One in six U.S. white women 50 years of age will sustain a hip fracture during her remaining life 1,2; the risk is even higher in Scandinavian countries. 3,4 Personal characteristics clearly affect this risk: low body weight and tall stature have consistently been identified as risk factors for hip fracture, 5–18 and in recent years, adult weight change has also been recognized to affect strongly osteoporotic fracture risk. 12–18 The manner in which these body measures affect hip fracture risk is uncertain; however, it is not clear whether body weight explains the effect of weight gain or vice versa, nor is it known whether body weight in early adult life is related to risk. Similarly, although it seems reasonable to assume that different weight-related biomechanical factors may operate in taller women than in shorter women, this interaction has never been analyzed. To clarify these issues, we used data from a large hip fracture study in postmenopausal Swedish women.
Subjects and Methods
This population-based case-control study was conducted in six counties of south-central Sweden: Stockholm, Uppsala, Västmanland, Örebro, Göteborg, and Malmöhus. These counties included nearly half of the 8.6 million inhabitants of Sweden during the study period October 1993 to February 1995. The study protocol was approved by the local ethics committees in the areas involved.
We aimed to ascertain all fractures of the cervical, intertrochanteric, or subtrochanteric regions of the proximal femur that occurred between October 1993 and February 1995 among native-born Swedish women resident in the study area who were born after 1913. All hospital records and x-ray reports were scrutinized to confirm eligibility and ascertain type of hip fracture. We identified 2,597 possible hip fracture cases, of which 1,610 were eligible. We excluded cases with a fracture due to malignancy (N = 26), high-energy trauma (mainly traffic accidents, N = 4), incorrect diagnosis (N = 41), previous fracture (N = 10), blindness (N = 5), birth outside Sweden (N = 202), severe alcohol abuse, psychosis or senile dementia (N = 576), and death within 3 months of the fracture (N = 123). Subjects were approached with a comprehensive questionnaire at a mean interval of 95 days after the fracture.
Potential controls were women born in Sweden in 1914 or later and randomly selected from a continuously updated population registry the month before the start of study. Questionnaires were sent to controls on six occasions evenly distributed during the study period. Two-thirds of the women selected (N = 3,307), all age 70–80 years, were frequency matched (2 controls:1 case) to the expected hip fracture age distribution, within county of residence. The remainder of the controls (age 50–69 years) were residents of the study area who were randomly selected from the population register as possible controls for a case-control study of breast cancer being conducted at the same time with the same questionnaire. Of the 4,872 candidate controls, 4,059 were eligible and 813 were excluded. The exclusions comprised 610 women born outside of Sweden, 157 who died before being approached, 44 who had senility or psychosis, and 2 who were blind.
We mailed a questionnaire to potential subjects asking about information on current height and weight, weight 1 year previously, and weight at age 18. We also requested information on education, leisure time physical activity at three different time periods (before age 18, between ages 18 and 30, and in recent years), dietary habits, recent alcohol consumption, and smoking habits. The women were also asked about medical history such as self-reported stroke, diabetes mellitus, cardiovascular and inflammatory bowel diseases, age at menarche and menopause, parity, and use of oral contraceptives and postmenopausal hormone replacement therapy.
Approximately 50% of participants were approached by telephone for completion of missing information. Some women refused the postal questionnaire but accepted a less extensive telephone interview. Of those eligible, 1,328 cases (82.5%) and 3,312 controls (81.6%) answered the questionnaire; of these, 202 (15.2%) of the cases and 497 (15.0%) of the controls responded solely by telephone.
We used odds ratios (ORs) with 95% confidence intervals (CIs) as measures of association. These were estimated by stratified analyses using Mantel-Haenszel estimates and test-based 95% CIs. 19 We categorized body height, current weight, and weight at age 18 into quintiles according to the distribution of the controls. In analyses stratified by adult weight change (lost or unchanged, or gained), we assessed the association between different categories of body measures and hip fracture risk.
