The operative goals of surgery in adolescent idiopathic scoliosis are to prevent progression while providing safe and optimal coronal correction, sagittal alignment, and axial derotation. Certainly, fusing the smallest number of spinal segments possible while obtaining those goals is also desirable to maximize motion segments both above and below the fused spinal segments. This philosophy holds true when fusing into the lumbar spine, because the incidence of back pain has been shown to increase with more distal fusion into the mid and lower lumbar spine. 5,6 Thus, the goal of treating a primary thoracic–compensatory lumbar curve (e.g., King II) is to perform selective thoracic fusion, leaving the lumbar spine unfused, in those cases amenable to the technique. 8
With the initial use of segmental spinal instrumentation systems, postoperative coronal decompensation after selective posterior thoracic fusion of King Type II curves was an occasional complication. 4,12–14,17 Although there are many theories espoused regarding the cause of this problem, the common pattern with postoperative coronal decompensation of the spine is progression of the unfused lumbar curve below a selective thoracic fusion. 4,7 As scoliosis surgeons have gained experience with segmental spinal instrumentation systems, the problem of postoperative decompensation appears to have decreased. However, little has been written about the fate of the residual unfused lumbar spine after successful selective thoracic fusion.
Recently, anterior correction of thoracic scoliosis has been investigated as an alternative to the posterior route. In a recent prospective comparison of anterior versus posterior instrumentation and fusion for thoracic adolescent idiopathic scoliosis, Betz et al 2 noted several advantages to the anterior thoracic approach, including usually fusing one or two levels short of the intended posterior fusion level; having the lowest level of fusion for most thoracic curvatures in the lower thoracic spine and thus completely avoiding fusion into the lumbar spine; and the absence of any coronal decompensation problems after this technique for primary thoracic–compensatory lumbar curves (e.g., King II).
Regardless of the chosen treatment of either anterior or posterior selective thoracic fusion, optimal spontaneous lumbar curve positioning is desirable. In this manner, maximum spontaneous improvement of the lumbar Cobb measurement, and apical translation from the midline, while maintaining, if not improving, global spinal balance, should be the goal. However, the ability to achieve this goal consistently with either posterior segmental or anterior thoracic instrumentation is unknown.
The purpose of this study was to compare similar groups of patients with thoracic adolescent idiopathic scoliosis who had undergone selective thoracic anterior or posterior spinal fusion with instrumentation with regard to the ultimate spontaneous positioning of the unfused lumbar spine. Our hypothesis was that spontaneous correction of the unfused lumbar spine is consistently achieved and is similar after either a selective thoracic anterior or posterior spinal fusion procedure that provides a mobile lumbar spine beneath the fusion.
Materials and Methods
One hundred twenty-three fusions for adolescent idiopathic thoracic scoliosis were investigated: 70 instrumented (Harms 3.2 or 4.0 mm threaded rod) thoracic anterior spinal fusions (ASF), and 53 instrumented (segmental hook–rod system) thoracic posterior spinal fusions (PSF), all with a minimum 2-year follow-up. All 70 anterior procedures were part of a multicenter prospective investigation of anterior versus posterior thoracic fusions. 2 Twenty of the 53 posterior thoracic fusions were part of this same prospective multicenter study group, whereas 33 cases were from the senior author’s institution treated by one of two attending scoliosis surgeons (LGL, KHB).
All cases in both groups had the following characteristics: the diagnosis of primary adolescent idiopathic thoracic scoliosis with varying degrees of compensatory lumbar curves; treatment with a single operative approach to the thoracic curve with the distal fusion ending at T11, T12, or L1; age 11 through 18 when the operations were performed; use of autogenous bone graft in the fusions, either from the ribs (in all anterior cases and in some posterior cases with concomitant thoracoplasties) or posterior iliac crest; and a minimum 2-year follow-up. In most anterior fusions, patients had rib bone obtained by thoracoplasty for fusion and also to allow increased chest wall mobility to perform the open anterior approach. This also may have aided in increasing curve correctability with the anterior approach.
The thoracic ASF procedures all had the distal fusion at the end vertebra of the Cobb measurement of the thoracic curve at either T11 (n = 29), T12 (n = 34), or L1 (n = 7). The thoracic PSF procedures had the lowest level of instrumentation and fusion at the stable vertebra, either T12 (n = 14) or L1 (n = 49). None of the PSF cases had fusion ending at T11.
The reason for using an additional 33 posterior cases from a single institution that was not part of the prospective study of anterior versus posterior thoracic fusions was that most thoracic fusions performed posteriorly in the prospective study group were fused to L2 and below. To establish as fair a comparison as possible between both groups of selective thoracic fusions, we sought to compare similar thoracic curves that had the fusion ending at the thoracolumbar junction, not extending into the lumbar spine below L1. All the PSF procedures were performed in a similar manner, with selective distraction and compression forces without derotation to avoid overcorrection, especially in lumbar modifier C curves (defined below). We thought that this would provide the fairest comparison for evaluating the spontaneous lumbar curve response below a selective thoracic fusion that left most of, if not all of, the lumbar spine mobile.
