Age at Initiation and Deformity Magnitude Influence Complication Rates of Surgical Treatment With Traditional Growing Rods in Early-Onset Scoliosis

Spine Deformity 4 (2016) 344e350 www.spine-deformity.org

Age at Initiation and Deformity Magnitude Influence Complication Rates of Surgical Treatment With Traditional Growing Rods in Early-Onset Scoliosis Vidyadhar V. Upasani, MD*, Kevin C. Parvaresh, MD, Jeff B. Pawelek, BS, Patricia E. Miller, MS, George H. Thompson, MD, David L. Skaggs, MD, MMM, John B. Emans, MD, Michael P. Glotzbecker, MD, Growing Spine Study Group Department of Orthopedic Surgery, Rady Children’s Hospital San Diego, University of California San Diego, 3030 Children’s Way, Suite 410, San Diego, CA 92123, USA Received 4 January 2016; revised 26 February 2016; accepted 3 April 2016

Abstract Study Design: Multi-center retrospective review. Objective: The purpose of this study was to identify preoperative variables associated with postoperative complications in early-onset scoliosis (EOS) patients treated with traditional growing rods (TGR); and to develop a model to predict the incidence of postoperative complications based on preoperative variables. Summary of Background Data: TGRs are commonly used to treat progressive EOS. Prior research has demonstrated a high rate of postoperative complications using this technique; however, few studies have identified preoperative factors that contribute to such complications. Methods: A total of 110 patients who initiated TGR treatment before 10 years of age and completed final treatment were identified from a multi-center database. Overall treatment effect was calculated for major curve size, thoracic kyphosis, thoracic height, and total spine height. Univariable and multivariable logistic regression identified preoperative predictors of complications. An algorithm was developed and validated to calculate the probability of complications based on preoperative data. Results: All patients completed TGR treatment (average follow-up 8.1 years). The overall treatment effect was a significant decrease in major curve magnitude, increase in thoracic height, increase in spine height, and no significant change in thoracic kyphosis. There were 263 total complications in 87 patients (79%) resulting in 84 unplanned surgeries. The most common complications were implant-related (49%), surgical site infection (23%), medical (19%), alignment (6%), and neurologic (3%). The significant independent preoperative predictors of complications were age at implantation and preoperative thoracic kyphosis. Multivariable regression showed that age less than 7.6 years,

The Growing Spine Foundation (GSF) provided financial support for this study; however, no authors or institutions received financial compensation for this work. In addition, the GSF provided the research data for this study. The GSF receives donations from the study group’s surgeon members, medical device industry, grateful patients, and other donors. Author disclosures: VVU (personal fees from OrthoPediatrics, grants from POSNA, outside the submitted work); KCP (none); JBP (none); PEM (none); GHT (personal fees from Shriner’s Hospital Medical Advisory Board, personal fees from Wolters Kluwer, other from OrthoPediatrics, outside the submitted work); DLS (nonfinancial support from Pediatric Orthopaedic Society of North America, nonfinancial support from Scoliosis Research Society, personal fees from Biomet, personal fees from Medtronic, personal fees from Zipline Medical, Inc., personal fees from Orthobullets, personal fees from Grand Rounds, grants from Pediatric Orthopaedic Society of North America, grants from Scoliosis Research Society, personal fees from Biomet, personal fees from Medtronic, personal fees from Johnson & Johnson, other from Wolters 2212-134X/$ - see front matter Ó 2016 Scoliosis Research Society. http://dx.doi.org/10.1016/j.jspd.2016.04.002

Kluwer HealtheLippincott Williams & Wilkins, other from Biomet Spine, other from Zipline Medical, Inc., outside the submitted work; in addition, DLS has a patent Medtronic issued, and a patent Biomet issued); JBE (other from journal Children’s Orthopedics, personal fees from Medtronics, personal fees from DePuy Synthes, other from DePuy Synthes, outside the submitted work); MPG (personal fees from Medtronic, personal fees from DePuy, other from Children’s Spine Study Group, outside the submitted work). Ethical approval: All participating institutions obtained IRB approval prior to study initiation. The medical devices mentioned in this paper have been cleared by the US Food and Drug Administration. *Corresponding author. Department of Orthopedic Surgery, Rady Children’s Hospital San Diego, University of California San Diego, 3030 Children’s Way, Suite 410, San Diego, CA 92123, USA. Tel.: 858-9666789; fax: 858-966-8519. E-mail address: [email protected] (V.V. Upasani).

