Fracture risk in patients treated with loop diuretics
L. Rejnmark, Department of Endocrinology and Metabolism C, Aarhus University Hospital, Aarhus Sygehus, Tage-Hansens Gade 2, DK-8000 Aarhus C, Denmark.
(fax: +45 8949 7684; e-mail: email@example.com).
Background. Loop diuretics (LD) increase renal calcium excretion. Discrepant results on associations between LD and fracture risk have been reported.
Objective. To assess the fracture risk in users of LD.
Design and subjects. A population-based pharmaco-epidemiological case–control design with fracture in year 2000 as outcome and use of LD during the previous 5 years as exposure variable. We used nationwide computerized registers to assess individual use of LD and related these data to individual fracture data and information on potential confounders. We compared 64 699 cases aged 40 years or more who sustained a fracture during year 2000 with 194 111 age- and gender-matched controls.
Results. A total of 44 001 subjects used LD. Ever use of LD was associated with a crude 51% (OR 1.51; 95% CI 1.48–1.55) increased risk of any fracture and a 72% (OR 1.72; 95% CI 1.64–1.81) increased risk of hip fracture. The risk estimates were reduced after confounder adjustment, i.e. adjusted risk of any fracture was increased by 4% (OR 1.04; 95% CI 1.01–1.07) and risk of hip fracture by 16% (OR 1.16; 95% CI 1.10–1.23). In current users, a tendency towards a decreased fracture risk with increased dose was observed, whereas in former users risk of fracture increased with increased dose. Use of furosemide was associated with higher risk estimates than use of bumetanide.
Conclusion. Treatment with LD affects fracture risk. Special attention should be paid to patients in whom treatment with LD is initiated or stopped, as they may be at an increased risk of fracture.
Loop diuretics (LD) are one of the most widely used groups of drugs. In addition to their effect on the renal and cardiovascular system, treatment with LD affects calcium homeostasis and bone metabolism. In response to treatment with LD, the renal calcium excretion and plasma PTH (parathyroid hormone) and 1,25(OH)2D levels are increased [1–5]. Thus, the increased urinary calcium losses may be compensated for by a 1,25(OH)2D-dependent augmented intestinal calcium absorption, mediated through increased PTH levels . However, there are only few data on whether these changes in calcium homeostasis affect bone metabolism and fracture risk. In a cross-sectional study by Ooms et al. , users of LD had a 5% lower BMD (bone mineral density) at the hip compared with non-users (P < 0.05), corrected for confounding by age, years since menopause, and body weight. However, in a recent study by our group, comparing 140 postmenopausal women treated with an LD for more than 2 years with 140 age-matched women not in diuretic therapy, BMD did not differ between groups after adjustment for differences in body weight . Nevertheless, our study showed a positive correlation between BMD and duration of therapy with an increase in BMD at the spine of 0.3% per 1 year of treatment .
The effects of LD on fracture risk have been studied in few epidemiological studies, with conflicting results. In a cohort study including 348 postmenopausal women followed for 5 ± 1.7 years, use of LD was associated with an increased risk of osteoporotic fractures (RR = 2.5, 95% CI 1.1–5.7) . Similarly, in a case–control study , current users of LD had a nearly fourfold increased risk of hip fractures (OR = 3.9, 95% CI 1.5–10.4). However, in two other case–control studies, no association was found between LD use and fracture risk [8, 9]. Thus, Rashiq and Logan  found use of LD to be associated with a nonsignificant reduced risk of fracture (OR = 0.68, 95% CI 0.28–1.7). Similarly, after adjustment for age, sex, and type of residence (community versus nursing home), Cumming and Klineberg  found no effect of treatment with LD on fracture risk (OR = 0.93, 95% CI 0.57–1.49). None of the four epidemiological studies examined associations between fracture risk and dosage and duration of use. As LD are frequently used in elderly in whom osteoporosis often occurs, there is a need for further evaluation of the potential effects of LD on calcium homeostasis and bone metabolism. The aim of the present study is to determine associations between treatment with LD and fracture risk.
