The surgical decision-making process for the treatment of adolescent idiopathic scoliosis (AIS) begins by clinical evaluation of the patient and the deformity, as well as careful review of appropriate long cassette spinal radiographs. The clinical components include a careful history, review of systems, and physical examination. The history should exclude other potential etiologies of scoliosis and also provide assessment of the degree of skeletal maturity. The physical examination carefully documents features such as shoulder heights, trunk shift, coronal decompensation, and evidence of upper thoracic, main thoracic, and thoracolumbar/lumbar prominences on forward bend evaluation. The pertinent preoperative radiographs include the upright coronal and lateral views of the entire spine, as well as right and left side bending flexibility radiographs at a minimum. Additional radiographs may include traction, push-prone, fulcrum bend in the coronal plane and hyperextension lateral radiographs in the sagittal plane. 13,16,19
When planning a scoliosis fusion operation, the surgeon initially must decide what particular region(s) of the spine require(s) surgical fusion versus those that may be left unfused. 3,15,17 Curve classification should help in this task for a classification system should help predict treatment. An ideal method would similarly classify curves that have radiographic similarities that promote consistent surgeon agreement on the decision of surgical fusion regions and specific fusion levels of the spine. Traditionally, the King classification has been the primary means of classifying adolescent idiopathic thoracic scoliosis. 5 It has provided important landmark principles such as the role of selective thoracic fusion in King II curves 9,14 and identification of structural proximal thoracic regions for the diagnosis of double thoracic curve patterns (King Type V). 5 Recently, several shortcomings of the King classification have been pointed out by two independent studies published simultaneously. 2,7 These shortcomings include poor intraobserver and interobserver reliability in using the classification system; the incompleteness of the system with primary thoracolumbar and lumbar, double major, and triple major curves being excluded; the coronal plane only evaluation of curve types; and the frequent difficulty in fully distinguishing between a King Type II and King Type III curve pattern, among others. 2,7
Recently, a new, comprehensive two-dimensional classification system of AIS has been developed and presented. 8 This new classification system consists of six curve types that are based on the regional columns of the spine (proximal thoracic [PT], main thoracic [MT], thoracolumbar/lumbar [TL/L]) that are divided into structural or nonstructural by specific, objective coronal and sagittal radiographic criteria (Table 1). In this manner six curve types are created: Type 1, MT, when only the MT region is structural (major); Type 2, double thoracic, when the PT and MT regions are both structural but the MT is the major curve; Type 3, double major, when both the MT and TL/L regions are structural but the MT region is the major curve (largest Cobb measurement); Type 4, triple major, when the PT, MT, and TL/L are all structural (MT or TL/L major); Type 5, TL/L, when only the TL/L region is structural (major); and Type 6, TL/L–structural MT, when both the TL/L and MT regions are structural but the TL/L region is the major curve (largest Cobb measurement) (Table 2). The classification is treatment based in that the structural regions as described above should be included in the instrumentation–fusion but the nonstructural regions should not.
After identification of the curve type, a lumbar curve modifier and sagittal thoracic modifier are added to form the complete triad classification system. The lumbar curve modifier (A, B, C) is based on the position of the apex of the lumbar spine to a center sacral vertical line (CSVL). The CSVL is drawn as a line bisecting the sacrum drawn vertical and parallel to the lateral edge of the radiograph. For lumbar modifier A this line falls between the lumbar pedicles up to the stable vertebra; for lumbar modifier B this line touches the apex of the lumbar curve between the medial edge of the pedicle to the lateral edge of the apical body or bodies; and for lumbar modifier C the CSVL falls completely medial to the apex of the lumbar curve whether the apex is a disc or a vertebral body (Figure 1). Lastly, a sagittal thoracic modifier is added based on the sagittal alignment between T5 and T12: “−” hypokyphotic when T5–T12 is <10°, “N” normal kyphosis for T5–T12 between 10° and 40°, and “+” hyperkyphotic for T5–T12 >40° (Figure 1). Curve classification thus combines the curve type (1–6), lumbar spine modifier (A, B, C), and sagittal thoracic modifier (−, N, +) (e.g., 1A−, Figure 1). Although this new classification system has been shown to be reliable among the developers, 11 it is unknown how reliable those not associated with the development would classify specific AIS cases.
