et al.  Systematic comparison of phenome-wide association study of electronic medical record data and genome-wide association study data.  Nat Biotechnol. 2013;31(12):1102-1110. doi:10.1038/nbt.2749PubMedGoogle ScholarCrossref
7.
Gottesman  O, Kuivaniemi  H, Tromp  G,  et al; eMERGE Network.  The Electronic Medical Records and Genomics (eMERGE) Network: past, present, and future.  Genet Med. 2013;15(10):761-771. doi:10.1038/gim.2013.72PubMedGoogle ScholarCrossref
8.
Carroll  RJ, Bastarache  L, Denny  JCR.  R PheWAS: data analysis and plotting tools for phenome-wide association studies in the R environment.  Bioinformatics. 2014;30(16):2375-2376. doi:10.1093/bioinformatics/btu197PubMedGoogle ScholarCrossref
9.
Roden  DM, Pulley  JM, Basford  MA,  et al.  Development of a large-scale de-identified DNA biobank to enable personalized medicine.  Clin Pharmacol Ther. 2008;84(3):362-369. doi:10.1038/clpt.2008.89PubMedGoogle ScholarCrossref
10.
Mosley  JD, Feng  Q, Wells  QS,  et al.  A study paradigm integrating prospective epidemiologic cohorts and electronic health records to identify disease biomarkers.  Nat Commun. 2018;9(1):3522. doi:10.1038/s41467-018-05624-4PubMedGoogle ScholarCrossref
11.
Kirby  JC, Speltz  P, Rasmussen  LV,  et al.  PheKB: a catalog and workflow for creating electronic phenotype algorithms for transportability.  J Am Med Inform Assoc. 2016;23(6):1046-1052. doi:10.1093/jamia/ocv202PubMedGoogle ScholarCrossref
12.
Denny  JC, Crawford  DC, Ritchie  MD,  et al.  Variants near FOXE1 are associated with hypothyroidism and other thyroid conditions: using electronic medical records for genome- and phenome-wide studies.  Am J Hum Genet. 2011;89(4):529-542. doi:10.1016/j.ajhg.2011.09.008PubMedGoogle ScholarCrossref
13.
Kho  AN, Cashy  JP, Jackson  KL,  et al.  Design and implementation of a privacy preserving electronic health record linkage tool in Chicago.  J Am Med Inform Assoc. 2015;22(5):1072-1080. doi:10.1093/jamia/ocv038PubMedGoogle ScholarCrossref
14.
Christophersen  IE, Rienstra  M, Roselli  C,  et al; METASTROKE Consortium of the ISGC; Neurology Working Group of the CHARGE Consortium; AFGen Consortium.  Large-scale analyses of common and rare variants identify 12 new loci associated with atrial fibrillation.  Nat Genet. 2017;49(6):946-952. doi:10.1038/ng.3843PubMedGoogle ScholarCrossref
15.
Mosley  JD, van Driest  SL, Wells  QS,  et al.  Defining a contemporary ischemic heart disease genetic risk profile using historical data.  Circ Cardiovasc Genet. 2016;9(6):521-530. doi:10.1161/CIRCGENETICS.116.001530PubMedGoogle ScholarCrossref
16.
Zuvich  RL, Armstrong  LL, Bielinski  SJ,  et al.  Pitfalls of merging GWAS data: lessons learned in the eMERGE network and quality control procedures to maintain high data quality.  Genet Epidemiol. 2011;35(8):887-898. doi:10.1002/gepi.20639PubMedGoogle ScholarCrossref
17.
Delaneau  O, Zagury  JF, Marchini  J.  Improved whole-chromosome phasing for disease and population genetic studies.  Nat Methods. 2013;10(1):5-6. doi:10.1038/nmeth.2307PubMedGoogle ScholarCrossref
18.
Howie  B, Fuchsberger  C, Stephens  M, Marchini  J, Abecasis  GR.  Fast and accurate genotype imputation in genome-wide association studies through pre-phasing.  Nat Genet. 2012;44(8):955-959. doi:10.1038/ng.2354PubMedGoogle ScholarCrossref
19.
Yang  J, Lee  SH, Goddard  ME, Visscher  PM.  GCTA: a tool for genome-wide complex trait analysis.  Am J Hum Genet. 2011;88(1):76-82. doi:10.1016/j.ajhg.2010.11.011PubMedGoogle ScholarCrossref
20.
Yang  J, Manolio  TA, Pasquale  LR,  et al.  Genome partitioning of genetic variation for complex traits using common SNPs.  Nat Genet. 2011;43(6):519-525. doi:10.1038/ng.823PubMedGoogle ScholarCrossref
21.
Purcell  SM, Wray  NR, Stone  JL,  et al; International Schizophrenia Consortium.  Common polygenic variation contributes to risk of schizophrenia and bipolar disorder.  Nature. 2009;460(7256):748-752.PubMedGoogle Scholar
22.
Burgess  S, Butterworth  A, Thompson  SG.  Mendelian randomization analysis with multiple genetic variants using summarized data.  Genet Epidemiol. 2013;37(7):658-665. doi:10.1002/gepi.21758PubMedGoogle ScholarCrossref
23.
Chugh  SS, Havmoeller  R, Narayanan  K,  et al.  Worldwide epidemiology of atrial fibrillation: a Global Burden of Disease 2010 Study.  Circulation. 2014;129(8):837-847. doi:10.1161/CIRCULATIONAHA.113.005119PubMedGoogle ScholarCrossref
24.
Salem  JE, El-Aissaoui  M, Alazard  M,  et al.  Modeling of amiodarone effect on heart rate control in critically ill patients with atrial tachyarrhythmias.  Clin Pharmacokinet. 2016;55(8):991-1002. doi:10.1007/s40262-016-0372-9PubMedGoogle ScholarCrossref
25.
Biondi  B, Bartalena  L, Cooper  DS, Hegedüs  L, Laurberg  P, Kahaly  GJ.  The 2015 European Thyroid Association guidelines on diagnosis and treatment of endogenous subclinical hyperthyroidism.  Eur Thyroid J. 2015;4(3):149-163. doi:10.1159/000438750PubMedGoogle ScholarCrossref
26.
Sawin  CT, Geller  A, Wolf  PA,  et al.  Low serum thyrotropin concentrations as a risk factor for atrial fibrillation in older persons.  N Engl J Med. 1994;331(19):1249-1252. doi:10.1056/NEJM199411103311901PubMedGoogle ScholarCrossref
27.
Cappola  AR, Fried  LP, Arnold  AM,  et al.  Thyroid status, cardiovascular risk, and mortality in older adults.  JAMA. 2006;295(9):1033-1041. doi:10.1001/jama.295.9.1033PubMedGoogle ScholarCrossref
28.
Selmer  C, Olesen  JB, Hansen  ML,  et al.  The spectrum of thyroid disease and risk of new onset atrial fibrillation: a large population cohort study.  BMJ. 2012;345:e7895. doi:10.1136/bmj.e7895PubMedGoogle ScholarCrossref
29.
Garber  JR, Cobin  RH, Gharib  H,  et al; American Association Of Clinical Endocrinologists And American Thyroid Association Taskforce On Hypothyroidism In Adults.  Clinical practice guidelines for hypothyroidism in adults: cosponsored by the American Association of Clinical Endocrinologists and the American Thyroid Association.  Thyroid. 2012;22(12):1200-1235. doi:10.1089/thy.2012.0205PubMedGoogle ScholarCrossref
30.
Alonso  H, Fernandez-Ruocco  J, Gallego  M,  et al Thyroid stimulating hormone directly modulates cardiac electrical activity.  J Mol Cell Cardiol. 2015;89(pt B):280-286. doi:10.1016/j.yjmcc.2015.10.019Google ScholarCrossref
31.
Strauss  DG, Vicente  J, Johannesen  L,  et al.  Common genetic variant risk score is associated with drug-induced QT prolongation and torsade de pointes risk: a pilot study.  Circulation. 2017;135(14):1300-1310. doi:10.1161/CIRCULATIONAHA.116.023980PubMedGoogle ScholarCrossref
32.
Salem  JE, Germain  M, Hulot  JS,  et al.  GENomE wide analysis of sotalol-induced IKr inhibition during ventricular REPOLarization, “GENEREPOL study”: lack of common variants with large effect sizes.  PLoS One. 2017;12(8):e0181875. doi:10.1371/journal.pone.0181875PubMedGoogle ScholarCrossref
Original Investigation
January 23, 2019