All ORs were controlled for age in five categories (<60, 60–64, 65–69, 70–74, and 75–81 years), as well as for current height or weight. Additional covariates were added one at a time to the model containing age, height or weight, and exposure variable of interest. The covariates considered were age at menopause (<50 or ≥50 years), menarche (<13 or ≥13 years), climacteric symptoms (yes or no), parity (0, 1, 2, or 3 or more children), use of oral contraceptives (ever or never), use of hormone replacement therapy (ever or never), stroke (yes or no), diabetes mellitus (yes or no), cardiovascular diseases (yes or no), inflammatory bowel diseases (yes or no), cigarette smoking (never, former, or current), current intake of alcohol (yes or no), usual intake of coffee and milk (fewer than two, or two or more, cups or glasses per day), use of vitamin supplements (multivitamin, vitamins C and E, calcium, and iron; yes or no), and leisure-time physical activity before age 18, between ages 18 and 30, and in recent years (less than two, or two or more, times per week). Controlling for these factors in addition to age, height, and body weight variables changed the findings only minimally; consequently, estimates from these models are not presented. Current weight was strongly correlated both with weight 1 year previously and with current body mass index (kg/m 2) (r = 0.94 and 0.91, respectively), and ORs for these variables were approximately the same as for current weight; therefore, we used only current weight here. Participants claiming natural menses were classified as premenopausal (50 controls and 1 case) and were excluded from the analysis.
The characteristics of cases and controls are shown in Table 1. Cases were on average 2 years older, slightly taller, and more slender than controls. Cases were 1.3 kg heavier at age 18, but 5.8 kg lighter at the time of study. The mean weight gain after age 18 was 11.0 kg in controls vs 4.3 kg for cases.
Separate Effects of Body Size Factors
Cases tended to be taller than controls, with mean heights (± standard deviation) of 164.1 ± 6.6 and 163.3 ± 5.9 cm. The tallest group (≥169 cm) had three times the hip fracture risk of women shorter than 159 cm (Table 2).
Current weight was also strongly associated with risk of hip fracture. Women who weighed less than 58 kg had an OR of 6.29 (95% CI = 4.82–8.18) compared with women heavier than 75 kg after controlling for age and height. This association was attenuated after additional control for weight change (OR = 3.54; 95% CI = 2.30–5.46). In contrast, weight at age 18 conferred a modestly increased risk, which disappeared after control for weight change (Table 2).
Weight gain between age 18 and the time of the study was strongly and inversely related to hip fracture risk. The effect of weight loss was almost identical in magnitude to that for weight gain, although opposite in sign. Compared with those with moderate weight change (–3 to 3 kg), those who gained more than 12 kg had an OR of 0.35 (95% CI = 0.27–0.45), and those with substantial weight loss (≥12 kg) had a markedly increased risk of hip fracture (OR = 3.29; 95% CI = 1.98–5.48) (Table 2).
Clarification of the independent effects of the body size factors was complicated by their interrelationships. Adult weight change and height retained strong effects even after control for current weight and weight at age 18; however, weight and height modified each other’s effects. Current weight was a stronger risk factor among shorter women: the OR (<53 vs ≥61 kg) was 5.31 (95% CI = 3.96–7.11) for women less than 161 cm tall, but 3.09 (95% CI = 2.01–4.73) for those taller than 165 cm. Conversely, height was a particularly strong risk factor among heavier subjects. For women weighing less than 53 kg, the OR (≥166 vs <161 cm) was 1.34 (95% CI = 0.82–2.18) compared with an OR of 2.35 (95% CI = 1.85–2.98) for women weighing more than 60 kg. As a consequence of these interactions, tall women with relatively low weight had particularly high risk of hip fracture. Compared with women less than 161 cm tall and weighing more than 70 kg, those who were taller than 165 cm and who weighed 65 kg or less had an OR of 8.81 (95% CI = 5.68–13.68).
The effects of current weight and weight at age 18 could not be interpreted without taking into account weight change. Whether or not a woman gained weight as an adult demarcated different patterns of effect for the other weight-related factors. Among women who gained weight as adults, current weight and weight at age 18 had no effect on hip fracture risk after controlling for the amount of gain (Table 3). Among the minority of women who did not gain weight, however, the effects were different. Current weight retained a weak protective effect after controlling for the amount of loss, and weight at age 18 became modestly protective (Table 3).
