Warfarin Use and Risk for Osteoporosis in Elderly Women

Methods

Study participants were ambulatory women 65 years of age and older who were enrolled in the Study of Osteoporotic Fractures and were recruited through population-based listings from four U.S. cities [14]. Our analyses were limited to 6201 women who participated in the fourth clinic visit (June 1992 to December 1994), during which a detailed history of medication use was first obtained, and to those who had technically adequate results on hip and heel densitometry. A subset of these women (n = 5122) also had technically adequate measurements of hip bone mineral density at the fifth clinic visit 2 years later.

Women were asked to bring medications to the clinic visit for an inventory that ascertained type of medication and daily dose. Hip bone density was measured by dual x-ray absorptiometry with Hologic QDR-1000 scanners (Waltham, Massachusetts), and calcaneal density was assessed by doing single-photon absorptiometry with Osteon Osteoanalyzers (Wahiawa, Hawaii). The mean coefficient of variation between centers was 0.9% for the hip and 1.2% for the calcaneus [15]. Every 4 months, women completed a survey about fractures. We excluded fractures that resulted from severe trauma, such as motor vehicle accidents, and self-reported vertebral fractures. All fractures were confirmed by review of radiographic reports. Fractures that occurred after the fourth visit through June 1997 were included in the analyses. Information on potential confounders was collected on questionnaires [14].

Differences in baseline characteristics between warfarin users and nonusers were assessed by using the chi-square test for dichotomous variables and the Student t-test for continuous variables. We used a general linear model to test whether warfarin use was associated with bone mineral density and a proportional hazards model to estimate age-adjusted fracture-free survival by warfarin use and to determine the relative hazard (with 95% CI) associated with warfarin use.

We adjusted for age, weight, estrogen use, clinic site, and characteristics for which warfarin users and nonusers differed significantly (P ≤ 0.1). These characteristics were self-rated health status (excellent or good, fair, and poor or very poor), involuntary loss of more than 2 kg of body weight since the last clinic visit, use of nonthiazide diuretics, frailty (weak, unsteady, or fragile) on physical examination, and inability to rise from a chair independently. We repeated our analyses, limiting warfarin users to those who had used warfarin for at least 2 years (at visits 4 and 5). We also compared the mean rate of bone loss at the total hip at visit 4 in women using warfarin with that in nonusers. Results are expressed as differences between warfarin users and nonusers, both as absolute amounts and as percentages (100% x mean difference in bone mineral density/mean bone mineral density in nonusers).

Analyses were performed with STATA, version 5.0 (College Station, Texas), and Statistical Analyses Software (Cary, North Carolina). A two-sided P value of 0.05 or less was considered significant for all statistical models.

This study was funded by the National Institutes of Health, which had no influence on the study or its publication.

Results

In our study, 6201 women had densitometry of the hip and 5999 women had densitometry of the heel. During an average of 3.5 years of follow-up, 576 nontraumatic, nonvertebral fractures occurred (in 15 of 149 warfarin users [10%] and 561 of 6052 nonusers [9.3%]). Of the 149 (2.4%) women who were using warfarin at visit 4, 102 were still using warfarin at visit 5, about 2 years later. The weekly dose of warfarin at visit 4 ranged from 2.5 mg to 70 mg (mean dose, 24 mg).

Compared with nonusers, more warfarin users had poor self-rated health (1.5% and 4.0%; P = 0.01), involuntary weight loss (19% and 32%; P < 0.001), use of nonthiazide diuretics (6.9% and 33%; P < 0.001), and frail appearance (15% and 23%; P = 0.005). Nonusers and users did not differ in age (76 years and 77 years; P = 0.17), weight (66 kg and 65 kg; P = 0.18), use of estrogen therapy (17% and 15%; P > 0.2), consumption of 1 or more alcoholic drinks in the past 30 days (46% and 41%; P > 0.2), current cigarette smoking (5.6% and 5.0%; P > 0.2), physical activity (5468.4 kJ/wk [1302 kcal/wk] and 4872.0 kJ/wk [1160 kcal/wk]; P > 0.2), women reporting a fall in the past 12 months (30% and 31%; P > 0.2), women reporting a fracture since age 50 years (44% and 40%; P > 0.2), use of thiazide diuretics (18% and 22%; P = 0.19), use of calcium supplements (45% and 38%; P = 0.12), or ability to rise from a chair independently (10% and 13%; P > 0.2).

