The effectiveness of estrogen in preventing fractures in elderly women is also uncertain. Estrogen has been shown to be effective in preserving bone mass in elderly women [15, 16], but recent data [17] from Framingham showed little protective effect of an average of 10 years of estrogen therapy on bone density among women 75 years of age and older. In other prospective [9, 10] and case–control studies [3], the protective effects of estrogen on fracture were greater in younger women and weaker [3, 9] or nonexistent [10] in older women.

In the Study of Osteoporotic Fractures, our prospective study of 9704 women who were 65 years of age or older, we assessed estrogen use and bone mass at baseline and ascertained incident fractures every 4 months to examine the association between estrogen use and fracture in elderly women.

Methods

Top Methods Results Discussion Author & Article Info References

Participants

From September 1986 through October 1988, women who were at least 65 years of age were recruited for the Study of Osteoporotic Fractures in Portland, Oregon; Minneapolis, Minnesota; Baltimore County, Maryland; and the Monongahela Valley near Pittsburgh, Pennsylvania. Age-eligible women were recruited from population-based lists of women (voter registration, driver's license, and Health Maintenance Organizations membership lists) [18]. The response to these mass mailings varied from 8% in Pittsburgh (voter registration lists) to 19% in Portland (Kaiser Health Plan membership lists). We excluded black women because of their lower incidence of fractures, women who were unable to walk without the assistance of another person, and women who had had bilateral hip replacements.

Estrogen Use

Detailed information on use of estrogen and progestin was collected at the baseline interview. Information on use of estrogen was missing in 136 women; these women were excluded from all analyses. Participants were asked to bring all medications to the clinic for verification of use, preparation, and dosage. In addition, pictures of tablets were presented to participants to assist them in the recollection of previously prescribed hormone preparations. Information was collected about oral and parenteral estrogens (skin patches, injections, vaginal creams, and suppositories) and oral progestins. Our analyses were confined to oral preparations. Data were also collected on age at initiation of use of hormone preparations and on whether a participant had used such preparations for the entire time period since initiation and, if not, when she had stopped. Initiation of estrogen use with respect to menopause was determined by comparing the age at last menstrual period with the age at initiation of estrogen. To estimate duration of use, women were asked to check all ages, from 40 to 100 years of age, at which they had used estrogen. Duration of use was calculated by adding the total number of years that a woman had used estrogen. Because we were interested in examining the effect of initiation of estrogen and the effect of duration of use with respect to menopause, we excluded from the initiation and duration analyses 889 women for whom age at menopause could not be determined and 231 women who had started estrogen therapy more than 5 years before menopause.

Measurement of Bone Mass

Bone mineral density (g/cm²) was measured using single-photon absorptiometry (OsteoAnalyzer, Siemens-Osteon, Wahiawa, Hawaii). Details of these methods have been reported elsewhere [18]. We scanned three sites, including the distal radius, the proximal radius, and the calcaneus. The distal radius is composed of about 60% cortical and 40% trabecular bone, the proximal radius is about 99% cortical bone [19], and the calcaneus is about 97% trabecular bone [20].

Other Measurements

Reported health status, type of menopause, alcohol consumption, physical activity, and cigarette smoking were assessed by a questionnaire that was reviewed with the participant by a trained interviewer. Women were considered to have had surgical menopause if they reported bilateral oophorectomy at the age they stopped menstruating. The measure of alcohol consumption was drinks per week adjusted for atypical drinking, especially heavy drinking during the previous 30 days. Dietary calcium intake was assessed by a food frequency questionnaire and by interview using standardized food models to estimate portion sizes [21]. Total calcium intake included dietary and supplemental calcium. Women were asked if they walked for exercise and if they had fallen in the previous 12 months. A modified Paffenbarger questionnaire was used to assess sports and recreation for the previous year, expressed in kcal/wk [22]. History of osteoporosis was ascertained by asking women if a physician had ever told them whether or not they had osteoporosis or a spine fracture. Women were also asked whether they had ever taken the following medications: thiazide diuretics, thyroid hormones, sedatives, anxiolytics, and Tums or calcium supplements. Cognitive function was assessed using the Modified Mini-Mental Status examination [23]. During the clinic examination, body weight was measured (after removal of shoes and heavy outer clothing) using a balance beam scale. Height was measured (after removal of shoes) using a Harpenden stadiometer (Holtain Ltd., Dyved, United Kingdom). Height and weight were used to calculate the body mass index (kg/m²).