The importance of adult weight gain on hip fracture risk is further illustrated in Figure 1, which presents hip fracture risks associated with various combinations of weight at age 18 and weight at current age. Women who were in the highest weight tertile at age 18 (≥59 kg) and lowest weight tertile at recent age (<61 kg) had a higher risk of hip fracture than women who were thin at age 18 and gained weight to reach the highest tertile at recent age (OR = 6.58; 95% CI = 3.99–10.86).
In this large, population-based study, we found that adult weight change and height were the dominant body size risk factors. Adult weight change (particularly weight loss vs weight gain) demarcated different patterns of hip fracture risk. Current weight had protective effects independent of weight change only among the minority of women who did not gain weight as adults. Whereas overall weight at age 18 appeared not to be an independent risk factor for hip fracture, it was modestly protective among women who did not gain weight. Height was an especially strong risk factor among heavy women and those with prominent weight gain.
Previous studies have reported an increased risk of hip fracture among tall women, 5–6,8–11,14,16,17 although our finding of a particularly large effect of height among heavy women is new. Presumably, the increased fracture risk among tall women is the consequence of biomechanical factors. One of these might be hip axis length, a risk factor for hip fracture that is correlated with height independently of age and bone density. 10 The interaction with weight implies that these height-related factors are exacerbated by body weight.
Although it is well known that high body weight is protective against hip fracture, 5–10,12,16,17 emerging evidence now points to an independent effect of weight change. 12–18 Our data, and that of some prior studies regarding hip fracture 17 or frailty fractures in general, 12 suggest that weight change explains much of the effect of current body weight.
In the Study of Osteoporotic Fractures, 12 the relation between frailty fractures and weight loss in women age 65 years or older was due to involuntary, not voluntary, loss and was probably mediated by lower bone mineral density. 13 A cohort study from Norway 14 showed a markedly increased risk of hip fracture among women who lost weight in the previous year as a result of poor appetite, implicating low body weight from nutritional deficiency.
Weight change probably reflects the time-integrated effects of a woman’s lifetime weight history and its impact on bone. Several recent studies have reported finding low levels of circulating or bone growth factors such as insulin-like growth factor-I in patients with osteoporosis. Hepatic production of insulin-like growth factor-I, under the influence of growth hormone, is dependent on nutritional status, which thus could underlie the association between weight loss, bone density, and hip fracture risk. 20–24 Weight loss may also be a marker of other health problems that increase risk: low bone density is associated with an increased risk of mortality. 25 Alternatively, weight gain may simply be a particularly good measure of adiposity, a protective factor against fractures.
The inverse relation between body weight and hip fracture risk has often been attributed to several hormonal and mechanical factors. Obese postmenopausal women have higher estrogen production and lower levels of sex hormone binding globulin than thin women; consequently, they have higher circulating levels of biologically available estrogen. 26,27 The larger muscle forces needed to move obese bodies during work might also be a stimulus to bone formation. 28 Soft tissue padding around the hip might also reduce trauma during a fall. 29,30 The dominance of weight gain on hip fracture risk suggests that these factors may be particularly pronounced in the context of weight gain.
The strengths of our study include its population-based design, large size, high participation rate among both cases and controls, and information on anthropometric measures at different ages. We restricted the study population to those born in Sweden to reduce genetic and cultural differences among the participants. Further, we excluded women with known senility, alcoholism, and psychosis because of the likelihood of inaccurate responses. We also had the opportunity to take into account possible confounding factors.
