Osteoporosis Int (1992) 2:241-246 © 1992European Foundationfor Osteoporosis
Osteoporosis International
Original Article No Secular Increase in the Prevalence of Vertebral Fractures Due to Postmenopausal Osteoporosis M. A. H a n s e n , K. Overgaard, V.-A. H. Nielsen, G. F. Jensen, A. G o t f r e d s e n and C. Christiansen Department of ClinicalChemistry,Glostrup Hospital, Universityof Copenhagen, Glostrup, Denmark
Abstract. We examined whether the prevalence of vertebral fractures in otherwise healthy, 70-year-old Danish women had increased during an interval of 10 years. The population-based epidemiological study included two age-matched groups of postmenopausal women. Group 1 consisted of 70-year-old women (n=386) living in a defined area of suburban Copenhagen recruited in 1979 for an epidemiological study. Of the 285 women who were entered, 173 were judged healthy, without secondary causes of osteoporosis. Group 2 was recruited by sending questionnaires to all women aged 68-72 years living in the same area in 1989. Of the 512 women who attended a medical screening, 387 were found to be without secondary causes of osteoporosis and had a spinal radiograph. Radiographs of the thoracolumbar spine were assessed for vertebral fracture by five radiological methods. There was no significant difference between the two groups in the prevalence of vertebral fractures and the 95% confidence intervals overlapped completely in all methods. The prevalence rates varied by method from about 35 % to more than 80% but the distribution of fracture types was similar in the two groups. We conclude that the prevalence of vertebral fractures due to postmenopausal osteoporosis has not increased since 1979 in otherwise healthy women residing in suburban Copenhagen, and that comparison of prevalences between studies requires that they use the same method of radiological assessment. Keywords:
Osteoporosis;
Prevalence;
Vertebral
fractures Correspondence and offprint requests to: Marc A. Hansen, Department of ClinicalChemistry,Glostrup Hospital, DK-2600 Glostrup, Denmark.
Introduction Osteoporotic fractures, particularly of the hip and vertebrae, are a major health care problem and an economic burden on societies in the West. The consequences are considerable disability, occupation of hospital beds, and many deaths [1]. As osteoporotic fractures are age-related, the increased life expectancy serves to increase the problem [1]. Moreover, reports of an increasing age-adjusted risk of hip fractures in Europe [2-4], the USA [5] and Canada [6] cause much concern. This trend has, however, not been generally confirmed, as data from the USA [7] have indicated that the age-adjusted rates of hip fractures for women levelled off in the mid-1950s and recent data from England and Wales have shown that rates of admission for hip fractures increased dramatically up to 1978 after which they levelled off [8]. Two recent Scandinavian studies have confirmed this trend in hip fracture rates [9,10]. However, a rising incidence in fractures at other sites has also been reported, although the body of evidence is smaller than that for hip fractures [11-13]. Only one study has been published on the possible secular trend in the incidence of vertebral fracture [14]. This Swedish study showed that from 19502 to 1982-3 it had increased fivefold in 70- to 79-year-old men and women. Few studies have been published on the epidemiology of vertebral fractures and their results vary somewhat [15,16]. Owing to widely different study designs the exact worldwide prevalence of such fractures in elderly postmenopausal women is unknown. Major obstacles to epidemiological studies are the lack of agreement on the definition of a vertebral fracture and the variety of diagnostic methods used. Several approaches have recently been suggested whereby adjustment could be
242
made for normal variations in the vertebral shape and interindividual variations in vertebral size [16-18]. A study of the trends in the prevalence of vertebral fractures requires the same method of assessing all radiographs and necessitates blind re-assessment of older ones. A design in which radiographs are selected according to the diagnosis does not take into account the changes over time in the clinical indication for radiography. Population-based study designs are therefore to be preferred. To examine the trends in the prevalence of vertebral fractures due to postmenopausal osteoporosis from 1979 to 1989, we studied two representative, wellmatched populations of otherwise healthy 70-year-old Danish women. Five methods of radiological assessment were used.
