Loop Diuretic Use and Fracture in Postmenopausal Women:  Findings From the Women's Health Initiative FREE

Osteoporosis and heart failure are particular health concerns for women.^1- 3 Half of all postmenopausal women will experience an osteoporosis-related fracture in their lifetime.^1- 2 By the age of 40 years, the lifetime risk for women of developing heart failure is 1 in 5.³

In 2007, in the United States, loop diuretics were among the top 200 prescribed medications, with 37 094 000 prescriptions for furosemide, 1 907 000 prescriptions for torsemide, and 1 741 000 prescriptions for bumetanide.⁴ Loop diuretics increase calcium excretion⁵ and, in some studies, hypercalciuria is associated with low bone mineral density (BMD),^6- 7 a risk factor for fractures.⁸ Loop diuretics may also cause orthostatic hypotension,⁹ which has been associated with falls in some,¹⁰ albeit not all,¹¹ studies. However, data are conflicted with regard to the association between loop diuretics and BMD, falls, and fractures. In a cross-sectional study of 140 postmenopausal women, no association was found between loop diuretic use and BMD¹²; in contrast, in a randomized clinical trial that included 84 postmenopausal women and was performed by the same investigators, the loop diuretic bumetanide decreased BMD.¹³ Similarly, findings are inconsistent with respect to the association of loop diuretics with falls—1 meta-analysis suggested that loop diuretics are weakly associated with falls,¹⁴ although cross-sectionally, in the British Women's Heart and Health Study, cardiovascular drugs were not independently associated with falls.¹⁵ Relative to fractures, a positive association of loop diuretic use with osteoporotic fractures was reported in a prospective cohort study that included more than 300 postmenopausal women¹⁶ and in a large population-based pharmacoepidemiologic case-control study of more than 44 000 users of loop diuretics.¹⁷ However, in another case-control study published in 1986, a small inverse association of loop diuretic use with femoral neck fractures was noted, although the confidence interval (CI) included 1.¹⁸

To our knowledge, no study to date has examined the relationship between loop diuretic use and BMD, falls, and fractures within the context of a large prospective study that uniformly collected medication exposure, BMD, fall, and fracture outcomes. In addition, the relationship between loop diuretics and fractures in women with congestive heart failure (CHF) has not been studied. The purpose of this article is to examine the association between loop diuretic use and changes in BMD, falls, and fractures in patients enrolled in the Women's Health Initiative (WHI), including women with and without CHF.

We hypothesized that, compared with loop diuretic nonusers, loop diuretic users would have greater loss of BMD, more falls, and more fractures and that these results would be a function of duration of use of this type of drug. Moreover, in experimental studies, aldosteronism, which accompanies heart failure, is associated with impairments in bone strength, which are accentuated by administration of a loop diuretic,¹⁹ and hip fractures are increased with CHF.²⁰ For this reason, we predicted that use of loop diuretics would be a particular risk for loss of BMD, falls, and fractures in the setting of CHF.

We conducted a prospective study of loop diuretic use and subsequent changes in BMD (mean follow-up of 3 years), falls (mean follow-up of 7.7 years), and fractures (mean follow-up of 7.7 years) in women aged 50 to 79 years enrolled at 40 clinical centers in the WHI from October 29, 1993 to December 31, 1998 and who were subsequently observed for an average of 7.7 years. The study population for the fracture and fall outcomes (3411 loop diuretic users and 130 444 nonusers) included all women in the WHI observational study and the WHI clinical trials who were not in the active arms of the hormone or calcium and vitamin D trials. The BMD was available in a subset (300 loop diuretic users and 9124 nonusers). Details of the WHI methods have been described elsewhere.²¹ All protocols were reviewed and approved by the institutional review board at each participating center.

There were 3820 women with CHF in these analyses, including 993 women with self-reported prevalent CHF (388 loop diuretic users and 605 nonusers) and 2827 with physician-adjudicated incident CHF (454 loop diuretic users and 2373 nonusers).

