Height and Lower Extremity Length as Predictors of Hip Fracture: Results of the NHANES I Epidemiologic Follow-up Study

Authors

  • Alexander R Opotowsky,

    1. Department of Medicine, College of Physicians and Surgeons, Columbia University, New York, New York, USA
    Search for more papers by this author
  • Brian W Su,

    1. New York Orthopedic Hospital, Trauma Training Center, College of Physicians and Surgeons, Columbia University, New York, New York, USA
    Search for more papers by this author
  • John P Bilezikian

    Corresponding author
    1. Department of Medicine, College of Physicians and Surgeons, Columbia University, New York, New York, USA
    2. Department of Pharmacology, College of Physicians and Surgeons, Columbia University, New York, New York, USA
    • Address reprint requests to: John P Bilezikian, MD Department of Medicine Division of Endocrinology College of Physicians and Surgeons 630 W. 168th Street New York, NY 10032, USA
    Search for more papers by this author

  • The authors have no conflict of interest

Abstract

The relationship between height, lower extremity length (LEL), and hip fracture was examined in 4264 women followed for 22 years. Greater height predicted fracture among younger women, whereas LEL predicted risk for all ages. LEL is a better predictor of hip fracture risk than height among older women.

Introduction: The significance of height as a risk factor for hip fracture is controversial. Age-related height loss, caused by nonpathological processes and vertebral deformities, may confound the relationship between standing height and hip fracture. In contrast, LEL might be less subject to age-related changes and thus be a better index of fracture risk for older individuals.

Methods: We evaluated data available from 4264 white women who participated in the NHANES I Epidemiologic Follow-up Survey. There were 203 incident hip fractures over the 22-year follow-up period. The effect of height and LEL on hip fracture risk was determined for three age groups: 40–59, 60–69, and 70–74 year olds.

Results: After controlling for potential confounders, the hazard ratio (HR) for hip fracture for each SD greater standing height was 1.81 for the youngest age group (p = 0.002), 1.29 for 60–69 year olds (p = 0.02), and 1.10 for 70–74 year olds (p = 0.49). In contrast, LEL predicted hip fracture risk for all age groups, with HRs of 1.43 (p = 0.02), 1.30 (p = 0.002), and 1.33 (p = 0.02), respectively. Similarly, higher tertile of standing height did not confer increased hip fracture risk among the oldest women, but higher tertile of LEL did. Among women that were 70–74 years old, those in the highest tertile of LEL had a 1.83-fold (p < 0.05) greater risk of hip fracture than those in the lowest tertile.

Conclusions: The results support the hypothesis that LEL is a better predictor of hip fracture than standing height among older women, the group at highest risk for these fractures.

INTRODUCTION

Anumber of risk factors for hip fracture have been identified consistently, most importantly bone mineral density (BMD), previous fractures, weight, body mass index (BMI), age, and bone turnover.(1–6) The relationship between standing height and hip fracture risk, however, is controversial. Some studies have reported significantly increased risk among taller women, while others have failed to confirm this finding.(3,7–14) Aging could account for these conflicting results, because it affects height by several mechanisms and is also independently associated with increased hip fracture incidence. The age-related increase in vertebral fracture incidence is particularly noteworthy. Because vertebral fractures are common in the elderly and are also a significant risk factor for hip fracture, they may confound the relationship between standing height and hip fracture risk.

Most data used to investigate risk factors for hip fracture come from studies of older women, a rational choice because older postmenopausal women are at the greatest risk for osteoporotic fractures. Studies focusing on older postmenopausal women have generally found no relationship between standing height and hip fracture, while those including younger women have suggested that height is a significant risk factor for hip fracture.(3,7–14) In the Study of Osteoporotic Fractures (SOF), which included 9516 white women at least 65 years old, recalled height at the age of 25 was shown to be a risk factor for hip fracture.(3) It was suggested that women who were tall in their youth may be at greater risk because they fall from a greater vertical distance, assuming that height is an invariant parameter.(3,15,16) However, the SOF study was not able to verify an association between height at exam and hip fracture within 5 years. Other studies also found no difference in height between subjects with fractures and controls without fractures.(7,8) Several studies suggest that height may be a risk factor for cervical, but not trochanteric, hip fractures.(17) Conversely, several studies, which included middle-aged women and men, have found that height predicts hip fracture overall.(10–14) In a prospective study of over 25,000 middle-aged women (35–49 years old), Meyer et al.(9) found that age-adjusted relative risk for fracture in women increases with height. Women taller than 1.7 m were 3.6 times more likely to sustain a hip fracture as those less than 1.55 m tall.

