Sei sulla pagina 1di 8


Cardiovascular Diseases and Risk of Hip Fracture

Ulf Sennerby, MD Context Recent studies indicate common etiologies for cardiovascular disease (CVD)
Håkan Melhus, MD, PhD and osteoporotic fractures.
Rolf Gedeborg, MD, PhD Objectives To examine the relation between CVD and risk of hip fracture in twins and
evaluate the relative importance of genetics and lifestyle factors in this association.
Liisa Byberg, PhD
Design, Setting, and Participants A cohort of all 31 936 Swedish twins born from
Hans Garmo, PhD
1914-1944 was followed up from the age of 50 years. The National Patient Registry
Anders Ahlbom, PhD identified twins with CVDs and fractures from 1964 through 2005. Time-dependent
Nancy L. Pedersen, PhD exposures using Cox proportional hazard regression models were evaluated.
Karl Michaëlsson, MD, PhD Main Outcome Measure Time to hip fracture after diagnosis of CVD.
Results The crude absolute rate of hip fractures was 12.6 per 1000 person-years after
a diagnosis of heart failure, 12.6 per 1000 person-years after a stroke, 6.6 per 1000 person-

years after a diagnosis of peripheral atherosclerosis, and 5.2 per 1000 person-years after
and osteoporosis, which are a diagnosis of ischemic heart disease compared with 1.2 per 1000 person-years for those
commoninelderlyindividuals, without a CVD diagnosis. The multivariable-adjusted hazard ratio (HR) of hip fracture
havebeenregardedasindepen- after a diagnosis of heart failure was 4.40 (95% confidence interval [CI], 3.43-5.63); af-
dentage-relateddisorders.1,2 Researchsug- ter a stroke, the HR was 5.09 (95% CI, 4.18-6.20); after a diagnosis of peripheral ath-
gests common etiologic mechanisms for erosclerosis, the HR was 3.20 (95% CI, 2.28-4.50); and after an ischemic heart disease
thesediseases.3,4 Boneandvasculatureare event, the HR was 2.32 (95% CI, 1.91-2.84). Identical twins without heart failure and
regulated by several shared factors,3 in stroke also had, after their co-twins had been exposed to these respective diseases, an
which calcification of the vascular walls increased rate of hip fracture. These sibling twins pseudoexposed for heart failure had a
multivariable-adjusted HR of 3.74 (95% CI, 1.97-7.10) for hip fracture, whereas pseu-
in many ways resembles the bone forma-
doexposure for stroke had an HR of 2.29 (95% CI, 1.20-4.35).
tion process. Interestingly, bisphospho-
nates not only decrease the progression Conclusions A diagnosis of CVD was significantly associated with risk of subse-
quent hip fracture. Increased risks in co-twins without an index diagnosis suggest ge-
of osteoporosis but also prevent the
netic factors in the association between CVD and osteoporotic fractures.
development of atherosclerosis and re-
JAMA. 2009;302(15):1666-1673
duce total mortality rate.5 In addition,
cholesterol-lowering statins that reduce
risk of CVD are thought to reduce the risk fore uncertain whether other CVDs may was attributable to genes and early en-
of osteoporotic fractures.6,7 increase the risk of future hip frac- vironmental sharing.
Stroke is a well-documented risk fac- ture. It is also unknown whether the
tor for hip fracture, which is the most dev- risk for hip fracture differs depending METHODS
astating consequence of osteoporosis.8 on CVD diagnosis and sex, as well as Individuals were ascertained from the
One explanation for the increased risk af- whether the risk reflects lifestyle and Swedish Twin Registry, currently the
ter this particular cardiovascular event in- individual environmental influences or
cludesanincreasedfallrisk,especiallybe- genetic constitution. A twin cohort pro- Author Affiliations: Department of Surgical Sciences,
Section of Orthopaedics (Drs Sennerby, Byberg, and Mi-
causeofhemiplegia.Anotherexplanation vides a framework for an ordinary co- chaëlsson) and Section of Anaesthesiology and Intensive
isthatimmobilityinducessarcopeniaand hort analysis while simultaneously ex- Care (Dr Gedeborg), Department of Medical Sciences,
boneloss.8,9 However,partoftheincreased amining whether the relation between Clinical Research Center (Drs Melhus, Gedeborg, Byberg,
risk might be a general underlying pre- cardiovascular events and hip fracture and Michaëlsson), and Regional Oncologic Center (Dr
disposition in common for hip fracture is explained by genetic and early envi- Garmo), Uppsala University, Uppsala Sweden, King’s Col-
and for CVD. ronmental factors.12 We therefore used College, London, UK (Dr Garmo); Department of Epide-
Prospective studies examining hip information from 31 936 twins in the miology,InstituteofEnvironmentalMedicine(DrAhlbom),
Department of Medical Epidemiology and Biostatistics,
fracture risk after cardiovascular events Swedish Twin Registry to investigate the Karolinska Institutet, Stockholm, Sweden (Dr Pedersen).
other than stroke have focused on aor- association between cardiovascular Corresponding Author: Karl Michaëlsson, MD, PhD,
Uppsala Clinical Research Center, Uppsala University,
tic calcification as the exposure and pre- events and future hip fracture risk and SE-751 85 Uppsala, Sweden (karl.michaelsson@surgsci
sent diverging results.10,11 It is there- to examine to what extent the relation