V.V. Upasani et al. / Spine Deformity 4 (2016) 344e350

345

thoracic kyphosis greater than 38 degrees, or major curve magnitude greater than 84 degrees significantly increased the probability of complications. Conclusions: Earlier age at implantation, greater thoracic kyphosis, and larger major curves increased the probability of complications following TGR instrumentation. These findings provide a valuable tool for predicting complications that may aid in surgical planning and shared decision making with patients and their families. Level of Evidence: IV. Ó 2016 Scoliosis Research Society. Keywords: Early-onset scoliosis; Traditional growing rods; Postoperative complications; Deformity magnitude

Introduction Early-onset scoliosis (EOS) is defined as a spinal deformity developing in patients 10 years old or younger [1]. Deformity progression is associated with multiple complications, with the majority involving pulmonary compromise [2]. Additionally, multiple late complications are common even with appropriate management, including infection, premature spinal fusion, and deformity progression [3]. Nonsurgical options include orthotic bracing and casting. The effectiveness of these techniques are often limited by thorax and spine compliance, rib cage plasticity, pulmonary compromise, and skin breakdown. Surgical management can improve chest wall and pulmonary development; however, complications including surgical site infection, implant failure, and curve progression are frequent. Various growth-friendly surgical options have been studied in EOS patients to both preserve spinal growth and control deformity. As suggested by Skaggs et al., these techniques may be broadly grouped into three main categories: distraction-based, tension-based, and guided-growth [4]. The most common distraction-based technique is the use of traditional growing rods (TGR), in which instrumentation is placed and sequentially lengthened as the child grows until skeletal maturity is approached and a final fusion performed. Prior research has shown the efficacy of such instrumentation [5,6], but has also reinforced the high incidence of complications (48%-90%) associated with initial instrumentation and subsequent repeated surgical lengthenings [7,8]. Few prior studies have investigated specific factors contributing to postoperative complications. Bess et al. reported a 58% overall complication rate in 81 of 140 EOS patients treated with TGR [9]. The authors found significantly more unplanned procedures due to implant complications in patients treated with single rather than dual rods but no overall difference in complication rate. Further, they showed that the complication risk increased with younger age at initial surgery and subsequent lengthening procedures. In a later commentary, Karol underscored the fact that many patients in this study had not yet undergone final fusion; thus, true complication rates are likely to even be higher [10]. The primary purpose of this study was to identify preoperative variables associated with postoperative complications in EOS patients treated with TGR. The secondary

goal was to develop a model to predict the incidence of postoperative complications based on preoperative variables. Materials and Methods A multicenter international database of EOS patients treated with single and dual TGR was retrospectively reviewed following Institutional Review Board approval. Patients who initiated TGR treatment before 10 years of age or less and completed treatment were included. Patients with incomplete preoperative medical records, inadequate imaging studies, or without completion of treatment were excluded. A total of 110 patients (49 male and 61 female) met our inclusion criteria. Demographic, radiographic, and complication data were recorded for each patient. Demographic data included diagnosis, age, date and details of index surgery (single or dual GR), American Society of Anesthesiologists (ASA) Physical Status Classification System at index surgery, dates and details of subsequent surgeries, and date of most recent follow-up. Radiographic data included measurement of the major and secondary curve magnitudes, sagittal plane measures (T5eT12 thoracic kyphosis, thoracolumbar kyphosis, lumbar lordosis, and maximum thoracic kyphosis), T1eT12 thoracic spine height (measured from the superior endplate of T1 to the inferior endplate of T12), and T1eS1 spinal height (measured from the superior endplate of T1 to the inferior endplate of S1). Time points analyzed included preoperative, initial postoperative, and most recent follow-up visits. Complication data included complication type and date. Statistical Analysis Complication rates along with 95% confidence intervals were estimated for surgical site infection, implant, alignment, neurological, and medical complications. Each complication rate was defined as the number of patients who experienced at least one complication out of the total number of patients in the study. All tests were two-sided, and p values less than .05 were considered significant. All analyses were conducted using SAS software, version 9.3 (SAS Institute, Inc., Cary, NC). Paired t tests were used to analyze the change in thoracic dimensions, including major curve magnitude, thoracic

