0021-972X/01/$03.00/0
The Journal of Clinical Endocrinology & Metabolism
Copyright © 2001 by The Endocrine Society
Vol. 86, No. 2
Printed in U.S.A.
Evidence for a Major Role of Heredity in Graves’
Disease: A Population-Based Study of Two Danish
Twin Cohorts*
THOMAS HEIBERG BRIX, KIRSTEN OHM KYVIK, KAARE CHRISTENSEN,
LASZLO HEGEDÜS
AND
Department of Endocrinology, Odense University Hospital (T.H.B., L.H.); and The Danish Twin
Registry, Epidemiology, Institute of Public Health, University of Southern Denmark, Odense University
(T.H.B., K.O.K., K.C.), DK-5000 Odense, Denmark
ABSTRACT
The etiology of Graves’ disease (GD), affecting up to 2% of a population in iodine-sufficient areas, is incompletely understood. According to current thinking, the development of GD depends on complex
interactions among genetic, environmental, and endogenous factors.
However, the relative contributions of the genetic and environmental
factors remain to be clarified.
In this study we report probandwise concordance rates for GD in a
new cohort of same sex twin pairs born between 1953 and 1976 (young
cohort), ascertained from the nationwide population-based Danish Twin
Register. To elucidate the magnitude of the genetic and environmental
influence in the etiology of GD, these new twin data were pooled with our
previously published twin data on GD (old cohort). The old cohort consisted of 2338 same sex twin pairs born between 1870 and 1920 who had
participated in questionnaire surveys during the 1950s, 1960s, and
1970s. The young cohort included 6628 same sex twin pairs born between
1953 and 1976 who had participated in a questionnaire survey in 1994.
In the young cohort there were four monozygotic (MZ) pairs and
one dizygotic (DZ) pair concordant for clinically overt GD, giving an
overall probandwise concordance rate of 0.35 [95% confidence interval (CI), 0.16 – 0.57] for MZ pairs and 0.07 (95% CI, 0.01– 0.24)
for DZ pairs (P ⬍ 0.02). In the combined twin cohorts there were
eight MZ pairs and one DZ pair concordant for clinically overt GD,
giving a crude concordance rate of 0.35 (95% CI, 0.21– 0.50) for MZ
pairs and 0.03 (95% CI, 0.01– 0.12) for DZ pairs (P ⬍ 0.02). Modelfitting analysis on the pooled twin data showed that 79% of the
liability to the development of GD is attributable to genetic factors.
Individual specific environmental factors not shared by the twins
could explain the remaining 21%. In conclusion, our study strongly
supports the idea that genetic factors play a major role in the
etiology of GD and suggest that a further search for susceptibility
genes is worthwhile. (J Clin Endocrinol Metab 86: 930 –934, 2001)
G
RAVES’ DISEASE (GD) is an organ-specific autoimmune thyroid disorder characterized by hyperthyroidism, various degrees of diffuse goiter, ophthalmopathy,
and, less commonly, pretibial myxedema (1). The etiology of
GD is incompletely understood, but seems to involve complex interactions among genetic, environmental, and endogenous factors (1, 2). The importance of genetic factors is
evident from clustering of GD within families (3, 4), and the
limited twin data available suggest a higher concordance rate
for GD in monozygotic (MZ) than in dizygotic (DZ) twins (2,
5). In addition, recent studies indicate the presence of a
number of genes or genetic markers with association and or
linkage with GD (2, 6 – 8). However, the relative contributions of genetic and environmental influences on disease
susceptibility remain to be defined. This can be investigated
in twins (9).
There are only a few studies concerning the occurrence of
GD or thyrotoxicosis in unselected twin populations (5, 10,
11). The two early studies (10, 11) dealt only with selfreported nonverified diagnoses of thyrotoxicosis, hampering
any conclusions with respect to GD (2). In our recent population-based twin study (5) in which all of the probands had
well defined phenotypes, the concordance rates for GD were
much lower than those previously reported, but were still
significantly higher for MZ than DZ pairs. This finding of a
lower overall concordance rate for confirmed GD has led the
authors of recent reviews to imply that the genetic contribution to disease etiology might not be as strong as previously believed (8, 12). However, such simple calculations of
twin data give an erroneous assessment of the genetic contribution to disease development. In complex diseases, the
concordance rates are a function of both the prevalence of the
disease and the heritability (13). Thus, it follows that if the
prevalence is low the concordance rate in twins will also be
low even though there is a high degree of heritability. More
sophisticated analyses of twin data give more precise estimates of the magnitude of the genetic component in disease
susceptibility (14). In rheumatoid arthritis, for example, the
heritability has been estimated to be approximately 60% despite a concordance rate of only 12% in MZ twins (15). Unfortunately, it requires a large population-based twin study
within a genetically homogeneous population to obtain a
reliable estimate of the heritability, which has not been available for GD.