1666 JAMA, October 21, 2009—Vol 302, No. 15 (Reprinted) ©2009 American Medical Association. All rights reserved.

largest twin registry in the world.12 Twin nonprescription medication use, occu- hip fracture patients that died instead
pairs born between 1914 and 1944 and pation, education, physical activity, an- had had a hip fracture or (2) all pa-
alive in 1972 were eligible for this study, thropometric measures, and consump- tients were alive at the end of the fol-
which resulted in 15 968 pairs. This co- tion of alcohol and tobacco. To diminish low-up period.
hort of twins has been described pre- the risk of period effects, an effort was Our twin design allowed us to evalu-
viously.13 Zygosity information (based made to interview members of a twin pair ate how genetic factors influence the
on questions about childhood resem- within a month of each other. association between CVD and hip frac-
blance) was obtained at the time of reg- The study was approved by the re- ture. In this additional analysis, the
istry compilation. search ethics committee of Karolinska In- co-twin (ie, the sibling) without CVD
Fractures, CVD diagnoses, and co- stitutet and the twins gave informed ver- was considered pseudoexposed to
morbidities until December 31, 2005, bal consent to participate in the study. CVD from the CVD event in the truly
were determined from the National CVD-affected co-twin until censoring
Patient Registry by International Classi- Statistical Analysis or until the pseudoexposed twin sib-
fication of Diseases (ICD) seventh Kaplan-Meier curves for time-to-hip ling experienced a true CVD. The rela-
through tenth edition codes. Cardiovas- fracture were plotted for each CVD cat- tive hip fracture rates were compared
cular disease subdiagnoses studied were egory. Crude overall rates of hip frac- between pseudoexposed and truly ex-
heart failure, stroke, ischemic heart dis- ture but also rates within 4 different posed twins. Similar estimates would
ease, and peripheral atherosclerosis (ex- time frames (ⱕ1, ⬎1-5, ⬎5-10, and suggest that genes and shared environ-
cluding cardiac and cerebral atheroscle- ⬎10 years) after each CVD diagnosis mental factors are the most important
rosis). We identified hypertension from are presented. determinants regarding risk of frac-
the registry and self-report data (de- Cox proportional hazards regression ture after the CVD event. Estimates in
scribed below) as a secondary out- models were used with time-varying ex- monozygotic and dizygotic twins were
come. The National Patient Registry, posure of CVD and covariate informa- compared to investigate the degree of
which was initiated in 1964, covered 83% tion to assess hazard ratios (HRs) for hip genetic influence.
of the Swedish population in 1972 and fracture with 95% confidence intervals We additionally estimated age-
all inpatient care in Sweden since 1987. (CIs). Co-twin dependences were specific 10-year absolute hip fracture
It is updated annually and valid in iden- handled using robust sandwich vari- risks. For each subpopulation with
tifying cases of fracture.14 The propor- ance estimates according to Lin and heart failure, stroke, atherosclerosis, or
tion of pathological fractures and frac- Wei.17 We modeled hip fracture risk af- ischemic heart disease, a Cox propor-
tures related to high-energy trauma in ter CVD exposure to risk of hip fracture tional hazards model for time to hip
elderly persons in Sweden is only about without a CVD diagnosis. First, we fit- fracture was fitted based on age and sex.
1%.15,16 Dates of death were based on data ted a model that included sex as a co- Predicted sex-specific survival curves
from the continuously updated Swed- variate (all twins were the same age). We were generated for age at CVD diagno-
ish National Population Register. then extended the model to include sepa- sis between 55 and 85 years using 1-year
Twins entered the study on the date rate marker variables (all dichoto- age increments. From each of these sur-
of their 50th birthday (ie, between 1964 mous) for concomitant disorders that vival curves the 10-year absolute risk
and 1994) and were followed up until were likely covariates of the association with 95% CIs was extracted and plot-
the date of a first hip fracture, date of between fracture and CVD, including any ted against age.
death, or end of the follow-up period form of diabetes mellitus, other endo- To demonstrate the importance of
(December 31, 2005). Twins with pre- crine disorders, hyperlipidemia, neuro- some lifestyle factors (identified from the
vious CVD or hip fracture were ex- logic disease, psychiatric disorder, res- telephone interview) on the HRs we ad-
cluded. All twins were considered un- piratory diseases, and musculoskeletal ditionally included in the multivari-
exposed at study entry and until the disorders, as well as subdiagnoses of able model body mass index (BMI) (con-
date of the first cardiovascular event. CVD. The same number of twins was in- tinuous, calculated as the weight in
Although we observed the twins for cluded in the sex-adjusted model as in kilograms divided by height in meters
more than 40 years, the diagnoses stud- the multivariable-adjusted model. Only squared), hormone therapy for women
ied in this article are age-related late- minor differences were observed in the (ever/never), impaired balance (yes/
onset diseases. Therefore, 92% of the HRs if we excluded twins with uncer- no), difficulty to rise from a chair (yes/
CVDs and 95% of the hip fractures were tain zygosity information. no), alcohol use (yes/no), smoking sta-
diagnosed during the past 20 years of Because CVDs are strongly associ- tus (never, former, current), leisure time
follow-up (1986-2005). ated with mortality, sensitivity analy- physical activity (low, medium, high),
A computer-assisted telephone inter- ses were conducted to check for 2 theo- previous fractures (yes/no), any pre-
view was conducted between 1998 and retical and extreme situations that scribed or nonprescribed medication
2000. Questions included items on dis- addressed the problem with compet- (yes/no), and use of thiazides, loop
eases and symptoms, prescription and ing risk from death in that (1) all non- diuretics, warfarin, angiotensin-
©2009 American Medical Association. All rights reserved. (Reprinted) JAMA, October 21, 2009—Vol 302, No. 15 1667