Two processes associated with aging may account for the discrepancy between studies of older and younger women. First, stature decreases with age through several nonpathological mechanisms such as changes in vertebral body shape and height, loss of intervertebral disc height, and postural changes.(18,19) The magnitude of this height loss is variable and unpredictable. This obscures the relationship between measured height in elderly individuals and peak height achieved in young adulthood. Second, the incidence and prevalence of vertebral fractures increases with age.(20–22) Unlike nonpathological height loss, vertebral fractures are associated with increased hip fracture risk.(1,4,5,), (23,24) The shorter stature of those with prevalent vertebral fractures and their concomitant higher risk of hip fracture are likely to obscure any increase in hip fracture risk because of taller peak standing height.(9)

To date, all studies investigating the relationship between height and hip fracture have focused on standing height. This study addresses the hypothesis that standing height predicts hip fracture risk among younger women, but not among older women. In addition, the study investigates the relationship between lower extremity length (LEL) and hip fracture. Because LEL is associated with peak standing height but is likely to be less affected by age, it may better predict hip fracture risk for older women than standing height.

MATERIALS AND METHODS

Data source

The First National Health and Nutrition Examination Survey (NHANES I), conducted by the National Center for Health Statistics between 1971 and 1975, is an extensive cross-sectional survey of over 25,000 individuals 1–74 years old. The 14,407 participants, ages 25–74, who had a medical examination, served as the baseline cohort for the NHANES I Epidemiologic Follow-up Study (NHEFS). These participants were followed intermittently until 1992. There were four follow-up periods: 1982–1984, 1986 (only those aged 55 or older at exam), 1987, and 1992. During each follow-up period, each participant was traced, vital status assessed, death certificates sought if the participant had died, and an interview was conducted with the participant or with a proxy for deceased or incapacitated participants. Details of the plan and operation of NHANES I and NHEFS have been published.(25–29)

NHEFS data have been used previously to investigate hip fracture risk factors, including various anthropometric risk factors. Langlois et al.(30) found that weight loss and maximum body weight are associated with an increased risk for hip fracture. Farmer used NHEFS data through 1984 to analyze the effect of BMI, triceps skinfold, and arm muscle area on hip fracture risk.(2)

This study includes 4264 white women aged 40–74 at baseline, with complete anthropometric data. Women who reported a previous hip fracture at baseline (n = 65) or with incomplete height, weight, or sitting height data (n = 5) were excluded from the analysis. These women were older, shorter, and weighed less (all p < 0.0001) than those included in the analysis, but there was no difference in risk of subsequent hip fracture.

Anthropometric measures and covariates

Height and sitting height were measured at baseline. For sitting height, the subject sat erect on an examining table, with eyes looking straight ahead, and with the infraorbital meatal line parallel to the tabletop. The participant sat back as far as possible, with feet on the appropriate step of a stool to ensure that her thighs were horizontal with the popliteal fossa at the table edge. A caliper arm was brought down firmly against the midline of the subject's head.(31) We subtracted sitting height from standing height and defined this measure as the LEL. Covariates were measured at baseline, unless otherwise noted, including (1) age at exam, (2) weight, (3) age at menopause, (4) history of chronic diseases based on self-reported doctor's diagnosis of emphysema, chronic bronchitis, thyroid disease, diabetes, heart failure, hypertension, arthritis, or malignancy, (5) alcohol use (drinking at least 2–3 drinks per week on average during the previous year), (6) nonrecreational physical activity (three levels) was reported by participants in response to the question, “In your usual day, aside from recreation, are you physically very active, moderately active, or quite inactive?,” (7) recreational physical activity reflected the participant's response to an equivalent question, (8) hormone replacement use, considered positive if the participant or their proxy reported use of hormone pills in any of the follow-up interviews, and (9) history of fracture, considered positive if a participant reported that a doctor had previously diagnosed a spine, wrist, or other fracture. Smoking history was available only for a subset of the participants and therefore was not included as a covariate in the model. Height and sitting height were neither directly measured nor queried about during any of the follow-up interviews. Thus, longitudinal data on height loss for each individual over the course of the study were not available.

