Bone, 12, pp. 107-llZ!, (1991) Printed in the USA. All rights reserved.
Copyright
8756-3282/91 $3.00 + .OO 0 1991 Pergamon Press plc
Vertebral Dimension Differences Between Caucasian Populations, and Between Caucasians and Japanese P. D. ROSS,
R. D. WASNICH,
Hawaii Osteoporosis
J. W. DAVIS
and J. M. VOGEL
Center, 932 Ward Ave., Suite 400, Honolulu, HI 96817, USA
Address for correspoderrce
and reprints: Philip D. Ross, Ph.D., Kuakini Osteoporosis Center, 347 North Ku&hi
Sweet, Honolulu, HI
96817, USA.
Abstract
These measurements were also compared with published values for Caucasian (‘ ‘white”) populations. Findings regarding the reproducibility of the measurements, and how best to define vertebral fractures, will appear in a subsequent report.
Various criteria have been proposed for using vertebral measurements to identify vertebral fractures. It is known that the normal distributions of vertebral heights and ratios vary with location within the spine. However, very little is known regarding the degree to which differences in these parameters may exist between populations. We report the vertebra-specific distributions of vertebral dimensions and ratios for Japanese-Americans, and compare these values to published data for Caucasians. The mean Japanese vertebral heights were 1 to 2 mm shorter than Caucasians, which may be due in part to the shorter stature of Japanese. However, differences in mean values were also observed between Caucasian populations. Furthermore, anterior/pasterior vertebral height ratios differed between Caucasian studies, and between races. Additional studies are needed to determine to what degree these differences are due to technical and biological factors before criteria derived from one population can Ibe used for identifying vertebral fractures in other populations of the same, or different, race.
Materials and Methods Selection of the cohort has been described previously (Yano et al. 1984). Briefly, a random subsample of men who had participated previously in the Honolulu Heart Study, and their wives, were invited to participate in the Kuakini Osteoporosis Study beginning in 1981. Only data on the wives will be discussed here. These women (n = 1098) ranged in age from 43 to 80 (mean = 63.3 years), and 1083 (99%) were postmenopausal at the time of the first examination. Additional examinations have been performed at scheduled intervals since the initial examination. Anterior-posterior (A-P) radiographs were performed with the subject lying supine, and lateral views were performed with the subject lying on her side, with knees bent. All radiographs were obtained using a tube to film distance of 41.3 inches (105 cm). A random subsample of approximately 50% of the original cohort received thoraco-lumbar spine radiographs (including all vertebrae below the approximate level of Tll, but only Ll through L5 were measured) during the initial examinations in 1982. The X-ray tube was positioned approximately over the level of L3 for both A-P and lateral views. Beginning in 1984, all subjects received radiographs at each exam that included all vertebrae below the approximate level of T8. These radiographs were performed with the tube positioned approximately over L2 for both A-P and lateral views, and only TlO through L6 were measured. Subsequent lateral radiographs of the entire spine were obtained at predetermined intervals beginning in late 1986 using the same conditions as in 1984-1985 for the lumbar view, plus a thoracic view (T3 through T12) with the tube positioned approximately over T8. The nomenclature used here for vertebral dimensions is as follows: A = anterior height, and P = posterior height. For ratios of posterior heights of adjacent vertebrae (P,lPi + , , Pi/Pi _ l)r the value of i increases moving caudally from T3 to L6 (Melton et al. 1989). Dimensions were measured using the same procedure described by Gallagher et al. (1988). Characteristics of the Caucasian groups used for comparison with this study were derived from the following reports: Davies et al. 1989, Minne et al. 1988, Smith et.al. 1987, Gallagher et
Key Words: Epidemiology-Vertebral gy-Fracture
fractures-Morpholoprevalence-Japanese-Americans-Caucasians.
Introduction Although vertebral fractures are the most common type of osteoporotic fracture, relatively little is known about their frequency and their consequences. The lack of an objective definition for vertebral fractures constitutes a major problem when attempting to investigate both risk factors for fracture and the resulting outcomes, such as pain and disability. There is often poor intra- and inter-observer agreement for qualitative diagnosis of vertebral fractures by radiologists (Jensen et al. 1984). The resulting misclassification errors can impair the ability to detect meaningful associations, and to compare results between studies. Several reports have recently recommended that vertebra-specific dimension criteria be used to define existing fractures (Davies et al. 1989; Minne et al. 1988; Smith et al. 1987; Gallagher et al. 1988). However, it has not been determined whether or not criteria derived from one population can be applied to other populations. In this report, we have examined the distributions of vertebral dimensions for a population of Japanese-American women. 107
108
al. 1988. These articles were identified by surveying journals devoted to bone and mineral research, and by a Medline search of articles published since 1980 concerning osteoporosis. In the study of Minne et al. (1988), lateral radiographs of 72 women were retrieved from archives of the radiology department in Heidelberg. Race was not specified, but it is assumed here that they were Caucasian. These films were judged to be normal by one of the investigators. Approximately 58% of the women were under 50 years old. Davies et al. (1989) used 568 lateral spine radiographs from a group of 191 active, generally healthy, Caucasian, Roman Catholic nuns, ages 35 to 45 at the initial examination. Serial radiographs were collected at five-year intervals beginning in 1967. Melton et al. (1989) contacted an age-stratified random sample of adult women in Rochester, Minnesota. From an overall response rate of 60%) they selected radiographs without clinical evidence of vertebral fractures from 52 women over age 50 who were not taking corticosteroids, anticonvulsants, thiazide diuretics, vitamin D in pharmacologic doses, or calcium supplements >500 mg/day, and who were free of diseases known to influence bone metabolism. Smith et al. (1987) used 114 normal, healthy 35 to 45year-old women; race was not specified. Gallagher et al. (1988) used 150 Caucasian women ages 34 to 67 years who were normal as determined from medical questionnaires. Subjects with current conditions