Identification of High-Risk Individuals for Hip Fracture: A 14-Year Prospective Study

Authors

  • Nguyen D Nguyen,

    1. Bone and Mineral Research Program, Garvan Institute of Medical Research, St Vincent's Hospital, University of New South Wales, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Chatlert Pongchaiyakul,

    1. Bone and Mineral Research Program, Garvan Institute of Medical Research, St Vincent's Hospital, University of New South Wales, Sydney, New South Wales, Australia
    2. Division of Endocrinology and Metabolism, Department of Medicine, Faculty of Medicine Khon Kaen University, Khon Kaen, Thailand
    Search for more papers by this author
  • Jacqueline R Center,

    1. Bone and Mineral Research Program, Garvan Institute of Medical Research, St Vincent's Hospital, University of New South Wales, Sydney, New South Wales, Australia
    Search for more papers by this author
  • John A Eisman,

    1. Bone and Mineral Research Program, Garvan Institute of Medical Research, St Vincent's Hospital, University of New South Wales, Sydney, New South Wales, Australia
    Search for more papers by this author
  • Tuan V Nguyen PhD

    Corresponding author
    1. Bone and Mineral Research Program, Garvan Institute of Medical Research, St Vincent's Hospital, University of New South Wales, Sydney, New South Wales, Australia
    • Bone and Mineral Research Program, Garvan Institute of Medical Research, St Vincent's Hospital, 384 Victoria Street, Darlinghurst, Sydney, NSW 2010, Australia
    Search for more papers by this author

  • The authors have no conflict of interest.

Abstract

In this 14-year prospective study, men and women were found to share a common set of risk factors for hip fracture: low BMD, postural instability and/or quadriceps weakness, a history of falls, and prior fracture. The combination of these risk factors accounted for 57% and 37% of hip fractures in women and men, respectively.

Introduction: Risk factors for hip fracture, including low BMD, identified in women, have not been shown to be useful in men. It is also not known whether fall-related factors (muscle strength and postural instability) predict hip fracture. This study examined the association between falls-related factors and hip fractures in elderly men and women.

Materials and Methods: This is an epidemiologic, community-based prospective study, which included 960 women and 689 men ≥60 years of age who have been followed for a median of 12 years (interquartile range, 6–13). The number of person-years was 9961 for women and 4463 for men. The outcome measure was incidence of hip fracture. Risk factors were femoral neck BMD (FNBMD), postural sway, quadriceps strength, prior fracture, and fall.

Results: Between 1989 and 2003, 115 (86 women) sustained a hip fracture. The risk of hip fracture (as measured by hazards ratio [HR]) was increased by 3.6-fold (95% CI: 2.6–4.5) in women and 3.4-fold (95% CI: 2.5–4.6) in men for each SD (0.12 g/cm2) reduction in FNBMD. After adjusting for BMD, the risk of hip fracture was also increased in individuals with the highest tertile of postural sway (HR: 2.7; 95% CI: 1.6–4.5) and low tertiles of quadriceps strength (HR: 3.0; 95% CI: 1.3–6.8). Furthermore, a history of fall during the preceding 12 months and a history of fracture were independent predictors of hip fracture. For each level of BMD, the risk of hip fracture increased linearly with the number of non-BMD risk factors. Approximately 57% and 37% of hip fracture cases in women and men, respectively, were attributable to the presence of risk factors, osteoporosis (BMD T score ≤ −2.5), and advancing age.

Conclusions: Men and women had a common set of risk factors for hip fracture: low BMD, postural instability and/or quadriceps weakness, a history of falls, and prior fracture. Preventive strategies should simultaneously target reducing falls and improvement of bone strength in both men and women.

INTRODUCTION

THE LIFE-TIME risk of hip fracture for a white woman 50 years of age is ∼15%, equivalent to the risk of developing breast cancer.(1) Hip fracture causes considerable disability, morbidity, mortality, and incurs significant costs,(2) and its incidence increases exponentially with advancing age.(3,4) Approximately 30% of hip fractures occur in men,(5,6) with a life-time risk being 5–6%.(6) Although the incidence and determinants of hip fracture in women have been studied extensively, comparable data have not been well documented in men. In women, measurement of BMD, particularly at the femoral neck, is a predictor of hip fracture, with each SD lower BMD being associated with a 2- to 4-fold increase in hip fracture risk.(7) However, there is a substantial overlap in BMD between women with and without hip fracture at any given age, making it difficult to use BMD alone as a discriminant of fracture.