Current medication use was ascertained by having the participants bring all the containers for medications taken for the 2 weeks prior to the baseline visit and the year 3 visit. Interviewers entered each medication into the database, which assigned drug codes using Medi-Span software (Wolters Kluwer Health; Conshohocken, Pennsylvania). Information on duration of use but not dose was recorded. For this study, we defined hormone therapy as use of an estrogen with or without progesterone (oral or patch formulations) and current loop diuretic use as any use at the baseline and/or at the time of the year 3 visit. Duration of use of loop diuretics was examined in 3 categories (<1 year, 1-3 years, or >3 years).

Dietary intake of calcium and vitamin D was measured with a semiquantitative food frequency questionnaire. Vitamin D intake was defined as the sum of vitamin D from supplements and diet, whereas calcium intake included all calcium from medications, supplements, and diet.

Questionnaires were used to collect information on age, ethnicity, smoking, parental history of hip fractures, prevalent fractures on or after the age of 55 years, age at menopause, health status, history of CHF, and history of coronary heart disease (myocardial infarction, angina, and/or coronary artery bypass graft or percutaneous coronary angioplasty). Alcohol consumption was estimated from the food frequency questionnaire. A construct of chronic conditions, including treated diabetes mellitus, stroke, any cancer, cardiovascular disease, arthritis, hypertension, 2 or more falls, emphysema, and hip fractures after the age of 55 years, was calculated in the WHI enrollees and included as the number of chronic conditions (n=0-4). Energy expenditure from recreational physical activity was used to determine physical activity levels. The Rand 36-Item Short Form Health Survey was used to calculate a physical function construct, with higher scores indicating better physical function. Height and weight were measured in the WHI enrollees²¹ and used to calculate body mass index (BMI) as weight in kilograms divided by height in meters squared. Geographic study sites and trial enrollment were adjusted for in the statistical analyses. All covariates used were from the baseline visit.

The BMD of the total hip, anterior-posterior lumbar spine, and total body was measured at baseline and at year 3 in participants at 3 of the 40 clinical centers of the WHI (Pittsburgh, Pennsylvania; Birmingham, Alabama; and Phoenix/Tucson, Arizona) by using a Hologic QDR densitometer Model 2000, 2000+, or 4500 Fan-beam (Hologic Inc, Waltham, Massachusetts). We determined change in BMD by loop diuretic use at these sites from baseline to year 3.

In the WHI clinical trials, total fractures were defined as all reported clinical fractures other than those of the ribs, sternum, skull or face, fingers, toes, and cervical vertebrae. Hip fractures were verified by review of radiologic reports or medical records by centrally trained and masked physician adjudicators at each clinical center. All fractures, including hip fractures, were adjudicated; however, all other fractures were ascertained by self-report.

A participant was identified as having a history of falls if she reported 2 or more falls in the year before baseline. Fall history was obtained by semiannual self-report questionnaires that asked the number of times the participant fell or landed on the ground (excluding falls due to sports participation). In epidemiologic studies, a recurrent faller has been defined as someone who has fallen twice or more during a defined period²²; for our study, we also defined a faller as a person who reported 2 or more falls in a year and determined during a mean follow-up of 7.7 years the number of times this event occurred in loop diuretic users compared with nonusers.

Descriptive analyses (Table 1 and Table 2) were stratified by loop diuretic use at baseline and are presented with the number of participants, means, and standard errors in each group for continuous variables and frequencies and percentages for categorical covariates. For continuous covariates, P values were taken from a loop diuretic term of an unadjusted linear model that modeled the continuous covariate as a function of baseline loop diuretic use, whereas for categorical variables, a χ² test was used. Modeling was performed with Cox proportional hazards models and includes results with both minimal adjustment (age, ethnicity, and BMI) and full adjustment (age; ethnicity; BMI; smoking status; alcohol intake; calcium and vitamin D intake; prevalent fracture at the age of 55 years or older; history of falls; number of chronic conditions; history of chronic heart disease; prevalent CHF; time-dependent incident CHF (to account for the presence or absence of the covariate, depending on the time in the model); physical function construct (36-Item Short Form Health Survey); β-blockers, thiazides, or bisphosphonates; past or current use of estrogen, corticosteroids, anticonvulsants, selective estrogen receptor modulators, angiotensin-converting enzyme inhibitors, statins, calcitonin, or heparin or warfarin sodium; age of menopause; physical activity levels; parental history of hip fractures; study site region; and self-reported health. All models were stratified within the models for enrollment in the WHI study arms (hormone therapy, Dietary Modification Trial, Calcium with Vitamin D Supplementation Trial, and observational study). The specified a priori analysis plan selected a stratified analysis to determine whether associations between loop diuretic use and falls and fractures were a function of CHF status. We, therefore, created a time-dependent variable for CHF and loop diuretic use. We did not determine the association between loop diuretic use and changes in BMD as a function of CHF status because of the small number of women with CHF who had these measurements taken.

