THYROID

Volume 12, Number 5, 2002

© Mary Ann Liebert, Inc.

Fractures in Patients with Hyperthyroidism

and Hypothyroidism: A Nationwide Follow-Up
Study in 16,249 Patients

Peter Vestergaard1,2 and Leif Mosekilde1,2

Aim: To study fracture risk in patients with hyperthyroidism and hypothyroidism. Subjects and methods: All pa-
tients with hyperthyroidism or autoimmune hypothyroidism diagnosed for the first time between 1983 and
1996 in Denmark were identified through the National Patient Discharge Register. Each patient was compared
with three age- and gender-matched controls randomly selected from the general population. Fracture occur-
rence before and after diagnosis was compared between patients and controls. Results: 11,776 patients with hy-
perthyroidism (6301 patients with diffuse toxic goiter, mean age, 52.1 6 18.6 years, and 5475 with nodular toxic
goiter, mean age, 60.4 6 15.9 years), and 4473 patients with hypothyroidism (mean age, 66.1 6 17.3) were iden-
tified. In patients with hyperthyroidism, fracture risk was only significantly increased around the time of di-
agnosis (incidence rate ratio [IRR] between 1.26 and 2.29), but decreased to normal levels after diagnosis. Sur-
gical treatment of hyperthyroidism was associated with a decreased fracture risk after diagnosis (RR 5 0.66,
95% confidence interval [CI]: 0.55–0.78). In hypothyroidism, fracture risk was significantly increased both be-
fore and after diagnosis with a peak around the time of diagnosis (IRR between 2.17 and 2.35). Conclusions:
Fracture risk is increased in hyperthyroidism and hypothyroidism. Thyroid surgery seems associated with a
decreased fracture risk in hyperthyroid patients.

Introduction risk did not seem to return to normal levels after diagnosis
and, thus, treatment (12). However, a decreased risk of foot

T HE BIOMECHANICAL COMPETENCE of bone depends on bone

mass, bone turnover, and bone structure. Excess thyroid
hormones are known stimulators of bone turnover (1,2). The
increased remodeling activity causes a reversible bone loss
because of an expansion of the remodeling space and an ir-
reversible loss because of a negative net bone balance and
eventually an increased risk of trabecular perforations (1). A
low bone mineral density (BMD) (3), an increased turnover
(4), and an increased number of trabecular perforations (1)
all increase the risk of fractures.
In untreated hyperthyroidism a decreased BMD has been
demonstrated (5–7). One year after treatment of hyperthy-
roidism, Tsai et al. (5) found a significant increase in BMD,
but the BMD did not reach normal levels. Langdahl et al.
(8,9) reported normal BMD in patients with hyperthyroidism
who had been euthyroid for many years.
Studies on fracture risk in patients with hyperthyroidism
have reported an increased risk of hip (10,11), forearm (12),
and spine fractures (12) after diagnosis (13), and the fracture
fractures has been reported (14).
In untreated hypothyroidism, histomorphometric studies
have disclosed a decreased bone turnover in trabecular as
well as cortical bone with an increased cortical thickness
(15,16). After substitution with levothyroxine, the hypothy-
roid state is reversed. This may result in an enhanced re-
modeling activity (16) to renew accumulated old bone. How-
ever, treatment may also lead to oversubstitution with
suppressed thyrotropin (TSH) levels. Both mechanisms may
be accompanied by increased turnover and decreased BMD
(2). Ribot at al. (17) found a loss of 5.4% in BMD in the spine
and 7% in the femoral neck after 1 year of substitution ther-
apy in 10 patients with hypothyroidism. However, during
the second year of substitution therapy, the BMD returned
to normal (18). Accordingly, BMD is usually normal in eu-
thyroid patients treated for hypothyroidism (19–21) but may
be decreased in the case of suppressed TSH after larger
dosages of levothyroxine (22–24).
In accordance with the reversible decreases in BMD, a tem-

1
Department of Endocrinology and Metabolism C, Aahus Amtssygehus, Aarhus University Hospital, Denmark.
2
Faculty of Health Sciences, University of Aarhus, Denmark.