or treatments known to affect calcium metabolism were excluded. Statistical methods Since many of the women in our cohort were older than 60 years, and may have had prevalent vertebral fractures, it was necessary to exclude some measurements in order to summarize reference range. In the absence of a consensus a “normal” individual vertebrae with A/P regarding a “gold standard,” and/or Pi/Pi-l values less than 75% of the median were excluded from analyses reported in this paper. This cutoff was based on the observation that four times the standard deviation (SD) is approximately equal in magnitude to 25% of the average value (0.25* average) for each type of ratio. Excluding extreme values using an 75% cutoff is therefore similar to excluding values more than 4 SD below the mean. Excluding the most extreme values in this manner yields a normal distribution by eliminating obvious fractures that otherwise skew the distribution. For analyses of dimension ratios, the ratio was calculated for each subject individually, prior to calculating means and SDS for the population, and converted to percent ratios by multiplying the result by 100. For the purpose of comparison, it was necessary to estimate Caucasian data from graphs in some reports. In these reports (Gallagher et al. 1988; Smith et al. 1987) values could not be estimated accurately; therefore, fewer significant digits were tabulated. Results Means and SDS of the anterior and posterior vertebral heights, and of ratios, for our cohort of Japanese-American women are provided in Table I. The ratios of anterior to posterior height, and of posterior heights for adjacent vertebrae, varied with position in the spine. A few subjects (n=7) had six lumbar vertebrae. Table II summarizes the differences in dimensions between our Japanese-American cohort and three Caucasian populations. Vertebral heights were approximately 1 to 2 mm lower for both A and P among Japanese-Americans, relative to Caucasians, varying somewhat with position in the spine. In the study of
P. D. Ross et al.: Comparison
of vertebral dimensions
Minne et al. (1988), values of A were approximately 2 to 3 mm greater than those of Japanese-Americans, and 1 mm greater than in the other two studies of Caucasians (Table II). However, values of P were similar among the three Caucasian studies. Three studies reported values of A/P for Caucasians (Gallagher et al. 1988; Melton et al. 1989; Davies et al. 1989). With the exception of T3 in one study (Gallagher et al. 1988), all values for Caucasians were equa1 to or greater than those for Japanese-Americans. The magnitude of these differences varied by location within the spine for each study, and there were also substantial differences among the Caucasian studies for a given location in the spine (Table III). Average differences in A/P percent ratios ranged from 1.4 to 4.3 among studies, compared to Japanese-Americans. Only two studies reported posterior height ratios for Caucasians (Melton et al. 1989; Davies et al. 1989). The average differences relative to Japanese-Americans were very small, varying only slightly between studies. Variation with location in the spine was much greater (ranging from - 2.9% to + 3.8%), both for comparisons of Caucasians with each other, and to the Japanese (Table IV).
Discussion The presence of vertebral fractures among older women, and the lack of an accepted “gold standard,” makes it difficult to determine normal values for older women. Therefore, one may be tempted to use younger women to develop a suitable “reference range.” However, some younger women may have vertebral abnormalities due to congenital defects, trauma, disease, or other conditions that are unrecognized or unreported by the individual. Also, as Minue et al. (1988) have shown, significant “cohort effects” may occur. In this context, the term cohort effect refers to underlying differences between successive generations. In the study by Minne et al. (1988), there was an apparent increase in vertebral heights in successive generations that are assumed to be a result of increases in body height. Thus, a “reference range” derived from young women may yield unreliable results if they are applied to older populations in the same community, the degree of error being related to the magnitude of differences in vertebral height and/or shape between the generations. Minne et al. recommend adjusting measured values by dividing by the height of T4. However, this may create additional problems, since it adds an additional source of measurement error. Furthermore, differences between communities may exist in addition to differences between generations and races. It may not be possible to adjust accurately for such differences, which may be exacerbated by both genetic and environmental differences between communities, even if they are of the same race. It is reasonable to suspect that women of Japanese ancestry might have vertebral dimensions that differ significantly from Caucasians, since the distributions of body physiques differ between these two races. The mean values of height and weight of our cohort were 151.3 cm and 53.9 kg (Ross et al. 1987), compared to 159.1 cm and 66.7 kg for U.S. whites (National Center for Health Statistics et al. 1989). This corresponds to a 7.8 cm (5.2%) difference in stature. Based upon regressions of vertebral height against stature for our Japanese population, adjusting for this difference in stature would increase the Japanese vertebral heights (A and P) by about 1.5 mm. Thus, it is possible that the average differences in vertebral height between Japanese and Caucasians may be explained in large part by the difference in body size. However, this cannot be demonstrated conclusively from the available data because of differences in age and race between our study and the others.
s
Vertebra T3 T4 T5 T6 T7 T8 T9 TlO Tll T12 Ll L2 L3 L4 L5 L6
$63 1506 1512 1495 1485 1485 1504 1864 1791 1798 2740 2835 2804 2803 2658 37 19.8 20.0 20.3 20.4 20.6 21.5 22.8 24.4 25.6 27.9 30.2 32.1 32.9 32.7 32.7 31.3
A 1.3 1.4 1.4 1.5 1.5 1.6 1.6 1.8 1.9 2.1 2.2 2.3 2.3 2.4 2.4 2.1
SD 21.3 21.7 22.4 23.0 23.6 23.9 24.5 26.0 28.4 31.0 33.2 33.8 33.3 31.5 29.5 28.6
P 1.6 1.5 1.5 1.5 1.5 1.5 1.6 1.8 2.0 2.1 2.1 2.1 2.2 2.4 2.2 2.7
SD 93.3 92.0 90.6 88.4 87.2 89.7 93.0 93.6 89.6 89.7 90.5 95.0 98.8 103.8 110.6 109.7
A/P 6.7 5.4 5.4 5.6 5.8 6.1 5.7 6.1 6.6 6.3 6.1 6.1 6.4 7.2 9.0 11.2
SD &.,
Table I. Means and standard deviations of vertebral dimensions (mm) and ratios (8) for Japanese-American because some subjects did not receive thoracic radiographs at the first examination.
SD ___ 9.3 5.4 4.6 4.5 4.5 4.7 5.1 5.9 6.0 5.9 5.3 4.3 4.8 6.1 6.4
102.3 103.2 102.3 102.4 101.5 102.7 106.4 109.2 109.3 107.0 101.7 98.8 94.6 93.6 93.6
in Hawaii.