Materials and Methods Subjects' Group 1. In 1979, all 70-year-old women (n = 386) living in nine municipalities near Glostrup Hospital were invited to participate in an epidemiological study [15]. The composition of the population in this area is representative of the Danish population as regards health and socio-economic conditions. A total of 285 women (75%) responded and, after an examination, a subgroup of 173 women was separated out. None of these women had any diseases or were taking medication known to influence calcium metabolism. The women in this subgroup were the ones mainly used for comparison with group 2. In nine of the 285 women the quality of the radiographs was too poor to permit detailed assessment of vertebral fractures. Group 2. Ten years later, in 1989, questionnaires were sent to all women aged 68-72 years living in the same area around Glostrup Hospital - a total of 2009. Of the 1522 women (75%) who returned the questionnaire 788 women appeared to be eligible for the study [19]. These women were invited to a medical screening, and 512 attended. A group of 387 women were without secondary causes of osteoporosis and gave their informed consent to have a radiograph of the thoracolumbar spine. None had any diseases or were taking medication known to influence calcium metabolism. On the selection criteria and state of health, the 387 women were comparable to the subgroup of 173 women in group 1. In order to obtain two groups of healthy 70-year-old women without secondary causes of osteoporosis, the following exclusion criteria were applied: severe immobility, severe cardiovascular disease, stroke or malignancy, and endocrinological diseases. Another exclusion criterion was intake of drugs with known influence on calcium metabolism, e.g. glucocorticoids, anticonvulsants, anticoagulants, anabolic steroids, thyroid hormones, antithyroid hormones.
M . A . Hansen et al.
Methods Medical histories were recorded, including the duration of the menopause and use of postmenopausal hormone replacement therapy. Radiographs of the thoracolumbar spine were taken with the participant lying in a fixed lateral position under standardized conditions with a fixed film-focus distance of 1,0 m. Two separate radiographs were taken of each patient, one with the X-ray tube focusing on T6 and one on L3. The vertebrae from T4 to L5 were well represented and, if necessary, supplemented by a radiograph of the thoracolumbar junction. The estimated radiation dose of the spinal radiographic procedure averages 0.4 cGy (400 mrad) [201. Five methods were used to assess the radiographs for fractures and all assessments were performed by the same experienced radiologist. In four of the methods three heights were measured on each vertebra to the nearest millimetre with a transparent ruler: anterior (ha), midvertebral (hm) and posterior (hp) (Fig. 1). The intra-observer variation in the height measurements was 2.5% (range 1.5-3.8%). Adjustments for normal variations in vertebral shape and size were made on the data of 32 healthy early postmenopausal women (age 52 +_ 2 years) who had no radiological evidence of vertebral fractures. The women were selected by questionnaires as a representative group of early postmenopausal women in the same geographic area as the two study groups. This group was also used to obtain normal ranges of vertebral heights.
Conventional "Naked Eye" Inspection. All the radiographs were assessed subjectively without taking quantitative measurements to confirm the presence or absence of (1) endplate deformities, (2) wedge fractures, or (3) compression fractures. The Method of Melton et al. [16]. This method basically employs the same criterion as the Kleerekoper method: a 15% reduction, at least, in vertebral height. But before this criterion was applied each height measurement was corrected by an adjustment factor that took into account the normal variations in vertebral shape and size throughout the spine between (1) the middle and posterior heights, (2) the anterior and posterior heights, and (3) the posterior heights of adjacent vertebrae. This method has a reported 94% sensitivity and an 82% specificity [211].
"~" INI:~iii ! N ~ I ~ l
J,a
..................... Normal
~ g N Endplate deformity
Wedge
fracture
Compression fracture
Fig. 1. Diagram illustrating vertebral heights and types of fractures.