411
412
porary increase in fracture risk within the first 2 years after
initiation of levothyroxine substitution therapy has been re-
ported in a questionnaire-based study (25).
Both BMD and bone turnover thus seem to return to nor-
mal in patients with hyperthyroidism as well as hypothy-
roidism on treatment (8,9,19). However, bone biomechanical
competence may still be altered because of persisting
changes in bone architecture (1).
The aim of the present study was to assess fracture risk in
a large population-based cohort of patients with hyperthy-
roidism and hypothyroidism at different time points before
and after diagnosis to evaluate if fracture risk returned to
normal after diagnosis and treatment.
Material and Methods
Patients
Denmark offers unique possibilities for this kind of study
(26) because of the extensive registration of diseases and
treatments. All patients diagnosed as new cases of hyper-
thyroidism (either hyperthyroidism with or without diffuse
toxic goiter: ICD8 codes 242.00, 242.08 or 242.09, and ICD10
code E05.0 or nodular toxic goiter: ICD8 code 242.19 and
ICD10 codes E05.1 and E05.2) or autoimmune hypothy-
roidism (ICD8 codes 244.03 and 244.09, and ICD10 code
E03.9) between January 1, 1983 and December 31, 1996 in
Denmark were included. Patients who had secondary hy-
pothyroidism were excluded. Patients were identified
through the National Patient Discharge Register, which is a
computer-based register of all admissions and discharges
from Danish hospitals (27) covering both diseases diagnoses
and surgical procedures. The registry covers inpatient ad-
missions until 1994 and outpatient and inpatient contacts
from 1995 and onwards. This did not limit the validity of the
data because most patients in Denmark are hospitalized
early in the course of their thyroid disease to undergo nec-
essary diagnostic tests and initiate therapy. On discharge the
patients are transferred to the outpatient clinics. A separate
analysis of the periods with inpatient and outpatient data
did not change the results in the following.
The register is used for monitoring disease occurrence and
for reimbursing hospital wards for treatments carried out.
Therefore, given that a patient was admitted for atrial fib-
rillation and it turns out that this was caused by hyperthy-
roidism, the discharge code 1 will be atrial fibrillation and
the discharge code 2 will be hyperthyroidism. The code of
hyperthyroidism will thus be linked to an admission start-
ing at say November 1, 1990 and ending at November 10,
1990, November 1 thus being the date of diagnosis. If the pa-
tient was later admitted again, November 1, 1990 would still
be the first date of diagnosis.
For insurance reasons, for example, even a fracture sus-
tained abroad is registered when the patient returns to the
country.
The register was founded in 1977 and is linked with other
registers through the nationwide Civil Registration Number.
This is a unique identifying code given to each inhabitant in
Denmark. On admission to a hospital, the dates of admis-
sion and discharge are recorded along with a diagnosis
(ICD8 or ICD10) for the actual admission. If surgical proce-
dures are performed, codes for these are also entered into
the register.
VESTERGAARD AND MOSEKILDE
For patients with hyperthyroidism, it was evaluated if sur-
gery to the thyroid had been performed (surgery codes: 0800,
08000, 0802, 08020, 0804, 08040, 0806, 08060, 0808, 08080,
0810, 08100, 0812, 08120, 0814, 08140, 0816, 08160, 0818,
08180, 0820, 08200, 0822, 08220, 0834, 08340, 08360, BAA00,
BAA05, BAA10, BAA20, BAA25, BAA30, BAA40, BAA50,
BAA60, or BAA99).
Patients who were not residing in Denmark were excluded
from the study. For each patient, the first date was recorded
when a diagnosis of a thyroid disorder had been made. The
observation period was subsequently divided into the pe-
riod before and after this date.
Controls
Each patient was compared with three age-, gender-, and
status-matched controls (35,300 for the hyperthyroid pa-
tients, and 13,410 for the patients with hypothyroidism). The
status-match included death or emigration. If a case subject
had died or emigrated, a control subject who had died or
emigrated on the same date or later was chosen. The Dan-
ish Ministry of the Interior performed this matching by stan-
dard procedures involving the civil registration number (30).
In a few cases it was not possible to find three matched con-
trols who had died or emigrated, and the patients were then
matched with one or two controls. This led to the loss of 28
controls (0.1%) among patients with hyperthyroidism, and 9
controls (0,1%) among patients with hypothyroidism. Each
control subject was assigned a “dummy date of diagnosis”
as the date when the corresponding case had his or her thy-
roid disorder diagnosed.
Validity of diagnoses
A screening of the files of 900 patients only revealed mis-
classification in less than 2% of the patients based on type of
diagnosis and surgical treatment. This is in accordance with
the general high validity of the register (27–29).
Changes in TSH caused by nonthyroidal illness (e.g., sick
euthyroid syndrome) did not lead to a diagnosis of hy-
pothyroidism in any of the patients screened. This is because
of the fact that such diseases are coded not as a thyroid dis-
ease but according to the extrathyroidal illness itself.
Outcome variable
For both patients and controls, information on any fracture
diagnosis (ICD8 codes 800.00–813.99, 820.00–824.99, and ICD10
codes S020–S02.9, S07.0–S07.9, S12.0–S12.9, S22.0–S22.9,
S32.0–S32.9, S42.0–S42.9, S52.0–S52.9, S72.0–S72.9, S82.0–S82.9)
registered between January 1, 1980 and December 31, 1996 was
then retrieved from the National Patient Discharge Register
under the Danish Board of Health (27).
In a random sample of 35 fracture diagnoses that were val-
idated, the precision of the diagnoses was 97% because one
case turned out to be a fracture of the proximal, and not the
distal forearm as stated in the diagnosis code.
DEXA scannings
Dual-energy x-ray absorptiometry (DEXA) scanning was
first introduced into general clinical practice in Denmark in
the mid-1990s. However, scanning results were available
from 216 patients (167 with hyperthyroidism and 49 with hy-
FRACTURES IN HYPERTHYROIDISM AND HYPOTHYROIDISM 413

pothyroidism) from one center (The Osteoporosis Clinic,
Aarhus Amtssygehus). They were all scanned several years
after diagnosis and treatment (Table 4). The patients were
scanned using a Hologic 1000/w, 2000, 4500 DEXA or a Nor-
land XR26 DEXA scanner. To account for the different scan-
ners, the BMD values were expressed as age-, and gender
adjusted Z scores.
Power
After diagnosis the power was 99% (2a 5 0.05) for demon-
strating an incidence rate ratio (IRR) of 1.15 in 11,776 patients
with hyperthyroidism versus 35,300 controls followed for a
mean of 6.4 years. Among the 4473 patients with hypothy-
roidism and 13,410 controls followed for a mean of 5.2 years,
the power was 88% for demonstrating an IRR of 1.15. Before
diagnosis, the power was higher because of the longer fol-
low-up time (Table 1). The study was thus well powered to
demonstrate even small increases in fracture risk and to an-
alyze subgroups.
Statistics
Mean and standard deviation (SD) were applied as de-
scriptive statistics. Numbers were compared by Mann-Whit-
ney test or x2 test for contingency tables where appropriate.
Incidence rates were calculated as number of subjects with
at least one fracture divided by the observation time.
Incidence rates were compared by IRR using Mantel-
Haenszel type x2 statistics. Relative risks were compared us-
ing Poisson regression. Risk factors were analyzed using Cox
regression or logistic regression (likelihood ratio method).
All calculations were performed using SPSS 6.1.3 for Win-
dows.
Results
Baseline characteristics
Table 1 shows baseline characteristics of the patients. Pa-
tients with nodular toxic goiter were significantly older than
patients with diffuse toxic goiter. Patients who later under-
went surgery were significantly younger at diagnosis than
those who did not (diffuse toxic goiter: 40.8 6 14.6 vs. 55.5 6
18.3 years, 2p , 0.01, nodular toxic goiter: 50.1 6 14.4 vs.
66.0 6 13.8, 2p , 0.01). Mean time from diagnosis to surgery
was 0.48 6 1.21 years in patients with hyperthyroidism who
underwent surgery.

