DIURETIC USE AND RISK OF VERTEBRAL FRACTURE IN WOMEN

Abstract

Background

Vertebral fracture is the most common type of osteoporotic fracture. While thiazide diuretics, which are commonly prescribed for the treatment of hypertension, decrease calciuria, they may also induce hyponatremia, which has been associated with increased vertebral fracture risk. Loop diuretics increase calciuria, which would reduce bone mineral density and increase vertebral fracture risk, but they rarely cause hyponatremia. Recent studies on diuretics and fractures did not include or specifically examine vertebral fracture. The few studies of diuretics and vertebral fracture have been limited by cases defined by self-report or administrative data, relatively small number of cases, study design that was not prospective, and lack of long-term follow-up with updated information on diuretic use.

Methods

We conducted a prospective cohort study of thiazide diuretic use, loop diuretic use, and risk of incident clinical vertebral fracture in 55,780 women, 55–82 years of age, participating in the Nurses’ Health Study, without a prior history of any fracture. Diuretic use was assessed by questionnaire every four years. Self-reported vertebral fracture was confirmed by medical record review. Cox proportional-hazards models were used to simultaneously adjust for potential confounders.

Results

Our analysis included 420 incident vertebral fracture cases documented between 2002 and 2012. The multivariate-adjusted relative risk of clinical vertebral fracture for women taking thiazides compared with women not taking thiazides was 1.47 (95% CI 1.18 to 1.85). The multivariate adjusted relative risk of vertebral fracture for women taking loop diuretics compared with women not taking loop diuretics was 1.59 (95% CI 1.12 to 2.25).

Conclusion

Thiazide diuretics and loop diuretics are each independently associated with increased risk of vertebral fracture in women.

Introduction

Vertebral fracture is the most common type of osteoporotic fracture¹ and is associated with significant disability,² morbidity ³ and mortality.⁴ Twenty-five percent of postmenopausal women in the United States are estimated to have a vertebral fracture⁵ and the prevalence increases with advancing age.⁶ In recent years, the incidence of vertebral fracture appears to be rising dramatically in men and women, especially after age 75 years, whereas the incidence of hip fracture is declining.⁷ Risk factors for vertebral fracture may differ from those for fractures at other sites due to different microarchitecture,^8–10 biomechanics,¹¹ and compressive loading.^{12, 13}

Thiazide diuretics, which are associated with decreased cardiovascular events and mortality,^{14, 15} are commonly prescribed for treating hypertension.^16,17 In the U.S., they are the fourth most frequent prescription drug and tenth most used drug overall.¹⁸ In older U.S. adults, hydrochlorothiazide is the second most commonly used prescription or over-the-counter drug.¹⁹ Thiazides decrease urinary calcium excretion²⁰ and improve calcium balance,²¹ which preserve bone mineral density at the vertebral body,^{22, 23} but thiazides can also induce hyponatremia and several recent studies suggest that hyponatremia is associated with increased fracture risk.^24–29 While thiazide use has been found to be protective for hip fracture in multiple studies,^30–35 the two studies of thiazide use and vertebral fracture^{36, 37} found no statistically significant association.

Loop diuretics increase urinary calcium excretion,³⁸ which could lower bone mineral density, a known risk factor for vertebral fracture,³⁷ but they rarely cause hyponatremia. However, results have been inconsistent on the association between loop diuretics and bone mineral density.^39–42 The two studies to date on the association between loop diuretics and vertebral fracture^43,44 found no statistically significant association.

The few studies to date on the association between diuretic use and vertebral fracture have been limited by cases that were defined by self-report or administrative data rather than medical record review, relatively small number of cases, study design that was not prospective, and lack of long-term follow-up with updated exposure information on diuretic use. Therefore, we studied the prospective association between thiazide diuretic use, loop diuretic use, and risk of incident clinical vertebral fracture over a 10-year period in the Nurses’ Health Study.