For these reasons and the rising number and proportion of the aged population, hip fracture is increasingly becoming an important public health concern. A better understanding of the risk factors for hip fracture may translate into better prophylactic strategies for reducing the problem. More than 90% of hip fracture cases are associated with falls.(8–10) Postural instability and quadriceps weakness are important predictors of falls,(11) and it has been shown previously that these factors were also predictive of all fractures in men and women.(12) It can therefore be hypothesized these are independent predictors of hip fracture risk. This study sought to assess the predictive value of fall-related factors and BMD in the identification of high-risk individuals to hip fracture and to ascertain the proportion of risk attributable to these factors.

MATERIALS AND METHODS

Setting and subjects

This study is part of an on-going longitudinal study, the Dubbo Osteoporosis Epidemiology Study (DOES); for which details of protocol and study design have been previously described.(5,12) Briefly, in 1989, all men and women ≥60 of age living in Dubbo, a city of ∼32,000 people 400 km north west of Sydney (Australia), were invited to participate in an epidemiological study. At that time, the population was comprised of 1581 men and 2095 women ≥60 years of age, of whom 98.6% were white and 1.4% were indigenous Aboriginal. These individuals were invited to participate in DOES. Dubbo was selected for the study because the age and gender distribution of the population closely resembled the Australian population, and it is relatively isolated in terms of medical care; virtually complete ascertainment of all fractures occurring in the target population is possible. This study was approved by the St Vincent's Campus Research Ethics Committee, and informed written consent was obtained from each participant.

Assessment of risk factors

Individuals were interviewed by a nurse coordinator who administered a structured questionnaire to obtain data including age, lifestyle factors such as duration of smoking intake and alcohol consumption, physical activity, any history of falls in the preceding 12 months, and any history of fractures in the past 5 years. Anthropometric variables (height, weight) were measured, and a dietary assessment was performed based on a frequency questionnaire for calcium intake as described elsewhere.(13)

Various risk factors for falling were tested on each subject at baseline. Quadriceps strength (maximum isometric contraction) was measured in the sitting position in the subject's dominant (stronger) leg with a horizontal spring gauge calibrated up to 50-kg force. This method has a reliability coefficient of 0.92.(11) Body sway was measured as displacements of the body at the level of the waist in 30-s periods. A 40-cm rod was rigidly attached to the waistband at the patient's back and, over 30 s, movements of a pen attached to the other end of the rod was recorded on a sheet of graph paper on an adjustable-height table. The area (mm2) encompassing all movements, forward and backward and left and right, was used as the sway. Four test conditions were used: eyes open and closed on a firm surface (wooden floor) and eyes open and closed on a compliant surface (high density foam 15 cm high). Full descriptions of these assessments and their test and retest reliability scores and CIs have been given elsewhere.(11) In this analysis, area of sway derived from the condition of eyes closed and on foam was used in the analysis, because it was found to be more sensitive than sway areas on other conditions.

BMD (g/cm2) was measured at the lumbar spine and femoral neck by DXA using a LUNAR DPX-L densitometer (GE-LUNAR Corp., Madison, WI, USA). The radiation dose with this method is <0.1 μGy. The CV of BMD in our institution in normal subjects was between 1.5% and 2% at the proximal femur and lumbar spine, respectively.(14) Based on the actual values of FNBMD obtained, each subject was classified as follows: osteoporotic, a BMD being 2.5 SD or more below the young normal level; osteopenic, a BMD between 2.5 and 1.1 SD below the young normal level; or normal. The “young normal” BMD was obtained from a sample of 52 Australian men and women between 20 and 32 years of age. These values are identical to those of the LUNAR white database. Volumetric BMD at the femoral neck (FNvBMD) was estimated as previously described.(15) Femoral neck volume (FNVOL) was estimated using the methods described by Faulkner et al.,(15) where it is assumed that the FNVOL is cylindrical in shape. The FNVOL was estimated by the formula π(d/2)2h, where d is the estimated diameter of the femoral neck, and h is the height of the femoral neck region of interest. As the projected area of femoral neck is based on a constant length along the axes of the neck of 1.5 cm (h), it is possible to estimate the FNvBMD at this site, by noting that the FNVOL can be expressed as a function of FNBMC and FNBMD. After some algebra, it can be shown that