TABLE 1. BASELINE CHARACTERISTICS : ACTUAL NUMBER (%), MEAN , AND SD

Group Parameter Women Men Combined P

Hyperthyroidism Patients (n) 10,126 (86%) 1,650 (14%) 11,776 —

Patient person years
Before diagnosis 97,194 15,917 113,112
After diagnosis 65,286 10,183 75,469
Controls (n) 30,354 (86%) 4946 (14%) 35,300 —
Controls person years —
Before diagnosis 291,347 47,696 339,042
After diagnosis 215,329 34,425 249,754

Diffuse toxic goiter Number (n) 5320 (53%) 981 (60%) 6301 (54%) —
Age at diagnosis (years) 52.1 6 18.7 52.3 6 18.0 52.1 6 18.6 ,0.73a
Thyroid surgery (n) 1235 (23%) 226 (23%) 1461 (23%) ,0.90b
Age at thyroid surgery (years) 41.7 6 14.3 43.9 6 15.1 42.0 6 14.4 ,0.04a
Fractures before diagnosis (n) 261 (4.9%) 45 (4.6%) 306 (4.9%) ,0.67b

Nodular toxic goiter Nodular toxic goiter (n) 4806 (48%) 669 (41%) 5475 (47%) —
Age at diagnosis (years) 60.4 6 16.1 60.7 6 14.6 60.4 6 15.9 ,0.65a
Thyroid surgery (n) 1669 (35%) 235 (35%) 1904 (35%) ,0.84b
Age at thyroid surgery (years) 50.5 6 14.5 53.2 6 13.4 50.9 6 14.4 ,0.01a
Fractures before diagnosis (n) 273 (5.7%) 20 (3.0%) 293 (5.4%) ,0.01b

Hypothyroidism Patients (n) 3872 (87%) 601 (13%) 4473 —

Patient person years —
Before diagnosis 36,813 5789 42,602
After diagnosis 20,387 2734 23,120
Controls (n) 11,610 (87%) 1800 (13%) 13,410 —
Controls person years —
Before diagnosis 110,398 17,345 127,743
After diagnosis 77,266 11,091 88,357
Age at diagnosis (years) 66.6 6 17.0 62.8 6 18.5 66.1 6 17.3 ,0.01a
Fractures before diagnosis 447 (12%) 30 (5%) 477 (11%) ,0.01b
a
Independent samples t test.
b 2
x test for contingency tables.
SD, standard deviation.
414 VESTERGAARD AND MOSEKILDE

Fracture risk nodular toxic goiter. Before diagnosis patients with diffuse
toxic goiter had a higher (2p , 0.05) relative risk of fractures
of the feet (IRR 5 1.92, 95% confidence interval [CI]:
Hyperthyroidism. Table 2 shows fracture risk in patients
1.04–3.55) than patients with nodular toxic goiter (IRR 5
compared to controls before and after diagnosis. Only pa-
0.75, 95% CI: 0.38–1.49, 2p 5 0.05). The same tendency was
tients with diffuse toxic goiter presented with a small, but
observed for fractures of the lower legs (IRR 5 1.00, CI:
significant increase in overall fracture risk before, but not af-
0.76–1.31 vs. IRR 5 0.62, 95% CI: 0.45–0.86, 2p 5 0.03).
ter diagnosis. The increase before diagnosis was linked to fe-
A comparison of patients with diffuse and nodular toxic goi-
mur and foot fractures. After diagnosis, there was a de-
ter demonstrated a lower fracture risk before diagnosis in pa-
creased risk of lower leg fractures compared to controls. The
tients with nodular compared to diffuse toxic goiter (Odds ra-
overall fracture risk decreased from before to after diagno-
tio: OR 5 0.82, 95% CI: 0.69–0.97) after adjustment for age,
sis in patients with diffuse toxic goiter (2p 5 0.03 by Poisson
gender, and subsequent thyroid surgery or not. After diagno-
regression; Table 2).
sis no difference was present between nodular and diffuse toxic
In patients with nodular toxic goiter no change in overall
goiter (hazard ratio: [HR] 5 1.08, 95% CI: 0.95–1.22).
fracture risk was present. However, there was an increase in
femur fractures both before and after diagnosis. Lower leg
Effects of surgery
fractures were less frequent than in normal controls before
diagnosis. Figure 1 shows the time-dependent variation in fracture
Only minor differences in skeletal localisation of fractions risk with a small peak around the time of diagnosis except
could be demonstrated between diffuse toxic goiter and in patients surgically treated for toxic nodular goiter.