In an attempt to standardize the positioning of the lumbar curve on the coronal radiograph, all curves were divided before and after surgery into Types A, B, or C according to the preoperative lumbar curve position. These three lumbar curve modifiers: A, B, and C were based on the relation of the center sacral vertical line (CSVL) to the apex of the lumbar spine after review of the standing long-cassette coronal radiograph. The CSVL was drawn with a fine-tip marker as a line that bisected the proximal sacrum, drawn vertically to parallel the lateral edge of the upright radiograph. Any pelvic obliquity less than 2 cm was accepted; if more than a 2-cm pelvic obliquity existed, then the upright coronal radiograph was obtained with an appropriate height lift under the patient’s short leg in an attempt to level the pelvis. The CSVL was drawn proximally until it reached the stable vertebra, which was the most proximal lumbar or lower thoracic vertebra most nearly bisected by the CSVL. If a disc was the most nearly bisected, then the next caudad vertebra was chosen as the stable vertebra. The stable vertebra was the distal level of instrumentation and fusion for all PSF cases. The lower thoracic end vertebra was the distal level of instrumentation and fusion for all ASF cases.
In Type A curves, the CSVL was located between the pedicles at the lumbar apex up to the stable vertebra indicating a nonstructural lumbar curve (Figure 1). If it was difficult to decide whether the CSVL actually touched the medial aspect of the lumbar apical pedicle, then the curve was classified as Type B. Thus, curve Type A included King TypesIII and IV.
In Type B curves, the CSVL touched the apex of the lumbar curve and thus fell between the medial border of the lumbar concave pedicle and the lateral margin of the apical vertebral body or bodies (if the apex was a disc). Lumbar curve modifier B thus demonstrated mild to moderate lateral deviation from the midline of the CSVL (Figure 2). If the physician was in doubt about whether the CSVL actually touched the lateral margin of the apical vertebral body or bodies (i.e., making it difficult to distinguish between Type B and C curves), then the Type B curve was chosen. Type B curves included King Types II and III. Often, we have considered these Type B curves to be King Types “II-and-one-half” reflecting the intermediary position between King Types II and III in the position of the lumbar spine apex relative to the midline. 9
In Type C curves, the CSVL fell completely medial to the concave medial aspect of the thoracolumbar and lumbar apical vertebral body or bodies (if the apex was a disc). Thus, in curve Type C, the lumbar spine at the apex fell completely lateral to the midline (Figure 3). If the physician was in doubt about whether the CSVL actually touched the lateral aspect of the apical thoracolumbar and lumbar vertebral body or bodies, then Type B was chosen as the curve identification. Curve Type C included King Type II curves and an occasional Type V curve that had a lumbar curve Type II pattern in addition to a positive T1 tilt of the proximal thoracic curve. In this study, only primary thoracic curves with a compensatory lumbar curve were evaluated. Thus, only King Types II–V curves were included, whereas King Type I, true double-major, triple-major, and primary thoracolumbar and lumbar curves were excluded. All King Type V curves were treated with posterior instrumentation and fusion because of the known risk of increased shoulder decompensation with an anterior approach for these curves. Of the 123 thoracic curves treated (70 ASF and 53 PSF) 58 were Type A, 48 Type B, and 17 Type C.
Long-cassette coronal radiographs were reviewed on all patients by a single reviewer (LGL). These included the preoperative upright coronal radiographs, right and left side bending radiographs, initial upright postoperative coronal radiographs, and 2-year postoperative coronal radiographs. All thoracic and lumbar curves were measured by the Cobb method. In addition, the preoperative, immediate postoperative, and 2-year postoperative coronal radiographs were distinguished according to whether the lumbar curves were Type A, B, or C based on the position of the CSVL in relation to the apex of the lumbar curve (see earlier description). In this respect, it was possible to determine whether there was a postoperative improvement, deterioration, or no change in the apical deviation of the lumbar curve in relation to the CSVL, after selective anterior or posterior thoracic fusion (i.e., whether a Type C curve before surgery remained a similar or worse Type C, or had less apical translation to achieve a “better” Type C, a Type B, or even a Type A lumbar curve position after surgery).
Statistical analysis was performed on several factors including comparing preoperative thoracic and lumbar curve measurements for ASF versus PSF patients by an independent group’s t test; the three preoperative classes were compared with an analysis of variance (ANOVA) with follow-up pair-wise comparisons by Tukey–Kramer tests; analysis of any differences between the two types of fusions (ASF versus PSF) was accomplished with a simple main-effects test after a factorial ANOVA. All tests results with P < 0.05 were considered statistically significant.