Received April 19, 2000. Revision received September 15, 2000. Accepted October 9, 2000.
Address all correspondence and requests for reprints to: Dr. Thomas
Heiberg Brix, The Danish Twin Register, Epidemiology, Institute of Public
Health, University of Southern Denmark, Odense University, Sdr. Boulevard 23A, DK-5000 Odense C, Denmark. E-mail: [email protected].
* This work was supported by grants from the Agnes and Knut Mørks
Foundation, the Dagmar Marshalls Foundation, the Novo Nordisk
Foundation Committee, the A. P. Møller and Hustru Chastine McKinney
Møllers Foundation, and the Clinical Research Institute, Odense
University.
930
ROLE OF HEREDITY IN GRAVES’ DISEASE
In this study we report concordance rates for GD in a new
cohort of Danish twins born between 1953 and 1976, ascertained from the nationwide population-based Danish Twin
Register. To estimate the magnitude of the genetic and environmental contributions to the etiology of GD, these new
twin data on GD were pooled with our previously published
twin data on GD (5).
Subjects and Methods
The twins were recruited from the Danish twin register. The ascertainment procedure, completeness, and validity of this nation-wide population-based register was described in detail previously (16). The
present study population was restricted to same sex twin pairs born
between 1870 and 1920 and between 1953 and 1976, named the old and
young cohorts, respectively. In both cohorts information on selfreported hyperthyroidism was obtained from nation-wide questionnaire surveys, and a diagnosis of GD was confirmed by information from
hospitals, out-patient clinics, specialists in endocrinology, and general
practitioners. All respondents were living in Denmark, which during the
last century has been regarded as a nonendemic goiter area. When
investigated, the median urinary iodine excretion has been between 70
and 100 ␮g/24 h (17).
Data for the old cohort have been published in detail previously (5).
The young cohort included 6628 same sex pairs (3556 female and 3072
male) who participated in a nation-wide questionnaire survey in 1994.
Three hundred and twelve subjects (201 females and 111 males) indicated present or previous hyperthyroidism. In 265 (85%) of these subjects
it was possible to obtain further information with respect to thyroid
disease from hospitals, out-patient clinics, specialists, or general practitioners. In 210 subjects the presence of previous or current hyperthyroidism was excluded; the main reasons were weight problems interpreted by the individual as hyperthyroidism (137 subjects), error when
filling out the questionnaire (36 subjects), or nontoxic thyroid disease (37
subjects). Thus, it follows that a total of 55 subjects with verified hyperthyroidism were identified in the young cohort. Of these, 5 subjects
had toxic nodular goiter and were therefore excluded from further
consideration. In the remaining 50 subjects (45 females and 5 males) the
hyperthyroidism was due to GD.
Overall, a total of 105 subjects with verified GD were identified from
the 2 cohorts. Fourteen of these 105 subjects were males (13%). Informed
consent was obtained from all of the participants, and the study was
approved by all regional scientific-ethical committees in Denmark (case
file 96/150 PMC).
Classification of the hyperthyroidism
All information from hospitals, out-patient clinics, specialists, and general practitioners was reviewed by two of us
(L.H. and T.H.B.), both blinded to the zygosity of the twins.
In the old cohort a diagnosis of GD was assigned on the basis
of clinical and histopathological evidence (5). In the young
cohort a diagnosis of GD was based on biochemical hyperthyroidism and a diffuse symmetrical goiter in combination
with either positive thyroid antibodies (thyroglobulin, thyroid peroxidase, or TSH receptor) and/or thyroid ophthalmopathy and/or diffuse hyperplasia on an isotope scan or
ultrasonography demonstrating homogenous echotexture.
Zygosity
In both cohorts determination of zygosity was primarily
based on self-reported answers to specific questions about
similarity and mistaken identity, which is a well established
and valid method (18). In addition, the zygosity of the four
concordant pairs from the young cohort was verified by
means of DNA typing of nine short tandem repeat systems
931
with the PE Applied Biosystems AmpFISTR Profiler Plus Kit
(19).
