Vol. 332 No.

12 RISK FACTORS FOR HIP FRACTURE IN WHITE WOMEN 767

RISK FACTORS FOR HIP FRACTURE IN WHITE WOMEN

STEVEN R. CUMMINGS, M.D., MICHAEL C. NEVITT, PH.D., WARREN S. BROWNER, M.D., M.P.H.,
KATIE STONE, M.A., KATHLEEN M. FOX, PH.D., KRISTINE E. ENSRUD, M.D., M.P.H., JANE CAULEY, DR.P.H.,
DENNIS BLACK, PH.D., AND THOMAS M. VOGT, M.D., M.P.H., FOR THE STUDY OF OSTEOPOROTIC FRACTURES
RESEARCH GROUP*
Abstract Background. Many risk factors for hip frac- treated with long-acting benzodiazepines or anticonvul-
tures have been suggested but have not been evaluated sant drugs, ingested greater amounts of caffeine, or spent
in a comprehensive prospective study. four hours a day or less on their feet. Examination find-
Methods. We assessed potential risk factors, includ- ings associated with an increased risk included the inabil-
ing bone mass, in 9516 white women 65 years of age or ity to rise from a chair without using one’s arms, poor
older who had had no previous hip fracture. We then fol- depth perception, poor contrast sensitivity, and tachycar-
lowed these women at 4-month intervals for an average dia at rest. Low calcaneal bone density was also an inde-
of 4.1 years to determine the frequency of hip fracture. All pendent risk factor. The incidence of hip fracture ranged
reports of hip fractures were validated by review of x-ray from 1.1 (95 percent confidence interval, 0.5 to 1.6) per
films. 1000 woman-years among women with no more than two
Results. During the follow-up period, 192 women had risk factors and normal calcaneal bone density for their
first hip fractures not due to motor vehicle accidents. In age to 27 (95 percent confidence interval, 20 to 34) per
multivariable age-adjusted analyses, a maternal history of 1000 woman-years among those with five or more risk
hip fracture doubled the risk of hip fracture (relative risk, factors and bone density in the lowest third for their age.
2.0; 95 percent confidence interval, 1.4 to 2.9), and the in- Conclusions. Women with multiple risk factors and
crease in risk remained significant after adjustment for low bone density have an especially high risk of hip frac-
bone density. Women who had gained weight since the ture. Maintaining body weight, walking for exercise, avoid-
age of 25 had a lower risk. The risk was higher among ing long-acting benzodiazepines, minimizing caffeine in-
women who had previous fractures of any type after the take, and treating impaired visual function are among the
age of 50, were tall at the age of 25, rated their own health steps that may decrease the risk. (N Engl J Med 1995;
as fair or poor, had previous hyperthyroidism, had been 332:767-73.)

O NE of every six white women will have a hip

fracture during her lifetime.1 Many potential risk
factors for hip fracture, such as lower body weight,
cause of their low incidence of hip fractures) and white women who
had undergone bilateral hip replacement or had an earlier hip frac-
ture were excluded. The study was approved by the appropriate com-
mittees on human research, and all the women provided written in-
cigarette smoking, caffeine intake, use of long-acting formed consent.
sedatives, and inactivity, have been identified in case–
control2-4 and prospective5-8 studies. However, the find- Assessment of Risk Factors
ings of case–control studies might be affected by selec- Questionnaire and Interview
tion and recall biases, and most previous prospective The women were questioned and examined in outpatient clinics.
studies have examined only a small number of the We ascertained their numbers of years of education, natural hair col-
many risk factors for hip fracture. To identify impor- or as young adults, height and nonpregnant weight at the age of 25,
ethnic origin, numbers of pregnancies and of children who were
tant risk factors, we assembled a cohort of older wom- breast-fed, ages at the last menstrual period, parental history of frac-
en, identified many potential risk factors, measured tures, falls during the previous year, smoking habits and alcohol
bone mass, and followed the women for hip fractures. intake, and self-rated health. We asked about physician-diagnosed
fractures since the age of 50, osteoporosis, spine fracture, hyperthy-
METHODS roidism, osteoarthritis, gastric surgery, hysterectomy, oophorectomy,
cataracts, and stroke. The women were also asked about current
Subjects therapy and therapy during the previous year with estrogen, diuret-
From 1986 to 1988, we recruited women who were able to walk ics, corticosteroids, thyroid hormones, anticonvulsant agents, antac-
and who were at least 65 years of age in Portland, Oregon; Min- ids, sleeping aids, and anxiolytic drugs. Long-acting benzodiazepines
neapolis; Baltimore; and the Monongahela Valley, Pennsylvania, were defined as those with half-lives of at least 24 hours.10
through mailings to women on lists such as voter-registration lists.9 The amount of dietary calcium was assessed by a food-frequency
The study group consisted of 9516 white women; black women (be- questionnaire,11 and caffeine intake was estimated.12 We also asked
about walking; exercise13; the number of hours spent sitting and lying
down per day; and the amount of difficulty experienced in walking,
From the Division of General Internal Medicine (S.R.C., W.S.B.) and the De- climbing and descending stairs, preparing meals, shopping, and do-
partment of Epidemiology and Biostatistics (S.R.C., M.C.N., W.S.B., K.S., D.B.), ing housework.14
University of California, San Francisco; the Department of Epidemiology and
Preventive Medicine, University of Maryland, Baltimore (K.M.F.); the Division Examinations
of Epidemiology, School of Public Health, University of Minnesota, Minneapolis
(K.E.E.); the Department of Epidemiology, University of Pittsburgh, Pittsburgh
We measured weight, height (by stadiometer), waist and hip cir-
(J.C.); and the Kaiser Permanente Center for Health Research, Portland, Oreg. cumferences, and knee height.15 Body-mass index (the weight in kil-
(T.M.V.). Address reprint requests to Dr. Cummings at the Prevention Sciences ograms divided by the square of the height in meters) was calculated
Group, 74 New Montgomery St., Suite 600, San Francisco, CA 94105. with knee height substituted for total height. Tests of neuromuscular
Supported by grants (1-RO1-AG05407, 1-RO1-AR35582, 5-RO1-AG05394, function included whether the subject could rise up from a chair
1-RO1-AM35584, and 1-RO1-AR35583) from the Public Health Service. (without using her arms) five times; the number of step-ups she com-
*The members of the Study of Osteoporotic Fractures Research Group are list- pleted in 10 seconds; the strength of her grip, triceps, knee exten-
ed in the Appendix. sions, and hip abduction (measured with a hand-held isometric dyna-

