Although many studies have examined aspects of the asso-ciation between physical activity and hip fracture1–21 import-ant details remain unclear. Few studies have focused on
whether physical activity early in life or later in life have similar effects,4,11,13,19although preliminary findings indicate that earlier physical activity is more protective than post-menopausal activity.4,19 If this is true, increased exercise in mid-life (or later) might not be effective in preventing hip fracture among older women. The importance of occupational physical activity is also unclear, since previous studies have found both increased13 and decreased12 risk among occupa-tionally active women. Finally, the association between exercise and other hip fracture risk factors has also not been clarified. For example, low body weight and weight loss are important fracture risk factors,13 but currently available data do not indicate whether physical activity should be recommended equally to women who have lost weight and to those who have gained weight during adult life. The aim of this study was to evaluate these questions in a large, population-based case-control study.
Physical activity and hip fracture:
a population-based case-control study
Bahman Y Farahmand,
aPer-Gunnar Persson,
aKarl Michaëlsson,
bJohn A Baron,
cAkke Alberts,
dTahere Moradi,
eSverker Ljunghall
ffor the Swedish Hip Fracture Study Group
gBackground A growing body of literature suggests that physical activity may be a protective factor against hip fracture.
Methods To study the association between hip fracture risk and recreational physical activity at various ages, changes in activity during adult life, occupational physical activity and how risks vary by adult weight change, we performed a population-based case-control study among postmenopausal women aged 50–81 years residing in six counties in Sweden in 1993–1995. The analysis consisted of 1327 women with hip fracture and 3262 randomly selected controls. Information on leisure physical activity before age 18, at 18–30 years and during recent years was based on a questionnaire. Data on occupational physical activity were collected through an independent classification of job titles obtained from record linkage with census data from 1960, 1970 and 1980.
Results There was a protective effect of recent leisure physical activity. Compared to women who reported no leisure activity, the odds ratios (OR) were 0.79 (95% CI : 0.62–1.00), 0.67 (95% CI : 0.54–0.84) and 0.48 (95% CI : 0.39–0.60) for women who exercised ,1 h per week, 1–2 h per week, and 3+ h per week, respectively. These decreased OR were more pronounced in women who had lost weight after 18 years of age than in those who had gained weight. Women with high physical activity at both 18–30 years and during recent years did not have a stronger protection than those with isolated high activity late in life, after accounting for recent activity. Occupational physical activity was not associated with hip fracture risk in this study.
Conclusions Recent physical activity is protective against hip fracture. The protective effect is most pronounced in women who had lost weight after age 18.
Keywords Case-control study, epidemiology, exercise, hip fracture, physical activity Accepted 20 October 1999
aDivision of Epidemiology, Karolinska Hospital, Stockholm County Council and Institute of Environmental Medicine, Karolinska Institutet, Sweden. bDepartment of Orthopaedics, University Hospital, Uppsala, Sweden. cDepartment of Community and Family Medicine, Dartmouth Medical
School, New Hampshire, USA.
dDepartment of Orthopaedics, Karolinska Hospital, Sweden.
eDepartment of Medical Epidemiology, Karolinska Institutet, Stockholm, Sweden.
fDepartment of Medicine, University Hospital, Uppsala, Sweden.
gThe Swedish Hip Fracture Study Group comprised Akke Alberts, John A Baron, Thomas Dolk, Bahman Y. Farahmand, Olof Johnell, Lena Lindén, Sverker Ljunghall, Karl Michaëlsson, Per-Gunnar Persson, K-G Thorngren, Mats Thorslund, Carl Zetterberg and Lena Zidén.
Reprint requests to: Bahman Y Farahmand, Division of Epidemiology/Norrbacka, S-171 76 Stockholm, Sweden. E-mail: [email protected]
Methods
The design of this Swedish population-based case-control study of hip fracture has previously been presented.22 Briefly, we attempted to include all new cases of fractures of the proximal femur that occurred during October 1993 to February 1995 in the Swedish counties of Stockholm, Uppsala, Västmanland, Örebro, Göteborg and Malmöhus among women born 1914 or later. Based on hospital discharge records or operation registers in all 24 hospitals in the study area, we identified 2597 possible cases. We found an additional 34 cases from the Swedish In-Patient Register who were also included. We reviewed medical records for all possible cases to ascertain type of fracture and previous hip fracture. We excluded women with a fracture due to malig-nancy (n = 26), high energy trauma (n = 4), incorrect diagnosis (n = 41), erroneous fracture date (n = 10), blindness (n = 5), birth outside of Sweden (n = 202), a diagnosis of severe alcoholic abuse, psychosis or senile dementia (n = 576) or death within 3 months of the fracture (n = 123), leaving 1644 eligible cases.