equation image

The estimated FNvBMD was derived as

equation image

Assessment of hip fracture

Hip fractures occurring during the study period were identified for residents of the Dubbo local government area through radiologists' reports from the two centers providing X-ray services as previously described.(5,12) Fractures were only included if the report of fracture was definite and, on interview, had occurred with minimal trauma (fall from standing height or less). Fractures clearly caused by major trauma (such as motor vehicle accidents) or underlying diseases (such as cancer or bone-related diseases) were excluded from the analysis. Any fractures >3 months before study entry were not considered in the analysis.

Statistical analysis

Incidence of fracture was calculated as the number of fracture cases per 1000 person-years. Because hip fracture is a relatively rare event, its incidence is assumed to follow a Poisson distribution in which the mean is equal the variance. The “frequentist” 95% CI of the incidence can therefore be constructed based on this assumption. The Cox's proportional hazards regression model was used to estimate relative risk and 95% CI for each SD change or in specified groups compared with reference group with categorized risk factors. The outcomes in this model were fracture incidence and the time to fracture from baseline BMD measurement. Stepwise and backward algorithms were used to search for a model with maximum discriminatory power. The significance of parameter estimates derived from the Cox's proportional hazards model(16) was tested with the likelihood ratio statistic.(17) The assumption of proportional hazards for the levels of each risk factor was tested by evaluating the linearity of plots of log[−log(S(ti)j], where S(ti)j describes the jth survival time for the ith level (i = 1,2) for each risk factor. The contribution of risk factors to hip fracture risk was further evaluated in a descriptive analysis, in which the measurement of each non-BMD risk factor was dichotomized into two categories (presence or absence). A risk score, or more specifically, the number of risk factors was derived as the sum of all individual risk factors for each individual. Incidence and relative risk of hip fracture were computed for each risk score independent of, or in combination with, BMD values.

To estimate the proportion of hip fracture that may be hypothetically reduced by the elimination of the risk factors, population attributable risk fraction (PARF) was calculated. The PARF is a function of two parameters: the prevalence of a risk factor, and the relative risk (RR) associated with the risk factor. To estimate the prevalence, each of the independent risk factors (obtained from the first stage of analysis) was dichotomized into high-risk and low-risk groups. The RR associated with risk of hip fracture for the high-risk category was estimated from the multiple logistic model. The statistical estimation of PARF was based on the “sequential attributable fractions.”(18) Briefly, for each threshold criterion used to define high-risk individuals, the expected probability of fracture was calculated from the estimated coefficients of the multivariate logistic analysis model. The expected probability was compared with the observed probability, and components of attributable fraction were subsequently estimated for each possible combination of risk factors. All database management and statistical analyses were performed using the SAS Statistical Analysis System version 8.01.(19)

RESULTS

Nine hundred sixty women and 689 men 70.6 ± 7.2 (SD) years of age were followed for a median of 12 years (interquartile range [IQR]: 6–13) with 9961 and 6643 person-years for women and men, respectively. During the follow-up period, 86 women and 29 men sustained a hip fracture, for an incidence of 9.4 (95% CI, 5.0, 17.6) per 1000 person-years for women and 4.4 (95% CI, 1.8, 10.8) for men. Fracture subjects were older, shorter, lighter, and had lower BMD than nonfracture subjects (Table 1). In men and women, there were no significant differences in physical activity or other lifestyle factors (e.g., dietary calcium intake, coffee intake, and cigarette smoking) between hip fracture and nonfracture groups. Alcohol consumption in women with hip fractures was significantly lower than in women without a fracture; however, such a trend was not evident in the men.

