Dr Jaglal served as a consultant for Theratotechnology Inc. Dr Adachi served as a consultant for Astra-Zeneca, Aventis, Eli Lilly and Company, Merck, Novartis, and Procter & Gamble. All other authors have no conflict of interest.
Department of Orthopaedics, Sunnybrook and Women's College Health Science Centre, Toronto, Canada
Department of Public Health Sciences, University of Toronto, Toronto, Canada
A worldwide epidemic of hip fractures has been predicted. Time trends in BMD testing, bone-sparing medications and hip and wrist fractures in the province of Ontario, Canada, were examined. From 1996 to 2001, BMD testing and use of bone-sparing medications increased each year, whereas despite the aging of the population, wrist and hip fracture rates decreased.
Introduction: If patients with osteoporosis are being diagnosed and effective treatments used with increasing frequency in the population, rates of hip and wrist fractures will remain stable or possibly decrease. We report here time trends in BMD testing, prescriptions for bone-sparing medications, hip and wrist fracture rates, and population projections of fracture rates to 2005 in the province of Ontario, Canada.
Materials and Methods: Ontario residents have universal access to Medicare. To examine time trends in BMD testing, all physician claims for DXA from 1992 to 2001 were selected from the Ontario Health Insurance Plan (OHIP) database. Trends in prescribing were examined from 1996 to 2003 using data from the Ontario Drug Benefit plan, which provides coverage to persons ≥65 years of age. Actual numbers of hip and wrist fractures were determined for 1992-2000 and population projections for 2001-2005 using time-series analysis. Wrist fractures were identified in the OHIP database and hip fractures through hospital discharge abstracts.
Results: From 1992 to 2001, the number of BMD tests increased 10-fold. There has been a steady increase in the number of persons filling prescriptions for antiresorptives (12,298 in 1996 to 225,580 in 2003) and the majority were for etidronate. For women, the rate of decline for wrist fractures is greater than that for hip fractures. The rate of hip fracture was fairly constant around 41 per 10,000 women ≥50 years between 1992 and 1996. In 1997, the hip fracture rate began to decrease, and the population projections suggest that this downward trend will continue to a rate of 33.1 per 10,000 in 2005.
Conclusions: Our findings suggest that fracture rates may be on the decline, despite the aging of the population, because of increased patterns of diagnosis and treatment for osteoporosis.
The primary aim of any intervention in osteoporosis is to reduce fracture. Of all the osteoporotic fractures, hip fractures cause an enormous burden to the health care system, patients, and their caregivers and are associated with more deaths, disability, and medical costs. In recent decades, age-specific incidence rates have increased substantially in most Western populations and now seem to have stabilized.(1–7) The aging of the population, however, has led many to predict a worldwide epidemic of hip fractures because of the increasing number of elderly people in the population who are at risk for fracture.(8–11) The incidence of hip fracture increases exponentially after age 70, and in Canada, population projections suggested that the numbers would double by 2010 and quadruple by 2040.(10,11)
Prevention of osteoporotic fractures depends on the identification of individuals at risk for fractures, followed by interventions to reduce this risk such as modification of lifestyle factors and use of bone-sparing medications. BMD testing with DXA is the gold standard for identifying individuals at risk, because BMD is the best quantifiable predictor of osteoporotic fracture.(12) A DXA test may help clarify the benefit-to-risk ratio for those women who are uncertain about initiating drug therapy. Several guidelines recommend the following preventative lifestyle measures: adequate amounts of calcium and vitamin D, regular weight-bearing exercise, avoid tobacco smoking, and keep alcohol intake moderate.(12–15) Treatment for those with osteoporosis typically involves use of bone-sparing medications and lifestyle changes. In recent years, there have been significant advances in the management of postmenopausal osteoporosis with several efficacious treatment options now being available. These include hormonal therapy (e.g., estrogen), bisphosphonates, raloxifene, and calcitonin in conjunction with calcium and vitamin D. Randomized trials show significant reduction in vertebral fractures as early as 6 months to 1 year after initiation of the newer bisphosphonates (alendronate or risedronate).(16–18) In women with existing vertebral fractures or diagnosed with osteoporosis, significant risk reduction has also been shown with raloxifene.(19)
Attention to osteoporosis and related fractures by health care providers and their patients have increased dramatically in recent years. This may be due to the increased availability of bone densitometry and treatment options and public awareness of osteoporosis as a health issue. For example, the role of primary care physicians in managing osteoporosis has increased considerably such that the proportion of BMD tests ordered by family physicians increased from 47% in 1992 to 80% in 1998, indicating that the management of osteoporosis is now part of mainstream medical practice.(20) Also, in terms of patient awareness, a review of eight women's magazines and two newspapers from 1998 to 2001 found 132 articles on osteoporosis.(21) We previously reported that the number of BMD tests performed in Ontario women increased ∼6-fold from 1992 to 1998.(20) However, whether this has translated into increased use of therapies to prevent osteoporotic fractures is unknown. For many women, decisions about drug therapy involve many considerations. If recommendations regarding treatment and lifestyle changes based on BMD test results are not followed or adhered to, fracture incidence will not be affected.