The New England Journal of Medicine

Downloaded from nejm.org on March 23, 2024. For personal use only. No other uses without permission.
Copyright © 1995 Massachusetts Medical Society. All rights reserved.
768 THE NEW ENGLAND JOURNAL OF MEDICINE March 23, 1995

mometer)16; her walking speed (over a 6-m course); and her ability Table 1. Characteristics of the 9516 Women Stud-
to walk and stand in a tandem position with her eyes open and ied to Determine the Risk of Hip Fractures.
closed. We assessed cognitive function with a modified version of the
Mini–Mental State Examination.17,18 We measured corrected visual CHARACTERISTIC VALUE*
acuity19 and contrast sensitivity 20 (separately averaging the scores for
Age (yr) 725
low and high spatial frequencies). We assessed depth perception us-
ing the Howard–Dohlman device21 and scored it as the standard de- White race (%) 99.7
viation of four trials. Blood pressure and pulse were measured with High-school graduate (%) 76.9
the subject supine and after she had been standing for one minute22; Mean weight (kg) 67.212.7
orthostatic hypotension was defined as a drop in systolic blood pres- Height (cm) 159.06.0
sure of 20 mm Hg or more on standing. History of any fracture since age of 50 (%) 35.9
We measured calcaneal bone mineral density, a measurement History of maternal hip fracture (%) 10.0
mainly of weight-bearing trabecular bone and a strong predictor of Previous hyperthyroidism (%) 9.2
hip fractures,23 using single-photon absorptiometry (OsteoAnalyzer,
Currently taking estrogen (%) 13.9
Siemens–Osteon, Wahiawa, Hawaii). During a second examination,
Currently taking anticonvulsant drugs (%) 1.1
conducted between 1988 and 1990, we measured the bone density of
the proximal femur, using dual x-ray absorptiometry (QDR 1000, Ho- Currently taking long-acting benzodiazepines (%) 9.2
logic, Waltham, Mass.) in 7786 women (82 percent of the survivors Currently taking thyroid hormone (%) 12.3
at that time).24 To determine the mean coefficient of variation for Self-reported health status (%)
these measurements, a few staff members were measured at the four Excellent or good 83.3
clinical centers, and the mean coefficient of variation was found to Fair 15.2
Poor or very poor 1.5
be 1.2 percent for both the calcaneus and femoral neck.24
Mean daily dietary calcium intake (mg) 713425
Ascertainment of Hip Fractures Mean daily caffeine intake (mg)† 172138
Walking for exercise (%) 50.5
We contacted the women about fractures by postcard or telephone
On feet 4 hr/day (%) 9.7
every four months and were able to complete 99 percent of these con-
tacts. We confirmed all hip fractures by reviewing radiographs.25 Currently a smoker (%) 9.9
Average weight gain since age of 25 (%) 2020
Statistical Analysis Average height at age of 25 (cm) 162.66.0
We used proportional-hazards analysis to identify potential predic- Resting pulse 80 beats/min (%) 12
tors of hip fracture and best subset analysis26 and backward stepwise Calcaneal bone mineral density (g/cm2) 0.410.10
analysis to identify independent risk factors. We found no interac- Femoral-neck bone mineral density (g/cm2)‡ 0.650.11
tions between risk factors and clinical centers. Unless specified, cor-
relations between variables in multivariable models were less than *Plus–minus values are means SD.
0.5. We report age-adjusted relative hazards as relative risks with 95 †Caffeine intake is reported for those drinking caffeinated beverages.
percent confidence intervals. Analyses were performed with the use ‡This characteristic was measured in 7834 women whose hip bone den-
sity was determined at the second visit.
of Statistical Analysis Software (SAS, Cary, N.C.).
We also explored whether the effects of certain risk factors could
be explained by their effect on bone density at the hip, limiting the
analyses to those variables — maternal history of hip fracture, height
women taking anticonvulsant drugs resulted from sei-
at age 25, and history of hyperthyroidism — that were unlikely to zures or loss of consciousness. As caffeine intake in-
change from base line to the measurement of hip bone density. creased, so did the risk of hip fracture. Women who
spent four hours per day or less on their feet had twice
RESULTS the risk of women who spent more than four hours per
Characteristics of the 9516 subjects are shown in Ta- day on their feet, whereas women who regularly walked
ble 1. During an average 4.1 years of follow-up, 192 for exercise had a 30 percent lower risk of hip fracture
women had first hip fractures not due to motor vehicle than women who did not walk regularly. Risk tended to
accidents, 565 died, and 92 were lost to follow-up. In decrease as the distance walked per day increased (rel-
multivariable analyses, we identified 16 independent ative risk, 0.9 per five blocks walked per day; 95 per-
risk factors for hip fracture (Table 2) besides bone den- cent confidence interval, 0.8 to 1.0).
sity. A woman whose mother had had a hip fracture Four characteristics observed in the physical exami-
had twice the risk of hip fracture of women without nation indicated an increased risk of hip fracture: the
such a maternal history, especially if her mother frac- inability to rise from a chair without using one’s arms,
tured her hip before the age of 80 (relative risk, 2.7; 95 a faster resting pulse rate, poorer depth perception,
percent confidence interval, 1.7 to 4.4) rather than at and poorer low-frequency contrast sensitivity.
80 or later (relative risk, 1.6; 95 percent confidence in-
terval, 1.0 to 2.7). Bone Density and History of Fracture
The more weight a woman had gained since the age Lower calcaneal bone density and a history of any
of 25, the lower her risk of hip fracture (Fig. 1). The type of fracture since the age of 50 independently in-
women who weighed less than they had at 25 had a creased the risk of hip fracture (Table 2). Wrist frac-
doubled risk of hip fracture (relative risk, 2.2; 95 per- tures (relative risk, 1.9; 95 percent confidence interval,
cent confidence interval, 1.6 to 3.0). Women who were 1.4 to 2.7), self-reported spine fractures (relative risk,
tall at the age of 25 also had a greater risk. 1.9; 95 percent confidence interval, 1.2 to 2.9), and all
Poorer self-rated health, a history of hyperthyroid- other types of fractures (relative risk, 1.5; 95 percent
ism, and therapy with long-acting benzodiazepines or confidence interval, 1.0 to 2.1) were all associated with
anticonvulsant drugs independently increased the risk an increased risk of hip fracture.
of hip fracture. None of the seven hip fractures among All risk factors remained significantly associated

