10.1 Introduction

The goals of surgery for adolescent idiopathic scoliosis (AIS) are to prevent further curve progression and to obtain deformity correction fusing as few motion segments as possible while keeping the patient balanced in the coronal and sagittal planes. Above all, we desire to keep the surgery safe without neurologic or other complications. A selective fusion refers to the situation where there is the potential to include in the fusion levels not only the primary curve but also the secondary curve. The term “selective thoracic fusion (STF)” is applied to patients who have a primary thoracic curve (Lenke 1 or 3) which is fused despite a lumbar or thoracolumbar/lumbar (TL/L) which crosses the midline (modifier C curves). The less common scenario is to perform a fusion of only the TL/L curve, known as a selective lumbar fusion (SLF) in the setting of a compensatory thoracic curve (Lenke 5 or 6). This chapter will outline the goals of surgery, the indications for performing an STF with several criteria evaluated in this process, the techniques to effectively perform a selective fusion, and the short- and long-term outcomes when a selective fusion is performed.

10.2 Indications and Criteria

It is generally accepted that preservation of motion segments is ideal to provide greater motion of the spine and, in turn, long-term spine health. Previous studies have demonstrated that preservation of lumbar motion segments decreases the incidence and severity of low back pain and has been most commonly studied in patients with fusions into the lumbar spine primarily focusing on fusion into the most distal levels. The traditional “dividing line” has been between the L3 and L4 fusion levels for double major scoliosis with the majority of studies demonstrating a significantly worse prognosis for those patients in which the fusion ended at L4 when compared to L3. ^{1 , 2 , 3} This can be primarily related to leaving the patient with three motion segments when fusing to L3 compared to leaving two motion segments when fusing to L4. ⁴ However, magnetic resonance imaging (MRI) studies of the lumbar spine with fusion to L3 or L4 do not necessarily identify changes consistent with these clinical findings. ⁵ Degenerative disc disease can occur and may be related to the size of the residual lumbar curve ⁶ and often is distal to the lowest instrumented vertebra (LIV) and most common at the L5–S1 junction. ⁷ The option of SLF does not offer as strong of an argument to preserve thoracic curve motion segments; mobility of this part of the spine is not as robust as the lumbar spine. However, as a general principle, it is important to fuse as few segments of the spine as possible. ^{8 , 9}

In an STF scenario, it is generally believed that preservation of the lumbar motion when stopping distally at T11, T12, or L1 is preferred over fusion into the lumbar spine, especially with fusion to L3 or L4. This assumes excellent coronal and sagittal plane balance and that the residual lumbar curve is small enough that it will not progress with time. Historically, the major challenge when performing an STF was to minimize imbalance in the coronal plane in which the patient is shifted to the left (in the typical right thoracic/left lumbar curve; Fig. 10‑1). This controversy was studied extensively to analyze risk factors for its occurrence and included parameters such as a larger lumbar curve, LIV which included the proximal aspect of the lumbar spine, residual distal lumbar tilt at the L4 level resulting in left trunk shift, and overcorrection of the thoracic curve. ^{10 , 11 , 12 , 13 , 14} The King–Moe classification made the distinction between a type II (primary thoracic curve with compensatory lumbar) and a type III (primary thoracic curve only) in which an STF had to be weighed against the fusion of both curves for the type II pattern. ^{12 , 15} The essential reason that patients had a coronal plane imbalance with a trunk shift to the left was due to the lower part of the lumbar curve having residual tilt to the left and the inability of the proximal aspect of the lumbar curve to bring the patient back to the right into balance. Richards ¹⁰ first reported residual obliquity of the L4 vertebra which could only be compensated for by an increase of the lumbar curve which essentially meant an increase in the proximal aspect of the curve. Others have recently studied this phenomenon describing the lumbosacral oblique takeoff angle (LSOTA) which improved by only 2 degrees in the patients who had an STF, while the patients who had a nonselective fusion had improvement by 11 degrees. ¹⁶ Sagittal imbalance can occur when the LIV is chosen incorrectly leading to distal junctional kyphosis (DJK) or when inappropriate restoration of thoracic kyphosis is achieved, which leads to junctional kyphosis occurring below the instrumentation. ^{17 , 18}

