11.7.1 Upper Thoracic and Main Thoracic UIV

Lee et al ⁴³ originally observed that fusions had to be extended through the UT curve to avoid shoulder asymmetry postoperatively. Although their data predated the Lenke classification system and included hybrid constructs, it became a fundamental principle in deciding UIV selection. They divided 246 patients into three groups: T1-positive tilt (tilted to the right in a right MT curve) with UT + MT fused, T1-positive tilt and only MT fused, and T1-negative tilt or neutral with MT fused. Based on preoperative shoulder imbalance and a high progression in certain groups postoperative, they recognized the importance of fusing the UT when the left shoulder is high preoperatively. That same principle has been advocated by others and simplified to the idea of fusing to T2 when the left shoulder is high in a right MT curve, T3 when they are even, and T4 if the right side is high. ⁴⁶ Other authors have advocated fusing the UT when it measures greater than 45 degrees with good outcomes as reliable bending X-rays of the UT curve are effort limited. ²³

Suk et al ²² divided 40 patients with pedicle-screw–only fusions into a group where the UT and MT curves were fused and another where only the MT curves were fused. They subsequently divided patients into three preoperative cohorts: left shoulder elevated greater than 5 mm, shoulders even, or right side elevated greater than 5 mm. If the left preoperative shoulder was up, all patients had left shoulder elevation postoperatively; if the shoulders were even, shoulder asymmetry correlated with not fusing the UT curve; and if the right shoulder was up before surgery, six of eight patients were still elevated postoperatively when both curves were fused. A mixed outcome was obtained when only the MT fused but all were within 1 cm of difference. They concluded that the UT curve should be fused when (1) the left side is up preoperatively, (2) they are even preoperatively, or (3) the right side is up but by less than 12 mm.

Other authors have advocated level selection based on the flexibility of the UT curve. Ilharreborde et al ⁴⁴ looked at Lenke 1 and 2 curves and fused the UT curve only if it was rigid, or if the T1 tilt and shoulder were in the same direction with a flexible UT curve. They defined flexible as a difference of 15 degrees on bending radiographs. Overall they reported 89% successful outcomes based on their protocol and advocated for their guidelines. However, they defined a successful outcome using parameters of T1 coronal translation, T1 tilt, and clavicular angle. Several other authors have shown that T1 tilt has a poor correlation to shoulder symmetry and clinical satisfaction. ^{20 , 43 , 45}

Bjerke et al ⁴⁷ compared Lenke, Trobisch (curve type and preoperative difference), and Ilharreborde selection criteria of UIV in a cohort of 263 patients. Using Lenke UIV selection criteria, 16% had imbalance postoperatively as compared to 14.9% using Ilharreborde criteria or 21.9% using Trobisch criteria. They noted that clavicle angle, T1 tilt, and radiographic shoulder height (RSH) were the only reliable measures, but T1 tilt was not associated with postoperative shoulder imbalance. Furthermore, they reported a new shoulder imbalance incidence of 8.8% using RSH cutoff of 15 mm.

Tang et al ⁴⁸ looked only at Lenke 1 type curves and fused to EV or EV –1 if the left shoulder was up and EV or EV +1 if the left shoulder was elevated preoperatively. 26.7% of UIV selected were within levels suggested by Rose and Lenke, ⁴⁶ but 56% were caudal to recommended levels. Furthermore, 61.3% of patients had improvement immediately postoperatively, but patients continued to self-correct over time and ultimately 93.3% had acceptable outcomes (< 1 cm shoulder height difference) at last the follow-up. ⁴⁸

Other authors have also reported varying degrees of spontaneous correction over time. Cil et al ⁴⁹ reviewed 37 patients instrumented with the Isola system having nonstructural UT curves and observed that 41% of patients with unfused UT curves improved over time. Lee et al ⁴³ similarly observed that spontaneous correction of 24.8 and 28.6% occurred in groups that had fusions of the MT but not UT curves. Although it is somewhat reassuring that mild asymmetry will correct, significant residual asymmetry may lead to further problems as the compensation has to occur across some unfused segments. Cao et al ⁵⁰ noted a correlation between shoulder imbalance and adding-on (p = 0.03). Interestingly, they did not observe an impact of postoperative shoulder imbalance between patients fused to T2 versus T3 and below.