We used the Swedish population register to recruit controls. A random sample of 3307 women aged 70–80 years was drawn, frequency matched to the expected distribution of cases with regard to age (5-year age groups) and county of residence. Potential controls aged 50–69 years and also residents of the study population (n = 1565) were selected in a parallel case-control study of breast cancer.23Of a total of 4872 candidate controls, 610 were born outside Sweden, 157 died before we contacted them, 44 had senility or psychosis and 2 were blind so 4059 were eligible for the study.
We mailed all subjects a questionnaire regarding medical history and lifestyle habits. The questionnaire also requested information regarding weight at age 18, current weight and height and recreational physical activity.
On average, cases were interviewed 3 months after the hip fracture. Cases and controls who did not respond to the postal questionnaire were recontacted first by mail, then by telephone. A supplementary telephone interview was performed to complete and clarify responses for about 50% of participating subjects. Some women refused the full questionnaire but ac-cepted a less extensive telephone interview. Among the women contacted, 1328 (82.5%) cases and 3312 (81.6%) controls par-ticipated, of whom 202 (15.2%) of the cases and 497 (15.0%) of the controls responded exclusively to the less extensive telephone interview. All women who reported having natural menses (1 case and 50 controls) were classified as premeno-pausal and excluded from analysis.
Information regarding recreational physical activity was requested for three periods of life (before age 18, 18–30 years, and in recent years). This was assessed by the following question adopted from the MEDOS hip fracture study:19,24,25 ‘Do you or did you ever practice sports or physical exercise during leisure time?’ Subjects were provided with four response options: (1) never; (2) ,1 h weekly; (3) 1–2 h weekly and (4)
.2 h weekly. In some analyses, we combined the response options (1) and (2) into one category. Physical exercise during ages 18–30 was compared with activity during recent years to classify adult changes in activity. Responses were divided into five categories: unchanged low, decreased, unchanged moderate, increased and unchanged high activity. Cigarette smoking was defined as never, former or current. Use of postmenopausal oestrogens was divided into ever and never use and included, in
addition to medium potency oestrogens, oral or local treatment with oestriol or unspecified oestrogens. Body mass index (BMI) was calculated as the ratio of weight (kg) and height squared (m2). Changes in body mass were measured by comparing recent weight with the reported weight at age 18. The women were classified as having unchanged weight (±3 kg), increased (gain of 3+ kg) or decreased (loss of >3 kg) weight.
Information on occupational physical activity was obtained through record linkage with data collected during the national censuses in 1960, 1970 and 1980. Personal identification numbers were used to link study subjects to the census data. Using job titles, three independent specialists in occupational medicine classified the occupational physical demands as very high, high (agricultural worker, founder, metalworker, store-man and woodstore-man, etc.), moderate (nursery school teachers, midwives, nurses, upholsterer and glazier, etc.), light (admin-istrators, architects, chemists, engineers and meteorologists,) or sedentary (accountant, dental technician, lawyer’s work, tele-phone operator, watchmaker) activity. As previously described,26 we included only occupations for which two experts agreed perfectly and the third was within one category. This resulted in a total of 202 rated occupations. We were unable to score activity levels for occupations in art, music, housework, military, etc. (46 titles).
We used unconditional logistic regression analysis to obtain unconditional maximum likelihood estimates of odds ratios (OR) and 95% CI.27All OR were adjusted for age in five categories according to Table 1. Adjustment of leisure physical activity for occupational exercise, and vice versa produced similar estimates and are not shown. In Tables 2–4 we also adjusted for BMI in five categories as described in Table 1. Further adjustment one at a time for use of postmenopausal oestrogens, smoking and previous self-reported diseases (diabetes, infarction, hyper-tension or stroke) left the results materially unchanged and these analysis will not be presented. Interactions were assessed through product terms in the logistic models.