Using the Lenke classification, the indications for performing an STF have traditionally been for patients with a type 1C or 2C in which the main thoracic (MT) curve is structural with a TL/L curve whose apex crosses the midline (lumbar modifier C) but does not bend to less than 25 degrees on a supine best bend radiograph. ¹⁹ Despite these curve patterns, which would indicate the prospect of performing a selective fusion, surgeons often elect not to perform an STF for the 1C curve patterns with up to 32% of patients having fusion into the lumbar spine in this setting. ²⁰ Other less clear indications include the 3C and 4C curves in which the TL/L curve is large and stiff enough that it does not bend to less than 25 degrees. ^{21 , 22 , 23} In these scenarios, a more careful assessment of the clinical and radiographic features of the patients is important as well as understanding the desires of the patient and the family, as coronal decompensation has been reported in up to 28% of patients. ²²

Criteria that make a selective fusion possible (either an STF or an SLF) should be assessed with a good physical examination as well as a critical analysis of radiographs, which generally and primarily assess the relative severity of deformity between the thoracic and lumbar curves. The physical examination findings to identify when an STF is being considered include a right trunk shift with an elevated right shoulder and a larger rotational prominence of the thoracic spine compared to the lumbar spine (Fig. 10‑2a, b). Patients who have preoperative decompensation to the left have a 57% chance of being decompensated postoperatively. ²⁴ These clinical features are similarly analyzed on radiographs with the general concept that the thoracic curve parameters are dominant over the lumbar curve. When the trunk shift is to the right as measured by a rightward position of the point at the mid-distance between the rib margins and the center sacral vertical line (CSVL), then a selective fusion is possible (Fig. 10‑3).

Radiographic criteria should focus on the degree of deformity of the thoracic and lumbar curves and the relationship between the two to identify which ones are dominant. The ratios comparing the thoracic and lumbar curves with respect to curve magnitude, apical vertebral translation (AVT), and apical vertebral rotation (AVR) should be determined to help with decision making for these patients. When the ratio of these measured parameters comparing the thoracic and lumbar curves is greater than 1.2, it is more likely that an STF will be successful ²⁵ (Fig. 10‑4). Other preoperative radiographic measurements which help predict a successful STF include lumbar curves less than 45 degrees which bend to less than 25 degrees. ²⁶ The lumbar gap (LG) is measured from the CSVL to the concave edge of the apical lumbar vertebra and helps distinguish C lumbar modifier curves and can aid in deciding whether to fuse the lumbar spine. ²⁷ In a matched series of 103 patients with lumbar modifier C curves, patients who underwent an STF had a smaller LG (9.6 mm) compared to those who had both curves fused (22.0 mm). A 47% correction of the thoracic curve and a 39% spontaneous correction of the lumbar curve was noted in those patients who had had an STF. ²⁷ Remaining growth always plays an important role in decision making, and when patients have open triradiate cartilage, the risk of curve progression secondary to the crankshaft and adding on is higher unless a combined anterior/posterior approach is used. ²⁸

