Risk of Major Osteoporotic Fracture After Cardiovascular Disease: A Population-Based Cohort Study in Taiwan

Background We investigated the association between cardiovascular disease (CVD) and the risk of major osteoporotic fracture in Taiwan. Methods Using the Taiwan National Health Insurance Database for the period 2000–2007, we classified 43 874 patients aged 50 years or older with newly diagnosed CVD (coronary artery disease, heart failure, cerebrovascular disease, or peripheral atherosclerosis) as the CVD group and 43 874 subjects without CVD (frequency-matched by sex, age, and date selected) as the non-CVD group. Incidence and hazard ratios (HRs) for major osteoporotic fracture of the spine, hip, humerus, and forearm/wrist were estimated for the period until the end of 2010. Results After adjustment for confounders, the overall HRs for major osteoporotic fracture were 1.24 (95% CI = 1.13, 1.36) in men with CVD and 1.18 (95% CI = 1.11, 1.25) in women with CVD, as compared with the non-CVD group. As compared with the non-CVD group, the adjusted HR for major osteoporotic fracture was highest among subjects with cerebrovascular disease (HR 1.31; 95% CI 1.23, 1.39), followed by those with heart failure (HR 1.18; 95% CI 1.11, 1.27), peripheral atherosclerosis (HR 1.12; 95% CI 1.04, 1.20), and coronary artery disease (HR 1.07; 95% CI 1.01, 1.12). Conclusions CVD is associated with risk of major osteoporotic fracture in men and women in Taiwan.


INTRODUCTION
Cardiovascular disease (CVD) and osteoporosis are major diseases that cause marked morbidity, disability, and mortality, and a very large socioeconomic burden worldwide. [1][2][3] In their review, Deaton et al noted that CVD accounts for about one-third of all deaths in the world. 1 Another review showed that osteoporosis is responsible for fractures in about 2 million people in the United States annually. 2 Recently, accumulating evidence shows a robust association between CVD and osteoporosis/osteoporotic fracture, which share risk factors and pathophysiologic pathways. [4][5][6][7][8][9] Aging, menopause, and chronic inflammation may partially explain this association. 6,7 A study in Australia by Chen et al found that fracture was 1.23 times (95% CI = 1.13-1.35) as likely among women with CVD than among those without CVD. 8 Gerber et al reported that US patients had a 1.32-fold risk of fracture (95% CI = 1.12-1.56) after myocardial infarction. 9 In 2010, CVD was the second leading cause of death in Taiwan, accounting for 10.8% of all deaths. 10 In a study using data from the Taiwan National Health Insurance database, Yang et al estimated that the average prevalence of osteoporosis in 1999-2001 was 11.4% in women, and 1.6% in men, aged 50 years or older. 11 To date, most studies of the association between CVD and osteoporosis/osteoporotic fracture were limited to white populations, and little evidence is available in Taiwan. Identifying risk factors for fracture is important in preventing development of fractures and improving outcomes. In this population-based cohort study, we used claims data from the National Health Insurance program in Taiwan to investigate the association between CVD and risk of major osteoporotic fracture.

Data sources
This study used claims data from the National Health Insurance program in Taiwan, which has been described in detail in previous studies. [12][13][14][15][16] In brief, the National Health Research Institute in Taiwan has used insurance claims data to establish several longitudinal data files on medical services provided to inpatients and outpatients. All data files were linked with scrambled identifiers to protect patient privacy. Disease diagnoses were coded according to the International Classification of Diseases 9th Revision-Clinical Modification (ICD-9-CM). In addition, the A-code was used to define diseases, because its use predated adoption of ICD-9 coding in Taiwan.