The New England Journal of Medicine

Downloaded from nejm.org on March 23, 2024. For personal use only. No other uses without permission.
Copyright © 1995 Massachusetts Medical Society. All rights reserved.
Vol. 332 No. 12 RISK FACTORS FOR HIP FRACTURE IN WHITE WOMEN 769

Table 2. Multivariable Models of Risk Factors for Hip Fracture fewer risk factors had an incidence of only 1.1 (95 per-
with and without Adjustment for Fractures and Calcaneal Bone cent confidence interval, 0.5 to 1.6) per 1000 woman-
Density among 9516 White Women.
years. The 6 percent of women who had five or more
RELATIVE RISK risk factors and calcaneal bone density in the lowest
MEASUREMENT (COMPARISON OR UNIT)* (95% CONFIDENCE INTERVAL)
third for their age had an incidence of hip fracture of
ADD FRACTURES AND
BASE MODEL† BONE DENSITY 27 (95 percent confidence interval, 20 to 34) per 1000
Age (per 5 yr) 1.5 (1.3–1.7) 1.4 (1.2–1.6)
woman-years and had 62 (32 percent) of the hip frac-
History of maternal hip fracture (vs. none) 2.0 (1.4–2.9) 1.8 (1.2–2.7) tures reported.
Increase in weight since age 25 (per 20%) 0.6 (0.5–0.7) 0.8 (0.6–0.9)
Factors Not Significant after Multivariate Adjustment
Height at age 25 (per 6 cm) 1.2 (1.1–1.4) 1.3 (1.1–1.5)
Self-rated health (per 1-point decrease)‡ 1.7 (1.3–2.2) 1.6 (1.2–2.1) For a number of factors that were initially associated
Previous hyperthyroidism (vs. none) 1.8 (1.2–2.6) 1.7 (1.2–2.5) with the risk of hip fracture in age-adjusted models,
Current use of long-acting benzodiazepines 1.6 (1.1–2.4) 1.6 (1.1–2.4) these associations were diminished and no longer sta-
(vs. no current use)
tistically significant after adjustment for other vari-
Current use of anticonvulsant drugs 2.8 (1.2–6.3) 2.0 (0.8–4.9)
(vs. no current use) ables (Table 3). For example, current smokers had
Current caffeine intake (per 190 mg/day) 1.3 (1.0–1.5) 1.2 (1.0–1.5) about twice as high a risk of hip fracture as nonsmok-
Walking for exercise (vs. not walking for 0.7 (0.5–0.9) 0.7 (0.5–1.0) ers or former smokers. The smokers had gained less (or
exercise) lost more) weight, had poorer health, had more diffi-
On feet 4 hr/day (vs. 4 hr/day) 1.7 (1.2–2.4) 1.7 (1.2–2.4)
culty rising from a chair, spent fewer hours on their
Inability to rise from chair (vs. no inability) 2.1 (1.3–3.2) 1.7 (1.1–2.7)
feet, were less likely to walk for exercise, and had faster
Lowest quartile for distant depth perception 1.5 (1.1–2.0) 1.4 (1.0–1.9)
(vs. other three) heart rates. Adjusting for these effects explained most
Low-frequency contrast sensitivity (per 1 SD 1.2 (1.0–1.5) 1.2 (1.0–1.5) of the association between smoking and hip fracture.
decrease) Alcohol ingestion was associated with a lower risk of
Resting pulse rate 80 beats/min (vs. 80 1.8 (1.3–2.5) 1.7 (1.2–2.4) hip fracture (Table 3). The risk was somewhat lower
beats/min)
Any fracture since age of 50 (vs. none) — 1.5 (1.1–2.0)
among those who drank seven or fewer drinks per week
Calcaneal bone density (per 1 SD decrease) — 1.6 (1.3–1.9) (relative risk as compared with nondrinkers, 0.7; 95
percent confidence interval, 0.5 to 0.9) than among
*For continuous variables, the relative risks are expressed as a change in risk for each spec-
ified change in the risk factor. those who drank more (relative risk, 0.9; 95 percent
†Base-model values are based on proportional-hazards analysis with backward stepwise confidence interval, 0.5 to 1.4). However, alcohol in-
elimination. Best subsets models yielded similar sets of risk factors, including the number of
steps in a 360-degree turn and the functional-status score; some did not include low-frequency
take was no longer significantly associated with a lower
contrast sensitivity, long-acting benzodiazepine therapy, or walking for exercise. risk of hip fracture after adjustment for the better self-
‡Health was rated as poor (1 point), fair (2 points), or good to excellent (3 points). reported health and ability to stand up from a chair
among those who drank alcohol.27
with the risk of hip fracture after bone density was add- Current thyroid-hormone therapy was no longer sig-
ed to the multivariable model. Therapy with anticon- nificantly associated with the risk of hip fracture after
vulsant drugs was no longer a significant factor after a adjustment for a history of hyperthyroidism, which
history of fractures was also added (Table 2). was reported by 36 percent of those taking thyroid
hormone.
Adjustment for Femoral-Neck Bone Density Greater weight, body-mass index, and modified body-
In analyses limited to the 7786 women in whom the
bone density of the hip had been measured, 83 of
whom later had hip fractures, a history of maternal hip 25