A similar strategy should be used when assessing whether to perform a selective fusion of the TL/L curve, comparing the relative deformity of the TL/L curve to the thoracic curve. The clinical finding suggesting that a selective TL/L fusion is appropriate is when the left shoulder is elevated compared to the right, a left trunk shift is present, and when the right waistline is more curved than the left. The radiographic assessment should include an analysis of the TL/L curve magnitude, AVT, and AVR relative to the respective thoracic measurement of these parameters, and if greater than 1.2, then selective fusion may be planned. In a multicenter study, 44 patients who had a TL/L:T Cobb ratio of 1.25 or greater and/or a thoracic curve which bent out to 20 degrees or less, 42 patients (93.3%) had a satisfactory result. The preoperative thoracic curve magnitude was also critical, with successful results seen when the average thoracic curve was 40 degrees compared to 49 degrees. ²⁹ Similar to other studies, continued patient spine growth should always be assessed, as an open triradiate cartilage often predicted a poor result with adding of the thoracic curve. ²⁹ Of 43 patients in whom the triradiate cartilages were closed, 42 had satisfactory results. It is important to recognize, however, that the ability of the thoracic curve to respond appropriately or the level necessary to maintain coronal balance is less than the lumbar curve response in an STF scenario. ³⁰

It is critical to always assess the sagittal plane as a final check to ensure that significant junctional kyphosis is not present between the T and TL/L curves. This has traditionally been thought to be an indication that both curves are structural and require inclusion of each in the fusion and instrumentation. More recently, it is proposed that this apparent junctional kyphosis seen preoperatively between curves may be a response to the hypokyphosis of the thoracic spine. If all or most of the other parameters indicate that a selective fusion of the thoracic spine is possible, then it is generally believed that an STF will be successful by restoring thoracic kyphosis (Fig. 10‑5a–d). This will recreate a more normal sagittal plane in the thoracic spine which subsequently results in increased lumbar lordosis. ³¹ Although not specifically studied, this improvement in the sagittal plane parameters should lead to a decreased incidence of junctional kyphosis. Recently, some have suggested using the stable sacral vertebra (SSV) defined by the vertebra best bisected by the posterior sacral vertical line on the lateral radiograph ¹⁷ ; however, further study is necessary in a larger cohort to determine the utility of this parameter.

10.3 Technical Aspects for Successful Selective Fusion

A successful selective fusion requires excellent preoperative planning and execution of a sound game plan to achieve the goals previously outlined. The general principles include achieving an adequate coronal plane correction without throwing the patient out of balance, the so-called decompensation in which the patient’s trunk shift is to the left for STF, and shifting to the right while creating a high right shoulder in the setting of an SLF. Similarly, the sagittal plane correction should achieve restoration of thoracic kyphosis while maintaining sagittal balance without creating junctional kyphosis.

10.3.1 Selection of Fusion Levels

The selection of fusion levels for an STF is critically important, especially when it comes to the LIV. The upper instrumented vertebra (UIV) is usually chosen based on whether there is a structural proximal thoracic curve. When the proximal curve is structural, the fusion level is usually T2 (Lenke 2 or 4 curve) and, rarely, T1. When there is no structural proximal thoracic curve (Lenke 1 or 3 curve), the UIV is usually T4, especially when the right shoulder is elevated preoperatively. Occasionally, it can be T3 when the left shoulder is elevated to allow for some control of the proximal thoracic spine, ensuring level shoulders at the completion of the surgery.

Careful selection of the LIV is necessary to achieve adequate correction of the MT curve and avoid junctional kyphosis while not influencing the lumbar curve in a manner that will lead to decompensation. Two parameters are generally assessed depending on the location of the vertebra relative to the following: the end vertebra (EV) and the stable vertebra (SV). When the EV and the SV are the same, the choice of LIV is easy and should be that shared vertebra (Fig. 10‑6a, b). When the SV is the disc below the EV, then the EV generally is the right choice to be the LIV when the curves are especially in smaller to mid-size curves as a more distal LIV will result in decompensation to the left. However, the LIV should be the EV + 1 in larger curves to maintain coronal balance. When the SV is distal to the EV, the LIV should be the SV (Fig. 10‑7a, b). ^{23 , 32}