Our study has several potential limitations. We cannot rule out the possibility of misclassification of height, because we used recent height in our calculations rather than maximum height. Height loss could have been more pronounced among patients with hip fractures than among controls, because the osteoporotic women may have had more prevalent spinal fractures and consequent loss of height. 31 This type of differential misclassification of height responses, as well as nondifferential misclassification, would have conservatively biased the associations we found. Our reliance on self-reported height may partially mitigate this potential bias, because self-reported height tends to be skewed toward maximum height. 32 In a study of long-term recall of childhood and adolescent body measures, 33 there were high correlations between measured and self-reported values (r = 0.84 for weight and r = 0.92 for height), although reported weight was more accurately recalled in lean than in obese adolescents (underestimate, 0.9 vs 2.3 kg). Similarly, differences between self-reported and measured weight and height among Swedish women have been found to be small, 32 although short and thin subjects have a tendency to overestimate their size, whereas tall and heavy subjects underestimate their measures. Consequently, it is likely that the associations we observed reflect true relations.
The subjects with missing data on weight at age 18 (cases 37.8% and controls 32.3%, respectively) did not differ much from other respondents with regard to the distributions of current weight and height. Analyses limited to the women providing data on weight at age 18 yielded point estimates regarding other anthropometric factors that were similar to point values observed before restriction, which suggests that women included in the analysis are not a biased subset of subjects.
We thank Hans-Olof Adami (Uppsala, Sweden), research nurse Lena Lindén (Uppsala, Sweden), and interviewer Birgit Wallander (Stockholm, Sweden). Members of the Swedish Hip Fracture Study Group are 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.
1. Cummings SR, Black DM, Rubin SM. Lifetime risks of hip, Colles’, or vertebral fracture and coronary heart disease among white postmenopausal women. Arch Intern Med 1989; 149:2445–2448.
2. Chrischilles EA, Butler CD, Davis CS, Wallace RB. A model of lifetime osteoporosis impact. Arch Intern Med 1991; 151:2026–2032.
3. Maggi S, Kelsey JL, Litvak J, Heyse SP. Incidence of hip fractures in the elderly: a cross-national analysis. Osteoporos Int 1991; 1:232–241.
4. Johnell O, Gullberg B, Allander E, Kanis JA. The apparent incidence of hip fracture
in Europe: a study of national register sources. Osteoporos Int 1992; 2:298–302.
5. Michaëlsson K, Holmberg L, Mallmin H, Sörensen S, Wolk A, Bergström R, Ljunghall S. Diet and hip fracture
risk: results from a case-control study
. Int J Epidemiol 1995; 24:771–782.
6. Meyer HE, Tverdal A, Falch JA. Risk factors for hip fracture
in middle-aged Norwegian women and men. Am J Epidemiol 1993; 137:1203–1211.
7. Kreiger N, Kelsey JL, Holford TR, O’Connor T. An epidemiologic study of hip fracture
in postmenopausal women. Am J Epidemiol 1982; 116:141–148.
8. Paganini 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.
9. Johnell O, Gullberg B, Kanis JA, Allander E, Elffors L, Dequeker J, Dilsen G, Gennari C, Vaz AL, Lyritis G, Mazzuoli G, Miravet L, Passeri M, Cano RP, Rapado A, Ribot C. Risk factors for hip fracture
in European women: the MEDOS Study. J Bone Miner Res 1995; 10:1802–1815.
10. Faulkner KG, Cummings SR, Black D, Palermo L, Gluer CC, Genant HK. Simple measurement of femoral geometry predicts hip fracture
: the study of osteoporotic fractures. J Bone Miner Res 1993; 8:1211–1217.
11. Hemenway D, Feskanich D, Colditz GA. Body height
and hip fracture
: a cohort study of 90,000 women. Int J Epidemiol 1995; 24:783–786.
12. Ensrud KE, Cauley J, Lipschutz R, Cummings SR. Weight change
and fractures in older women. Arch Intern Med 1997; 157:857–863.
13. Ensrud KE, Lipschutz RC, Cauley JA, Seeley D, Nevitt MC, Scott J, Orwoll ES, Genant HK, Cummings SR. Body size
and hip fracture
risk in older women: a prospective study. Am J Med 1997; 103:274–280.
14. Meyer 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. Osteoporos Int 1995; 5:239–246.
15. Langlois JA, Harris T, Looker AC, Madans J. Weight change
between age 50 years and old age is associated with risk of hip fracture
in white women aged 67 years and older. Arch Intern Med 1996; 156:989–994.