Association of Thyroid Function Genetic Predictors With Atrial Fibrillation: A Phenome-Wide Association Study and Inverse-Variance Weighted Average Meta-analysis

Author Affiliations
  • 1Department of Medicine, Vanderbilt University Medical Center, Nashville, Tennessee
  • 2Sorbonne Université, Institut National de la Santé et de la Recherche Médicale (INSERM) CIC Paris-Est, AP-HP, Institute of Cardio metabolism and Nutrition (ICAN), Pitié-Salpêtrière Hospital, Department of Pharmacology, Paris, France
  • 3Department of Pharmacology, Vanderbilt University Medical Center, Nashville, Tennessee
  • 4Biomedical Informatics, Vanderbilt University Medical Center, Nashville, Tennessee
  • 5Department of Medicine (Medical Genetics), University of Washington, Seattle
  • 6Department Genome Sciences, University of Washington, Seattle
  • 7Kaiser Permanente Washington Health Research Institute, Seattle
  • 8Department of Obstetrics and Gynecology, Vanderbilt University Medical Center, Nashville, Tennessee
  • 9Department of Biomedical Informatics, Vanderbilt University Medical Center, Nashville, Tennessee
  • 10Divisions of Human Genetics and Pulmonary Medicine, The Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania
  • 11Department of Biomedical Informatics, Columbia University, New York
  • 12Cardiovascular Research Center, Massachusetts General Hospital, Boston
  • 13The Broad Institute of Harvard and MIT, Cambridge, Massachusetts
JAMA Cardiol. 2019;4(2):136-143. doi:10.1001/jamacardio.2018.4615
Key Points

Question  Is there any association between a polygenic predictor of thyroid function and atrial fibrillation (AF) risk?

Findings  This phenome-wide association study of 37 154 individuals applied a genetic predictor of thyrotropin levels to an electronic health record cohort and identified expected associations with thyroid disorders as well as a significant inverse association with AF, even after dropping cases of overt thyroid disease. This finding was replicated in an AF genome-wide association study (17 931 AF cases and 115 142 controls).

Meaning  This study suggests a potential role for genetically determined variation in thyroid function within a physiologically accepted normal range as a risk factor for this increasingly common arrhythmia.

Abstract

Importance  Thyroid hormone levels are tightly regulated through feedback inhibition by thyrotropin, produced by the pituitary gland. Hyperthyroidism is overwhelmingly due to thyroid disorders and is well recognized to contribute to a wide spectrum of cardiovascular morbidity, particularly the increasingly common arrhythmia atrial fibrillation (AF).

Objective  To determine the association between genetically determined thyrotropin levels and AF.

Design, Setting, and Participants  This phenome-wide association study scanned 1318 phenotypes associated with a polygenic predictor of thyrotropin levels identified by a previously published genome-wide association study that included participants of European ancestry. North American individuals of European ancestry with longitudinal electronic health records were analyzed from May 2008 to November 2016. Analysis began March 2018.

Main Outcomes and Measures  Clinical diagnoses associated with a polygenic predictor of thyrotropin levels.

Exposures  Genetically determined thyrotropin levels.