TABLE 2. FRACTURE RISK IN PATIENTS COMPARED TO CONTROLS BEFORE AND AFTER DIAGNOSIS

IRR (95% CI) IRR (95% CI)

Disorder Fracture site Fractures before diagnosis Fractures before diagnosis pa

Diffuse All fracture types 306/785 1.17 (1.02–1.33)* 448/1502 0.97 (0.87–1.08) 0.03
toxic Skull and jaws 10/26 1.15 (0.56–2.39) 8/14 1.86 (0.79–4.37) 0.40
goiter Spine 20/51 1.18 (0.70–1.97) 45/99 1.48 (1.04–2.10)* 0.47
Upper arm 31/92 1.01 (0.67–1.52) 47/177 0.86 (0.63–1.19) 0.56
Forearm 62/150 1.24 (0.92–1.66) 95/347 0.89 (0.71–1.12) 0.08
Colles’ fracture 41/110 1.12 (0.78–1.60) 80/281 0.93 (0.72–1.19) 0.40
Hands 8/24 1.00 (0.45–2.22) 26/112 0.76 (0.49–1.16) 0.55
Femur 121/280 1.30 (1.05–1.60)* 197/610 1.05 (0.90–1.23) 0.13
Femoral neck 105/243 1.30 (1.03–1.63)* 179/547 1.07 (0.90–1.26) 0.18
Lower leg 70/210 1.00 (0.76–1.31) 52/234 0.72 (0.54–0.98)* 0.12
Ankle 37/134 0.83 (0.57–1.19) 37/170 0.71 (0.50–1.01) 0.55
Feet 16/25 1.92 (1.04–3.55)* 28/83 1.10 (0.72–1.69) 0.15
Nodular All fracture types 293/864 1.02 (0.89–1.16) 571/1817 1.06 (0.96–1.16) 0.62
toxic Skull and jaws 3/19 0.47 (0.14–1.56) 4/18 0.75 (0.25–2.21) 0.58
goiter Spine 25/67 1.12 (0.71–1.77) 43/139 1.04 (0.74–1.47) 0.81
Upper arm 29/92 0.94 (0.62–1.43) 61/243 0.85 (0.64–1.12) 0.67
Forearm 43/153 0.84 (0.60–1.18) 110/420 0.88 (0.72–1.09) 0.82
Colles’ fracture 40/115 1.04 (0.73–1.49) 95/253 0.91 (0.72–1.14) 0.52
Hands 11/11 3.00 (1.35–6.64)* 17/112 0.51 (0.31–0.84)* ,0.01
Femur 160/335 1.43 (1.19–1.73)* 307/852 1.21 (1.07–1.38)* 0.16
Femoral neck 146/303 1.44 (1.19–1.76)* 282/761 1.25 (1.09–1.43)* 0.23
Lower leg 43/208 0.62 (0.45–0.86)* 61/250 0.82 (0.62–1.09) 0.20
Ankle 24/144 0.50 (0.33–0.76)* 48/185 0.87 (0.64–1.20) 0.04
Feet 10/40 0.75 (0.38–1.49) 23/59 1.31 (0.81–2.12) 0.19
Hypo- All fracture types 477/1003 1.43 (1.28–1.59)* 611/1812 1.29 (1.18–1.41)* 0.14
thyroidism Skull and jaws 11/19 1.74 (0.83–3.61) 3/22 0.52 (0.16–1.70) 0.09
Spine 62/84 2.21 (1.61–3.05)* 63/124 1.94 (1.44–2.62)* 0.56
Upper arm 72/128 1.69 (1.27–2.24)* 101/230 1.68 (1.33–2.12)* 0.97
Forearm 75/170 1.32 (1.01–1.73)* 88/328 1.03 (0.81–1.30) 0.17
Colles’ fracture 60/127 1.42 (1.04–1.92)* 65/277 0.90 (0.68–1.17) 0.02
Hands 7/25 0.84 (0.36–1.94) 32/75 1.63 (1.08–2.46)* 0.16
Femur 216/446 1.45 (1.24–1.71)* 333/1039 1.22 (1.08–1.39)* 0.11
Femoral neck 191/391 1.46 (1.23–1.74)* 301/940 1.22 (1.07–1.39)* 0.05
Lower leg 95/209 1.36 (1.07–1.74)* 54/165 1.25 (0.92–1.70) 0.67
Ankle 55/128 1.29 (0.94–1.77) 37/121 1.19 (0.81–1.69) 0.74
Feet 8/22 1.09 (0.49–2.45) 20/50 1.53 (0.91–2.56) 0.49

*p , 0.05.
a
Comparison of the IRR before and after diagnosis by Poisson regression—if p , 0.05 the IRR decreased or increased significantly.
Incidence rate ratios (IRR) and 95% confidence intervals (95% CI). Number of subjects with at least one incident fracture.
FRACTURES IN HYPERTHYROIDISM AND HYPOTHYROIDISM 415

A B

C D

FIG. 1. Fracture risk in patients compared to controls before and after diagnosis stratified by goiter type and surgery. IRR,
incidence rate ratio. A: Diffuse toxic goiter, surgically treated. B: Diffuse toxic goiter, not surgically treated. C: Nodular
toxic goiter, surgically treated. D: Nodular toxic goiter, not surgically treated.

Before surgery, no significant difference in overall fracture
risk was present between those who subsequently had sur-
gery and those who did not (Table 3). However, after sur-
gery, fracture risk was lower among the surgically treated
than the nonsurgically treated after adjustment for age, gen-
der, and fractures before diagnosis. A fracture prior to di-
agnosis was a significant risk factor for a fracture after di-
agnosis (Table 3).
Hypothyroidism
Figure 2 shows fracture risk in hypothyroidism. Fracture
risk was increased up to 10 years prior to diagnosis with
a peak around the time of diagnosis. More than 5 years af-
ter diagnosis fracture risk returned to normal. The in-
creased fracture risk was present in most skeletal sites
(Table 2).
Female gender and age at diagnosis were significant risk
factors before and after diagnosis. Furthermore, previous
fractures increased the risk of new fractures.
Bone mineral measurements
Table 4 shows scanning results from the limited number
of patients who had DEXA scans several years after diag-
nosis and treatment. Patients with nodular toxic goiter
tended to have lower age-adjusted BMD (Z scores) than pa-
tients with diffuse toxic goiter. In general, Z scores in pa-
tients with diffuse toxic goiter were close to expected values
except for the femoral neck in those who did not have sur-
gery. Patients with nodular toxic goiter tended have reduced
Z scores.
In diffuse toxic goiter, those who underwent surgery
tended to have higher Z scores than those who did not. In
416 VESTERGAARD AND MOSEKILDE

