Evaluation of Wedging of Lower Thoracic and Upper Lumbar Vertebral Bodies in the Pediatric Population

Pediatric Imaging Original Research Gaca et al. Evaluation of Wedging of Lumbar Vertebral Bodies in Children Pediatric Imaging Original Research Ana Maria Gaca 1 Huiman X. Barnhart 2 George S. Bisset, III 1 Gaca AM, Barnhart HX, Bisset GS Keywords: compression, pediatric, trauma, vertebra DOI:10.2214/AJR.09.3065 Received May 18, 2009; accepted after revision August 9, 2009. 1 Department of Radiology, Duke University Medical Center, Erwin Rd., Box 3808, Durham, NC 27710. Address correspondence to A. M. Gaca (ana.gaca@duke.edu). 2 Department of Biostatistics and Bioinformatics, Duke University School of Medicine, Durham, NC. AJR 2010; 194:516 520 0361 803X/10/1942 516 American Roentgen Ray Society Evaluation of Wedging of Lower Thoracic and Upper Lumbar Vertebral Bodies in the Pediatric Population OBJECTIVE. Anterior wedging of thoracolumbar vertebral bodies is often seen in children. The purpose of our study was to show whether mild anterior wedging of pediatric thoracolumbar junction vertebral bodies can be seen as a normal variant, rather than as the sequela of trauma. MATERIALS AND METHODS. A retrospective review was performed of pediatric abdomen and pelvis CT performed between January 2004 and March 2006, excluding children at high risk for compression fractures. Twenty CT studies were randomly selected for each of five arbitrary age groups: < 3, 4 7, 8 11, 12 14, and 15 17 years (100 total CT studies). Using sagittal reformations, anterior and posterior heights of all T10 L3 vertebral bodies were measured to determine anterior-to-posterior vertebral body height ratio (A:P ratio). Inter- and intraobserver agreement was determined. The lower limit of normal (lower fifth percentile of the distribution) was estimated using quantile regression. RESULTS. A:P ratio at the thoracolumbar junction was greater than 0.893 in 95% of children. There was no statistically significant correlation between age and the A:P ratio. There was strong intra- and interobserver agreement. CONCLUSION. From T10 through L3, 95% of children have an A:P ratio greater than 0.893. This suggests that an A:P ratio less than 0.893 should raise the possibility of vertebral body injury. Because age was not statistically significant with respect to the A:P ratio, this value can be used across all pediatric age groups. I njuries to the thoracolumbar spine are rare in children, accounting for approximately 3 4% of traumatic pediatric injuries in pediatric trauma registries (including patients requiring admission or specialist care or trauma fatalities) [1, 2]. These injuries are most commonly related to motor vehicle collisions, with sports-related injuries becoming more common with aging and injuries related to falls and nonaccidental injury becoming less common [3, 4]. The mechanism will determine the injury pattern seen, with compression fractures of the spine typically related to moderate flexion injuries [3]. Compression fractures in children are most common at the thoracolumbar junction and will more commonly involve the superior end plate than the inferior end plate [3]. These are typically treated conservatively, with activity restrictions or possibly orthotic treatment [3]. Mild anterior wedging of lower thoracic and upper lumbar vertebral bodies is of- ten seen in children with no clear history of trauma. This raises the question whether a mild degree of anterior vertebral body wedging may be a normal developmental variant, rather than solely the sequela of trauma. A review of the literature revealed two studies that addressed this question in adults. The first, by Fletcher [5] in 1947, described anterior wedging as a normal variant after measuring vertebral bodies on lumbosacral spine radiographs of 575 male World War II veterans. In this investigation, however, approximately 90% of the radiographs were performed for back pain. The second investigation, by Lauridsen et al. [6], was published in 1984 and studied 164 adult oncology patients. This study suggested that anterior wedging can be seen as a normal variant without clinical significance. Neither of these investigations, however, included children. In 1976, Hegenbarth and Ebel [7] described the radiographic findings of vertebral body fractures in 36 children with a history of trauma. The authors calculated an index of 516 AJR:194, February 2010