The number and position of the thoracic screws is important to achieve a good result. This continues to be controversial, with both sides of the argument generating good-quality reasons for their viewpoint. ³³ The proponents for a higher screw density feel that more screws allow one to control the instrumented curve better and can dial in the desired correction. Those who argue for a lower screw density suggest that a high-density construct is not necessary as the goal is to achieve a bit of under correction of the thoracic curve to avoid decompensation due to overcorrection of the MT curve with too much horizontalization of the MT LIV. In a small series of patients comparing consecutive pedicle screws and those who had screws skipped at various levels, there were no differences between the two groups with respect to coronal plane correction, sagittal balance, or clinical outcomes. ³⁴ It has been the practice of the author to utilize approximately a 1.5-screw density with strategic placement of the screws to allow for control of the ends of the construct with two segmental screws on the left rod with apical control by two to four screws placed at the apex. The right rod has excellent fixation at the ends, with one to two screws placed at the apex to assist in direct apical derotation during left rod placement and indirectly during placement of the undercontoured right rod.

For an SLF, the fusion levels for a 5C curve are generally the same whether you perform an anterior or a posterior surgery and should include the proximal and distal EVs (Fig. 10‑8a–d). In the past, when anterior fusion was more often utilized, fusing “end to end vertebrae” was the general rule, while doing a short-segment fusion was an option for smaller curves. ³⁵ With the greater use of the posterior approach for these curves, there appears to be a similar correction in the coronal and sagittal planes with some exceptions; however, the number of fusion levels tends to be higher with the posterior approach. ^{36 , 37 , 38 , 39 , 40} The LIV is generally thought to be the distal EV of the TL/L curve and works well to create coronal plane correction without significant disc wedging below the LIV. The best outcomes occur when the disc below the chosen LIV is “open” to the opposite side of the convexity of the lumbar curve. In other words, it is part of the fractional lumbosacral curve, and in this case, the distal EV will work well to be the LIV. If the disc is parallel below the intended LIV, then it is expected that some disc wedging will occur below the LIV. ^{41 , 42 , 43} When the LIV is the EV, the results are very good with small-to-medium curves; however, some have suggested that fusing one level below the EV is appropriate for curves greater than 60 degrees. ⁴⁴

The UIV selection for 5C curves is dependent on the curve magnitude and flexibility of the TL/L and MT curves and the position of the proximal EV relative to the SV. In general, when the EV and the SV are the same, then fusion to this shared vertebra is the correct choice. When the EV is distal to the SV, then fusion to the SV should be the UIV.

10.4 Correction Mechanics and Desired Correction

10.4.1 Selective Thoracic Fusion

The correction of the MT curve can be performed in several ways using traditional rod rotation or with more of an apical correction maneuver. My preferred technique is to place the left rod into the proximal and distal segments with a stiff 6.0 cobalt chromium (CoCr) rod in which hyperkyphosis has been placed. The rod is locked at the LIV to prevent the rod from migrating within the LI screw while leaving the UIV unlocked to allow for continued lengthening of the spine. Some would argue to fully tighten the UIV to allow for greater correction of the spine; however, this may be too constrained of a system, and the desired correction is most often less than fully corrected. The apex of the spine is then brought to the left rod through reducers which are attached to the apical and periapical screws. While this is occurring, a tower is attached to the left LIV screw and a counter-clockwise rotational maneuver is placed on this screw while the right-sided apical screws have towers which push down on that side to impart a derotation of the apex to improve the axial plane deformity. The force on the distal left LIV is used to minimize the forces imparted to the lumbar curve which assists in preventing lumbar decompensation. At the same time, derotation towers at the apex on the convex side can be used to help with apical derotation of the curve. Traditional left (concave) rod derotation can be performed in this setting; it may create risk for decompensation, as this maneuver influences the proximal aspect of the lumbar spine due to the mechanics of correction. ⁴⁵