Methods

Study Population

The Nurses’ Health Study (NHS) is an ongoing, prospective cohort study which began in 1976, enrolling 121,700 female registered nurses 30–55 years of age. The cohort is followed with biennial mailed questionnaires that ask about lifestyle practices, medications and newly diagnosed diseases. The follow-up rate has been >90% of the eligible person-time. Deaths are confirmed through the National Death Index. Approximately 98% of the cohort is white.

This analysis includes 55,780 women who answered the 2012 questionnaire, which included the question on lifetime history of vertebral fracture, and who also answered the 2002 questionnaire, which included questions on thiazide and loop diuretic use, and serves as the baseline year for this analysis. Updated information on diuretic use was obtained every four years during the follow-up period. The study protocol was approved by the Brigham and Women’s Hospital Institutional Review Board.

Assessment of Diuretic Use

On the 2002, 2006 and 2010 questionnaires, nurses were asked if they regularly took “Thiazide diuretic” or “Lasix” (the brand name for a loop diuretic) in the past two years. The question did not ask about diuretic dose or frequency. Thiazide use and loop diuretic (“Lasix”) use were categorized as binary variables (yes/no) in our analyses.

Assessment of Covariates

Potential confounders included age, body mass index (BMI) (<22 kg/m², 22–24.9 kg/m², 25–29.9 kg/m², and ≥30 kg/m²), race (white or non-white), smoking status (never, past, current), physical activity (quintiles of metabolic equivalent task scores), self-reported history of falls, hypertension, diabetes, osteoporosis, postmenopausal hormone use, anti-hypertension medication use, anti-depressant medication use, bisphosphonate use, oral steroid use, and physical exam during the previous two years, ascertained from the questionnaires. Self-reported weight was highly reliable (r=0.97) among a subset of participants who underwent direct measurement of their weight.⁴⁵ Physical activity reported on the questionnaires has been previously validated in a similar cohort when compared with physical activity diaries (r=0.79).⁴⁶ Self-reported hypertension⁴⁷ and diabetes⁴⁸ were previously validated in this cohort. Race was self-reported and categorized in this analysis as white and non-white.

Diet was assessed by extensively validated⁴⁹ semiquantitative food-frequency questionnaires that inquired about the average intake of over 130 individual food items and 22 individual beverages as well as vitamins and supplements during the previous year. The participants were asked to complete food frequency questionnaires in 2002, 2006, and 2010. The variables considered in our models were alcohol intake (none, 0.1–4.9 g/day, 5–14.9 g/day, ≥ 15 g/day), caffeine intake, supplemental calcium intake (none, 1–500 mg/day, >500 mg/day), supplemental vitamin D intake (none, 1–400 IU/day, >400 IU/day), and quintiles of dietary intakes of calcium, vitamin D, vitamin A, protein, phosphorus, and magnesium.

Ascertainment of Vertebral Fracture

Participants were first asked about lifetime history of a diagnosis of vertebral fracture on the 2012 questionnaire. They were asked whether they had ever had a clinician-diagnosed “vertebral (spine) fracture, x-ray confirmed” and the year of first diagnosis. We mailed a supplemental questionnaire to nurses who reported a vertebral fracture in 2002 or afterwards and asked permission to obtain their medical records related to the vertebral fracture. Among the participants who gave consent to obtain their medical records and for whom we were able to obtain the medical records that contained sufficient information to make a diagnosis, we confirmed cases of vertebral fracture by radiology report (e.g. x-ray, computed tomography scan, or magnetic resonance imaging) or medical report (e.g. clinic visit note, operative note, hospital discharge summary).

A self-reported vertebral fracture was confirmed as a case if the radiology or medical report contained the word “fracture” (e.g. “vertebral fracture”, “spine fracture”, “compression fracture”, “wedge fracture”) or language to suggest a vertebral fracture (e.g. “severe wedge compression”, “vertebral collapse”, “acute compression”). Medical records that contained less definitive language for a vertebral fracture were adjudicated by one of the authors (HNR), blinded to exposure status, who is an International Society for Clinical Densitometry-trained expert on the reading and interpretation of vertebral fractures. We coded participants as “probable” cases when a diagnosis of vertebral fracture was less certain (e.g. “mild compression deformity”, “stable” or “chronic” “compression deformity”); “probable” cases were not included in our analysis.