Table Table 1.. Baseline Characteristics of Participants at Study Entry
original image

Univariate analysis

As expected, advancing age and low femoral neck BMD were each significantly associated with hip fracture risk for both genders. The magnitude of association was similar for women and men. Each 5-year increase in age was associated with a 2.0-fold (95% CI: 1.7–2.3) and 2.6-fold (95% CI: 2.0–3.5) increase in the risk of hip fracture for women and men, respectively. Each SD (0.12 g/cm2) lower FNBMD was associated with a 3.6-fold (95% CI: 2.9–4.5) and 3.4-fold (95% CI: 2.5–4.6) increase in risk of hip fracture for women and men, respectively. When femoral neck BMD was expressed in g/cm3 units (i.e., estimated volumetric BMD, FNvBMD), each 0.03 g/cm3 lower FNvBMD was associated with a 2-fold (95% CI: 1.8–2.4) and 2.7-fold (95% CI: 2.0–3.7) increase in the risk of hip fracture for women and men, respectively.

Although femoral neck BMD in men was significantly higher than that in women (mean difference = 0.13g/cm2, p < 0.001), the difference in FNBMD between fractured and nonfractured men (mean difference = −0.20 g/cm2; p < 0.001) was comparable with that in women (mean difference = −0.24 g/cm2; p < 0.001). It seems that the BMD - fracture risk relationship was age- and sex-dependent. For example, among those between 60 and 75 years of age with low femoral neck BMD (<0.85 g/cm2), the cumulative incidence of hip fracture in men was virtually identical to that in women, but among those with BMD being >0.85 g/cm2; the incidence in women was higher than in men. On the other hand, among those 76 years of age or older, the incidence in men was substantially higher than in women when BMD fell below 0.75 g/cm2, but the trend was reversed for BMD >0.75 g/cm2 (Fig. 1).

Figure FIG. 1..

Cumulative incidence of hip fracture in women and men classified by age (A, ≤75 years; B, >75 years) and femoral neck BMD level.

Postural instability, quadriceps weakness, a history of falls, and prior low trauma fracture were each associated with increased risk of hip fracture in men and women (Table 2). Among women with hip fracture, 18 had already sustained a prior fractures (6 occurred at the hip, 5 at the vertebrae, 2 at the forearm, and 5 in other sites). In women, a prior hip fracture was associated with a 5.4-fold (95% CI: 2.3–1.5), whereas a prior vertebral fracture was associated with a 4.6-fold (95% CI: 1.9–11.6) increase in the risk of subsequent hip fracture (data not shown). Among men with hip fracture, nine had previously sustained a prior fracture, including six hip, one forearm, and two in other sites.

Table Table 2.. Hazard Ratio (HR) for Predictive Factor: Univariate Cox's Proportional Hazard Model
original image

Multivariate analysis

There was no significant difference in the risk of hip fracture after adjusting for femoral neck BMD between men and women. Therefore, gender was considered a covariate in subsequent analyses. After adjusting for FNBMD, a history of fracture, a fall during the previous 12 months, quadriceps weakness, and postural instability were each associated with increased risk of hip fracture in men and women (Tables 3 and 4). Each SD (60 cm2) higher postural sway was associated with a 1.3-fold (95% CI: 1.2–1.5) and each SD (kg) lower in quadriceps strength was associated with 1.7-fold (95% CI: 1.4–2.2) increase in the risk of hip fracture after adjusting for femoral neck BMD and gender.

Table Table 3.. Hazard Ratio (HR) With 95% CI for Predictive Factor (Cox's Proportional Hazard Model)
original image
Table Table 4.. Hazard Ratio (HR) With 95% CI for Predictive Factor Categorized (Cox's Proportional Hazard)
original image

Postural instability, quadriceps weakness, fall history, fracture history, and age were each dichotomized in two categories: risk presence (with a score of 1) and risk absence (with a score of 0). For postural sway values and quadriceps weakness, the dichotomization was based on the gender-specific tertile distribution. Thus, any woman with a postural sway measurement >16.2 cm2 or with a quadriceps strength being lower than 15 kg was scored as 1; otherwise, a score of 0 was used. Likewise, any man with a postural sway measurement >14.5 cm2 or with a quadriceps strength <28 kg was scored as 1; otherwise a score of 0 was used. A prior fracture or a previous fall was also scored as 1. The sum of these risk factors was derived for each individual and was termed risk score. In the entire sample, the median score was 2, with minimal and maximal scores being 0 and 4, respectively.