Ascertainment of hip fracture

At each follow-up interview, participants reported all hospital stays since the previous interview. Hospitals listed by the participant were contacted, and discharge summaries were requested for all hospital stays occurring during that period, irrespective of whether the participant had reported that specific stay. Each discharge summary could list up to 10 diagnoses, and a hip fracture case was defined as an ICD-9-CM code of 820 for any of the diagnoses listed. The date of the admission for hip fracture was considered the date of the fracture and for women who had more than one record with an ICD-9-CM code 820 listed as a diagnosis only the earliest record of hip fracture was used. Two hundred and three first incident hip fractures were identified.

Statistical analysis

Participants were divided into three age groups: 40–59, 60–69, and 70–74 years old. Unequal age groups were used to ensure a more evenly distributed number of incident hip fractures among the groups. More women were included in the younger age groups because hip fracture incidence increases sharply with age. The participants were then, by age group, categorized into height and LEL tertiles. For each age group descriptive statistics for participants in each tertile of height and LEL were calculated using ANOVA.

Mean standing height and LEL were calculated for each age group to evaluate the hypothesis that age-related changes in standing height mainly reflect changes in sitting height in the setting of relatively stable LEL. Pearson correlation coefficients were calculated between standing height, sitting height, and LEL for each age group. ANOVA was performed to evaluate differences in the ratio of sitting height to LEL between age groups.

Hazard ratios (HRs) for hip fracture were calculated using Cox proportional hazard regression models with the PHREG procedure in SAS Release 8.02 (SAS Institute, Cary, NC, USA). For participants who sustained a hip fracture, follow-up time was defined as the time from the baseline medical examination to the date of admission for hip fracture. Follow-up time for participants who did not sustain a hip fracture was the time from the date of the baseline medical exam to either the date of death, if the participant died during the study, or the date of the last follow-up interview.

HRs were calculated for a model including all the women and using separate models for each of the three age groups. All models controlled for age at exam, weight, age at menopause, history of hormone replacement (ever), history of chronic disease, alcohol use, recreational and nonrecreational physical activity, and history of fracture. Cox regression was performed using height and LEL as tertiles and as continuous variables. Results for the analysis of height and LEL as continuous variables are presented per SD increase (SDs of 6.34 and 4.16 cm, respectively).

The statistical sample weights provided by the National Center for Health Statistics are highly variable and skewed and were not used because their inclusion substantially inflates variance estimates.(32) The results presented reflect relationships seen in a large national sample but are not necessarily reflective of the entire U.S. population.

RESULTS

Descriptive statistics, by height and LEL tertile, are provided in Table 1. For all age groups, height tertile was positively associated with LEL and with weight. Similarly, higher LEL tertile was associated with taller height and greater weight. In the youngest age group, but not in the older groups, height and LEL were negatively associated with age at exam. In the oldest age group, LEL tertile, but not height tertile, was positively associated with level of nonrecreational physical activity. There were no significant differences in age at menopause, prevalence of previously diagnosed fracture, or hormone replacement use between tertiles of height or LEL for any age group. Table 2 shows hip fracture incidence rates by age group. As expected, the hip fracture incidence in this cohort increased sharply with age.

Table Table 1. Characteristics by Baseline Standing Height and Lower Extremity Length Tertiles
original image
Table Table 2. Hip Fracture Incidence by Age Group
original image

Analysis of mean standing height and LEL by age group supported the hypothesis that age-related height loss mainly affects the upper body. Mean standing height for the three age groups, respectively, was 161.7, 158.8, and 157.5 cm. The 70–74 year olds were shorter than the 40–59 year olds by an average of 4.2 cm, an absolute difference of 2.6%. LEL averaged 75.7, 75.2, and 75.2 cm for each age group, respectively, differing 0.5 cm between the youngest and oldest age, an absolute difference of only 0.7%. Thus, changes in sitting height account for virtually all (88%) of the overall height difference between the youngest and oldest age groups. Although such differences may be due either to secular (generational) differences or to the effects of aging, the discrepancy between the age-related differences in standing height and LEL show the relative constancy of the latter measure. Furthermore, while height and LEL were highly correlated for all age groups, that correlation decreased somewhat with age (r = 0.87, r = 0.84, r = 0.82, all p < 0.0001 versus no correlation). Likewise, the correlation between LEL and sitting height was greatest for the youngest age group and declined by age group (r = 0.35, r = 0.31, r = 0.22, all p < 0.0001 versus no correlation). The ratio of sitting height to LEL decreased with age: 1.14 in the youngest group, 1.11 in the middle group, and 1.09 in the oldest group (p < 0.0001 for a difference between the three). Taken in concert, these findings suggest strongly that age-related height loss predominantly affects the upper body and spares the lower extremities.