Significant variability still exists when deciding which level to choose for the UIV. Generally, the Lenke classification helps us choose which curves to fuse and we can then refine our decision based on the available literature. However, universal rules may not always be applicable as varying factors influence our outcomes. Inherent factors to the curve such as a large but flexible primary thoracic curve will allow greater correction which will drive the left shoulder higher as will surgical choices such as metal type, possible use of osteotomies, and implant density. A realistic understanding of an individual surgeon’s expected correction applied to preoperative parameters will often help a surgeon obtain more reliable and favorable outcomes.

11.7.2 Thoracolumbar Upper Instrumented Vertebra

For thoracolumbar curves, classically UIV fusion levels are chosen as the UEV. This most commonly is T10 or T11. Shufflebarger et al ⁵¹ originally demonstrated comparable outcomes using posterior screw-based instrumentation to anterior results, and many subsequent studies have continued to use the same selection criteria with excellent outcomes. ^{41 , 52 , 53} Okada et al ⁵⁴ reported on 29 patients with Lenke 5C curves where the UIV was either UEV –1 or UEV and noted no difference in correction initially (8.1 ± 5.1 vs. 7.5 ± 6.8, p = 0.580). However, even with shorter follow-up, the UEV –1 group had a greater loss of correction (Cobb 12.2 ± 6.7 vs. 7.2 ± 0.9, p = 0.033) without differences in sagittal and coronal balance. Generally, the UIV selected remains the UEV, but caution should be used to avoid selecting a UIV at the apex of kyphosis.

11.8.1 Main Thoracic Lower Instrumented Vertebra

Typically, the LIV for an MT curve will span between T10 and L2. Criteria for choosing a selective fusion where the compensatory structural curve is not fused will be discussed in another chapter and therefore not addressed here. As discussed in section “Background/Historical Context” of this chapter, the principles of NVs and stable zone described by Goldstein, ¹ Moe et al, ⁴ and Harrington ⁸ are still fundamentally used but slightly more refined.

Suk et al ²² looked at LIV with respect to NV and EV in Lenke 1 and Lenke 1 AR-type curves. They observed unsatisfactory outcomes (T1 translation > 2 cm from CSVL or adding-on) in 28% (Lenke 1) and 50% (Lenke 1 AR) of patients. The NV is typically caudal to the EV and can be further distal especially with Lenke 1 AR curve patterns. In patients where NV = EV + 3, they noted unsatisfactory results in patients fused short to NV – 2 (75%) or NV – 3 (71%). However, if the NV = EV or was EV + 1, patients did well if fused to NV or NV + 1. With the 5-year follow-up, Clément et al ⁵⁵ similarly suggested selecting LIV as lowest between NV and the stable vertebra. Erdemir et al ⁵⁶ further tried to subclassify the lumbar rotation as turning into the thoracic curve (Lenke 1 AR), neutral, or turning away from it and recommended fusing to NV – 1 (L2–L3), NV (T12–L1), and NV + 1 (L1–L2), respectively. However, some of the difficulty in standardizing level selection lies in the variation in selecting the NV. In one study, 50 consecutive surgically treated cases of AIS were reviewed by 16 scoliosis surgeons. Interobserver reliability for the end, neutral, and stable vertebrae had kappa values of 0.45, 0.32, and 0.52, respectively. ⁵⁷ The poor agreement between surgeons in selecting the NV limits the utility of the NV as a common factor in level selection. Furthermore, Rizkallah et al ⁵⁸ compared three different LIV selection criteria: Suk, Lenke, and Dubousset criteria. In mostly Lenke 1A/B curves, using these three parameters, they did not find a significant difference in outcomes between groups.

