FRAX predicts cardiovascular risk in women undergoing osteoporosis screening: the Manitoba bone mineral density registry

Abstract

Osteoporosis and cardiovascular disease (CVD) are highly prevalent in older women, with increasing evidence for shared risk factors and pathogenesis. Although FRAX was developed for the assessment of fracture risk, we hypothesized that it might also provide information on CVD risk. To test the ability of the FRAX tool and FRAX-defined risk factors to predict incident CVD in women undergoing osteoporosis screening with DXA, we performed a retrospective prognostic cohort study which included women aged 50 yr or older with a baseline DXA scan in the Manitoba Bone Mineral Density Registry between March 31, 1999 and March 31, 2018. FRAX scores for major osteoporotic fracture (MOF) were calculated on all participants. Incident MOF and major adverse CV events (MACE; hospitalized acute myocardial infarction [AMI], hospitalized non-hemorrhagic cerebrovascular disease [CVA], or all-cause death) were ascertained from linkage to population-based healthcare data.

The study population comprised 59 696 women (mean age 65.7 ± 9.4 yr). Over mean 8.7 yr of observation, 6021 (10.1%) had MOF, 12 277 women (20.6%) had MACE, 2274 (3.8%) had AMI, 2061 (3.5%) had CVA, and 10 253 (17.2%) died. MACE rates per 1000 person-years by FRAX risk categories low (10-yr predicted MOF <10%), moderate (10%–19.9%) and high (≥20%) were 13.5, 34.0, and 64.6, respectively. Although weaker than the association with incident MOF, increasing FRAX quintile was associated with increasing risk for MACE (all P-trend <.001), even after excluding prior CVD and adjusting for age. HR for MACE per SD increase in FRAX was 1.99 (95%CI, 1.96–2.02). All FRAX-defined risk factors (except parental hip fracture and lower BMI) were independently associated with higher non-death CV events.

Although FRAX is intended for fracture risk prediction, it has predictive value for cardiovascular risk.

Osteoporosis and cardiovascular disease are common in older individuals and have shared risk factors. FRAX is a tool that was developed to assess an individual’s risk of fracture. In this study, we tested the ability of FRAX to predict heart attacks and stroke.

We found that although weaker than the association with incident fracture, increasing FRAX score was associated with increasing risk for heart attack and stroke. Thus, although FRAX is intended for fracture risk prediction, it has predictive value for cardiovascular disease including heart attack and stroke.

Introduction

Chronic noncommunicable diseases have emerged as the predominant risk to global health and account for nearly three-quarters of deaths worldwide, with cardiovascular diseases (CVDs), including heart attacks and strokes, at the top of this list.¹ Osteoporotic fractures are more common than heart attacks and strokes combined and are not only associated with significant morbidity but also increased mortality risk, two-thirds of which is due to CVD.^2-6

There are well established treatments that reduce the risk of CVD or osteoporotic fractures, and treatment is guided by individual risk.^7,8 Because no single risk factor predicts future CV events or osteoporotic fractures well, risk assessments combining risk factors have been developed, which provide individuals with a quantitative estimate of their future risk of developing a disease or event within a specific time frame.^9,10 Primary and secondary prevention guidelines for both diseases recommend using risk prediction tools to determine those at highest risk and most likely to benefit from pharmacological interventions.^11-17

However, despite the recommended use of validated risk prediction tools and availability of evidence-based treatments to prevent CV events and osteoporotic fractures, screening rates for both diseases are low. BMD testing rates for those potentially at high risk for osteoporosis have been found to be between 12% and 56%, with only 29% of those tested having had a fracture risk calculated,^18-20 and survey results of primary care practitioners found that most do not routinely use CV risk scores.^21,22 An important barrier that has been identified by physicians to the adoption of risk prediction scores is the lack of time and high workload.^23,24 Risk prediction scores are now available for numerous diseases beyond CVD and osteoporosis/fractures, including diabetes mellitus, depression, dementia, chronic kidney disease, falls.²⁵ With the current demands put on primary care physicians, the implementation of all or even a majority of risk prediction tools is likely not feasible, even for the most motivated.