Ascertainment of Incident Fractures

Details of our method for identifying new fractures during follow-up have been published [24]. Briefly, we contacted participants every 4 months by postcard or telephone to ask whether they had sustained a fracture or fall. More than 99.5% of these follow-up contacts were completed. We interviewed participants about the way in which the fracture occurred. To confirm fractures, we obtained a copy of the radiographic report, which had to specifically mention the occurrence of an acute fracture. Hip fractures were also confirmed by radiologic review of copies of radiographs. We excluded fractures that occurred because of major trauma such as motor vehicle accidents. Most vertebral fractures do not come to medical attention; these fractures must be discovered by systematically obtaining radiographs from all participants and by comparing them with previous radiographs. Hence, self-reported vertebral fractures were not included.

Duration of fracture follow-up was calculated as the time to first occurrence of a fracture. Follow-up for fractures ranged from 0.02 years to 6.5 years. All nonspinal fractures that occurred before 1 April 1993 were included. For women who died during the follow-up period, date of death was used as the end of follow-up when fracture follow-up was not appropriate. The category "all nonspinal fractures" included hip and wrist fractures. Fractures of the distal radius or ulna were considered wrist fractures. Hip fractures were defined as those of the proximal femur.

Statistical Analysis

Estrogen use was classified as never, previous, or current. Chi-square tests of homogeneity and analyses of variance and covariance were used to compare baseline characteristics by estrogen use. Proportional hazard regression models were used to assess the relation between estrogen use and fracture. Women who had never used any type of estrogen formed the reference group for all analyses. Separate models were done for current and previous users. To test the hypothesis that the protective effect of estrogen use may be underestimated if a history of osteoporosis is not taken into account, we stratified by history of osteoporosis or spine fracture (or both). We also stratified patients by estrogen regimen (unopposed estrogen compared with estrogen and progestin) and by age ( 75 years or >75 years). Stratification by estrogen regimen was confined to those with wrist and all nonspinal fractures because combination therapy was not used frequently in our cohort and because few participants in this group had hip fractures.

The multivariate model included age, body mass index, total calcium intake (supplemental and dietary), physical activity (kcal/wk), surgical menopause (yes or no), history of smoking (yes or no), history of thyroid medication use, current use of thiazide diuretics, history of osteoporosis or spine fracture or both (yes or no), current use of sedatives or anxiolytics, alcohol consumption (drinks/wk), cognitive function (Mini-Mental Status examination 23), and falls in the previous year (yes or no). To test the hypothesis that bone mass accounts for the effect of estrogen on fracture risk, we subsequently added the bone mass of the distal radius to the wrist and nonspinal fracture models. Calcaneus bone mineral density was added to the hip fracture models because we have shown this measure to be superior to radial measurements in predicting hip fractures [25].

To determine if estrogen is protective in cigarette smokers, we formed three strata: 1) never-smokers and previous smokers who quit before menopause, 2) previous smokers who quit during or after the onset of menopause, and 3) current smokers. Within these strata, we compared current users of estrogen with never-users. For analyses of the effects of duration of estrogen use with respect to menopause, we formed two strata for duration of therapy: "short duration" (< 10 years) and "long duration" (10 years or more). For analyses of the effects of initiation of estrogen with respect to menopause, we formed two strata for initiation of therapy: "early" (2 years before to less than 5 years after menopause) and "late" (> 5 years after menopause). Because duration of use and initiation of use are confounded, we stratified current long-term users by when estrogen was initiated with respect to menopause, and we analyzed the risk for all nonspinal fractures. All analyses were done using SAS Software (SAS Institute, Inc., Cary, North Carolina).

