|
|
| |||||||||||||||||||||||||||||||
Background Although more than 1 percent of black women 80 years of age or older have hip fractures each year, little is known about risk factors for hip fracture in these women.
Methods We carried out a case-control study involving 144 black women admitted with a first hip fracture to 1 of 30 hospitals in New York and Philadelphia. The controls were 218 black women living in the community who were matched to the case patients according to age and ZIP Code or telephone exchange and 181 hospitalized black women matched according to age and hospital. Information was obtained through personal interviews and was studied by multivariable logistic-regression analysis.
Results When the case patients were compared with the control subjects from the community, the women in the lowest quintile for body-mass index had a markedly increased risk of hip fracture as compared with the women in the highest quintile (odds ratio, 13.5; 95 percent confidence interval, 4.2 to 43.3). Postmenopausal estrogen therapy for one year or more was protective for women under 75 years of age (odds ratio, 0.1; 95 percent confidence interval, <0.1 to 0.5). Factors associated with an increased risk of hip fracture included a history of stroke (odds ratio, 3.1; 95 percent confidence interval, 1.2 to 8.1), use of aids in walking (odds ratio, 5.6; 95 percent confidence interval, 2.7 to 11.5), and consumption of seven or more alcoholic drinks per week (odds ratio, 4.6; 95 percent confidence interval, 1.5 to 14.1). The results were similar when the case patients were compared with the hospitalized control subjects.
Conclusions Among black women thinness, previous stroke, use of aids in walking, and alcohol consumption are associated with an increased risk of hip fracture. Postmenopausal estrogen therapy protects against hip fracture in women under 75 years of age.
Methods
Study Design and Subjects
The case patients were black women 45 years of age or older with a radiologically confirmed diagnosis of a first hip fracture who were admitted to 1 of 30 participating hospitals in New York City or Philadelphia between September 1987 and August 1992. The results of a parallel study of white women have been published previously9,10,11. The study protocol was approved by the institutional review committee at each hospital, and all the study subjects gave written informed consent.
Of 315 women eligible as case patients, 221 (70 percent) participated in the study, 67 (21 percent) declined to participate, 7 (2 percent) had physicians who declined on their behalf, and 20 (6 percent) could not be followed after discharge from the hospital or were patients for whom a proxy respondent was needed but not available. A median of 11 days elapsed between admission to the hospital and the time of the study interview.
The results in the case patients were compared with the results in two control groups, a group of black women living in the community (community controls) and a group of black women who were hospitalized (hospital controls).
The community control subjects were recruited from randomly selected lists of Medicare recipients generated by the Health Care Financing Administration12 (for matching with case patients 65 years of age or older) and by telephone through random-digit dialing (for matching with case patients less than 65 years of age). The community controls were matched to the case patients according to 10-year age group and either ZIP Code or telephone exchange. No community controls were recruited for five case patients who were less than 65 years of age and did not live in a household with a telephone. In the comparison of case patients with community controls, these five case patients were excluded. Of 431 eligible community controls, 236 (55 percent) were interviewed, 157 (36 percent) declined to participate, 27 (6 percent) could not be contacted, and 11 (3 percent) required a proxy respondent but none was available.
The hospital control subjects were matched to the case patients according to 10-year age group and hospital. Of 354 eligible hospital controls, 218 (62 percent) were interviewed, 63 (18 percent) declined to participate, 49 (14 percent) had physicians who declined on their behalf, 14 (4 percent) could not be followed after discharge from the hospital, and 10 (3 percent) required a proxy respondent but none was available.
Proxy respondents were interviewed when the women were medically incapacitated or cognitively impaired. Cognitive impairment was considered to be present when a woman scored four or more errors on a modified Kahn-Goldfarb Mini-Mental Status Examination13. Proxy respondents were interviewed on behalf of 77 case patients, 37 hospital controls, and 18 community controls. The remaining 144, 181, and 218 women in the respective groups formed the basis of this analysis.
Data Collection
All the interviews were conducted in person, except for 16 proxy interviews that were conducted by telephone.
Alcohol and Smoking
Alcohol consumption was assessed by asking the women about their consumption during the past year and the usual frequency of consumption during their adult years. The number of alcoholic drinks consumed per week was classified as follows: grade 1, one or less; grade 2, two to six; and grade 3, seven or more14. A smoker was defined as a person who smoked at least one cigarette per day for six months or more. The average number of cigarettes smoked per day and the duration of smoking were also determined.