Evaluation of Wedging of Lumbar Vertebral Bodies in Children the anterior vertebral body height to posterior vertebral body height, measured from lateral radiographs of the spine. The authors considered an index value less than 0.95 to be pathologic. The basis of this value as a cutoff for normal or abnormal is unclear and appears to be based on the authors experience [7]. Additionally, because some of these patients were asymptomatic at the site of radiographic abnormality, it is unclear whether the radiographic findings represent acute injuries. The goal of this investigation was to determine whether mild anterior wedging of vertebral bodies at the thoracolumbar junction can be seen as a normal developmental variant in children with no identifiable risk factors for compression fractures. Materials and Methods Patients This investigation was approved by our institutional review board, and a waiver of consent was obtained for the retrospective collection of patient information. This study followed HIPAA guidelines regarding patient privacy rights. The PACS at our institution was searched for all abdomen/ pelvis or chest/abdomen/pelvis CT examinations performed on patients 17 years old or younger between January 1, 2004, and March 29, 2006, yielding a total of 1,327 abdomen/pelvis and chest/abdomen/ pelvis CT studies. The electronic records of each patient were reviewed for pertinent medical history, indication for the examination, and age at the time of the examination. All patients considered at high risk for vertebral body wedge compression deformity, including those with a history of trauma, malignancy, or steroid use, were excluded. The remaining 182 patients were then divided into five arbitrary groups on the basis of age at the time of the CT examination: < 3, 4 7, 8 11, 12 14, and 15 17 years. To determine intraobserver measurement agreement, each reader performed two sets of measurements on a subset of 20 patients (four in each age group), with each read separated in time by approximately 4 weeks (process designed to minimize recall bias). Each reader then performed vertebral body measurements on an additional 80 patients (16 randomly selected from each age group), for a total of 100 patients, The average age of the patient at the time of CT was 9.2 years (age range, 1 day 17 years). All CT examinations were performed on either a 16- or 64-MDCT scanner using standardized pediatric weight-based scanning protocols. Indications for the examinations are listed in Table 1. The CT examination for each patient was sent to a workstation (Vitrea 2, Vital Images). Anterior and posterior vertebral body height measurements were TABLE 1: Indications for CT Indication Fig. 1 Screenshot from Vitrea workstation (Vital Images) shows orthogonal views of spine and sagittal reformation used to perform anterior and posterior vertebral body height measurements. (Screenshot courtesy of Vital Images, Inc.) performed on vertebral bodies from the T10 to the L3 levels. Sagittal reformations with bone windows were used, with orthogonal planes to ensure the measurements were obtained at the most midline aspect of the vertebral body (Fig. 1). An anterior-to-posterior vertebral body height (A:P) ratio was recorded. Measurements were performed independently by two staff pediatric radiologists with American Board of Radiology certificates of added No. of Examinations Rule out appendicitis 41 Abdominal pain, not otherwise specified 21 Rule out abscess 14 Pancreatitis 8 Rule out neuroblastoma (all examinations normal) 3 GI bleeding 2 Rule out obstruction 2 Follow-up lesion on prenatal ultrasound (one congenital hepatic cyst, one ovarian cyst) Rule out perforation 2 Fever of unknown origin 1 Precocious puberty 1 Foreign body ingestion 1 Evaluate for Crohn s disease 1 Weight loss 1 qualification in pediatric radiology. These radiologists had an average of 12.5 years of experience. Data and Statistical Analysis Intra- and interobserver agreement was assessed by using a limits-of-agreement method [8]. Specifically, the mean and SD of the differences between the two readings (by the same reader for intraobserver agreement and by two readers for interob- 2 AJR:194, February 2010 517

Gaca et al. server agreement) on the same subject were calculated for each of the six vertebral body locations. The 95% limits of agreement were then constructed based on the mean and SD for each location. If there was no difference on the limits of agreement across locations, the overall limits of agreement were then computed for data with repeated measurements as illustrated by Bland and Altman [8, 9]. The limits of agreement provide two values within which we expect the differences between two single readings to fall for 95% of subjects. If these values for the difference are small relative to the subject matter, then the agreement is acceptable. Intraobserver agreement was determined from the initial 20 subjects. Both readers made measurement twice on each subject. Interobserver agreement was determined from all 100 subjects. If the limits of agreement for interobserver agreement were similar to the limits of agreement for intraobserver agreement, then interobserver agreement was considered strong because the readings by different readers were similar to the multiple readings by the same reader. The fifth percentile of the A:P ratio and its 95% CI were estimated, and the lower limit of the 95% CI was used as the lower limit of normal. Quantile regression was performed to assess the effect of age and vertebral body location on the fifth percentile of the distribution of the A:P ratio. All analyses were performed in SAS version 9.1 (SAS Institute). TABLE 2: Agreement Between Readers in Anterior-to-Posterior Ratio Results Intra- and interobserver agreement were calculated