Subjects and methods
Due to the extensive nature of registers covering contacts to the health sector, Denmark offers good possibilities for studies on the occurrence of fractures . This case–control study was performed within the Danish population that constituted approximately 5.3 million individuals during the study period. In Denmark, the National Health Service provides tax-supported health care for all inhabitants, allowing free access to general practitioners and hospitals. In addition, reimbursement is provided for a variable proportion of the costs of prescribed drugs. Using the unique 10-digit civil registry number that is assigned to all Danish citizens shortly after birth, a complete hospital discharge and prescription history can be established for each individual, and valid linkage between population-based registries can be obtained.
Identification of fracture cases
In Denmark, The National Hospital Discharge Register covers all contacts (on in- or outpatient basis) to the hospitals . The register was founded in 1977, but outpatient records were first completely incorporated from 1995. The files of The National Hospital Discharge Register include information on the civil registry number of the patient, date of discharge, and discharge diagnoses, assigned exclusively by the physician at discharge. The Danish version of the International Classification of Diseases, 8th revision (ICD-8) was used until the end of 1993, followed by the 10th revision (ICD-10) hereafter. The register has nationwide coverage of public hospitals with an almost 100% completeness of recordings and a high precision of diagnoses , especially of fracture diagnoses . Using The National Hospital Discharge Register we identified all subjects older than 40 years of age who had sustained a fracture between 1 January 2000 and 31 December 2000 (n = 64 699).
Selection of population-based controls
Using the Civil Registration System, which has electronic records on all changes in vital status, including change of address and date of death for the entire Danish population since 1968, we randomly selected three controls for each case, matched by gender and year of birth. The controls were selected using the incidence-density sampling technique , i.e. the controls had to be alive and at risk for fracture diagnosis at the time the corresponding case was diagnosed. A total of 194 111 controls were included in the study.
Data on use of loop diuretics
In Denmark, pharmacies are equipped with a computerized accounting system through which data are sent directly to a prescription database at The Danish Medicines Agency with key information on prescriptions for refundable drugs. The prescription database includes information on the patient's civil registry number, the type and amount of drug prescribed according to the Anatomical Therapeutical Chemical classification system [14, 15], and the date the prescription was filled.
Within the Prescription Database, we identified all prescriptions for LD amongst cases and controls redeemed within 5 years before the date of hospitalization of cases. LD are available in Denmark only by prescription, and the public health authority provides reimbursement for expenses on LD. Dose–effects relationships were investigated in terms of Defined Daily Dosages (DDD), i.e. one DDD is equal to 40 mg of furosemide or 1 mg of bumetanide. Current users were defined as subjects who had redeemed at least one prescription for an LD the year preceding fracture. Former users were defined as those who had used LD within 5 years before the date of hospitalization of cases, but not the year preceding fracture.
Data on confounding factors
In order to adjust for potential differences in comorbidity between cases and controls, we used the Charlson comorbidity index . The index includes 19 diseases, which have been selected and weighted on the basis of the strength of their association with mortality . Thus, for cases and controls, we identified all discharge diagnoses covering the 19 diseases included in the comorbidity index, using The National Hospital Discharge Register. Moreover, using this register, number of days spent in hospital the year preceding fracture (year 1999) and a history of a prior fracture in the period 1977–2000 were also included as confounders.
In addition, the prescription database was used to identify the use of other medications believed to affect fracture risk. We retrieved data on whether subjects, during the period of 1996–2000 had redeemed a prescription (‘yes’ or ‘no’) for the following drugs: any diuretic other than LD (thiazide diuretics, potassium sparing, and other diuretics), antihypertensive drugs (except diuretics, i.e. beta-blockers, calcium channel blockers, angiotensin-converting enzyme inhibitors, and other types of antihypertensive drugs), antiresorptive drugs (oestrogens, bisphosphonates and/or raloxifene), antiepileptic drugs, oral anticoagulants, anxiolytics, sedatives, neuroleptics, antidepressants, systemic and topical corticosteroids, thyroid hormones, and antithyroid drugs.
Furthermore, our confounder analysis included 1999 data from the National Bureau of Statistics on income, social status, working status, and educational status, as well as data from The National Health Organisation Register on number of contacts to general practitioners and practising specialists for the period 1996–2001.