Analysis of data
We defined a proband as a subject with verified GD who
was ascertained through one of the questionnaire surveys
independently of disease status in the cotwin. The similarity
in MZ and DZ twins was assessed by probandwise concordance rates, which is defined as the proportion of affected
cotwins of probands. It gives the risk that a twin is affected
given that the cotwin is affected and is, thus, directly comparable to risk estimates reported for other relatives or in the
background population (20). Estimation of the 95% confidence intervals (CI) for the concordance rates were based on
the binomial distribution.
For a complex trait such as GD there is no simple method
by which to assess heritability (proportion of variance of a
disease attributable to additive genetic effects) (21). We analyzed the data by structural equation modeling for twin
data as described in detail by Neale and Cardon (22). In this
approach, the phenotypic variance of the liability to GD is
partitioned into genetic and environmental components. The
genetic variance may be due to additive (A) or dominant (D)
genetic influence. The environmental variance can be divided into variance due to common environmental factors
(C) shared by twins reared in the same family and variance
due to individual nonshared environmental factors (E). Five
different etiological models, ACE, ADE, AE, CE, and E, were
fitted to the data using the computer software Mx, programmed for analysis of categorical twin data (23). The DE
model is not taken into account, because it is biologically rare
to have genetic dominance in the absence of additive genetic
factors. The fit of each model was assessed by a ␹2 goodness
of fit test that tested the agreement between the observed and
the predicted statistics (a small ␹2 value and a high P value
indicate a good agreement between the model and the observed data). The goal in model fitting is to explain the
observed data as well as possible with as few parameters as
possible (parsimony). We used Akaike’s Information Criterion, which equals the ␹2 value minus twice the degrees of
freedom (24). The model with the lowest value of Akaike’s
Information Criterion reflects the best balance of goodness of
fit and parsimony.
Results
Concordance rates
Results regarding concordance rates are summarized in
Table 1. For comparison, the number of cases and the concordance rate for GD in the old cohort is also shown. In the
young cohort there were 4 MZ pairs (3 female) and 1 DZ pair
(female) concordant for clinically overt GD, giving a crude
probandwise concordance rate of 0.35 (95% CI, 0.16 – 0.57) for
MZ pairs and 0.07 (95% CI, 0.01– 0.24) for DZ pairs (P ⫽
0.015). When the twins from the young cohort were stratified
according to gender (females, 45 twin individuals; males, 5
twin individuals), the difference in the probandwise concordance rate between MZ and DZ pairs did not reach statistical
significance. However, among females there was a tendency
toward a higher concordance rate in MZ than in DZ pairs,
932
JCE & M • 2001
Vol. 86 • No. 2
BRIX ET AL.
TABLE 1. Concordance rates for Graves’ disease in two cohorts of Danish twins, according to zygosity
Cohort
Zygosity
No. of pairs
Concordant
Discordant
Both unaffected
Probandwise
concordance rate
Old
MZ
DZ
4
0
14
33
919
1368
0.36 (0.17– 0.59)a
0.00 (0.00 – 0.11)
Young
MZ
DZ
4
1
15
25
2952
3631
0.35 (0.16 – 0.57)a
0.07 (0.01– 0.24)
Combined
MZ
DZ
8
1
29
58
3871
4999
0.35 (0.21– 0.50)a
0.03 (0.01– 0.12)
Values in parentheses represent 95% confidence intervals.
a
Monozygotic vs. dizygotic, P ⬍ 0.02.
although this was not, strictly speaking, statistically significant (MZ vs. DZ, 0.29 vs. 0.08; P ⫽ 0.08). Due to the low
number of affected male subjects, no tests of the difference
between male MZ and DZ pairs has been performed. Concordance rates for the combined twin cohort are shown in
Table 1.
Discordance time, length of follow-up, and thyroid status of
the cotwins to twins with GD
Among the 4 concordant MZ pairs in the young cohort, the
time from the diagnosis of GD in the first affected twin until
diagnosis in the cotwin was less than 4 yr (1, 1, 2, and 4 yr,
respectively). The discordance time in the concordant DZ
pair was 1 yr. Regardless of zygosity and gender, the overall
mean follow-up time of the healthy cotwins was 10.3 yr
(range, 3–24 yr). The mean follow-up time of the proband’s
cotwin (time span between disease onset in the proband and
last contact with the cotwin) was similar in MZ and DZ
cotwins (MZ vs. DZ, 10.8 vs. 10.1 yr; P ⫽ 0.78). In 4 of the 40
discordant pairs (2 MZ and 2 DZ pairs), the cotwin of the twin
with GD had a history of goiter. In all 4 cases the subject was,
according to their general practitioner, biochemically and
clinically euthyroid. Data for the old cohort have been published previously (5).
