Early-onset scoliosis (EOS) is defined as scoliosis with onset less than the age of 10 years, regardless of etiology.1–3 Growth friendly surgery allows for spine and chest growth while controlling the spine and chest deformity.3 To assess ongoing spine growth of patients undergoing growth friendly surgery, the most common approach is to serially measure the standard-of-care vertical spine height (SoCVH) measured on coronal radiographs. These length measurements are 1-dimensional and are usually performed from first thoracic vertebra to first sacral vertebra (T1-S1), although alternatively measured from T1 to T12 if examining only the thoracic spine region.4,5 An example of a thoracic SoCVH measurement is shown in Figure 1.
The Children’s Spine Study Group (2015) has published that rib-based distraction surgeries maintain 75% of the expected T1-S1growth up to the age of 10 years.6 It is hypothesized that traditional coronal plane measurements of spine length may not account for out-of-plane growth (ie, kyphosis or scoliosis), underestimating of the growth effects of the surgical interventions. Sankar and colleagues found that the use of spine-based distraction implants performed well at controlling the coronal plane deformity of the scoliosis, but were found to exhibit a “law of diminishing returns” in the increase in vertical spine height imparted at each follow-up spine expansion surgery. His group attributed the diminishing returns solely to autofusion that limited the ability for the spine to continue to grow with repeated surgical expansions.7 This paper provoked a trend toward the use of serial casting as a “delay tactic” for the treatment of EOS8; however, it did not comment on the effects of repetitive distractions on sagittal plane growth or alignment. As it has been demonstrated that spine-based distraction surgeries are kyphogenic by nature,6,9,10 there may be increases in length (other than in the coronal plane) that is not represented by traditional measurements, underrepresenting the growth effects of spine-based distraction surgery.
We hypothesize that, due to these limitations of the standard of care vertical height measurement, the increase in kyphosis of these patients is in fact growth out of the coronal plane, which may be occurring at the expense of vertical height. Therefore, the “law of diminishing returns” may not only be a loss of spine growth related to autofusion, it may in fact be a diminishing return of the measurement technique; that the tissues of the spinal column are still increasing in length (ie, the vertebrae are growing, with their endplates growing farther apart). To account for this potential “loss of growth,” we developed a custom software program to measure the true 3-dimensional curved arc length of the spine [3-dimensional true spine length (3DTSL)]. The purpose of this study was to validate this new technique for measuring the true 3-dimensional length of the spine for patients with EOS.
Using the LabView software environment (National Instruments, Austin, TX), a research engineer created a custom 3-dimensional spine curve length measurement program. The first iteration of this experimental software has been validated by the research group and has been adopted in the EOS field for 2-dimensional sagittal spine length (SSL) measurements.11–13 The current iteration of the spine length measurement program added a module for 3-dimensional curve measurement functionality. The program calibrates the radiographic images and allows the user to select the desired spine section to be measured (any contiguous section between T1 and S1). The accuracy and impact of measuring the spine in 3 dimensions has not been previously investigated.
The calculation for measuring a spine section is shown in following equation:
where X is the lateral pixel position of the point on the coronal radiograph, Y is the vertical pixel position of the point on the coronal and lateral radiograph (C or L subscript, respectively), Z is the lateral pixel position of the point on the lateral radiograph, C is the pixel size at the point on the coronal radiograph, and L is the pixel size at the point on the lateral radiograph. Each segment is summed to calculate the full 3-dimensional length of the spine section.
Physical Model Measurements
To assess the absolute accuracy and usefulness of the 3DTSL measurement tool and dual calibration process, accuracy, and interrater reliability (IRR) tests were conducted on physical rod model. Ten alignment configurations with varying degrees of “coronal” and “sagittal” curvatures were constructed using contoured 1/8 inch (3.2 mm) brass rods. Notches were inscribed around the circumference of the rods segmenting them into 9 measurement segments of 30 mm, for a total rod length of 270 mm. The configurations of these rods can be seen in Figure 2. For each model, the curved rod length was measured, segment by segment, from top to bottom, using high accuracy digital linear callipers (±0.01 mm; Mastercraft Tools, sold through Canadian Tire Ltd, Welland, Canada). The length measurement was repeated 4 times around the circumference (at 90 degrees intervals) of the rod with the average of the 4 segment measurements taken as the true length of the model segment.