Currently, in the surgical treatment of AIS, there are three distinct choices for the operative approach (anterior spinal fusion [ASF], posterior spinal fusion [PSF], or both) and multiple choices for the selection of specific fusion levels. In two previous studies much variability occurred for both the selection of operative approaches and especially for choice of fusion levels in AIS cases. 7,12
The purpose of this multisurgeon assessment was to: 1) evaluate the ability of a group of scoliosis surgeons, not involved in the development of a new classification system of AIS, to choose the corresponding curve type and classification as described in a new classification system of AIS; 2) and to evaluate the variability of a group of scoliosis surgeons for the selection of an operative approach and both proximal and distal fusion levels in accordance with a new classification system with a radiographic and clinical case study method.
At a recent spinal surgery meeting, an AIS roundtable discussion was held. Twenty-eight scoliosis surgeons were chosen and presented seven preselected cases of operative AIS chosen by the lead author via good quality slides. The slides projected included the following: long cassette upright coronal and lateral, right and left side bending radiographs, along with posterior upright and forward bend clinical photographs. Cobb measurements of all coronal curves were provided, as well as sagittal thoracic kyphosis (T5–T12) and lumbar lordosis (T12-sacrum). Surgeons were given a brief (5 minutes) introduction to a new classification system of AIS and had a handout outlining details of the classification system for their own individual use (Figure 2). Ample time was provided to review the radiographs and clinical pictures, as directed by the scoliosis reviewer audience. A handout was provided that allowed each surgeon assessor to circle the specific curve type (1–6), lumbar spine modifier (A, B, C), and sagittal thoracic modifier (−, N, +). In addition, each surgeon reviewer circled the operative approach they would use (ASF, PSF, or both) and wrote in the proximal as well as distal fusion levels (Figure 1). The sheets were then collected and each of the seven cases reviewed. The appropriate classification was then discussed and postoperative posteroanterior and lateral upright radiographs and postoperative clinical photographs were presented for each case. This highlighted the surgical treatment as recommended by the classification system with structural regions instrumented/fused and nonstructural regions left unfused.
The results for the triad classification system assessment of the seven cases presented can be broken down into specific curve type (1–6), lumbar spine modifier (A, B, C), and sagittal thoracic modifier (−, N, or +). Table 3 shows the results for the curve type as presented, the percent of reviewers choosing the corresponding curve type for that specific case; the lumbar modifier as well as the percent choosing the corresponding lumbar modifier for that specific case; and lastly, the sagittal thoracic modifier as well as the percent of reviewers choosing the corresponding sagittal thoracic modifier. Of the seven cases presented, four were MT, Type 1 (cases 1, 2, 5, and 7), one case was a double thoracic, Type 2 (case 3), one was TL/L, Type 5 (case 4), and one was a TL/L–structural MT, Type 6 (case 6). Overall, the reviewers successfully chose the curve type 84% of the time, with a range from 58% (case 2, MT Type 1), to 100% (case 1, MT Type 1; and case 3, double thoracic Type 2). The lowest percent selection of curve types (58%, case 2) belongs to a curve where both the thoracic and lumbar curves cross the midline, with the thoracic curve the major curve and the lumbar curve a minor nonstructural curve. This is often considered a double major curve pattern, but by this new classification system because the lumbar curve was a minor nonstructural curve, this is classified as a Type 1C MT curve that underwent a selective thoracic fusion.