fracture was significantly associated with an increased

(per 1000 woman-years)

risk of hip fracture before (relative risk, 2.0; 95 per- 20

Rate of Hip Fracture

cent confidence interval, 1.2 to 3.6) and after (relative

risk, 1.9; 95 percent confidence interval, 1.1 to 3.2) ad- 15.0
15
justment for femoral-neck bone density. Similarly, ad-
justment for femoral-neck bone density did not affect
the significant associations between height at the age 10
of 25 or previous hyperthyroidism and the risk of hip 6.3
fracture.
5 4.0 3.4
Number of Risk Factors, Bone Density, and Prediction of Hip 1.1
Fracture 0
10 10 to 10 10 to 30 30 to 50 50
Fifteen percent of the women had five or more of the
risk factors listed in Table 2 (including older age and Change in Weight (%)
previous fracture, but not low bone density); their inci- Figure 1. Association between the Change in Body Weight after
dence of hip fracture was 19 (95 percent confidence in- the Age of 25 and the Risk of Hip Fracture.
terval, 15 to 22) per 1000 woman-years (Fig. 2). By The rate of hip fracture is adjusted for age. The T bars denote
comparison, the 47 percent of women who had two or the upper 95 percent confidence limits.

The New England Journal of Medicine

Downloaded from nejm.org on March 23, 2024. For personal use only. No other uses without permission.
Copyright © 1995 Massachusetts Medical Society. All rights reserved.
770 THE NEW ENGLAND JOURNAL OF MEDICINE March 23, 1995

mass index were all associated with Table 3. Predictors Significantly Associated with Hip Fractures in Age-Adjusted Mod-
a decreased risk of hip fracture, but els but Not in Multivariable Models.
not quite as strongly as was the per- RELATIVE RISK
centage of weight change since the PREDICTOR (95% CONFIDENCE INTERVAL) VARIABLES IN MULTIVARIABLE MODEL*
age of 25. Neither current weight AGE-ADJUSTED MULTIVARIABLE WEIGHT HEALTH
ABILITY TO
RISE
nor modified body-mass index re- MODEL MODEL CHANGE STATUS FROM CHAIR OTHER

mained significantly associated with Current smoking (vs. never 2.1 (1.4–3.3) 1.4 (0.9–2.3) X X X X†
the risk of hip fracture after ad- having smoked)
justment for weight gain. However, Any alcohol in the previous 0.7 (0.5–0.9) 0.8 (0.6–1.1) X X
year (vs. none)
the percentage of weight change
Current thyroid hormone 1.6 (1.1–2.3) 1.3 (0.9–2.0) X‡
was correlated with current weight therapy (vs. none)
(r  0.75) and body-mass index Functional-status score 1.4 (1.2–1.5) 1.1 (1.0–1.3) X X
(r  0.76). (per 3 points)
A history of falling indicated an Fall in the previous year (vs. 1.6 (1.2–2.1) 1.4 (1.0–2.1) X X X§
none)
increased risk of hip fracture (Table Gait speed (per 0.22 m/sec) 1.4 (1.2–1.6) 0.9 (0.8–1.0) X
3); the risk increased 30 percent Current weight (per 12.7 kg) 0.8 (0.7–0.9) 1.2 (0.9–1.4) X
with each additional fall (relative
risk, 1.30 per fall from 0 to  5; 95 *When these variables were included in a proportional-hazards model, the predictor variable was no longer significantly
associated with the risk of hip fracture.
percent confidence interval, 1.1 to †Other variables included are hours on one’s feet, walking for exercise, and pulse rate.
1.5). After adjustment for the inabil- ‡The other variable included is a history of hyperthyroidism.
ity to rise from a chair, spending §The other variable included is hours on one’s feet.
four hours or less on one’s feet, and
poor health, however, a history of falling and the num- cluded the inability to rise from a chair without using
ber of falls were no longer significantly associated with one’s arms, no other measurement of neuromuscular
hip fracture. function remained significantly associated with the risk
Poor performance on almost every test of neuromus- of hip fracture.
cular function, such as gait speed, was associated with The risk decreased with increasing numbers of births
an increased risk of hip fracture. When the models in- (relative risk, 0.9 per birth; 95 percent confidence inter-
val, 0.8 to 1.0). This association was not quite statis-
tically significant after adjustment for weight change
since the age of 25.
27.3 Factors Not Significantly Associated with the Risk of Hip
Fracture
(per 1000 woman-years)

