Low bone mineral density and fracture burden in postmenopausal women

Study design and patient population

In this retrospective cohort study, we examined records for 16 505 women 50 years of age or older who underwent baseline bone mineral densitometry of the lumbar spine (first through fourth lumbar vertebrae) and the proximal femur (total hip, femoral neck and trochanter) between May 1998 and October 2002. All of the women had medical coverage with Manitoba Health throughout the observation period, which ended Mar. 31, 2004, and were included in the clinical database of the Manitoba Bone Density Program.8 The database includes all test results obtained by dual-energy x-ray absorptiometry since 1990 and is more than 99% complete and accurate, as judged by a chart audit.9 All clinical bone densitometry in the province of Manitoba, Canada, is performed within a single program structure that maintains uniform testing indications and reporting. The criteria for testing are consistent with published guidelines and highlight the importance of female sex, age 65 years or older, prior fragility fractures, prolonged corticosteroid use and other clinical risk factors in the decision of whether or not to test.4 Men were not included in the study group, because of small numbers and potential selection bias. The study was approved by the Research Ethics Board of the University of Manitoba and the Health Information Privacy Committee of Manitoba Health.

Measurement of bone mineral density

The lumbar spine T-scores and Z-scores (number of standard deviations above or below age-matched mean) used the US reference values for white women provided by the manufacturer of the absorptiometer. Hip T-scores and Z-scores were based on the revised reference data from the National Health and Nutrition Examination Survey (NHANES III) (calculated using the Prodigy version 8.8 densitometer, GE Lunar, Madison, Wis.).10 Bone densitometry was performed with 1 of the program's 2 primary instruments (DPX or Prodigy, GE Lunar). The instruments were cross-calibrated in vivo with 59 volunteers, and no clinically significant differences were identified (all T-score differences < 0.2). Therefore, all analyses are based on the unadjusted numeric results generated by the instrument. For women with more than one densitometry record, the first measurement that included information for both spine and hip was used. Densitometers underwent daily assessment of stability with a spine phantom, and each demonstrated stable long-term performance (coefficient of variation [standard deviation/mean] < 0.5%).

Ascertainment of incident fractures

Physician claims processed by Manitoba Health and hospital databases record all health care utilization in the province. From these sources, we extracted information about patients' demographic characteristics, date and type of service provided, and diagnoses, coded according to the International Classification of Diseases, 9th revision (Clinical Modification) (ICD-9-CM). Anonymous linkage of these databases to the bone mineral density database was accomplished by means of a unique scrambled health identification number and was successful in 99.4% of cases.9 Each subject's longitudinal health record was assessed from the date of bone density measurement to Mar. 31, 2004, for the presence of the following ICD-9-CM fracture codes: hip (ICD-9-CM 820–821), clinical spine (ICD-9-CM 805, defined below), forearm (ICD-9-CM 813) or proximal humerus (ICD-9-CM 812). We excluded traumatic fractures (ICD-9-CM E800–E879 and E890–E999). Nontraumatic fractures of the hip, clinical spine, forearm and proximal humerus were collectively designated as “osteoporotic fractures” and were included in our analysis. Clinical spine fracture refers to a fracture identified when a patient presents with back pain, and spinal radiography confirms fracture of the vertebral body. The accuracy of the data from Manitoba Health has previously been established for a number of clinical disorders, including osteoporotic fractures.11 Our use of administrative health data to define fractures in this way showed that the relation between bone mineral density and fractures in our clinical cohort was similar to that reported in a large meta-analysis.12

Analysis

Bone mineral density values were categorized as normal, osteopenic or osteoporotic according to the system of the World Health Organization;3 T-scores were also grouped in increments of 0.5 across a wide range, from above 0 to –3.5 and below. The percentage of women with fractures and the fracture incidence rates per 1000 person-years (with 95% confidence intervals [CIs]) were derived for the various T-score categories for each region of interest and for the minimum T-score (the lowest T-score of the 4 skeletal sites). The results were analyzed for the total sample and were then stratified into 2 age groups, 50–64 years and 65 years or older. Excess fracture risk was examined by site of bone mineral density testing and was defined as the fracture rate for those with reduced bone mineral density (osteopenic or osteoporotic T-score) minus the fracture rate for those with normal bone mineral density.13 Similarly, odds ratios (OR) for fracture were derived for those with reduced bone mineral density (osteopenic or osteoporotic T-score) relative to those with normal values. Continuous data are expressed as mean (and SD) unless otherwise stated. A p value of less than 0.05 was used to determine statistical significance.