Data from the different registers were merged at the National Bureau of Statistics, and for each subject the 10-digit civil registry number was substituted by a unique case number, i.e. as investigators we had no access to personally identifiable information. Mean and standard deviation were used as descriptive statistics. Crude odds ratios (ORs) were calculated and 95% confidence intervals (CIs) approximated using the method of Miettinen . A conditional logistic regression analysis was used to assess the association between any fracture and the exposure variable. Analyses were also performed gender- and age stratified. Tests for trend were performed by rank-sum method. Analyses were performed using stata 8.1 (STATA Corp., College Station, TX, USA) and spss 10.1.0 (SPSS Inc., Chicago, IL, USA) – both in the Unix version.
Cases and controls were well matched on age and gender, whereas most other characteristics differed significantly between groups (Table 1). Compared with the controls, more cases were single, unemployed, had previous fractures, and in general had used more drugs and had more concurrent diseases (Table 1). In addition, use of LD was more frequent in cases (21.7%) than in the controls (15.5%, Table 1). Thus, ever use of LD was associated with a 51% (OR 1.51; 95% CI 1.48–1.55) increased risk of any fracture as well as an increased crude risk of fractures at the hip, spine and forearm (Table 2). Risk estimates were lowered by adjustment for potential confounders, i.e. the variables shown in Table 1. However, after adjustment risk of any fracture (OR 1.04; 95% CI 1.01–1.07) and risk of hip fracture (OR 1.16; 95% CI 1.10–1.23) were significantly increased (Table 2).
Table 1. Characteristics of patients with a fracture during year 2000 (cases) and controls
|Age (years)||66.5 ± 15.4||66.5 ± 15.4||–|
| Men||22 312 (34.5)||66 943 (34.5)||–|
| Women||42 387 (65.5)||127 168 (65.5)||–|
|Annual income (DKR)||172 771 ± 144 807||186 074 ± 217 985||<0.01|
| Widowed||18 298 (28.3)||52 372 (27.0)||<0.01|
| Divorced||9283 (14.3)||20 929 (10.8)|
| Married||29 772 (46.0)||102 110 (52.6)|
| Unmarried||7078 (10.4)||17 939 (9.2)|
| Other||56 (0.1)||162 (0.1)|
| Independent||2234 (3.5)||8048 (4.2)||<0.01|
| Assisting wife||193 (0.3)||876 (0.5)|
| Working||17 556 (27.2)||61 732 (31.8)|
| Retired||39 267 (60.8)||107 809 (55.6)|
| Other||5329 (8.3)||15 405 (7.9)|
| 0 (no comorbidity)||40 938 (63.3)||142 416 (73.4)||<0.01|
| 1–2||16 372 (25.3)||40 132 (20.7)|
| 3–4||5154 (8.0)||8668 (4.5)|
| ≥5||2235 (3.5)||2895 (1.5)|
|Previous fracture||24 481 (37.8)||31 503 (16.2)||<0.01|
|Number of bed days in 1999||14.9 ± 49.0||6.2 ± 26.1||<0.01|
|Contacts to GP or specialists in 1999||34.7 ± 54.0||25.6 ± 38.8||<0.01|
|Ever use of anti-osteoporotic drugs (including HRT)||12 114 (18.7)||25 934 (13.4)||<0.01|
|Ever use of corticosteroid||40 124 (62.0)||112 427 (57.9)||<0.01|
|Ever use of levothyroxine||2625 (4.1)||7028 (3.6)||<0.01|
|Ever use of antithyroid drug||1171 (1.8)||3404 (1.8)||0.35|
|Anxiolytics, sedatives and hypnotics||31 789 (49.1)||75 029 (38.7)||<0.01|
|Neuroleptics||8624 (13.3)||15 346 (7.9)||<0.01|
|Ever use of antidepressant||16 509 (25.5)||30 612 (15.8)||<0.01|
|Ever use of antiepileptic drug||5558 (8.6)||8533 (4.4)||<0.01|
|Ever use of AK treatment||2576 (4.0)||5867 (3.0)||<0.01|
|Ever use of any diuretics||25 896 (40.0)||67 731 (34.9)||<0.01|