Design
We used claims data covering 1 million insured persons and established 2 study groups. The CVD group included patients who received new diagnoses during 2000-2007 of coronary artery disease (ICD-9 410-414; A-code A270, A279), heart failure (ICD-9 428), cerebrovascular disease (ICD-9 430-438; A-code A290-A294, A299), or peripheral atherosclerosis (ICD-9 440-448; A-code A300). All people younger than 50 years on the day of diagnosis were excluded. For each selected CVD patient, 1 adult without medical claims for CVD was randomly selected for the non-CVD group, after frequencymatching for sex, age (every 5-year span), and index data. The index date was defined as the date of diagnosis for CVD patients and as the middle date of the same index month as their matched CVD patients for non-CVD subjects. To decrease confounding effects, subjects previously diagnosed with fracture at any site (ICD-9 800-829, E887, and A-code A470-A476, A479) before the index date were excluded from the study. Diagnosis of a major osteoporotic fracture, including fracture of the spine (ICD-9 805 and 806), hip (ICD-9 820), humerus (ICD-9 812), and forearm (ICD-9 813)/ wrist (ICD-9 814), were used as the study end-point. These fracture sites have been defined by the World Health Organization fracture risk assessment tool, FRAX. 17 Both the CVD and non-CVD groups were followed up to determine fracture incidence until a subject received a fracture diagnosis, death, withdrawal from the insurance program, loss to followup, or December 31, 2010.
Comorbidities potentially associated with major osteoporotic fracture before the index date were hypertension, arrhythmia, hyperlipidemia, obesity, diabetes mellitus, dementia, Parkinson's disease, depression, chronic kidney disease, osteoporosis, menopause, cancer, thyrotoxicosis, tobacco use, and alcoholism. All were diagnosed by using ICD-9 code and A-code.

Definition of exposure
Medical therapy for osteoporosis was defined as at least 1 recorded prescription for bisphosphonate, calcitonin, estrogen, or raloxifene before the index date. Use of other bonerelated medications was defined as at least 1 prescription for thiazolidinedione, proton-pump inhibitor, glucocorticoid, statin, warfarin, or heparin before the index date.

Statistical analysis
Demographic status, comorbidities, and medications were compared between the CVD and non-CVD groups among men and women separately. The chi-square test and Student t test were used to compare differences between the CVD and non-CVD groups in sociodemographic characteristics, comorbidities, and medications. The incidence rate of fracture was calculated as number of fracture cases identified during follow-up divided by total person-years for each group. Cox proportional hazard models were used to estimate hazard ratios (HRs) and 95% CIs for sex-specific risk of hip fracture and major osteoporotic fracture in association with CVD. Subanalysis evaluated whether different types of CVD were associated with risk of major osteoporotic fracture. All analyses were performed using SAS software version 9.1 (SAS Institute Inc., Cary, NC, USA), and the statistical significance level was defined as a 2-sided P value less than 0.05.

RESULTS
Baseline characteristics of the study population Analysis of claims data from 2000-2007 revealed 43 874 patients with CVD (CVD group) and 43 874 subjects free from disease (non-CVD group) who met the eligibility criteria. Men in the CVD group had a significantly shorter mean duration of follow-up than did men in the non-CVD group (6.03 vs 6.44 years, P < 0.0001; Table 1). Comorbidities and medications were more frequent in the CVD group than in the non-CVD group, with the exception of obesity, cancer, and use of bisphosphonate and raloxifene. Women in the CVD group had a significantly shorter mean duration of follow-up than did women in the non-CVD group (6.25 vs 6.64 years, P < 0.0001; Table 1). As with men, most comorbidities and medications were more frequent in the CVD group than in the non-CVD group, with the exception of obesity, cancer, tobacco use, alcoholism, and use of bisphosphonate and raloxifene. There was no significant difference in medical therapy for osteoporosis between the CVD group and non-CVD group in either sex (data not shown). Table 2 shows the sex-specific risk of hip fracture by duration of follow-up for the CVD and non-CVD groups. The incidence of hip fracture was highest among women with CVD (4.27 per 1000 person-years), followed by women without CVD, men with CVD, and men without CVD (1.96 per 1000 person-years). Multivariate Cox regression analysis revealed that the HR for hip fracture associated with CVD was not significant (HR, 1.02; 95% CI = 0.92, 1.13).