Results

Top Methods Results Discussion Author & Article Info References

Overall, 13.7% of participants reported current use of oral estrogens and 27.4% of participants reported previous use (Table 1). Seventy-nine percent of women currently taking oral preparations were taking unopposed estrogen and 21% were taking estrogen plus progestin. Current users tended to have initiated use later than did previous users. The average duration of estrogen use (±SD) in current users was 12.7 ±9.7 years (median, 12 years) and 4.8 ±5.8 years (median, 2 years) in previous users. The mean age at which previous users had stopped taking estrogen was 55.5 ±8.3 years. We have previously shown that estrogen users tend to be younger, better educated, and less obese than nonusers [26]. Estrogen users had better cognitive function. Calcium intake, physical activity, alcohol consumption, and history of smoking were greater among current users. Current users were more likely to report a history of osteoporosis, surgical menopause, use of thyroid hormone, and current use of sedatives or anxiolytics. Self-reported health status and history of falls were similar among current users of estrogen, previous users, and never-users. Previous users were less likely to use thiazide diuretics than were never-users. During an average of 4.6 years, 1552 women had at least one nonspinal fracture, including 246 women with a hip fracture and 361 with a wrist fracture.

Current Use of Estrogen Compared with Previous Use

Compared with women who had never used estrogen, current users of estrogen had a decreased risk for wrist fracture and for any nonspinal fracture (Table 2). The relative risk for hip fracture was decreased but not statistically significant. No association was noted between previous use of estrogen and the risk for either hip fracture or for all nonspinal fractures. Previous users had a 20% decrease in the risk for wrist fracture, but confidence intervals included 1.0.

History of Osteoporosis

No evidence was found to show that the effect of estrogen on the risk for wrist fracture and all nonspinal fractures differed in women with a history of osteoporosis (Table 2). On the other hand, we found a statistically significant decrease in the relative risk for hip fracture among current users who had no history of osteoporosis, but we found no decrease in risk for women with osteoporosis.

Unopposed Estrogen Compared with Estrogen plus Progestin

Among current users, the effect of unopposed estrogen was similar to that of estrogen plus progestin for wrist fracture and all nonspinal fractures. The multivariate-adjusted relative risk for wrist fracture among current unopposed estrogen users was 0.42 (CI, 0.25 to 0.70), and this risk among current users of estrogen plus progestin was 0.31 (CI, 0.11 to 0.84). For all nonspinal fractures, the multivariate-adjusted relative risk was 0.69 (CI, 0.56 to 0.86) among users of unopposed estrogen and 0.51 (CI, 0.33 to 0.78) among users of estrogen plus progestin.

Age

The association between estrogen use and risk for all nonspinal fractures was similar in those younger and older than 75 years of age (Table 3). The effect of estrogen on wrist fracture was similar in younger and older women, but the confidence intervals were wider in the older age stratum because of the smaller number of women. We found an 80% decrease in the risk for hip fractures among women older than 75 years of age, and we found no effect on hip fracture in those 75 years age or younger. Once we had excluded women with a history of osteoporosis, the relative risk for hip fracture associated with current estrogen use was decreased in older and younger women, but confidence intervals included 1.0.

Estrogen and Bone Mass

If current use of estrogen protects women from fracture by preserving bone mass, then adjustment for bone mass should substantially decrease the association between estrogen use and the risk for fracture. Table 2 shows that among current users of estrogen, the multivariate-adjusted relative risk for hip fracture was 0.60 (CI, 0.36 to 1.02). Inclusion of calcaneal bone mineral density in the multivariate model decreased the effect of estrogen use by about 20% (relative risk, 0.73; CI, 0.43 to 1.24). Similarly, the multivariate relative risk for wrist fracture adjusted for bone mineral density was 0.50 (CI, 0.31 to 0.83) compared with the multivariate relative risk unadjusted for bone mass (relative risk, 0.39; CI, 0.24 to 0.64). The relative risk for all nonspinal fractures using the multivariate model adjusted for bone mass was 0.80 (CI, 0.65 to 0.98); the risk was also attenuated by about 20% when compared with the multivariate relative risk unadjusted for bone mass (relative risk, 0.66; CI, 0.54 to 0.80). Nevertheless, the effect of estrogen, even after adjusting for bone mineral density, remained statistically significant for wrist fracture and all nonspinal fractures.