One potential solution is to find risk prediction tools that can predict more than one disease or event, thereby increasing efficiency in clinical practice. There are several shared risk factors for CVD and osteoporotic fractures such as age, sex, estrogen deprivation status, smoking, and concomitant diseases such as rheumatoid arthritis and diabetes,^26-28 raising the possibility that fracture prediction tools may also predict CVD. In this study, we evaluate the utility of the most widely-used fracture risk prediction tool, FRAX®,²⁹ to predict incident major adverse cardiovascular events (MACEs).

Methods

Data sources

This retrospective population-level cohort study was conducted in Manitoba, the fifth most populous province in Canada with a population of 1.4 million in 2022.³⁰ Canada has a publicly funded and administered healthcare system that is managed by each province or territory.³¹ In Manitoba, Manitoba Health is responsible for the administration of public healthcare, which covers the full cost of DXA scans. Each resident is assigned a unique personal health identification number that is used to access the public healthcare system.³²

The cohort was defined using the Manitoba Bone Mineral Density (BMD) Program, which performs all DXAs in Manitoba. Since 1990, the linked Manitoba BMD Registry, which has previously been validated to be highly accurate and complete, has maintained a database of all DXA results and associated demographic and clinical data, including personal health identification number.³³We included all women aged 50 yr or older who had a first (baseline) fan-beam hip DXA (GE Lunar Prodigy or iDXA) recorded in the Manitoba BMD Registry between February 28, 1999 and March 29, 2018. We limited our analysis to women as they represent the vast majority of the Manitoba BMD registry (~90%).³⁴ We excluded those not registered for health care in Manitoba, with less than 3 yr of coverage prior to DXA measurement, or who were without coverage at the time of DXA as we would not be able to accurately ascertain relevant baseline covariates or outcomes in these individuals.

The cohort was linked to province-wide healthcare administrative databases using the anonymized personal health identification number. Information on healthcare visits were obtained from the Physician Claims Database (PCD) and the Discharge Abstract Database (DAD). The PCD records the date and type of service and the associated diagnosis codes using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes. The DAD, which captures up to 25 diagnoses per acute-care hospitalization, used the ICD-9-CM prior to 2004 and International Classification of Diseases, Tenth Revision, Canadian Enhancements (ICD-10-CA) after 2004. Medication use was ascertained from the provincial pharmacy database, which captures all outpatient medications dispensed in the province.³⁵

Deaths were ascertained from the Vital Statistic registry, which records all deaths that take place in Manitoba.³⁶ The study was approved by the Health Research Ethics Board for the University of Manitoba, and approval for data access was provided by the Manitoba Health Information Privacy Committee.

Fracture probability assessment

FRAX estimates 10-yr probability of major osteoporotic fractures (MOFs) based on age, sex, weight, height, previous fracture, parental history of hip fracture, current smoking, high alcohol use, prolonged glucocorticoid exposure, comorbid conditions such as rheumatoid arthritis and other secondary causes of osteoporosis, and femoral neck BMD and considers death as a competing risk.³⁷ Ten-year probabilities of MOF were calculated for each individual using the country-specific (Canadian) FRAX tool (FRAX Desktop Multi-Patient Entry version 3.8).³⁸ Height and weight were measured at each DXA encounter. Age, sex, smoking and alcohol exposure, family history, and comorbidities were reported by patients through the intake questionnaire that is completed prior to each DXA. Data on comorbidities and prior fracture were supplemented by data from the above-described linked provincial population-based healthcare databases. Prolonged glucocorticoid exposure, defined as exposure for more than 3 mo in the prior year, was ascertained from both the intake questionnaire and the provincial pharmacy system.³⁵

Incident fractures

MOFs were defined as non-traumatic fractures of the hip, vertebrae, humerus or forearm. Incident MOFs were ascertained from the DAD and the PCD up to March 30, 2018. We used previously validated fracture-site specific algorithms which required a combination of inpatient and outpatient fracture-specific encounters and excluded high trauma and pathologic fractures codes.^39,40 Hip and forearm fractures required site-specific reduction, fixation or casting codes and codes of the same fracture type within 6 mo of an incident fracture were considered to represent the same fracture. Fracture date was defined as the first clinical encounter for the fracture.