converting enzyme inhibitors, ␤- FIGURE 1. If the twin had had 2 or more cluded 3487 (21.8%) monozygotic,
blockers, and cortisone. In this particu- cardiovascular diseases (eg, heart fail- 5921 (37.1%) same sex dizygotic, and
lar analysis, the outcome variable was ure and stroke), this twin was in- 5607 (35.1%) opposite sex dizygotic
hip fracture cases occurring between the cluded in both the heart failure analy- twin pairs. Another 953 (6.0%) twin
date of the telephone interview and the sis and in the stroke analysis but if the pairs had uncertain zygosity. Of the base
end of the study (December 31, 2005). cardiovascular diseases had occurred at sample (n=31 936), 3663 died before
Because most persons with hyperten- different dates they contributed to dif- the telephone interview, leaving 28 273
sion, in contrast to the other cardiovas- ferent exposure times. This twin was, eligible twins to be interviewed. Of
cular diseases in our study, are not iden- however, not included in the analyses these, 24 598 (87%) participated in the
tified from the National Patient Registry of the other cardiovascular diseases. Sta- interview. Before censoring, 35% of all
(with diagnoses from in-patient care), a tistical analyses were performed using women and 43% of all men were diag-
separate secondary analysis was done for SAS software version 9.1 (SAS Insti- nosed with CVD. Characteristics as a
this disorder. We therefore combined the tute Inc, Cary, North Carolina), Stata function of cardiovascular diagnosis are
exposure information on hypertension 10.1 (Stata Corporation Inc, College presented in TABLE 1. Individuals with
from self-reports during the telephone in- Station, Texas), and R (R Foundation a CVD were, on average, born earlier
terview with the information collected for Statistical Computing, Vienna, Aus- and had more concomitant diseases.
from the patient registry. The twins with tria 2008). However, only modest differences were
hypertension were considered exposed observed in anthropometric measures
from the first date of this diagnosis in the RESULTS and smoking habits between persons
registry or in other cases from the date Four separate validation studies using with and without a CVD diagnosis.
of the telephone interview. genotyping have shown that 95% to Hip fracture after the age of 50 years
The number of twins included and 98% of the twin pairs in the study are was diagnosed in 1442 twins (70% of
excluded in the analyses is given in correctly classified.12 Our cohort in- these cases were women). Kaplan-

Figure 1. Flow Diagram of Those Included and Excluded from the Study

31 936 Twins who were 50 years old at study entry (15 968 pairs)

Heart failure analysis Stroke analysis Peripheral atherosclerosis analysis Ischemic heart disease analysis Hypertension analysis

10 577 Individual twins excluded 9885 Individual twins excluded 11 820 Individual twins excluded 7978 Individual twins excluded 6846 Individual twins excluded
from heart failure analysisa from stroke analysisa from peripheral from ischemic heart disease from hypertension analysisa
10 540 CVD other than 9743 CVD other than atherosclerosis analysisa analysisa 6586 CVD other than
heart failure stroke 11 683 CVD other than 7741 CVD other than hypertension
22 Hip fracture 22 Hip fracture peripheral ischemic heart disease 32 Hip fracture
<50 years old <50 years old atherosclerosis 25 Hip fracture <50 years old
15 Heart failure 120 Stroke <50 29 Hip fracture <50 years old 228 Hypertension
<50 years old years old <50 years old 212 Ischemic heart <50 years old
108 Peripheral disease <50
atherosclerosis years old
<50 years old

21 359 Individual twins included 22 051 Individual twins included 20 123 Individual twins included in 23 958 Individual twins included 24 550 Individual twins included
in heart failure analysis in stroke analysis peripheral atherosclerosis in ischemic heart disease in hypertension analysis
2270 Individual twins 2962 Individual twins analysis analysis 8506 Individual twins
experienced heart experienced stroke 1061 Individual twins 4869 Individual twins experienced
failure 1605 Men experienced experienced ischemic hypertension
1254 Men 1357 Women peripheral heart disease 3755 Men
1016 Women atherosclerosis 2944 Men 4751 Women
597 Men 1925 Women
464 Women

7493 Twin pairs included in 8004 Twin pairs included in 6864 Twin pairs included in 9164 Twin pairs included in
pseudoexposure analysisb pseudoexposure analysisb pseudoexposure analysisb pseudoexposure analysisb
1035 Co-twins had heart 1497 Co-twins had stroke 435 Co-twins had peripheral 2612 Co-twins had ischemic
failure pseudoexposure pseudoexposure atherosclerosis heart disease
pseudoexposure pseudoexposure

The number of twins included and excluded in the analyses of heart failure, stroke, peripheral atherosclerosis, ischemic heart disease, and hypertension. CVD indicates
cardiovascular disease.
a Discordant for disease of interest.
b Exclusion algorithm: 1) excluded twins without CVD of interest, 2) excluded twins with hip fracture before the age of 50 years, and 3) excluded twins with a diagnosis
of heart failure before the age of 50 years.