Of the 123 curves treated, 70 were treated with anterior thoracic fusion and 53 with posterior thoracic fusion. For the ASF group, the mean preoperative thoracic Cobb measurement was 57 ± 12.5° (range, 42–104°), decreasing to 33 ± 12.3° on side bending, with instrumented thoracic correction measuring 19 ± 9.1° immediately after surgery and 24 ± 11.7° (range, 6–41°) at 2-year follow-up, for a mean 58% instrumented thoracic correction. For the 53 PSF cases, the mean preoperative thoracic curve was 59 ± 12.4° (range, 38–90°), decreasing to 32 ± 12.2° on side bending, 32 ± 11.4° immediately after surgery, and 36 ± 14.0° (range, 15–57°) at 2-year follow-up, for a mean ultimate 38% thoracic curve correction. The mean percentage of instrumented thoracic correction was statistically superior in the ASF (58%) compared with the PSF (38%) group (P < 0.05).
For the 70 cases treated with ASF, the preoperative lumbar curve averaged 34 ± 10.5° (range, 15–58°), decreasing with side bending to 9 ± 7.6°, with immediate postoperative correction averaging 16 ± 9.9° and the 2-year correction averaging 15 ± 9.2° (range, 0–43°), for a mean 56% spontaneous lumbar curve correction. For the 53 PSF cases, the preoperative lumbar curve averaged 38 ± 12.8° (range, 13–60°), decreasing with side bending to 8 ± 8.8°, with immediate postoperative correction averaging 23 ± 11.6° and the 2-year correction averaging 24 ± 11.6° (range, 0–55°), for a mean 37% spontaneous lumbar curve correction. The mean percentage spontaneous lumbar curve correction was statistically better for the ASF (56%) than for the PSF (37%) group (P < 0.05).
Curves With a Lumbar A Modifier
There were 58 curves that before surgery had a Type A lumbar curve modifier, with 40 of the thoracic curves treated with ASF and 18 treated with PSF. For the ASF group (n = 40), the mean preoperative thoracic Cobb measurement was 57 ± 13.3°, decreasing to 29 ± 12.9° on side bending, with instrumented thoracic correction measuring 18 ± 9.2° immediately after surgery and 22 ± 11.1° at 2-year follow-up, for a mean 61% instrumented thoracic correction. For the 18 PSF Type A cases, the mean preoperative thoracic curve was 52 ± 10.8°, decreasing to 24 ± 10.4° on side bending, and measuring 26 ± 9.1° immediately after surgery and 28 ± 9.1° at 2-year follow-up, for a mean 46% instrumented thoracic correction. The mean percentage of instrumented thoracic correction was statistically superior for the ASF (61%) versus the PSF (46%) group (P < 0.05).
In the 40 Type A cases treated with ASF, the preoperative lumbar curve averaged 30 ± 9.0°, decreasing with side bending to a mean 6 ± 5.1°, with spontaneous lumbar correction to 13 ± 9.5° immediately after surgery and 13 ± 7.3° at 2-year follow-up, for a mean 57% spontaneous lumbar curve correction. In the 18 Type A cases treated with PSF, the preoperative lumbar curve averaged 25 ± 6.6°, decreasing with side bending to 2 ± 5.1°, with spontaneous lumbar curve correction averaging 13 ± 6.0° after surgery and 14 ± 7.1° at 2-year follow-up, for a mean 40% spontaneous lumbar curve correction. The mean percentage of spontaneous lumbar curve correction was statistically better for the ASF (57%) than the PSF (40%) group (P < 0.05).
Curves With a Lumbar B Modifier
There were 48 cases with preoperative lumbar B modifiers, with 23 cases treated with thoracic ASF and 25 cases with thoracic PSF. For the 23 ASF cases, the mean preoperative thoracic curve was 55 ± 10.5°, decreasing with side bending to 35 ± 9.5°, with instrumented thoracic correction of 21 ± 8.5° immediately after surgery and 25 ± 12.6° at 2-year follow-up, for an average 54% instrumented ASF thoracic correction. For the 25 thoracic Type B curves treated with PSF, the mean preoperative thoracic curve was 59 ± 10.9°, decreasing with side bending to 33 ± 11.4°, with instrumented thoracic correction of 33 ± 11.9° immediately after surgery and 34 ± 12.6° at 2-year follow-up, for a mean 42% instrumented PSF thoracic correction. The mean percentage of instrumented thoracic correction was statistically superior for the ASF (54%) versus the PSF (42%) group (P < 0.05).