Association between CVD and hip fracture
Association between CVD and major osteoporotic fractures Table 3 shows the sex-specific risk of major osteoporotic fracture by duration of follow-up among adults with and without CVD. The incidence of fracture was highest among women with CVD (21.5 per 1000 person-years), followed by women without CVD, men with CVD, and men without CVD (7.02 per 1000 person-years). Multivariate Cox regression analysis revealed that the HR for major osteoporotic fracture associated with CVD was 1.16 (95% CI 1.10, 1.22). The sex-specific HRs were similar for men and women (HR, 1.24 vs 1.18). The risk of fracture declined slightly after 1 year of follow-up.
Subanalysis of the association between type of CVD and major osteoporotic fracture The association between type of CVD and risk of major osteoporotic fracture was also analyzed ( Table 4). As compared with the non-CVD group, the adjusted HR of major osteoporotic fracture was highest among patients with cerebrovascular disease (HR 1.31, 95% CI 1.23, 1.39), followed by those with heart failure (1.18; 1.11, 1.27), peripheral atherosclerosis (1.12; 1.04, 1.20), and coronary artery disease (1.07; 1.01, 1.12). The chi-square test and a t-test were used to compare subjects with and without cardiovascular disease.

DISCUSSION
Although this study is not novel, it is to our knowledge the first population-based cohort study to examine the association between CVD and risk of major osteoporotic fracture in Taiwan. We selected only adults aged 50 years or older, and the prevalences of most comorbidities and medications were therefore greater among the CVD group than among the non-CVD group. The incidence of major osteoporotic fracture among women was more than 2-fold that among men,  regardless of CVD status. Our finding of a risk of major osteoporotic fracture was somewhat comparable to the results of previous studies (risk rates, 1.2 to 1.5). [7][8][9] The present study found that the overall risk of fracture in the CVD group was approximately 1.2-fold that of the non-CVD group. However, there were variations in men and women with regard to duration of follow-up and interaction between comorbidities. Additional analysis by CVD type showed that cerebrovascular disease (HR 1.31), heart failure (HR 1.18), peripheral atherosclerosis group (HR 1.12), and coronary artery disease (HR 1.07) were substantially associated with increased risk of major osteoporotic fracture. Two studies by Sennerby et al showed that Swedish patients with cerebrovascular disease had higher fracture risk (OR = 2.76 and HR = 5.09), 4,5 and our findings are similar. The authors suggested that the increased likelihood of falling after cerebrovascular disease partially explained this trend. 4,5 Variation in fracture risk during follow-up is important in preventive care. Though the absolute incidence rate of major osteoporotic fracture was higher after 1 year of follow-up, the hazard was lower during subsequent follow-up. In contrast, Sennerby et al found a higher risk of hip fracture within the first year after a CVD diagnosis in Sweden. 4,5 In the present study, the absolute incidence rate of hip fracture was higher among the CVD group, but the HR was not statistically significant (HR 1.02, 95% CI 0.92, 1.13). Although the timecourse effect differed between previous studies and the present study, we believe that clinicians should be mindful of the risk of major osteoporotic fracture even at 1 year after a CVD diagnosis, as risk did not substantially vary during the followup period.
Some limitations should be addressed. First, because claims data were used, some CVD diagnoses might not fulfill all criteria for each disease. Also, instances of fracture at hip or major skeletal sites could not be diagnosed as fragility fracture. This limitation may have led to underestimation of fracture risk. Second, because claims data did not provide information on bone mineral density (BMD), we were unable to evaluate fracture risk in relation to presence of osteoporosis, osteopenia, or normal BMD. Therefore, we were unable to examine the relation between CVD and BMD. Third, although there was no record of body mass index (BMI) in the claims data, we were able to include information on obesity (ie, ICD-9 278.00 and 278.01 and A-code A183) as a comorbidity in the analysis.

Conclusion
After careful adjustment for covariates and consideration of duration of follow-up, CVD was associated with increased risk of major osteoporotic fracture in both sexes, although the incidence of such fractures was higher among women than among men.