30 Factors that were not significantly associated with

Rate of Hip Fracture

the risk of hip fracture included hair color, ethnic an-

20
14.7 cestry, whether there was a maternal history of frac-
tures other than hip fracture, the number of children
breast-fed, the timing of menopause, past smoking sta-
4.0 9.4
10 tus, whether the subject had cataracts, and whether she
5.6
2.6 had used short-acting benzodiazepines (Table 4).
0 Dietary calcium intake was not related to the risk of
1.1 1.9
Lowest 5 hip fracture. There was no increased risk among the 11
Third Middle 1.1 3–4 percent of women who ingested 400 mg of calcium or
Ca less per day (relative risk, 1.1; 95 percent confidence in-
lc ane Third
ors
al B
Highest 0 –2
F act terval, 0.5 to 2.3), even when women taking calcium
one Third isk
De fR supplements or estrogen were excluded.
nsi .o
ty No Estrogen is often prescribed for osteoporosis, a fact
that could lead to an underestimation in observational
Figure 2. Annual Risk of Hip Fracture According to the Number studies of the effectiveness of estrogen to prevent hip
of Risk Factors and the Age-Specific Calcaneal Bone Density. fracture. To minimize this bias, we analyzed the effect
The risk factors (from Table 2) are as follows: age 80; maternal of estrogen in women with no previous diagnosis of os-
history of hip fracture; any fracture (except hip fracture) since the teoporosis or fracture. In this group, estrogen therapy
age of 50; fair, poor, or very poor health; previous hyperthyroid-
ism; anticonvulsant therapy; current long-acting benzodiazepine appeared to have a strong protective effect (Table 4),
therapy; current weight less than at the age of 25; height at the but the confidence limits were wide.
age of 25 168 cm; caffeine intake more than the equivalent of Diabetes mellitus was not associated with a signif-
two cups of coffee per day; on feet 4 hours a day; no walking icantly increased risk (relative risk, 1.3; 95 percent
for exercise; inability to rise from chair without using arms; lowest
quartile (standard deviation 2.44) of depth perception; lowest
confidence interval, 0.8 to 2.1). After adjustment for
quartile (0.70 unit) of contrast sensitivity; and pulse rate obesity, there was a trend toward an increased risk of
80 per minute. hip fracture in diabetic women not taking insulin (rel-

The New England Journal of Medicine

Downloaded from nejm.org on March 23, 2024. For personal use only. No other uses without permission.
Copyright © 1995 Massachusetts Medical Society. All rights reserved.
Vol. 332 No. 12 RISK FACTORS FOR HIP FRACTURE IN WHITE WOMEN 771

Table 4. Factors Not Significantly Associated with Hip Fracture in characteristics of the proximal femur besides density,
Age-Adjusted Models. or perhaps a propensity to fall on the hip, may account
RELATIVE RISK (95%
for this familial predisposition.
VARIABLE COMPARISON CONFIDENCE INTERVAL) Many of the other risk factors are also believed to act
Hair color (as young adult) by reducing bone mass. However, adjustment for calca-
Black or brown Reference category 1.0 neal bone density did not substantially affect the risk of
Red Black or brown 0.9 (0.4–1.8) hip fracture associated with caffeine intake, a change
Blond Black or brown 1.1 (0.8–1.6)
Ancestry
in weight, walking for exercise, anticonvulsant drug
Northern European All others 1.2 (0.9–1.6) therapy, or a history of fracture. These factors may af-
Southern European All others 0.4 (0.2–1.1) fect the risk of hip fracture in other ways, perhaps by
Maternal fracture other than No maternal fracture 1.0 (0.6–1.7) influencing characteristics of bone other than density or
hip fracture
Children breast-fed Per child (none to 5) 0.9 (0.8–1.0)
by affecting the risk of falling.
Natural menopause before Menopause at or after 1.3 (0.9–1.8) Gaining weight reduces a woman’s risk of hip frac-
age of 45 age of 45 ture, and losing weight increases it. Previous studies
Bilateral oophorectomy All others 1.0 (0.7–1.6) have found that heavier women have a lower risk of hip
Past cigarette smoking Never smoked 1.3 (1.0–1.8) fracture.4,5,7 Weight loss may also be a marker for an
Daily dietary calcium intake Per 425 mg* 0.9 (0.8–1.1) underlying illness that increases risk. The high degree
Self-reported osteoarthritis No osteoarthritis 1.1 (0.8–1.6)
of correlation between weight change, current weight,
Unextracted cataracts No cataracts 1.1 (0.8–1.5)
and obesity limits our ability to determine whether the
Short-acting benzodiazepine No benzodiazepines 1.2 (0.8–2.1)
use risk of hip fracture is affected more by previous change
Aluminum-containing antacid None or less frequent 1.1 (0.7–1.7) or by current weight.
use at least weekly We confirmed that women who were tall when they
Current estrogen use Never used estrogen 1.0 (0.6–1.5) were young have a greater risk of hip fracture,8 per-
No history of osteoporosis Never used estrogen 0.3 (0.1–1.1)
or fracture haps because they fall farther.28,29 Taller women also
History of osteoporosis or Never used estrogen 1.3 (0.8–2.2) have a longer hip-axis length (the distance from the
fracture
greater trochanter to the inner pelvic brim), which has
Current thiazide diuretic use Never used thiazides 0.8 (0.6–1.2)
been associated with a greater risk of hip fracture.30
Current height Per 6.0 cm* 1.0 (0.8–1.1)
Modified Mini–Mental State 5 errors 1.5 (0.9–2.3)
Hyperthyroidism may or may not reduce bone
score mass.12,31-34 In our study, reduced bone mass did not ac-
Poorer visual acuity with Per 7.4 units* 1.1 (1.0–1.2) count for the strong association between previous hy-
usual correction perthyroidism and the risk of hip fracture. Hyperthy-
Orthostatic hypotension† No orthostatic change 1.2 (0.8–1.7)
roidism may cause long-lasting impairments of bone
*Value is equivalent to 1 SD. strength not detected by densitometry, or impairments
†Orthostatic hypotension was defined as a decrease in systolic blood pressure of 20 mm Hg of neuromuscular function not detected by our exami-
or more on standing.
nations. It can also reduce muscle strength. The asso-
ciation between resting tachycardia and hip fracture
ative risk, 1.6; 95 percent confidence interval, 0.9 to may be due to undiagnosed hyperthyroidism, although
2.7). There was also a trend toward an increased risk a faster pulse may also indicate decreased physical fit-
of hip fracture among those with a history of gastric ness or impaired cardiac function.
surgery (relative risk, 2.0), Parkinson’s disease (rela- Our finding that caffeine consumption increases the
tive risk, 1.6), and previous stroke (relative risk, 1.6), risk of hip fracture agrees with the findings of two pre-
but these conditions were uncommon and the lower 95 vious studies.6,35 Although a high caffeine intake has
percent confidence limit was less than 1.0 in each in- also been associated with reduced bone mass,12,36,37 our
stance. results suggest that caffeine may influence the risk of
hip fracture in other ways.
DISCUSSION Women who spent four hours or less per day on their
We found that many factors influence the risk of hip feet had a substantially increased risk of hip fracture,
fracture in older women and that the assessment of risk whereas walking for exercise reduced the risk. Exer-
factors and the measurement of bone density have com- cise may be a marker for health and functional status,
plementary value for the prediction of hip fracture. A but the association between activity and hip-fracture
small number of women with multiple risk factors and risk remained significant after adjustment for self-rated
low bone density have an especially high risk. They ac- health and findings on tests of neuromuscular function.
count for a large proportion of hip fractures and should We confirmed that therapy with long-acting benzodi-
be the focus of intensive efforts to prevent them. azepines increases the risk of hip fracture4,10; older
A woman whose mother had a hip fracture, especial- women should avoid these drugs. We also confirmed
ly before the age of 80, is at least twice as likely to have that women taking anticonvulsant drugs have a very
a hip fracture herself as a woman without such a ma- high risk of hip fracture,38 although none of these wom-
ternal history. Other types of maternal fractures did en had a hip fracture during a seizure. We previously
not increase hip-fracture risk, and the risk was inde- found no association between the use of anticonvulsant
pendent of bone mass, height, and weight. Inherited drugs and lower appendicular bone mass in this co-