Warfarin users and nonusers had similar age-adjusted bone density measurements at the heel and hip. Adjusting for additional confounders and limiting analyses to women who had used warfarin for at least 2 years had little effect (Table 1). Warfarin users and nonusers did not differ for age-adjusted fracture rates (relative hazard, 1.2 [95% CI, 0.7 to 1.9]) (Figure 1). Limiting our analyses to women who had used warfarin for at least 2 years had little effect (relative hazard, 1.3 [CI, 0.7 to 2.3]). After adjusting for additional confounders, the risk for fracture was similar among women who were using warfarin at the initial assessment (relative hazard, 1.0 [CI, 0.60 to 1.7]) and women who had used warfarin for at least 2 years (relative hazard, 1.1 [CI, 0.61 to 2.0]) compared with nonusers.

View larger version:

• In this window
• In a new window

Figure 1. The thin line represents patients who received warfarin; the thick line represents patients who did not receive warfarin. Fracture-free survival by warfarin use, adjusted for age.

The annual rate of age-adjusted bone loss at the hip was similar in warfarin users (1.1% [CI, 0.7% to 1.4%]) and nonusers (0.8% [CI, 0.7% to 0.9%]); the mean difference in rates of loss was 0.3% ([CI, −0.1% to 0.6%]; P = 0.18). Adjusting for additional confounders did not change these results.

Discussion

Regular use of warfarin was not associated with less initial bone mass, greater bone loss, or increased risk for fracture. These findings suggest that the vitamin K deficiency induced by warfarin administration is not associated with increased risk for osteoporosis.

Five previous studies suggested that warfarin use may be associated with lower bone density [6, 7, 9, 10, 13]. Three other studies showed no effect [8, 11, 12]. All of these studies were limited by small sample sizes and referral bias. One of the studies was prospective [9], but this study had a short follow-up period and no control group. Our large study had the power to detect even small effects of warfarin on bone mass and fracture risk. Warfarin users and nonusers came from the same community-based sample, eliminating referral bias and broadening the applicability of our findings. Finally, we used a prospective design and a long follow-up period, which allowed us to assess clinically important end points without recall bias.

Vitamin K deficiency has been associated with increased risk for osteoporotic fractures [1-5]. One explanation for this observation is that low levels of vitamin K are indicative of poor nutritional status and malnutrition is a risk factor for osteoporosis [16]. Another explanation is that vitamin K deficiency influences osteoporosis by its effects on osteocalcin, a noncollagenous bone protein. Osteocalcin must be γ-carboxylated, a vitamin K-dependent process, to be incorporated into bone [17]. In bone, γ-carboxylated osteocalcin binds to hydroxyapatite and may be important for mineralization [17]. Under-γ-carboxylated osteocalcin in serum has been associated with decreased bone mineral density and increased risk for fracture [18, 19].

We found that warfarin use, which greatly increases under-γ-carboxylated osteocalcin [18, 20], did not lead to increased risk for fracture. This does not support the hypothesis that elevated under-γ-carboxylated osteocalcin levels increase susceptibility to fracture. Our results suggest that endogenous vitamin K deficiency is a marker of malnutrition or that elevated under-γ-carboxylated osteocalcin levels result from intrinsic problems in bone matrix.

Our study has limitations. We studied white, ambulatory women older than 65 years of age. Therefore, our findings may not apply to men or to women who are younger, are from other ethnic groups, or are institutionalized. Information was not available about duration of warfarin use, although we assessed the effect of warfarin use for at least 2 years by performing our analyses in women who were using this medication at both visit 4 and visit 5. Indices of vitamin K status, such as undercarboxylated osteocalcin levels and international normalized ratios, were not available. However, other studies have confirmed that warfarin causes a functional deficiency of vitamin K [20], and there is no reason to suspect that this effect would be different in our population. Finally, selection bias may have occurred if women who use warfarin and are osteoporotic were less likely to be included in this study.

We show that warfarin, the most commonly prescribed oral anticoagulant in postmenopausal women, has no substantial adverse effects on skeletal health. Postmenopausal women who are receiving warfarin do not need increased surveillance or treatment for osteoporosis.

Dr. Browner: Suite 602, 74 New Montgomery Street, San Francisco, CA 94105.

Dr. Bauer: Suite 616, 74 New Montgomery Street, San Francisco, CA 94105.

Dr. Cummings: Suite 610, 74 New Montgomery Street, San Francisco, CA 94105.