Results  Of 37 154 individuals, 19 330 (52%) were men. The thyrotropin polygenic predictor was positively associated with hypothyroidism (odds ratio [OR], 1.10; 95% CI, 1.07-1.14; P = 5 × 10−11) and inversely associated with diagnoses related to hyperthyroidism (OR, 0.64; 95% CI, 0.54-0.74; P = 2 × 10−8 for toxic multinodular goiter). Among nonthyroid associations, the top association was AF/flutter (OR, 0.93; 95% CI, 0.9-0.95; P = 9 × 10−7). When the analyses were repeated excluding 9801 individuals with any diagnoses of a thyroid-related disease, the AF association persisted (OR, 0.91; 95% CI, 0.88-0.95; P = 2.9 × 10−6). To replicate this association, we conducted an inverse-variance weighted average meta-analysis using AF single-nucleotide variant weights from a genome-wide association study of 17 931 AF cases and 115 142 controls. As in the discovery analyses, each SD increase in predicted thyrotropin was associated with a decreased risk of AF (OR, 0.86; 95% CI, 0.79-0.93; P = 4.7 × 10−4). In a set of AF cases (n = 745) and controls (n = 1680) older than 55 years, directly measured thyrotropin levels that fell within the normal range were inversely associated with AF risk (OR, 0.91; 95% CI, 0.83-0.99; P = .04).

Conclusions and Relevance  This study suggests a role for genetically determined variation in thyroid function within a physiologically accepted normal range as a risk factor for AF. The clinical decision to treat subclinical thyroid disease should incorporate the risk for AF as antithyroid medications to treat hyperthyroidism may reduce AF risk and thyroid hormone replacement for hypothyroidism may increase AF risk.

Introduction

Thyroid hormone levels (free thyroxine [FT4] and free triiodothyronine [FT3]) are tightly regulated through feedback inhibition by thyrotropin, produced by the pituitary gland.1 Thyrotropin is measured clinically to assess thyroid function as it can detect thyroid hormone abnormalities with higher sensitivity and specificity than FT4 or FT3.1 Overt hyperthyroidism is overwhelmingly due to thyroid disorders including Graves disease, toxic adenoma goiter, or toxic multinodular goiter1 and is well recognized to contribute to a wide spectrum of morbidity including atrial fibrillation (AF), hypertension, heart failure, and cerebrovascular diseases.2 Owing to feedback inhibition, hyperthyroidism due to thyroid gland disorders is characterized by low serum thyrotropin levels.1 Conversely, hypothyroidism is associated with high thyrotropin levels. Thyroid hormone levels are highly heritable, and multiple common single-nucleotide variants (SNVs) have been associated with thyrotropin levels.3,4

Longitudinal electronic health record (EHR) data capture diverse clinical outcomes in large sample sizes and, when coupled with large-scale DNA collections, have been used to define the disease spectrum associated with single or multiple genomic variants.5,6 We report here a phenome-wide association study (PheWAS) identifying clinical diagnoses associated with a polygenic predictor of thyrotropin levels identified by a previously published genome-wide association study (GWAS) that included North American and European participants.3 We performed phenomewide scanning of 1318 phenotypes using a cohort of 37 154 North American individuals of European ancestry with EHR data (Table 1).7,8

Methods
Study Populations

The primary EHR population was derived from the eMERGE (Electronic Medical Records and Genetics) Phase I & II Network (n = 16 924), a consortium of medical centers using EHRs as a tool for genomic research, and from Vanderbilt University Medical Center’s BioVU resource (n = 20 230).7,9 We have previously found that this data set performs robustly for biomarkers discovery and validation.10 The participating eMERGE sites were Geisinger Health System, Vanderbilt University Medical Center, Marshfield Clinic, Northwestern University, Mayo Clinic, and Kaiser Permanente/University of Washington. BioVU is Vanderbilt University Medical Center’s deidentified collection of patients whose DNA was extracted from discarded blood and linked to phenotypes through a deidentified EHR.9 All individuals were born prior to 1990 and fell within 4 SDs for each of the first 2 principal components based on common SNVs for the subset of individuals self-identified as white/non-Hispanic. The eMERGE study was approved by the local institutional review board at each eMERGE site, and participant consent was obtained per institutional review board–approved protocols at each site.7 Between-site variability in terms of the positive predictive value of phenotyping algorithms (in particular, for hypothyroidism) and replication of published associations between SNVs and clinical phenotypes has been shown to be high and have been published elsewhere.6,10-13

We identified 6797 individuals of European ancestry in BioVU who did not have any International Classification of Diseases, Ninth Revision, or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, codes for thyroid diseases and who had thyrotropin measurements that fell within the clinically normal reference range.

Summary Statistics

Two sets of SNVs previously found to be associated with thyrotropin from 2 independent GWAS studies were used in this study (Table 1).3,4 The international SNVs set (n = 20) was previously identified in 26 420 patients with euthyroid of European ancestry from Europe and North America.3 This set was used for both discovery PheWAS and the replication inverse-variance weighted average (IVWA) meta-analysis. To further confirm replication of our results using IVWA meta-analysis, we also used a second Iceland thyrotropin SNVs set (n = 24) previously identified in 27 758 Icelanders with an insular genetic background.4 For the discovery PheWAS, we used the European ancestry international rather than the Icelandic thyrotropin SNVs predictor because its extrinsic validity in a North American EHR cohort was, a priori, more suitable considering that approximately 25% of participants were from North America in the international data set.3

Summary statistics for AF were from a prior meta-analysis of 31 GWAS studies performed by the Atrial Fibrillation Genetics (AFGen) consortium.14 The AFGen consortium is a collaboration among multiple studies with the aim of investigating the genetic causes of AF. Atrial fibrillation cases had either documented AF on an electrocardiogram and/or 1 inpatient or 2 outpatient diagnoses of AF. Controls were free of AF.14 The majority of the participants were of European ancestry (15 979 cases, 102 776 referents) but also included African American (3 studies; 641 cases, 5234 referents), Japanese (1 study; 837 cases, 3293 referents), Hispanic (1 study; 277 cases, 3081 referents), and Brazilian (1 study; 197 cases, 758 referents) individuals.14

Genetic Data

In the BioVU eMERGE EHR cohort, SNVs genotype data were acquired on the Illumina Human660W-Quadv1_A, HumanOmni1-Quad, HumanOmni5-Quad, MEGA-EX, Human610, Human550, HumanOmniExpressExome-8v1.2A, and Affymetrix 6.0 SNV array platforms. Quality control steps for the EHR data sets were performed per previously published protocols.15,16 For each data set, SNVs were prephased using SHAPEIT,17 version 2.r837 (Oliver Delaneau), and imputed using IMPUTE2,18 version 2.3.0 (University of Oxford), in conjunction with the October 2014 release of the 1000 Genomes cosmopolitan reference haplotypes. All platforms were imputed to the 1000 Genomes standard, which comprises approximately 80 000 000 common and rare SNVs. Single-nucleotide variants used to construct the thyrotropin predictors were drawn from this data and were imputed on 1 or more platforms.