Several studies have examined time trends and incidence rates of hip fractures.(8–11,22–27) The main goal of these studies was to characterize geographic and temporal variations in fracture incidence by age and sex. None of these studies examined trends in BMD testing or medication use in conjunction with fracture rates, and only one(23) included hip fracture projections. We report here an ecologic study of time trends in BMD testing from 1992 to 2001, prescriptions for bone-sparing medications from 1996 to 2003, hip and wrist fracture rates from 1992 to 2000, and population projections of fracture rates to 2005 for men and women residing in the province of Ontario, Canada. We hypothesize that if patients with osteoporosis are being diagnosed and effective treatments are being used with increasing frequency in the population, rates of hip and wrist fractures will remain stable or possibly decrease.
MATERIALS AND METHODS
The province of Ontario has a population of ∼11 million residents and represents 40% of the population of Canada. The three primary data sources used for this study were the Ontario Health Insurance Plan (OHIP) claims database and the Canadian Institute for Health Information (CIHI) hospital discharge abstract database for the period of January 1, 1992 to December 31, 2001 and the Ontario Drug Benefit (ODB) plan database for the period of January 1, 1996 to December 31, 2003, in addition to census population estimates for 1992-2001. Citizens of the province of Ontario have universal access to publicly funded medicare. The OHIP database contains all physician fee-for-service billings or claims in the province.(28) The claims submitted to the database each include a unique identifying number or health card number of the patient, date of the claim, physician code, and fee code for service provided. Unlike the OHIP database, which provides coverage to all Ontarians, the ODB plan provides coverage for medications to all persons ≥65 years of age, residents of homes for special care, and those <65 years of age receiving family benefits, general welfare assistance, extended health care benefits, or home care benefits.(29) The prescription claims submitted to the ODB program each include the unique identifying number of the beneficiary, the type of drug dispensed, the date the prescription was filled, the quantity of the drug dispensed, and the total cost of the prescription. The third database, the CIHI hospital discharge abstracts, provides information on all hospitalizations in the province, including admission date, most responsible diagnosis (defined as the diagnosis most responsible for admission), secondary diagnoses, procedures performed during the inpatient stay, whether the patient had been transferred to another acute care facility, age, sex, length of hospital stay, discharge destination, and in-hospital mortality. These data have been validated and have been shown to be of high quality.(30)
To examine time trends in BMD testing, all physician claims for BMD measurement by DXA from January 1, 1992 to December 31, 2001 were selected from the OHIP database. To obtain the age and sex of the patient for each test ordered, the OHIP file was linked using a unique identifier to the Registered Persons Database that contains these demographic data. Only BMD tests for those ≥50 years at the time of the claim were retained for analysis. The proportion of men and women who received the test by 5-year age groups from 1992 to 2001 was also calculated.