Results
Table 1 presents the distribution of cases and controls by age, BMI, cigarette smoking, use of postmenopausal oestrogen and previous diseases.
Recent physical exercise was associated with a decreased risk of hip fracture: women who were active >3 h per week had about a 50% reduction in risk as compared to sedentary women. This decreased risk was not affected by adjustment for activity before 18 years of age and at ages 18–30 years. In the minimally adjusted analysis, exercise at these earlier intervals was also associated with reduced hip fracture risk. However, these OR became close to unity after adjustment for recent recreational activity (Table 2).
The protective effect of recent physical activity varied modestly with adult weight change. Among women who lost >3 kg after age 18, recreational exercise for >3 h per week was associated with a 76% reduction in risk (OR = 0.24, 95% CI : 0.14–0.42) compared with those who were inactive. The corresponding estimate among those who gained >3 kg, was only a 38% reduction in risk (OR = 0.62, 95% CI : 0.46–0.82) (Table 3). The interaction of physical activity and weight change was statistic-ally significant (P = 0.05).
In Table 4 the age- and BMI-adjusted results appear to indicate that continued physical activity conferred a stronger protection than isolated activity late in life. However, after further adjustment for recent activity these decreased OR were all close to unity.
Occupational exercise had no consistent association with risk. Compared with subjects in the most sedentary jobs in 1960, 1970 or 1980, those reporting more active work had OR close to unity. Even the most active occupations did not confer a materially decreased risk. Analyses incorporating occupational activity reflected by two consecutive censuses also did not show any relationship (data not shown).
Discussion
This large population-based case-control study found a decreased risk of hip fracture in women with high physical activity in the years before interview. This decreased risk was particularly pronounced in women who had not gained weight
after early adulthood. Change in physical activity after age 18–30 was not associated with hip fracture risk after adjustment for recent activity. Occupational physical activity based on job titles from censuses 1960, 1970 and 1980 was not associated with hip fracture.
Although many previous studies have reported a decreased risk of hip fracture associated with high recent physical activity,2–21there are relatively limited data regarding exercise earlier in life. In the present study no effect of early activity was seen after adjustment for recent activity. A previous inves-tigation11found a more pronounced protective effect of earlier activity than that in the years before the fracture, but two other studies4,19observed slightly stronger effects for current activity. In these analyses, however, earlier activity was not taken into account in the evaluation of current activity, and vice versa.
The information we obtained enabled us to evaluate whether changes in physical exercise patterns influence hip fracture risk. To our knowledge, this has not been investigated previously. Women with increased physical activity and those who main-tained high activity at both younger and older ages experienced the lowest risks of hip fracture. However, these decreased risks were all explained by recent activity. Thus, it is never too late to start to be physically active.
The importance of exercise for bone mass can be seen from two perspectives. First, although the importance of peak mass for fracture risk in the elderly remains to be clarified, it is likely to affect bone mass late in life. Peak bone mass, achieved in early adult life, might be influenced by the skeletal loading associated with physical activity during childhood, puberty, and early adult life. Some cross-sectional studies have demonstrated a slightly positive association between bone mass in children and adolescents and their physical activity28–30although others have been unable to find such correlations.31Second, even after peak bone mass is achieved, bone tissue adapts to mechanical loading; bone mass will be reduced unless there are daily physical stresses in similar amounts as previously. Indeed, weight-bearing exercise in sedentary postmenopausal women prevents bone loss and may induce modest increases in bone mass.32These basic considerations indicate that only continued and on-going activity would be expected to be of benefit, even though present activity is the most important. Strength training exercise is effective among postmenopausal women not only in preserv-ing bone mass but also in improvpreserv-ing muscle mass, strength and balance.32,33 Among elderly individuals, risks for falls is an important determinant of hip fracture risk in addition to bone density. The severity of the fall may be reduced by improved muscle strength and balance.34