Medications
Information was obtained on the frequency and duration of any estrogen, progestin, or thiazide diuretic therapy and the age at which the therapy began. The women were also asked about any treatment with anxiolytic, hypnotic, antidepressant, or antipsychotic drugs during the past month. Psychotropic medications were classified as long-acting if they had half-lives of 24 hours or more, as specified in the American Medical Association's Drug Evaluations, fifth edition15.
Adjusted Weight
The women's estimates of their current height (in meters) and weight (in kilograms) were used to calculate the body-mass index,16 which is the weight in kilograms divided by the square of the height in meters. The women were also asked their weight at 25 years of age.
Lower-Limb Function before the Fracture
The women were asked whether they needed assistance in performing four activities related to lower-limb function: walking across a small room, getting out of a chair, walking outside on level ground, and walking up or down stairs. Their scores were calculated as the number of tasks for which assistance was needed. Women totally confined to a bed or chair were given a score of 5. The degree of physical impairment was determined separately by assessing the use of ambulatory aids, such as a cane or walker.
Visual Impairment
The women were asked whether they had glaucoma or cataracts. Those who reported that they could not see well enough (with their eyeglasses or contact lenses) to recognize a friend across the room were classified as visually impaired.
Neurologic Disease
The women were asked whether they had a history of Parkinson's disease, stroke, or epilepsy.
Proxy Questionnaire
Because of concern about potentially inaccurate reporting, the questionnaire given to the proxy respondents excluded items on the duration of smoking, the duration of therapy with thiazide diuretics or psychotropic drugs, and postmenopausal estrogen therapy. In addition, no information was obtained from the proxy respondents about the women's past use of alcohol or their current or former height and weight.
Statistical Analysis
Odds ratios and 95 percent confidence intervals were used to estimate relative risks, with conditional logistic-regression analysis because of the large number of strata that resulted from matching17,18. The initial analyses controlled for the effects of the study design (age group and ZIP Code, telephone exchange, or hospital) and also included age as a continuous variable. The final multivariable models included all statistically significant variables, as well as those whose removal substantially changed the estimates of the effect of other factors.
The relation between body-mass index and the risk of hip fracture was assessed by fitting polynomials of increasing order. On the basis of the proportional increase in value of the maximum-likelihood function, a model containing both linear and quadratic terms best described the relation. Thus, multivariable models include terms for both body-mass index and body-mass index squared.
We assessed whether there was a difference in the odds ratio between the subjects interviewed directly and those for whom a proxy was required, by obtaining separate estimates and comparing them. Because the magnitude and direction of the odds ratios did not differ substantially and no interaction with proxy status was detected, only the results from the direct interviews are presented here, since they contained more complete information.
Analyses were conducted in which the case patients were compared with both the community and the hospital controls. Detailed results are presented only for the comparison with the community controls, because the results of the two analyses were similar, and we thought that for most variables the community controls provided a more appropriate comparison group than the hospital controls.
Results
The characteristics of the case patients and control subjects who were interviewed directly are shown in Table 1. Over half the case patients were 75 years old or older. There were no significant differences between the case patients and the control subjects with regard to age, years of education, or marital status.
|
Table 2 shows the odds ratios for the variables identified in other studies19,20,21,22,23,24,25 as possible risk factors or protective factors for hip fracture in white women. Women in the lowest quintile for body-mass index ( 
22.6) had a greatly increased risk of hip fracture as compared with women in the heaviest quintile (body-mass index, 
31.6) (odds ratio, 5.6; 95 percent confidence interval, 2.3 to 13.8). No association was found between the risk of hip fracture and body-mass index at 25 years of age as recalled by the women. Sixty-seven percent of the control subjects reported that their current weight was more than 4.5 kg greater than their weight at the age of 25, as compared with 40 percent of the case patients (odds ratio, 0.5; 95 percent confidence interval, 0.3 to 0.9). In contrast, 34 percent of the case patients reported a net loss of 4.5 kg or more as compared with only 13 percent of the control subjects (odds ratio, 2.6; 95 percent confidence interval, 1.2 to 6.0). The effect of weight loss was independent of the number of chronic illnesses.
|
Current consumption of seven or more drinks per week containing alcohol was associated with an increased risk of hip fracture as compared with consumption of one drink or less per week. Women who reported decreasing their alcohol consumption from a former level of seven or more drinks per week did not have an increased risk of hip fracture as compared with those who reported always consuming little or no alcohol.