Phenotypic Data

Electronic health record clinical phenotypes used for the PheWAS analysis were based on phecodes (https://phewascatalog.org/), which are collections of related International Classification of Diseases, Ninth Revision, Clinical Modification, diagnosis codes.8 For each phenotype, cases are individuals with 1 or more instances of the code appearing in their medical record. Any eMERGE site that had fewer than 10 cases for a given phenotype was excluded for that phenotype (421 phenotypes had 1 or more sites excluded for the analysis). Phenotypes that only affected a single sex (eg, uterine prolapse, prostatitis) were not included in the primary analyses but were examined in secondary analyses. There were 1318 clinical phenotypes with more than 150 cases in the EHR data set that were used in the PheWAS analyses. Controls were individuals without the clinical phenotype or any closely related phecodes (as defined by the phecode) and whose decade of birth fell within the range of birth decades observed among cases. For the phenotype with the smallest number of controls (n = 1098), the study had 80% power to detect an association with an odds ratio (OR) of more than 2.9 for a phenotype with 150 cases (at α = 1.89 × 10−5), assuming a baseline risk of disease of 10% (power calculation used the Vanderbilt biostatistics power and sample size application, version 3.0).

To determine the validity of the significant PheWAS associations, clinical record review was performed by 2 clinicians (J.D.M. and J.-E.S.) for PheWAS phenotypes representative of hypothyroidism (phecode 427.21), hyperthyroidism/goiter (phecode 242.2), and hypothyroidism (phecode244). For each phenotype, 30 records of BioVU individuals who were cases in the PheWAS analyses were reviewed to determine whether the clinical documentation supported the diagnosis identified by International Classification of Diseases, Ninth Revision, codes. The positive predictive value for each phenotype was then determined.

Among the BioVU individuals with thyrotropin measurements, there were 745 AF cases and 1680 controls based on the PheWAS AF definition and who were older than 55 years. These individuals were used to directly test for an epidemiologic association between thyrotropin measurements within the normal range and AF.

Statistical Analysis

Linear mixed models, as implemented in the program Genome-wide Complex Trait Analysis, version 1.26.0,19 were used to estimate the proportion of the phenotypic variance in thyrotropin levels explained by common SNVs using a subset of individuals with measured thyrotropin and no thyroid disease (n = 6797).19 The analyses were based on 1 888 974 SNVs with minor allele frequency more than 0.01 and were adjusted for age, sex, and 20 principal components, per standard protocols.20

Genetically predicted values for thyrotropin levels within the BioVU eMERGE EHR data set were computed using weighted genetic risk scores according to the following formula:

Image description not available.

where the SNV genotype is coded as 0, 1, or 2 and wi is the β coefficient (representing effect size) from the published thyrotropin GWAS analysis.3,4

A combined thyrotropin predictor was created by combining the SNVs from the 2 predictors (international and Icelandic data sets) and using a clumping algorithm to identify a set of the most strongly associated, independent SNVs in low linkage disequilibrium (r2 < 0.2).21 This final combined predictor comprised 29 SNVs (Table 1).

The proportion of variance in measured thyrotropin levels explained by the thyrotropin genetic predictor was based on the partial correlation between the predicted (using 20 international thyrotropin SNVs) and measured thyrotropin levels, adjusted for age and sex. These analyses used the subset of BioVU individuals with measured thyrotropin.

For the PheWAS analyses, the thyrotropin predictor included 19 of 20 SNVs that passed quality control (rs11624776 was not included owing to high rates of missing values among eMERGE individuals).3 An association between each PheWAS phenotype and each genetically predicted biomarker (thyrotropin) was performed using multivariable logistic regression, adjusting for 3 principal components, birth decade, sex, eMERGE site, and genotyping platform. The genetically predicted thyrotropin value was standardized to have an SD of 1, so ORs reflect the risk per SD increase in the genetically predicted thyrotropin level. Association analyses used SAS version 9.3 (SAS Institute). To adjust for multiple testing, we applied a strict Bonferroni correction and associations with (P < .05) / (2 × 1318) = approximately 1.89 × 10−5 were considered significant. The AF association was also tested excluding 9801 individuals with a phecode related to thyroid disease (phecodes 193.*, 194.*, 226.*, 227.*, 240.*, and 247.*).

The AF-thyrotropin association was replicated using summary statistics data for AF from the AFGen consortium.14 Two independent SNV sets for thyrotropin were used: the international meta-analysis set (n = 20 SNVs)3 and the Iceland set (n = 24 SNVs).4 Genome-wide association studies such as AFGen report regression coefficients summarizing the associations between each SNV and a trait (AF for AFGen study) and not the individual-level data. These summarized data comprise the coefficients (β) and standard errors from univariate regression models of the risk factor on each genetic variant (here, association of thyrotropin with each SNV of international or Icelandic thyrotropin polygenic set) and of the outcome on each genetic variant (here, association of AF with each SNV of international or Icelandic thyrotropin polygenic set). Genetic association studies use combinations of genetic variants as instrumental variables to assess whether a risk factor is associated with a disease outcome. An estimate of association for each individual genetic variant and a combined estimate can be obtained by IVWA meta-analysis of these estimates (Figure, B for association between genetically determined thyrotropin levels and AF). Inverse-variance weighted average meta-analysis was performed using the mendelian randomization R package.22 An association of 2-sided P ≤ .05 was considered significant in the replication data set.