The amount of the desired correction of the thoracic curve has evolved throughout the years when the concept of decompensation was first identified using hook-rod instrumentation and the King–Moe AIS classification. The patients were often fused one or two segments longer; overcorrection of the thoracic curve occurred, and the patient’s trunk shift was to the left which was displeasing to the patient and family. It was suggested that the tilt of the L4 vertebra remained following surgery which was the driver for the left body shift and could not be compensated for by the upper aspect of the lumbar curve, as this was corrected indirectly with the STF. It is generally believed that rod rotation without some stabilization of the LIV on the left to prevent this from occurring will continue to leave patients imbalanced to the left. More recently, some have advocated for greater correction during STF; however, it should be noted that some patients in these studies may not have been true Lenke C modifier curves and instead were either B modifiers or very close to being a B modifier, where more aggressive correction of the thoracic spine can be achieved. ⁴⁶ This author would caution against attempts at complete correction of the MT curve, as the LIV of this curve is the UEV of the TL/L curve and if brought to a horizontal position will not allow the patient to return to the right since the lower part of the TL/L curve remains with some tilt of L3 and L4 (Fig. 10‑9a–d).

The correction techniques when performing an SLF from the anterior approach are traditional rod rotation which is performed from the convex side of the spine following segmental screw fixation. This achieves correction of the coronal and axial plane while maintaining lumbar lordosis. The correction mechanics posteriorly vary and depend on the size of the curve and are, in part, particular to each surgeon. However, in general, most curves can be corrected with some form of convex side shortening with compression on the convex side with subsequent distraction on the concave side. A rod rotation maneuver may be included as the initial step to place the left rod first followed by rod rotation to improve the coronal and axial planes followed by convex compression. This correction is made easier when Ponte osteotomies are used on the convex side to provide room for the posterior shortening, ³⁶ while the concave side osteotomies allow for the spine to become untethered. When a larger thoracic curve is present proximal to the TL/L curve, it is important to avoid overcorrection which leaves the UIV of the TL/L too horizontal. This potentially reveals the right thoracic curve which may manifest as a right trunk shift and right shoulder elevation, requiring further surgery (Fig. 10‑10a, b).

10.5 Outcomes Following Selective Fusion

The outcomes of STF through the years has been outstanding in the short and long term, and it is very uncommon to revise an STF to include the lumbar curve later, as the vast majority of patients will compensate for the thoracic curve correction. ⁴⁷ This generally occurs with an initial decrease in the lumbar curve postoperatively followed by an increase in its size to allow the proximal aspect of the curve to tilt in the horizontal plane without much effect of the lower part of the lumbar curve, resulting in good coronal plane balance. The anterior approach has been successful in achieving a good thoracic curve correction followed by good responses from the lumbar curve, with improved balance compared to posterior hybrid systems. ⁴⁸ Yong et al ⁴⁹ reported excellent correction of average 53 degrees thoracic curves to 25 degrees at 2 years with improvement of the lumbar spine by 41% to maintain balance. Others have compared the anterior approach directly to the posterior approach, demonstrating similar results. ^{50 , 51 , 52} Patel et al ⁵⁰ reported a matched group of patients treated anteriorly compared to posteriorly with similar 48 and 49% correction, respectively, of the lumbar curve following an STF. The vast majority of surgery today is through the posterior approach for an STF, as the procedure is much more familiar to all spine deformity surgeons, the surgical duration and hospital stay are shorter, and the return to activities happens sooner. With the use of screws and, more importantly, different correction mechanics, the TL/L curves improve, and this occurs most commonly in the distal segments of the curve. In the acute period, the patients who often get decompensation to the left are those in which correction was overdone so that the TL/L curves cannot compensate for the MT correction, patients who have significant growth remaining (Risser 0 with open triradiate cartilage), and patients in which the LIV was too distal. It should be remembered that a left trunk shift of some degree is nearly always seen postoperatively, and families should be aware of this occurrence and that correction of this happens over time and may take as long as 1 year to fully resolve. The traditional strategies to leave significant tilt of the LIV has been questioned in a multicenter study of 33 patients with 1C curves who had an STF in which the authors found no correlation between the postoperative LIV tilt and coronal balance. ⁴⁶ The sagittal plane outcome is something that is getting paid more attention to, and studies to identify the outcome in this plane have demonstrated overall good balance when the correct LIV is chosen and thoracic kyphosis has been restored. Measured junctional kyphosis of greater than 10 degrees is rarely clinically significant. ^{17 , 18 , 31 , 53 , 54} Some have suggested that thoracic hypokyphosis persists or is worsened following an STF and is accompanied by similar changes in the noninstrumented lumbar curve and these findings require long-term follow-up. ⁵⁵ STF for Lenke 3 curves, in which the thoracic and lumbar curves are deemed to be structural, should be performed with caution, as these patients are often out of balance radiographically despite overall good patient-reported outcomes. The long-term outcomes need further study. ⁵⁶