For the lumbar curve modifiers there were three Type A modifiers (cases 1, 3, and 4), one Type B modifier (case 7), and three Type C modifiers (cases 2, 5, and 6). Overall, 86% of the reviewers chose the lumbar modifier with a range from 58% (case 6, Type C lumbar modifier) to 100% (case 1, Type A lumbar modifier). The lowest percent selection of lumbar modifier (58%, case 6) was a case with a major lumbar curve and a minor structural thoracic curve. Forty-two percent of the reviewers selected the lumbar A modifier, which can only be used as the thoracic curve is the major curve. Overall, there were two hypokyphotic or “−” sagittal thoracic modifiers presented (cases 1 and 2), and five normal kyphotic or “N” modifiers presented (cases 3, 4, 5, 6, and 7). Overall, 90% of sagittal thoracic modifiers were chosen by the reviewers with a range from 79% (case 6, normal sagittal thoracic modifier “N”) to 96% (case 4, with a normal sagittal thoracic modifier “N”). Thus, this multisurgeon review by surgeons not involved in the development of this new system found a fairly high rate of selecting the curve classifications with 84% of the curve types, 86% of lumbar modifiers, and 90% of the sagittal thoracic modifiers chosen by the reviewers. The overall success rate of complete curve classification combining the curve type, lumbar modifier, and sagittal modifier approached 80% (79.4.%).
Surgical Decision Making
Twenty-eight surgeon reviewers chose the operative approach consisting of ASF, PSF, or both, as well as specific proximal and distal fusion levels that they would perform for these seven AIS cases. In five of the cases the PSF approach was the most popular, ranging from 42% for case 7 (Type 1, MT) to 80% for case 6 (Type 6, TL/L–structural MT). The ASF approach was the most popular for case 4 (Type 5, TL/L), with 80% of the reviewers choosing this approach. Lastly, both ASF and PSF were chosen by 58% of the reviewers for case 3 (Type 2, double thoracic). In no case was the same approach chosen by all the reviewers, four cases had two different approaches chosen, and three cases had three different approaches, with the most variability given for case 7 (ASF 29%, PSF 42%, both 29%) (Table 4).
There were a variety of proximal fusion levels chosen for the seven individual cases. There was an average of five different proximal (range 4–8) fusion levels chosen for the seven cases. The individual choice of proximal fusion levels ranged from as low as 5% to as high as 69% for individual proximal fusion levels in the seven cases reviewed. Case 5 had the greatest variability with eight different proximal fusion levels chosen, with case 3 having the least amount of variability with only four different proximal fusion levels chosen and 78% of the reviewers choosing T2 as the proximal fusion level (Table 5).
For the distal fusion levels chosen by the 28 reviewers, there was also great variability with an average of four distal (range 3–5) fusion levels chosen by the reviewers for the seven operative cases presented. The least amount of agreement was for cases 2 and 7, in which five different distal fusion levels were chosen with a range from 10% (L2) to 30% (T12) for specific fusion levels for case 2, and 6% (L2) to 33% (T12) for specific distal fusion level for case 7 (Figure 2). Overall, the greatest agreement was noted for L3 being the distal fusion level chosen by 77% of reviewers for case 4; this case had three different distal fusion levels chosen (Table 5). An example of one of the cases presented (case 7) highlights the excellent agreement on classification of this Type 1BN curve (91% chose curve Type 1, 95% lumbar modifier B, and 90% sagittal modifier N) with highly variable choices for the operative approach (ASF 29%, PSF 42%, both 29%) and proximal (T3– 6%, T4– 33%, T5– 44%, T6– 11%, T8– 6%), and distal (T11– 22%, T12– 33%, L1– 12%, L2– 6%, L3– 28%) fusion levels.