Data for treatment with antiresorptives are reported for the period 1996-2003 because etidronate, the only antiresorptive medication fully covered by the ODB plan, was first approved for use in 1996. The newer preparations, including second- and third-generation bisphosphonates (alendronate, risedronate) and selective estrogen receptor modulators (raloxifene), are available under limited use criteria, that is, persons are only covered if they fail to respond to etidronate (continued loss of BMD of >3% after 2 years of etidronate therapy), experience a new osteoporosis-related fracture after 1 year of etidronate therapy, or experience intractable side effects with etidronate or documented allergy that precludes continuation with therapy. Alendronate was approved under limited use criteria in 1996, raloxifene in 1999, and risedronate in 2000. Fluoride and salmon calcitonin are also covered, but their use is limited in Ontario and was therefore not included in the analysis. Estrogen is also fully covered, but in the ODB database, information on indication is not provided; therefore, it was not possible to determine if estrogen was prescribed for bone health, leading to misclassification. As a result, estrogen is not included in the analyses. Also, calcium and vitamin D are not covered by the ODB. Trends in number of users of antiresorptive medications and the percent of use of etidronate from 1996 to 2003 among men and women are presented.
The CIHI database provided information on all hospitalizations for hip fractures; the OHIP database provided information on all physician claims for treatment of wrist fractures. Hip fracture was defined according to the International Classification of Diseases (ICD-9) category 820 and includes transcervical fractures (820.0, 820.1), pertrochanteric fractures (820.2, 820.3), and fracture of an unspecified part of the neck of the femur (820.8, 820.9). Only those records with hip fracture as the most responsible diagnosis for the hospital stay were included. Excluded were records of hip fracture for patients <50 years of age, non-Ontario residents, invalid health care numbers, records with missing sex codes, neoplasm diagnosis (ICD-9 codes 140.0-239.9) in any diagnosis field, and duplicate records, indicating a transfer for the same episode of hip fracture. Unlike vertebral fractures, there are specific billing codes to identify wrist fractures in the OHIP database that specify type of treatment: F028 (closed reduction of wrist), F027 (no reduction of the wrist), and F030 (open reduction of wrist). Excluded were records of wrist fractures for patients <50 years of age and duplicate records on the same day. Census population estimates supplied by Statistics Canada were used in computing age-adjusted rates. Estimates were obtained for each calendar year from 1992 to 2000. Data were stratified by 10-year age groups, beginning with 50 years and ending with 90+ years by sex. Age standardization was performed by the direct method. Actual numbers of hip and wrist fractures were determined for 1992-2000 and population projections for 2001-2005.
Fracture projections analysis
Time-series analysis was used to make projections based on current trends.(31) Observations having a temporal sequence are often autocorrelated (i.e., the value at time x is affected by the value at time x − 1). As a result, the error terms are not independent, making simple regression inappropriate for their analysis. Time-series analysis, a collection of techniques for modeling autocorrelation in temporally sequenced data, was conducted using exponential smoothing models and autoregressive, integrated, moving average (ARIMA) models to model quarterly data from January 1, 1992 through December 31, 2001. These data were used to form the basis of quarterly use forecasts from January 1, 2002 through December 31, 2005. To adequately assess trends over time, data were examined on a quarterly basis, and the results were aggregated to annual estimates. Forecasts were determined using naturally occurring patterns and models that most optimally fit the available data. All projections take into consideration the natural, stable increase in population over time.
The smoothing models are largely exponential in nature and were derived from established algorithms. Numerous smoothing models were fit to the data using the SAS Econometric Time Series system software package for Windows, version 6.11 (SAS Institute, Cary, NC, USA). Both linear and nonlinear models were considered. The selection of a particular smoothing model is based on an iterative process that is computer-driven according to a set of criterion that optimizes model fit. To optimize model selection, several steps were taken. The α level, trend, damping, and seasonality smoothing weights were set to optimize the fit of the model to the data for exponential smoothing models. The autocorrelation, partial autocorrelation, and inverse autocorrelation functions were assessed for model parameter appropriateness and seasonality. When appropriate exponential smoothing models could not be established, ARIMA models were constructed and guided by autocorrelation functions for initial parameter estimation. Standard tests for time-series analyses were conducted. Stationarity was assessed using the autocorrelation function and the augmented Dickey-Fuller test.(32) The presence of white noise was assessed by examining the autocorrelations using the Ljung-Box χ2 statistic.(33)
Overall, time trends in BMD testing are similar for men and women, but the numbers of tests done in women are about 10 times greater (Table 1). From 1992 to 2001, the number of physician billings for BMD tests in Ontario increased 10-fold among men and women (34,611 in 1992 to 375,712 in 2001 among women compared with 2196 to 28,407 among men). Each year the numbers of BMD tests have increased in both men and women, with the largest increase occurring between 1997 and 1998 (36% in men; 29% in women). Figure 1 shows the proportion of women receiving a BMD test by 5-year age groups. Within each age group, the proportion of women who received a BMD test increased in each of the years from 1992 to 2001. Across the age groups, women 55-69 years of age were the group most likely to have a BMD test. In the 55-69-year-old age group, ∼20 percent of women had a BMD test in 2001 compared with 2% in 1992.