Prevalence of Vertebral Fractures
The Method of Kleerekoper et al. [22]. Endplate and wedge deformities were defined as a 15% reduction, at least, in the midvertebral and anterior heights compared with the posterior height; compression deformities had to have a ]5% reduction, at least, of the posterior height compared with adjacent vertebrae. This method has a reported sensitivity of 100% and a specificity of 26% for diagnosis of wedge fractures, and 22% sensitivity and 95% specificity for diagnosis of compression fractures [21]. The Method of Minne et al. [17]. The three heights (ha, hm, hp) of each vertebra from T5 to L5 were separately normalized by dividing each height by the corresponding height of T4 and then comparing the ratios with the normal ranges. A vertebra was defined as fractured when any of the normalized heights were below the lower limit of the normal range. The Method of Hedlund and Gallagher [18]. Fracture, but not the type, was identified on the following criteria: (1) the percentage reduction in anterior compared with posterior height (PRH) below the normal range (mean - 2 S D ) , and (2) the percentage difference in the anterior height (PDAH) between adjoining vertebrae outside the normal range (mean _+SD). A third criterion was applied when criterion 2 was invalid owing to three or more adjacent compression fractures: (3) anterior height below the normal range (mean -2SD). Criteria 2 and 3 combined had a reported 86.6% sensitivity and 100% specificity for identification of osteoporotic vertebral fractures except endplate deformities [18]. Bone Mineral Content of the Distal Forearm (gMCarm). This was measured both in 1979 and in 1989 by single photon absorptiometry with a t25I source (3.7 GBq) with a photopeak at 27 keV. The BMC is determined as the mean of six scans just proximal to the site where the distance between the radius and ulna is 8 mm. Since the mid-1970s the system has been calibrated every week and a control scan performed every day using the same standard. In our department the long-term precision in vivo is 1% [23].
243
Results Table 1 shows that the two populations were not significantly different in terms of clinical data and BMCarm- The women in group 2 (1989) were significantly taller than those in group 1 (1979) and had used postmenopausal hormone replacement therapy for a longer period (p
Osteoporosis International
Original Article No Secular Increase in the Prevalence of Vertebral Fractures Due to Postmenopausal Osteoporosis M. A. H a n s e n , K. Overgaard, V.-A. H. Nielsen, G. F. Jensen, A. G o t f r e d s e n and C. Christiansen Department of ClinicalChemistry,Glostrup Hospital, Universityof Copenhagen, Glostrup, Denmark
Abstract. We examined whether the prevalence of vertebral fractures in otherwise healthy, 70-year-old Danish women had increased during an interval of 10 years. The population-based epidemiological study included two age-matched groups of postmenopausal women. Group 1 consisted of 70-year-old women (n=386) living in a defined area of suburban Copenhagen recruited in 1979 for an epidemiological study. Of the 285 women who were entered, 173 were judged healthy, without secondary causes of osteoporosis. Group 2 was recruited by sending questionnaires to all women aged 68-72 years living in the same area in 1989. Of the 512 women who attended a medical screening, 387 were found to be without secondary causes of osteoporosis and had a spinal radiograph. Radiographs of the thoracolumbar spine were assessed for vertebral fracture by five radiological methods. There was no significant difference between the two groups in the prevalence of vertebral fractures and the 95% confidence intervals overlapped completely in all methods. The prevalence rates varied by method from about 35 % to more than 80% but the distribution of fracture types was similar in the two groups. We conclude that the prevalence of vertebral fractures due to postmenopausal osteoporosis has not increased since 1979 in otherwise healthy women residing in suburban Copenhagen, and that comparison of prevalences between studies requires that they use the same method of radiological assessment. Keywords:
Osteoporosis;
Prevalence;
Vertebral
fractures Correspondence and offprint requests to: Marc A. Hansen, Department of ClinicalChemistry,Glostrup Hospital, DK-2600 Glostrup, Denmark.