For the same Type B cases treated with thoracic ASF (n = 23), the mean preoperative lumbar curve was 37 ± 9.2°, decreasing with side bending to 12 ± 7.6°, with spontaneous lumbar curve correction averaging 20 ± 10.2° immediately after surgery and 20 ± 10.1° at 2-year follow-up, for a mean 46% spontaneous lumbar curve correction. For the 25 Type B cases treated with PSF, the mean preoperative lumbar curve was 40 ± 7.1°, decreasing with side bending to 7 ± 7.2°, with spontaneous lumbar curve correction averaging 24 ± 7.8° immediately after surgery and 25° at 2-year follow-up, for a mean 38% spontaneous lumbar curve correction. The ultimate mean percentage of lumbar curve correction was statistically superior for the ASF (46%) versus the PSF (38%) group (P < 0.05).
Curves With a Lumbar C Modifier
There were 17 cases with preoperative lumbar C modifiers, with 7 treated with thoracic ASF and the remaining 10 treated with thoracic PSF. In the seven ASF-treated cases, the preoperative thoracic curve was 65 ± 15.1°, side bending to 43 ± 12.7°, with immediate postoperative correction averaging 24 ± 11.6° and 2-year correction averaging 27 ± 12.9°, for a mean 59% instrumented ASF thoracic correction. For the 10 cases treated with thoracic PSF, the mean preoperative thoracic curve was 67 ± 13.3°, side bending to 40 ± 10.9°, with immediate postoperative correction of 40 ± 8.9° and 2-year follow-up correction averaging 49 ± 14.5°, for a mean 27% correction. The ultimate mean percentage thoracic instrumented correction was statistically superior for the ASF (59%) versus the PSF (27%) group (P < 0.05).
For the Type C curves treated with thoracic ASF (n = 7), the preoperative lumbar curve averaged 42 ± 16.2°, decreasing with side bending to 12 ± 11.6°, with spontaneous lumbar correction averaging 24 ± 4.8° immediately after surgery and 21 ± 9.8° at 2-year follow-up, for a mean 50% spontaneous lumbar curve correction. For the thoracic PSF Type C curves (n = 10), the preoperative lumbar curve averaged 53° ± 9.1°, side bending to 18 ± 9.0° with spontaneous lumbar correction averaging 38 ± 7.9° immediately after surgery and 37 ± 8.3° at 2-year follow-up, for a mean 30% spontaneous lumbar correction. The ultimate mean percentage of spontaneous lumbar curve correction was statistically superior for the ASF (50%) versus the PSF (30%) group (P < 0.05).
Lumbar Curve Apical Positioning
Specifically, when examining lumbar curve apical deviation for Type A curves, 48% of ASF cases and 30% of PSF cases had lumbar curves with a more proximal stable vertebra after surgery, indicating improved lumbar curve coronal alignment. For Type B curves, 75% of the ASF group and 65% of the PSF group had decreased lumbar apical translation to create an improved Type B or even a Type A lumbar curve pattern after surgery. For Type C curves, 71% of the ASF cases and 70% of the PSF cases had decreased lumbar apical translation to create an improved Type C, Type B, or even in a few instances, a Type A lumbar curve pattern after surgery (Figures 4 and 5).
Overall, only one ASF-treated and three PSF-treated lumbar curves had increased apical translation after surgery versus before surgery, indicating an unbalanced lumbar spine. The single ASF-treated case involved a Type A modifier and was fused one level short of the proper lower thoracic end vertebra, with adding on to the thoracic curve after surgery. Of the three PSF-treated cases, all involved Type B modifiers and underwent mild to moderate coronal decompensation of the lumbar curve after instrumented posterior thoracic fusion, which is a known risk of treating a primary thoracic–compensatory lumbar curve with posterior segmental spinal instrumentation of the thoracic curve alone. None of these decompensations was severe enough clinically or radiographically to require revision extension of the instrumentation and fusion into the lower lumbar spine.
When examining mean instrumented thoracic correction, the ASF group had statistically more instrumented thoracic correction for all three groups A, B, and C versus the PSF groups. When evaluating the mean spontaneous lumbar curve correction, the ASF group also had statistically more correction than the PSF group for all three groups A, B, and C. When evaluating the percentage of lumbar curves that had an improved true lumbar apical translation (i.e., preoperative Type B to postoperative Type A), the results in the instrumented ASF group were superior for curve Types A and B, whereas in the ASF and PSF groups, results were nearly equal for Type C curves.
Although most reports of operative treatment of thoracic adolescent idiopathic scoliosis have concentrated on spinal correction achieved over the instrumented thoracic levels, potentially part of the long-term results of the surgical fusion depend on the residual position, balance, and mobility of the unfused lumbar spine. Thus, the goal of any instrumented selective fusion for thoracic scoliosis should be not only to limit the number of fused lumbar segments, but also to optimize the coronal and sagittal position of the unfused lumbar spine. Theoretically, this should minimize both transitional breakdown immediately below the fused region and the rate of development of degenerative changes throughout the unfused lumbar spine.