A more reliable tool derived from Harrington’s stable zone is the principle of the last touched vertebra (LTV). Lenke et al ⁵⁹ described the LTVs as the caudal level where the CSVL touches the vertebrae. Others have expanded on it describing a last substantially touched vertebra where the CSVL at least touches the pedicle or is even more medial (Fig. 11‑1). Looking at 116 Lenke 2A patients, Cao et al ⁵⁰ divided patients based on whether the LIV was proximal, at, or distal to the LTV. Fifty percent of patients fused proximal to the LTV developed adding-on, whereas only 7.3% developed adding-on if the LIV was at the LTV and 7% if LIV was distal to the LTV. Similarly, Matsumoto et al ⁶⁰ found in a multivariate analysis of 169 Lenke 1A curves that only apical translation and LIV–LTV distance were significantly correlated with adding-on.

Qin et al ⁶¹ assessed 104 Lenke 1A patients and divided them into three cohorts based on the LIV and noted a higher rate of adding-on (%) in the nonsubstantially touched cohort: LIV = nonsubstantially touched (66.7%), LIV = nonsubstantially touched +1 (11.6%), or substantially touched vertebra (10%). Similarly, Murphy et al ⁶² noted three factors associated with adding-on in 160 patients: LIV–LSTV (last substantially touched vertebra) distance, Risser, and LTV–CSVL less than 2 cm. The risk of adding-on was 16% with one factor, 27% with two, and 80% with all three.

It is also important to always consider the sagittal profile of patients when selecting fusion levels. A thoracolumbar kyphosis needs to be bridged when selecting levels just as a rostral UIV should not stop at the apex of a kyphosis. Yang et al ⁶³ built on the sagittal stable vertebra (SSV) principle (the vertebra touched by a line drawn from the posterior corner of the sacrum extending rostrally in a lateral film (Fig. 11‑2) described by Lenke. They observed distal junctional kyphosis (DJK) in 22% of patients when the LIV selected was above the SSV as compared to 4% or below. Lowe et al ⁶⁴ reported that 4.5% of patients had DJK (10 degrees of focal kyphosis) preoperatively and recommended the LIV be chosen as EV +2 if preoperative DJK was present. As scoliosis is a three-dimensional problem, surgical consideration should always include all three planes to minimize decompensation.

11.8.2 Thoracolumbar Lower Instrumented Vertebra

When considering LIV selection for lumbar/thoracolumbar curves, additional consideration should be given to the preservation of motion segments. Hypothetically, preservation of more motion segments may avoid accelerated degenerative changes and minimize pain as well as the need for further surgery in the future. Some studies have shown that preserving more unfused levels allows a greater distribution of motion across those segments. ⁶⁵ However, a paucity of data correlating clinical outcome scores to an LIV of L3 or L4 exists. ^{66 , 67} The majority of our long-term follow-up studies are limited by the use of outdated instrumentation, poor follow-up, and uncontrolled variables such as indications for reoperation, smoking status, and other factors impacting pain and function. Limitations notwithstanding, our surgical objectives are to achieve optimal correction while preserving as many motion segments as possible.

Shufflebarger et al ⁵¹ described the use of pedicle screws for the correction of Lenke 5C curves with comparable outcomes to anterior approaches. They reported good outcomes with the LIV as EV or EV + 1. Using a unilateral screw and concave top/bottom only construct, Roberts et al ⁶⁸ identified five factors that determined the need to extend beyond EV–EV: lowest instrumented vertebra disc angle (LIVA), TL/L curve magnitude, lumbar convex bending angle, lumbar apical vertebral translation, and compensatory thoracic curve magnitude. Others have similarly identified preoperative lowest instrumented vertebral tilt (LIVT) of greater than 25 degrees as a possible risk factor for coronal imbalance. ^{52 , 53} Wang et al ⁵³ further noted continued improvement of global and trunk balance from immediate postoperative to 2 years and derived a formula for 2-year lumbar translation: 14.1 + 1.2 (preoperative LIV–CSVL distance). They concluded that the LIV should not exceed 28 mm of coronal translation and 25 degrees of tilt.