Photographic Physical Model Measurements
The photographic images of the rods were assessed by 5 reviewers using the developed custom software and a proprietary advanced image calibration method to remove magnification error associated with measuring distances on plain film radiographs.14 The rod models were sequentially placed vertically in the center of the dual calibration imaging area and photographed using high resolution Canon Rebel T4i (EOS 650D) digital camera (Canon Inc., Tokyo, Japan) mounted on a stationary tripod approximately 1.4 m from the center of the rod stand. The rods were photographed once in the “coronal” plane and then rotated 90 degrees clockwise and photographed again in the “sagittal” plane.
In each of the images, the reviewers used the etched marks on the rod models as measurement points on the “coronal” and “sagittal” images. The images and measurements were calibrated using the novel advanced calibration method developed for the 3DTSL measurement. This advanced calibration creates a set of volumetric calibration equations which varies pixel dimensions throughout the volume, removing any discrepancy between the magnification of the measured object and the calibration plane. This removes a persistent issue with plain film radiography which can account for a significant source of error in radiographic distance measurements.14 An analysis of variance test was used to examine the difference between the 5 reviewer measurements and those of the physical caliper measurements. Interclass coefficients (ICC) were used to assess agreement of measurements between reviewers. Analysis of variance and ICC tests were conducted using IBM SPSS Statistics version 20 (IBM Corporation, Armonk, NY).
Clinical Radiographic Measurements
To examine the IRR of the measurement under clinical radiographic conditions, 6 reviewers measured the curved spine lengths of the coronal and sagittal radiographs of 23 preoperative patients with EOS from our institution. The vertical distance between the start and end points of the curved spine paths recorded on coronal radiographs were used to determine the SoCVH. The generation of the 3DTSL from the 2 biplanar curved spine lengths is demonstrated in Figure 3. The SoCVH and 3DTSL were measured from T1 to L1 using our custom measurement software to standardize the level selection between all reviewers and limit the measured section to a single sagittal curve. All clinical radiographs were anonymized and reviewers were blinded to the radiographs clinical prominence to remove any impact of previous clinical experience the reviewers may have with the radiographs. Standardized image pixel sizes were used to calibrate all distance measurements. ICC assessments were used to determine the agreement and repeatability of the clinical radiographic measurements. Measurements were then repeated by all reviewers at a minimum of 2 weeks after initial data collection to examine measurement repeatability.
Examining the impact of a curved spine measurement, the difference between the standing SoCVH and the 3DTSL was examined using the matched measurements recorded simultaneously by the custom software. Reviewer spine length measurements were matched to focus on the difference between the height and curved lengths. This eliminates potential errors associated with the selection of measurement end points. Paired t tests were used to assess agreement of measurements between the SoCVH and 3DTSL measurements.
Photographic Physical Model Measurement Assessment
Five reviewers measured the 3DTSL of the 10 physical rod models. The mean physical length of the rods was 267.70 mm (SD=0.86; range, 266.73 to 269.21), whereas the mean photographically measured 3DTSL was 268.0 mm (SD=1.8; range, 264.3 to 270.3). The mean absolute measurement error between the reviewer’s photographic 3DTSL and the physical measurements was 1.2 mm (SD=0.9; range, 0.0 to 3.0). This difference was not statistically significant between the physically measured rod lengths and the photographically measured rod lengths (P=0.998). Nor were there significant differences between any of the reviewer’s measured rod lengths. Relatively, the error rate between the photographic and physical measurements was only 0.4% (SD=0.5%; range, 0.0% to 1.1%).
The physical models showed excellent IRR of 0.999 [95% confidence interval (CI), 0.996-1.000], whereas the mean IRR was 0.997 (95% CI, 0.990-0.999).