The incidence rate of hip fracture was highest among individuals with a risk score ≥3 and osteoporotic BMD (45.8 per 1000 person-years; 95% CI: 34.3–61.1) compared with those with normal BMD and risk score being 0 (0 per 1000 person-years; 95% CI: 0–3.8). For a given BMD level, the incidence of hip fracture increased exponentially with the number of risk factors. For a given risk score category, the incidence of hip fracture also increased exponentially with the lowering of BMD levels (Fig. 2).

Figure FIG. 2..

Hip fracture incidence (per 1000 person-years) stratified by femoral neck BMD T scores and number of risk factors.

Attributable fraction

Results of the analysis of PARF suggested that in both sexes the attributable risk fraction generally increased as a function of the presence of cumulative risk factors (Table 5). Fourteen percent of women and 7% of men had all risk factors, and the probability of hip fracture associated with this risk group was 33% (in women) and 23.5% (in men). Consequently, ∼57% and 37% of hip fracture cases in women and men, respectively, were attributable to the presence of all risk factors, osteoporosis, and advancing age. When postural instability and quadriceps weakness were removed from the analysis, the presence of the remaining risk factors (i.e., a prior fracture and a previous fall), advancing age, and osteoporosis together accounted for 47% and 28% of hip fracture cases in women and men, respectively (data not shown).

Table Table 5.. Estimation of Population Attributable Risk Fraction (PARF) of Hip Fracture for Risk Factors, Osteoporosis, and Age Exposure
original image

DISCUSSION

Hip fracture, one of the most serious consequences of osteoporosis, is difficult to predict because of its relatively low incidence and multiplicity of risk factors. Our findings indicate that, apart from the established risk factor of low BMD, postural instability, quadriceps weakness, a history of fall, and a prior fracture independently contribute to the identification of the “at-risk” individuals. Indeed, men and women with multiple risk factors had the highest risk of hip fracture, at any given BMD level.

Although it has been shown previously that postural sway and quadriceps weakness were independent predictors of any fracture,(12) to our knowledge, this is the first empirical demonstration that postural instability and quadriceps weakness are specifically independent predictors of hip fracture risk in elderly men and women. The magnitude of association between these two risk factors and hip fracture risk was slightly lower than the magnitude of association between BMD and fracture risk. However, given the BMD-independent contribution of the fall-related factors to hip fracture risk, it could be argued that the proportion of women and men who are at-risk of hip fracture at any age and time-point is greater than the proportion identified by low BMD alone. These data suggest that the concept that hip fracture prevention should be expanded beyond the improvement of BMD to include falls prevention as a primary component.

The association between postural sway and quadriceps weakness and fracture is presumably mediated through falls. Increased body sway is a measure of postural instability(20,21) and a predictor of falls.(22) This is presumably because of increased sway being related to impairment of tactile sensitivity, joint position sense, reaction time alone, or in combinations.(11) Quadriceps strength and muscle functions generate force-producing movement and also have a role in proprioception.(20) Moreover, muscle weakness is associated with a decreased muscle mass, which has been reported to have an independent relationship to the risk of hip fracture.(23)

This study showed that a history of preceding fall was also an independent risk for hip fracture. Indeed, fallers had 1.5-fold increased risk for hip fracture compared with nonfallers even after adjustment for FNBMD. Alcohol consumption or sedative use was not associated with an increased risk of fall. In this study the risk of fall increased with advancing age, such that by the age of 70+, 35% of men and 52% of women reported to have fallen at least once in the previous 12 months. As a result, in the presence of age, the effect of fall on fracture risk was only marginally significant.