Thiazide diuretic therapy was commonly used by these women. This therapy had no significant protective effect after adjustment for body-mass index (odds ratio, 0.8; 95 percent confidence interval, 0.5 to 1.6). Regardless of their status as case patients or control subjects, heavy women (body-mass index, 31.6) were more likely to have taken a thiazide diuretic (34 percent) than thin women (body-mass index, 22.6) (16 percent; P = 0.01). Factors not associated with hip fracture included cigarette-smoking status, parity, and age at menopause.
Several variables were assessed that have been identified as risk factors for falls. An increased risk of hip fracture was associated with a previous stroke, the use of ambulatory aids, lower-limb dysfunction, and a diagnosis of epilepsy (Table 2). There was a possibly increased risk associated with therapy with long-acting psychotropic drugs (odds ratio, 2.8; 95 percent confidence interval, 0.8 to 9.4), but not with short-acting ones. We compared the characteristics of falls in the case patients with those in a subgroup of 29 control subjects who reported a fall in the past six months that did not result in a fracture. The control subjects who had fallen recently were more likely to have landed on a soft surface (39 percent) than the case patients (22 percent; P = 0.02), although there was no difference in the height of the falls.
We constructed a multivariable conditional logistic-regression model, retaining factors that were independent predictors of hip fracture and factors whose removal substantially altered the effect of one or more of the other factors (Table 3). The independent predictors included body-mass index, estrogen therapy, use of ambulatory aids, alcohol consumption, and history of stroke.
|
We compared the case patients and the hospital controls with respect to the variables shown in Table 3. The results were similar, although the strength of the associations varied slightly from those in the comparison with the community controls. Body-mass index was defined in quintiles based on the distribution in the hospital controls. Women in the lowest quintile ( 22.6) had an increased risk of hip fracture as compared with women in the heaviest quintile ( 31.6) (odds radio, 13.9; 95 percent confidence interval, 4.4 to 44.1); a history of stroke (odds ratio, 2.2; 95 percent confidence interval, 0.8 to 5.6), use of ambulatory aids (odds ratio, 2.4; 95 percent confidence interval, 1.2 to 4.9), and current consumption of seven or more alcoholic drinks per week (odds ratio, 4.0; 95 percent confidence interval, 1.0 to 15.8) were all associated with an increased risk of hip fracture. There was also a trend toward a protective effect of estrogen therapy lasting one year or more (odds ratio, 0.3; 95 percent confidence interval, 0.1 to 1.3).
Discussion
A number of important risk factors for hip fracture emerged in this study of black women. Thinner women were at substantially higher risk for hip fracture than their heavier counterparts, a finding consistent with those of studies of white women19,20,21 and a previous study of black women22. The protective effect of increased body mass appears to result primarily from weight gained since the early adult years, not from having been heavy in early adulthood. The protection due to increased body mass has been postulated to be a result of increased adipose-based production of estrogen, greater gravitational forces on bone mass, and increased padding around the hips that may decrease the transmission of energy from the impact of the fall to the proximal femur5,19,23. Although 95 percent of hip fractures in the women in these studies occurred as a result of a fall, no information was collected about the site of impact or the distribution of body fat.
Estrogen therapy had a protective effect on women under the age of 75 years, as has also been reported for white women24,25,26,27. Few women reported taking estrogens (8 percent of the control subjects), although there was a rather high rate of hysterectomy (33 percent). Although the power of the study to evaluate estrogen therapy in detail was limited, our results are consistent with the hypothesis that there is a protective effect of recent use and of prolonged therapy27.
We found no protective effect of thiazide diuretic drugs, which many women reported taking. The latter were not found to protect against hip fracture. Whether thiazides offer such protection is a particularly important question for black women, given the high prevalence of hypertension in this group28. Among white women, some studies have reported a protective effect of thiazides, but most did not control for body mass29,30,31.