TABLE 3. RISK FACTORS FOR FRACTURES BEFORE AND AFTER DIAGNOSIS

Time Disease Group Parameter Risk estimate (95% CI)

Beforea Hyperthyroidism Diffuse Thyroid surgery (yes/no) 0.81 (0.54–1.20)

Gender (male vs. female) 0.96 (0.69–1.33)
Age at diagnosis (years) 1.05 (1.04–1.06)*
Nodular Thyroid surgery (yes/no) 0.74 (0.51–1.06)
Gender (male vs. female) 0.53 (0.33–0.85)*
Age at diagnosis (years) 1.07 (1.05–1.08)*
Hypothyroidism All Gender (male vs. female) 0.44 (0.30–0.65)*
Age at diagnosis (years) 1.04 (1.03–1.05)*
Afterb Hyperthyroidism Diffuse Fracture before diagnosis (yes/no) 2.80 (2.09–3.75)*
Thyroid surgery (yes/no) 0.71 (0.54–0.95)*
Gender (male vs. female) 0.46 (0.32–0.66)*
Age at diagnosis (years) 1.06 (1.05–1.06)*
Nodular Fracture before diagnosis (yes/no) 1.88 (1.41–2.50)*
Thyroid surgery (yes/no) 0.63 (0.50–0.79)*
Gender (male vs. female) 0.70 (0.52–0.96)*
Age at diagnosis (years) 1.06 (1.05–1.07)*
Hypothyroidism All Fracture before diagnosis (yes/no) 2.01 (1.62–2.50)*
Gender (male vs. female) 0.62 (0.45–0.86)*
Age at diagnosis (years) 1.04 (1.03–1.05)*

*p , 0.05.
a
Logistic regression (likelihood ratio model), odds ratios (OR), and 95% confidence intervals (CI).
b
Multivariate Cox regression analysis, hazard ratio (HR) and 95% confidence intervals (CI).

nodular toxic goiter, no differences in Z score could be dem-
onstrated between those who underwent surgery and those
who did not.
In patients with hypothyroidism, the mean Z scores did
not deviate from that expected in the general population.
a decrease in fracture rate in hyperthyroidism. In the inter-
pretation of the results it should be noted that it is only pos-
sible reliably to identify the time of diagnosis, and that there
may have been a subclinical course before the time of diag-
nosis.

DISCUSSION
In a large population-based study, we have demonstrated
an increased risk of femur fractures in patients with hyper-
thyroidism and an overall increase in fracture risk in patients
with hypothyroidism. Thyroid surgery was associated with
FIG. 2. Fracture risk in patients with hypothyroidism com-
pared to controls. IRR, incidence rate ratio.
Hyperthyroidism
Fracture risk. The observed changes in fracture risk in hy-
perthyroidism are small and clearly need large study sam-
ples to document (10,11,14).
In the present study, the increase in hip fracture risk after
diagnosis (RR 5 1.17, 95% CI: 1.05–1.30) was moderate, and
within that reported by a previous questionnaire-based
study (RR 5 1.1, 95% CI: 0.4–3.4) (12). However, it appears
that the excess risk of hip fractures in hyperthyroidism in-
creases with age (10–12). The cohorts of Wejda et al. (OR 5
2.5, 95% CI: 1.2–5.3, mean age 76 years) (11) and Cummings
et al. (10) (OR 5 1.7, 95% CI: 1.2–2.5, mean age 72 years) were
significantly older than the actual study group, which may
explain the differences in risk estimates. The lack of any sig-
nificant effect on this skeletal site in the questionnaire-based
study (12) can thus be explained by the lower mean age (55
years) and lower number of participants.
The reduced risk of lower leg and ankle fractures before
and after diagnosis in the present study remains unex-
plained. Seeley et al. (14) reported a similar decrease in
foot and toe fractures (but not ankle fractures). However,
Seeley et al. (14) did not state the case-mix between dif-
fuse and nodular toxic goiter precluding a comparison
with our material, which demonstrated a difference be-
tween the two types of hyperthyroidism. Hyperthy-
roidism with nodular goiter is much more common in low
iodine intake areas such as Denmark (30) than in high io-
dine intake areas such as the United States (31), and this
FRACTURES IN HYPERTHYROIDISM AND HYPOTHYROIDISM 417

TABLE 4. RESULTS OF DEXA SCANNINGS

Age at scanning Time from Lumbar spine Femoral neck

Group n (years) diagnosis (years) Z score Z score

Hyperthyroidism
Diffuse toxic goiter 96 57.3 6 11.3 7.0 6 4.2 0.04 6 1.33 20.15 6 1.01*
Nodular toxic goiter 71 65.5 6 11.3 5.6 6 4.6 21.08 6 1.27* 21.04 6 0.94*
2pa — ,0.01 0.05 ,0.01 ,0.01
Diffuse toxic goiter
No surgery 71 58.9 6 11.4 6.5 6 4.1 20.10 6 1.25* 20.33 6 0.93*
Surgery 25 52.9 6 9.90 8.9 6 3.8 0.46 6 1.48 0.31 6 1.08
2pa 0.02 ,0.01 0.07 ,0.01
Nodular toxic goiter
No surgery 52 67.6 6 11.0 4.5 6 4.3 21.11 6 1.29* 21.11 6 1.00*
Surgery 19 59.8 6 10.3 8.7 6 3.9 21.00 6 1.23* 20.81 6 0.74*
2pa — ,0.01 ,0.01 0.74 0.15
Hypothyroidism 48 61.9 6 13.0 6.5 6 4.5 20.16 6 1.44* 20.31 6 1.39*
a
Independent samples t test (two-tailed p value). *2p , 0.05 for comparison with a Z score of 0.
Mean 6 SD expressed as age-adjusted Z scores.
DEXA, dual-energy x-ray absorptiometry.