Table Table 1.. Number of Billings for BMD Testing Among Males and Females in Ontario 1992 to 2001
Table 2 displays the total number of men and women >65 years of age who filled prescriptions for antiresorptive bone-sparing medications under the government sponsored plan and the percentage that was for etidronate. Overall, about 10 times more women filled prescriptions for antiresorptive medications than men from 1996 to 2003. From 1996 to 2003, there has been a steady increase in the number of persons filling a prescription for an antiresorptive medication. Almost all of these prescriptions up until the year 2000 were for etidronate. From 2001 onward, there has been a steady increase in the proportion of persons filling prescriptions for the newer bisphosphonates, alendronate and risedronate, and for raloxifene. In 2003, 60.6%, 20.1%, 16.1%, and 3.2% of prescriptions filled for antiresorptives among women were for etidronate, alendronate, risedronate, and raloxifene, respectively, whereas for men, the percentage of antiresorptive medications filled were 69.1% for etidronate, 17.5% for alendronate, and 13.4% for risedronate.
Table Table 2.. Number of Males and Females ≥65 Years of Age Filling Prescriptions for Antiresorptive Medications and the Percent of These for Etidronate From 1996 to 2003
When examined as the proportion of the female population ≥65 years of age filling prescriptions for osteoporosis medications, it increased steadily from 1.5% in 1996 to 23.3% in 2003.
Table 3 shows the number and Fig. 2 shows the age-adjusted rate of hip and wrist fractures in men and women ≥50 years of age in the province of Ontario from 1992 to 2000 and population projections to 2005. Overall, in Ontario, the number and rate of wrist fractures consistently exceeds the number of hip fractures at all time periods and in both sexes studied. In the year 2000, there were 14,417 wrist and hip fractures in women and 4557 in men. The rate of hip fracture was fairly constant at ∼41 per 10,000 women ≥50 years of age between 1992 and 1996. In 1997, the hip fracture rate began to decrease, and this downward trend is reflected in the population projection of a rate of 33.1 per 10,000 in 2005 for women. In terms of actual numbers of hip fractures in women, the number has remained fairly constant at ∼6500 per year and is not expected to increase substantially, even though there will be larger numbers of elderly in the future because of the aging population. For wrist fractures, there were more fluctuations in the rate between 1992 and 1996, but from 1997 to 2000, the rate decreased compared with the earlier period. Overall, there is a 1:3 male to female ratio in the number of fractures, and in men, the decline in incidence for hip or wrist fractures is much less pronounced than in women. Population projections suggest that the rate of wrist fractures will also decrease, whereas the number of wrist fractures remains fairly constant, similar to the pattern seen for hip fractures. However, among women, the rate of decline for wrist fractures is greater than that seen for hip fractures but not so for men.
Table Table 3.. Number of Hip Fractures and Wrist Fractures in Men and Women ≥50 Years of Age by Year in Ontario, Canada, From 1992 to 2000 and Population Projections to 2005
When trends in BMD testing, treatment, and fracture rates are considered simultaneously, from our cross-sectional analyses, the data suggest that the increases in BMD testing and use of bone-sparing medications between 1997 and 2003 is correlated with a decrease in the rates of wrist and hip fracture. Based on our population projections, this decrease is expected to continue. From the trends reported, we cannot definitively conclude that increases in BMD testing and use of bone-sparing medications have resulted in decreased fracture rates because we did not follow an inception cohort of patients that linked BMD testing to treatment and fracture rates at an individual level.