346

V.V. Upasani et al. / Spine Deformity 4 (2016) 344e350

kyphosis, thoracic height, and total spine height at various time points throughout the study. The immediate effect of surgery on thoracic dimensions was assessed by measurements taken preoperatively and immediately postoperatively; the effect of lengthening over time was assessed with measurements taken postoperatively and prior to final fusion; the effect of fusion was assessed with measurements taken before and after final fusion; and the overall effect of treatment was assessed by preoperative measurements compared to measurements from the most recent follow-up. Student t test, analysis of variance, Spearman, and Pearson correlation analyses were used to assess the relationship between preoperative characteristics and age at implantation. Univariable and multivariable logistic regression were used to identify preoperative predictors of complication. Any variables that were significantly associated with age were adjusted for in multivariable analysis and then tested for effect on complication outcomes. The level of significance was set at p 5 .1 for univariate regression analysis and at p 5 .05 for multivariable regression and all subsequent analyses. Stepwise model selection procedures were implemented, in which variables were removed and added to models based on Akaike information criteria (AIC) and change in pseudoR2 value. Pseudo-R2 values were calculated using Cragg and Uhler’s pseudo-R2. Final model fit was assessed using the likelihood ratio test and maximum pseudoR2 value. Significant continuous covariates from adjusted analysis were dichotomized by identifying an optimal cutoff value that could separate subjects at higher versus lower risk for complication. Cutoffs were determined using receiver operating characteristic (ROC) analysis with Youden index. Youden index identifies the point on the ROC curve that is farthest from the line of nondiscrimination, maximizing both sensitivity and specificity simultaneously. The multivariable logistic regression model using binary covariates was tested for model fit, and odds ratios with 95% confidence intervals were estimated for significant independent predictors. A prediction algorithm based on the final multivariable logistic regression model was tested and validated to predict the likelihood of complication given age, preoperative major curve size, and preoperative sagittal kyphosis. Crossvalidated predicted probabilities were estimated using a 10-fold cross-validation procedure. ROC curves were constructed for model-predicted probabilities as well as for crossvalidated predicted probabilities to assess the diagnostic performance of the multivariable predictors in identifying when a subject might be at higher risk for complication. To quantify the diagnostic utility, the area under the curve (AUC) was estimated for both the predicted and the cross-validated predicted probabilities, along with exact 95% confidence intervals, and the AUCs were compared using a bootstrap method with 2,000 repeated samples.

Results There were 110 consecutive patients in the study including 49 boys and 61 girls, with a mean age of 6.3  2.1 years (range, 1.3 to 9.9 years) at initial implantation. There were 38 patients (35%) with a single TGR and 72 patients (65%) with a dual TGR at index surgery. The mean preoperative major curve magnitude was 76  20 degrees (range, 33 to 135 degrees) and mean preoperative thoracic kyphosis was 56  27 degrees (range, 11 to 124 degrees). Mean follow-up time was 8.1 years (range, 1.3 to 17.3 years). All patients were followed until completion of treatment. Ninety-nine patients (90%) underwent final fusion, whereas the remaining 11 patients (10%) had their growing rods retained without a final fusion after completing the lengthening process. Patient preoperative characteristics are summarized in Table 1. The treatment effects are summarized in Table 2. The immediate effect of surgery (preoperative to postoperative) was a significant decrease in major curve magnitude, significant decrease in thoracic kyphosis, significant increase in thoracic height, and a significant increase in spine height (p ! .01). At the end of the lengthenings (postoperative to prefinal fusion), we observed a significant increase in major curve magnitude, significant increase in thoracic kyphosis, significant increase in thoracic height, and significant increase in spine height (p ! .01). The effect of final fusion (prefinal to postfinal fusion) was a significant decrease in major curve magnitude, significant decrease in thoracic kyphosis, significant increase in thoracic height, and significant increase in spine height (p ! .01). The overall effect (preoperative to latest visit) of surgery was a significant decrease in major curve magnitude, significant increase in thoracic height, significant increase in spine height (p ! .01), and no significant change in thoracic kyphosis (p 5 .36). Table 1 Preoperative patient characteristics (N 5 110). Demographics Sex (% male), n (%) Age at implantation (years), M  SD Etiology, n (%) Congenital Idiopathic Neuromuscular Syndromic ASA level at implantation (n 5 96), n (%) 1 2 3 4 Preoperative radiographic measures, M  SD Major curve magnitude (degrees) Maximum sagittal kyphosis (degrees) Thoracic height (cm) Spine height (cm)