Results for each outcome are presented with event totals, annualized percentages, and hazard ratios (HRs), with their corresponding 95% CIs. Fracture and fall outcomes were individually modeled first by any loop diuretic use at baseline and then by duration of loop diuretic use at baseline. Results are presented for both sets of models, with duration of loop diuretic use stratified into groups of less than 1 year, 1 to 3 years, and more than 3 years. Missing data for the covariates of parental history of hip fractures (33.6%), prevalent fractures after the age of 55 years (9.5%), and prevalent falls (3.3%) were set to 0 for all proportional hazards models. For the covariate of number of chronic conditions, if a participant had more than 4 of these chronic conditions at baseline, her value was set to 4.

In all fracture- and fall-adjusted analyses, data on 11.3% of the total sample of 133 855 women were not included because of missing data on covariates. The data analyses for fractures and falls were censored at the date of last follow-up, first fracture, or death, whichever came first.

The mean BMDs by loop diuretic use are presented with sample sizes and standard errors. Comparisons between means in the loop diuretic user and nonuser groups were performed in both minimally and fully adjusted linear models using the baseline covariates described herein. All analyses were conducted with SAS statistical software, version 9.1 (SAS Institute Inc, Cary, North Carolina).

Our data set included 3411 loop diuretic users; 88.9% of these were users of furosemide, 7.8% bumetanide, 2.9% torsemide, and 0.4% ethacrynic acid. Compared with nonusers, loop diuretic users were older, more likely to have had a fracture on or before the age of 55 years, more likely to have a lower physical function construct, more likely to have CHF or chronic heart disease, and more likely to be in the observational study rather than the clinical trial arms of the WHI. Significant differences were found in the WHI enrollees between the loop diuretic users and nonusers with respect to ethnicity (P < .001), smoking status (P < .001), self-reported health (P < .001), number of chronic health conditions (P < .001), and alcohol use (P < .001). Among the clinical trial arms, loop diuretic users were less likely to be enrolled in the Dietary Modification Trial or the Calcium with Vitamin D Supplementation Trial (P ≤ .045) (Table 1). Loop diuretic users were shorter and heavier on average and had a higher BMI, a younger age at menopause, a higher unadjusted BMD of the lumbar spine and total hip, lower levels of physical activity, and lower intakes of vitamin D and calcium than did nonusers (P < .001 for all; Table 1). No significant differences were found in parental history of hip fractures (P = .46) or unadjusted whole-body BMD (P = .23) at baseline between loop diuretic users and nonusers (Table 1). Baseline characteristic patterns similar to those of the whole population were present in women with CHF who were loop diuretic users compared with women with CHF who were nonusers, except that no significant differences were found in personal history of fracture on or after the age of 55 years, baseline vitamin D intake, or trial enrollments (other than the hormone therapy trial). However, there were significant differences in the whole-body bone mineral density between women with CHF who were loop diuretic users compared with women with CHF who were nonusers of loop diuretics (Table 2). Compared with nonusers of loop diuretics, users were significantly more likely to use β-blockers, calcitonin, anticonvulsants, corticosteroids, heparin or warfarin sodium, angiotensin-converting enzyme inhibitors, and statins (P < .001 for all) and significantly less likely to use thiazides (P < .001). Use of hormone therapy also differed significantly between the loop diuretic users and nonusers (P < .001). No significant differences were found between the groups with respect to use of bisphosphonate (P = .38) or selective estrogen receptor modulators (P = .95) (Table 3). Among women with CHF, users of loop diuretics were significantly more likely to use β-blockers, corticosteroids, heparin or warfarin sodium, angiotensin-converting enzyme inhibitors, and statins (P ≤ .03) and significantly less likely to use thiazides (P < .001). No significant differences were found for loop diuretic use in women with CHF with respect to use of hormone replacement therapy (P = .15), bisphosphonates (P = .52), selective estrogen receptor modulators (P = .63, calcitonin (P = .65), or anticonvulsants (P = .23) (Table 4).