We included vertebral fractures that were related to low or moderate trauma (e.g. tripping, slipping, falling from the height of a chair or lower). We excluded vertebral fractures due to high trauma (e.g. fall from a ladder, fall down a flight of stairs), motor vehicle accidents, bicycle accidents, or horseback riding accidents. We also excluded cases of cervical or sacral fracture. We included in the analysis only cases of vertebral fracture that were confirmed by medical record review and diagnosed during the 10 years of follow-up between 2002 and May 31, 2012. Participants with a history of hip or wrist fracture at baseline were excluded from the analysis.

Statistical Analyses

The study design was prospective; information on thiazide and loop diuretic use was collected before the diagnosis of clinical vertebral fracture. For each participant, person-time of follow-up was counted from the date on which the 2002 questionnaire was returned to 1) the date on which the vertebral fracture was diagnosed, 2) death, or 3) May 31, 2012, whichever occurred first. We allocated person-time of follow-up according to the updated exposure status at the start of each follow-up period. We used Cox proportional-hazards models to simultaneously adjust for potential confounders as listed above. All P values are two-tailed.

Results

Thiazide Use and Vertebral Fracture Risk

During 543,209 person-years of follow-up over a 10-year period, there were 420 confirmed cases of incident vertebral fracture. The characteristics of the cohort according to thiazide use in 2002 are shown in Table 1. For our analyses, however, the updated responses to thiazide use were included for each time period.

Table 1.

Age-Standardized Baseline Characteristics of Women According to Thiazide Use in 2002

	Thiazide Use
	No (n=47,520)	Yes (n=8,260)	p-value
Age, years*	66.1 (6.6)	67.5 (6.6)	< 0.001
Body Mass Index, kg/m2	26.5 (5.1)	28.8 (5.7)	< 0.001
Physical Activity (METS/week) †	18.9 (22.5)	16.6 (19.9)	< 0.001
Dietary Calcium (mg/day) #	857.3 (320.4)	850.6 (312.9)	< 0.001
Calcium Supplement (mg/day)	630.4 (527.3)	630.7 (523.7)	0.78
Calcium Supplement Use (yes/no), %	80	81
Total (Dietary and Supplemental)   Vitamin D Intake (IU/day) #	567.7 (336.8)	572.0 (330.5)	< 0.001
Phosphorus Intake (mg/day) #	1,254 (257)	1,256 (254)	< 0.01
Magnesium Intake (mg/day) #	374.0 (123.2)	373.7 (119.9)	< 0.001
Total Protein Intake (gm/day) #	69.9 (13.0)	71.3 (13.1)	< 0.001
Animal Protein Intake (gm/day) #	46.1 (13.6)	47.7 (13.5)	< 0.001
Total (Dietary and Supplemental)   Vitamin A Intake (mcg/day) #	1,931 (1,403)	1,911 (1,297)	< 0.001
Alcohol Intake (gm/day)	6.2 (10.7)	5.8 (10.4)	< 0.001
Smoking status			< 0.001
Never smoker, %	47	45
Past smoker, %	46	50
Current smoker, %	7	5
History of Falls	8	10	< 0.001
Self-Reported Diabetes, %	7	12	< 0.001
Self-Reported Hypertension, %	43	94	< 0.001
Self-Reported Osteoporosis, %	16	14	< 0.001
Postmenopausal Hormone Use, %	38	41	< 0.001
Bisphosphonate Use, %	13	10	< 0.001
Oral Steroid Use, %	2	3	< 0.01

^*Value is not age adjusted.

^†Physical activity and history of falls was not asked about in 2002 so data is from the 2000 questionnaire.

^#Energy adjusted.

In 2002, 14.8% of the participants were taking thiazides and in 2010 it was 19.4%. Those taking thiazides had higher BMI, less physical activity, and were more likely to have hypertension and diabetes. In 2002, 99.5% of the women were postmenopausal and there were relatively similar rates of postmenopausal hormone use amongst the thiazide and non-thiazide users.