|Ever use of loop diuretics||14 008 (21.7)||29 993 (15.5)||<0.01|
|Ever use of furosemide||13 594 (21.0)||28 824 (14.8)||<0.01|
|Ever use of bumetanide||808 (1.2)||1936 (1.0)||<0.01|
|Current use of loop diuretics||10 393 (16.1)||21 599 (11.1)||<0.01|
|Former use of loop diuretics||3615 (5.6)||8394 (4.3)||<0.01|
|Ever use of potassium sparing diuretics||2780 (4.3)||5049 (2.6)||<0.01|
|Ever use of thiazides||15 316 (23.7)||42 647 (22.0)||<0.01|
|Ever use of other diuretics||4599 (7.1)||13 893 (7.2)||0.68|
|Ever use of antihypertensives (other than diuretics)||18 825 (29.1)||54 135 (27.9)||<0.01|
Table 2. Crude and adjusted odds ratios (OR) for fractures in ever users of loop diuretics
|Any fracture||1.51 (1.48–1.55)||1.04 (1.01–1.07)|
|Hip fracture||1.72 (1.64–1.81)||1.16 (1.10–1.23)|
|Spine fracture||1.63 (1.47–1.81)||1.03 (0.91–1.17)|
|Forearm fracture||1.20 (1.13–1.28)||0.98 (0.91–1.05)|
Table 3 shows dose–effect relationships stratified by former and current use. In former users, risk of fracture increased with increased dose of LD. Thus, risk of any fracture as well as risk of hip fracture was significantly increased with doses >1 DDD day−1 in former users.
Table 3. Dose–effect relationships in former and current users on associations between average daily dose of loop diuretics and risk of fracture at different skeletal sites; all subjects and stratified by gender
|<0.25||0.96 (0.91–1.01)||1.43 (1.28–1.59)|
|0.25–0.99||1.08 (0.99–1.18)||1.23 (1.18–1.30)|
|1–1.99||1.69 (1.37–2.09)||1.00 (0.95–1.04)|
|>2||2.62 (1.72–3.99)||1.00 (0.95–1.06)|
|<0.25||0.83 (0.73–0.93)||1.75 (1.40–2.18)|
|0.25–0.99||1.13 (0.94–1.34)||1.56 (1.42–1.71)|
|1–1.99||1.75 (1.20–2.55)||1.10 (1.01–1.20)|
|>2||7.42 (2.34–23.5)||1.13 (1.03–1.25)|
|<0.25||0.92 (0.72–1.18)||1.94 (1.19–3.19)|
|0.25–0.99||1.05 (0.72–1.53)||1.16 (0.92–1.44)|
|1–1.99||2.06 (0.58–7.41)||0.99 (0.81–1.22)|
|>2||1.64 (0.21–12.7)||1.07 (0.84–1.37)|
|<0.25||0.99 (0.88–1.13)||1.10 (0.84–1.45)|
|0.25–0.99||1.13 (0.91–1.40)||0.96 (0.85–1.09)|
|1–1.99||0.93 (0.51–1.68)||1.02 (0.91–1.15)|
|>2||5.09 (1.25–20.8)||0.89 (0.77–1.02)|
However, in current users the analyses showed a tendency towards a decreased fracture risk as number of DDD increased. Thus, in patients who had used <1 DDD day−1 the risk of any fracture as well as the risk of fractures at the hip and spine was increased, whereas no significant increase in risk of any fracture or risk of vertebral fractures was found in users of more than 1 DDD day−1. However, current use of high doses of LD (>1 DDD day−1) was associated with an increased hip fracture risk (Table 3). Subanalysis in women and men, and in subjects younger or older than 65 years of age, as well as performing the analyses calculating accumulated dose instead of average daily dose revealed a similar pattern on risk of fracture in users of LD (data not shown).
Furosemide versus bumetanide
Furosemide was used more often than bumetanide (Table 1). A direct comparison of risk estimates showed a significantly higher risk of fractures at the hip (P = 0.02) and forearm (P = 0.04), as well as a borderline significantly higher risk of any fracture (P = 0.05) in users of furosemide compared with users of bumetanide (Table 4). Dose–effect analyses revealed similar results, with higher risk estimates in users of furosemide than in users of bumetanide (data not shown).
