Radiol Phys Technol (2015) 8:312–319 DOI 10.1007/s12194-015-0322-5
Absolute reliability of adipose tissue volume measurement by computed tomography: application of low-dose scan and minimal detectable change—a phantom study Takahiro Onuma1 • Tamotsu Kamishima2 • Tsukasa Sasaki3 • Motomichi Sakata2
Received: 27 March 2015 / Revised: 28 May 2015 / Accepted: 29 May 2015 / Published online: 9 June 2015 Japanese Society of Radiological Technology and Japan Society of Medical Physics 2015
Abstract Metabolic syndrome increases the risk of developing diabetes and cardiovascular disease, particularly heart failure. Abdominal obesity is commonly assessed by measurement of the waist circumference, which exhibits a positive correlation with the visceral fat area measured on computed tomography (CT). CT is an excellent technique for measurement of cross-sectional areas of adipose tissue, but the exposure to ionizing radiation limits broad and repeated application in healthy subjects. Our purpose in this study was to determine the reliability of low-dose CT for abdominal fat quantification as compared with a standard CT protocol. A phantom was scanned by use of changes in the volume of vegetable oil, simulating visceral and subcutaneous adipose tissue, and by changes in the tube current–time products (25–300 mAs). We measured the volume of vegetable oil for each mAs value, and we calculated the minimal detectable change (MDC) in the volume by making repeated measurements. The measured volume of vegetable oil at 50 mAs and higher was not significantly different (p [ 0.05), but that at 25 mAs was significantly different (p \ 0.001), from that at 300 mAs. The MDC was less than 0.4 ml regardless of the mAs value at all mAs values assessed. We suggest that the adipose tissue volume is
& Tamotsu Kamishima
[email protected] 1
Graduate School of Health Sciences, Hokkaido University, North-12 West-5, Kita-ku, Sapporo 060 0812, Japan
2
Faculty of Health Sciences, Hokkaido University, North-12 West-5, Kita-ku, Sapporo 060 0812, Japan
3
Department of Clinical Support for Medical Practice, Hokkaido University Hospital, North-14 West-5, Kita-ku, Sapporo 060 0812, Japan
determined accurately by CT at 50 mAs (75 % reduction of radiation exposure compared with the standard dose). Keywords Computed tomography (CT) Low-dose scan Minimal detectable change (MDC) Phantom study
1 Introduction In recent years, the relationship of obesity to various diseases has been unraveled, and the concept of metabolic syndrome was introduced in defining people at high risk for developing lifestyle-related diseases. Metabolic syndrome increases the risk of developing diabetes and cardiovascular disease, particularly heart failure [1–3]. International diagnostic criteria for metabolic syndrome have been established by the World Health Organization (WHO) [1], the National Cholesterol Education Program (NCEP) [2], and the International Diabetes Federation (IDF) [3]. Abdominal obesity is commonly assessed by measurement of the waist circumference, and which is positively correlated with the visceral fat area measured on computed tomography (CT) [4]. There are some studies on adipose tissue volume measurements [5–7], made by use of lowdose scans [8, 9], and in the relationship between fat volume and the risk of various diseases [10–12]. However, no research about determining the lowest radiation dose and its reliability for fat volume measurement by use of CT has been published. An intraclass correlation coefficient (ICC) is widely used when reliability is examined. This method can assess ‘‘relative reliability’’ [13, 14]. In contrast to relative reliability, a method of examining what kind of and how many variations and errors are mixed into the measurement is ‘‘absolute reliability’’ [13, 14]. One of the methods for
Absolute reliability of adipose tissue volume measurement by computed tomography: application…
evaluating absolute reliability is the minimal detectable change (MDC). MDC is interpreted as the amount of change that is not likely to be due to change variation in measurement [15, 16]. It can be considered as a ‘‘real change,’’ which exceeds the measurement error by a wide enough margin so as not to be considered as a chance result. A change in an observed value that is less than the MDC can be considered similar to an error in measurement [14]. For example, in a comparison of abdominal fat volume between two measurements, where the MDC for abdominal fat volume is calculated to be 10 cm3: when the abdominal fat volume difference is less than 10 cm3, the change is within the measurement error; if the volume difference is more than 10 cm3, the change is considered to be a real change. Our aim in this study was to determine the absolute reliability and MDC of low-dose CT for abdominal fat quantification as compared with those of the standard-dose CT protocol, with use of a phantom.