Smoking

Current smoking did not attenuate the reduction in the relative risk for fractures associated with estrogen use (Table 4). The relative risks for all nonspinal, wrist, and hip fractures were similar among women who had never smoked and among those who currently smoked. Confidence intervals were wider for current smokers because of the smaller number of women in this group. Previous smokers who were currently using estrogen were at a decreased risk for hip fracture compared with previous smokers who did not use estrogen.

Duration and Initiation of Estrogen Use with Respect to Menopause

The relative risks for all nonspinal fractures were similar among current short- and long-term users of estrogen (P = 0.66). In short- and long-term current users, estrogen was associated with a 30% to 40% reduction in the risk for all nonspinal fractures (Table 5). On the other hand, current long-term use was associated with a 75% reduction in the risk for wrist fractures whereas short-term use was associated with a 25% reduction (P = 0.03). We were unable to formally test for statistical differences in the relative risks for hip fracture between short- and long-term current users because few participants in these groups had hip fractures. Nevertheless, our results suggest that current long-term use was associated with a reduction in all hip fractures (relative risk, 0.27; CI, 0.08 to 0.85). Long duration of use among previous users was not associated with a reduction in the risk for hip, wrist, or all nonspinal fractures.

Current users who initiated estrogen replacement therapy early in menopause had a significantly reduced risk for wrist and all nonspinal fractures (Table 6). The magnitude of the effect of estrogen replacement therapy was diminished if therapy was initiated more than 5 years after menopause. For example, the risk for wrist fracture was reduced by 71% among women who started estrogen earlier compared with 23% among those who started later, but these relative risks were not significantly different from each other (P > 0.05). Women who had used estrogen in the past within 5 years of menopause had a somewhat lower risk for wrist and all nonspinal fractures than did women who had never used estrogen, but the confidence intervals included 1.0.

Early initiation among long-term users was associated with a 50% reduction in the risk for all nonspinal fractures, but no effect was noted if estrogen was initiated late, even among long-term users. The age-adjusted risk for all nonspinal fractures among current long-term users of estrogen who initiated estrogen within 5 years of menopause was 0.53 (CI, 0.37 to 0.74), and this risk was 0.97 (CI, 0.56 to 1.68) in current long-term users who initiated estrogen more than 5 years after menopause.

Discussion

Top Methods Results Discussion Author & Article Info References

Our results suggest that current use of estrogen reduces the risk for hip, wrist, and all nonspinal fractures but that estrogen is more effective if initiated within 5 years of menopause and if used for longer than 10 years. Our study confirms the findings of previous studies that showed that current use of estrogen was associated with a significant reduction in the risk for hip and wrist fractures, and our study extends these findings to include the overall risk for all nonspinal fractures. Because most fractures are related to low bone mass [27], this finding is not surprising. The observation that estrogen prevents all fractures broadens the public health effect of this therapy as a preventive intervention and needs to be considered when evaluating the risks and benefits of estrogen for older women.

Our study supports the finding of previous case–control [6, 7] and prospective studies [9] that initiation of estrogen use soon after menopause is associated with the lowest risk for fracture. Because initiation and duration of use are confounded, we examined long-term users of estrogen separately by also examining when they initiated therapy relative to menopause. Of importance, long-term users who had initiated therapy after 5 years of menopause had no significant reduction in the risk for all nonspinal fractures, despite an average duration of use of 16 years. This suggests that the time at which estrogen is initiated with respect to menopause may be more important than the total duration of use.