Introduction Osteoporotic fractures, particularly of the hip and vertebrae, are a major health care problem and an economic burden on societies in the West. The consequences are considerable disability, occupation of hospital beds, and many deaths [1]. As osteoporotic fractures are age-related, the increased life expectancy serves to increase the problem [1]. Moreover, reports of an increasing age-adjusted risk of hip fractures in Europe [2-4], the USA [5] and Canada [6] cause much concern. This trend has, however, not been generally confirmed, as data from the USA [7] have indicated that the age-adjusted rates of hip fractures for women levelled off in the mid-1950s and recent data from England and Wales have shown that rates of admission for hip fractures increased dramatically up to 1978 after which they levelled off [8]. Two recent Scandinavian studies have confirmed this trend in hip fracture rates [9,10]. However, a rising incidence in fractures at other sites has also been reported, although the body of evidence is smaller than that for hip fractures [11-13]. Only one study has been published on the possible secular trend in the incidence of vertebral fracture [14]. This Swedish study showed that from 19502 to 1982-3 it had increased fivefold in 70- to 79-year-old men and women. Few studies have been published on the epidemiology of vertebral fractures and their results vary somewhat [15,16]. Owing to widely different study designs the exact worldwide prevalence of such fractures in elderly postmenopausal women is unknown. Major obstacles to epidemiological studies are the lack of agreement on the definition of a vertebral fracture and the variety of diagnostic methods used. Several approaches have recently been suggested whereby adjustment could be
242
made for normal variations in the vertebral shape and interindividual variations in vertebral size [16-18]. A study of the trends in the prevalence of vertebral fractures requires the same method of assessing all radiographs and necessitates blind re-assessment of older ones. A design in which radiographs are selected according to the diagnosis does not take into account the changes over time in the clinical indication for radiography. Population-based study designs are therefore to be preferred. To examine the trends in the prevalence of vertebral fractures due to postmenopausal osteoporosis from 1979 to 1989, we studied two representative, wellmatched populations of otherwise healthy 70-year-old Danish women. Five methods of radiological assessment were used.
Materials and Methods Subjects' Group 1. In 1979, all 70-year-old women (n = 386) living in nine municipalities near Glostrup Hospital were invited to participate in an epidemiological study [15]. The composition of the population in this area is representative of the Danish population as regards health and socio-economic conditions. A total of 285 women (75%) responded and, after an examination, a subgroup of 173 women was separated out. None of these women had any diseases or were taking medication known to influence calcium metabolism. The women in this subgroup were the ones mainly used for comparison with group 2. In nine of the 285 women the quality of the radiographs was too poor to permit detailed assessment of vertebral fractures. Group 2. Ten years later, in 1989, questionnaires were sent to all women aged 68-72 years living in the same area around Glostrup Hospital - a total of 2009. Of the 1522 women (75%) who returned the questionnaire 788 women appeared to be eligible for the study [19]. These women were invited to a medical screening, and 512 attended. A group of 387 women were without secondary causes of osteoporosis and gave their informed consent to have a radiograph of the thoracolumbar spine. None had any diseases or were taking medication known to influence calcium metabolism. On the selection criteria and state of health, the 387 women were comparable to the subgroup of 173 women in group 1. In order to obtain two groups of healthy 70-year-old women without secondary causes of osteoporosis, the following exclusion criteria were applied: severe immobility, severe cardiovascular disease, stroke or malignancy, and endocrinological diseases. Another exclusion criterion was intake of drugs with known influence on calcium metabolism, e.g. glucocorticoids, anticonvulsants, anticoagulants, anabolic steroids, thyroid hormones, antithyroid hormones.
M . A . Hansen et al.