The New England Journal of Medicine

Downloaded from nejm.org on March 23, 2024. For personal use only. No other uses without permission.
Copyright © 1995 Massachusetts Medical Society. All rights reserved.
772 THE NEW ENGLAND JOURNAL OF MEDICINE March 23, 1995

hort,12 suggesting that the increased risk might be due younger women, nursing home residents, or possibly
to impairments of neuromuscular function we did not women of other races.48 The study had limited power
measure. to evaluate risk factors that are relatively uncommon,
The inability to rise from a chair without using one’s such as Parkinson’s disease. Other risk factors, such as
arms is associated with an increased risk of falls39 and osteoarthritis and calcium intake, were based only on
a twofold increase in the risk of hip fracture. None of the women’s own reports. Measurement of hip instead
the other assessments of neuromuscular function add- of calcaneal bone density may improve the prediction
ed significantly to the prediction of subsequent hip of hip fracture. In addition, some of the risk factors
fracture. that are independent of calcaneal bone density may not
Poor depth perception and a reduced ability to per- be independent of hip bone density. Predictive models
ceive contrast (but not poor visual acuity) increased derived in one group may not perform as well in other
the risk of hip fracture independently. This suggests groups; ideally, our findings should be tested prospec-
that treatment or prevention of ophthalmologic condi- tively in other populations.
tions that impair depth perception and contrast sensi- We conclude that many factors increase the risk of
tivity, such as cataracts, diabetic retinopathy, and glau- hip fracture in older white women living in the commu-
coma, may help prevent hip fracture. nity. The effect of most individual factors is moderate,
Women who have had wrist fractures have a greater but together their impact is substantial. Women with
risk of hip fractures than those who have not had wrist multiple risk factors and low bone density are at espe-
fractures.40,41 Our results indicate that any type of post- cially high risk. A woman may be able to minimize her
menopausal fracture signals an increased risk of hip risk of hip fracture in a number of ways, notably by
fracture. This increased risk is independent of bone walking for exercise, avoiding long-acting sedative–hyp-
mass, implying that a history of fractures may indicate notic agents, reducing caffeine intake, quitting smoking,
an increased risk of falling or defects in bone strength treating impaired vision, and taking measures that
not detected by densitometry. maintain bone density.
Our analysis suggests that smoking increases the We are indebted to Stephen B. Hulley and Jennifer Kelsey for their
risk of hip fracture by limiting normal weight gain and support in the development of this study.
by its adverse effects on the health, neuromuscular fit-
ness, and exercise patterns of older women.27 The lower APPENDIX
risk for former smokers implies that quitting smoking The following investigators were members of the Study of Os-
diminishes the risk of hip fracture. teoporotic Fractures Research Group: University of California, San
Francisco (Coordinating Center): S.R. Cummings (principal investiga-
Although consistent with the view that current estro- tor), M.C. Nevitt (project director), D. Black (study statistician),
gen therapy protects against hip fracture,42 these re- H.K. Genant (director, central radiology laboratory), C. Arnaud,
sults are limited by the small number of women in our W. Browner, L. Christianson, M. Dockrell, C. Fox, C. Glüer, S. Har-
study who were taking estrogen. Another analysis of es- vey, M. Jergas, L. Palermo, A. Pressman, R. San Valentin, D. Seeley,
trogen therapy and fractures in this cohort has con- P. Steiger, and K. Stone; University of Maryland, Baltimore: R. Sherwin
(principal investigator), J. Scott (co-investigator), K. Fox (co-inves-
firmed that estrogen therapy protects against osteo- tigator), J. Lewis (project coordinator), G. Greenberg (clinic coordi-
porotic fractures.43 Our analysis illustrates the point nator), M. Bahr, S. Trusty, L. Finazzo, S. Snyder, E. Oliner, B. Hoh-
that observational studies will underestimate the effec- man, and T. Page; University of Minnesota, Minneapolis: K. Ensrud
tiveness of estrogen if they do not account for the fact (principal investigator), R. Grimm, Jr. (co-investigator), C. Bell
(project director), E. Mitson (study coordinator), I. Chavier, K. Ja-
that many women take estrogen for osteoporosis. Our cobson, S. Fillhouer, C. Shoberg, D. Michel, S. Estill, J. Hansen, and
study may underestimate the benefit in those with os- M. Baumhover; University of Pittsburgh, Pittsburgh: J.A. Cauley (prin-
teoporosis to the degree that estrogen is taken by wom- cipal investigator), L.H. Kuller (co–principal investigator), L. Harp-
en with more severe osteoporosis. er (project director), M. Nasim (clinic coordinator), C. Bashada,
The results of this study do not support widely held L. Buck, A. Githens, A. McCune, D. Medve, S. Rudovsky, and
N. Watson; Kaiser Permanente Center for Health Research, Portland, Oreg.:
beliefs that fair hair color, northern European ancestry, T.M. Vogt (principal investigator), W.M. Vollmer and E. Orwoll (co-
earlier natural menopause, and antacid therapy are as- investigators), J. Blank (project director), F. Heinith (clinic coordi-
sociated with an increased risk of hip fracture. In con- nator), R. Bright, J. Downing, B. Packer, C. Souvanlausky, L. Pud-
trast to other investigators,44,45 we found that moderate erbauth, and D. Franco.
alcohol intake did not increase the risk of hip fracture. REFERENCES
Although breast-feeding temporarily decreases bone
1. Cummings SR, Black DM, Rubin SM. Lifetime risks of hip, Colles’, or ver-
mass,46 it did not increase the risk of hip fracture. Like tebral fracture and coronary heart disease among white postmenopausal
most3,5,7 but not all47 prospective observational studies, women. Arch Intern Med 1989;149:2445-8.
ours found no relation between calcium intake and pro- 2. Paganini-Hill A, Ross RK, Gerkins VR, Henderson BE, Arthur M, Mack
TM. Menopausal estrogen therapy and hip fractures. Ann Intern Med 1981;
tection from hip fracture, even in women with very low 95:28-31.
intakes. However, we assessed calcium intake only once 3. Cooper C, Barker DJP, Wickham C. Physical activity, muscle strength, and
calcium intake in fracture of the proximal femur in Britain. BMJ 1988;297:
with a short questionnaire; repeated assessments may 1443-6.
be more accurate. 4. Grisso JA, Kelsey JL, Strom BL, et al. Risk factors for falls as a cause of
This study had several limitations. The participants hip fracture in women. N Engl J Med 1991;324:1326-31.
5. Farmer ME, Harris T, Madans JH, Wallace RB, Cornoni-Huntley J, White
were community-dwelling white women over the age of LR. Anthropometric indicators and hip fracture: the NHANES I epidemio-
65, so these findings are not generalizable to men, logic follow-up study. J Am Geriatr Soc 1989;37:9-16.