In our study, the protective effect of physical activity was particularly strong in women who lost weight after age 18. These results support findings from a recent cohort study in which physical activity had a beneficial effect on bone density in women who lost weight or whose weight remained stable, but not in women who gained weight.35The most important mechanisms thought to explain increased osteoporosis risk in women with weight loss are reduced mechanical loading of the skeleton and decreased extraglandular production of oestrogens in adipose tissue—both well-known stimuli for bone gain and preservation.36 Furthermore, women with low mechanical bone loading as a result of low physical activity level, low body weight, or a low adult weight gain seem to benefit the most
Table 1 Number of study women by case-control status
Cases Controls No. % No. % Age (years) ,60 101 7.6 291 8.9 60–64 162 12.2 368 11.3 65–69 337 25.4 838 25.7 70–74 640 48.2 1352 41.4 75–81 87 6.6 413 12.7
Body mass index (kg/m2)
,21.75 532 41.1 641 19.9 21.75–23.75 291 22.5 649 20.2 23.76–25.50 209 16.2 620 19.3 25.51–28.00 170 13.1 663 20.6 28.01+ 92 7.1 643 20.0 Cigarette smoking Never 719 54.3 1872 60.7 Former 260 19.6 619 20.0 Current 345 26.1 595 19.3
Use of postmenopausal oestrogena
Never 956 72.0 2073 63.5 Ever 371 28.0 1189 36.5 Previous diseases Diabetes mellitus no 115 10.2 194 7.0 yes 1012 89.8 2569 93.0 Myocardial infarction no 82 8.6 133 5.5 yes 872 91.4 2278 94.5 Hypertension no 349 33.8 828 32.5 yes 685 66.2 1717 67.5 Stroke no 77 8.1 72 3.0 yes 872 91.9 2290 97.0
Table 2Odds ratios (OR) and 95% CI of hip fracture with physical exercise before age 18 years, 18–30 years and during recent years No. of
Physical exercise at certain ages Controls Cases ORa(95% CI) ORb(95% CI)
Before age 18 (hours/week)
Never 708 342 1.00 (reference) 1.00 (reference)
,1 289 130 0.85 (0.65–1.10) 1.03 (0.76–1.39)
1–2 841 357 0.83 (0.69–1.00) 1.10 (0.86–1.40)
3+ 1332 470 0.68 (0.57–0.81) 1.06 (0.83–1.37)
χ2test for trend (P-value) P = 0.63
At ages 18–30 (hours/week)
Never 716 376 1.00 (reference) 1.00 (reference)
,1 399 174 0.78 (0.61–0.98) 0.88 (0.66–1.16)
1–2 880 350 0.72 (0.60–0.87) 0.89 (0.69–1.15)
3+ 1174 397 0.57 (0.48–0.69) 0.85 (0.65–1.11)
χ2test for trend (P-value) P = 0.28
Recent years (hours/week)
Never 870 511 1.00 (reference) 1.00 (reference)
,1 454 192 0.75 (0.60–0.93) 0.79 (0.62–1.00)
1–2 805 300 0.64 (0.53–0.77) 0.67 (0.54–0.84)
3+ 1045 306 0.45 (0.38–0.54) 0.48 (0.39–0.60)
χ2test for trend (P-value) P = 0.0001
aAdjusted for body mass index and age. b
Adjusted for other variables in the Table in addition to body mass index and age.
Table 3Odds ratios (OR) and 95% CI of hip fracture with physical exercise during recent years among subgroups of weight change after age 18 years
Subgroups of weight change and No. of
physical exercise during recent years Controls Cases ORa(95% CI)
Lost weight, >3 kg:
Recent physical activity (hours/week)
Never 43 85 1.00 (reference)
,1 25 29 0.64 (0.32–1.27)
1–2 53 48 0.45 (0.26–0.80)
3+ 77 36 0.24 (0.14–0.42)
χ2test for trend (P-value) P = 0.0001
Unchanged weight, –3 to 3 kg:
Recent physical activity (hours/week)
Never 64 60 1.00 (reference)
,1 42 30 0.79 (0.43–1.44)
1–2 83 46 0.61 (0.36–1.02)
3+ 115 35 0.33 (0.20–0.57)
χ2test for trend (P-value) P = 0.0001
Gained weight, >3 kg:
Recent physical activity (hours/week)
Never 418 133 1.00 (reference)
,1 257 61 0.79 (0.55–1.12)
1–2 470 119 0.79 (0.59–1.05)
3+ 530 115 0.62 (0.46–0.82)
χ2test for trend (P-value) P = 0.02
from hormone replacement therapy to reduce the risk of hip fracture.37We thus suggest that interventions such as oestrogen supplementation or increased physical activity should best be targeted to those with greatest risk of fracture to gain optimal relative benefit from the treatment.