Incident cardiovascular events

MACE included hospitalized acute myocardial infarction (AMI), hospitalized non-hemorrhagic cerebrovascular disease (CVA), and all-cause mortality, per previous administrative database studies’ definitions.^41-45 Incident MACE were ascertained from the DAD up to March 30, 2018. We examined each component of MACE separately and aggregated. The date of MACE was defined as the date of first AMI diagnosis, CVA diagnosis, or death from any cause. Prior CVD (occurring before the index DXA date) was ascertained in the same way.

Statistical analyses

Baseline characteristics were compared between those with MACE and those without using one-way ANOVA for continuous variables that met ANOVA assumptions (age, BMI, BMD) and the non-parametric Mann–Whitney test for those that did not (FRAX score and observation time). Chi-square tests of independence were used for categorical variables. Observed crude rates of MOF, MACE, AMI, CVA, and death per 1000 person-years were calculated by FRAX risk category: low (MOF 10-yr risk <10%), moderate (10%–20%), and high (>20%).^12,46 Pearson Chi-square test was used to examine correlation between MOF and MACE events in the full cohort and stratified by FRAX MOF quintile and risk category.

We estimated hazard ratios (HRs) and 95% CIs for FRAX MOF and individual FRAX-defined risk factors to predict MOF, MACE, AMI, CVA, and death using multivariable survival regression models that considered the competing risk of death (in all the models except the one exploring outcome death).^47,48 FRAX MOF was log-transformed and modeled on the continuous scale and categorized by SD and quintiles. Proportional hazards assumption was assessed using Schoenfeld residuals. We specifically chose to model HR per SD and quintiles of FRAX over the area under the curve due to the limitations of using receiver operating characteristic curves for the evaluation of risk prediction scores.⁴⁹

To examine FRAX in a primary prevention setting, we performed sensitivity analysis excluding those with prior CVD. Extending this conservative approach, we further adjusted for age, since age is a strong predictor of both MOF and MACE, to evaluate the performance of FRAX beyond its dependence on age. Statistical analyses were performed with Statistica (Version 13.0, StatSoft Inc., Tulsa, OK).

Results

The study population contained 59 696 women (mean age 65.7 ± 9.4 yr). Over a mean of 8.7 yr of observation, 6021 women (10.1%) had MOF, 12 277 (20.6%) had MACE, 2274 (3.8%) had AMI, 2061 (3.5%) had CVA, and 10 253 (17.2%) died. About 2290 women (3.84%) experienced both MOF and MACE during the follow-up period, which exceeded the fraction expected if these events were independent (2.07%). The ratio of the observed overlap percentage compared to the expected overlap percentage if these two events were not correlated was 1.85 (Pearson Chi-square P-value <.001). A similar pattern was seen when stratified by FRAX MOF quintiles and risk categories, all P-values <.001. Those who had incident MACE were older, had lower BMI, were more likely to be current smokers, have high alcohol intake, been exposed to prolonged glucocorticoids, and have rheumatoid arthritis and other secondary causes of osteoporosis (all P-values <.02, Table 1). They were also more likely to have a lower femoral neck T-score, a higher FRAX predicted 10-yr risk of MOF, and a history of prior clinical fracture, prior MOF, AMI, and CVA. Observation time was on average 0.4 yr shorter in those with incident MACE (P-value <.001).

Table 1.

Baseline characteristics.

	No incident MACE N = 47 419 n (%)	Incident MACE N = 12 277 n (%)	P-value
Age (yr), mean ± SD	64.0 ± 8.6	71.9 ± 9.6	<.001
BMI (kg/m2), mean ± SD	27.5 ± 5.6	26.9 ± 5.6	<.001
Prior clinical fracture	8912 (18.8)	3232 (26.3)	<.001
Parental hip fracture	4262 (9.0)	529 (4.3)	<.001
Current smoker	4440 (9.4)	1519 (12.4)	<.001
Prolonged glucocorticoid use	1337 (2.8)	754 (6.1)	<.001
Rheumatoid arthritis	1284 (2.7)	380 (3.1)	.020
Secondary causes of osteoporosis	6106 (12.9)	2179 (17.7)	<.001
High alcohol intake	150 (0.3)	115 (0.9)	<.001
Femoral neck T-score, mean ± SD	−1.3 ± 0.9	−1.8 ± 1.0	<.001
FRAX MOF %, mean ± SD	9.7 ± 6.2	14.7 ± 8.7	<.001
Prior MOF	7237 (15.3)	2734 (22.3)	<.001
Prior AMI	788 (1.7)	690 (5.6)	<.001
Prior CVA	568 (1.2)	596 (4.9)	<.001
Observation time (yr), mean ± SD	8.8 ± 5.4	8.5 ± 4.7	<.001