2 Materials and methods 2.1 Phantom study A phantom was created that simulated the distribution of visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) in the abdomen (Fig. 1a, b). The phantom was constructed of a glass cylinder (diameter 28 cm; length 22 cm), two types of airtight plastic bags (ziploc freezer bag, Asahi Kasei Home Products Corporation, Tokyo, Japan; large size; height 27.3 cm, width 26.8 cm; small size; height 12.7 cm, width 17.7 cm) filled with vegetable oil (mass density was 0.92 g/cm3) or sugar water, stone powder clays, and bubble wrap (height 30 cm; width 30 cm). In this study, because the scans were repeated with increasing amounts of fat, the diameter of the phantom increased as the fat mass was increased. The visceral fat compartment consisted of airtight plastic bags (height 12.7 cm; width 17.7 cm) filled with vegetable oil (5 bags of 400 ml and 1 bag of 300 ml with a total of 2300 ml, to simulate the VAT), an airtight plastic bag (height 27.3 cm; width 26.8 cm) filled with a solution of 200 g of sugar in 1000 ml of water to simulate paravertebral muscles, and 3 columns of stone powder clays (diameter 4 cm; length 15 cm, to simulate vertebral bones). Also, 30 rolls of bubble wrap (rolled diameter of 2 cm to simulate bowel interposed between VATs) were inserted into the cylinder. The subcutaneous fat compartment consisted of oil-filled bags attached to the exterior of the phantom (7 bags of 400 ml and 1 bag of 300 ml with a total of 3100 ml, to simulate SAT). Also, an airtight plastic bag (height
313
Fig. 1 a Photograph, b schematic diagram, c axial image scanned by c computed tomography (CT) with a 120-kVp/300 mAs protocol. d, e Fat volume measurement (300 mAs) for VAT (d) and SAT (e). f Schematic diagram of the phantom after completion of all of the experimental scans (total VAT/SAT of 4300/5100 ml, respectively). In a, b, vegetable oil-filled plastic bags (VO) simulate visceral adipose tissue (VAT, yellow color) or subcutaneous adipose tissue (SAT, yellow color), separated by the wall of the glass cylinder. Sugar water-containing bags (SW, green color) and bubble wrap (*, blue color) and stone powder clay (SPC, red color) simulate muscle, intestine, and vertebral bones, respectively. All of these structures are visualized in the axial CT image (c). d, e Fat volume measurement (300 mAs) for VAT d and SAT e. VAT regions (internal cylinder) can be displayed by mouse clicking on the monitor d. Similarly, SAT (outside the cylinder) can be chosen (e). f Schematic diagram of the phantom after completion of all of the experimental scans (total VAT/ SAT of 4300/5100 ml, respectively; VO of the shaded area). VAT and SAT were arranged in the upper portion inside the phantom and on the outside surface of the phantom, respectively
27.3 cm; width 26.8 cm) filled with a solution of 200 g sugar in 1000 ml of water to simulate the abdominal muscle wall was put on the exterior of the phantom. Furthermore, additional 20 oil-filled bags (200 ml in small plastic bags) were prepared for later increments. The resulting volumes of oil in the internal cylinder (VAT) and the external bags (SAT) were 2300–4300 mL and 3100–5100 mL, respectively. 2.2 CT scan The phantom was scanned with an 80-detector-row CT scanner (AquilionTM PRIME/Beyond Edition, Toshiba Medical Systems Corporation, Tokyo, Japan). The scans were obtained at varying tube current–time products (mAs) from 25 to 300 mAs (7 conditions: 25, 50, 100, 150, 200, 250, and 300 mAs). Other than the mAs, the scanning parameters were kept constant for all spiral CT scans (tube voltage, 120 kVp; gantry rotation time, 1.0 s; detector pitch, 65; beam pitch, 0.8125; reconstruction thickness, 5 mm; field of view (FOV), 460 mm; matrix size, 512 9 512 pixels). The procedure of this experimental CT scanning is demonstrated below. The phantom (VAT 2300 ml and SAT 3100 ml) was scanned at 300 mAs, and then at decreasing mAs. Subsequently, both VAT and SAT were increased, each by 200 ml at the same time, and scanned at the 7 conditions of different mAs. This was repeated until the VAT/SAT increment was up to 2000 ml or a total VAT/SAT of 4300/5100 ml, respectively (Fig. 1). Thus, 77 kinds of images were obtained. Figure 1a shows the location of the phantom in relation to the bed. Figure 1c shows the representative CT image (VAT 2300 ml, SAT 3100 ml). Figure 1f shows the phantom after completion of all of the experimental scans (total VAT/SAT of 4300/5100 ml, respectively). The increment of fat (VAT/