After adjusting for age, thiazide use was associated with an increased risk of clinical vertebral fracture (RR 1.44, 95% CI 1.16 to 1.80) (Table 2). The multivariable-adjusted results were materially unchanged (RR 1.47, 95% CI 1.18 to 1.85).

Table 2.

Age- and Multivariable-Adjusted Relative Risks for Incident Vertebral Fracture According to Diuretic Use^*

	Thiazide Use
	No	Yes
Cases of Vertebral Fracture (n)	316	104
Person-years (n)	451,178	92,032
Age-adjusted Relative Risk (95% CI)	1.0	1.44 (1.16, 1.80)
Multivariate Relative Risk (95% CI) †	1.0	1.47 (1.18, 1.85)
	Furosemide Use
	No	Yes
Cases of Vertebral Fracture (n)	382	38
Person-years (n)	521,248	21,961
Age-adjusted Relative Risk (95% CI)	1.0	1.63 (1.16, 2.28)
Multivariate Relative Risk (95% CI) †	1.0	1.59 (1.12, 2.25)

^*Thiazide and furosemide use were updated throughout the analysis period (2002–2012). Relative risks are for the risk of vertebral fracture compared with the group that did not use diuretics.

^†The multivariate model includes: body mass index, race, physical activity, history of falls, smoking status, alcohol intake, supplemental calcium intake, quintiles of diet calcium intake, total vitamin D intake, vitamin A intake, total protein intake, self-reported diabetes or osteoporosis, history of beta-blocker use, bisphosphonate use, oral steroid use, or postmenopausal hormone use, and recent physical exam.

Loop Diuretic Use and Vertebral Fracture Risk

The characteristics of the cohort according to loop diuretic use in 2002 are shown in Table 3. For our analyses, however, the updated responses to loop diuretic use were included for each time period. In 2002, 2.8% of the participants were taking a loop diuretic and in 2010 it was 6.5%. Compared with the women not taking a loop diuretic, the women taking a loop diuretic were slightly older, had higher BMI, less physical activity, and were more likely to have diabetes and hypertension. The loop diuretic users had slightly lower supplemental calcium intake and lower alcohol intake.

Table 3.

Age-Standardized Baseline Characteristics of Women According to Furosemide Use in 2002

	Furosemide Use
	No (n=54,208)	Yes (n=1,572)	p-value
Age, years*	66.2 (6.6)	68.8 (6.7)	< 0.001
Body Mass Index, kg/m2	26.7 (5.1)	31.4 (7.3)	< 0.001
Physical Activity (METS/week) †	18.7 (22.2)	13.7 (20.8)	< 0.001
Dietary Calcium (mg/day) #	856.1 (319.1)	864.8 (328.4)	< 0.001
Calcium Supplement (mg/day)	632.4 (526.7)	569.0 (526.5)	< 0.001
Calcium Supplement Use (yes/no), %	80	76
Total (Dietary and Supplemental)   Vitamin D Intake (IU/day) #	568.6 (335.3)	561.3 (360.4)	< 0.001
Phosphorus Intake (mg/day) #	1,254 (256)	1,276 (260)	< 0.001
Magnesium Intake (mg/day) #	374.0 (122.6)	374.4 (128.6)	< 0.001
Total Protein Intake (gm/day) #	70.0 (13.0)	72.6 (13.6)	< 0.001
Animal Protein Intake (gm/day) #	46.2 (13.6)	49.3 (14.0)	< 0.001
Total (Dietary and Supplemental)   Vitamin A Intake (mcg/day) #	1,926 (1,382)	2,010 (1,634)	< 0.001
Alcohol Intake (gm/day)	6.2 (10.7)	4.3 (9.4)	< 0.001
Smoking status			< 0.001
Never smoker, %	47	43
Past smoker, %	46	51
Current smoker, %	7	6
History of Falls	8	12	< 0.001
Self-Reported Diabetes, %	8	25	< 0.001
Self-Reported Hypertension, %	50	83	< 0.001
Self-Reported Osteoporosis, %	15	16	< 0.001
Postmenopausal Hormone Use, %	38	38	< 0.001
Bisphosphonate Use, %	13	10	< 0.001
Oral Steroid Use, %	2	6	< 0.001

^*Value is not age adjusted

^†Physical activity and history of falls was not asked about in 2002 so data is from the 2000 questionnaire.