In models adjusted for age, ethnicity, and BMI, a significant association was found between loop diuretic use and total fractures (HR, 1.31; 95% CI, 1.20-1.42), hip fractures (1.75; 1.34-2.28), clinical vertebral fractures (1.68; 1.35-2.10), other fractures (1.27; 1.15-1.41), and 2 or more falls (1.37; 1.30-1.45), and no statistically significant association was found between loop diuretic use and lower arm or wrist fractures (1.17; 0.97-1.41) (Table 5). However, in fully adjusted models, the association between loop diuretic use (ever) and total fractures (HR, 1.09; 95% CI, 1.00-1.19; P = .052), hip fractures (1.21; 0.91-1.60), clinical vertebral fractures (1.17; 0.92-1.48; P = .20), and falls (1.01; 0.96-1.08; P = .62) was no longer statistically significant (Table 5). Although the trends for duration of use of loop diuretics and falls and fractures were not statistically significant, we observed modest increased risks for other clinical fractures (1.16; 1.01-1.33) and total fractures (HR, 1.16; 95% CI, 1.03-1.31) in women who had used loop diuretics for more than 3 years.

In minimally adjusted models in the 2827 women with incident CHF, loop diuretic use was significantly inversely associated with clinical vertebral fractures but not hip fractures (HR, 0.69; 95% CI, 0.35-1.37). In fully adjusted models in women with CHF, loop diuretic use was also significantly inversely associated with both hip (HR, 0.44; 95% CI, 0.21-0.94) and clinical vertebral (0.46; 0.21-0.97) fractures although there were only 16 hip-fracture cases (3 in loop diuretic users and 13 in nonusers) and only 23 clinical vertebral fractures (3 in loop diuretic users and 20 in nonusers). In fully adjusted models, in a subgroup analysis of 1146 of these women with adjudicated incident CHF who, in addition to a clinical diagnosis of CHF, had imaging procedures that documented impaired systolic or diastolic function, there was similar directionality of findings (hip fractures: HR, 0.27; 95% CI, 0.06-1.23; P = .09; clinical vertebral fractures: HR, 0.39; 95% CI, 0.10-1.50; P = .17), although the results were not statistically significant.

The change in BMD analysis included 300 loop diuretic users and 9124 nonusers. In this subset, in the fully adjusted models, no significant differences were found between loop diuretic users and nonusers in baseline BMD of the total hip (P = .45), lumbar spine (P = .31), total body (P = .29), year 3 BMD of the total hip (P = .34), lumbar spine (P = .05), total body (P = .05), or change in BMD measurements from baseline to year 3 (Table 6).

In this large, prospective, population-based study of postmenopausal women, after adjustment for potential confounding variables, no significant association was found between ever use of loop diuretics and total or site-specific fractures, including hip, clinical vertebral, lower arm or wrist, and other clinical fractures. A small positive association was found between prolonged duration of use (>3 years) of loop diuretics and other clinical and total fractures. After adjustment for potential confounding variables, no significant association was found between loop diuretics and falls or change in BMD at multiple skeletal sites.

A small positive relationship between ever use of loop diuretics and hip fractures¹⁷ and total fractures^17,23 has been reported. In the WHI, large differences in health status were found between users and nonusers of loop diuretics. Loop diuretic users were at high risk of fracture, independent of frequency of use; after controlling for confounding by indication,²⁴ we found no statistically significant association between ever use of loop diuretics and fractures. In support of this concept, it has been suggested that the relationship between thiazides and fractures may reflect selection bias, rather than being an independent association of the drug with fracture.²⁵ Indeed, in a small case-control study, in which, similar to our results, no significant association was found between ever use of loop diuretics and femoral neck fractures, the authors postulated that health status itself, rather than medication use, was the critically important factor.¹⁸ However, in the WHI, women who had used loop diuretics for a prolonged period had a modest increased risk for other clinical and total fractures, even after controlling for other covariates, including health status. Therefore, it may be important to consider fracture prevention measures in women who receive loop diuretic therapy.