Few previous studies have focused on the association between occupational physical activity and hip fracture, and the results have been conflicting. A case-control study from Australia13 found that high levels of occupational physical activity at age 20 increased fracture risk, while activity at age 50 was not associated with risk. In contrast, a case-control study from England8found that sedentary work at age 50 was asso-ciated with a markedly increased risk of hip fracture, and a large case-control study19reported that occupational physical activity during the recent past was protective, but not earlier activity.
In a case-control study from Toronto, women with more than 20 years experience in physically active jobs had about 50% the hip fracture risk of women with more than 20 years of less active work.17In a large cohort study from Norway, intensive physical activity at work was also associated with more than a 50% reduction in the incidence of hip fracture.12
Census data has the advantage of not being biased by case-control status or perceptions of physical activity. Nonetheless, information on physical activity from job titles is somewhat crude and it is likely that misclassification biased our findings toward unity. Moreover, less than 50% of our subjects had employment outside the home and could be classified with regard to occupational physical activity. Thus the analyses using these data had limited statistical power. Moreover, since we found no effect of early recreational activity, it is possible that
Table 4 Odds ratios (OR) and 95% CI of hip fracture with changes in physical exercise between ages 18–30 and during recent years
Change in physical exercise No. of
between age 18–30 and recent years Controls Cases ORa(95% CI) ORb(95% CI)
Unchanged low (,1 h/week) 814 438 1.00 (reference) 1.00 (reference)
Decreased 782 346 0.81 (0.68–0.97) 1.05 (0.84–1.31)
Unchanged moderate (1–2 h/week) 354 140 0.70 (0.55–0.89) 1.23 (0.83–1.82)
Increased 498 174 0.66 (0.53–0.82) 1.40 (0.90–2.19)
Unchanged high (3+ h/week) 713 197 0.44 (0.36–0.54) 1.05 (0.62 –1.78)
χ2test for trend (P-value) P= 0.0001 P = 0.70
aAdjusted for body mass index and age. b
Adjusted for recent physical activity in addition to body mass index and age.
Table 5 Age-adjusted odds ratios (OR) and 95% CI of hip fracture with occupational physical activity based on occupational titles in census 1960, 1970 and 1980
Occupational physical activity No. of
in 1960, 1970 and 1980 Controls Cases OR (95% CI)
1960 Sedentary 385 171 1.00 (reference) Light 365 134 0.78 (0.59–1.02) Medium 247 119 1.03 (0.77–1.37) High/very high 136 48 0.71 (0.49–1.04) 1970 Sedentary 368 150 1.00 (reference) Light 405 180 1.09 (0.84–1.42) Medium 479 174 0.84 (0.65–1.09) High/very high 283 127 1.04 (0.78–1.38) 1980 Sedentary 631 224 1.00 (reference) Light 433 147 0.97 (0.76–1.24) Medium 416 120 0.82 (0.64–1.06) High/very high 265 101 1.06 (0.80–1.40)
No. of controls (cases)
1960 1970 1980
Work at home 1787 (698) 1184 (455) 1272 (631)
No scorea 230 (115) 308 (151) 217 (92)
Unknownb 112 (42) 235 (90) 28 (12)
aExperts were not agreed on occupational physical demand. bClassified as unspecified by Statistics Sweden.
the time periods referred to by census data (ages 25–66) are sufficiently early to be unrelated to hip fracture risk at later ages.
The misclassification of information on leisure physical activity is also a potential source of error in this study. Moreover, it is likely that data on earlier physical exercise may be less accurate than that for recent activity and this differential measurement error could explain why no association with early activity was found after adjustment for current activity. Few studies have specifically investigated the magnitude of agreement between originally reported information on physical activity and corres-ponding retrospective information. Retrospective information on vigorous activity would probably be more reliable than that for less intensive activity,38,39and retrospective information on physical activity up to 10 years previously may be reported relatively well.38Thus, we believe that information on physical exercise was non-differentially misclassified, and if anything a conservative bias may be present in our data.