With the usage of posterior segmental spinal instrumentation systems, there have been several reports discussing the potential adverse effects on the lumbar curve relating to postoperative coronal decompensation after selective fusions for thoracic scoliosis. 4,10,15 These studies have all involved posterior instrumented thoracic correction in primarily King Type II curves. 8 These are primary thoracic curves with compensatory lumbar curves that in a strict sense cross the midline but are not structural enough to warrant inclusion in the thoracic instrumentation and fusion. Ideally, fusing the thoracic curve alone maintains coronal balance and also maximizes long-term lumbar mobility. But occasionally, postoperative decompensations of the lumbar curve below have occurred requiring inclusion of the residual lumbar curve to produce acceptable coronal alignment. This has required extension of the instrumentation and fusion usually into the mid to low lumbar spine (L3 or L4) with fewer residual mobile lumbar segments below the fusion. 9 However, the incidence of postoperative coronal decompensation appears to be decreasing as experience has accrued with both proper curve identification of those Type II curves that can successfully undergo a selective thoracic instrumentation and fusion, as well as appropriate posterior segmental instrumentation techniques that maintain coronal and sagittal balance. 11,15,16
Recently, anterior instrumentation and fusion of thoracic curves has been prospectively investigated by a group of scoliosis surgeons, 2 in a prospective study in which similar types of thoracic curves treated either anteriorly or posteriorly, with different distal levels of instrumentation and fusion into the lower thoracic and lumbar spine were compared. The results of that study show that coronal balance was equal in both the anterior and posterior groups, despite shorter overall fusion levels for the anterior group. However, the resultant effect of these shorter fusions on the unfused lumbar spine was not specifically evaluated.
The current study was designed to determine the spontaneous correction of the unfused lumbar spine comparing an anterior and a posterior approach for similar thoracic curves. The preoperative lumbar curves were grouped according to the preoperative position of the CSVL to the apex of the lumbar spine (lumbar curve modifiers A, B, and C).
Separation of the lumbar curve into modifiers A, B, and C is a means of quantifying the position of the lumbar spine to a more detailed extent than is provided by the King classification alone. 9 In this respect, lumbar modifier A curves are those with minimal apical deviation and rotation indicating nonstructural characteristics to the lumbar curve and thus include King Type III, Type IV, and a few Type V curves. Lumbar curve modifier B has a mild to moderate amount of lumbar apical deviation and rotation, so that the CSVL touches the apex of the lumbar curve in some fashion. These are intermediate curves between true King Type III and IV curves which have no structural characteristics of the lumbar spine and true King Type II curves that have definite structural characteristics with complete lateral deviation of the apex of the lumbar spine from the midline. Type B curves have been classified as either King Type II or Type III, depending on the amount of structural characteristics present in the lumbar spine compared with the thoracic spine; we often have classified those as Types II-and-one-half. 9 Lumbar curve modifier C is a lumbar curve in which the apex completely deviates off the midline and thus has significant structural components of its own. However, in the Type C curves investigated in this study, the thoracic curve had more significant structural characteristics, such as a larger Cobb measurement and greater apical deviation and rotation than the lumbar curve, so that selective thoracic fusion would be appropriate. Thus, these Type C curves were true Type II curves as defined by King 8 and redefined by Lenke et al. 10 There were no King Type I curves in this study group, because the thoracic curve was always the primary (largest Cobb measurement) and most structural curve.
Preoperative classification into Types A, B, and C was also helpful in defining after surgery whether the lumbar spine had changed position with respect to deviation from the midline. In this manner, an improved lumbar A modifier would have a more proximal stable vertebra and thus a straighter lumbar spine; an improved lumbar B modifier would have less apical deviation, so that potentially the lumbar spine at the apex would not touch the apical pedicle and thus become a lumbar A modifier; and a lumbar C modifier would have less deviation so that the CSVL would touch the apex to become a lumbar B modifier, or even in a rare circumstance, would be improved to the point that the CSVL fell entirely within the pedicles at the apex to create a lumbar A modifier.
Our results document that spontaneous lumbar curve correction occurred consistently after both selective thoracic anterior and posterior fusions. Only four spines (one ASF and three PSF) had lumbar curves with greater apical translation after surgery than before. All of the other curves had decreased apical translation and improved Cobb measurements and thus an improved lumbar curve position after surgery. For some of the curves, the spontaneous lumbar curve correction was dynamic and actually improved from the immediate postoperative radiograph to the 2-year postoperative radiograph (47% of ASF cases, and 32% of PSF cases). Interestingly, this improvement in spontaneous lumbar curve correction occurred consistently with a slight loss of instrumented thoracic correction noted 2 years after surgery. It thus appeared that as thoracic curve correction settled into its definitive fused position with slight loss of thoracic correction, spontaneous lumbar curve positioning concomitantly improved as an adjustment. This is precisely the finding the Kalen and Conklin 7 noted in their radiographic evaluation of selective thoracic fusions performed either with Harrington instrumentation or segmental spinal instrumentation. They found that the behavior of the lumbar curve echoed that of the thoracic curve in the frontal plane, with both having nearly equal percentages of correction to maintain proper alignment. They recommended continued use of selective thoracic fusion in King Type II curves in appropriate cases.