To ascertain whether there was an epidemiologic association between measured thyrotropin levels that fall within the normal range, a logistic regression analysis, adjusting for sex and age, was conducted on individuals older than 55 years. Thyrotropin levels were standardized, so OR estimates reflect the change in risk per SD changes in thyrotropin levels.

Results

Using a discovery population of 37 154 North American individuals of European ancestry (19 330 men [52%]) (Table 2) that we have previously shown to be effective for biomarkers discovery and characterization,10 we generated a weighted polygenic risk score using SNVs weights from a published international GWAS meta-analysis of thyrotropin based on data of individuals with euthyroid of European ancestry (Table 1).3 For each individual, we computed the expected thyrotropin level based on the polygenic risk score. In a subset of 6797 individuals with thyrotropin measurements and without thyroid disease, the estimated additive thyrotropin heritability was 0.31 ± 0.07. The polygenic predictor accounted for 5.8% of phenotypic variation and 18.6% of additive genetic heritability. Predicted thyrotropin values were used to conduct a PheWAS5,6 of 1318 clinical diagnoses (phecodes).8 In the PheWAS, the thyrotropin polygenic predictor was positively associated with hypothyroidism (OR, 1.10; 95% CI, 1.07-1.14; P = 5 × 10−11) and inversely associated with diagnoses related to hyperthyroidism (OR, 0.64; 95% CI, 0.54-0.74; P = 2 × 10−8 for toxic multinodular goiter and thyroid goiter and OR, 0.78; 95% CI, 0.73-0.83; P = 8 × 10−17 for nontoxic multinodular goiter) (Figure A and eTable in the Supplement). Among nonthyroid associations, the top association was AF/flutter (OR, 0.93; 95% CI, 0.9-0.95; P = 9 × 10−7) (eTable in the Supplement). When we repeated the analyses excluding 9801 individuals without any diagnoses of a thyroid-related disease, the AF association persisted (OR, 0.91; 95% CI, 0.88-0.95; P = 2.9 × 10−6). There were no associations in the PheWAS analyses with sex-specific phenotypes. A record review among a random sample of cases showed that the positive predictive values of the PheWAS case definitions were 0.87, 0.97, and 0.73 for diagnoses of hypothyroidism, atrial fibrillation, and toxic multinodular goiter, respectively.

To replicate the genetic association between AF and the thyrotropin polygenic predictor, we conducted an IVWA meta-analysis22 using AF SNVs weights from a GWAS of 17 931 AF cases and 115 142 controls.14 As in the discovery analyses, each increase in predicted thyrotropin was associated with a decreased risk of AF (OR, 0.91; 95% CI, 0.84-0.99; P = .03) (eFigure 1A in the Supplement). A second SNV thyrotropin predictor derived from an independent Icelandic population (based on data from 27 758 individuals and 24 SNVs)4 gave similar results (OR, 0.86; 95% CI, 0.79-0.93; P = 4.7 × 10−4) (Figure, B), and the results did not change when 2 SNVs that did not reach genomewide significance in the original Icelandic population were excluded (OR, 0.85; 95% CI, 0.78-0.93; P = 3.6 × 10−4). The 2 thyrotropin risk scores were linearly correlated (r = 0.68) and had similar overall distributions (eFigure 2 in the Supplement). Interestingly, none of the individual SNVs used to build the 2 studied thyrotropin polygenic risk scores was significantly associated with AF (Table 1). A thyrotropin predictor comprising 29 independent SNVs taken from both international and Icelandic data sets showed similar results (OR, 0.90; 95% CI, 0.84-0.97; P = .006) (eFigure 1B in the Supplement).

To ascertain whether there was a similar epidemiologic association, we examined whether AF was associated with directly measured thyrotropin in individuals without diagnosed thyroid diseases and thyrotropin levels that fell within the normal range. Among individuals older than 55 years, there was an inverse association between thyrotropin and AF (OR, 0.91 per SD change in thyrotropin levels; 95% CI, 0.83-0.99; P = .04).

Discussion

Atrial fibrillation is the most common arrhythmia worldwide with increasing frequency noted with age.23 Atrial fibrillation affects more than 33 million people worldwide and progressive increases in overall burden, incidence, prevalence, and AF-associated mortality have been reported between 1990 and 2010.23,24 Cardiovascular symptoms are the most sensitive indicators of mild thyroid hormone excess (subclinical hyperthyroidism) particularly in elderly persons, similar to overt hyperthyroidism.25-28 Subclinical hyperthyroidism, defined as decreased thyrotropin levels despite normal FT4 and FT3 levels, has been associated with AF in observational studies.25-28 Recommendations on treatment of subclinical hyperthyroidism to decrease AF risk in patients with low or undetectable thyrotropin are controversial and rely on longitudinal observational studies that have provided low quality of evidence.1 Therefore, how subclinical hyperthyroidism should be managed is still a matter of debate.1 Similarly, treatment guidelines for subclinical hypothyroidism have vacillated between treating subclinical disease and leaving hypothyroidism untreated.29

Our findings inform this debate as we observed that genetically mediated variation in thyrotropin levels modulates AF risk.1 This association between AF and thyrotropin genetic risk scores persisted even after we repeated the analyses excluding individuals with a diagnosis of thyroid disease, indicating that among individuals not meeting clinical thresholds for diagnosis, variation in genetically determined thyrotropin levels predispose to AF risk. We were also able to recapitulate the AF association using directly measured thyrotropin levels that fell within the normal ranges. Thus, the clinical decision to treat subclinical thyroid disease should incorporate the risk for AF, as antithyroid medications to treat hyperthyroidism may reduce AF risk and thyroid hormone replacement for hypothyroidism may increase AF risk. Interestingly, it has also been shown recently that thyrotropin may have direct electrophysiological effects in rat ventricular myocytes,30 although the precise mechanisms for the association are not well defined.