We found an increased risk of hip fracture in women who reported consuming seven or more alcoholic drinks per week as compared with those who reported minimal or no consumption of alcohol. This association has not been reported consistently in white women19,20. The association between alcohol consumption and the risk of hip fracture was independent of body weight, the number of chronic illnesses, and cigarette smoking and was limited to current consumption. These results are consistent with those reported by the Framingham Study, in which heavy current, but not former, consumption of alcohol was associated with an increased risk of hip fracture14. This may indicate that alcohol consumption increases the risk of hip fracture by increasing the risk of falls.
As we found in our earlier study of white women,5 risk factors for falls (e.g., previous stroke and the use of ambulatory aids) were important independent predictors of hip fracture in black women.
Some limitations of the present study should be mentioned. As compared with studies of younger persons, it may seem to have a low response rate (70 percent of case patients, 62 percent of hospital controls, and 55 percent of community controls). However, very old women (median age, 76 years) participated, and lower response rates are common in studies of persons in this age group32,33. Because some women with severe cognitive impairment were excluded and certain factors could not be assessed accurately in proxy interviews, the results of this study cannot be generally applied to women with severe cognitive impairment. It was not possible to validate the information reported by the subjects. Finally, several potential risk factors were not assessed, such as exposure to sunlight, vitamin D intake, and thyroid hormone therapy.
In conclusion, a major determinant of the risk of hip fracture in black women is body mass. Given the magnitude of the association and the prevalence of obesity in black women (60 percent),34 increased body mass probably explains some of the racial differences in rates of hip fracture. Estrogen therapy may be important for the prevention of such fractures in black women. Finally, the risk factors for falls also appear to be important risk factors for hip fracture. Programs to prevent falls in black women are likely to result in a reduced incidence of hip fracture.
Supported by a grant (R01-AR35409) from the National Institutes of Health.
We are indebted to the late Ms. Christy Golden for her support of this project.
Source Information
From the Center for Clinical Epidemiology and Biostatistics, Division of General Internal Medicine, School of Medicine, University of Pennsylvania, Philadelphia (J.A.G., B.L.S., L.A.O., G.M., K.L.); the Department of Health Research and Policy, School of Medicine, Stanford University, Stanford, Calif. (J.L.K.); and the Division of Epidemiology, Columbia University, New York (L.S., S.H.). The Northeast Hip Fracture Study Group includes, in Philadelphia, Dr. James Anthony and Dr. Alan Hibberd, Mercy Catholic Medical Center; Dr. Michael Clancy, Temple University Health Sciences Center; Dr. Jerome Cotler, Jefferson University Hospital; Dr. William DeLong, Cooper Hospital/ University Medical Center; Dr. Malcolm Ecker, Chestnut Hill Hospital; Dr. Z.B. Friedenberg, Presbyterian-University of Pennsylvania Medical Center; Dr. Robert Good, Bryn Mawr Hospital; Dr. Charles Hummer, Crozer-Chester Medical Center and Sacred Heart Hospital; Dr. David Junkin, Abington Hospital; Dr. William Markman, Jeanes Hospital; Dr. Pekka Mooar, Medical College of Pennsylvania; Dr. Herbert Stein, Frankford Hospital; Dr. Milton Wohl, Albert Einstein Medical Center; Dr. Brendan Wynne, Philadelphia College of Osteopathic Medicine; and Dr. Dennis Zaslow, St. Joseph's Hospital; and in New York, Dr. David Andrews, Allen Pavilion; Dr. Harold Dick, Columbia-Presbyterian Medical Center; Dr. Edward Habermann, Montefiore Medical Center; Dr. Harvey Insler, Bronx-Lebanon Hospital Center; Dr. Placido Menezes, Interfaith Medical Center; Dr. Ronald Rosenthal, Long Island Jewish Medical Center; Dr. Arthur Sadler, Jacobi Hospital; Dr. Marvin Shelton, Harlem Hospital; Dr. Joel Teicher, Brookdale Hospital Medical Center; and Dr. Robert Zickel, St. Luke's-Roosevelt Hospital Center.
Address reprint requests to Dr. Grisso at the Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania, 317R Nursing Education Bldg., 420 Service Dr., Philadelphia, PA 19104-6095.
References
| |||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