using 95% limits of agreement (Table 2). The limits of agreement were similar across the six vertebral body locations, and the overall limits of agreement were computed combining information from all six locations by using the method for repeated measures [8, 9]. On the basis of the initial 20 subjects, the intraobserver agreement was similar for both readers, with limits of agreement of 0.091, 0.082 and 0.120, 0.104, respectively. This implies confidence that the difference between readings on the same subject by the same reader will be within 0.12 in 95% of the subjects. This value is relatively small, and thus the intraobserver agreement is strong. On the basis of a total of 100 subjects, the interobserver agreement was also very strong, with limits of agreement of 0.128, 0.121, which was similar to the limits of agreement for intraobserver agreement. This implies that the measurements by the two readers on the same subject were similar to the two measurements by the same reader. The mean A:P ratio for vertebral bodies from T10 to L3 ranged from 0.977 to 1.008. There was a statistical difference by vertebral body position (p < 0.001). Although age was statistically significant with respect to the absolute vertebral body measurements, age was not statistically significant with respect to the A:P ratio. Quantile regression showed that vertebral body location was statistically associated with the fifth percentile of the A:P ratio; however, age was not significantly associated. The A:P ratio for each vertebral body level followed a bell-shaped distribution, as seen, for example, at the L3 level (Fig. 2). The average A:P ratio, the estimated fifth percentile, and corresponding 95% CIs at each of the six vertebral body locations TABLE 3: Anterior-to-Posterior (A:P) Ratio Vertebral Body Level Agreement Mean of Two Readings (SD) 95% Limits of Agreement Intrareader difference Reader 1 0.004 (0.044) 0.091, 0.082 Reader 2 0.008 (0.057) 0.120, 0.104 Interreader difference 0.003 (0.064) 0.128, 0.121 Fig. 2 Chart shows distribution of anteriorto-posterior (A:P) ratios for L3 level. A:P ratio follows bell-shaped distribution. Average A:P ratio at L3 level is 1.01 and fifth percentile is 0.93 with 95% CI, 0.90 0.95. No. of Patients Average A:P Ratio are listed in Table 3. The fifth percentiles are similar for vertebral body location, T10, T11, T12, L1, and L2, but are different from the fifth percentile at L3. The quantile regression model showed the overall fifth percentile to be 0.909 (95% CI, 0.893 0.924) and 0.932 (95% CI, 0.903 0.961) for non-l3 (T10 L2) and L3 locations, respectively. By using the lower limit of the 95% CI, the lower limit of normal was 0.893 for non-l3 locations and 0.903 for L3 locations. Because there was no statistical difference in the A:P ratio by age group, the lower limit of normal for the A:P ratio can be used for children of all ages. Estimated Fifth Percentile (95% CI) T10 0.986 0.909 (0.893 0.924) T11 0.977 0.887 (0.861 0.913) T12 0.980 0.905 (0.890 0.921) L1 0.981 0.893 (0.974 0.911) L2 0.991 0.913 (0.893 0.934) Estimated Fifth Percentile Combined (95% CI) 0.909 (0.893 0.924) L3 1.008 0.932 (0.903 0.961) 0.932 (0.903 0.961) All levels 0.990 8 6 4 2 0 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 A:P Ratio 518 AJR:194, February 2010

Evaluation of Wedging of Lumbar Vertebral Bodies in Children Fig. 3 Sagittal reformations of spine in three children. A, Standard quality reformation in 14-year-old boy. B, Sagittal reformation in 4-year-old boy shows blurring of vertebral body end plates. C, Reformation image in 15-year-old girl shows stair-step artifact. Discussion Anterior wedging of vertebral bodies is an important finding in the setting of trauma. It is not rare, however, to see mild anterior wedging of vertebral bodies at the thoracolumbar junction in asymptomatic children. If mild anterior wedging can be seen as a normal variant, misdiagnosis as the sequela of trauma may result in unnecessary imaging of these patients, with unnecessary radiation exposure. This investigation shows that although the average A:P ratio for vertebral bodies at the thoracolumbar junction is close to 1, there is a bell-shaped distribution to these A:P ratios at all levels, with the ratio in some patients closer to 0.9. This suggests that some anterior wedging of vertebral bodies from the T10 to the L3 level can be seen in normal, asymp tomatic patients. This finding is in keeping with those by Masharawi et al. [10] who found anterior wedging of lower thoracic and upper lumbar vertebral bodies when normal adult spine vertebral bodies were measured directly. A bell-shaped distribution of A:P ratios was seen in all the age groups we investigated and was independent of the child s age. Therefore, guidelines for the lower limits of normal for these vertebral A bodies can be used in children of all ages. For the sake of simplicity, an A:P ratio of 0.9 may be used as a lower limit of normal for the vertebral bodies at the thoracolumbar junction (T10 L3), keeping in mind that, as with any estimate for the lower limit of normal, healthy patients may