difference may have affected the case-mix in the study of
Seeley et al. (14).
Furthermore, previous studies have disclosed that obesity
and weight gain are predictors of ankle fractures (14,32).
Hence, the reduction in body weight often seen in the hy-
perthyroid state may protect against ankle fractures. Other
possibilities would be that hyperthyroidism differentially in-
creases bone strength in the weight-bearing trabecular bone
in the ankle and foot.
The reason for the increased overall fracture risk before
diagnosis could be prolonged subclinical hyperthyroidism,
diagnostic delay, or a delay in registering diagnosis in the
hospital system. A prolonged subclinical course in nodular
toxic goiter is supported by Figure 1 (right lower panel)
where fracture risk was increased for up to 4 years before
diagnosis in those not treated surgically. However, the re-
sults for those not treated surgically do not support this con-
cept. In diffuse toxic goiter, fracture risk was only increased
for limited periods before (surgically treated) or after (non-
surgically treated) diagnosis. The increase in fracture risk be-
fore diagnosis is supported by previous studies reporting a
decreased BMD in patients with untreated hyperthyroidism
(5–7).
In the questionnaire-based study, the number of years of
follow-up was much smaller (4975 [12]) versus 75,469 in the
actual study and this provides much better options for long-
term follow-up than in the questionnaire-based study. The
fact that there was a significant increase in fracture risk be-
fore diagnosis but not in the previous study (12) may be be-
cause of the much larger study sample in the present study,
i.e., it may not have been possible to attain statistical signif-
icance in the previous study (12).
Bone mineral density. The lower BMD Z score observed
in our study in patients after treatment for nodular toxic
goiter in comparison with patients with diffuse toxic goi-
ter is in agreement with a longer subclinical course in el-
derly subjects with nodular toxic goiter. Furthermore, in el-
derly people, the increase in bone turnover may occur at a
time when the skeleton is in a negative balance because of
aging and lack of sex hormones. Hyperthyroidism may ac-
celerate this bone loss (3) and lead to irreversible lesions
caused by increased endocortical resorption and trabecular
perforations.
The normal BMD Z score observed in patients with dif-
fuse toxic goiter is in agreement with three previous studies
reporting normal BMD in patients treated for hyperthy-
roidism (8,9,33). These observations support the reversible
nature of the bone loss induced by a short increase in bone
remodeling (1) and the return to normal of fracture risk seen
in our study. Unfortunately, the previous studies (8,9,33) did
not provide data on the distribution between diffuse and
nodular toxic goiter, precluding a detailed comparison with
our results.
If previous studies of BMD in patients with hyperthy-
roidism are summarized in a meta-analysis they indicate a
mean 6 standard error of the mean (SEM) Z score in the
spine of 20.94 6 0.09 and in the femoral neck of 20.90 6
0.09 in untreated patients (5–9,34). In treated patients the cor-
responding figures are 20.52 6 0.09 in the spine and
20.22 6 0.06 in the hip (5–9,34).
The figures from these other studies for the lumbar spine
were close to our observations after treatment (Table 4) in
the combined group of patients with Graves’ disease and
nodular toxic goiter (mean 6 SEM Z score 20.43 6 0.11). On
the other hand, our study group displayed a somewhat
larger deficit in the hip (mean 6 SEM, Z score 20.55 6 0.08)
than expected from previous studies.
By using the estimates presented by Marshall et al. (3), it
is possible to transform the BMD values into fracture risk.
In the untreated the IRR for all fractures would be 1.50.94 5
1.5 with the spine data and 1.60.90 5 1.5 with the hip data.
This is actually between that observed within the first 2 years
after diagnosis in the questionnaire-based study (12) (IRR 5
1.9, 95% CI: 1.3–2.9) and the IRR observed within the first
years after diagnosis in the present study (IRR 5 1.44, 95%
CI: 1.26–1.65).
After treatment the IRR should be 1.1–1.2 using the com-
bined BMD results of the previous studies (5–9,34), i.e., no
major change in fracture risk should be expected. This is
418
probably why no major change in fracture risk was seen af-
ter diagnosis in Table 2.
It may thus seem that the present study probably gives a
fracture estimate, which is close to that found in population
based samples of patients with hyperthyroidism.
Effect of surgery in hyperthyroidism. The decreased frac-
ture risk after diagnosis in diffuse toxic goiter (Table 2) sup-
ports the concept that controlling the hyperthyroid state may
reduce fracture risk. This is further supported by the risk re-
duction following surgery after diagnosis in both diffuse and
nodular toxic goiter using multivariate models that also in-
clude age, gender, and previous fractures (Table 3). In pa-
tients with diffuse toxic goiter, this is explained by differ-
ences in BMD levels: those who had had surgery had
significantly higher BMD levels after surgery than those who
had not (Table 4). However, this was not the case in patients
with nodular toxic goiter, where BMD was equally reduced
in those who had and had not had surgery. The lower rela-
tive fracture risk before surgery in patients with nodular