The New England Journal of Medicine

Downloaded from nejm.org on March 23, 2024. For personal use only. No other uses without permission.
Copyright © 1995 Massachusetts Medical Society. All rights reserved.
Vol. 332 No. 12 RISK FACTORS FOR HIP FRACTURE IN WHITE WOMEN 773

6. Kiel DP, Felson DT, Hannan MT, Anderson JJ, Wilson PWF. Caffeine and 28. Hayes WC, Myers ER, Morris JN, Gerhart TN, Yett HS, Lipsitz LA. Impact
the risk of hip fracture: the Framingham Study. Am J Epidemiol 1990;132: near the hip dominates fracture risk in elderly nursing home residents who
675-84. fall. Calcif Tissue Int 1993;52:192-8.
7. Paganini-Hill A, Chao A, Ross RK, Henderson BE. Exercise and other fac- 29. Nevitt MC, Cummings SR. Type of fall and risk of hip and wrist fractures:
tors in the prevention of hip fracture: the Leisure World study. Epidemiology the study of osteoporotic fractures. J Am Geriatr Soc 1993;41:1226-34.
1991;2:16-25. 30. Faulkner KG, Cummings SR, Black D, Palermo L, Glüer CC, Genant HK.
8. Meyer HE, Tverdal A, Falch JA. Risk factors for hip fracture in middle-aged Simple measurement of femoral geometry predicts hip fracture: the study of
Norwegian women and men. Am J Epidemiol 1993;137:1203-11. osteoporotic fractures. J Bone Miner Res 1993;8:1211-7.
9. Cummings SR, Black DM, Nevitt MC, et al. Appendicular bone density and 31. Fraser SA, Anderson JB, Smith DA, Wilson GM. Osteoporosis and fractures
age predict hip fracture in women. JAMA 1990;263:665-8. following thyrotoxicosis. Lancet 1971;1:981-3.
10. Ray WA, Griffin MR, Downey W. Benzodiazepines of long and short elim- 32. Fallon MD, Perry HM, Bergfeld M, Droke D, Teitelbaum SL, Avioli LV. Ex-
ination half-life and the risk of hip fracture. JAMA 1989;262:3303-7. ogenous hyperthyroidism with osteoporosis. Arch Intern Med 1983;143:
11. Cummings SR, Block G, McHenry K, Baron RB. Evaluation of two food 442-4.
frequency methods of measuring dietary calcium intake. Am J Epidemiol 33. Paul TL, Kerrigan J, Kelly AM, Braveman LE, Baran DT. Long-term L-thy-
1987;126:796-802. roxine therapy is associated with decreased hip bone density in premeno-
12. Bauer DC, Browner WS, Cauley JA, et al. Factors associated with appendic- pausal women. JAMA 1988;259:3137-41.
ular bone mass in older women. Ann Intern Med 1993;118:657-65. 34. Franklyn JA, Betteridge J, Daykin J, et al. Long-term thyroxine treatment
13. Kriska AM, Sandler RB, Cauley JA, LaPorte RE, Hom DL, Pambianco G. and bone mineral density. Lancet 1992;340:9-13.
The assessment of historical physical activity and its relation to adult bone 35. Hernández-Avila M, Colditz GA, Stampfer MJ, Rosner B, Speizer FE,
parameters. Am J Epidemiol 1988;127:1053-63. Willett WC. Caffeine, moderate alcohol intake, and risk of fractures of
14. National Center for Health Statistics, Fitti JE, Kovar MG. The supplement the hip and forearm in middle-aged women. Am J Clin Nutr 1991;54:157-
on aging to the 1984 National Health Interview Survey. Vital and health sta- 63.
tistics. Series 1. No. 21. Washington, D.C.: Government Printing Office, 36. Hernández-Avila M, Stampfer MJ, Ravnikar VA, et al. Caffeine and other
1987:49. (DHHS publication no. (PHS) 87-1323.) predictors of bone density among pre- and perimenopausal women. Epide-
15. Lohman TG, Roche AF, Martorell R. Anthropometric standardization refer- miology 1993;4:128-34. [Errata, Epidemiology 1993;4:279 and Epidemiol-
ence manual. Champaign, Ill.: Human Kinetics Books, 1988:177. ogy 1993;4:563.]
16. Bohannon RW. Manual muscle test scores and dynamometer test scores of 37. Cooper C, Atkinson EJ, Wahner HW, et al. Is caffeine consumption a risk
knee extension strength. Arch Phys Med Rehabil 1986;67:390-2. factor for osteoporosis? J Bone Miner Res 1992;7:465-71.