Model fitting and heritability
The outcome of the etiological modeling for the pooled
twin data (old plus young cohorts) is summarized in Table
2. Models that only included environmental factors (CE and
E models) provided a poor fit to the data (P ⬍ 0.05) and could
be excluded. The three models including both genetic and
environmental effects (ACE, ADE, and AE) all explained the
observed data statistically well. However, according to the
AIC (24), the AE model provided the best overall fit for the
data. In this model 79% of the liability to clinically overt GD
is attributable to genetic factors, whereas individual-specific
environmental factors not shared by cotwins explain the
remaining 21% of the phenotypic variance. No cohort effect
was observed, as biometric modeling, allowing the variance
components (a2, d2, c2, and e2) to vary across the two cohorts,
did not change the overall results (data not shown).
Discussion
The goal of this population-based twin study was to confirm our previous results on GD and, if possible to investigate
the magnitude of genetic and environmental factors in the
etiology of GD. Clearly, the concordance rates for clinically
overt GD found in the present study in a new cohort of
Danish twins with well defined phenotypes are fully comparable with our previous results. Our findings (significantly
higher concordance rates in MZ than in DZ twins) are supported by preliminary results from a very recent populationbased twin study from California (25). In the latter study the
pairwise concordance rate for GD was estimated to approximately 0.17 in MZ and 0.02 in DZ twins. Thus, in all three
population-based twin studies using well defined phenotypes the concordance rate for GD was significantly higher
in MZ than in DZ twins, confirming the presence of genetic
factors in the etiology of GD.
Accepting, therefore, that there is a genetic component in
the etiology of GD, how large is its effect? In the model-fitting
analyses we found that a model that comprised only additive
genetic effects and nonshared unique environmental effects
had the best overall fit to the observed data. In this model (AE
model), approximately 80% of the liability to develop clinically overt GD is attributable to genetic factors, whereas
individual-specific environmental factors not shared by the
twins explain the remaining 20%. The AE model is a simple
one, and it does not include shared environmental effects (C),
suggesting that all familial resemblance for GD is due solely
to genetic factors. These findings, which were consistent
across two twin cohorts, not only confirm the presence of a
genetic component in the etiology of GD, but also suggest
that the magnitude of the genetic contribution in GD may be
at the same level as that seen in other common autoimmune
diseases, such as rheumatoid arthritis and insulin-dependent
diabetes mellitus (15, 26).
The results of this study should be interpreted in the context of potential limitations. The data in this study were
obtained from Caucasians living in Denmark, among whom
cultural background and living conditions are generally homogeneous. Thus, the results of this study cannot uncritically
be extrapolated to other groups or populations. Heritability
estimates are specific to the population in which they were
estimated. Other populations may differ in genetic or environmental variance, or both, and hence the ratio of genetic to
total phenotypic variance will differ too. It is also important
to point out that due to inadequacy of power, the heritability
estimates based on the genetic modeling used here do not
take a possible gene-environment interaction into account.
Clearly, the prevalence of GD changes with variations in
environmental factors such as iodine intake (27) and smoking
(28, 29). Thus, gene-environment interactions may be im-
ROLE OF HEREDITY IN GRAVES’ DISEASE
933
TABLE 2. Model fitting on Graves’ disease, estimated by means of structural equation modeling
Model
ACE
ADE
AE
CE
E
Goodness of fit tests
Components of variance
␹2 (df)
P
AIC
a2
d2
0.76 (2)
0.03 (2)
0.76 (3)
11.5 (3)
58.5
0.69
0.98
0.86
0.01
⬍0.01
⫺3.2
⫺3.9
⫺5.2
5.6
50
0.79 (0.38 – 0.90)
0.24 (0.00 – 0.89)
0.79 (0.64 – 0.90)
0.57 (0.00 – 0.91)
c2
e2
0.00 (0.00 – 0.37)
0.21 (0.10 – 0.37)
0.19 (0.09 – 0.35)
0.21 (0.10 – 0.36)
0.38 (0.26 – 0.54)
1.00
0.62 (0.46 – 0.74)
Note, a small ␹2 value and a high P value indicate a good agreement between the model and the observed data. AIC, Akaike’s Information
Criterion (24). Values in parentheses are likelihood-based 95% confidence intervals. a2, Proportion of variance in liability to GD due to additive
genetic factors, equals heritability; d2, proportion of variance in liability to GD due to dominant genetic factors; c2, proportion of variance in
liability to GD due to shared environmental factors; e2, proportion of variance in liability to GD due to individual-specific (unique) environmental
factors.
portant in GD, making the results of heritability difficult to
interpret.