1668 JAMA, October 21, 2009—Vol 302, No. 15 (Reprinted) ©2009 American Medical Association. All rights reserved.

Meier curves for the proportion of per- tion, an increased risk was also seen eases. Thus, a pseudoexposure of heart
sons with hip fracture at different ages (HR, 2.42; 95% CI, 1.85-3.17). Cardio- failure in identical twin siblings was as-
after a diagnosis of heart failure, stroke, vascular disease was associated with an sociated with a 4-fold increased hip frac-
ischemic heart disease, or peripheral ath- increased relative hip fracture rate to a ture rate as compared with those with-
erosclerosis are displayed in FIGURE 2. comparable level in both men and out CVD (TABLE 3). Moreover, identical
The curve for stroke started to diverge women except after stroke, for which twins pseudoexposed to stroke had a
from the curve for individuals with- the fracture rate tended to be higher in doubled hip fracture rate after the stroke
out a CVD diagnosis early after study men (HR, 6.65; 95% CI, 4.82-9.19) than event in the co-twin. These increased
entry, whereas the curve for ischemic in women (HR, 4.42; 95% CI, 3.49- risks were less pronounced in pseudo-
heart disease started to diverge from the 5.61). exposed dizygotic siblings. Overall, we
baseline curve 25 years later when the Identical twins without heart fail- also found that sibling pseudoexpo-
persons were 75 years of age. The crude ure and stroke also had an increased sure to ischemic heart disease and pe-
absolute rate of hip fractures was 12.6 rate of hip fracture after their co-twins ripheral atherosclerosis conferred in-
per 1000 person-years after a diagno- were exposed to these respective dis- creased hip fracture rates; however, we
sis of heart failure, 12.6 per 1000 per-
son-years after a stroke, 6.6 per 1000
person-years after a diagnosis of pe- Table 1. Selected Characteristics as a Function of Cardiovascular Disease
ripheral atherosclerosis and 5.2 per No. (%) of Participants
1000 person-years after a diagnosis of Ischemic
ischemic heart disease, compared with Peripheral Heart
No CVD Heart Failure Stroke Atherosclerosis Disease
1.2 per 1000 person-years for those (n = 19 089) (n = 2270) (n = 2962) (n = 1061) (n = 4869)
without a CVD diagnosis. Birth year (%) 1934 (8) 1925 (4) 1926 (8) 1927 (8) 1928 (8)
Within the first year after diagnosis Age at the CVD, mean (SD), y a NA 74.4 (8.2) 70.5 (9.0) 69.0 (9.2) 67.9 (9.0)
of both heart failure and stroke BMI, mean (SD) 25.0 (3.3) 25.7 (3.7) 25.3 (3.0) 25.1 (3.3) 25.6 (3.1)
(FIGURE 3), the hip fracture rate was Men 8408 (44.0) 1254 (55.2) 1605 (54.2) 597 (56.3) 2944 (60.5)
higher than the more distant time Women 10 681 (56.0) 1016 (44.8) 1357 (45.8) 464 (43.7) 1925 (39.5)
frames despite that the twins had Cigarette smoking a
achieved an older age and hence were Former 4163 (27.2) 533 (32.4) 613 (28.0) 278 (35.9) 1363 (36.0)
more likely to experience hip fracture Current 2244 (14.7) 203 (12.3) 286 (13.1) 162 (20.9) 469 (12.4)
during the latter periods. The higher Leisure physical activity a
Sedentary 3001 (19.6) 458 (27.8) 494 (22.6) 234 (30.2) 881 (23.2)
rates after a recent diagnosis, how-
Moderate exercise b 10740 (70.3) 1135 (69.0) 1576 (72.1) 498 (64.3) 2667 (70.4)
ever, were not observed in ischemic
Heavy exercise b 1540 (10.1) 53 (3.2) 116 (5.3) 42 (5.4) 243 (6.4)
heart disease and peripheral athero-
Impaired balance a 1197 (7.8) 323 (19.6) 482 (22.0) 160 (20.7) 629 (16.6)
sclerosis. Difficulty rising from a chair a 1469 (9.6) 310 (18.8) 383 (17.5) 155 (20.0) 600 (15.8)
All cardiovascular diagnoses were as- Ever use of 3771 (43.0) 177 (24.0) 240 (23.8) 91 (26.0) 496 (32.8)
sociated with increased hip fracture risk hormone therapy
independent of the other CVD diag- in women a
noses investigated and other comor- Heart failure c NA 596 (20.1) 319 (30.1) 1387 (28.5)
bidities (TABLE 2). Thus, in compari- Cardiovascular diseases c
Cerebrovascular lesion NA 586 (25.8) NA 309 (29.1) 1034 (21.2)
son with individuals without CVD, Ischemic heart disease NA 1416 (62.4) 1053 (35.6) 504 (47.5) NA
heart failure gave a multivariable- Acute myocardial infarction NA 904 (39.8) 633 (21.4) 317 (29.9) 2640 (54.2)
adjusted HR of 4.40 (95% CI, 3.43- Peripheral atherosclerosis NA 336 (14.8) 321 (10.8) NA 515 (10.6)
5.63) for hip fracture; individuals with Hypertension NA 800 (35.2) 1092 (36.9) 422 (39.8) 1532 (31.5)
a stroke had an HR of 5.09 (95% CI, Cancer c 2757 (14.4) 463 (20.4) 530 (17.9) 228 (21.5) 905 (18.6)
4.18-6.20). The elevated hip fracture Endocrine disease c 1224 (6.4) 984 (43.4) 1023 (34.5) 432 (40.7) 1798 (36.9)
rate was present after ischemic stroke Musculoskeletal disorder c 2605 (13.6) 747 (32.9) 832 (28.1) 348 (32.8) 1333 (27.4)
(HR, 4.95; 95% CI, 4.07-6.02) and hem- Psychiatric disease c 1742 (9.1) 537 (23.7) 662 (22.4) 223 (21.0) 855 (17.6)
orrhagic stroke (HR, 5.48; 95% CI, 3.68- Neurologic disorder c 907 (4.8) 375 (16.5) 728 (24.6) 207 (19.5) 698 (14.3)
8.15). Peripheral atherosclerosis (HR, Respiratory disease c 1511 (7.9) 1152 (50.7) 918 (31.0) 405 (38.2) 1666 (34.2)
3.20; 95% CI, 2.28-4.50) and ische- Abbreviation: BMI, body mass index, calculated as the weight in kilograms divided by height in meters squared; NA, not
mic heart disease (HR, 2.32; 95% CI, applicable.
a Among those who participated at the telephone interview (n=15 281 for those with no CVD, n=1646 for heart failure,
1.91-2.84) also conferred higher risk of n=2186 for stroke, n=774 for atherosclerosis, and n=3791 for ischemic heart disease. The number of responding women
was respectively 8765, 738, 1008, 350, and 1513.
hip fracture. If the diagnosis of ische- b Moderate exercise was defined as regular walking or gardening. Heavy exercise was defined as regularly engaging in
mic heart disease was restricted to in- hard physical training.
c Identified from the National Patient Registry.
dividuals with acute myocardial infarc-
©2009 American Medical Association. All rights reserved. (Reprinted) JAMA, October 21, 2009—Vol 302, No. 15 1669