^#Energy adjusted.

After adjusting for age, loop diuretic use was associated with an increased risk of clinical vertebral fracture (RR 1.63, 95% CI 1.16 to 2.28) (Table 2) and the multivariable-adjusted results were similar (RR 1.59, 95% CI 1.12 to 2.25).

Additional Analyses

In secondary analyses, we attempted to examine the association between duration of thiazide or loop diuretic use and risk of clinical vertebral fracture, but there were too few cases with longer duration of diuretic use to draw meaningful conclusions. We adjusted for other classes of anti-hypertensive medications, including angiotensin-converting enzyme inhibitors, calcium channel blockers, and beta-blockers and the results were not substantially changed. When we restricted the analysis to participants with hypertension, the point estimates were not substantially changed. When we further adjusted for the use of antidepressants, including selective serotonin reuptake inhibitors and tricyclic antidepressants, the results were materially unchanged. There was no statistically significant interaction for either diuretic with age (p for interaction ≥0.34) or history of osteoporosis (p for interaction ≥0.17).

Discussion

In this large, prospective study of women, we observed an independent increased risk of incident clinical vertebral fracture for thiazide use and loop diuretic use. Our study has several strengths that distinguish it from other reports, including the large number of incident clinical vertebral fracture events that were confirmed by medical record review and repeated assessment of diuretic use over time.

The higher risk of vertebral fracture with thiazide use was unexpected. Our ascertainment of thiazide use appears to be reasonably accurate. A prior study of thiazide use and hip fracture risk in the Nurses’ Health Study³⁰ reported a lower risk, findings consistent with other studies in the literature.^31–35

This is the first prospective study to report an increased risk of clinical vertebral fracture with thiazide use in women. The Danish population-based case-control study,³⁶ which used national registry data for thiazide use and fracture cases, did not observe a statistically significant association between former or current thiazide use and vertebral fracture. However, after age stratification, there was an increased odds for current thiazide users among those >65 years of age (OR 1.20, 95% CI 1.04 to 1.39). Of note, the validity of fracture reports from the national registry was based on a random sample of diagnoses (n=35) and the authors did not specify how many of the 35 reports were vertebral fractures versus fractures at other sites.⁵⁰ The prospective Study of Osteoporotic Fractures³⁷ did not find a statistically significant association between thiazide use and risk of vertebral fracture. While the study was prospective and vertebral fractures were defined radiographically, there were only 181 incident vertebral fractures among the women who did not have a baseline prevalent fracture, thiazide use was ascertained only at baseline, and follow-up averaged less than 4 years.

In contrast to the consistently reported protective relation between thiazides and hip fracture,^30–35 the higher risk of vertebral fracture with thiazide use might be mediated by hyponatremia (information on serum sodium in NHS was not available), which directly affects bone mainly via activation of osteoclasts.⁵¹ With raised activity of osteoclasts, bone resorption is increased without concomitant bone formation, impairing local repair of microdamage and decreasing bone quality, thereby increasing fracture risk, all of which is not captured by bone mineral density measurements.^{52, 53} Skeletal bone is a rich reservoir of sodium⁵⁴ and may play a key role in maintaining sodium homeostasis,⁵⁵ but possibly at the expense of the bone’s structural integrity. Two recent studies, one in rats⁵⁶ and the other in cell culture,⁵⁷ support this possibility. The spine is particularly susceptible to microdamage (e.g. linear microcracks and diffuse damage)⁵⁸ compared with other fracture sites, given its different biomechanics, loading parameters, microarchitecture, composition, geometry, and structural integrity,^{52, 53, 59} and the microdamage accumulates with aging.⁶⁰ The potential harmful effect of hyponatremia on the spine’s ability to repair local microdamage may explain the increased vertebral fracture risk seen with thiazide use.