This study also shows the power of PheWAS to associate clinically relevant candidate diagnoses with biomarker variation. Genetic-based association approaches used in PheWAS are effective because heritable genetic variation represents a lifelong exposure to disease risk. Thus, even modest genetic perturbations in homeostatic levels of a causative biomarker may be detectable given the long duration of exposure. Moreover, the AF meta-analysis, also known as mendelian randomization, avoids some of the limitations of observational studies. In mendelian randomization, since genotypes are not affected by disease status, any reverse causation bias should be avoided. These analyses also highlight the use of coupling polygenic predictors, which better estimate genetic risk as compared with individual SNVs,31 with PheWAS to enable virtual biomarker studies whereby a biomarker is measured in 1 population and then imputed into and evaluated in a larger, more deeply phenotyped population. Such an approach can vastly increase sample sizes and the number of diseases that can be evaluated for a given biomarker. The use of polygenic predictors also increases power. As an example, while no single SNV has been associated with high effect size explaining drug-induced QT prolongation or torsade de pointes in GWAS studies,32 Strauss et al31 demonstrated that a genetic QT score comprising 61 common SNVs explained a significant proportion (approximately 20% to 30%) of the variability in drug-induced QT prolongation and was a significant predictor of drug-induced torsade de pointes.

Limitations

There are limitations to this study. Phenome-wide association study phenotypes are derived from EHR billing codes and may be underascertained or false positives, which can attenuate the strength of PheWAS associations but also result in missed diagnoses of overt thyroid disease. While the positive predictive value of the AF phenotype was high, we validated the association in an independent, well-adjudicated data set to ensure the reproducibility of our discovery analyses. Some individuals may have been taking medications such as thyroid hormone replacement, which could bias or attenuate results; however, the strength of the association was preserved even after individuals with diagnosed thyroid disorders were dropped from the analysis. The SNVs used in the thyrotropin predictors in the study and replication populations were derived predominantly from individuals of European ancestry. Development and validation of predictors in other ancestries are needed to validate the robustness of the findings in these groups. The finding of an association even in the absence of overt thyroid disease reinforces the potential role for genetically determined variation in thyroid function within a physiologically accepted normal range as a risk factor for this increasingly common arrhythmia. Further prospective studies are required to assess whether thyrotropin polygenic predictors increase the predictive accuracy in AF risk assessment. We considered examining the association between FT4 or FT3 and AF, but existing published data did not allow us to construct robust genetic predictors. Studies examining the relationship between genetically determined levels of these hormones and AF risk will be informative to more fully characterize how genetic regulation of the thyroid hormone endocrine axis affects AF risk.

Conclusions

Applying a genetic predictor of thyrotropin levels to an EHR cohort (n = 37 154) identified expected associations with thyroid disorders as well as a significant inverse association with AF, even after dropping cases of overt thyroid disease. This finding was replicated in an AF GWAS study (17 931 AF cases and 115 142 controls) and epidemiologically using directly measured thyrotropin levels and suggests a role for genetically determined variation in thyroid function within a physiologically accepted normal range as a risk factor for this increasingly common arrhythmia.

Back to top
Article Information

Corresponding Author: Joe-Elie Salem, MD, PhD, Hospital Pitié-Salpétrière, Department of Clinical Pharmacology, Boulevard de l’hôpital, 75013 Paris, France (joe-elie.salem@aphp.fr).

Accepted for Publication: November 5, 2018.

Published Online: January 23, 2019. doi:10.1001/jamacardio.2018.4615

Author Contributions: Dr Mosley had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Salem, Glazer, Knollmann, Denny, Roden, Mosley.

Acquisition, analysis, or interpretation of data: Salem, Shoemaker, Bastarache, Shaffer, Glazer, Kroncke, Wells, Straub, Shi, Jarvik, Larson, Velez Edwards, Edwards, Davis, Hakonarson, Weng, Fasel, Wang, Ellinor, Denny, Mosley.

Drafting of the manuscript: Salem, Shi.

Critical revision of the manuscript for important intellectual content: Salem, Shoemaker, Bastarache, Shaffer, Glazer, Kroncke, Wells, Straub, Jarvik, Larson, Velez Edwards, Edwards, Davis, Hakonarson, Weng, Fasel, Knollmann, Wang, Ellinor, Denny, Roden, Mosley.

Statistical analysis: Shaffer, Glazer, Shi, Davis, Mosley.

Obtained funding: Jarvik, Larson, Ellinor, Denny, Roden.

Administrative, technical, or material support: Shoemaker, Wells, Straub, Jarvik, Larson, Edwards, Hakonarson, Knollmann, Ellinor, Denny, Roden, Mosley.

Supervision: Salem, Hakonarson, Denny, Roden, Mosley.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Salem reported grants from Fondation Recherche Medicale (SPE20170336816) during the conduct of the study. Dr Shoemaker reported a grant from the National Institutes of Health (grant K23HL127704)Dr Larson reported grants from National Institutes of Health/National Human Genome Research Institute during the conduct of the study. Dr Davis reported grants from Vanderbilt University Medical Center during the conduct of the study. Dr Ellinor reported grants from Bayer AG to the Broad Institute focused on the genetics and therapeutics of atrial fibrillation, personal fees from Novartis, and personal fees from Quest Diagnostics outside the submitted work. Dr Denny reported grants from National Institutes of Health during the conduct of the study. No other disclosures were reported.