___
women
98.1 97.2 97.2 98.0 98.7 97.6 94.3 91.7 91.8 93.7 98.5 101.4 105.9 107.3 107.8
I!&+,
6.5 5.1 4.6 5.4 4.5 4.4 4.5 4.9 5.1 5.1 4.6 4.4 5.7 7.3 9.3
SD
The sample size varies with region of the spine
P. D. Ross et al.: Comparison
110
Table II. Differences Differences
of vertebral dimensions
in anterior and posterior heights between Japanese-American women and Caucasian women. are given in units of millimeters, calculated as (Caucasian value) minus (Japanese-American
value). Values at the bottom of the table represent the means and SDS of the 10 to 15 values in each COlUlM.
Study: Vertebra T3 T4 T5 T6 T7 T8 T9 TlO Tll T12 Ll L2 L3 L4 L5 Mean difference: SD of differences0.5
P Davies Minng Gallaaher
A Davies
Minne Bllaaher
__ __ -_ __ 1.8 1.4 1.4 1.5 1.8 2.2 2.0 1.6 2.1 3.1 __
__ 1.9 2.0 2.1 2.4 2.2 2.2 2.5 2.9 3.0 2.7 2.3 2.8 2.9 4.0
2 2 2 2 1 0 0 2 1 1 2 2 2 1 1
__
__
__ __ __ 1.5 1.5 1.5 1.4 1.1 1.2 1.7 0.5 1.5 2.7 __
1.2 1.1 1.1 1.0 1.1 1.5 1.6 1.3 1.1 1.7 1.2 2.3 3.1 1.9
2 1 2 2 1 2 1 1 2 1 1 1 2 1 0
1.9
2.6 0.5
1.4 0.7
1.5 0.6
1.5 0.6
1.3 0.6
Furthermore, no relationship between stature and A/P or (Pi/ P, _ ,) ratios could be demonstrated. Thus, the differences in A/P ratio among studies do not appear to be explained by the differences in stature between races. The differences in specific vertebral heights were sometimes greater between Caucasian studies than between Japanese and Caucasians (for example, the values of A for T8 and T9, Table II). These data suggest that each study may have used a different
procedure to select the vertebral margins. Differences in margin selection have been reported to influence the calculated vertebral area (Nelson et al. 1990), but have not been reported for vertebral height measurements. For the study of Minne et al., differences in the values of A were somewhat greater than in the other Caucasian studies, but the average values of P were not, suggesting that the method used to mark the anterior dimensions may have been different. Details of the procedures used to mark
Table III. Differences
in anterior/posterior percent ratios between Japanese-American women and Caucasian women. Differences were calculated as (Caucasian A/P % value) minus (Japanese-American A/P % value). Values at the bottom of the table represent the means and SDS of the 10 to 15 values in each column.
Caucasian Group: Vertebra T3 T4 T5 T6 T7 T8 T9 TIO Tll T12 Ll L2 L3 L4 L5
Davies
Melton
Smith
Gallaaher
__ __ __ __ 2.2 0.9 0.0 1.1 3.5 3.8 4.5 3.9 1.8 1 .o __
__ 4.9 4.9 4.7 5.0 3.5 1.7 2.4 3.8 4.5 6.5 6.3 4.6 3.3 3.7
__
-1 1 1 1 1 1 2 2 1 2 3 2 0 1
Mean difference: SD of differences:
2.3 1.6
4.3 1.3
2.6 2.0
1 2 1 4 2 0 1 2 6 5 4 4 0 5
4
1.4 1.2
P. D. Ross et al.: Comparison
111
of vertebral dimensions
Table IV. Differences in posterior height percent ratios between Japanese-American women and Caucasian women. Differences were calculated as (Caucasian % value) minus (Japanese-American % value). Values at the bottom of the table represent the means and SDS of the 10 to 15 values in each column.
Caucasian Group: Vertebra! T3 T4 T5 T6 T7 T8 T9 TIO Tll T12 Ll L2 L3 L4 L5 Mean difference: SD of differences:
y&l-
&&+1-
Melton
Melton
__ __ __ __ __ -0.3 0.1 -1.2 -1.3 0.0 -1.6 -0.5 2.9 3.8 _-
__ _-
__
2.1 1.7 0.5 1.6 0.9 -0.8 1.0 -1.8 -2.9 0.6 3.1 2.2 -2.5*
-1.9 -0.8 -0.5 -1.4 -0.7 0.7 -0.8 1.4 2.8 -0.3 -2.7 -2.3 0.9* __
0.2 1.9
0.4 1.9
-0.4 1.6
*It was assumed that ratios in the original report were inverted (see reference for explanation).
vertebral borders on radiographs were not specified in most reports. Based on the illustration in the paper, it appears that the procedure used by Gallagher et al. (1988) should yield the same results as the method used in our study, which differed only in that the borders were marked with points, and lines were not drawn. However, the possibility remains that interobserver bias exists between the two studies, which would only be detected by exchanging and marking many films, possibly hundreds. It may be worth noting that the smallest differences between Japanese and Caucasians were consistently noted for the study of Gallagher et al. In contrast to our study and that of Gallagher et al., Hurxthal recommended that posterior prominences be excluded from measurements, which would reduce values of P by about 1 to 2 mm. Although the exact procedure was not specified, if Minne et al. had used IHurxthal’s procedure, this might explain in part why the values elf A, but not P, were greater than those of the other Caucasian studies. Alternatively, true differences in vertebral heights (and ratios) may exist among Caucasian populations because of differences in genetic and/or environmental factors. Our findings suggest that vertebral shapes may also differ between Japanese and Caucasians, since the A!P and Pi/Pi- 1 ratios were different. Within one comparison population, these differences were not consistent, varying in magnitude with location in the spine (‘Tables III and IV). Moreover, both the average and vertebra-specific differences varied considerably, depending on the Caucasian study chosen for comparison. This suggested that there could also be differences in radiographic technique, position of markings for dimensions, and/or shapes of the vertebrae among the Caucasian studies, as well as between this study of Japanese-Americans and the Caucasian studies.