Methods Medical histories were recorded, including the duration of the menopause and use of postmenopausal hormone replacement therapy. Radiographs of the thoracolumbar spine were taken with the participant lying in a fixed lateral position under standardized conditions with a fixed film-focus distance of 1,0 m. Two separate radiographs were taken of each patient, one with the X-ray tube focusing on T6 and one on L3. The vertebrae from T4 to L5 were well represented and, if necessary, supplemented by a radiograph of the thoracolumbar junction. The estimated radiation dose of the spinal radiographic procedure averages 0.4 cGy (400 mrad) [201. Five methods were used to assess the radiographs for fractures and all assessments were performed by the same experienced radiologist. In four of the methods three heights were measured on each vertebra to the nearest millimetre with a transparent ruler: anterior (ha), midvertebral (hm) and posterior (hp) (Fig. 1). The intra-observer variation in the height measurements was 2.5% (range 1.5-3.8%). Adjustments for normal variations in vertebral shape and size were made on the data of 32 healthy early postmenopausal women (age 52 +_ 2 years) who had no radiological evidence of vertebral fractures. The women were selected by questionnaires as a representative group of early postmenopausal women in the same geographic area as the two study groups. This group was also used to obtain normal ranges of vertebral heights.
Conventional "Naked Eye" Inspection. All the radiographs were assessed subjectively without taking quantitative measurements to confirm the presence or absence of (1) endplate deformities, (2) wedge fractures, or (3) compression fractures. The Method of Melton et al. [16]. This method basically employs the same criterion as the Kleerekoper method: a 15% reduction, at least, in vertebral height. But before this criterion was applied each height measurement was corrected by an adjustment factor that took into account the normal variations in vertebral shape and size throughout the spine between (1) the middle and posterior heights, (2) the anterior and posterior heights, and (3) the posterior heights of adjacent vertebrae. This method has a reported 94% sensitivity and an 82% specificity [211].
"~" INI:~iii ! N ~ I ~ l
J,a
..................... Normal
~ g N Endplate deformity
Wedge
fracture
Compression fracture
Fig. 1. Diagram illustrating vertebral heights and types of fractures.
Prevalence of Vertebral Fractures
The Method of Kleerekoper et al. [22]. Endplate and wedge deformities were defined as a 15% reduction, at least, in the midvertebral and anterior heights compared with the posterior height; compression deformities had to have a ]5% reduction, at least, of the posterior height compared with adjacent vertebrae. This method has a reported sensitivity of 100% and a specificity of 26% for diagnosis of wedge fractures, and 22% sensitivity and 95% specificity for diagnosis of compression fractures [21]. The Method of Minne et al. [17]. The three heights (ha, hm, hp) of each vertebra from T5 to L5 were separately normalized by dividing each height by the corresponding height of T4 and then comparing the ratios with the normal ranges. A vertebra was defined as fractured when any of the normalized heights were below the lower limit of the normal range. The Method of Hedlund and Gallagher [18]. Fracture, but not the type, was identified on the following criteria: (1) the percentage reduction in anterior compared with posterior height (PRH) below the normal range (mean - 2 S D ) , and (2) the percentage difference in the anterior height (PDAH) between adjoining vertebrae outside the normal range (mean _+SD). A third criterion was applied when criterion 2 was invalid owing to three or more adjacent compression fractures: (3) anterior height below the normal range (mean -2SD). Criteria 2 and 3 combined had a reported 86.6% sensitivity and 100% specificity for identification of osteoporotic vertebral fractures except endplate deformities [18]. Bone Mineral Content of the Distal Forearm (gMCarm). This was measured both in 1979 and in 1989 by single photon absorptiometry with a t25I source (3.7 GBq) with a photopeak at 27 keV. The BMC is determined as the mean of six scans just proximal to the site where the distance between the radius and ulna is 8 mm. Since the mid-1970s the system has been calibrated every week and a control scan performed every day using the same standard. In our department the long-term precision in vivo is 1% [23].
243
Results Table 1 shows that the two populations were not significantly different in terms of clinical data and BMCarm- The women in group 2 (1989) were significantly taller than those in group 1 (1979) and had used postmenopausal hormone replacement therapy for a longer period (p