314
T. Onuma et al.
Absolute reliability of adipose tissue volume measurement by computed tomography: application…
SAT of 2000/2000 ml, respectively) is shown shaded. VAT and SAT were arranged in the upper portion inside the phantom and on the outside surface of the phantom, respectively.
315
standard CT radiation dose descriptor was displayed on the CT console just after the CT scan was finished. The CTDIvol values were averaged for each mAs value. 2.5 Statistical analysis
2.3 Fat volume measurements Adipose tissue volumes were measured in each scan with use of commercially available image analysis software (volume analyzer SYNAPSE VINCENT, FUJIFILM Medical Co., Ltd., Tokyo, Japan). A detailed description of the fat volume measurements is given in Fig. 1d, e. Images were reconstructed three dimensionally, and the fat volume was measured after extraction of fat by use of the CT value threshold [-190 to -30 Hounsfield units (HU)] [8, 11, 17– 19]. The fat volume was measured twice at an interval of 1 week by the same radiological technologist. Furthermore, we evaluated the noise level for each condition. We measured the standard deviation (SD) of the CT value for fat in the phantom using a circular region of interest (ROI), as is shown in Fig. 2. The relationship of the SD and fat volume percentage [the percentages of adipose tissue volume on CT images to fat volume prepared in the phantom (%)] was evaluated. 2.4 Radiation dose measurement For radiation dose measurement, the CT dose index volume (CTDIvol) was assessed for each examination. This
All statistical analyses were performed with commercially available software (Microsoft Excel and PASW Statistics 18). The percentages of adipose tissue volume on CT images to fat volume prepared in the phantom at each mAs were compared by Kruskal–Wallis test (7 groups of different mAs settings; each group consisted of 11 data with different fat volumes for the same mAs setting), and multiple comparisons were carried out if significant differences were found. For the MDC calculation, the fat volume was measured twice at an interval of 1 week to prevent contamination of the bias. Subsequently, Bland– Altman analysis [20] was carried out for judging systematic errors, and the MDC was calculated if there were no systematic errors. The MDC generally employed MDC95, which indicates the 95 % confidence interval (CI), and which was calculated from the following formula [13, 21]: MDC95 ¼ standard error of measurement ðSEMÞ 1:96 pffiffiffi 2 pffiffiffi where 1.96 reflects the 95 % CI, and 2 accounts for the additional uncertainty introduced by use of difference scores from measurements obtained at two different time points [15]. The SEM provides a value for measurement error in the same units as the measurement itself [13] and is calculated with the following formula [13, 16, 21]: pffiffiffiffiffiffiffiffiffiffiffi SEM ¼ s 1 r where s is the SD of time 1 and time 2, and r is the reliability coefficient for the test, i.e., Pearson’s correlation coefficient between test and retest values. For all statistical analyses, a p value of \0.05 was considered to indicate significance.