had a limited number of hip frac- Hypertension, as diagnosed based on ing had a hip fracture, multivariable-
tures in the subgroup of identical self-reports and register data, was also adjusted HRs for hip fracture ranged
twins. associated with an increased hip frac- from 9.78 (95% CI, 8.99-10.64) for
Ten-year absolute risks of hip frac- ture risk with a sex-adjusted HR of 1.42 heart failure to 3.69 (95% CI, 3.42-
ture in relation to sex and age at CVD (95% CI,1.23-1.64) and a multivariable- 3.98) for ischemic heart disease. At the
diagnosis are depicted in FIGURE 4. A adjusted HR of 1.59 (95% CI, 1.36- other extreme, when those who died
higher risk of hip fracture was ob- 1.85). without experiencing a hip fracture
served in women at all ages and in all We analyzed hip fracture risk in twins were observed to the end of the fol-
CVD categories. The average 75-year- who had participated in the telephone in- low-up period, the HRs ranged from
old woman had an 18% (95% CI, 13%- terview. Using this approach, we had the 3.89 (95% CI, 3.21-4.70) for stroke to
22%) risk of hip fracture within 10 years possibility to include lifestyle factors, 1.97 (95% CI, 1.60-2.43) for ischemic
after diagnosis of heart failure, whereas medication use, and body stature to our heart disease.
the corresponding estimate was 10% multivariable model. This comprehen-
(95% CI, 7%-14%) in a man of the same sive multivariable model only margin- COMMENT
age. A 75-year-old woman with a stroke ally affected our parameter estimates of Our population-based twin study dem-
had a 10-year risk of hip fracture of 19% hip fracture risk after an event of CVD; onstrated an increased rate of hip frac-
(95% CI, 15%-23%), whereas a man of thus, these results are not presented. ture after different major cardiovascu-
the same age had a risk of 15% (95% In sensitivity analyses, when twins lar events in both women and men. The
CI, 12%-19%). who had died were classified as hav- high absolute 10-year risk of hip frac-

Figure 2. Cumulative Proportion of Hip Fracture by Cardiovascular Disease Status

Heart failure Stroke

40 40
Heart failure Stroke
30 30
Cumulative Proportion

Cumulative Proportion
With Hip Fracture, %

With Hip Fracture, %

20 20

10 10

0 0
50 55 60 65 70 75 80 85 50 55 60 65 70 75 80 85
Age, y Age, y
No. at risk No. at risk
Heart failure 0 25 120 205 321 416 387 245 Stroke 0 144 363 556 676 774 633 280
No CVD 21 359 21 023 20 494 14 695 9653 5796 2716 945 No CVD 22 051 21 585 20 925 14 915 9723 5756 2630 905

Peripheral atherosclerosis Ischemic heart disease

40 40
Peripheral atherosclerosis Ischemic heart disease
30 30
Cumulative Proportion

Cumulative Proportion
With Hip Fracture, %

With Hip Fracture, %

20 20

10 10

0 0
50 55 60 65 70 75 80 85 50 55 60 65 70 75 80 85
Age, y Age, y
No. at risk No. at risk
Peripheral Ischemic heart
atherosclerosis 0 71 181 256 283 274 188 88 disease 0 335 959 1412 1544 1419 956 460
No CVD 20 123 19 747 19 221 13 470 8616 5030 2309 811 No CVD 23 958 23 299 22 210 15 673 10 139 5926 2695 912

Kaplan-Meier curves of hip fracture for twins with and without cardiovascular disease (CVD).

1670 JAMA, October 21, 2009—Vol 302, No. 15 (Reprinted) ©2009 American Medical Association. All rights reserved.

ture after CVD was more prominent in a stroke,26 which could explain our constitutes a competing risk for frac-
women even though we found no dif- higher relative rates of hip fracture in ture. The sensitivity analysis, which ad-
ferences in relative rates for hip fracture men. The genetic influence measured dressed the problem with competing
between the sexes, and even a higher as pseudoexposure was less dominant risk from death, indicates that our es-
relative risk for hip fracture after a diag- after a diagnosis of stroke than for heart timates are not likely to be exagger-
nosis of stroke in men. The previously failure, which could, in part, be ex- ated; rather, the contrary is true.
recognized lifetime risk of hip fracture plained by the increased propensity to Our results extend the findings from
independent of CVD status in women is falls after a stroke. Additionally, bal- epidemiological studies examining the
twice that of men.18,19 We advocate that ance disturbances may have occurred association between CVD and frac-
individuals with a recent diagnosis of as adverse effects of medication, lead- ture risk. In a prospective study in
CVD should have their future fracture ing to falling accidents. However, this women, Bagger et al10 verified the cross-
risk evaluated with clinical risk factors is not likely the most plausible expla- sectional finding of Schulz et al33 that
and bone scans (eg, by the recently es- nation because there is no increased risk severe aortic calcifications increased the
tablished 10-year probability using the of fracture with cardiovascular drug relative risk of hip fracture 2-fold to
World Health Organization fracture risk treatment27 and because most of the in- 3-fold. In contrast, a recent Framing-
assessment tool [FRAX] algorithm).20 creased rate of hip fracture after CVD ham Heart Study11 report indicated that
Furthermore, it should be emphasized was mediated by genetic causes in com- the severity of aortic calcifications did
that our estimates were independent of mon for the diseases. not significantly affect hip fracture rate.
previously recognized clinical risk fac- We found the highest hip fracture The 35-year-long duration of the follow-
tors of hip fracture and a diagnosis of rates early after diagnosis of heart fail- up, however, might have concealed a
CVD would most probably add risk to ure or stroke. Such rates have previ-
those factors. ously been observed after stroke8 but
An increased hip fracture risk for the not after heart failure. The time course Figure 3. Rate of Hip Fracture After
Diagnosis of Cardiovascular Disease
pseudoexposure in the co-twin analy- of the hip fracture rate may have been
ses, particularly in identical twins, is an partially related to immobilization af- 30