Funding/Support: This study was supported by FRM (grant SPE20170336816, Dr Salem) and the American Heart Association (grant 16FTF30130005, Dr Mosley). Vanderbilt University Medical Center’s BioVU projects are supported by numerous sources: institutional funding, private agencies, and federal grants. These include the National Institutes of Health–funded Shared Instrumentation Grant S10RR025141 and Clinical and Translational Science Awards grants UL1TR002243, UL1TR000445, and UL1RR024975. Genomic data are also supported by investigator-led projects that include U01HG004798, R01LM010685, R01NS032830, RC2GM092618, P50GM115305, U01HG006378, U19HL065962, and R01HD074711 and the following additional funding sources, accurate as of publication date: NIH K07 CA172294, AHA 14GRNT20460090, NIH P01 DK038226; NIH R24 DK 96527, NIH U01 HG004798, NIH R01 LM010685, NIH R01 NS032830, NIH R01 EY012118, NIH K12 HD043483, NIH R01 DK078616, NIH RC2 GM092618, NHMRC APP1064524, NIH R01 CA162433, NIH P01 HL056693, NIH P50 GM115305, NIH U01 HG006378, NIH U19 HL065962, NIH U01 HG004603, PCORI (private), NIH P50 CA09813, NIH R01 HD074711, NIH R03 HD078567, NIH R01 DK080007, and NIH P50 HL081009 (an updated list can be found at https://victr.vanderbilt.edu/pub/biovu/?sid=229). The eMERGE (Electronic Medical Records and Genetics) Network was initiated and funded by National Human Genome Research Institute through grants from Children’s Hospital of Philadelphia (grant U01HG006830); Essentia Institute of Rural Health, Marshfield Clinic Research Foundation, and Pennsylvania State University (grant U01HG006389); Geisinger Clinic (grant U01HG006382); Kaiser Permanente/University of Washington (grant U01HG006375); Mayo Clinic (grant U01HG006379); Icahn School of Medicine at Mount Sinai (grant U01HG006380); Northwestern University (grant U01HG006388); Brigham and Women’s Hospital (grants U01HG006378 and U01HG8685); Vanderbilt University Medical Center (grant U01HG8672); and Vanderbilt University Medical Center serving as the Coordinating Center (grant U01HG006385); Center for Inherited Disease Research (grant U01HG004438) and the Broad Institute (grant U01HG004424) serve as genotyping centers. Dr Ellinor is supported by National Institutes of Health awards (grants 1RO1HL092577, R01HL128914, and K24HL105780) and by the Fondation Leducq (grant 14CVD01).

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We wish to acknowledge the expert technical support of the VANTAGE and VANGARD core facilities, supported in part by the Vanderbilt-Ingram Cancer Center and Vanderbilt Vision Center. We also wish to acknowledge Sara Van Driest, MD, PhD, Vanderbilt University Medical Center, for her editorial input. Dr Van Driest was not compensated outside of her standard salary.