AMI = hospitalized acute myocardial infarction; BMI = body mass index; CVA = hospitalized non-hemorrhagic cerebrovascular disease; MACE = major adverse cardiovascular events; MOF = major osteoporotic fracture.

Observed MOF incidence rates per 1000 person-years stratified by FRAX 10-yr MOF risk category fell within the 10-yr FRAX predicted MOF percentages: 6.9/1000 person-years in low risk (<10%), 16.3/1000 person-years in moderate risk (10%–20%), and 30.9/1000 person-years in high risk (>20%). Observed MACE incidence rates per 1000 person-years stratified by FRAX 10-yr MOF risk category were above the FRAX-predicted MOF percentages: 13.5 in low risk, 34.0 in moderate risk, and 64.6 in high risk (Table S1). This was largely driven by all-cause death, although AMI and CVA incidence rates increased similarly with each FRAX risk category (Figure 1).

Figure 1.

MOF and cardiovascular outcome rates per 1000 person-years (with 95% upper CI error bars) by FRAX risk category. MOF = major osteoporotic fracture

As presented in Table 2, the HR for MOF increased with increasing FRAX quintile (reference, 1.34, 2.19, 3.24, 6.29, P-value for trend <.001). Similarly, the HR ratio for MACE increased with increasing FRAX quintile (reference, 1.65, 2.42, 3.77, 7.23, P-value for trend <.001). The HR for each component of MACE also increased with increasing FRAX quintile (P-value for trend <.001 for AMI, CVA, and death). Sensitivity analysis performed on those who did not have prior diagnosed CVD showed similar HRs to the original analysis. After adjustment for age, the HR for MACE and its component diagnoses continued to increase with increasing FRAX quintile. Although the estimates were lower than the unadjusted estimates, the P-value for trend for the outcome CVA was no longer statistically significant after the exclusion of those with prior CVD and adjustment for age (P-value .053).

Table 2.

Hazard ratios (HR) for MOF and cardiovascular outcomes by FRAX quintile.

Outcome		HR (95% CI)
	Unadjusted	Excluding prior CVD	Adjusted for age, excluding prior CVD
MOF
First (lowest)	1 (referent)	1 (referent)	1 (referent)
Second	1.34 (1.20–1.50)	1.34 (1.20–1.50)	1.17 (1.05–1.32)
Third	2.19 (1.98–2.43)	2.13 (1.92–2.37)	1.74 (1.56–1.95)
Fourth	3.24 (2.94–3.57)	3.08 (2.79–3.40)	2.34 (2.09–2.62)
Fifth (highest)	6.29 (5.75–6.89)	6.03 (5.49–6.63)	4.10 (3.64–4.62)
P-trend	<.001	<.001	<.001
MACE
First (lowest)	1 (referent)	1 (referent)	1 (referent)
Second	1.65 (1.53–1.79)	1.65 (1.52–1.80)	0.98 (0.90–1.07)
Third	2.42 (2.24–2.61)	2.35 (2.17–2.55)	1.05 (0.97–1.15)
Fourth	3.77 (3.52–4.05)	3.64 (3.37–3.93)	1.22 (1.11–1.33)
Fifth(highest)	7.23 (6.76–7.73)	6.92 (6.44–7.44)	1.50 (1.36–1.65)
P-trend	<.001	<.001	<.001
AMI
First (lowest)	1 (referent)	1 (referent)	1 (referent)
Second	1.68 (1.42–1.99)	1.62 (1.35–1.95)	1.12 (0.92–1.35)
Third	2.08 (1.77–2.45)	1.98 (1.66–2.38)	1.13 (0.93–1.37)
Fourth	3.02 (2.59–3.53)	2.95 (2.49–3.5)	1.36 (1.11–1.66)
Fifth (highest)	4.72 (4.08–5.48)	4.36 (3.7–5.14)	1.47 (1.18–1.84)
P-trend	<.001	<.001	<.001
CVA
First (lowest)	1 (referent)	1 (referent)	1 (referent)
Second	1.86 (1.52–2.26)	1.82 (1.47–2.25)	1.02 (0.82–1.27)
Third	2.76 (2.29–3.33)	2.61 (2.13–3.2)	1.08 (0.87–1.35)
Fourth	4.12 (3.45–4.93)	3.87 (3.19–4.7)	1.16 (0.93–1.46)
Fifth (highest)	6.90 (5.81–8.19)	6.53 (5.42–7.87)	1.21 (0.94–1.54)
P-trend	<.001	<.001	.053
Death
First (lowest)	1 (referent)	1 (referent)	1 (referent)
Second	1.61 (1.47–1.77)	1.61 (1.46–1.77)	0.91 (0.83–1.01)
Third	2.54 (2.33–2.77)	2.46 (2.24–2.70)	1.03 (0.93–1.14)
Fourth	4.24 (3.92–4.60)	4.05 (3.71–4.42)	1.23 (1.11–1.36)
Fifth (highest)	8.53 (7.91–9.21)	8.12 (7.48–8.82)	1.56 (1.40–1.74)
P-trend	<.001	<.001	<.001