Hip Fracture Rate per 1000 Person-Years

Follow-up period, y
indication that genes predispose to the ter the CVD event. A concomitant de- ≤1 >5-10
development of CVD and fractures. Most crease in muscle strength and pos- 25 >1-5 >10

of the overall increased rate of hip frac- tural stability may increase the risk for
ture after heart failure (and part of the falls8 and subsequently for fractures.
increased risk after stroke) appears to be Moreover, immobilization increases the 15
explained by genes or by early environ- rate of bone loss that can further in-
mental sharing (ie, not individual life- crease the risk of fracture.28 Increased 10

style habits or other individual-specific bone loss has been observed after the 5
environmental factors). diagnosis of heart failure, hyperten-
Our study was not designed to fully sion, and stroke.29 The severity of the 0
elucidate the common pathophysiologi- cardiovascular diagnoses, which we Heart Stroke Peripheral Ischemic
Failure Atherosclerosis Heart
cal mechanisms that CVDs and hip frac- were not able to address in this study, Disease
ture share. The genetic factors that can be of importance in relation to frac-
Hip fracture rates with 95% confidence intervals dur-
might explain our associations in- ture risk. A more severe CVD is asso- ing different periods after diagnosis of cardiovascular
clude not only telomere length21,22 but ciated with higher mortality,30-32 which disease.
also specific genes involved in cellular
mechanisms shared by the vascula- Table 2. Hazard Ratios of Hip Fracture Associated With Different Cardiovascular Diseases a
ture and bone. Matrix proteins sup- Peripheral Ischemic Heart
porting bone, vessel walls, and the myo- Heart Failure Stroke Atherosclerosis Disease
cardium can be of special relevance.23,24 No. of hip fracture 113 218 45 185
Other potential factors are calcifica- cases with CVD
tion regulatory hormones, sex ste- Exposed, person-years 8931 17 313 6828 35 935
roids, proteins related to lipid metabo- No CVD, person-years 430 041 437 461 396 356 464 172
lism, oxidative stress, and chronic Sex-adjusted model, 3.04 (2.42-3.81) 3.86 (3.25-4.59) 2.04 (1.50-2.79) 1.85 (1.54-2.21)
HR (95% CI)
inflammation.3,25 Multivariable model, 4.40 (3.43-5.63) 5.09 (4.18-6.20) 3.20 (2.28-4.50) 2.32 (1.91-2.84)
In addition to their vulnerability to HR (95% CI) b
fractures through direct influences on Abbreviations: CI, confidence interval; CVD, cardiovascular disease; HR, hazard ratio.
a No CVD group was used as reference group (includes 526 twins with hip fracture).
bone, persons with CVDs could have b Adjusted for sex, endocrine disorder, neurologic disease, psychiatric disorder, respiratory disease, musculoskeletal dis-
a greater propensity for falls. There is order, hyperlipidemia, diabetes mellitus, heart failure, stroke, peripheral atherosclerosis, ischemic heart disease, and hy-
pertension (all dichotomous).
a higher tendency for men to fall after
©2009 American Medical Association. All rights reserved. (Reprinted) JAMA, October 21, 2009—Vol 302, No. 15 1671

Table 3. Hazard Ratios of Hip Fracture for Pseudoexposure of Different Cardiovascular Diseases by Zygosity a
Type of Pseudoexposure

Monozygotic Twins (n=3487) Dizygotic Twins (n= 11 528)

Peripheral Ischemic Heart Peripheral Ischemic Heart

Heart Failure Stroke Atherosclerosis Disease Heart Failure Stroke Atherosclerosis Disease
No. exposed 196 304 78 534 839 1193 375 2078
No. of hip fracture cases 12 12 2 9 33 43 13 63
among exposed
No. at risk, exposed, 1598 2736 632 4209 6513 10 619 3653 18 560
No. of hip fracture cases 87 103 79 94 244 307 214 300
among nonexposed
No. at risk, nonexposed, 70 722 75 757 62 998 84 299 211 019 227 427 184 090 259 335
Sex-adjusted model, 3.41 1.99 1.66 1.17 1.96 1.62 1.24 1.68
HR (95% CI) (1.80-6.45) (1.05-3.76) (0.41-6.81) (0.57-2.37) (1.34-2.89) (1.15-2.29) (0.70-2.20) (1.25-2.25)
Multivariable model, 3.74 2.29 1.91 1.35 2.10 1.88 1.58 2.02
HR (95% CI) b (1.97-7.10) (1.20-4.35) (0.47-7.85) (0.66-2.78) (1.43-3.09) (1.33-2.64) (0.89-2.80) (1.50-2.72)
Abbreviations: CI, confidence interval; HR, hazard ratio.
a No cardiovascular disease group was used as reference group (HR=1.0).
b Adjusted for sex, endocrine disorder, neurologic disease, psychiatric disorder, respiratory disease, musculoskeletal disorder, hyperlipidemia, diabetes mellitus, heart failure, stroke, pe-
ripheral atherosclerosis, ischemic heart disease, and hypertension (all dichotomous).

demonstrated that the association be-

Figure 4. Risk of Hip Fracture by Age and Sex tween CVD and hip fracture risk in the
Heart failure Stroke case-control study was dependent on
60 60 the diagnosis of CVD. Furthermore, in
10-Year Cohort Risk for Hip Fracture, %

10-Year Cohort Risk for Hip Fracture, %

the same study, the excess risk of hip
50 50
fracture, independent of comorbidity
40 40 and lifestyle habits, was most promi-
nent after a diagnosis of heart failure,
30 30 hypertension, or stroke and less obvi-
20 20
ous after ischemic heart disease or pe-
ripheral atherosclerosis. The present co-
10 10 hort study confirms that these CVDs are
associated with an increased hip frac-
0 0
ture risk.
55 60 65 70 75 80 85 55 60 65 70 75 80 85
Age at Diagnosis, y Age at Diagnosis, y The reliability of our results de-
Atherosclerosis Ischemic heart disease
pends on the quality of the register data.
60 60 The unique personal identity number
10-Year Cohort Risk for Hip Fracture, %