Clinical Radiographic Measurement Assessment
Average participant age was 5.6 years (range, 1.3 to 9.5 y) with 12 males and 11 females. The group had a mean Cobb angle of 68 degrees (range, 22 to 102 degrees) and a mean kyphosis of 37 degrees (range, 5 to 85 degrees). The participants were generally heterogeneous with the following etiologies using Vitale’s Classification for Early Onset Scoliosis: 7 syndromic, 7 congenital, 6 idiopathic, and 3 neuromuscular.1
The clinical radiographic image analysis showed excellent IRR with an average ICC of 0.952 (95% CI, 0.882-0.982) for the 3DTSL and 0.975 (95% CI, 0.913-0.989) for the SoCVH. The average ICCs for IRR were 0.944 (95% CI, 0.826-0.979) for the 3DTSL and 0.965 (95% CI, 0.910-0.986) for the SoCVH. All P-values for all ICC values were <0.001.
The mean curved lengths of the participant’s spines were 193.9 mm (SD=30.0; range, 142.8 to 276.8) and 156.1 mm (SD=29.7; range, 74.7 to 207.3) for the 3DTSL and SoCVH, respectively. The mean difference for all reviewers between the matched measured 3DTSL and SoCVH on the clinical radiographs was 37.8 mm (SD=21.4; range, 1.3 to 95.4) and was statistically significant (P<0.0001). On average, the 3DTSL of the measured spines was 124.2% of the measured SoCVH.
The spine length data (kyphotic and scoliotic ranges) is displayed in Table 1. The data show an increasing discrepancy (between 3DTSL and SoCVH), with increasing sagittal and coronal curvatures. This is clearly demonstrated in Figures 4 and 5. Owing to only having 1 patient in the range, the 0.0 to 9.9 degrees range of kyphosis curvatures was excluded from Figure 5.
The current standard-of-care, gold standard vertical height measurement (SoCVH) used to assess on-going growth of the spine during growth friendly EOS surgical treatment examines surgical expansion of the spine in only 1 dimension. This 1-dimensional measurement has demonstrated a “law of diminishing returns” in the return on investment from these ongoing surgical procedures.7 The purpose of this study was to validate a new technique for measuring the 3DTSL for patients with EOS. The use of the 3DTSL measurement technique produced accurate and repeatable results during the physical model study while demonstrating very good interrater and intrarater agreements, under clinical imaging conditions. The use of the novel 3DTSL measurement provides researchers and clinicians with unprecedented information on the true length of the spine, unconstrained by the plane of radiographic observation or the verticality of patient positioning. This novel technique will allow for the study of the true growth of the spine during growth friendly surgical EOS treatments which may or may not show that the “law of diminishing returns” is in fact an artifact of the SoCVH measurement’s failing to accurately measure spines which are becoming increasingly kyphotic.
Limitations of this study are mainly related to the radiographic portion of the measurements. Clinically, we are limited by the quality of lateral radiographs to determine landmarks for our measurements. As can be seen in Figure 3, landmarks are much more visible on the coronal image than the sagittal view. However, although we believe that higher image quality of lateral radiographs will increase usability and accuracy of the software in these conditions,15 the most important aspect to generate a highly reliable and repeatable measurement is to identify and select the correct caudal and cranial vertebrae for the spine section and to accurately follow the coronal and sagittal curves of the spine during point selection. This is leeway generated due to the summation of the individual vertebral segments into the spine section whole. An error on a single vertebral segment endpoint creates a condition where 1 segment is measured slightly longer than its physical length while the next segment in the chain is measured slightly shorter by an identical amount. These coupled errors cancel each other out across the entirety of the spine section. Future iterations of the software are planned to use a spline-based line measurement line constructions where the user would have to select the caudal and cranial endplates of the spine section and ensure that the measurement path follows the spinal curves. This will remove these coupled internal length errors in the future.
The use of retrospective radiographic images also excluded the use of the advanced calibration method, whereas the impact of the image magnification was specifically addressed in the physical model using the advanced calibration method. For the clinical images we used the detector pixel size, which will cause all lengths measured to be overestimated by an unquantifiable amount. However, the lack of magnification correction did not affect the interreviewer and intrareviewer reliabilities as a standard calibration was used for all measurements.