AMI = hospitalized acute myocardial infarction; CVA = hospitalized non-hemorrhagic cerebrovascular disease; MACE = major adverse cardiovascular event; MOF = major osteoporotic fracture.

The HR for MOF and MACE, per SD in FRAX was similar (MOF, HR 1.97, 95% CI, 1.92–2.02; MACE, HR 1.99, 95% CI, 1.96–2.02; Table 3). The HRs for each component of MACE per SD in FRAX were all significant at an alpha of 0.05 (AMI 1.68, CVA 1.87, death 2.13). These remained significant even after excluding those with prior CVD, with the estimates remaining nearly unchanged. After adjusting for age, the HRs for MOF and MACE, per SD in FRAX, decreased for all measured outcomes, but remained significant.

Table 3.

Hazard ratios (HR) for MOF and cardiovascular outcomes per SD in FRAX.

Outcome	HR (95% CI)
	Unadjusted	Excluding prior CVD	Adjusted for age, excluding prior CVD
MOF	1.97 (1.92–2.02)	1.95 (1.90–2.00)	1.78 (1.72–1.85)
MACE	1.99 (1.96–2.02)	1.97 (1.94–2.01)	1.23 (1.19–1.26)
AMI	1.68 (1.62–1.75)	1.67 (1.59–1.75)	1.18 (1.10–1.26)
CVA	1.87 (1.79–1.95)	1.87 (1.78–1.96)	1.08 (1.01–1.17)
Death	2.13 (2.09–2.17)	2.11 (2.06–2.15)	1.28 (1.24–1.32)

AMI = hospitalized acute myocardial infarction; CVA = hospitalized non-hemorrhagic cerebrovascular disease; MACE = major adverse cardiovascular event; MOF = major osteoporotic fracture.

When examining the individual components of FRAX in predicting MACE (Table 4), high alcohol intake was the most important predictor of MACE (HR 3.00, 95% CI, 2.50–3.61), followed by age (HR per 10 yr of age 2.38, 95% CI, 2.33–2.43) and smoking (HR 2.16, 95% CI, 2.05–2.28). Other FRAX variables associated with increased hazard of MACE included increasing BMI, prior fracture, glucocorticoid use, rheumatoid arthritis, secondary causes of osteoporosis, and decreasing femoral neck T-score. Parental hip fracture was the only component of FRAX that was not independently associated with increased hazard of MACE and in fact was associated with reduced risk of MACE (HR 0.80, 95% 0.73–0.87). These associations remained nearly unchanged after excluding those with prior CVD.

Table 4.

Hazard ratio (HR) for major adverse cardiovascular event (MACE) by FRAX variable.