16. Cumming RG, Klineberg RJ. Case-control study
of risk factors for hip fracture
in the elderly. Am J Epidemiol 1994; 139:493–503.
17. Cummings SR, Nevitt MC, Browner WS, Stone K, Fox KM, Ensrud KE, Cauley J, Black D, Vogt TM. Risk factors for hip fracture
in white women: study of Osteoporotic Fractures Research Group. N Engl J Med 1995; 332:767–773.
18. Grisso JA, Kelsey JL, Strom BL, O’Brien LA, Maislin G, LaPann K, Samelson L, Hofman S. Risk factors for hip fracture
in black women. N Engl J Med 1994; 330:1555–1559.
19. Ahlbom A. Biostatistics for Epidemiologists. Boca Raton, FL: Lewis Publishers, 1993; 111–118.
20. Nicolas V, Prewett A, Bettica P, Mohan S, Finkelman RD, Baylink DJ, Farley JR. Age related decreases in insulin-like growth factor-I and transforming growth factor B in femoral cortical bone from both men and women: implications for bone loss with aging. J Clin Endocrinol Metab 1994; 78:1011–1016.
21. Pfeilschrifter J, Schneidt-Nave C, Leidig-Bruckner G, Woitge HW, Blum WF, Wuster C, Haack D, Ziegler R. Relationship between circulating insulin-like growth factor components and sex hormones in a population-based sample of 50 to 80 year old men and women. J Clin Endocrinol Metab 1996; 81:2534–2540.
22. Koshla S. Idiopathic osteoporosis: is the osteoblast to blame? (Editorial) J Clin Endocrinol Metab 1997; 82:2792–2794.
23. Kurland ES, Rosen CJ, Cosman F, McMahon D, Chan F, Shane E, Lindsay R, Dempster D, Bilezikian JP. Insulin-like growth factor I in men with idiopathic osteoporosis. J Clin Endocrinol Metab 1997; 82:2799–2805.
24. Johansson AG, Eriksen EF, Lindh E, Langdahl B, Blum WF, Lindahl A, Ljunggren Ö, Ljunghall S. Reduced serum levels of the growth hormone-dependent insulin-like growth factor binding protein and a negative bone balance at the level of individual remodeling units in idiopathic osteoporosis in men. J Clin Endocrinol Metab 1997; 82:2795–2798.
25. Browner WS, Seeley DG, Vogt TM, Cummings SR. Non-trauma mortality in elderly women with low bone mineral density: study of Osteoporotic Fractures Research Group. Lancet 1991; 338:355–358.
26. Siiteri PK, Hammond GL, Nisker JA. Increased availability of serum estrogens in breast cancer: a new hypothesis. In: Pike MC, Siiteri PK, Welsch CW, eds. Hormones and Breast Cancer (Banbury Report No. 8). Cold Spring Harbor, NY: Cold Spring Harbor Laboratory, 1981; 87–101.
27. Davidson BJ, Ross RK, Paganini-Hill A, Hammond GD, Siiteri PK. Total and free estrogens and androgens in postmenopausal women with hip fractures. J Clin Endocrinol Metab 1982; 54:115–120.
28. Frost HM. Obesity, and bone strength and “mass”: a tutorial based on insights from a new paradigm. Bone 1997; 21:211–214.
29. Lauritzen JB, Petersen MM, Lund B. Effect of external hip protectors on hip fractures. Lancet 1993; 341:11–13.
30. Ekman A, Mallmin H, Michaëlsson K, Ljunghall S. External hip protectors to prevent osteoporotic hip fractures. Lancet 1997; 350:563–654.
31. Mautalen CA, Vega EM, Einhorn TA. Are the etiologies of cervical and trochnteric hip fractures different? Bone 1996; 18:133.
32. Kusowska-Wolk A, Karlsson P, Stolt M, Rössner S. The predictive validity of body mass index
on self-reported weight
. Int J Obes 1989; 13:441–453.
33. Must A, Willett WC, Dietz WH. Remote recall of childhood height
, and body build by elderly subjects. Am J Epidemiol 1993; 138:56–64.