The position of the X-ray tube would be expected to influence the values of A and P, because the magnification effect will increase with distance of the vertebra from the hypocenter of the tube. The A/P ratio would be expected to be influenced much less, since both the A and P measurements of a single vertebra should have a similar degree of magnification. The ratio of Pi/Pi - 1 might be influenced somewhat, because since adjacent vertebrae will be subject to different magnification factors, especially at the margins of the film. Increasing distance from the hypocenter will also increase the problem of parallax, making it more difficult to define vertebral margins, and possibly biasing the measured dimensions. Davies et al. (1989) used a tube-to-film distance of 102 cm (40 inches), centered over L3 for lumbar views, and over T7 for thoracic views. In the study of Minne et al. (1988), the tube was 105 cm (41.3 inches) from the film, centered over L3 and T9, respectively; similar conditions were used in our study of Japanese-Americans (see Methods). A tube-to-film distance of 105 cm (41.3 inches) was used by Melton et al. (1989), but centering position was not specified. Gallagher et al. (1988) used a tube-to-film distance of 40 inches, centered over T9 for thoracic views, and L2 for lumbar views. Information regarding tube position was not available for the study of Smith et al. (1987). Gallagher et al. (1988) reported that an increase of 10 cm (4 inches) in tube-to-film distance reduced the measured dimensions by approximately 6%, or 1.9 mm for an initial measurement of 32 mm. Thus, the differences in tube-to-film distance of 2.5 cm (1 inch) or less, do not completely explain the observed variations in A and P values among studies. Likewise, there are no obvious patterns in any of the vertebral dimension parameters related to the differences in position of the X-ray beam hypocenter on the spine between studies.
112
P. D. Ross et al.: Comparison
Differences in patient thickness among studies could have influenced the measured dimensions, by affecting the distance between the vertebrae and the film, resulting in different magnification factors. Unfortunately, characteristics regarding patient thickness were not specified in any of the studies compared here. Some of the observed differences in A/P ratios (Table III) were comparable in magnitude to the SD values shown for A/P measurements in Table I. If a cutoff of -2 SD from one population were to be applied to a second population, an error of this magnitude (1 SD) would result in misclassifying approximately 14% of the population as having vertebral fractures. In other situations, the proportion of people misclassified could be greater or smaller, depending upon both the cutoff selected and the true magnitude of the difference between the populations. As an example, we applied the “adjusted minimum” cutoff values reported by Davies et al. (1989) to our population. Using their criteria resulted in an apparent prevalence of crush fractures at L4 and L5 of 8% and 18%, respectively. These figures are much higher than expected, based on the report of Melton et al. (1989). Furthermore, a radiologist who reviewed our films from the first examination found the prevalence of crush fractures to be less than 1% at L4 and L5 (data not shown). In condlusion, it appears that the vertebral heights of our Japanese-Americans are 1 to 2 mm shorter than Caucasians. It remains uncertain how Japanese and Caucasian vertebrae may differ in shape. Further research is needed regarding technical, genetic, and other factors (e.g., birth cohort, age) that may influence vertebral dimensions and ratios. At present, it is not advisable to use reference ranges developed for one population for classifying fracture prevalence in other populations.
Acknowledgmenrs: We thank the for excellent technical and clerical
entire staff of the Osteoporosis Center support.
of vertebral dimensions
References Davies, K. M.; Reeker, R. R.; Heaney, R. P. Normal vertebral dimensions and normal variation in serial measurements of vertebrae. J. Bone Min. Res. 4:341-349; 1989. Gallagher, J. C.; Hedlund, L. R.; Stoner, S.; Meeger, C. Vertebral morphometry: Normative data. Bone Miner. 4189-196; 1988. Jensen. G. F.: McNair. P.; Boesen, J.; Hegedus, V. Validity in diagnosing osteoporosis. Eur. J. Radial. &l-3; 1984. Melton, L. J. III; Km, S. H.; Frye, M. A.; Wahner, H. W.; O’Fallon, W. M.; Riggs, B. L. Epidemiology of vertebral fractures in women. Am. .l. Epidemiol. 129:1000-1011; 1989. Minne, H. W.; Leidig, G.; Waster, C.; Siromachkostov, L.; Baldauf, G.; Bickel, R.; Saw, P.; Lojen, M.; Ziegler, R. A newly developed spine deformity index (SDI) to quantitate vertebral crush fractures in patients with osteoporosis. Bone Miner. 3:335-349; 1988. National Center for Health Statistics; Najjar, M. F.; Rowland, M. Anthropometric reference data and prevalence of overweight, United States 1976-80. Vital and health statistics. Series 11, No. 238. DHHS F’ubl. No. (PHS)87-1688. Public Health Service. Washington, DC.: U.S. Government Printing Office, Oct. 1987. Nelson, D.; Peterson, E.; Tilley, B .; O’Fallon, W. M.; Chao, E.; Riggs, B. L.: Kleerkoper, M. Measurement of vertebral area on spine X-rays in osteoporosis: Reliability of digitizing techniques. J. Bone Min. Res. 5707-716; 1990. Ross, P. D., Wasnich, R. D.; Heilbmn, L. K.; Vogel, .I. M. Definition of a spine fracture threshold based upon prospective frachlre risk. Bone 8:271-278; 1987. Smith, R.; Cummings, S. R.; Genant, H. K.; Steiger, P.; Reeker, R. The study of osteoporotic fractures study group: Overdiagnosis of vertebral fractures: the need for vertebra specific criteria. Christiansen, C.; Johansen. J. S.; Riis, B. I.; eds., Osreoporosis 1987. Copenhagen: Osteopress; 1987; 106-108. Yano, K.; Wasnich, R. D.; Vogel. J. M.; Heilbnm. L. K. Bone mineral measurements among middle-aged and elderly Japanese residents in Hawaii. Am. J. Epidemiol. 119~751-764, 1984.