FRAX variable	HR (95% CI)
	Adjusteda	Excluding prior CVD
Age per 10 yr	2.38 (2.33–2.43)	2.40 (2.34–2.46)
BMI per 5 kg/m2	1.05 (1.03–1.07)	1.04 (1.02–1.06)
Prior fracture	1.15 (1.10–1.20)	1.13 (1.08–1.19)
Parental hip fracture	0.80 (0.73–0.87)	0.82 (0.74–0.90)
Smoker	2.16 (2.05–2.28)	2.19 (2.06–2.33)
Glucocorticoid use	1.67 (1.55–1.80)	1.78 (1.63–1.94)
Rheumatoid arthritis	1.40 (1.26–1.55)	1.33 (1.18–1.50)
Secondary osteoporosis	1.79 (1.71–1.88)	1.72 (1.62–1.82)
High alcohol intake	3.00 (2.50–3.61)	2.98 (2.41–3.68)
Femoral neck T-scoreb	1.13 (1.10–1.15)	1.11 (1.08–1.14)

BMI = body mass index; CVD = cardiovascular disease.

^aAdjusted for all variables listed above.

^bPer SD decrease.

Discussion

Although FRAX is intended for fracture risk prediction, we found that it can concurrently predict cardiovascular risk. In fact, the HR for MACE by increasing FRAX quintiles is higher than that for MOF at each FRAX quintile and there is approximately a 2-fold increase in hazard of both MOF and MACE for every SD increase in FRAX. Importantly, FRAX performed equally well in sensitivity analysis excluding those with prior CVD, exhibiting its utility in a primary prevention setting, analogous to that of Framingham Risk Score (FRS).⁵⁰

MACE rates per 1000 person-years were well stratified by FRAX clinical risk categories. Of greater clinical relevance is the absolute risk over 10 yr, which is the output generated by both FRS and FRAX. Interestingly, the cut off in most healthcare settings for high risk is >20% for both risk prediction tools. The event rate per 1000 person-years tracked closely with the 10-yr risk, as evidenced by the MOF incidence rate per 1000 person-years tracking closely with the FRAX-predicted 10-yr MOF risk. FRAX discriminated individuals at low, moderate, and high risk of MACE, but FRAX-predicted MOF probabilities were lower than MACE incidence rates and could not be directly used to predict MACE 10-yr risk.

CVD and osteoporosis were once considered unrelated and independent diseases. There is now an established body of evidence that CVD and osteoporosis share epidemiologic associations, common risk factors, and pathophysiologic pathways. One Taiwanese study found a 2-fold increase in the incidence of osteoporosis in those with CAD (adjusted HR 2.04, 95% CI, 1.99–2.08).⁵¹ The MrOS Sweden study found that men with peripheral arterial disease were almost two times more likely to suffer a hip fracture after adjustment for age and that this association is independent of hip BMD.⁵² Results of the Bushehr Elderly Health Program showed a significant association between WHO CVD risk score and osteosarcopenia.⁵³ Lower BMD has been shown to be associated with CVD, including subclinical carotid atherosclerosis, arterial stiffness and aortic calcification, independent of other clinical risk factors.^54-57 Previous studies have demonstrated an association between higher FRAX-assessed 10-yr MOF risk and higher odds of having coronary artery calcification and CVD in women.^58-60

The wide adoption of FRS and other CVD risk prediction tools is limited by studies reporting a lack of generalizability to populations outside of Framingham, given that the probability of CVD varies around the world.^25,61,62 Studies outside of the USA have found that FRS and newer CVD risk prediction tools modeled on US data tend to overestimate CVD risk.^61,62 To this point, FRAX models are calibrated to the specific epidemiology of fracture and death for that country/population.²⁹ However, whether the gradient of CVD risk is similar between countries is unknown. Although FRS only considers age, sex, cholesterol level, systolic blood pressure, smoking status, and diabetes, it appears that the variables in FRAX, other than parental hip fracture, independently predict MACE as well and may augment FRS performance. These variables include traditionally considered osteoporosis-specific risk factors such as prior fracture, secondary osteoporosis, and femoral neck T-score. Furthermore, we showed that FRAX was associated with MACE not only through its inclusion of age or age-associated risk factors, as evidenced by the significance of our age-adjusted model. Expectantly, the effect estimates were reduced without age as a predictive factor, as would all risk prediction scores that include age if adjusted for age.