Kim et al ⁶⁹ reviewed 66 patients with Lenke 5C curves using pedicle-screw–only constructs and compared LIV based on the flexibility of the bending radiographs. Patients were divided into three cohorts: fused to L3 with flexible L3 (L3 crosses the CSVL, and Nash–Moe 2 or less on bending), fused to L3 with a rigid L3, or fused to L4. Patients with a less flexible L3 had more LIVT and coronal translation than the flexible L3 or L4 cohort. Based on LIVT greater than 10 degrees or coronal translation of 15 mm, 9.1% of the flexible L3 cohort had unsatisfactory results as compared to 12.5% in the L4 and 68.2% in the rigid L3 group. Based on these results, they recommended selecting L3 when flexible, but otherwise to fuse to L4. Chang et al ⁷⁰ replicated these findings. They reported 15.2% of patients with flexible curves had unsatisfactory results (adding-on 5 degrees and distalization or increased disc angulation of 5 degrees or LIV tilt > 10 degrees or coronal translation > 15 mm) as compared to 61.1% in the rigid cohort.

11.9.1 Type 1: Main Thoracic Curves

Type 1 MT curves are the most commonly treated form of AIS. ⁷¹ Although these curves can be treated with either ASF or PSF, posterior instrumentation and fusion remain the gold standard. ⁷²

To determine the UIV, clinical shoulder balance needs to be assessed. As previously discussed, the clavicle angle appears to be the most reliable preoperative indicator for shoulder assessment. As a general rule, for a patient with a right MT curve and right shoulder elevation, the proximal extension of spinal fixation to T4 or T5 is appropriate. Exclusion of the UT segments will allow left shoulder elevation to occur with correction and will produce level shoulders postoperatively. For a patient with level shoulders, the extension of fixation to T3 or T4 is often indicated. Cranial extension of the posterior spinal segmental instrumentation (PSSI) to include the UT segments will facilitate control over the left shoulder height and maintain shoulder balance postoperatively. For a patient with left shoulder elevation, the extension of fixation to T2 is usually necessary to eliminate postoperative shoulder elevation. This proximal extension of the spinal fixation levels will allow intraoperative compression of the upper left thoracic segments to lower the left shoulder and correct the patient’s preoperative imbalance (Fig. 11‑3).

As discussed in detail previously, an alternate approach is to leave the UT curve unfused and allow self-correction to achieve shoulder balance. ^{43 , 49 , 73} Brooks et al ⁷³ reviewed 626 patients and noted that a UIV of T2 was associated with improvement of the UT curve magnitude, but patients with a UIV of T4 had better shoulder balance (34 vs. 45/48% for T2/T3, p = 0.008).

The Lenke system suggests selecting the LIV as the lowest vertebra touched by the CSVL for lumbar curves with an A modifier. Most commonly, this is the vertebra proximal to the stable vertebra (stable-1) or occasionally the vertebra two levels proximal to the stable vertebra (stable-2). PSSI in type 1 curves is best suited for patients with a normal or hyperkyphotic sagittal modifier. For type 1B curves, the LIV (the most cephalad vertebra intersected or bisected by the CSVL) will usually be located in the thoracolumbar junction. Selective fusion criteria for Lenke 1C curves are discussed in Chapter 10.

11.9.2 Type 2: Double Thoracic Curves

Type 2 double thoracic (DT) curve includes a major structural MT and a structural minor PT curve. ²² The general guideline calls for posterior arthrodesis of both curves. The LIV is chosen in a manner similar to that for treating type 1 curves. Generally, the proximal level of fusion should be either T2 or T3. Consideration of the patient’s clinical shoulder balance and radiographic clavicle angle is critical. ¹⁸ Patients with a high left shoulder preoperatively will require fusion extending to T2 (Fig. 11‑4). This will allow greater control in balancing the shoulders. For patients with level shoulders, the UIV may be either T2 or T3. When assessing the patient’s deformity to determine the UIV, the rigidity and magnitude of the PT curve will need to be considered. Large or inflexible curves will probably generate a significant shoulder imbalance as the MT curve is corrected. For patients with a high right shoulder, the UIV may be T3. Excluding the most proximal vertebra will allow spontaneous left shoulder elevation with correction of the MT curve. Often, the PT curve is hyperkyphotic, and compression of the convexity is therefore applied first. In general, when correcting the PT curve, one must consider the correction that will be attained in the MT curve. Greater correction of the MT curve will further increase the elevation of the left shoulder.