The decision-making process for the surgical treatment of AIS can be broken down into a basic algorithm starting with initial clinical patient and radiographic evaluation. Next, curve classification is performed to isolate regions of the spine to be included in the instrumentation–fusion. Finally, the surgeon must select the appropriate operative treatment consisting of the approach (ASF, PSF, or both) and the selection of both proximal and distal fusion levels. A logical goal would be to objectively characterize each step along the way as an attempt to create a consistent and reliable method of patient and radiographic evaluation leading to appropriate curve classification and finally the selection of the “best” surgical treatment. This is a form of evidence-based analysis whereby treatment decisions are made based on outcomes. In the past this has been nearly impossible to consider because of certain flaws in the initial radiographic evaluation and classification. 2,7 The two important findings of this study are that it is possible to have scoliosis surgeons come to a reasonable consistent agreement on the specific criteria necessary for curve classification with a new, comprehensive, two-dimensional system and that the variability in surgical decision making in operative AIS as we enter the 21st century is still extremely variable. It would be helpful if a classification system could not only be developed and used appropriately but also help direct specific surgical treatments. To select the best treatment, a scoring or grading system will be required to objectively separate out various treatments of the same or similar curve types. Only then will a dedicated surgical treatment algorithm be able to objectively determine the “best” operative treatment for a particular curve pattern. 1,4,6,10,11
We had several specific goals when starting out to produce a new classification system of AIS: that it would be reliable, treatment based, two-dimensional to include the sagittal plane, have objective criteria to separate out the curve types, be reliable both for intraobserver and interobserver testing, be comprehensive for all AIS curves not just thoracic curves, and be simple and practical for scoliosis surgeons. 8 In this current study we tested this system on a group of scoliosis surgeons not involved in the development and found surprisingly good accuracy of curve classification of the individual components and entire system. Thus, based on this current AIS case presentation model, it is fairly easy for surgeons not previously versed in this system to learn it quickly.
Evaluating variability in the selection of operative approach and fusion levels has been performed in the past. At the Scoliosis Research Society meeting in 1991, Lonstein presented a series of 28 problematic cases reviewed by 12 different surgeons who selected the King classification system and chose appropriate fusion levels. 12 Although reliability data were not presented, much variability occurred in both. In a previous study Lenke et al presented the results of an eight-member Scoliosis Research Society group who reviewed 27 operative cases of AIS. 7 They found that high variability occurred in the selection of both the appropriate King classification as well as the proximal and distal fusion levels. As an example, only one of 27 cases had the same operative approach and fusion levels chosen by all eight Scoliosis Research Society reviewers. Our current report further reinforces the need for a classification and grading system of AIS that helps group similar curves together to critically and objectively evaluate the variable treatments that are used for each particular curve pattern. The goal is to determine the “best” treatment method for particular curve patterns that can be objectively scored from multiple perspectives including radiographs, cosmesis, a patient self-administered questionnaire, and functional aspects (PFTs, gait, trunk range of motion, etc.). However, this can only be accomplished if agreement exists on curve classification that groups similar curves in a similar fashion. After accurate and reliable curve classification, the best choice for the specific approach, as well as the proximal and distal fusion levels, can then be objectively determined.
The interobserver and intraobserver reliability of this new system has recently been tested by two groups: the five originators of this system and seven randomly selected Scoliosis Research Society active members. Both groups viewed the same 27 operative cases used to assess the reliability of the King system. The interobserver and intraobserver kappa results for the curve type for the originator group were 0.85 and 0.83, respectively. The interobserver and intraobserver reliabilities for the randomly selected surgeon group were 0.74 and 0.80, respectively. A recognized standard of good to excellent reliability using kappa values is κ > 0.75; thus, all these values fell within this good to excellent range except the interobserver reliability of the randomly selected group (κ = 0.74), which fell just below. 18 Because of the large number of reviewers and small number of cases reviewed, the generation of kappa statistics for this study is suboptimal because of the inherent potential variability of the reviewer group (large number) versus the actual cases reviewed (smaller number).
In conclusion, in this multisurgeon review by those not part of the development of a new classification system of AIS, there was a high rate of agreement on selecting specific criteria necessary for curve classification, with individual components approaching 90% and the entire classification approaching 80%. In addition, the high variability of operative approaches and fusion levels further reinforces the need for both a classification and grading system of AIS that allows similar curves to be grouped together to critically and objectively evaluate the variable treatments used for each particular curve pattern. Only then will objective criteria be able to determine the “best” surgical treatment for each AIS patient.
- A new classification system of AIS was used by surgeons not involved in the development of the system.
- There is currently high variability in the selection of the operative approach and both proximal and distal fusion levels for AIS cases among various scoliosis surgeons.