Besides their report, most of the literature on spontaneous lumbar curve response after selective thoracic fusion was written in response to problems associated with coronal decompensation after segmental spinal instrumentation. One of the main points of King’s original article describing the five types of thoracic curves (King Types I–V) was to distinguish King Type II curves that would be appropriate for selective thoracic fusion and thus avoid fusing into the lumbar spine. 8 Their criteria appeared to work well for posterior Harrington instrumentation for consistent overall spinal balance, and compensation would ensue. However, when treating King Type II curves with posterior segmental spinal instrumentation, it appears important to select those Type II curves appropriate for selective thoracic fusion. Lenke et al 10 described ratio criteria for thoracic-to-lumbar Cobb measurements, apical translation, and apical rotation that would indicate a successful outcome after selective thoracic fusion. Richards 14 described two types of lumbar curves for a Type II curve, a Type A and a Type B. The Type A curve is more flexible and thus is appropriate for selective thoracic fusion, whereas the Type B curve is stiffer and with potential for complication after selective thoracic fusion. The main emphasis in both of these articles is that for selective thoracic fusion to be successful, the thoracic curve should be the largest and most structural curve with the lumbar curve less structural and thus able to compensate for the thoracic correction achieved.
After a suitable curve is identified for selective thoracic fusion, the scoliosis surgeon must apply appropriate instrumentation principles when using posterior segmental spinal instrumentation. These include using the stable vertebra as the safest distal fusion point and reversing the hooks between neutral and stable on the left-side concave correcting rod (for main right thoracic scoliosis). 3,4,16 Limiting thoracic curve correction so that the lumbar curve can accommodate is also advantageous. 10 Excessive posterior thoracic curve correction may by itself be a causative factor in producing coronal decompensation. 4,10,14 In the current study, however, those curves at highest risk of postoperative decompensation (preoperative lumbar modifier C or true King II curves) had a statistically greater thoracic correction with ASF with no lumbar curve decompensation.
In contrast to the wealth of information on selective thoracic fusion after posterior segmental spinal instrumentation, there is a paucity of information on selective anterior thoracic instrumentation and fusion. However, there are two principles that in theory may allow as good, if not better, lumbar curve response to instrumented anterior thoracic fusion. The first is that the distal level of fusion for anterior thoracic instrumentation is the lower thoracic end vertebra which is normally one or two levels more proximal than the stable vertebra in various types of thoracic curves. Thus, there are usually more unfused thoracolumbar and lumbar segments to accommodate the instrumented anterior thoracic fusion.
Secondly, the correction of thoracic curves anteriorly occurs by convex compression forces, which theoretically, lift up the convex lower thoracic spine and subsequently pull up the concavity of the upper lumbar curve, thereby translating it to the midline. In contrast, the correction forces applied for a posteriorly treated thoracic curve involve concave distraction, derotation, and/or translation forces applied on the lower thoracic region that are transferred to the lumbar spine, potentially causing decompensation. However, it has been advised that reversing the distal hooks into a claw configuration and thereby compressing between the neutral and stable vertebrae at the thoracolumbar junction may lessen the risk of coronal decompensation and improve the sagittal profile and spinal balance. 3,4,16 Although intuitively plausible, no definitive spinal or geometric model has proved or disproved these theories.
The results of the current study demonstrate that both instrumented thoracic and spontaneous lumbar curve correction was statistically better for anteriorly versus posteriorly treated thoracic curves. Also, the percentage of thoracic correction correlated with the percentage of spontaneous lumbar correction in all curves investigated. This was true for each preoperative lumbar curve modifier A, B, and C, with the most difference noted for modifier C curves that are classic King Type II curves. However, it must be noted that lumbar modifier C also included the fewest curve types with only 7 cases treated anteriorly and 10 treated posteriorly. Also, the amount of instrumented thoracic correction in the posterior group was intentionally limited by the surgeon to avoid overcorrecting the thoracic curve beyond that which the lumbar curve could accommodate to avoid postoperative coronal decompensation. These factors probably played a role in the lesser amount of instrumented thoracic correction and potentially the lesser degree of spontaneous lumbar correction in the posteriorly treated Type C curves. Certainly, with more modern instrumentation constructs consisting of distal pedicle screws at the thoracolumbar junction, more posterior correction can be obtained without a high rate of postoperative lumbar curve decompensation. However, it was interesting to note the absence of decompensation problems after selective ASF for lumbar C modifier curves, despite a mean 59% instrumented thoracic correction (versus 29% with posterior instrumentation).