However, there are several factors indicating that these are not spurious findings. First, the rate of hip and wrist fractures remained constant between 1992 and 1996 during a time period when BMD testing rates in Ontario were much lower. We previously reported an exponential increase in the number of BMD tests ordered between 1992 and 1998, with the greatest increases between 1996 and 1998.(20) We also reported that in the early 1990s, specialists were mostly responsible for ordering BMD tests, but by the late 1990s, >80% of BMD tests were being ordered by primary care physicians. We could hypothesize that the higher rates of diagnosis for osteoporosis and the shift from specialist to primary care in the late 1990s led to a greater number of women with osteoporosis being detected and treated, which coincides with the reduction in fracture rates.
There are a number of possible factors contributing to the increasing use of BMD testing. These include increased awareness of osteoporosis among primary care physicians and the public, release of clinical practice guidelines for osteoporosis in 1996, and availability of densitometry and therapeutic options for treating low BMD.(14) Perhaps the most influential of these factors has been the introduction of the bisphosphonate family of medications to the marketplace in 1995 as an alternative to hormone replacement therapy, which coincides with a sharp jump in BMD testing rates after 1995.(20)
A BMD test may help clarify the benefit-to-risk ratio for those women who are uncertain about initiating drug therapy. A number of studies have found that knowing the result of a BMD test influences drug prescribing and also lifestyle changes.(34–41) In these studies, women were between 3 and 16 times more likely to start or receive a prescription for hormone replacement therapy if the BMD report classified the patient as having low bone mass or osteoporosis. Our data on drug use indicate an increase in the use of bone-sparing medications in the elderly female population ⩾65 years of age from 4.3% in 1997 to 18.5% in 2001. This increase in use of medications coincides with increased BMD testing rates in this age group. Therefore, it stands to reason the increases in BMD testing and treatment will result in decreased fracture rates. Recent trials have shown that hip fractures may be reduced with therapy.(41,42) Therapy significantly reduces the risk of hip fracture among women with confirmed osteoporosis.
In 1981, there were 6872 hip fractures in Ontario.(11) This number increased steadily up until 1996 and now has stabilized, thereby negating previous projections.(10,11) This has resulted in a reduction in the rate of hip fractures, because the elderly population has increased over that time period. We also showed that wrist fracture rates are on the decline. This begs the question of whether the treated part of the population is large enough to explain the reduction in fractures. We note that the proportion of the female population ⩾65 years of age filling prescriptions for osteoporosis medications increased 5-fold from 4.3% in 1997 to 23.3% in 2003, whereas at the same time, the fracture rate began to decrease. This might suggest that the reduction in fractures may be due in part to the increase in the proportion of the population being treated with antiresorptive medications. Data from the Canadian Multicentre Osteoporosis Study estimate that the prevalence of osteoporosis of the femoral neck using DXA manufacturer standards ranges from ∼30% for women 60-69 years of age to 50% for those 70-79 years of age and 70% for those ⩾80 years of age.(43) Therefore, if approximately one-quarter of women over age 65 were being treated, this would represent about one-half of those with osteoporosis. This is most likely an overestimate, because our data do not examine adherence to treatment; we only know if at least one prescription was filled. On the other hand, Chang et al.,(22) in a study examining secular changes in the incidence rates of hip and other osteoporotic fractures in the Dubbo Osteoporosis Epidemiology Study, reported a significant downward trend in total osteoporotic fracture incidence over three successive 4-year birth cohorts for the period 1989-2000 in both men and women. However, there was no significant downward trend in hip fracture incidence in either men or women. Similar to Chang et al., our data indicate no downward trend in the rate of hip fractures in men from 1992 to 2000. For women, our data indicate a decrease over the time period. In the Dubbo study, only 5% of all participants were on antifracture medication at the end of 2000 compared with ∼18% in our study, and this may explain why we see a decrease.
Two recent articles summarizing data on fracture risk concluded that the presence of pre-existing, osteoporosis-related fractures is associated with a marked increase in future fractures of two to four times.(44,45) Prior fracture is one of the single most important predictors of future fracture.(12) Our findings are particularly important because the occurrence of a wrist fracture is a strong predictor of future fracture risk, and this is where we saw a decreasing incidence, particularly among women. Several cohort studies have shown that women who experience a wrist fracture have almost double the expected risk of future hip fracture.(45–48) The risk of hip fracture after a wrist fracture was increased 1.4-fold in women in Rochester, MN,(46) 1.5-fold in Swedish women,(47) and 1.8-fold in Danish women.(48) Similarly, the risk of hip fracture after a wrist fracture was also increased in men; Swedish men had a 2.7-fold increase and Rochester men a 2.3-fold increase. Therefore, if our results of declining wrist fracture rates continue into the future, we should expect to see even greater reductions in the number of hip fractures. Indeed, of those sustaining an incident hip fracture, 10% will have yet another clinical fracture in the subsequent year.(49,50) Pharmacologic or other interventions in the first critical year after a hip fracture may potentially blunt this accelerated rate of bone loss and lessen the risk of subsequent fractures. Our results are intriguing and need to be confirmed in other populations.