- There is a need for a method to critically and objectively evaluate these variable surgical treatments to determine the best radiographic and clinical outcome.
The authors thank Karen Steger-May, MA, Department of Biostatistics, WA University School of Medicine, St. Louis, Missouri.
1. Betz RR, Harms J, Clements DH, et al. Comparison of anterior and posterior instrumentation for correction of adolescent idiopathic scoliosis. Spine 1999; 24: 225–39.
2. Cummings RJ, Loveless EA, Campbell J, et al. Interobserver reliability and intraobserver reproducibility of the system of King et al for the classification of adolescent idiopathic scoliosis. J Bone Joint Surg Am 1998; 80: 1107–11.
3. Kalen V, Conklin M. The behavior of the unfused lumbar spine following selective thoracic fusion for idiopathic scoliosis. Spine 1990; 15: 271–4.
4. Kaneda K, Shono Y, Satoh S, et al. Anterior correction of thoracic scoliosis with Kaneda anterior spinal system. Spine 1997; 22: 1358–68.
5. King HA, Moe JH, Bradford DS, et al. The selection of fusion levels in the thoracic idiopathic scoliosis. J Bone Joint Surg Am 1983; 65: 1302–13.
6. Lenke LG, Betz R, Harms J, et al. Spontaneous lumbar curve coronal correction after selective anterior or posterior thoracic fusion in adolescent idiopathic scoliosis. Spine 1999; 24: 225–39.
7. Lenke LG, Betz RR, Betz R, et al. Intraobserver and interobserver reliability of the classification of thoracic adolescent idiopathic scoliosis. J Bone Joint Surg Am 1998; 80: 1097–106.
8. Lenke LG, Betz RR, Harms J, et al. A new and comprehensive classification system of adolescent idiopathic scoliosis. J Bone Joint Surg Am 2000; 83: 1169–81.
9. Lenke LG, Bridwell KH, Baldus C, et al. Preventing decompensation in King type II curves treated with Cotrel-Dubousset instrumentation. Spine 1992; 17 (suppl): 274–81.
10. Lenke LG, Bridwell KH, Baldus C, et al. Cotrel-Dubousset instrumentation for adolescent idiopathic scoliosis. J Bone Joint Surg Am 1992; 74: 1056–67.
11. Lenke LG, Bridwell KH, Blanke K, et al. Radiographic results of arthrodesis with Cotrel-Dubousset instrumentation for the treatment of adolescent idiopathic scoliosis. J Bone Joint Surg Am 1998; 80: 807–14.
12. Lonstein J. Decompensation with Cotrel-Dubousset instrumentation: a multi-center study. Orthop Trans 1992; 16: 158.
13. Luk KD, Cheung KMC, Lu DS, et al. Assessment of scoliosis correction in relation to flexibility using the fulcrum bending correction index. Spine 1998; 23: 2303–7.
14. McCane SE, Denis F, Lonstein JE, et al. Coronal and sagittal balance in surgically treated adolescent idiopathic scoliosis with the King II curve pattern: a review of 67 consecutive cases having selective thoracic arthrodesis. Spine 1998; 23: 1063–73.
15. Moe JH. A critical analysis of methods of fusion for scoliosis: an evaluation in two hundred and sixty-six patients. J Bone Joint Surg Am 1958; 40: 529–54.
16. Polly DW, Sturm PF. Traction versus supine side bending: which technique best determines curve flexibility? Spine 1998; 23: 804–8.
17. Shuffelbarger HL, Clark CE. Fusion levels and hook patterns in thoracic scoliosis with Cotrel-Dubousset instrumentation (CDI). Spine 1990; 15: 916–20.
18. Svanholm H, Starklint H, Gundersen HJ, et al. Reproducibility of histomorphologic diagnoses with special reference to the kappa statistic. APMIS 1989; 97: 689–98.
19. Vaughan JJ, Winter RB, Lonstein JE. Comparison of the use of supine bending and traction radiographs in the selection of the fusion area in adolescent idiopathic scoliosis. Spine 1996; 21: 2469–73.