Limitations

Since our cohort is derived from those presenting for BMD testing, our elderly (mean age 66 yr) cohort represents those at risk of osteoporosis and is therefore limited in its generalizability to the general population. Conversely, our cohort is a referral population which would reflect an appropriate real-world clinical setting. We limited our analysis to women³⁴ and thus these results cannot be generalized to men. Interestingly, parental hip fracture appeared to be protective against MACE. This may reflect a selection bias of patients presenting for BMD testing due to a parental history of hip fracture, but who did not have other personal risk factors for osteoporosis or MACE, resulting in the appearance of a protective effect of parental hip fractures. Moreover, parents who sustain hip fractures may have lower risk for CVD and death, as they must live long enough to sustain a hip fracture, which typically occurs in the eighth decade of life.⁶³ These CVD protective factors may be passed onto their offspring.

Another limitation of our study is that our MACE definition includes all-cause death as one of the three component events instead of cardiac death. This will drive the MACE incidence rates higher than when only considering cardiac death, likely explaining the high MACE rates seen in our study. If we only consider the AMI and CVA rates, the sum of these rates in each of the FRAX risk categories (5.2/1000-person years in low risk, 11.8 in moderate, and 20.1 in high) is just under the observed MOF rates and within the 10-yr FRAX-predicted MOF risk categories (low <10%, moderate 10%–20%, high >20%). Furthermore, non-hospitalized CVD was not ascertained in this study as we only considered hospitalized AMI and CVA. Unfortunately, we did not have access to clinical data on cholesterol level and systolic blood pressure, which would have allowed us to calculate FRS and directly compare FRAX performance to that of FRS.

Conclusions

Shared risk factors are not unique to CVD and osteoporosis. In fact, epidemiologic studies have found that there are several key risk factors associated with non-communicable diseases such as physical inactivity, poor nutrition, socioeconomic status, smoking, and excessive alcohol intake.^64-67 Theoretically, these findings support the development of a single risk prediction score for multiple chronic NCDs. Practically, our study shows that FRAX, a tool developed specifically to stratify an individual’s risk of one specific NCD (osteoporotic fractures), has predictive value for another NCD (CVD). Further studies are needed to compare the performance of FRAX against established CVD prediction tools such as FRS, as well as to examine other areas of overlap, with the goal to develop a shared risk prediction tool for multiple NCDs. Such a tool may improve the uptake and integration of risk prediction scoring into clinical practice across multiple chronic diseases.

Author contributions

Carrie Ye (Methodology, Writing—original draft, Writing—review & editing), John T. Schousboe (Conceptualization, Methodology, Writing—review & editing), Suzanne N. Morin (Writing—review & editing), Lisa M. Lix (Writing—review & editing), Eugene V. McCloskey (Writing—review & editing), Helena Johansson (Writing—review & editing), Nicholas C. Harvey (Writing—review & editing), John A. Kanis (Writing—review & editing), and William D. Leslie (Conceptualization, Data curation, Formal analysis, Methodology, Writing—review & editing)

Funding

None declared.

Conflicts of interest

C.Y., J.T.S., S.N.M., L.M.L., H.J., W.D.L. have nothing to disclose.

E.V.M.: Nothing to disclose for the context of this paper, but numerous ad hoc consultancies/ speaking honoraria and/or research funding from Amgen, Bayer, Fresenius Kabi, General Electric, GSK, Hologic, Lilly, Merck Research Labs, Novartis, Novo Nordisk, Nycomed, ObsEva, Pfizer, and UCB.

N.C.H.: Nothing to disclose for the context of this paper, but has received consultancy/ lecture fees/ honoraria/ grant funding from Alliance for Better Bone Health, Amgen, MSD, Eli Lilly, Servier, Shire, UCB, Kyowa Kirin, Consilient Healthcare, Theramex and Internis Pharma.

J.A.K.: Director of Osteoporosis Research Ltd, a company that develops and maintains FRAX.

Data availability

Data sharing is not permitted under the Researcher Agreement with Manitoba Health and Seniors Care (MHASC). However, researchers may apply for data access through the Health Research Ethics Board for the University of Manitoba and the Health Information and Privacy Committee of MHASC.