In addition, one of the basic assumptions of the classical
twin study, that MZ twins carry identical genes, may not be
fulfilled in twin studies of immune-mediated diseases, such
as GD (2, 30). The generation of Igs and T cell receptors by
somatic mutations of germline genes and rearrangements
that take place during the differentiation of the immune
system are a potential source of difference between the two
twin individuals in a given MZ pair (2, 15). Indeed, there is
now evidence for a bias in the use of T cell receptor V genes
by T cells involved in autoimmune diseases, including autoimmune thyroid disease (30). Therefore, the classic twin
study will tend to underestimate the magnitude of the genetic component in GD as well as in other autoimmune
diseases.
Although the Danish Twin Register is population based,
our final study sample is unlikely to be completely representative of the entire twin population. Twins who did not
answer/return the questionnaires were not included. However, as we used the probandwise concordance rate, it is not
crucial that all twins should be studied as long as there is no
systematic bias in the ascertainment procedure (20). The information on the presence or absence of GD was based on
self-reports. This may induce a bias, because of problems in
recall or validity of the questionnaire. However, the diagnosis was confirmed by review of medical records. Additionally, in the young cohort, a record linkage between the
Twin Register and the National Discharge Register did not
suggest systematic under reporting of thyroid disease.
In the young cohort, the follow-up period of the healthy
cotwins was rather short. That is, not much time was allowed
for the cotwin to develop GD, and hence the concordance
rates may increase with increasing follow-up time. This is,
however, a very complex process, because new discordant
pairs may also appear with increasing observation time.
Moreover, it seems unlikely that such a potential increase in
concordance should be restricted to DZ compared with MZ
twins. This is supported by the fact that the concordance rates
found in the young cohort were nearly identical to those
found in our previously study, in which there was 25-yr
follow-up of the cotwins (5).
In twin studies it is assumed that the intrapair difference
in environment is the same for MZ and DZ twins. It has,
however, been argued that twins in MZ pairs are more likely
to share more similar environments than twins in DZ pairs
(31). This could result in an overestimation of the genetic
influence in disease etiology. In our study the ADE and AE
models fitted the observed data better than the model that
included shared environmental effects (ACE model). On the
other hand, this may be due to inadequacy of power to detect
a small to modest effect of shared environment in this study.
Therefore, we can only conclude that we were unable to
identify major shared environmental influences on the risk
for GD.
In conclusion, our study suggests a multifactorial etiology
of GD among Caucasians living in areas with borderline
iodine deficiency. It supports a major etiological role for
genetic factors in the development of GD, but also indicates
that environmental factors are of importance. It is likely that
genetic predisposition (susceptibility) is necessary for the
development of GD, and that environmental factors possibly
have a predominant role in controlling whether a genetically
predisposed subject progresses to clinically overt GD.
References
1. Weetman AP, McGregor AM. 1994 Autoimmune thyroid disease: further
developments in our understanding. Endocr Rev. 15:788 – 830.
2. Brix TH, Kyvik KO, Hegedüs L. 1998 What is the evidence of genetic factors
in the etiology of Graves’ disease? A brief review Thyroid. 8:627– 634.
3. Bartels ED. 1941 Heredity in Graves‘ disease. With remarks on heredity in toxic
adenoma in the thyroid, non-toxic goitre, and myxoedema. PhD Thesis, Einar
Munksgaard, Copenhagen.
4. Stenszky V, Kozma L, Balazs C, Rochlitz S, Bear JC, Farid NR. 5. The genetics
of Graves’ disease: HLA and disease susceptibility. J Clin Endocrinol Metab.
61:735–740.
5. Brix TH, Christensen K, Holm NV, Harvald B, Hegedüs L.1998. A population
based study of Graves’ disease in Danish twins. Clin Endocrinol (Oxf).
48:397– 400.