Finally, this analysis was performed only in the coronal plane, and certainly analysis of sagittal plane alignment—specifically, the instrumented thoracic region, the thoracolumbar junction, and the unfused lumbar spine—is required for a thorough analysis of the results of selective anterior versus posterior thoracic fusion. 1,3 Although Betz et al 2 and Lenke et al 10 reported the results of these patients in large-scale studies without finding any difference in thoracolumbar junctional sagittal alignment in comparing anterior versus posterior thoracic fusions.
It is noteworthy that spontaneous lumbar curve correction was consistent and favorable after both selective thoracic anterior and posterior fusions among all three curve modifiers investigated (A, B, and C), with the ASF procedure statistically better. This reinforces the notion that there are many curves appropriate for selective thoracic fusion in which sparing the lumbar spine may position the lumbar curve in a better alignment. Certainly, longer follow-up is required to see whether this improved spontaneous lumbar curve correction is maintained over time and what effects it may have on the development of lumbar degenerative disc disease below the fusion.
There are certainly some weaknesses to this study, including the fact that it was multicentered with six different surgeons participating in the operative treatment. Also, although these curves were prospectively observed, they were not randomized, and thus the individual surgeons decided on whether they would treat a curve by the anterior or posterior route. However, this bias was somewhat minimized by having a homogenous group of preoperative curves that were similar in Cobb measurements in both the thoracic and the lumbar regions, whether treated anteriorly or posteriorly (Tables 1 and 2). Also, the position of the lumbar curve could be quantified objectively by dividing all curves into lumbar modifiers A, B, or C based on the preoperative position of the apical lumbar spine to the CSVL.
Another potential criticism is that distal fusion levels for these curves were intentionally different for those treated anteriorly or posteriorly. The anterior groups all had the lowest fusion level at the lower end vertebra of the thoracic curve, which was T11 (n = 29), T12 (n = 34), or L1 (n = 7). In contrast, the posteriorly treated curves were fused to the most proximal stable vertebra, which was T12 (n = 14) or L1 (n = 49). Thus, most anteriorly treated curves were fused above L1, whereas most posteriorly treated curves were fused to L1. It is unknown how the posteriorly treated lumbar curve would have responded if the same distal fusion levels had been treated as in the anterior procedures. However, previous experience with posterior instrumentation and fusion of thoracic curves with segmental spinal instrumentation has shown the safest correction and balance is obtained when stopping at the stable vertebra. 11 Failure to adhere to this leads to a risk of decompensation, which occurred at a rate of 22% in one study. 4
Although favorable spontaneous lumbar curve correction occurred consistently after both selective thoracic anterior and posterior spinal fusions, the thoracic ASF procedure provided the best percentage of thoracic correction and spontaneous lumbar curve correction in all curve types investigated. Overall, the amount of thoracic correction paralleled the spontaneous correction of the unfused lumbar curve. In addition, the ultimate lumbar curve position based on the lumbar apical translation from the midline was as good, if not slightly better, in the anteriorly treated group for most curve types investigated. It is incumbent on scoliosis surgeons to consider either anterior or posterior selective thoracic fusion for those adolescent idiopathic curves amenable to the technique to achieve a balanced and mobile lumbar spine. However, the anterior approach appears to have a distinct advantage in allowing improved thoracic correction with corresponding improved lumbar spontaneous correction, especially for lumbar curve modifier C (true King Type II curves).
The authors thank Jack Baty for statistical analysis.
1. Bernhardt M, Bridwell KH. Segmental analysis of the sagittal plane alignment of the normal thoracic and lumbar spine and thoracolumbar junction. Spine 1989; 14: 717–21.
2. Betz RR, Harms J, Clements DH, Lenke LG, Lowe TG, Shufflebarger H. Prospective comparison of anterior versus posterior fusion for thoracic adolescent idiopathic scoliosis
. Presented at the Scoliosis Research Society, 31st Annual Meeting, Ottawa, Canada, September, 1996.
3. Bridwell KH, Betz RR, Capelli AM, Huss G, Harvey C. Sagittal plane analysis in idiopathic scoliosis
patients treated with Cotrel–Dubousset instrumentation. Spine 1990; 15: 921–6.
4. Bridwell KH, McAllister JW, Betz RR, Huss G, Clancy M, Schoenecker PL. Coronal decompensation produced by Cotrel–Dubousset “derotation” maneuver for idiopathic right thoracic scoliosis. Spine 1991; 16: 769–77.
5. Cochran T, Irstam L, Nachemson A. Long term anatomic and functional changes in patients with adolescent idiopathic scoliosis
treated by Harrington rod fusion. Spine 1983; 8: 576–84.