Dare Received: April 9, 1990 Dare Revised: October 3, 1990 Dare Accepted: October 19, 1990
Copyright
8756-3282/91 $3.00 + .OO 0 1991 Pergamon Press plc
Vertebral Dimension Differences Between Caucasian Populations, and Between Caucasians and Japanese P. D. ROSS,
R. D. WASNICH,
Hawaii Osteoporosis
J. W. DAVIS
and J. M. VOGEL
Center, 932 Ward Ave., Suite 400, Honolulu, HI 96817, USA
Address for correspoderrce
and reprints: Philip D. Ross, Ph.D., Kuakini Osteoporosis Center, 347 North Ku&hi
Sweet, Honolulu, HI
96817, USA.
Abstract
These measurements were also compared with published values for Caucasian (‘ ‘white”) populations. Findings regarding the reproducibility of the measurements, and how best to define vertebral fractures, will appear in a subsequent report.
Various criteria have been proposed for using vertebral measurements to identify vertebral fractures. It is known that the normal distributions of vertebral heights and ratios vary with location within the spine. However, very little is known regarding the degree to which differences in these parameters may exist between populations. We report the vertebra-specific distributions of vertebral dimensions and ratios for Japanese-Americans, and compare these values to published data for Caucasians. The mean Japanese vertebral heights were 1 to 2 mm shorter than Caucasians, which may be due in part to the shorter stature of Japanese. However, differences in mean values were also observed between Caucasian populations. Furthermore, anterior/pasterior vertebral height ratios differed between Caucasian studies, and between races. Additional studies are needed to determine to what degree these differences are due to technical and biological factors before criteria derived from one population can Ibe used for identifying vertebral fractures in other populations of the same, or different, race.
Materials and Methods Selection of the cohort has been described previously (Yano et al. 1984). Briefly, a random subsample of men who had participated previously in the Honolulu Heart Study, and their wives, were invited to participate in the Kuakini Osteoporosis Study beginning in 1981. Only data on the wives will be discussed here. These women (n = 1098) ranged in age from 43 to 80 (mean = 63.3 years), and 1083 (99%) were postmenopausal at the time of the first examination. Additional examinations have been performed at scheduled intervals since the initial examination. Anterior-posterior (A-P) radiographs were performed with the subject lying supine, and lateral views were performed with the subject lying on her side, with knees bent. All radiographs were obtained using a tube to film distance of 41.3 inches (105 cm). A random subsample of approximately 50% of the original cohort received thoraco-lumbar spine radiographs (including all vertebrae below the approximate level of Tll, but only Ll through L5 were measured) during the initial examinations in 1982. The X-ray tube was positioned approximately over the level of L3 for both A-P and lateral views. Beginning in 1984, all subjects received radiographs at each exam that included all vertebrae below the approximate level of T8. These radiographs were performed with the tube positioned approximately over L2 for both A-P and lateral views, and only TlO through L6 were measured. Subsequent lateral radiographs of the entire spine were obtained at predetermined intervals beginning in late 1986 using the same conditions as in 1984-1985 for the lumbar view, plus a thoracic view (T3 through T12) with the tube positioned approximately over T8. The nomenclature used here for vertebral dimensions is as follows: A = anterior height, and P = posterior height. For ratios of posterior heights of adjacent vertebrae (P,lPi + , , Pi/Pi _ l)r the value of i increases moving caudally from T3 to L6 (Melton et al. 1989). Dimensions were measured using the same procedure described by Gallagher et al. (1988). Characteristics of the Caucasian groups used for comparison with this study were derived from the following reports: Davies et al. 1989, Minne et al. 1988, Smith et.al. 1987, Gallagher et
Key Words: Epidemiology-Vertebral gy-Fracture
fractures-Morpholoprevalence-Japanese-Americans-Caucasians.
Introduction Although vertebral fractures are the most common type of osteoporotic fracture, relatively little is known about their frequency and their consequences. The lack of an objective definition for vertebral fractures constitutes a major problem when attempting to investigate both risk factors for fracture and the resulting outcomes, such as pain and disability. There is often poor intra- and inter-observer agreement for qualitative diagnosis of vertebral fractures by radiologists (Jensen et al. 1984). The resulting misclassification errors can impair the ability to detect meaningful associations, and to compare results between studies. Several reports have recently recommended that vertebra-specific dimension criteria be used to define existing fractures (Davies et al. 1989; Minne et al. 1988; Smith et al. 1987; Gallagher et al. 1988). However, it has not been determined whether or not criteria derived from one population can be applied to other populations. In this report, we have examined the distributions of vertebral dimensions for a population of Japanese-American women. 107
108
al. 1988. These articles were identified by surveying journals devoted to bone and mineral research, and by a Medline search of articles published since 1980 concerning osteoporosis. In the study of Minne et al. (1988), lateral radiographs of 72 women were retrieved from archives of the radiology department in Heidelberg. Race was not specified, but it is assumed here that they were Caucasian. These films were judged to be normal by one of the investigators. Approximately 58% of the women were under 50 years old. Davies et al. (1989) used 568 lateral spine radiographs from a group of 191 active, generally healthy, Caucasian, Roman Catholic nuns, ages 35 to 45 at the initial examination. Serial radiographs were collected at five-year intervals beginning in 1967. Melton et al. (1989) contacted an age-stratified random sample of adult women in Rochester, Minnesota. From an overall response rate of 60%) they selected radiographs without clinical evidence of vertebral fractures from 52 women over age 50 who were not taking corticosteroids, anticonvulsants, thiazide diuretics, vitamin D in pharmacologic doses, or calcium supplements >500 mg/day, and who were free of diseases known to influence bone metabolism. Smith et al. (1987) used 114 normal, healthy 35 to 45year-old women; race was not specified. Gallagher et al. (1988) used 150 Caucasian women ages 34 to 67 years who were normal as determined from medical questionnaires. Subjects with current