All radiographs were captured by a single site, using their standard clinical radiographic patient positioning protocol and there were still significant differences in patient positioning. This diminishes the validity of using a simple, 1-dimensional measurement to assess the complicated 3-dimensional configuration and growth of the spine. With the use of the 3DTSL and the implementation of the advanced calibration method, the standardization of patient positioning will be of less importance as all spinal length, in all directions, will be taken into account, not just vertical height. The 3-dimensional assessment of true spine length will provide a highly accurate determination of “growth” for patients undergoing growth friendly treatments for EOS.
In an effort to keep the measured spine segment standardized for all patients and reviewers, we chose to examine only the thoracic spine in this study. In actual usage, there is no limitation to the spinal levels, which can be measured with the program, as long as they remain contiguous for a single length measurement, and identical in both fields of view.
On the clinical radiographs, there was a significant difference (37.8 mm) in the measured thoracic spine length between the SoCVH and 3DTSL. To put this into clinical context, this represents almost 21% of the average coronal T1-T12 height at the age of 5 years (180 mm)5 whereas at its most extreme, the difference in measurement was 53% of the 5-year thoracic height. In addition, the mean difference between the 2 measurements represents a length more than the total of the average thoracic spine height gained after 15 lengthening surgeries (31 mm) for a group of patients treated with rib-based distraction surgeries.6 As a relative error, disassociated from image magnification effects, this equated to an error of almost one quarter the length of the SoCVH measurement (24.2%).
The difference between the 3DTSL and SoCVH in the clinical radiographs follows an increasing pattern as the kyphosis or Cobb angle increased. There were no statistical differences between the 3DTSL lengths of the different kyphosis or Cobb angle bins in our similar aged cohort, but there was a significant reduction in SoCVH height as the deformities were stratified. This result is a good indication that although deformities are occurring in these patients, their rate of spine growth is maintained. The use of traditional method to measure spine height is likely to have been under representing the “true” spine length as measured in the 3-dimensional environment.
Posterior distraction-based growth friendly implants and expansion surgeries appear to control the coronal deformity well but appear to be kyphogenic by nature.6,9,10,16 The ongoing reduction of apparent growth noticed in these and the Sankar et al7 study is attributed solely to autofusion of the spine defining the “law of diminishing returns.” Much of the ongoing behavior of the SoCVH during distraction-based growth friendly treatments can be corroborated or explained by taking the 3DTSL and spine curvature angles into account. Two specific cases are discussed below.
• The significant height gained at initial implantation demonstrated in the literature may be caused by decreasing the Cobb and kyphosis angles.6,7,11 This height increase does not constitute an increase in spine length, just a better estimation by the limited SoCVH measurement.
Previous research completed by our group found that the large increase in spine height correlated to very little increase in the SSL (the curved length of the spine measured on the lateral radiograph, a precursor to the 3DTSL).11,13
• The reduction in the yield from a given expansion surgery, or “law of diminishing returns” may be correlated with the increase in sagittal curvature.
Our previous research demonstrated that while the yields on the coronal measurements decreased over time, those of the SSL were maintained and followed the expected growth trends of a child.4,11 We theorize that this trend will hold with the 3DTSL as it maintains the measurement of out of plane growth missing from the SoCVH measurement.
The impact of the 3DTSL as a serial growth measurement for EOS patients, undergoing surgical treatment on pulmonary development and function is currently unknown. A study by Dreimann et al17 has shown that patients with large thoracic kyphosis (>60 degrees) have an odds ratio of 20 of having a FVC of 50% or less, whereas Zeng et al18 demonstrated that correction of a kyphotic deformity did not significantly increase the thoracic volume. This seems to show that while thoracic volume is maintained, the ability to utilize it is diminished in the hyperkyphotic spine. Future research focusing on pulmonary function and 3DTSL growth may allow for the optimization implant design or expansion protocols to better maintain or improve pulmonary function.