We found similar effects of estrogen on the reduction of all nonspinal fractures and wrist fractures in younger and older women. Current estrogen use was, however, more effective in reducing hip fractures among the oldest women. These results are consistent with studies of estrogen and bone loss in elderly women [15, 16], suggesting that estrogen can be used to prevent hip fractures, even among the very old. On the other hand, the Framingham Study recently suggested that estrogen use for at least 7 years does not preserve bone mass in women older than 75 years of age, but it did not assess the effectiveness of estrogen on risk for fracture within these age strata [17]. Kanis and colleagues [3] reported a lower relative risk for hip fracture among estrogen users younger than 80 years of age compared with users 80 years or older, but even among the latter, estrogen decreased hip fracture by 30% (P > 0.05). Finally, Naessen and colleagues [10] found little effect of estrogen among women older than 60 years of age, although the older women tended to take less potent estrogens.

We found no difference in the effect of unopposed estrogen and combination therapy with estrogen plus progestin on fracture risk. These results support findings from a prospective study from Sweden [10], which found that combination therapy is effective in reducing the risk for hip fracture. Our findings are also consistent with the observation that combination therapy and monotherapy (unopposed estrogens) have similar effects on bone mass [28]. Progestins given alone have also been shown to reduce bone loss, particularly loss of cortical bone, but there is little evidence that the combination therapy has additive or synergistic effects on bone mass [29].

Case-control studies [1, 2, 4, 6] have suggested that at least 5 years of use may be needed to reduce fractures. Recent data from Framingham [17] suggest that at least 7 years of use may be needed for the preservation of bone density, at least in women younger than 75 years of age. We found little effect of duration of use among current users on the occurrence of all nonspinal fractures, but more than 10 years of use was associated with a substantial reduction in the risk for wrist and hip fractures.

Inclusion of bone mass in the models decreased the effect of current estrogen use by 20% to 28%. These results suggest that estrogens prevent fracture, in part, by preserving bone mass. However, even after adjusting for bone mass, we found that estrogen users had a significantly reduced risk for wrist and all nonspinal fractures, suggesting that estrogen prevents fractures by other additional mechanisms. Our baseline measurements of bone density included only appendicular sites; axial bone mass was measured at a later visit. We have shown that measurement of bone mass at the hip better predicts which women are more likely to fracture their hips than does measurement of bone mass at other sites [25]. It is possible that inclusion of hip bone mass in our multivariate model may have led to greater attenuation of the effect of estrogen on risk for hip fracture. Nevertheless, estrogens could have beneficial effects on bone quality, for example, by preventing complete perforation of trabeculae. Moreover, estrogen could influence neuromuscular function and muscle strength, thereby reducing the risk for falls. Previous studies of younger women [30, 31] have shown that estrogen users have greater muscle strength than do nonusers. In elderly women, however, we found no association between estrogen use and improved neuromuscular function [32].

We did not confirm previous findings that current smoking eliminates the protective effect of estrogens on the risk for fracture [33]. The relative risk for hip, wrist, and all nonspinal fractures associated with current estrogen use was similar in nonsmokers and current smokers. Smoking has antiestrogen effects; smokers tend to have menopause approximately 2 years earlier than do nonsmokers, and smokers have a lower risk for endometrial cancer [34]. Michnovicz and colleagues [35] suggested that smoking may influence the metabolism of estrogen into less potent estrogen metabolites. However, we found no difference in serum estrone levels between smoking and nonsmoking women (mean age, 58 years [36]). In addition, if smoking negates the protective skeletal effects of estrogen, then smoking may be expected to have similar biological effects on other disease end points. In the Lipid Research Clinics follow-up study [37], estrogen was effective in reducing the risk for cardiovascular disease among smokers and nonsmokers. Estrogen was also shown to be effective in decreasing the risk for cerebrovascular disease among smokers [38]. This suggests that smoking does not negate the positive effect of estrogen on cardiovascular disease. Our results suggest that the same is true for fractures.

Our study confirms that previous estrogen use does not offer much protection against all nonspinal fractures or, specifically, against hip fractures [1, 9], even in women who had previously used estrogen for more than 10 years or who had initiated estrogen replacement therapy early in menopause. Previous users had a somewhat lower risk for wrist fractures. These results imply that women need to continue estrogen therapy indefinitely to reduce the risk for fracture.