conditions or treatments known to affect calcium metabolism were excluded. Statistical methods Since many of the women in our cohort were older than 60 years, and may have had prevalent vertebral fractures, it was necessary to exclude some measurements in order to summarize reference range. In the absence of a consensus a “normal” individual vertebrae with A/P regarding a “gold standard,” and/or Pi/Pi-l values less than 75% of the median were excluded from analyses reported in this paper. This cutoff was based on the observation that four times the standard deviation (SD) is approximately equal in magnitude to 25% of the average value (0.25* average) for each type of ratio. Excluding extreme values using an 75% cutoff is therefore similar to excluding values more than 4 SD below the mean. Excluding the most extreme values in this manner yields a normal distribution by eliminating obvious fractures that otherwise skew the distribution. For analyses of dimension ratios, the ratio was calculated for each subject individually, prior to calculating means and SDS for the population, and converted to percent ratios by multiplying the result by 100. For the purpose of comparison, it was necessary to estimate Caucasian data from graphs in some reports. In these reports (Gallagher et al. 1988; Smith et al. 1987) values could not be estimated accurately; therefore, fewer significant digits were tabulated. Results Means and SDS of the anterior and posterior vertebral heights, and of ratios, for our cohort of Japanese-American women are provided in Table I. The ratios of anterior to posterior height, and of posterior heights for adjacent vertebrae, varied with position in the spine. A few subjects (n=7) had six lumbar vertebrae. Table II summarizes the differences in dimensions between our Japanese-American cohort and three Caucasian populations. Vertebral heights were approximately 1 to 2 mm lower for both A and P among Japanese-Americans, relative to Caucasians, varying somewhat with position in the spine. In the study of
P. D. Ross et al.: Comparison
of vertebral dimensions
Minne et al. (1988), values of A were approximately 2 to 3 mm greater than those of Japanese-Americans, and 1 mm greater than in the other two studies of Caucasians (Table II). However, values of P were similar among the three Caucasian studies. Three studies reported values of A/P for Caucasians (Gallagher et al. 1988; Melton et al. 1989; Davies et al. 1989). With the exception of T3 in one study (Gallagher et al. 1988), all values for Caucasians were equa1 to or greater than those for Japanese-Americans. The magnitude of these differences varied by location within the spine for each study, and there were also substantial differences among the Caucasian studies for a given location in the spine (Table III). Average differences in A/P percent ratios ranged from 1.4 to 4.3 among studies, compared to Japanese-Americans. Only two studies reported posterior height ratios for Caucasians (Melton et al. 1989; Davies et al. 1989). The average differences relative to Japanese-Americans were very small, varying only slightly between studies. Variation with location in the spine was much greater (ranging from - 2.9% to + 3.8%), both for comparisons of Caucasians with each other, and to the Japanese (Table IV).
Discussion The presence of vertebral fractures among older women, and the lack of an accepted “gold standard,” makes it difficult to determine normal values for older women. Therefore, one may be tempted to use younger women to develop a suitable “reference range.” However, some younger women may have vertebral abnormalities due to congenital defects, trauma, disease, or other conditions that are unrecognized or unreported by the individual. Also, as Minue et al. (1988) have shown, significant “cohort effects” may occur. In this context, the term cohort effect refers to underlying differences between successive generations. In the study by Minne et al. (1988), there was an apparent increase in vertebral heights in successive generations that are assumed to be a result of increases in body height. Thus, a “reference range” derived from young women may yield unreliable results if they are applied to older populations in the same community, the degree of error being related to the magnitude of differences in vertebral height and/or shape between the generations. Minne et al. recommend adjusting measured values by dividing by the height of T4. However, this may create additional problems, since it adds an additional source of measurement error. Furthermore, differences between communities may exist in addition to differences between generations and races. It may not be possible to adjust accurately for such differences, which may be exacerbated by both genetic and environmental differences between communities, even if they are of the same race. It is reasonable to suspect that women of Japanese ancestry might have vertebral dimensions that differ significantly from Caucasians, since the distributions of body physiques differ between these two races. The mean values of height and weight of our cohort were 151.3 cm and 53.9 kg (Ross et al. 1987), compared to 159.1 cm and 66.7 kg for U.S. whites (National Center for Health Statistics et al. 1989). This corresponds to a 7.8 cm (5.2%) difference in stature. Based upon regressions of vertebral height against stature for our Japanese population, adjusting for this difference in stature would increase the Japanese vertebral heights (A and P) by about 1.5 mm. Thus, it is possible that the average differences in vertebral height between Japanese and Caucasians may be explained in large part by the difference in body size. However, this cannot be demonstrated conclusively from the available data because of differences in age and race between our study and the others.
s
Vertebra T3 T4 T5 T6 T7 T8 T9 TlO Tll T12 Ll L2 L3 L4 L5 L6
$63 1506 1512 1495 1485 1485 1504 1864 1791 1798 2740 2835 2804 2803 2658 37 19.8 20.0 20.3 20.4 20.6 21.5 22.8 24.4 25.6 27.9 30.2 32.1 32.9 32.7 32.7 31.3
A 1.3 1.4 1.4 1.5 1.5 1.6 1.6 1.8 1.9 2.1 2.2 2.3 2.3 2.4 2.4 2.1
SD 21.3 21.7 22.4 23.0 23.6 23.9 24.5 26.0 28.4 31.0 33.2 33.8 33.3 31.5 29.5 28.6
P 1.6 1.5 1.5 1.5 1.5 1.5 1.6 1.8 2.0 2.1 2.1 2.1 2.2 2.4 2.2 2.7
SD 93.3 92.0 90.6 88.4 87.2 89.7 93.0 93.6 89.6 89.7 90.5 95.0 98.8 103.8 110.6 109.7
A/P 6.7 5.4 5.4 5.6 5.8 6.1 5.7 6.1 6.6 6.3 6.1 6.1 6.4 7.2 9.0 11.2
SD &.,
Table I. Means and standard deviations of vertebral dimensions (mm) and ratios (8) for Japanese-American because some subjects did not receive thoracic radiographs at the first examination.
SD ___ 9.3 5.4 4.6 4.5 4.5 4.7 5.1 5.9 6.0 5.9 5.3 4.3 4.8 6.1 6.4
102.3 103.2 102.3 102.4 101.5 102.7 106.4 109.2 109.3 107.0 101.7 98.8 94.6 93.6 93.6
in Hawaii.