toxic goiter may indicate that patients selected for surgery
could be in better general health than those who were not.
We have previously shown that radioactive iodine, which
may take some time to control hyperthyroidism, is associ-
ated with an increased risk of fractures (12). Surgery, how-
ever, results in prompt control of the hyperthyroid state pro-
vided sufficient tissue is removed. These results emphasize
the importance of early detection and effective treatment of
hyperthyroidism.
However, in the present study, data on treatment with ra-
dioactive iodine were not available, and caution is advised
in the interpretation of these data.
Hypothyroidism
In the present study, we observed an increased risk of frac-
tures in untreated patients with primary hypothyroidism be-
ginning up to 8 years before diagnosis. In untreated hy-
pothyroidism, bone turnover is reduced and bone mass is
often normal or slightly increased (1). The only skeletal ex-
planation for the increase in fracture risk would be a reduced
renewal of bone leading to accumulation of stress fractures.
However, it seems more likely that the increase in fracture
risk before diagnosis is caused by an increased risk of falling
because of compromised neuromuscular function, compli-
cating heart disease, increased sensitivity to drugs, etc.
After initiation of thyroid substitution therapy, an increase
in bone turnover, a decrease in bone mass (17,18), and an in-
creased risk of falls (12) may contribute to the continuously
increased fracture risk observed for up to 4 years after di-
agnosis.
In our patients with treated hypothyroidism no decrease
in Z scores could be demonstrated in accordance with a pre-
vious study (19). The normal Z scores after diagnosis (Table
4) are in accordance with the normalization of fracture risk
5 years after diagnosis (Fig. 2). In Denmark, substitution ther-
apy with levothyroxine is generally aimed at keeping serum
TSH within the normal range, and this may explain why no
decrease in bone mineral was found (15).
The increase in fracture risk within the first 2 years after
diagnosis in a previous questionnaire-based study (IRR 5
2.9, 95% CI: 1.9–4.3) (25) was a little higher than in the pre-
VESTERGAARD AND MOSEKILDE
vious actual study (1.98, 95% CI: 1.77–2.23), again probably
demonstrating that the present study yields a more reliable
population based estimate. In the previous questionnaire-
based study (25), the power may have been inadequate to
demonstrate the small significances, which could be dem-
onstrated in the present study.
However, it should be noted that the observation of a tem-
porary increase in fracture risk is basically the same in the
present study and the previous study (25).
Study limitations
The general advantage of the study is the large number of
participants, which made it possible to analyze subgroups
with a high power in the analyses. Furthermore, the study
was population-based using secondary data and thus limit-
ing information bias as seen in questionnaire-based studies.
The disadvantages are the lack of information on treatment
with radioactive iodine.
Selection bias seems unlikely, because all patients in the
relevant diagnosis groups were included. Information bias
also seems limited because the validity of the fracture diag-
noses and the thyroid disorder diagnoses were high.
Although the peak in fracture risk around the time of di-
agnosis may be linked to the underlying disease, a Berkson
bias is possible: patients with fractures being more likely to
undergo screening for other diseases including thyroid dis-
orders.
Conclusions
Fracture risk is increased in both hyperthyroidism and hy-
pothyroidism. Age, female gender, and previous fractures
are common risk factors. Thyroid surgery seems associated
with a reduced risk in hyperthyroid patients.
Acknowledgments
This study was made possible through a grant from
Aarhus University Research Foundation (E-2000-SUN-1-
125), the Danish Research Council (#9600822), and The Novo
Nordic Centre for Research in Growth and Regeneration.
REFERENCES
1. Mosekilde L, Eriksen EF, Charles P 1990 Effects of thyroid
hormones on bone and mineral metabolism. Endocrinol
Metab Clin North Am 19:35–63.
2. Uzzan B, Campos J, Cucherat M, Nony P, Boissel JP, Perret
GY 1996 Effects on bone mass of long term treatment with
thyroid hormones: A meta-analysis. J Clin Endocrinol Metab
81:4278–4289.
3. Marshall D, Johnell O, Wedel H 1996 Meta-analysis of how
well measures of bone mineral density predict occurrence
of osteoporotic fractures. BMJ 312:1254–1259.
4. Garnero P, Sornay-Rendu E, Claustrat B, Delmas PD 2000
Biochemical markers of bone turnover, endogenous hor-
mones and the risk of fractures in postmenopausal women:
The OFELY study. J Bone Miner Res 15:1526–1536.
5. Tsai KS, Lai SM, Huang KM, Chieng PU, Su CT, Chen FW
1991 Decreased bone mineral density in patients with pro-
longed thyrotoxicosis before and after treatment. J Formos
Med Assoc 90:250–255.
6. Földes J, Tarján G, Szathmari M, Varga F, Krasznai I, Hor-
vath CS 1993 Bone mineral density in patients with en-
FRACTURES IN HYPERTHYROIDISM AND HYPOTHYROIDISM
dogenous subclinical hyperthyroidism: is the thyroid status
a risk factor for osteoporosis? Clin Endocrinol 39:521–527.
7. Lee MS, Kim SY, Lee MC, Cho BY, Lee HK, Koh C-S, Min
HK 1990 Negative correlation between the change in bone
mineral density and serum osteocalcin in patients with hy-