6. Tomer Y, Barbesino G, Greenberg DA, Concepcion E, Davies TF. 1999 Mapping the major susceptibility loci for familial Graves’ and Hashimoto’s diseases: evidence for genetic heterogeneity and gene interactions. J Clin Endocrinol Metab. 84:4656 – 4664.
7. Vaidya B, Imrie H, Perros P, et al. 1999 The cytotoxic T lymphocyte antigen-4
is a major Graves’ disease locus. Hum Mol Genet. 8:1195–1199.
8. Allahabadia A, Gough SCL. 1999 The different approaches to the genetic
analysis of autoimmune thyroid disease. J Endocrinol. 163:7–13.
9. Hawkes CH. 1997 Twin studies in medicine: what do they tell us ? Q J Med.
90:311–321.
10. Harvald B, Hauge M. 1996 A catamnestic investigation of Danish twins: a
preliminary report. Dan Med Bull. 3:150 –158.
11. Verschuer OV. 1998 Die zwillingsforschung im dienste der inneren medizin.
Verh Dtsch Ges Med. 64:262–273.
12. Philippou G, McGregor AM. 1998 The aetiology of Graves’ disease: what is
the genetic contribution? Clin Endocrinol (Oxf). 48:393–395.
13. Smith C. 1994 Concordance in twins: methods and interpretation. Am J Hum
Genet. 26:454 – 466.
14. Martin N, Boomsma D, Machin G. 1997 A twin-pronged attack on complex
traits. Nat Genet. 17:387–392.
15. Worthington J, Silman AJ. 1995 Genetic control of autoimmunity, lessons from
twin studies. Clin Exp Immunol. 101:390 –392.
934
BRIX ET AL.
16. Kyvik KO, Christensen K, Skytthe A, Harvald B, Holm NV. 1996 The Danish
twin register. Dan Med Bull. 43:467– 470.
17. Rasmussen LB, Andersson G, Haraldsdóttir J, et al. 1996 Iodine. Do we need
an enrichment program in Denmark? Int J Food Sci Nutr. 47:377–381.
18. Hauge M. 1981 The Danish twin registre. In: Mednick SA, Baert AE, Bachmann
BP, eds. Prospective longitudinal research. An empirical basis for the primary
prevention of psychosocial disorders. Oxford: Oxford University Press;
217–221.
19. Perkin-Elmer Corp. 1997 User’s manual. Foster City: Perkin-Elmer Corp.
20. McGue M. 1992 When assessing twin concordance, use the probandwise not
the pairwise rate. Schizophrenia Bull. 18:171–176.
21. Gao XJ, Klar N, Donner A. 1997 Comparison of methods for analyzing binary
data arising from two-sample twin studies. Gen Epidemiol. 14:349 –363.
22. Neale MC, Cardon LR. 1992 Methodology for genetic studies of twins and
families. Dordrecht: Kluwer.
23. Neale MC. 1997 MX: statistical modeling. Richmond: University of Richmond;
vol 4, 1–159.
24. Akaike H. 1987 Factor analysis and AIC. Psychometrika. 52:317–332.
25. Ringold DA, Nicoloff JT, Kesler H, Davis H, Mack T. Graves’ disease,
26.
27.
28.
29.
30.
31.
JCE & M • 2001
Vol. 86 • No. 2
genetics and environment: revised estimates of the concordance rates in
monozygotic and dizygotic twin populations [Abstract 13]. Proc of the 72nd
Annual Meet of the Am Thyroid Assoc. 1999.
Kyvik KO, Green A, Beck-Nielsen H.1995. Concordance rates of insulin
dependent diabetes mellitus: a population based study of young Danish twins.
Br Med J. 311:913–917.
Vanderpump MPJ, Tunbridge WMG.1996. The epidemiology of thyroid diseases. In: Braverman LE, Utiger RD, eds. Werner and Ingbar’s the thyroid.
Philadelphia: Lippincott-Raven; 474 – 482.
Bertelsen JB, Hegedüs L. 1994 Cigarette smoking and the thyroid. Thyroid.
4:327–331.
Brix TH, Hansen PS, Kyvik KO, Hegedüs L. 2000 Cigarette smoking and risk
of clinically overt thyroid disease: a population based twin case-control study.
Arch Intern Med. 160:661– 666.
Davies TF. 1998 Autoimmune tjyroid disease genes come in many styles and
colors. J Clin Endocrinol Metab. 83:3391–3393.
Hrubec Z, Robinette CD. 1984 The study of human twins in medical research.
N Engl J Med. 310:435– 441.