There are strengths and limitations to our study. The main strength is that this is a population-based study in a province of ∼11 million inhabitants. As a result, there is no selection bias because the findings reflect the impact of all practice patterns. The main limitation of this study is that we were only able to examine trends and do not have sufficiently detailed data to examine the impact of BMD testing or drug use on fracture rates at the individual level. The information on BMD testing does not include the result of the test nor the reason for the test. For example, we have no information on what proportion of women had normal BMD or had osteopenia or osteoporosis and how this may have influenced physicians and patients or whether the test was ordered to diagnose osteoporosis or to monitor treatment. Similarly, with respect to our information on drug use, general preventive measures for elderly women with low BMD include appropriate calcium intake or supplementation, adequate vitamin D intake or supplementation, and weight-bearing exercise.(49) Calcium and vitamin D supplementation are not covered by the ODB. Also, we do not have drug use data on women <65 years of age because they do not have government-sponsored drug coverage and we did not receive information from private drug plans. Therefore, we cannot extrapolate our findings on medication use to those <65 years of age. Another concern is the lack of drug data from private insurance plans. This limitation would result in an underestimate of prescriptions filled for bone-sparing medications in those ⩾65 years of age if these persons were obtaining prescriptions from private drug plans. Because the majority of those over the age of 65 years would be retired and most private insurance plans are an employee benefit, it is unlikely that private insurance plans would account for much use in this age group.
A more important limitation is that we only have information on whether prescriptions were filled but not whether they took the drug or what was the percentage of adherence. There is also the possibility that not all hip fractures were recorded on the hospital discharge abstracts, resulting in an underestimate of hip fracture rates. This is unlikely because hip fracture is one of the most reliably coded hospital diagnoses.(30)
Several studies have reported on evaluation and treatment rates for osteoporosis in older individuals with fractures and have all concluded that these rates fall below what is recommended by current guidelines.(51–61) This study did not examine the subgroup of patients with fracture but instead focused on all those >65 years of age, some of whom would have had fractures before initiation of treatment and others who are being treated for osteoporosis. We also have more recent data on treatment than the other studies, which might explain why the proportion treated in our study is higher than that reported in the literature. It is also difficult to compare the exact proportions on medications to these other studies because they collected information regardless of type of drug coverage and had information on calcium and vitamin D. This study should not be used for direct comparisons but instead as an examination of time trends in BMD testing, treatment, and fracture incidence.
In summary, this is the first study using North American data to show that the numbers of hip and wrist fractures are not increasing as predicted. One other study in Sweden using data from 1982 to 1996 also forecasted a total decrease of 11% in hip fractures up to the year 2010, corresponding to a decrease of 19% in the age-adjusted incidence in women and a 7% increase in men.(23) The trends noted indicate that fracture rates are on the decline, despite the aging of the population, and this is may be due in part to increases in diagnosis and treatment for osteoporosis. This study is an important first step in examining the impact of management efforts to reduce osteoporosis and related fractures from a population perspective. These results need to be confirmed in other populations. This study highlights that fracture rates are declining and that we need to continue with present efforts to improve detection and treatment of osteoporosis to reduce the human and economic burden caused by fractures.
This study was funded by a grant from the Ontario Women's Health Council, and access to the data was provided by the Institute for Clinical Evaluative Sciences. Dr Jaglal is a Career Scientist of the Ontario Ministry of Health and Long-Term Care. Dr Mamdani is supported by a New Investigator award from the Canadian Institutes of Health Research (CIHR). Dr Hawker is a Canadian Institutes of Health Research Scientist and the FM Hill Chair in Academic Women's Medicine at the University of Toronto.