10-Year Cohort Risk for Hip Fracture, %

used in Sweden made it possible to link

50 50 individual prospective data on dis-
40 40
eases treated during hospitalization
through the nationwide and complete
30 30 National Patient Registry. With this reg-
istry, there is virtually no loss to follow-
20 20
up. The overall quality of this registry
10 10 as well as the validity of the CVD data
are considered high.35,36 After the first
0 0 event of acute stroke, the diagnosis
55 60 65 70 75 80 85 55 60 65 70 75 80 85 could be confirmed in 94% of the cases:
Age at Diagnosis, y Age at Diagnosis, y
for heart failure, the corresponding fig-
Ten-year probabilities of hip fracture by age at time of diagnosis of cardiovascular disease. Shaded areas in- ure was 95%. Thus, the impact of di-
dicate 95% confidence intervals.
agnostic misclassification is probably
modest. However, in addition to no in-
potential association between aortic cal- study in women by our team,34 we ob- formation on severity of the CVD, the
cification and risk of hip fracture. In the served the greatest fracture rates after ability to detect CVDs, diagnostic cri-
present study and in a case-control a recent diagnosis of CVD. We also teria for diagnoses, and treatments have
1672 JAMA, October 21, 2009—Vol 302, No. 15 (Reprinted) ©2009 American Medical Association. All rights reserved.