3 Results 3.1 Radiation dose measurement Figure 3 shows CTDIvol for each mAs value. The CTDIvol was proportional to the mAs value (r = 0.999, p \ 0.001). 3.2 Fat volume measurement Fig. 2 Region of interest (ROI) for fat in the phantom. Standard deviation (SD) of the CT value in the fat was measured at the same point using five consecutive images. Yellow circle shows ROI of fat
The percentages of adipose tissue volume on CT images to fat volume prepared in the phantom (%) are shown in
CTDIvol (mGy)
316
T. Onuma et al. 40.00 35.00 30.00 25.00 20.00 15.00 10.00 5.00 0.00
0
50
100
150
200
250
300
350
mAs value
Fig. 3 Relationship of mAs value and CTDIvol. There is a significant positive correlation between mAs value and CTDIvol (r = 0.999, p \ 0.001); The CTDIvol was proportional to the mAs value
Fig. 4 (a, VAT; b, SAT). Data of the first time measurement were used for this analysis. CT images of the phantom at 25 and 300 mAs are shown in Fig. 5. For both VAT and SAT, the percentages of adipose tissue volume on CT images to fat volume prepared in the phantom at each mAs were significantly different (Kruskal–Wallis, p \ 0.001). Multiple comparisons revealed that the percentages of adipose tissue volume on CT images at 25 mAs were significantly smaller than those at the other mAs values for both VAT and SAT (p \ 0.001). There was no significant difference in measured fat volume between 50 mAs and mAs values above 100 mAs (p [ 0.05). Bland–Altman analysis demonstrated that there were no systematic errors between the two sets of measured fat volume at a 1-week interval for both VAT and SAT (Tables 1, 2 VAT, SAT). For the MDC calculation, the fat volume was measured twice at an interval of 1 week to prevent contamination of the bias (the same data as those used for Bland–Altman analysis). The SEM and MDC at each mAs value for VAT and SAT are listed in Table 3. When we consider that the volume of fat prepared for this study was above 2300 ml (VAT, 2300–4300 ml; SAT, 3100–5100 ml), the SEM and MDC values of less than 0.144 and 0.399, respectively, are small enough for all mAs values, when we take the amount of the fat volume (hundreds to thousands milliliters) into consideration in this experiment. Figure 6 shows the inverse relationship of the mAs value and mean SD of the fat in the phantom. Figure 7 shows the inverse relationship of the SD of the fat and fat volume percentage (the determination coefficient is 0.9983).
4 Discussion Our aim in this study was to determine the absolute reliability and MDC of low-dose CT for abdominal fat quantification as compared with those of the standard-dose CT
Fig. 4 Percentage of measured adipose tissue volume to true volume at each mAs (a VAT; b SAT). Upper horizontal line of box, 75th percentile; lower horizontal line of box, 25th percentile; horizontal bar within box, median. VAT visceral adipose tissue, SAT subcutaneous adipose tissue, VAT percentage the percentage of measured VAT volume to true volume at each mAs, SAT percentage the percentage of measured SAT volume to true volume at each mAs. *p \ 0.001; **p [ 0.05
protocol, with use of a phantom. Among the various doses, the lowest dose at which the visceral and subcutaneous fat volume could be measured as accurately as at the standard dose was about 75 % lower than the standard dose. The MDC for repeated measurement of visceral and subcutaneous fat volume was less than 0.4 ml for all doses; small enough compared with the measured fat volume. Therefore, our phantom study suggested that the minimal required dose for abdominal fat volume measurement was about 75 % lower than the standard dose. This result is similar to the results of prior research. Yoon et al. [8] reported that accurate measurement of abdominal adipose tissue volumes is possible with 90 kVp
Absolute reliability of adipose tissue volume measurement by computed tomography: application…
317
Fig. 5 CT images of the phantom at 25 (a and c) and 300 (b and d) mAs. The regions in green show the area recognized and measured as fat by the workstation (a and b, visceral adipose tissue; c and d,
subcutaneous adipose tissue). The fat regions are less homogenously colored at 25 mAs compared with those at 300 mAs
Table 1 Summary of Bland–Altman analysis for VAT
Table 2 Summary of Bland–Altman analysis for SAT
MAs value
Mean of averaged difference (ml)
95 % CI (ml)
r*
p value
MAs value
Mean of averaged difference (ml)
95 % CI (ml)
r*
p value
300
0.016
-0.125 to 0.157
-0.461
0.154
300
0.083
-0.170 to 0.337
0.314
0.348
250
0.025
-0.120 to 0.170
-0.475
0.140
250
0.036
-0.256 to 0.328
-0.112
0.742
200
0.021
-0.138 to 0.179
-0.078
0.819
200
0.043
-0.176 to 0.262
-0.209
0.537
150
0.001
-0.182 to 0.185
0.304
0.364
150
0.016
-0.187 to 0.219
-0.145
0.670
100
-0.008
-0.163 to 0.147
0.506
0.112
100
0.000
-0.158 to 0.159
-0.221
0.513
50
-0.024
-0.333 to 0.285
-0.155
0.650
50
0.000
-0.090 to 0.089
0.164
0.630
25
0.025
-0.218 to 0.269
0.109
0.750
25
0.019
-0.148 to 0.185
-0.068
0.842
For each mAs value, there was no correlation on the Bland–Altman analysis (p [ 0.05)
For each mAs value, there was no correlation on the Bland–Altman analysis (p [ 0.05)
VAT visceral adipose tissue, CI confidence interval, r* Pearson’s correlation coefficient between values of first and second measurements
VAT visceral adipose tissue, CI confidence interval, r* Pearson’s correlation coefficient between values of first and second measurements
(about 5 mGy, or a 73.4 % dose reduction compared with 140 kVp), and Rogalla et al. [9] reported that fat volume can be measured at 4 mGy. Hence, the lowest dose at
which fat volume can be measured accurately is around 4–5 mGy. However, these studies did not evaluate the absolute reliability of low-dose CT such as the MDC.