perthyroidism. J Clin Endocrinol Metab 70:766–770.
8. Langdahl BL, Loft AGR, Eriksen EF, Mosekilde L, Charles
P 1996 Bone mass, bone turnover, calcium homeostasis, and
body composition in surgically treated and radioiodine
treated former hyperthyroid patients. Thyroid 6:169–175.
9. Langdahl BL, Loft AGR, Eriksen EF, Mosekilde L, Charles
P 1996 Bone mass, bone turnover, body composition, and
calcium homeostasis in former hyperthyroid patients treated
by combined medical therapy. Thyroid 6:161–168.
10. Cummings SR, Nevitt MC, Browner WS, Stone K, Fox KM,
Ensrud KE, Cauley J, Black D 1995 Risk factors for hip frac-
ture in white women. Study of Osteoporotic Fractures Re-
search Group. N Engl J Med 332:767–773.
11. Wejda B, Hintze G, Katschinski B, Olbricht T, Benker G 1995
Hip fractures and the thyroid: a case-control study. J Intern
Med 237:241–247.
12. Vestergaard P, Rejnmark L, Weeke J, Mosekilde L 2000 Frac-
ture risk in patients treated for hyperthyroidism. Thyroid
10:341–348.
13. Cooper C, Shah S, Hand DJ, Adams J, Compston J, Davie M,
Woolf A 1991 Screening for vertebral osteoporosis using indi-
vidual risk factors. The Multicentre Vertebral Fracture Study
Group. Osteoporos Int 2:48–53.
14. Seeley DG, Kelsey J, Jergas M, Nevitt MC 1996 Predictors of
ankle and foot fractures in older women. The Study of Os-
teoporotic Fractures Research Group. J Bone Miner Res
11:1347–1355.
15. Mosekilde L, Melsen F 1978 Morphometric and dynamic
studies of bone changes in hypothyroidism. Acta Pathol Mi-
crobiol Scand [A] 86:56–62.
16. Coindre JM, David JP, Riviere L, Goussot JF, Roger P,
de Mascarel A, Meunier PJ 1986 Bone loss in hypothyroidism
with hormone replacement. A histomorphometric study.
Arch Intern Med 146:48–53.
17. Ribot C, Tremollieres F, Pouilles JM, Louvet JP 1990 Bone
mineral density and thyroid hormone therapy. Clin En-
docrinol (Oxf) 33:143–153.
18. Trémollières F, Pouillès JM, Louvet JP, Ribot C 1991 [Tran-
sitory bone loss during substitution treatment for hypothy-
roidism. Results of a two year prospective study]. Rev Rhum
Mal Osteoartic 58:869–875.
19. Langdahl BL, Loft AG, Eriksen EF, Mosekilde L, Charles P
1996 Bone mass, bone turnover and body composition in for-
mer hypothyroid patients receiving replacement therapy.
Eur J Endocrinol 134:702–709.
20. Fowler PB, McIvor J, Sykes L, Macrae KD 1996 The effect of
long-term thyroxine on bone mineral density and serum
cholesterol. J R Coll Physicians Lond 30:527–532.
21. Toh SH, Brown PH 1990 Bone mineral content in hypothy-
roid male patients with hormone replacement: A 3-year
study. J Bone Mineral Res 5:463–467.
22. Demeester-Mirkine N, Bergmann P, Body J-J, Corvilain J
1990 Calcitonin and bone mass status in congenital hy-
pothyroidism. Calcif Tissue Int 46:222–226.
419
23. Pioli G, Pedrazzoni M, Palummeri E, Sianesi M, Del Frate
R, Vescovi PP, Prisco M, Ulietti V, Costi D, Passeri M 1992
Longitudinal study of bone loss after thyroidectomy and
suppressive thyroxine therapy in premenopausal women.
Acta Endocrinol 126:238–242.
24. Schneider DL, Barret-Connor EL, Morton DJ 1994 Thyroid
hormone use and bone mineral density in elderly women:
Effects of estrogen. JAMA 271:1245–1249.
25. Vestergaard P, Weeke J, Hoeck HC, Nielsen HK, Rungby J,
Rejnmark L, Laurberg P, Mosekilde L 2000 Fractures in pa-
tients with primary idiopathic hypothyroidism. Thyroid
10:335–340.
26. Frank L 2001 Epidemiology. When an entire country is a co-
hort. Science 287:2398–2399.
27. Andersen TF, Madsen M, Jørgensen J, Mellemkjær L, Olsen
JH 1999 The Danish National Hospital Register. Danish Med
Bull 46:263–268.
28. Jürgensen HJ, Frølund C, Gustafsen J, Mosbech H, Guld-
hammer B 1984 [Registration of diagnoses in a national pa-
tient register. Preliminary assessment of the validity of the
register]. Ugeskr Læger 146:3303–3308.
29. Mosbech J, Jørgensen J, Madsen M, Rostgaard K, Thornberg
K, Poulsen TD 1995 [The Danish National Patient Register:
Evaluation of Data Quality]. Ugeskr Læger 157:3741–3745.
30. Laurberg P, Pedersen KM, Vestergaard H, Sigurdsson G
1991 High incidence of multinodular toxic goiter in the el-
derly population in a low iodine intake area vs. high inci-
dence of Graves’ disease in the young in a high iodine
intake area: Comparative surveys of thyrotoxicosis epi-
demiology in East-Jutland Denmark and Iceland. J Intern
Med 229:415–420.
31. Hollowell JG, Staehling NW, Hannon WH, Flanders DW,
Gunter EW, Maberly GF, Braverman LE, Pino S, Miller DT,
Garbe PL, DeLozier DM, Jackson RJ 1998 Iodine nutrition in
the United States. Trends and public health implications: io-
dine excretion data from National Health and Nutrition Ex-
amination Surveys I and III (1971–1974 and 1988–1994). J
Clin Endocrinol Metab 83:3401–3408.
32. Daly PJ, Fitzgerald Jr RH, Melton III LJ, Ilstrup DM 1987
Epidemiology of ankle fractures in Rochester, Minnesota.
Acta Orthop Scand 58:539–544.
33. Orwoll ES, Bauer DC, Vogt TM, Fox KM 1996 Axial bone
mass in older women. Study of Osteoporotic Fractures Re-
search Group. Ann Intern Med 124:187–196.
34. Campos-Pastor MM, Muñoz-Torres M, Escobar-Jiménez F,
Ruiz de Almodóvar R, Gimeno EJ 1993 Bone mass in females
with different thyroid disorders: Influence of menopausal
status. Bone Mineral 21:1–8.
Address reprint requests to:
Peter Vestergaard
The Osteoporosis Clinic
Aarhus Amtssygehus
Tage Hansens Gade 2
DK-8000 Aarhus C
Denmark
E-mail: p-vest@post4.tele.dk