The strengths of the present study include its large size and population-based design, and our aggressive attempts to identify all cases. Some possible cases and controls (e.g. those with known senility, alcoholism and psychosis) were excluded because of the relatively complex questionnaire. We were also able to account for many potential confounding factors.
Our study suggests that recent leisure physical activity has a protective influence on risk of hip fracture. This protective effect varied moderately with weight change after age 18. Women who lose weight during adult life have the highest risks of frac-ture and so are also those who may benefit most from exercise. We found that even among women who have been inactive previously, the potential preventive effect of exercise may be substantial. Preventive actions should stimulate sedentary women to start to exercise and these actions should be targeted specifically on women who have lost weight during adult life.
Acknowledgements
This study was supported by grants from the Swedish Council for Social Research (project 93-0029) and US National Institutes of Health (NR 1 RO 1 CA58427).
References
1Meredith CN. Exercise in the prevention of osteoporosis. In: Munro H,
Schlierf G (eds). Nutrition of the Elderly. New York: Vevey/Raven Press, 1992;29:169–75.
2Paganini-Hill A, Ross RK, Gerkins VR, Henderson BE, Arthur M,
Mack TM. Menopausal estrogen therapy and hip fracture. Ann Intern Med 1981;95:28–31.
3Åström J, Ahnqvist S, Beertema J, Jonsson B. Physical activity in
women sustaining fracture of the neck of the femur. J Bone Joint Surg 1987;69:381–83.
4Lau E, Donnan S, Barker DJP. Physical activity and calcium intake
in fracture of the proximal femur in Hong Kong. Br Med J1988;297: 1441–43.
5Cooper C, Barker DJP, Wickham C. Physical activity, muscle strength,
and calcium intake in fracture of the proximal femur in Britain. Br Med J 1988;297:1443–46.
6Wickham CAC, Walsh K, Cooper C et al.Dietary calcium, physical
activity, and risk of hip fracture: a prospective study. Br Med J 1989;299:889–92.
7Farmer ME, Harris T, Madans JH, Wallace RB, Cornoni-Huntley J,
White LR. Anthropometric indicators and hip fracture. The NHANES I epidemiologic follow-up study. J Am Geriatr Soc 1989;37:9–16.
8Cooper C, Wickham C, Coggon D. Sedentary work in middle life and
fracture of the proximal femur. Br J Ind Med 1990;47:69–70.
9Paganini-Hill A, Chao A, Ross RK, Henderson BE. Exercise and other
factors in the prevention of hip fracture: the Leisure World study. Epidemiology 1991;2:16–25.
10Nieves JW, Grisso JA, Kelsey JL. A case-control study of hip fracture:
evaluation of selected dietary variables and teenage physical activity. Osteoporosis Int 1992;2:122–27.
11Jaglal SB, Kreiger N, Darlington GA. Past and recent physical activity
and risk of hip fracture. Am J Epidemiol 1993;138:107–18.
12Meyer HE, Tverdal A, Falch JA. Risk factors for hip fracture in
middle-aged Norweigian women and men.Am J Epidemiol 1993;137:1203–11.
13Cumming RG, Klineberg RJ. Case-control study of risk factors for hip
fractures in the elderly. Am J Epidemiol 1994;139:493–503.
14Coupland C, Wood D, Cooper C. Physical inactivity is an independent
risk factor for hip fracture in the elderly. J Epidemiol Community Health 1993;47:441–43.
15Cummings SR, Nevitt MC, Browner WSet al.Risk factors for hip
fracture in white women. N Engl J Med 1995;332:767–73.
16Meyer HE, Henriksen C, Falch JA, Pedersen JI, Tverdal A. Risk factors
for hip fracture in a high incidence area: a case-control study from Oslo, Norway. Osteoporosis Int 1995;5:239–46.
17Jaglal SB, Kreiger N, Darlington GA. Lifetime occupational physical
activity and risk of hip fracture in women. Ann Epidemiol 1995;5: 321–24.