318
T. Onuma et al.
Table 3 SEM and MDC at each mAs value for SAT and VAT MAs value
VAT (ml)
SAT (ml)
SEM
MDC
SEM
MDC
300
0.027
0.075
0.093
0.258
250
0.028
0.078
0.130
0.360
200
0.038
0.106
0.072
0.199
150
0.049
0.136
0.062
0.172
100
0.032
0.088
0.037
0.104
50
0.144
0.399
0.012
0.033
25
0.090
0.251
0.042
0.118
SEM and MDC of less than 0.144 and 0.399, respectively, are small enough for all mAs values, when we take the amount of the fat volume (hundreds to thousands milliliter) into consideration in this experiment SEM standard error of measurement, MDC minimal detectable change, VAT visceral adipose tissue, SAT subcutaneous adipose tissue
16 14
mean SD
12 10 8 6 4 2 0
25
50
100
150
200
250
300
mAs value
Fig. 6 Relationship of mAs value to mean standard deviation of fat. CT values were measured at the same point in five continuous images. Yellow outline shows region of interest (ROI) of fat
fat volume percentage (%)
99.5 99.0 98.5 98.0
Conflict of interest of interest.
97.5 y = -0.0313x2 + 0.401x + 97.921 R² = 0.9983
97.0 96.5
Our study is clearly different from the previous studies. The main purpose in our study was to determine the absolute reliability of low-dose CT; the scans were obtained at various fat volumes as well as mAs values; moreover, for MDC calculation, the fat volume was measured twice at an interval of 1 week to prevent contamination of the bias. Although the resulting dose reduction was similar to that in previous studies, in the case of assessment methods for clinical applications, evaluation of the absolute reliability would provide better information.Absolute reliability is important because it is able to evaluate what kind of and how many variations and errors are mixed into the measured value [13, 14]. Hence, the assessment of the interval change between the baseline value and the follow up or post-intervention value tends to become more reliable. There are several reasons why the measured fat volume was underestimated at 25 mAs. First, artifacts with higher HU values from simulated vertebrae and ribs were prominent over the fat. The second factor is increased noise, which is in inverse proportion to the square root of the mAs. The limitation of this study is the difference in constitution between our phantom and the human body; we created a phantom of a simplified model in terms of adipose tissue distribution using vegetable oil in a plastic bag; however, in the human body, adipose tissue is distributed in a more complex manner, especially for visceral fat. Moreover, the effect of artifacts from the motion of muscles, the bowel wall, and the abdominal wall was not reproduced in our phantom study. We need to conduct a human study of adipose tissue volume measurements using low-dose CT. In conclusion, we determined the absolute reliability and MDC of low-dose CT for abdominal fat quantification as compared with those of the standard-dose CT protocol with use of a phantom. Our results suggested that the adipose tissue volume is accurately and reliably determined by CT with a 75 % reduction of the radiation exposure compared with the standard dose.
0
2
4
6
8
10
12
The authors declare that they have no conflict
References 14
16
mean SD
Fig. 7 Relationship of standard deviation of fat to fat volume percentage. R2 is determination coefficient. Fat volume percentage decreases as the inverse square of the SD value (determination coefficient is 0.9983)
1. Alberti KG, Zimmet PZ. Definition, diagnosis and classification of diabetes mellitus and its complications. Part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation. Diabet Med. 1998;15:539–53. 2. Report F. National Cholesterol Education Program. Arch Intern Med. 1991;151:1071.