Population characteristics

The mean age of the 16 505 women in the cohort was 65 (SD 9) years (Table 1). Forty-seven percent (n = 7737) of the women were 65 years of age or older. The majority of the women were white (98.2%), with small percentages of other ethnic groups (1.3% Asian and 0.3% black). Mean weight was 67.9 (SD 13.8) kg, and mean height was 160.4 (SD 6.7) cm. The mean T-scores for all measurement sites were within the osteopenic interval and were significantly lower among women 65 years of age or older than among women 50–64 years of age (p < 0.001). Z-scores for all skeletal sites were close to zero (range –0.1 to +0.1), which indicates that the average bone mineral density was close to the manufacturer's age-matched reference.

Outcomes

Over the mean follow-up period of 3.2 (SD 1.5) years, a total of 765 (4.6%) women experienced a fracture, which corresponds to an overall fracture rate of 14.5 (95% CI 13.5–15.5) per 1000 person-years. Of these 765 osteoporotic fractures, 520 (68%) occurred in women 65 years of age or older. Although fracture rates in both age groups were highest among women with osteoporosis, most fractures occurred in women with normal bone mineral density or osteopenia. Figure 1 illustrates the continuous relation between higher fracture rates and lower bone mineral density, the highest fracture rates being associated with the lowest T-scores. Depending on the measurement site used, 59.7% to 67.8% of fractures that occurred after assessment of bone mineral density involved women who had nonosteoporotic values. The percentage of fractures that occurred in association with nonosteoporotic T-scores was greater among women 50–64 years of age (75.5%–86.5%) than among women over 65 years of age (50.6%–59.0%). When we used the threshold T-score for intervention of the National Osteoporosis Foundation (–2.0), the percentage of osteoporotic fractures occurring among those with T-scores above the threshold was 40% to 50%.5

• Download figure
• Open in new tab
• Download powerpoint

Figure 1: Number of fractures and fracture rate (per 1000 person-years, with 95% confidence intervals) by lumbar spine (A) and total hip (B) bone mineral density T-score and World Health Organization category.

We used the lumbar spine results to define World Health Organization categories for the overall sample and found that 22% of the fractures (8.2 per 1000 person-years) occurred in women with normal bone mineral density, 38% (14.3 per 1000 person-years) in women with osteopenia and 40% (26.2 per 1000 person-years) in women with osteoporosis. Fracture incidence rates in women with osteoporotic bone mineral density at different hip sites ranged from 33.4 (95% CI 29.6–37.2) to 40.6 (95% CI 35.6–45.7) per 1000 person-years (Figure 1, Table 2), compared with 26.2 (95% CI 23.2–29.1) per 1000 person-years for those with osteoporotic values for the lumbar spine. Excess fracture rates for osteoporotic or osteopenic bone mineral density values relative to normal values were much higher at the total hip than at the lumbar spine (34.0 [95% CI 28.8–39.2] v. 18.0 [95% CI 14.9–21.2) per 1000 person-years) (Table 3).

Overall fracture rates were significantly higher among women at least 65 years of age than among younger women (21.6 [95% CI 19.7–23.4 v. 8.6 [95% CI 7.5–9.7] per 1000 person-years) (Table 2). Even within each World Health Organization category, older women had significantly greater fracture rates (Table 2). Excess fracture rates among those with osteoporotic values (relative to those with normal values) were higher for women over 65 years of age (min–max: 19.4–37.1 per 1000) than for those 50–64 years of age (min–max: 9.1–17.0 per 1000 person-years) (Table 3). Overall odds ratios for fracture among women with osteoporotic values (relative to those with normal values) were significantly greater (3.52 to 6.85 per 1000 person-years) than those for women with osteopenic values (1.83 to 2.59 per 1000 person-years).