The outcomes for an SLF are similar to that of an STF, with generally excellent coronal balance and sagittal balance; however, the ability for a thoracic curve to respond to correction of the TL/L curve is less. It is important to realize that the compensatory thoracic curve responds less to correction of the TL/L spine than the lumbar curve in STF and may demonstrate increased rotational deformity postoperatively, especially if the patient has significant residual spine growth. ³⁰ The importance of the flexibility of the thoracic curve should be emphasized, as it plays a critical role in the postoperative balance of the patient. ⁵⁷ When direct comparisons of STF and SLF are performed, the results generally suggest that the acceptable compensatory curve that will allow for a selective fusion is larger for STF (lumbar) than for SLF (thoracic). This was demonstrated in a multicenter study comparing selective fusion for 1C and 5 C curves in which excellent coronal balance was seen at 2 and 5 years for both curve patterns; however, it is important to recognize that the preoperative compensatory curves were smaller for the 5C curves (25 degrees for thoracic curve) than the 1C curves (40 degrees for lumbar curve). ⁵⁸

The long-term outcomes of STF have generally been very good, with the overall maintenance of the correction and coronal balance over time. Larson et al ⁵⁹ compared patients who had an STF with a long fusion for 1C curves over a 20-year follow-up and showed STF patients having 43% correction of the compensatory lumbar curve which was maintained at the final follow-up when compared to the 2-year results, and there was no worsening of the L4 obliquity. The STF patients were more likely to recognize their deformity when compared to patients who had both curves fused. The longest follow-up for patients undergoing an STF was by Lonstein, ⁶⁰ who reported at 33 years on 40 patients with an overall good correction of the thoracic (56–38 degrees) and compensatory lumbar (44–37 degrees) curves, with 5 patients having additional surgery but overall good Oswestry Disability Index (8.7—minimal disability) and SRS-30 (3.8) scores.

Long-term results of SLF demonstrate similar results with a recent 9-year follow-up study of patients who underwent a single-rod anterior spinal fusion/instrumentation with curve correction of the TL/L curve from 59 to 23 degrees with the compensatory thoracic curve going from 39 to 29 degrees. ⁶¹ Delfino et al ⁹ demonstrated similar results with longer follow-up, averaging 17 years with 72% correction of the TL/L curve with compensatory thoracic curve correction of 31 to 18 degrees without the need for reoperation. At a similar 17-year follow-up, Kelly et al ⁸ demonstrated good functional scores without progression of the instrumented or uninstrumented curves following an anterior spinal fusion/instrumentation; however, two patients required revision for implant failure. The long-term results following a posterior fusion and instrumentation are still lacking, as the posterior approach for this curve pattern has been only recently performed.

Combined series of both STF and SLF fusions have recently reported 10-year results with good coronal plane correction of the instrumented segments of 51 and 60% for the T and TL/L curves, respectively, with a compensatory improvement of the compensatory curves to nearly the same levels and overall excellent SRS outcome scores, which did not change over time. ⁶²