The use of the true 3-dimensional spine length measurement can provide an accurate and repeatable measure of spine growth that is not currently captured by standard-of-care techniques. By neglecting 2 dimensions, the traditional measurements underrepresents the true length of the spine and any nonvertical growth imparted to EOS patients during spine expansion surgeries. Until the 3DTSL measurement is taken into consideration, the “law of diminishing returns” for spine-based growing rods should be interpreted with caution. This novel measurement technique should complement the current assessment of growth during the treatment of EOS and the inclusion of the 3DTSL into clinical assessment is a first step into examining the full picture of the true spine length, true spine growth, and a more complete analysis of EOS treatment outcomes.
The authors thank the IWK Health Center for its support in our ongoing research and Sara Skentelbery for her help in generating Figure 3.
1. Williams BA, Matsumoto H, McCalla DJ, et al. Development and initial validation of the Classification of Early-Onset Scoliosis (C-EOS). J Bone Joint Surg Am. 2014;96:1359–1367.
2. El-Hawary R, Akbarnia BA. Early onset scoliosis—time for consensus. Spine Deform. 2015;3:105–106.
3. Skaggs DL, Guillaume T, El-Hawary R, et al. Early onset scoliosis consensus statement, SRS Growing Spine Committee, 2015. Spine Deform. 2015;3:107.
4. Dimeglio A. Growth of the spine before age 5 years. J Pediatr Orthop B. 1993;1:102–107.
5. Dimeglio A. Growth in pediatric orthopaedics. J Pediatr Orthop. 2001;21:549–555.
6. El-Hawary R, Samdani A, Wade J, et al. Children’s Spine Study Group. Rib-based distraction surgery maintains total spine growth. J Pediatr Orthop. 2016;36:841–846.
7. Sankar WN, Skaggs DL, Yazici M, et al. Lengthening of dual growing rods and the law of diminishing returns. Spine (Phila Pa 1976). 2011;36:806–809.
8. Fletcher ND, McClung A, Rathjen KE, et al. Serial casting as a delay tactic in the treatment of moderate-to-severe early-onset scoliosis. J Pediatr Orthop. 2012;32:664–671.
9. Schroerlucke SR, Akbarnia BA, Pawelek JB, et al. How does thoracic kyphosis affect patient outcomes in growing rod surgery? Spine (Phila Pa 1976). 2012;37:1303–1309.
10. Dede O, Motoyama EK, Yang CI, et al. Pulmonary and radiographic outcomes of VEPTR (vertical expandable prosthetic titanium rib) treatment in early-onset scoliosis. J Bone Joint Surg Am. 2014;96:1295–1302.
11. Chukwunyerenwa CK, Johnston C, McClung AM, et al. Paper #45: the effect of growth-friendly surgery on coronal and sagittal plane spine growth in idiopathic scoliosis. Spine Deform. 2014;2:517.
12. Mendonca RGM, Samdani A, Smith JT, et al. Children’s Spine Study Group. Paper #7 complications and radiographic outcomes of posterior spinal fusion (PSF) and observation (OB) in patients who have undergone distraction-based treatment for early onset scoliosis (EOS). Spine Deform. 2015;3:614–615.
13. Spurway AJ, Chukwunyerenwa CK, Kishta W, et al. Sagittal spine length measurement: a novel technique to assess growth of the spine. Spine Deform. 2016;4:331–337.
14. Aubin C-E, Bellefleur C, Parent S, et al. Paper #10 importance of calibration for the radiographic assessment of fusionless treatment of pediatric scoliosis. Spine Deform. 2015;3:616.
15. Kubat O, Gauthier LE, Yasin MS, et al. Poster 16-low radiation radiographs improve reliability of upper thoracic sagittal measurements for children with scoliosis. In 9th International Congress on Early Onset Scoliosis.
16. Shah SA, Karatas AF, Dhawale AA, et al. The effect of serial growing rod lengthening on the sagittal profile and pelvic parameters in early-onset scoliosis. Spine (Phila Pa 1976). 2014;39:E1311–E1317.
17. Dreimann M, Hoffmann M, Kossow K, et al. Scoliosis and chest cage deformity measures predicting impairments in pulmonary function. Spine (Phila Pa 1976). 2014;39:2024–2033.
18. Zeng Y, Chen Z, Ma D, et al. The influence of kyphosis correction surgery on pulmonary function and thoracic volume. Spine (Phila Pa 1976). 2014;39:1777–1784.