Table 4. Risk of fracture in current users of bumetanide and furosemide; all subjects and stratified by gender
|Any fracture||0.95 (0.86–1.06)||1.06 (1.03–1.09)||0.05|
|Hip fracture||0.97 (0.80–1.19)||1.24 (1.17–1.32)||0.02|
|Spine fracture||1.41 (0.90–2.24)||1.04 (0.90–1.20)||0.21|
|Forearm fracture||0.74 (0.57–0.98)||0.99 (0.91–1.07)||0.04|
In a nationwide study, including 64 699 patients who had sustained a fracture and 194 111 age- and gender-matched controls, we found use of LD to influence fracture risk. After adjustment for important confounders an inverse relationship was found between fracture risk and number of redeemed DDD of LD. However, stratification by former and current use revealed important interactions, as fracture risk only decreased with increased dose in current users. Finally, our analyses showed a higher fracture risk in users of furosemide than in users of bumetanide. These findings may help explain the conflicting results of previous studies on associations between use of LD and fracture risk.
Until now, only three case–control and one cohort study have reported associations on fracture risk in users of LD. Rashiq and Logan  compared 102 hip fracture patients with 204 age- and sex-matched controls, and found a nonsignificant reduced hip fracture risk in users of LD (OR 0.68; 95% CI 0.28–1.7). Similarly, in 209 hip fracture patients compared with 207 controls, Cumming and Klineberg  found use of furosemide to be associated with a nonsignificantly reduced hip fracture risk (OR 0.93; 95% CI 0.57–1.49), after adjustment for age, sex, and type of residence (community versus nursing home). In contrast, an increased fracture risk has been reported in two studies. Thus, in a case–control study, including 462 hip fracture patients and an equal number of age- and sex-matched controls, Heidrich et al.  found a nonsignificant 60% increased risk of hip fractures (OR 1.6; 95% CI 0.8–2.9), after adjustment for different medical conditions (alcoholism, organic brain syndrome, leg paralysis, history of stroke, days of hospitalization in the year preceding fracture, nursing home residence), body mass index, and use of phenobarbital, corticosteroids, and thiazides. In the study by Heidrich et al. , stratification by former and current use was performed; 52 subjects were defined as current users [i.e. they had received a prescription for furosemide within 3 months of the index (fracture) date], and 21 subjects were defined as former users. Whereas no significant association was found in former users (adjusted OR 1.2; 95% CI 0.2–7.0), an approximately fourfold increased hip fracture risk was found in current users of LD (adjusted OR 3.9; 95% CI 1.5–10.4). Similarly, Tromp et al.  found use of LD to be associated with an increased fracture risk in 348 elderly women followed prospectively for 5 years. In the cohort, 50 women reported use of LD at baseline, and 83 fractures (including 16 hip fractures) occurred during follow-up. After adjustment for age, the study showed a trend towards an increased risk of hip fractures (RR 2.2; 95% CI 0.7–6.7), osteoporotic fractures (RR 2.1; 95% CI 1.0–4.7), and nonvertebral fractures (RR 1.3; 95% CI 0.6–2.7). Moreover, after further adjustment for body mass index, a previous postmenopausal fracture, low mobility, and differences in levels of sex hormone binding globulin, baseline use of LD was associated with a significantly increased risk of any osteoporotic fracture (RR 2.5; 95% CI 1.1–5.7).
In addition to differences in study size and confounder adjustment, the discrepant results of previous studies may be explained by the lack of dose–effect analyses. According to our results, dose and duration of use influence the association between use of LD and fracture risk. Thus, one explanation for the discrepant results of previous studies could be that the dose of LD used by studied subjects has differed between study populations. Accordingly, if most subjects in a study group have received a high dose of LD, the overall result may show no effect or a decreased fracture risk, whereas the opposite may be the case if a high proportion of studied subjects has been treated with only a low dose of LD.