___
women
98.1 97.2 97.2 98.0 98.7 97.6 94.3 91.7 91.8 93.7 98.5 101.4 105.9 107.3 107.8
I!&+,
6.5 5.1 4.6 5.4 4.5 4.4 4.5 4.9 5.1 5.1 4.6 4.4 5.7 7.3 9.3
SD
The sample size varies with region of the spine
P. D. Ross et al.: Comparison
110
Table II. Differences Differences
of vertebral dimensions
in anterior and posterior heights between Japanese-American women and Caucasian women. are given in units of millimeters, calculated as (Caucasian value) minus (Japanese-American
value). Values at the bottom of the table represent the means and SDS of the 10 to 15 values in each COlUlM.
Study: Vertebra T3 T4 T5 T6 T7 T8 T9 TlO Tll T12 Ll L2 L3 L4 L5 Mean difference: SD of differences0.5
P Davies Minng Gallaaher
A Davies
Minne Bllaaher
__ __ -_ __ 1.8 1.4 1.4 1.5 1.8 2.2 2.0 1.6 2.1 3.1 __
__ 1.9 2.0 2.1 2.4 2.2 2.2 2.5 2.9 3.0 2.7 2.3 2.8 2.9 4.0
2 2 2 2 1 0 0 2 1 1 2 2 2 1 1
__
__
__ __ __ 1.5 1.5 1.5 1.4 1.1 1.2 1.7 0.5 1.5 2.7 __
1.2 1.1 1.1 1.0 1.1 1.5 1.6 1.3 1.1 1.7 1.2 2.3 3.1 1.9
2 1 2 2 1 2 1 1 2 1 1 1 2 1 0
1.9
2.6 0.5
1.4 0.7
1.5 0.6
1.5 0.6
1.3 0.6
Furthermore, no relationship between stature and A/P or (Pi/ P, _ ,) ratios could be demonstrated. Thus, the differences in A/P ratio among studies do not appear to be explained by the differences in stature between races. The differences in specific vertebral heights were sometimes greater between Caucasian studies than between Japanese and Caucasians (for example, the values of A for T8 and T9, Table II). These data suggest that each study may have used a different
procedure to select the vertebral margins. Differences in margin selection have been reported to influence the calculated vertebral area (Nelson et al. 1990), but have not been reported for vertebral height measurements. For the study of Minne et al., differences in the values of A were somewhat greater than in the other Caucasian studies, but the average values of P were not, suggesting that the method used to mark the anterior dimensions may have been different. Details of the procedures used to mark
Table III. Differences
in anterior/posterior percent ratios between Japanese-American women and Caucasian women. Differences were calculated as (Caucasian A/P % value) minus (Japanese-American A/P % value). Values at the bottom of the table represent the means and SDS of the 10 to 15 values in each column.
Caucasian Group: Vertebra T3 T4 T5 T6 T7 T8 T9 TIO Tll T12 Ll L2 L3 L4 L5
Davies
Melton
Smith
Gallaaher
__ __ __ __ 2.2 0.9 0.0 1.1 3.5 3.8 4.5 3.9 1.8 1 .o __
__ 4.9 4.9 4.7 5.0 3.5 1.7 2.4 3.8 4.5 6.5 6.3 4.6 3.3 3.7
__
-1 1 1 1 1 1 2 2 1 2 3 2 0 1
Mean difference: SD of differences:
2.3 1.6
4.3 1.3
2.6 2.0
1 2 1 4 2 0 1 2 6 5 4 4 0 5
4
1.4 1.2
P. D. Ross et al.: Comparison
111
of vertebral dimensions
Table IV. Differences in posterior height percent ratios between Japanese-American women and Caucasian women. Differences were calculated as (Caucasian % value) minus (Japanese-American % value). Values at the bottom of the table represent the means and SDS of the 10 to 15 values in each column.
Caucasian Group: Vertebra! T3 T4 T5 T6 T7 T8 T9 TIO Tll T12 Ll L2 L3 L4 L5 Mean difference: SD of differences:
y&l-
&&+1-
Melton
Melton
__ __ __ __ __ -0.3 0.1 -1.2 -1.3 0.0 -1.6 -0.5 2.9 3.8 _-
__ _-
__
2.1 1.7 0.5 1.6 0.9 -0.8 1.0 -1.8 -2.9 0.6 3.1 2.2 -2.5*
-1.9 -0.8 -0.5 -1.4 -0.7 0.7 -0.8 1.4 2.8 -0.3 -2.7 -2.3 0.9* __
0.2 1.9
0.4 1.9
-0.4 1.6
*It was assumed that ratios in the original report were inverted (see reference for explanation).
vertebral borders on radiographs were not specified in most reports. Based on the illustration in the paper, it appears that the procedure used by Gallagher et al. (1988) should yield the same results as the method used in our study, which differed only in that the borders were marked with points, and lines were not drawn. However, the possibility remains that interobserver bias exists between the two studies, which would only be detected by exchanging and marking many films, possibly hundreds. It may be worth noting that the smallest differences between Japanese and Caucasians were consistently noted for the study of Gallagher et al. In contrast to our study and that of Gallagher et al., Hurxthal recommended that posterior prominences be excluded from measurements, which would reduce values of P by about 1 to 2 mm. Although the exact procedure was not specified, if Minne et al. had used IHurxthal’s procedure, this might explain in part why the values elf A, but not P, were greater than those of the other Caucasian studies. Alternatively, true differences in vertebral heights (and ratios) may exist among Caucasian populations because of differences in genetic and/or environmental factors. Our findings suggest that vertebral shapes may also differ between Japanese and Caucasians, since the A!P and Pi/Pi- 1 ratios were different. Within one comparison population, these differences were not consistent, varying in magnitude with location in the spine (‘Tables III and IV). Moreover, both the average and vertebra-specific differences varied considerably, depending on the Caucasian study chosen for comparison. This suggested that there could also be differences in radiographic technique, position of markings for dimensions, and/or shapes of the vertebrae among the Caucasian studies, as well as between this study of Japanese-Americans and the Caucasian studies.