11.9.3 Type 3: Double Major Curves

Lenke type 3 double major curves are defined by a major structural MT curve as well as a minor structural TL/L curve. The general rule calls for the posterior fusion of both curves (Fig. 11‑5). The instrumentation begins at T3 to T5, depending on shoulder position, as with type 1 curves. The LIV is often the most cephalad lumbar vertebra touched by the CSVL when the lumbar curve is flexible and secure pedicle fixation is achieved and is usually L3 or L4. The dilemma is in selecting the appropriate distal fusion level while attempting to maintain as many motion segments as possible in the lumbar spine. The LIV should have near-neutral rotation, the disc below the LIV preoperatively should be parallel or wedged at the convexity, and the apex of the lumbar curve should be at least one disc level above the LIV. The TL/L curve is often more flexible than the MT curve. ⁷⁴ The goal is to render the LIV horizontal on the AP film. Often, type 3 curves will have TL kyphosis, and it is important to correct this regional deformity during the surgical procedure. ⁷⁵ Instrumentation levels should not end in this area of kyphosis but should be extended distally to avoid postoperative sagittal decompensation.

A small percentage of type 3 curves show dominant thoracic curve characteristics. Careful evaluation of these curves may allow the surgeon to perform a selective thoracic fusion, which will be discussed elsewhere.

11.9.4 Type 4: Triple Major Curves

The Lenke type 4 triple major curve pattern involves all three regions of the spine. The majority of patients with this pattern require fusion from T2 or T3 to L3 or L4. Again, the evaluation of shoulder balance is critical for determining the proximal level of fusion. Optimal horizontalization of the LIV is the primary goal corresponds with the previous discussion of type 1, 2, and 3 curves, respectively. Rarely, a selective thoracic or double thoracic fusion can be performed, leaving the lumbar curve unfused. In these cases, the clinical and radiographic ratios of the thoracic-to-lumbar curve are greater than 1.2 in favor of the larger thoracic curve (Fig. 11‑6).

11.9.5 Type 5: Thoracolumbar/Lumbar Curves

The Lenke type 5 curve pattern demonstrates a major curve in the TL/L region with a minor nonstructural curve in the MT region. Curves of this pattern may be treated with either ASF or PSF. Occasionally, we still utilize anterior spinal instrumentation and fusion from the upper to the lower end vertebrae of the curvature. The surgical levels include all convex discs within the curve. ^{76 , 77} Instrumentation is achieved with a dual-rod system. Hurford et al ⁷⁸ reviewed 48 TL/L curves treated with dual-screw/dual-rod constructs and found coronal correction of the TL/L curve averaging 75%, excellent sagittal alignment, and no instrumentation failure or pseudarthrosis at a minimum follow-up of 2 years. However, a trend toward posterior instrumentation is developing because of disadvantages of the anterior thoracolumbar approach. In selecting posterior fusion levels, the UIV and LIV are usually identical to those in ASF for the same curves. For small, very flexible curves, the minimal fusion technique of Hall and colleagues, ³⁹ as previously described, may be used. It is extremely important to evaluate the MT and PT regions as well as shoulder balance. If the left shoulder is elevated, some residual tilt must be maintained at the UIV to aid with postoperative shoulder balance. Correction of the lumbar curve and secondary correction of the compensatory thoracic curve will cause further elevation of the left shoulder. Therefore, careful evaluation of shoulder balance is necessary, including examination of the scapulae and thoracic prominence. If there is no MT component to the curve, the UIV and LIV may be horizontalized for correction. The goal is minimal deformity above and below the thoracolumbar fusion levels (Fig. 11‑7).

Excellent results can be achieved with selective lumbar fusion. Sanders et al ⁷⁹ reviewed 49 AIS patients who had undergone selective fusion for TL/L curves of 30 to 55 degrees. All of the patients were followed up for 2 years. Satisfactory results were achieved in patients with a TL/L-to-MT Cobb-angle ratio of greater than 1.25 and a thoracic curve that bent to 20 degrees or less. These satisfactory results were defined as a thoracic curve at follow-up of less than 40 degrees, acceptable balance and sagittal alignment, and no need for further procedures.