17. Magaziner J, Bassett SS, Hebel JR. Predicting performance on the Mini- 38. Annegers JF, Melton LJ, Sun CA, Hauser WA. Risk of age-related fractures
Mental State Examination: use of age- and education-specific equations. in patients with unprovoked seizures. Epilepsia 1989;30:348-55.
J Am Geriatr Soc 1987;35:996-1000. 39. Nevitt MC, Cummings SR, Kidd S, Black D. Risk factors for recurrent non-
18. Teng EL, Chui HC. The Modified Mini-Mental State (3MS) examination. syncopal falls: a prospective study. JAMA 1989;261:2663-8.
J Clin Psychiatry 1987;48:314-8. 40. Finsen V, Benum P. Colles’ fracture as an indicator of increased risk of hip
19. Bailey IL, Lovie JE. New design principles for visual acuity letter charts. fracture: an epidemiological study. Ann Chir Gynaecol 1987;76:114-8.
Am J Optom Physiol Opt 1976;53:740-5. 41. Mallmin H, Ljunghall S, Persson I, Naessen T, Krusemo UB, Bergström R.
20. Ginsburg AP. A new contrast sensitivity vision test chart. Am J Optom Phys- Fracture of the distal forearm as a forecaster of subsequent hip fracture: a
iol Opt 1984;64:403-7. population-based cohort study with 24 years of follow-up. Calcif Tissue Int
21. Gibson JJ. In: The perception of the visual world. Boston: Houghton Mifflin, 1993;52:269-72.
1950. 42. Grady D, Rubin SM, Petitti D, et al. Hormone therapy to prevent disease and
22. Ensrud KE, Nevitt MC, Yunis C, Hulley SB, Grimm RH, Cummings SR. prolong life in postmenopausal women. Ann Intern Med 1992;117:1016-37.
Postural hypotension and postural dizziness in elderly women: the study of 43. Cauley JA, Seeley DG, Ensrud K, Ettinger B, Black D, Cummings SR. Es-
osteoporotic fractures. Arch Intern Med 1992;152:1058-64. trogen replacement therapy and fractures in older women. Ann Intern Med
23. Cummings SR, Black DM, Nevitt MC, et al. Bone density at various sites 1995;122:9-16.
for prediction of hip fractures. Lancet 1993;341:72-5. 44. Felson DT, Kiel DP, Anderson JJ, Kannel WB. Alcohol consumption and hip
24. Steiger P, Cummings SR, Black DM, Spencer NE, Genant HK. Age-related fractures: the Framingham Study. Am J Epidemiol 1988;128:1102-10.
decrements in bone mineral density in women over 65. J Bone Miner Res 45. Hemenway D, Colditz GA, Willett WC, Stampfer MJ, Speizer FE. Fractures
1992;7:625-32. and lifestyle: effect of cigarette smoking, alcohol intake, and relative weight
25. Nevitt MC, Cummings SR, Browner WS, et al. The accuracy of self-report on the risk of hip and forearm fractures in middle-aged women. Am J Public
of fractures in elderly women: evidence from a prospective study. Am J Ep- Health 1988;78:1554-8.
idemiol 1992;135:490-9. 46. Sowers M, Corton G, Shapiro B, et al. Changes in bone density with lacta-
26. Hosmer DW Jr, Lemeshow S. Applied logistic regression. New York: John tion. JAMA 1993;269:3130-5.
Wiley, 1989:118-24. 47. Holbrook TL, Barrett-Connor E, Wingard DL. Dietary calcium and risk of
27. Nelson HD, Nevitt MC, Scott JC, Stone KL, Cummings SR. Smoking, al- hip fracture: 14-year prospective population study. Lancet 1988;2:1046-9.
cohol, and neuromuscular and physical function of older women. JAMA 48. Grisso JA, Kelsey JL, Strom BL, et al. Risk factors for hip fracture in black
1994;272:1825-31. women. N Engl J Med 1994;330:1555-9.

Massachusetts Medical Society

Registry on Continuing Medical Education
To obtain information on continuing medical education courses in the New England area,
call between 9:00 a.m. and 12:00 noon, Monday through Friday, (617) 893-4610 or in
Massachusetts 1-800-322-2303, ext. 1342.

The New England Journal of Medicine

Downloaded from nejm.org on March 23, 2024. For personal use only. No other uses without permission.
Copyright © 1995 Massachusetts Medical Society. All rights reserved.