18Michaëlsson K, Holmberg L, Mallmin Het al.Diet and hip fracture
risk: a case-control study. Int J Epidemiol 1995;24:771–82.
19Johnell O, Gullberg B, Kanis JAet al.Risk factors for hip fracture in
European women: the Medos study. J Bone Miner Res1995;10:1802–15.
20Huang Z, Himes JH, McGovern PG. Nutrition and subsequent hip
fracture risk among a national cohort of white women. Am J Epidemiol 1996;144:124–34.
21Grisso JA Kelsey JL, O’Brien LA et al.Risk factors for hip fracture in
men. Am J Epidemiol 1997;145:786–93.
22Michaëlsson K, Baron JA, Farahmand BY et al.Hormone replacement
therapy and hip fracture risk. Br Med J1998;316:1858–63.
23Magnusson C, Baron J, Persson I et al. Body size in different periods
of life and breast cancer risk in post-menopausal women. Int J Cancer 1998;76:29–34.
24Dequeker J, Ranstam J, Valsson J, Sigurgevisson B, Allander E, the
Medos study group. The Mediterranean osteoporosis (MEDOS) study questionnaire. Clin Rheumatol1991;10:54–72.
25Dequeker J, Tobing L, Rutten V, Geusens P, Medos study group.
Relative risk factors for osteoporotic fracture: a pilot study of the Medos questionnaire. Clin Rheumatol1991;10:49–53.
26Moradi T, Nyrén O, Bergström R et al.Risk for endometrial cancer in
relation to occupational physical activity: a nationwide cohort study in Sweden. Int J Cancer 1998;76:665–70.
27SAS Institute Inc. SAS Technical Report, SAS/STAT Software: The
GENMOD Procedure, Version 6. Cary, NC: SAS Institute Inc., 1994.
28Ruiz JC, Mandel C, Garabedian M. Influence of spontaneous calcium
intake and physical exercise on the verebral and femoral bone mineral density of children and adolescents. J Bone Miner Res 1995;10:675–82.
29Slemenda CW, Miller JZ, Hui SL, Reister TK, Johnston CC Jr. Role of
physical activity in the development of skeletal mass in children. J Bone Miner Res 1991;6:1227–33.
30Kroger H, Heikkinen J, Laitinen K, Kotaniemi A. Dual-energy X-ray
absorptiometry in normal women: a cross-sectional study of 717 Finnish volunteers. Osteoporosis Int 1992;2:135–40.
31Young D, Hopper JL, Nowson CAet al.Determinants of bone mass in
10- to 26-year-old females: a twin study. J Bone Miner Res1995;10: 558–67.
32Dalsky GP, Stocke KS, Ehsani AA, Slatopolsky E, Lee WC, Birge SJ.
Weight-bearing exercise training and lumbar bone mineral content in postmenopausal women. Ann Intern Med 1988;108:824–28.
33Nelson ME, Fiatarone MA, Morganti CM, Trice I, Greenberg RA,
Evans WJ. Effects of high-intensity strength training on multiple risk factors for osteoporotic fractures. A randomized controlled trial. JAMA 1994;272:1909–14.
34Greenspan SL, Myers ER, Maitland LA, Resnick NM, Hayes WC. Fall
severity and bone mineral density as risk factors for hip fracture in ambulatory elderly. JAMA 1994;271:128–33.
35Nguyen TV, Sambrook PN, Eisman JA. Bone loss, physical activity,
and weight change in elderly women: the Dubbo osteoporosis epidemiology study. J Bone Miner Res 1998;13:1458–67.
36Frost MH, Obesity, and bone strength and ‘mass’: a tutorial based on
insight from a new paradigm. Bone 1997;21:211–14.
37Michaelsson K, Baron JA, Johnell O, Persson I, Ljunghall S. Variation
in the efficacy of hormone replacement therapy in the prevention of hip fracture. Osteoporosis Int 1998;8:540–46.
38Blair SN, Dowda M, Pate RRet al.Reliability of long-term recall of
participation in physical activity by middle-aged men and women. Am J Epidemiol 1991;133:266–75.
39Slattery ML, Jacobs DR Jr. Assessment of ability to recall physical