Absolute reliability of adipose tissue volume measurement by computed tomography: application… 3. Alberti KGMM, Zimmet P, Shaw J. Metabolic syndrome–a new world-wide definition. A consensus statement from the International Diabetes Federation. Diabet Med. 2006;23:469–80. 4. Pouliot MC, Despre´s JP, Lemieux S, Moorjani S, Bouchard C, Tremblay A, et al. Waist circumference and abdominal sagittal diameter: best simple anthropometric indexes of abdominal visceral adipose tissue accumulation and related cardiovascular risk in men and women. Am J Cardiol. 1994;73:460–8. 5. Kobayashi J, Tadokoro N, Watanabe M, Shinomiya M. A novel method of measuring intra-abdominal fat volume using helical computed tomography. Int J Obes Relat Metab Disord. 2002;26:398–402. 6. Jensen MD, Kanaley JA, Reed JE, Sheedy PF. Measurement of abdominal and visceral fat with computed tomography and dualenergy X-ray absorptiometry. Am J Clin Nutr. 1995;61:274–8. 7. Nemoto M, Yeernuer T, Masutani Y, Nomura Y, Hanaoka S, Miki S, et al. Development of automatic visceral fat volume calculation software for CT volume data. J. Obes. 2014;2014:495084. 8. Yoon DY, Moon JH, Kim HK, Choi CS, Chang SK, Yun EJ, et al. Comparison of low-dose CT and MR for measurement of intraabdominal adipose tissue: a phantom and human study. Acad Radiol. 2008;15:62–70. 9. Rogalla P, Meiri N, Hoksch B, Boeing H, Hamm B. Low-dose spiral computed tomography for measuring abdominal fat volume and distribution in a clinical setting. Eur J Clin Nutr. 1998;52:597–602. 10. Nam SY, Choi IJ, Ryu KH, Park BJ, Kim HB, Nam B-H. Abdominal visceral adipose tissue volume is associated with increased risk of erosive esophagitis in men and women. Gastroenterology. 2010;139(1902–11):e2. 11. O’Leary DP, O’Neill D, McLaughlin P, O’Neill S, Myers E, Maher MM, et al. Effects of abdominal fat distribution parameters on severity of acute pancreatitis. World J Surg. 2012;36:1679–85.
319
12. Rosenquist KJ, Pedley A, Massaro JM, Therkelsen KE, Murabito JM, Hoffmann U, et al. Visceral and subcutaneous fat quality and cardiometabolic risk. JACC Cardiovasc Imaging. 2013;6:762–71. 13. Donoghue D, Stokes EK. How much change is true change? The minimum detectable change of the Berg Balance Scale in elderly people. J Rehabil Med. 2009;41:343–6. 14. Fritz SL, Blanton S, Uswatte G, Taub E, Wolf SL. Minimal detectable change scores for the Wolf Motor Function Test. Neurorehabil Neural Repair. 2009;23:662–7. 15. Steffen T, Seney M. Test-retest reliability and minimal detectable change on balance and ambulation tests, the 36-item short-form health survey, and the unified Parkinson disease rating scale in people with parkinsonism. Phys Ther. 2008;88:733–46. 16. Haley SM, Fragala-Pinkham MA. Interpreting change scores of tests and measures used in physical therapy. Phys Ther. 2006;86:735–43. 17. Sjo¨stro¨m L, Kvist H, Cederblad A, Tyle´n U. Determination of total adipose tissue and body fat in women by computed tomography, 40 K, and tritium. Am J Physiol. 1986;250:E736–45. 18. Brandberg J, Lonn L, Bergelin E, Sjostrom L, Forssell-Aronsson E, Starck G. Accurate tissue area measurements with considerably reduced radiation dose achieved by patient-specific CT scan parameters. Br J Radiol. 2008;81:801–8. 19. Dicker D, Herskovitz P, Katz M, Atar E, Bachar GN. Computed tomography study of the effect of orlistat on visceral adipose tissue volume in obese subjects. Isr. Med. Assoc. J. 2010;12:199–202. 20. Martin Bland J, Altman D. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;327:307–10. 21. DE Beaton. Understanding the Relevance of Measured Change Through Studies of Responsiveness. Spine (Phila. Pa. 1976). 2000;25:3192–9.