References
1.
De Leo  S, Lee  SY, Braverman  LE.  Hyperthyroidism.  Lancet. 2016;388(10047):906-918. doi:10.1016/S0140-6736(16)00278-6PubMedGoogle ScholarCrossref
2.
Ryödi  E, Salmi  J, Jaatinen  P,  et al.  Cardiovascular morbidity and mortality in surgically treated hyperthyroidism: a nation-wide cohort study with a long-term follow-up.  Clin Endocrinol (Oxf). 2014;80(5):743-750. doi:10.1111/cen.12359PubMedGoogle ScholarCrossref
3.
Porcu  E, Medici  M, Pistis  G,  et al.  A meta-analysis of thyroid-related traits reveals novel loci and gender-specific differences in the regulation of thyroid function.  PLoS Genet. 2013;9(2):e1003266. doi:10.1371/journal.pgen.1003266PubMedGoogle ScholarCrossref
4.
Gudmundsson  J, Sulem  P, Gudbjartsson  DF,  et al.  Discovery of common variants associated with low TSH levels and thyroid cancer risk.  Nat Genet. 2012;44(3):319-322. doi:10.1038/ng.1046PubMedGoogle ScholarCrossref
5.
Denny  JC, Ritchie  MD, Basford  MA,  et al.  PheWAS: demonstrating the feasibility of a phenome-wide scan to discover gene-disease associations.  Bioinformatics. 2010;26(9):1205-1210. doi:10.1093/bioinformatics/btq126PubMedGoogle ScholarCrossref
6.
Denny  JC, Bastarache  L, Ritchie  MD,  et al.  Systematic comparison of phenome-wide association study of electronic medical record data and genome-wide association study data.  Nat Biotechnol. 2013;31(12):1102-1110. doi:10.1038/nbt.2749PubMedGoogle ScholarCrossref
7.
Gottesman  O, Kuivaniemi  H, Tromp  G,  et al; eMERGE Network.  The Electronic Medical Records and Genomics (eMERGE) Network: past, present, and future.  Genet Med. 2013;15(10):761-771. doi:10.1038/gim.2013.72PubMedGoogle ScholarCrossref
8.
Carroll  RJ, Bastarache  L, Denny  JCR.  R PheWAS: data analysis and plotting tools for phenome-wide association studies in the R environment.  Bioinformatics. 2014;30(16):2375-2376. doi:10.1093/bioinformatics/btu197PubMedGoogle ScholarCrossref
9.
Roden  DM, Pulley  JM, Basford  MA,  et al.  Development of a large-scale de-identified DNA biobank to enable personalized medicine.  Clin Pharmacol Ther. 2008;84(3):362-369. doi:10.1038/clpt.2008.89PubMedGoogle ScholarCrossref
10.
Mosley  JD, Feng  Q, Wells  QS,  et al.  A study paradigm integrating prospective epidemiologic cohorts and electronic health records to identify disease biomarkers.  Nat Commun. 2018;9(1):3522. doi:10.1038/s41467-018-05624-4PubMedGoogle ScholarCrossref
11.
Kirby  JC, Speltz  P, Rasmussen  LV,  et al.  PheKB: a catalog and workflow for creating electronic phenotype algorithms for transportability.  J Am Med Inform Assoc. 2016;23(6):1046-1052. doi:10.1093/jamia/ocv202PubMedGoogle ScholarCrossref
12.
Denny  JC, Crawford  DC, Ritchie  MD,  et al.  Variants near FOXE1 are associated with hypothyroidism and other thyroid conditions: using electronic medical records for genome- and phenome-wide studies.  Am J Hum Genet. 2011;89(4):529-542. doi:10.1016/j.ajhg.2011.09.008PubMedGoogle ScholarCrossref
13.
Kho  AN, Cashy  JP, Jackson  KL,  et al.  Design and implementation of a privacy preserving electronic health record linkage tool in Chicago.  J Am Med Inform Assoc. 2015;22(5):1072-1080. doi:10.1093/jamia/ocv038PubMedGoogle ScholarCrossref
14.
Christophersen  IE, Rienstra  M, Roselli  C,  et al; METASTROKE Consortium of the ISGC; Neurology Working Group of the CHARGE Consortium; AFGen Consortium.  Large-scale analyses of common and rare variants identify 12 new loci associated with atrial fibrillation.  Nat Genet. 2017;49(6):946-952. doi:10.1038/ng.3843PubMedGoogle ScholarCrossref
15.
Mosley  JD, van Driest  SL, Wells  QS,  et al.  Defining a contemporary ischemic heart disease genetic risk profile using historical data.  Circ Cardiovasc Genet. 2016;9(6):521-530. doi:10.1161/CIRCGENETICS.116.001530PubMedGoogle ScholarCrossref
16.
Zuvich  RL, Armstrong  LL, Bielinski  SJ,  et al.  Pitfalls of merging GWAS data: lessons learned in the eMERGE network and quality control procedures to maintain high data quality.  Genet Epidemiol. 2011;35(8):887-898. doi:10.1002/gepi.20639PubMedGoogle ScholarCrossref
17.
Delaneau  O, Zagury  JF, Marchini  J.  Improved whole-chromosome phasing for disease and population genetic studies.  Nat Methods. 2013;10(1):5-6. doi:10.1038/nmeth.2307PubMedGoogle ScholarCrossref
18.
Howie  B, Fuchsberger  C, Stephens  M, Marchini  J, Abecasis  GR.  Fast and accurate genotype imputation in genome-wide association studies through pre-phasing.  Nat Genet. 2012;44(8):955-959. doi:10.1038/ng.2354PubMedGoogle ScholarCrossref
19.
Yang  J, Lee  SH, Goddard  ME, Visscher  PM.  GCTA: a tool for genome-wide complex trait analysis.  Am J Hum Genet. 2011;88(1):76-82. doi:10.1016/j.ajhg.2010.11.011PubMedGoogle ScholarCrossref
20.
Yang  J, Manolio  TA, Pasquale  LR,  et al.  Genome partitioning of genetic variation for complex traits using common SNPs.  Nat Genet. 2011;43(6):519-525. doi:10.1038/ng.823PubMedGoogle ScholarCrossref
21.
Purcell  SM, Wray  NR, Stone  JL,  et al; International Schizophrenia Consortium.  Common polygenic variation contributes to risk of schizophrenia and bipolar disorder.  Nature. 2009;460(7256):748-752.PubMedGoogle Scholar
22.
Burgess  S, Butterworth  A, Thompson  SG.  Mendelian randomization analysis with multiple genetic variants using summarized data.  Genet Epidemiol. 2013;37(7):658-665. doi:10.1002/gepi.21758PubMedGoogle ScholarCrossref
23.
Chugh  SS, Havmoeller  R, Narayanan  K,  et al.  Worldwide epidemiology of atrial fibrillation: a Global Burden of Disease 2010 Study.  Circulation. 2014;129(8):837-847. doi:10.1161/CIRCULATIONAHA.113.005119PubMedGoogle ScholarCrossref
24.
Salem  JE, El-Aissaoui  M, Alazard  M,  et al.  Modeling of amiodarone effect on heart rate control in critically ill patients with atrial tachyarrhythmias.  Clin Pharmacokinet. 2016;55(8):991-1002. doi:10.1007/s40262-016-0372-9PubMedGoogle ScholarCrossref
25.
Biondi  B, Bartalena  L, Cooper  DS, Hegedüs  L, Laurberg  P, Kahaly  GJ.  The 2015 European Thyroid Association guidelines on diagnosis and treatment of endogenous subclinical hyperthyroidism.  Eur Thyroid J. 2015;4(3):149-163. doi:10.1159/000438750PubMedGoogle ScholarCrossref
26.
Sawin  CT, Geller  A, Wolf  PA,  et al.  Low serum thyrotropin concentrations as a risk factor for atrial fibrillation in older persons.  N Engl J Med. 1994;331(19):1249-1252. doi:10.1056/NEJM199411103311901PubMedGoogle ScholarCrossref
27.
Cappola  AR, Fried  LP, Arnold  AM,  et al.  Thyroid status, cardiovascular risk, and mortality in older adults.  JAMA. 2006;295(9):1033-1041. doi:10.1001/jama.295.9.1033PubMedGoogle ScholarCrossref
28.
Selmer  C, Olesen  JB, Hansen  ML,  et al.  The spectrum of thyroid disease and risk of new onset atrial fibrillation: a large population cohort study.  BMJ. 2012;345:e7895. doi:10.1136/bmj.e7895PubMedGoogle ScholarCrossref
29.
Garber  JR, Cobin  RH, Gharib  H,  et al; American Association Of Clinical Endocrinologists And American Thyroid Association Taskforce On Hypothyroidism In Adults.  Clinical practice guidelines for hypothyroidism in adults: cosponsored by the American Association of Clinical Endocrinologists and the American Thyroid Association.  Thyroid. 2012;22(12):1200-1235. doi:10.1089/thy.2012.0205PubMedGoogle ScholarCrossref
30.
Alonso  H, Fernandez-Ruocco  J, Gallego  M,  et al Thyroid stimulating hormone directly modulates cardiac electrical activity.  J Mol Cell Cardiol. 2015;89(pt B):280-286. doi:10.1016/j.yjmcc.2015.10.019Google ScholarCrossref
31.
Strauss  DG, Vicente  J, Johannesen  L,  et al.  Common genetic variant risk score is associated with drug-induced QT prolongation and torsade de pointes risk: a pilot study.  Circulation. 2017;135(14):1300-1310. doi:10.1161/CIRCULATIONAHA.116.023980PubMedGoogle ScholarCrossref
32.
Salem  JE, Germain  M, Hulot  JS,  et al.  GENomE wide analysis of sotalol-induced IKr inhibition during ventricular REPOLarization, “GENEREPOL study”: lack of common variants with large effect sizes.  PLoS One. 2017;12(8):e0181875. doi:10.1371/journal.pone.0181875PubMedGoogle ScholarCrossref
×