The most obvious explanation for the increased fracture risk in users of LD is the hypercalciuria induced by treatment [1–3]. In patients with hypercalciuria histomorphometric analyses have revealed an increased bone turnover [20, 21]. Accordingly, if bone turnover is increased in response to treatment with LD, fracture risk may increase. Our findings of a relative decrease in fracture risk in long-term users of LD may be explained by several mechanisms. Importantly, the relatively decreased fracture risk in patients on long-term treatment with LD agrees with the results of a recent cross-sectional study from our group. Comparing 140 postmenopausal women treated with an LD for more than 2 years with 140 age-matched women not in diuretic therapy, we found duration of therapy to be positively correlated with BMD at the spine and whole body, i.e. BMD at the spine increased by 0.3% per year of treatment . Whether these changes in BMD and fracture risk are due to the consequences of long-term changes in calcium homeostasis with increased plasma PTH and 1,25(OH)2D levels needs further studies [4, 22]. An additional explanation for the changes in fracture risk during long-term treatment with LD is that during initiation of treatment, the blood pressure-lowering effect of LD may cause dizziness and increase fall- and fracture tendency [23, 24]. Patients who experience side effects may subsequently stop therapy, whereas those who continue treatment tolerate it without major side effects. Similarly, patients may have contacted their doctor for various diseases requiring treatment with LD, or for other diseases and coincidentally been diagnosed with conditions necessitating treatment with LD. The disease causing patients to contact their doctor may per se be associated with an increased fracture risk, and as the disease is treated fracture risk may decline. If so, the reduced fracture risk in patients who have used a high dose of LD may rather be due to a healthy drug user, rather than a pharmacological, effect.
The main effect of furosemide and bumetanide is caused by an inhibition of the Na+/K+/2Cl− transport system in the luminal membrane of the thick ascending limb of the loop of Henle. Thereby, the normal lumen-positive potential is diminished causing an increased renal calcium excretion . However, in addition to the inhibition of the Na+/K+/2Cl− transport system in the loop of Henle, bumetanide (but not furosemide) is known to inhibit the sodium phosphate (Na-Pi) cotransporter in the proximal tubule of the renal nephron . Interestingly, the Na-Pi cotransporter is also expressed by osteoclastic cells and inhibition of the cotransporter has been shown to cause a decreased osteoclast-mediated bone resorption [27, 28]. Thus, our findings of a lower fracture risk in users of bumetanide compared with users of furosemide could be due to a direct inhibition of bone resorption caused by bumetanide.
Strengths and limitations to the study
The main strengths of our study are the uniformly organized healthcare system allowing a large-scale population-based design, and the use of data on exposure and confounders that are collected before the date of fracture. Thus, recall bias did not influence data collection.
The weaknesses include potential selection bias, e.g. the use of routine hospital discharge diagnoses of fractures coded by hospital doctors to ascertain case status. Some coding errors have probably occurred. However, misclassification of case status is unlikely to be related to prescription of LD before hospitalization, and any nondifferential misclassification will lead to underestimation of our risk estimates. Moreover, the positive predictive value of hip fracture discharge diagnoses from Danish hospitals has previously been shown to be as high as 93% , and the risk of misclassification of case status is thus most likely of minor importance.
We cannot exclude that information biases may have influenced our results, e.g. we had no information on patient compliance in our study, as redeeming a prescription was used as a proxy for actual use of a drug. However, the patients had paid for part of the cost of the drug, which increases the likelihood of compliance. Furthermore, data on drugs, including LD, administered during hospitalization are not registered in the prescription database and therefore not included in our study. Both of these uncertainties could have lead to misclassification of the exposure. Very few subjects had used specific anti-osteoporotic drugs (most had used oestrogen-containing compounds), and the higher use amongst fracture cases was because of confounding-by-indication.
Although we adjusted for several potential confounding factors in the statistical analyses, our results may still be influenced by potential confounding factors not included in the analyses, e.g. smoking, physical activity, differences in body weight, use of calcium/vitamin D supplements, or by residual confounding due to the use of crude measures (risk of falling).
In a large-scale nationwide case–control study, use of LD was associated with an increased risk of any fracture as well as an increased hip fracture risk. Dose–effect analysis revealed that with increased averages daily dose risk estimates increased in former users but decreased in current users. In addition, we found furosemide to be more harmful to bone than bumetanide. As the changes in fracture risk in patients treated with LD may be due to circumstances related to the disease necessitating treatment, as well as changes in calcium homeostasis caused by therapy, further studies are needed to explore causal relationships. Until the results of such studies are available, special attention should be paid to patients in whom treatment with LD is initiated or stopped, as they may be at an increased risk of fracture.
Conflict of interest statement
No conflict of interest was declared.
The National Bureau of Statistics, Denmark, is acknowledged for its help in providing data. Financial support was provided by the Danish Medical Research Council.