The relative risk of hip fracture for each SD lower FNBMD in men was comparable with women, suggesting that absolute BMD predicts hip fracture risk equally in men and women, which is consistent with the cross-sectional results of de Laet et al.(24) The finding that the predictive value of BMD in men was equivalent to that in women implies that intervention to increase BMD in men could reduce fracture risk. Indeed, Orwoll et al.(25) found that alendronate (an antiresorptive agent) treatment men with osteoporosis increased BMD and reduced vertebral fracture risk from 7.1% in the placebo group to 0.8% in the treated group.

The BMD-fracture risk relationships in men and women deserves a comment. In the younger age group (≤75 years old) and low BMD group there was virtually no sex-related difference in the cumulative incidence of fracture; however, among those whose BMD >0.9 g/cm2, the incidence of fracture in women was higher than in men. In the older age group, the cumulative incidence of fracture in men was higher than in women for BMD <0.75 g/cm2, whereas for BMD <0.75 g/cm2, the trend was reversed. This paradox is apparently because of the fact that BMD in men, on average, is higher than in women by 20% (p < 0.001).

In this study, a history of prior fracture was a strong predictor of subsequent hip fracture, consistent with previous studies in women.(26–28) However, this study further showed that that this is also an important risk factor for hip fracture in men. The mechanism of this relationship is not clear. In this study, subjects with prior fracture fell more often, had lower quadriceps strength, and worse postural sway. Therefore, it seems likely that the association is mediated through fall-related factors. In this study, the risk of hip fracture increased almost exponentially with the number of risk factors. However, the gradient of risk was highest in individuals with low (osteoporotic) BMD. This suggests that the combination of multiple risk factors with BMD measurement could greatly improve the sensitivity and predictive value for hip fracture.

When osteoporosis was defined in terms of BMD measurements (i.e., T scores ≤ −2.5), it is important to note that individuals with osteoporosis often had other risk factors such as advancing age and the presence of at least one non-BMD risk factor. For example, ∼25% of women and 15% of men in this study were classified as having osteoporosis; however, among this osteoporotic group, 14% of women and 7% of men also had at least one risk factor and advancing age. Therefore, the estimated population attributable risk fraction was not perfectly additive in terms of the number of risk factors. Nevertheless, the fact that the group with the highest number of risk factors (i.e., osteoporosis and advancing age and at least one non-BMD risk factor) had the highest risk of hip fracture (33% in women and 23.5% in men) suggests that the incorporation of non-BMD information could potentially improve the predictive value of a case-finding model for identifying high-risk individuals.

The study was based on a relatively large sample size with a long duration of follow-up, which allowed the detection of associations not possible in smaller or shorter duration studies. However, the number of hip fractures in men is relatively small (n = 29), which could limit the statistical power to detect truly independent predictors. In addition, the effects of potential confounders (such as impairment of vision, functional status, and daily living activities) not considered in the study may have compromised the interpretation. The population is of white background; therefore, extrapolation to other populations should be made with caution. Measurements of BMD, strength, and sway at any single time-point will include measurement errors and hence could have underestimated any true association between these factors and hip fracture risk. Selection bias was also present in this study, in that participants were healthier than nonparticipants.(2)

In summary, this study suggests that postural instability, quadriceps weakness, and a history of fall and prior fracture are predictions independent from BMD of hip fracture in both men and women. These risk factors could be incorporated in an assessment model for prediction of hip fracture in the elderly population. The prevention of hip fracture should move beyond the “paradigm” of reducing bone loss and include some aspects on practical issues of fall prevention and cognitive factors.

Acknowledgements

The authors thank the assistance of Janet Watters and Donna Reeves for the interview, data collection, and measurement of BMD. We also appreciate the invaluable help of the staff of Dubbo Base Hospital. The authors thank Natasa Ivankovic and Jim McBride for the management of the database. We acknowledge Khon Kaen University (Thailand) for providing financial support to CP to undertake work at the Garvan Institute of Medical Research. This work has been supported by the National Health and Medical Research Council of Australia and untied educational grants from GE-Lunar, Merck Australia, Eli Lilly International, and Aventis Australia. NDN is a recipient of an untied grant from Merck Australia and a recipient of the postgraduate Research Award from the National Health and Medical Research Council of Australia.

Ancillary