Table 3 shows the relationship between height and LEL tertiles with hip fracture risk by age group. Among the youngest age group, those in the highest height tertile had 3.91 (95% CI, 1.37, 11.13) times higher hip fracture risk than those in the lowest tertile. In the 60–69 year old group, the risk of hip fracture was 2.01 (95% CI, 1.18, 3.44) times higher in the highest tertile of height compared with the lowest tertile. The oldest age group was the exception, with no change in risk associated with height tertile. In all three age groups, however, those with the longest lower extremities had significantly higher risk of hip fracture than individuals in the lowest tertile of LEL. For all three groups, those in the lowest tertile of LEL had the lowest risk of hip fracture. At first glance it seems that LEL also confers greater hip fracture risk among the youngest age group. However, the small sample size in the youngest age group precludes any conclusion to be drawn, as demonstrated by the wide and overlapping confidence intervals. As described below, analysis of LEL as a continuous variable also failed to support the hypothesis that LEL confers a higher hip fracture risk among the youngest age group. The relationship of height and LEL tertile with hip fracture incidence, by age group, is shown in Fig. 1. The figure shows that age is a more powerful predictor of hip fracture risk than height or LEL. Still, the results suggest that LEL predicts hip fracture risk independently of age group.

Table Table 3. Univariate and Multivariate Hazard Ratios for Hip Fracture for Each Tertile of Height and Lower Extremity Length by Age Group
original image
Figure FIG. 1..

Hip fracture incidence rates per 10,000 person-years, by age group and (A) height tertile and (B) LEL tertile. Hip fracture incidence increased with increasing age, as expected. (A) Hip fracture risk increased with height tertile for women in the youngest age group, but there was no clear pattern in the relationship between hip fracture incidence and height tertile for 60–69 year olds and 70–74 year olds. (B) Within each age group, subjects in the lowest tertile of LEL had the lowest incidence of hip fracture. The importance of LEL as a predictor of hip fracture is made apparent when considering that 60–69 year olds in the highest tertile of LEL had a greater hip fracture risk than 70–74 year olds in the lowest tertile. Rates depicted are unadjusted. For adjusted HRs, see Table 3.

HRs, per SD increase in height and LEL, are shown in Table 4 and Fig. 2. Increased standing height was associated with an increased risk for hip fracture only for women under 70 years old. Each SD increase in height increased the risk of hip fracture by 81% among 40–59 year olds (p = 0.002), but by only 29% for 60–69 year olds (p = 0.02). Standing height did not predict an increased hip fracture risk among the oldest women (p = 0.49). In marked contrast, LEL was positively related to hip fracture risk for all age groups, without any difference in its relationship to fracture with age. For all age groups, each SD increase in LEL conferred a 30–43% increased risk of hip fracture (p = 0.02–0.002).

Table Table 4. Hip Fracture HRs Per SD Increase in Height and Lower Extremity Length by Age Group
original image
Figure FIG. 2..

HR (±SEM) for hip fracture for each SD increase in (A) height and (B) LEL, by age group controlling for age, weight, age at menopause, hormone use, recreational activity, nonrecreational activity, alcohol use, history of fracture, and history of chronic diseases. LEL consistently predicted hip fracture risk across all age groups, while height only predicted hip fracture risk among women in the two lower age groups; *p < 0.05.

DISCUSSION

This analysis of the NHEFS data show that the relationship between height and hip fracture risk is limited to middle-aged women. On the other hand, LEL predicts higher hip fracture risk among all age groups. The explanation likely lies in the increasing prevalence of both nonpathological height loss and vertebral fractures with age. Any relationship between peak standing height and hip fracture may be obscured by random, nondifferential height loss, and confounded by the increased risk of hip fracture among women who sustain vertebral fractures and thereby lose height.