fall below this limit and injured patients may fall above this limit. There are several limitations to our investigation. The total number of patients was small. It is impossible to completely exclude trauma in active children. However, if repeated microtrauma caused the gradual onset of vertebral body wedging, one would expect to see the A:P ratio decrease as a child ages. In this investigation, there was no significant difference, or even trend toward significance, in the average A:P ratio across all age groups or between the youngest and oldest age groups. Because this was a retrospective investigation, we were measuring vertebral body heights on CT scans that were performed up to 3 years before the initiation of our study. Because the original raw data sets were no longer available for these examinations, sagittal reformatted images were obtained using less-detailed data sets. As a consequence, the reformatted images were B occasionally suboptimal, with sporadic artifacts that made vertebral body height measurements more difficult (Fig. 3). Currently at our institution, children undergoing chest, abdomen, and pelvis CT in the setting of trauma typically undergo sagittal and coronal reformations at the time of the examination, and these images are used to evaluate for skeletal injury, avoiding the need for multiple radiographs. The quality of these images would be slightly higher than those used for this investigation, and evaluation for vertebral body wedging would be more accurate. It is speculative to translate measurements from CT to radiographs of the thoracolumbar spine. Obtaining accurate vertebral body height measurements could potentially be more difficult using spine radiographs, particularly if the vertebral bodies are oblique. Radiographs of the thoracolumbar spine are not performed on asymptomatic children; therefore, this investigation would be impractical to repeat using radiographs rather than CT. It is important to note that this information should only be used within the clinical context of the particular child. The L3 vertebral body has the highest average A:P ratio of the measured vertebral bodies at 1.01, with a range of C AJR:194, February 2010 519

Gaca et al. 0.87 1.14. The calculated 95% CI for the L3 vertebral body is 0.90 0.96. Theoretically, a child with an L3 vertebral body A:P ratio at the upper end of the spectrum could suffer a compression wedge deformity that results in a decrease in the A:P ratio from 1.1 to 0.9. Although the A:P ratio would still be within normal limits, it would be abnormal for this patient and result in nearly 20% loss of anterior vertebral body height. In the setting of focal spinal tenderness or neurologic deficit, clinical and radiologic evaluation for vertebral body and spinal cord injury must continue even if the A:P ratio measurements are within the 95% CIs. In conclusion, mild anterior wedging of vertebral bodies can be seen at the thoracolumbar junction in children without a history of trauma, suggesting this may be a normal variant rather than the sequela of trauma. The lower limits of normal for the A:P vertebral body height from the T10 to the L2 levels is 0.89 FOR YOUR INFORMATION and for the L3 level is 0.90. These lower limits of normal are independent of patient age and can be used for all children. Because it may be possible to sustain a wedge compression fracture of a thoracolumbar vertebral body with an A:P ratio that remains above the lower limits of normal described in this article, these limits should be used as guidelines and within the clinical context for any particular child. References 1. Cirak B, Ziegfeld S, Knight VM, Chang D, Avellino AM, Paidas CN. Spinal injuries in children. J Pediatr Surg 2004; 39:607 612 2. Martin BW, Dykes E, Lecky FE. Patterns and risks in spinal trauma. Arch Dis Child 2004; 89:860 865 3. Reilly CW. Pediatric spine trauma. J Bone Joint Surg Am 2007; 89[suppl 1]:98 107 4. Bilston LE, Brown J. Pediatric spinal injury type and severity are age and mechanism dependent. Spine 2007; 32:2339 2347 5. Fletcher GH. Anterior vertebral body wedging: frequency and significance. Am J Roentgenol Radium Ther 1947; 57:232 238 6. Lauridsen KN, De Carvalho A, Andersen AH. Degree of vertebral wedging of the dorso-lumbar spine. Acta Radiol Diagn (Stockh) 1984; 25:29 32 7. Hegenbarth R, Ebel KD. Roentgen findings in fractures of the vertebral column in childhood examination of 35 patients and its results. Pediatr Radiol 1976; 5:34 39 8. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999; 8:135 160 9. Bland JM, Altman DG. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat 2007; 17: 571 582 10. Masharawi Y, Salame K, Mirovsky Y, et al. Vertebral body shape variation in the thoracic and lumbar spine: characterization of its asymmetry and wedging. Clin Anat 2008; 21:46 54 Unique customized medical search engine service from ARRS! ARRS GoldMiner is a keyword- and concept-driven search engine that provides instant access to radiologic images published in peer-reviewed journals. For more information, visit http://goldminer.arrs.org. 520 AJR:194, February 2010