The position of the X-ray tube would be expected to influence the values of A and P, because the magnification effect will increase with distance of the vertebra from the hypocenter of the tube. The A/P ratio would be expected to be influenced much less, since both the A and P measurements of a single vertebra should have a similar degree of magnification. The ratio of Pi/Pi - 1 might be influenced somewhat, because since adjacent vertebrae will be subject to different magnification factors, especially at the margins of the film. Increasing distance from the hypocenter will also increase the problem of parallax, making it more difficult to define vertebral margins, and possibly biasing the measured dimensions. Davies et al. (1989) used a tube-to-film distance of 102 cm (40 inches), centered over L3 for lumbar views, and over T7 for thoracic views. In the study of Minne et al. (1988), the tube was 105 cm (41.3 inches) from the film, centered over L3 and T9, respectively; similar conditions were used in our study of Japanese-Americans (see Methods). A tube-to-film distance of 105 cm (41.3 inches) was used by Melton et al. (1989), but centering position was not specified. Gallagher et al. (1988) used a tube-to-film distance of 40 inches, centered over T9 for thoracic views, and L2 for lumbar views. Information regarding tube position was not available for the study of Smith et al. (1987). Gallagher et al. (1988) reported that an increase of 10 cm (4 inches) in tube-to-film distance reduced the measured dimensions by approximately 6%, or 1.9 mm for an initial measurement of 32 mm. Thus, the differences in tube-to-film distance of 2.5 cm (1 inch) or less, do not completely explain the observed variations in A and P values among studies. Likewise, there are no obvious patterns in any of the vertebral dimension parameters related to the differences in position of the X-ray beam hypocenter on the spine between studies.
112
P. D. Ross et al.: Comparison
Differences in patient thickness among studies could have influenced the measured dimensions, by affecting the distance between the vertebrae and the film, resulting in different magnification factors. Unfortunately, characteristics regarding patient thickness were not specified in any of the studies compared here. Some of the observed differences in A/P ratios (Table III) were comparable in magnitude to the SD values shown for A/P measurements in Table I. If a cutoff of -2 SD from one population were to be applied to a second population, an error of this magnitude (1 SD) would result in misclassifying approximately 14% of the population as having vertebral fractures. In other situations, the proportion of people misclassified could be greater or smaller, depending upon both the cutoff selected and the true magnitude of the difference between the populations. As an example, we applied the “adjusted minimum” cutoff values reported by Davies et al. (1989) to our population. Using their criteria resulted in an apparent prevalence of crush fractures at L4 and L5 of 8% and 18%, respectively. These figures are much higher than expected, based on the report of Melton et al. (1989). Furthermore, a radiologist who reviewed our films from the first examination found the prevalence of crush fractures to be less than 1% at L4 and L5 (data not shown). In condlusion, it appears that the vertebral heights of our Japanese-Americans are 1 to 2 mm shorter than Caucasians. It remains uncertain how Japanese and Caucasian vertebrae may differ in shape. Further research is needed regarding technical, genetic, and other factors (e.g., birth cohort, age) that may influence vertebral dimensions and ratios. At present, it is not advisable to use reference ranges developed for one population for classifying fracture prevalence in other populations.
Acknowledgmenrs: We thank the for excellent technical and clerical
entire staff of the Osteoporosis Center support.
of vertebral dimensions
References Davies, K. M.; Reeker, R. R.; Heaney, R. P. Normal vertebral dimensions and normal variation in serial measurements of vertebrae. J. Bone Min. Res. 4:341-349; 1989. Gallagher, J. C.; Hedlund, L. R.; Stoner, S.; Meeger, C. Vertebral morphometry: Normative data. Bone Miner. 4189-196; 1988. Jensen. G. F.: McNair. P.; Boesen, J.; Hegedus, V. Validity in diagnosing osteoporosis. Eur. J. Radial. &l-3; 1984. Melton, L. J. III; Km, S. H.; Frye, M. A.; Wahner, H. W.; O’Fallon, W. M.; Riggs, B. L. Epidemiology of vertebral fractures in women. Am. .l. Epidemiol. 129:1000-1011; 1989. Minne, H. W.; Leidig, G.; Waster, C.; Siromachkostov, L.; Baldauf, G.; Bickel, R.; Saw, P.; Lojen, M.; Ziegler, R. A newly developed spine deformity index (SDI) to quantitate vertebral crush fractures in patients with osteoporosis. Bone Miner. 3:335-349; 1988. National Center for Health Statistics; Najjar, M. F.; Rowland, M. Anthropometric reference data and prevalence of overweight, United States 1976-80. Vital and health statistics. Series 11, No. 238. DHHS F’ubl. No. (PHS)87-1688. Public Health Service. Washington, DC.: U.S. Government Printing Office, Oct. 1987. Nelson, D.; Peterson, E.; Tilley, B .; O’Fallon, W. M.; Chao, E.; Riggs, B. L.: Kleerkoper, M. Measurement of vertebral area on spine X-rays in osteoporosis: Reliability of digitizing techniques. J. Bone Min. Res. 5707-716; 1990. Ross, P. D., Wasnich, R. D.; Heilbmn, L. K.; Vogel, .I. M. Definition of a spine fracture threshold based upon prospective frachlre risk. Bone 8:271-278; 1987. Smith, R.; Cummings, S. R.; Genant, H. K.; Steiger, P.; Reeker, R. The study of osteoporotic fractures study group: Overdiagnosis of vertebral fractures: the need for vertebra specific criteria. Christiansen, C.; Johansen. J. S.; Riis, B. I.; eds., Osreoporosis 1987. Copenhagen: Osteopress; 1987; 106-108. Yano, K.; Wasnich, R. D.; Vogel. J. M.; Heilbnm. L. K. Bone mineral measurements among middle-aged and elderly Japanese residents in Hawaii. Am. J. Epidemiol. 119~751-764, 1984.
Dare Received: April 9, 1990 Dare Revised: October 3, 1990 Dare Accepted: October 19, 1990