Conversely, in women with CHF in the WHI, a small inverse association was found between loop diuretics and hip and clinical vertebral fractures. Interestingly, loop diuretics have been reported to improve functional capacity²⁶ and reduce oxidative stress.²⁷ Oxidative stress has been linked with osteoporosis²⁸ and CHF.²⁷ However, only 3.2% of our study population had CHF and, of these, only 19.3% were loop diuretic users. In contrast, national data suggest that 6% to 10% of elderly persons have CHF,²⁹ and most are prescribed loop diuretics.³⁰ Although it is likely that these findings in women with CHF, which were based on small numbers, were simply attributable to chance, additional studies of the relationship between loop diuretic use in women with CHF and osteoporosis are needed.

The relationship of falls to medication use is complex. We hypothesized that loop diuretics might be positively associated with falls because they may cause orthostatic hypotension. However, in a large meta-analysis,¹¹ gait and balance problems and prior falls were the principal risk factors for future falls; no independent relationship was found between orthostatic hypotension and falls. In support of our lack of association between loop diuretic use and falls, another study¹⁵ states that cardiovascular drugs are not independently associated with falls.

Cross-sectionally, in our study, in agreement with what has been previously published,^12,31 unadjusted BMD was higher at the total hip site in users of loop diuretics compared with nonusers; these differences in BMD disappeared in adjusted models. Longitudinally, after adjustments, no significant differences were found in the changes in BMD at any skeletal site measured between loop diuretic users and nonusers in the WHI. In contrast, in a randomized clinical trial, significant losses in BMD were reported in loop diuretic users¹³ and, in men in the Osteoporotic Fractures in Men Study, loop diuretic use was associated with increased rates of hip bone loss.³²

Our study has a number of limitations. Prefracture health, which may change in older populations, is a predictor of fracture,^33- 34 and we only included baseline levels of prefracture health. Low levels of 25-hydroxyvitamin D may be associated with BMD,³⁵ falls,³⁶ and fractures³⁷; however, we could not adjust for this. A particularly important limitation of our study is the lack of information on doses of loop diuretics used; dose effects of loop diuretics on calcium homeostasis³⁸ and fracture risk have been reported.¹⁷ We could not determine whether there was a residual risk for changes in bone health after use of diuretics had been stopped because we did not have the exact date that use of these drugs was discontinued. There were few users of loop diuretics other than furosemide. Our fall history was from self-report, which may be inaccurate.³⁹ There were few hip and vertebral fractures in those with CHF. The mean calcium intake was 1083 mg/d in the total loop diuretic population and 1139 mg/d in those with incident CHF. Our results may not be applicable to populations with lower calcium intakes, those primarily being treated with loop diuretics, or men. Finally, this study is not applicable to thiazide-type diuretics, which have been associated with increases in BMD⁴⁰ and lower forearm and hip fracture risk.⁴¹

In conclusion, no significant association was found between ever use of loop diuretics and changes in BMD, falls, or fractures in postmenopausal women in the WHI. However, prolonged use of loop diuretics was associated with higher fracture risk in postmenopausal women. Loop diuretics are most commonly used by women in poor health who are already at risk for falls, fractures, and loss of BMD.

Correspondence: Laura D. Carbone, MD, MS, University of Tennessee Health Science Center, 956 Court Ave, Memphis, TN 38163 (LCarbone@utmem.edu).

Accepted for Publication: July 11, 2008.

Author Contributions:Study concept and design: Carbone and Johnson. Acquisition of data: Johnson and Robbins. Analysis and interpretation of data: Carbone, Johnson, Bush, Larson, Thomas, and LaCroix. Drafting of the manuscript: Carbone and Johnson. Critical revision of the manuscript for important intellectual content: Carbone, Bush, Robbins, Larson, Thomas, and LaCroix. Statistical analysis: Bush, Larson, and LaCroix. Obtained funding: Johnson and Robbins. Administrative, technical, and material support: Carbone, Johnson, Robbins, and Thomas. Study supervision: LaCroix.

Financial Disclosure: Dr Carbone has received grant support from Procter & Gamble and honorariums from Merck, Novartis, Procter & Gamble, and Aventis.

Funding/Support: The WHI Program is funded by the National Heart, Lung, and Blood Institute of Health, US Public Health Service contract N01-WH-3-2118.