changed over our long follow-up, with the data; and preparation, review, or approval of the 19. Oden A, Dawson A, Dere W, Johnell O, Jonsson
manuscript. B, Kanis JA. Lifetime risk of hip fractures is
the strong possibility of conservative bi- underestimated. Osteoporos Int. 1998;8(6):599-
ased estimates as a consequence. Our 603.
REFERENCES 20. Kanis JA, Johnell O, Oden A, Johansson H,
study has additional strengths, includ- McCloskey E. FRAX and the assessment of fracture
ing a population-based design, a large 1. Dennison E, Mohamed MA, Cooper C. Epidemi- probability in men and women from the UK. Osteo-
ology of osteoporosis. Rheum Dis Clin North Am. 2006; poros Int. 2008;19(4):385-397.
number of CVD events and fractures, 32(4):617-629. 21. Wong LS, de Boer RA, Samani NJ, van Veldhuisen
inclusion of men and women, com- 2. Mathers CD, Loncar D. Projections of global mor- DJ, van der Harst P. Telomere biology in heart failure.
plete follow-up, and extensive covari- tality and burden of disease from 2002 to 2030. PLoS Eur J Heart Fail. 2008;10(11):1049-1056.
Med. 2006;3(11):e442. 22. Valdes AM, Richards JB, Gardner JP, et al. Telo-
ate data. Information on body weight, 3. McFarlane SI, Muniyappa R, Shin JJ, Bahtiyar G, mere length in leukocytes correlates with bone min-
smoking, medication, physical fit- Sowers JR. Osteoporosis and cardiovascular disease: eral density and is shorter in women with osteoporosis.
brittle bones and boned arteries, is there a link? Osteoporos Int. 2007;18(9):1203-1210.
ness, and physical activity level, how- Endocrine. 2004;23(1):1-10. 23. Baldini V, Mastropasqua M, Francucci CM,
ever, was obtained in only a subgroup 4. Tankó LB, Christiansen C, Cox DA, Geiger MJ, D’Erasmo E. Cardiovascular disease and osteoporosis.
McNabb MA, Cummings SR. Relationship between J Endocrinol Invest. 2005;28(10)(suppl):69-72.
and by telephone interview. Changes osteoporosis and cardiovascular disease in postmeno- 24. Hofbauer LC, Brueck CC, Shanahan CM, Schoppet
in these variables during follow-up were pausal women. J Bone Miner Res. 2005;20(11): M, Dobnig H. Vascular calcification and osteoporosis—
not considered in our study. Another 1912-1920. fromclinicalobservationtowardsmolecularunderstanding.
5. Lyles KW, Colón-Emeric CS, Magaziner JS, et al; Osteoporos Int. 2007;18(3):251-259.
limitation is that the pseudoexposure HORIZON Recurrent Fracture Trial. Zoledronic acid and 25. Theill LE, Boyle WJ, Penninger JM. RANK-L and
analyses in the identical twins in- clinical fractures and mortality after hip fracture. N Engl RANK: T cells, bone loss, and mammalian evolution.
J Med. 2007;357(18):1799-1809. Annu Rev Immunol. 2002;20:795-823.
cluded few hip fracture cases that led 6. Bauer DC, Mundy GR, Jamal SA, et al. Use of stat- 26. Nyberg L, Gustafson Y. Fall prediction index for
to unstable estimates. The analyses with ins and fracture: results of 4 prospective studies and patients in stroke rehabilitation. Stroke. 1997;28
cumulative meta-analysis of observational studies and (4):716-721.
a higher number of fractures in dizy- controlled trials. Arch Intern Med. 2004;164(2): 27. Wiens M, Etminan M, Gill SS, Takkouche B. Ef-
gotic twins, however, gave supportive 146-152. fects of antihypertensive drug treatments on fracture
7. Rejnmark L, Olsen ML, Johnsen SP, Vestergaard outcomes: a meta-analysis of observational studies.
results, ie, the increased risk of hip frac- P, Sørensen HT, Mosekilde L. Hip fracture risk in statin J Intern Med. 2006;260(4):350-362.
ture after a CVD seems to be partially users—a population-based Danish case-control study. 28. Heaney RP, Valent DJ, Barton IP. Hospitalization-
explained by the contribution of either Osteoporos Int. 2004;15(6):452-458. related bone loss and the protective effect of
8. Kanis J, Oden A, Johnell O. Acute and long-term risedronate. Osteoporos Int. 2006;17(2):212-216.
genetic constitution or early shared en- increase in fracture risk after hospitalization for stroke. 29. Farhat GN, Strotmeyer ES, Newman AB, et al. Volu-
vironment. Regrettably, we did not have Stroke. 2001;32(3):702-706. metric and areal bone mineral density measures are
9. Myint PK, Poole KE, Warburton EA. Hip fractures associated with cardiovascular disease in older men
the possibility to adjust our estimates after stroke and their prevention. QJM. 2007;100 and women: the Health, Aging, and Body Composi-
for bone mineral density. (9):539-545. tion Study. Calcif Tissue Int. 2006;79(2):102-111.
10. Bagger YZ, Tankó LB, Alexandersen P, Qin G, 30. Prosser J, MacGregor L, Lees KR, Diener HC, Hacke
Clinicians should be aware of the Christiansen C; Prospective Epidemiological Risk Fac- W, Davis S; VISTA Investigators. Predictors of early
considerably increased rate of hip frac- tors Study Group. Radiographic measure of aorta cal- cardiac morbidity and mortality after ischemic stroke.
ture in both sexes, especially after a re- cification is a site-specific predictor of bone loss and Stroke. 2007;38(8):2295-2302.
fracture risk at the hip. J Intern Med. 2006;259 31. Jacobs DR Jr, Kroenke C, Crow R, et al. PREDICT:
cent hospitalization for CVD. Genetic (6):598-605. a simple risk score for clinical severity and long-term
predisposition is probably a major de- 11. Samelson EJ, Cupples LA, Broe KE, Hannan MT, prognosis after hospitalization for acute myocardial in-
O’Donnell CJ, Kiel DP. Vascular calcification in middle farction or unstable angina: the Minnesota heart survey.
terminant of the excess fracture rate. age and long-term risk of hip fracture: the Framing- Circulation. 1999;100(6):599-607.
ham Study. J Bone Miner Res. 2007;22(9):1449- 32. Dickstein K, Cohen-Solal A, Filippatos G, et al; Task
Author Contributions: Dr Michaëlsson had full ac- 1454. Force for Diagnosis and Treatment of Acute and
cess to all of the data in the study and takes respon- 12. Lichtenstein P, De Faire U, Floderus B, Svartengren Chronic Heart Failure 2008 of European Society of
sibility for the integrity of the data and the accuracy M, Svedberg P, Pedersen NL. The Swedish Twin Regsitry: Cardiology; ESC Committee for Practice Guidelines;
of the data analysis. a unique resource for clinical, epidemiologic and genetic Document Reviewers. ESC Guidelines for the diag-
Study concept and design: Sennerby, Melhus, studies. J Intern Med. 2002;252(3):184-205. nosis and treatment of acute and chronic heart fail-
Michaëlsson. 13. Michaelsson K, Melhus H, Ferm H, Ahlbom A, ure 2008: the Task Force for the Diagnosis and Treat-
Acquisition of data: Melhus, Pedersen, Michaëlsson. Pedersen NL. Genetic liability to fractures in the elderly. ment of Acute and Chronic Heart Failure 2008 of the
Analysis and interpretation of data: Sennerby, Melhus, Arch Intern Med. 2005;165(16):1825-1830. European Society of Cardiology. Developed in col-
Gedeborg, Byberg, Garmo, Ahlbom, Pedersen, 14. Naessén T, Parker R, Persson I, Zack M, Adami HO. laboration with the Heart Failure Association of the
Michaëlsson. Time trends in incidence rates of first hip fracture in the ESC (HFA) and endorsed by the European Society of
Drafting of the manuscript: Sennerby, Michaëlsson. Uppsala Health Care Region, Sweden, 1965-1983. Am Intensive Care Medicine (ESICM). Eur Heart J. 2008;
Critical revision of the manuscript for important in- J Epidemiol. 1989;130(2):289-299. 29(19):2388-2442.
tellectual content: Sennerby, Melhus, Gedeborg, 15. Michaëlsson K, Baron JA, Farahmand BY, et al; 33. Schulz E, Arfai K, Liu X, Sayre J, Gilsanz V. Aortic
Byberg, Garmo, Ahlbom, Pedersen, Michaëlsson. The Swedish Hip Fracture Study Group. Hormone re- calcification and the risk of osteoporosis and fractures.
Statistical analysis: Gedeborg, Byberg, Garmo, Ahlbom, placement therapy and risk of hip fracture: popula- J Clin Endocrinol Metab. 2004;89(9):4246-4253.
Michaëlsson. tion based case-control study. BMJ. 1998;316 34. Sennerby U, Farahmand B, Ahlbom A, Ljunghall
Obtained funding: Melhus, Pedersen, Michaëlsson. (7148):1858-1863. S, Michaëlsson K. Cardiovascular diseases and future
Administrative, technical, or material support: 16. Michaëlsson K, Lithell H, Vessby B, Melhus H. Se- risk of hip fracture in women. Osteoporos Int. 2007;
Sennerby, Pedersen, Michaëlsson. rum retinol levels and the risk of fracture. N Engl J Med. 18(10):1355-1362.
Study supervision: Pedersen, Michaëlsson. 2003;348(4):287-294. 35. Ingelsson E, Arnlöv J, Sundström J, Lind L. The va-
Financial Disclosures: None reported. 17. Lin DY, Wei LJ. The robust inference for the Cox lidity of a diagnosis of heart failure in a hospital discharge
Funding/Support: This work was supported by grant proportional hazards model. J Am Stat Assoc. 1989; register. Eur J Heart Fail. 2005;7(5):787-791.
AG 08754 from the National Institutes of Health and 84(408):1074-1078. 36. Merlo J, Lindblad U, Pessah-Rasmussen H, et al. Com-
another from the Swedish Research Council. 18. Nguyen ND, Ahlborg HG, Center JR, Eisman JA, parison of different procedures to identify probable cases
Role of the Sponsor: The funding organizations had Nguyen TV. Residual lifetime risk of fractures in of myocardial infarction and stroke in two Swedish pro-
no role in the design and conduct of the study; col- women and men. J Bone Miner Res. 2007;22(6): spective cohort studies using local and national routine
lection, management, analysis, and interpretation of 781-788. registers. Eur J Epidemiol. 2000;16(3):235-243.

©2009 American Medical Association. All rights reserved. (Reprinted) JAMA, October 21, 2009—Vol 302, No. 15 1673