We found that current use of estrogen was associated with similar reductions in the risk for wrist and all nonspinal fractures among women with and without a history of osteoporosis. Current use of estrogen was not associated with a statistically significant reduction in the risk for hip fracture in women reporting a history of osteoporosis, although the reduction in risk in those without osteoporosis was statistically significant. This may be related to the observation that the relation between history of osteoporosis and fracture is greater for hip fracture than for wrist and all fractures. Women with osteoporosis are more likely to take estrogen [26]. Use of estrogen may be a marker of more severe osteoporosis, and this could lead to an underestimation of the effect of estrogen in that group. For example, women without a history of osteoporosis had a reduced risk for hip fracture even if they initiated estrogen replacement therapy late with respect to menopause or if they had used estrogen for less than 10 years, although none of these observations was statistically significant. Other studies have found estrogen to be effective in slowing bone loss [39-41] and in preventing spine fractures in women with established osteoporosis [42].

Although our prospective design avoids several pitfalls of case–control studies, it has several limitations. Women in our cohort were primarily white, ambulatory volunteers, who may be at lower risk for hip fracture. We excluded black women, but our results can be generalized to the large population of community-dwelling elderly white women. The prevalence of estrogen use was low in this population, and we had limited power to examine its association with hip fractures, especially in stratified analyses. Our information on estrogen use, especially duration of use and initiation of use with respect to menopause, was based on self-report. However, we confirmed current use by interview and review of medication bottles brought to the clinic, and inaccurate recall of duration of use and initiation of use would bias our results toward finding no association. Finally, selection bias for estrogen use is well established [26, 43]. Hormone therapy may be a marker of socioeconomic, clinical, or lifestyle variables that place users at lower risk for fracture. We controlled for many factors that differ markedly between women who used estrogen compared with women who never used estrogen, but other factors that we did not measure may have confounded our results. Only a randomized trial comparing hormone use with placebo can provide conclusive evidence that treatment causes a reduction in fractures. Such a trial, the Women's Health Initiative, is in progress.

Current use of estrogen replacement therapy appears to reduce the risk for fracture in older women. Our results suggest that estrogen should be initiated early in menopause and continued indefinitely. Although the overall benefit of estrogen replacement therapy may outweigh the risks for most women, we have few data on the effects of such long-term use, which could amount to 30 years of use if the average age of menopause is 50 years and if life expectancy exceeds 80 years. Further studies are needed to evaluate the overall risks and benefits of the long-term use of estrogen.

Appendix

Investigators in the Study of Osteoporotic Fractures Research Group were as follows: University of California, San Francisco (Coordinating Center): S.R. Cummings (principal investigator), M.C. Nevitt (project director), D. Black (study statistician), H.K. Genant (director, central radiology laboratory), C. Arnaud, W. Browner, L. Christianson, M. Dockrell, C. Fox, C. Gluer, S. Harvey, S.B. Hulley, M. Jergas, L. Palermo, A. Pressman, P. San Valentin, D. Seeley, and K. Stone; University of Maryland: R. Sherwin (principal investigator), J. Scott (co-investigator), K. Fox (co-investigator), J. Lewis (project coordinator), G. Greenberg (clinic coordinator), M. Bahr, S. Trusty, L. Finazzo, S. Snyder, E. Oliner, B. Hohman, and T. Page; University of Minnesota: K. Ensrud (principal investigator), R. Grimm Jr. (co-investigator), C. Bell (project director), E. Mitson (study coordinator), I. Chavier, K. Jacobson, S. Fillhouer, C. Shoberg, D. Michel, S. Estill, J. Hansen, and J. Baumhover; University of Pittsburgh: J.A. Cauley (principal investigator), L.H. Kuller (co-principal investigator), L. Harper (project director), M. Nasim (clinic coordinator), C. Bashada, L. Buck, A. Githens, A. McCune, D. Medve, S. Rudovsky, and N. Watson; The Kaiser Permanente Center for Health Research, Portland Oregon: T.M. Vogt (principal investigator), W.M. Vollmer, E. Orwoll (co-investigators), J. Blank (project director), F. Heinith (clinic coordinator), R. Bright, J. Downing, B. Packer, C. Souvanlausky, L. Puderbauth, and D. Franco.