Height loss is a nearly ubiquitous sequel of aging and occurs predominantly caused by changes in trunk length, while LEL remains relatively constant.(33) There is no evidence that this natural aging process is associated with hip fracture risk. On the other hand, vertebral fractures are not only associated with height loss, but are also a major risk factor for hip fracture. Furthermore, vertebral fractures are associated with a variety of other adverse health outcomes including back pain and disability.(34–36) There is a strong relationship between the number of vertebral deformities and height loss, even after adjusting for age. Women with two or more vertebral fractures lose 8 cm of height on average, while those with one fracture lose 4 cm and those without fractures lose an average of only 3 cm.(23) Thus, women with vertebral fractures have greater height loss and higher risk of hip fracture than women with no evidence of vertebral fractures.

Studies have consistently found that individuals with vertebral fractures are at a higher risk of hip fracture than those without. Melton et al.(4) found that clinically recognized prevalent vertebral fractures increased the risk of subsequent hip fracture 2.3-fold. Black et al.(1) confirmed this finding, noting that subjects with radiographic evidence of vertebral fracture had a 3-fold risk of subsequent hip fracture. Even after controlling for age and BMD, vertebral fractures independently predicted a 1.86-fold higher risk of hip fracture.

While often overlooked, vertebral fractures are the most common osteoporotic fracture. They usually occur earlier in the natural history of osteoporosis than hip fractures, and their prevalence increases with age.(21,22) Melton et al.(21) found that 7.6% of 50- to 54-year-old women have radiographic evidence of vertebral fracture, while 64.3% of women aged 90 years and older have such evidence. Cooper et al.(22) found that 25.3% of women over 50 years old have radiographic evidence of vertebral fractures, with an estimated incidence of 17.8 per 1000 person-years. However, only 30% of radiographically defined vertebral fractures are clinically detected.

There are two mechanisms by which height and LEL might affect hip fracture risk that are consistent with the findings presented. First, the aphorism, “The bigger they are the harder they fall,” may have biomechanical significance. Taller people have further to fall and may hit the ground with more force. Falls from greater height may also be associated with greater torque across the cervical or trochanteric regions of the femur. Nevitt and Cummings(15) found that among women who fall, taller women have a greater risk of hip fracture than shorter women. Our results do not support this observation across all age groups, because standing height was not a hip fracture risk factor for women 70 or older. It is more likely that the distance of the proximal femur from the ground is a better predictor of torque than standing height. This is consistent with our finding that increased LEL bestows higher hip fracture risk among all age groups.

A more appealing explanation lies in the relationship between height and femoral geometry. Various measures of femoral geometry have been found to be predictive of hip fracture.(37–41) In particular, hip axis length (HAL), the distance from the greater trochanter to the inner pelvic brim, is an independent risk factor for hip, even after controlling for height, weight, and BMD.(37,38) Several studies have reported correlations between height and measures of femoral geometry. Height and HAL are correlated highly, with estimates of the correlation coefficient ranging from 0.42 to 0.57.(40,42) These studies included older women; the relationship between height and HAL may be even stronger before the age of related height loss and the incidence of vertebral fractures. LEL is also likely to be strongly related to femoral geometry, and the increased risk of hip fracture seen among women with longer LEL may reflect this relationship.

These results raise the question of whether height and LEL are simply predictors of bone density or are independent risk factors for hip fracture. The value of BMD for ascertaining fracture risk derives in large part from its ability to integrate the many disparate risk factors for fracture such as estrogen exposure, physical activity, genetics, and weight. Height seems to predict hip fracture independently from BMD. In the SOF study, controlling for bone density enhanced the relationship between recalled height at the age of 25 and hip fracture, suggesting that height and BMD affect hip fracture risks by different mechanisms.(3) In addition, previous studies suggest height is positively correlated with BMD.(43,44) These results support the conclusion that height and LEL are likely to be independent measures of hip fracture risk.

This is the first prospective study to investigate how age affects the relationship between standing height and hip fracture and to evaluate the relationship between LEL and hip fracture risk. Our results help explain the puzzling discrepancies between studies measuring height earlier and later in life. The fracture risk bestowed by peak standing height is confounded by age, but LEL retains its predictive value. While further studies are needed to confirm the independence of LEL and BMD as hip fracture risk factors, evidence suggests that they independently predict hip fracture. LEL is a novel, noninvasive, inexpensive measure that shows promise as a clinically useful adjunct to BMD for evaluating hip fracture risk.

Ancillary