The association between thiazide use and hyponatremia is well-established^{61, 62} with an estimated 3 in 10 patients exposed to thiazides expected to develop hyponatremia.⁶³ Our study’s finding of a positive association between thiazide use and clinical vertebral fracture is consistent with two recent prospective reports demonstrating the possible role of even mild hyponatremia increasing the risk of developing both vertebral and non-vertebral fractures, independent of falls and bone mineral density.^{24, 26} The Rotterdam study²⁴ reported an association between hyponatremia and higher prevalence of vertebral fracture (OR 1.78, 95% CI 1.04 to 3.06), combined prevalent and incident vertebral fracture (OR 1.61, 95% CI 1.00 to 2.59), but not incident vertebral fracture alone.²⁴ However, there were only 8 incident vertebral fracture cases among the 136 participants with hyponatremia. More compelling data come from the prospective Osteoporotic Fractures in Men Study,²⁶ where hyponatremia was associated with an increased risk of both prevalent (OR 2.46, 95% CI 1.22 to 4.95) and incident (OR 3.53, 95% CI 1.35 to 9.19) vertebral fractures, with the results unchanged even after adjusting for bone mineral density.

Although our findings of an association between furosemide use and vertebral fracture was in the expected direction,^64,65 this is also the first prospective study to report an increased risk of clinical vertebral fracture with loop diuretic use in women. The mechanisms by which loop diuretic use increase vertebral fracture risk are likely different than the mechanisms for thiazide diuretics. Possible mechanisms include the calciuric effect of loop diuretics,³⁸ subsequent increase in plasma parathyroid hormone level,⁴⁰ and decreased bone mineral density.^{40, 41} In contrast to prior studies, we confirmed vertebral fractures by medical record review rather than by self-report or from administrative data. In the Danish population-based case-control study⁴³ that defined loop diuretic use and vertebral fractures from registry data, the association between loop diuretics and vertebral fracture in adjusted models was not statistically significant. In the prospective Women’s Health Initiative (WHI),⁴⁴ no statistically significant association was found between ever use of loop diuretics and clinical vertebral fracture. However, vertebral fracture was ascertained by self-report and a prior validation study in WHI found a relatively lower rate of agreement between self-report and medical records for clinical vertebral fracture (51%), compared with 78% for hip and 81% for forearm/wrist fractures.⁶⁶

There are several limitations to our observational study. There is the possibility of residual or unmeasured confounding. For example, we did not have data on bone mineral density or morphometric fracture assessment in our participants. Our definition of clinical vertebral fracture was based on medical record review, which means that some participants who might have had a vertebral fracture could not be confirmed. In these cases, we did not have permission or were not able to obtain their medical record, or there was insufficient evidence in the medical record to make a definitive diagnosis of vertebral fracture. We recognize that the incidence rate therefore is likely low because of our method of case ascertainment. However, this misclassification would likely be non-differential since we would not expect vertebral fracture cases who were exposed to diuretics to be identified differently from cases who were not exposed to diuretics so the findings presented are internally valid. We were unable to assess adherence to diuretic use. We were unable to examine a dose-response relation between diuretic use and vertebral fracture risk. While we postulate that hyponatremia, even mild, may be responsible for the association between thiazide use and vertebral fracture risk, we did not have information on serum sodium concentration for our participants. Finally, since our study population was female and almost entirely white, our findings are not necessarily generalizable to men or other races.

In this large prospective cohort study, thiazides and loop diuretics were independently associated with an increased risk of clinical vertebral fracture in women. Further research is warranted on diuretics and vertebral fracture risk, especially since thiazides are so commonly prescribed for the treatment of hypertension.

Clinical Significance.

• Thiazide diuretic use is independently associated with an increased risk of vertebral fracture in women.

• Loop diuretic use is independently associated with an increased risk of vertebral fracture in women.

• Among women with hypertension who are at high risk for vertebral fracture, diuretics should be used cautiously. Further research is warranted on diuretics and vertebral fracture risk, especially since thiazides are so commonly prescribed for the treatment of hypertension.