6. Ginsburg HH, Goldstein L, Haake PW, Perkins S, Gilbert K. Longitudinal study of back pain in postoperative idiopathic scoliosis
: Long term follow-up, phase IV. Presented at the Scoliosis Research Society, Asheville, North Carolina, 1995.
7. Kalen V, Conklin M. The behavior of the unfused lumbar spine following selective thoracic fusion for idiopathic scoliosis
. Spine 1990; 15: 271–4.
8. King HA, Moe JH, Bradford DS, Winter RB. The selection of fusion levels in thoracic idiopathic scoliosis
. J Bone Joint Surg [Am] 1983; 65: 1302–13.
9. Lenke LG, Betz RR, Bridwell KH, et al. Using the King method to classify adolescent idiopathic scoliosis
: Is it reliable? Presented at the Scoliosis Research Society, 31st Annual Meeting, Ottawa, Canada, September, 1996.
10. Lenke LG, Bridwell KH, Baldus C, Blanke K. Preventing decompensation in King type II curves treated with Cotrel–Dubousset instrumentation: Strict guidelines for selective thoracic fusion. Spine 1992; 17: 274–81.
11. Lenke LG, Bridwell KH, Baldus C, Blanke K, Schoenecker PL. Cotrel–Dubousset instrumentation for adolescent idiopathic scoliosis
. J Bone Joint Surg [Am] 1992; 74: 1056–67.
12. Lonstein J. Decompensation with Cotrel–Dubousset instrumentation: A multicenter study. 25th Anniversary Meeting of the Scoliosis Research Society, Minneapolis, Minnesota, September, 1991.
13. Richards BS, Birch JG, Herring JA, Johnston CE, Roach JW. Frontal plane and sagittal plane balance following Cotrel–Dubousset instrumentation for idiopathic scoliosis
. Spine 1989; 14: 733–7.
14. Richards BS. Lumbar curve response in type II idiopathic scoliosis
after posterior instrumentation of the thoracic curve. Spine 1992; 17: S282–6.
15. Roye DP Jr, Farcy JP, Rickert JB, Godfried D. Results of spinal instrumentation of adolescent idiopathic scoliosis
by King type. Spine 1992; 17(Suppl): S270–3.
16. Shufflebarger HL, Clark CE. Fusion levels and hook patterns in thoracic scoliosis with Cotrel–Dubousset instrumentation. Spine 1990; 15: 916–20.
17. Thompson JP, Transfeldt EE, Bradford DS, Ogilvie JW, Boachie–Adjei O. Decompensation after Cotrel–Dubousset instrumentation of idiopathic scoliosis
. Spine 1990; 15: 927–31.
Point of View
John Anthony Herring MD
Texas Scottish Rite Hospital for Children
This article addresses what I would call the King II dilemma, a tempest (in a teapot?) that has raged on for several years. King et al observed that distraction instrumentation of the thoracic curve in a patient with both a thoracic and a lumbar curve would be followed by spontaneous correction of the noninstrumented lumbar component, with a well balanced spine as the result. When this was applied in the modern era using rotational instrumentation, a trunk imbalance developed to the left in some patients, which sometimes resolved and at other times persisted. In an effort to avoid this, surgeons began to undercorrect the thoracic curves, and found that the problem largely went away (Authors’ references 8, 4, and 7).
The authors of this report show that they can obtain maximum thoracic curve correction with anterior instrumentation without being bothered by postoperative decompensation. The fact that the anterior surgery produced better correction than the posterior procedure was predetermined by the intentional undercorrection performed posteriorly to avoid decompensation. We do not know whether maximum correction of the thoracic curve posteriorly is more or less corrective than maximum correction anteriorly. We also do not know if the concomitant thoracoplasty in the anterior cases added to the flexivity of the deformity. The absence of decompensation is attributed to the difference in forces applied, but this is conjectural without confirming biomechanical studies.
At this point, I believe that the authors have substantiated their claim that anterior instrumentation of these curves can produce better coronal correction and balance compared with posterior instrumentation. I am not certain whether this applies to curves of the type A modifier. Perhaps more vigorous posterior correction in this group would have resulted in better correction without decompensation. The advantage seems valid, however, for groups B and C.
To me, the question is, “Should I now adopt the anterior approach to these curves?” At this point, I am unsure of the answer. What happens to the sagittal profile after anterior instrumentation? Is there increased kyphosis in younger patients? What happens if the hardware fails; will it impinge on a vital structure, or will it be impossible to retrieve? How about pseudarthrosis repair; wouldn’t that be difficult to accomplish anteriorly? What about an infection anteriorly, early or late; how difficult would the management be? Is the smaller rod anteriorly headed for fatigue failure? I personally will wait a bit longer for the answers to some of these questions before jumping on the bandwagon, but maybe not too long.