49 (45) 6.3  2.08 16 29 36 29

(15) (26) (33) (26)

6 44 43 3

(6) (46) (45) (3)

75.5  19.95 55.6  26.58 163.5  37.14 262.7  48.72

ASA, American Society of Anesthesiologists classification; M, mean; SD, standard deviation.

V.V. Upasani et al. / Spine Deformity 4 (2016) 344e350 Table 2 Effects of treatment.

347

Table 4 Univariate analysis. Mean % change 95% CI change

Major curve magnitude (degrees) Immediate effect of 32.5 surgery After completion of 13.3 lengthenings Effect of final fusion 15.8 Overall effect 27.2 Thoracic kyphosis (degrees) Immediate effect of 11.6 surgery Effect of lengthenings 11.3 Effect of final fusion 7.9 Overall effect 5.1 Thoracic height (mm) Immediate effect of 25 surgery Effect of lengthenings 26.2 Effect of final fusion 8.7 Overall effect 60.6 Spine height (mm) Immediate effect of 39.2 surgery Effect of lengthenings 32.5 Effect of final fusion 20.7 Overall effect 99.9

p

Variable

29.48 !.001

43

35.58,

31

9.2, 17.48

26 36

20.04, 33.11,

11.47 !.001 21.3 !.001

22

17.69,

5.56

.001

28 13 9

5.21, 17.48 13.29, 2.43 15.85, 5.62

.001 .007 .357

15

19.24, 30.73

!.001

14 4 37

18.14, 34.26 2.85, 14.46 49.21, 71.91

!.001 .005 !.001

15

33.38, 44.99

!.001

11 6 38

22.86, 42.12 13.96, 27.49 84.88, 114.94

!.001 !.001 !.001

!.001

CI, confidence interval.

There were 263 complications total (Table 3). The most common complications were implant related (49%), followed by surgical site infection (23%), medical (19%), alignment (6%), and neurologic (3%). At least one complication occurred in 87 patients (79%, 95% confidence interval 5 70%-86%). At least one complication occurred in 74% of single TGR patients and 82% of dual TGR patients, which was not significantly different even with Table 3 Postoperative complication details. Frequency (%) Surgical site complications Infectionsdsuperficial Infectionsddeep Unplanned surgery Implant complications Hook/screw dislodgement Rod fracture Prominent implants Unplanned surgery Alignment complications Junctional kyphosis Curve decompensation Unplanned surgery Neurologic complications Medical complications Cardiopulmonary Dural tear/gastrointestinal/other Unplanned surgery Total complications

60 (23%) 34 26 21 129 (49%) 43 57 29 53 16 (6%) 10 6 4 7 (3%) 51 (19%) 38 13 6 263

At Least One Complication (n 5 87)

Demographics Sex (% male), n (%) 40 (46%) Age at implantation 6  2.06 (years) Etiology, n (%) Congenital 14 (16%) Idiopathic 14 (28%) Neuromuscular 24 (28%) Syndromic 25 (29%) ASA level at implantation, n (%) 1 5 (7%) 2 33 (44%) 3 34 (45%) 4 3 (4%) Number of lengthenings 6 (1-19) Duration of follow-up 8.2  3.48 (years), M  SD Construct type, n (%) Single 28 (32%) Dual 59 (68%) Preoperative radiographic measures, M  SD Maximum curve size 77.8  20.24 (degrees) Maximum sagittal 58.7  25.73 kyphosis (degrees) Thoracic height (mm) 158.9  34.46 Spine height (mm) 256.1  47.32

No Complications (n 5 23)

p

9 (35%) 7.5  1.83

.56 .008

2 5 12 4

d .68 .13 .90

(9%) (22%) (52%) (17%)

1 (5%) 11 (52%) 9 (43%) 0 (0%) 3 (2-12) 7.5  3.69

d .66 .81 .99 .04 .41

10 (43%) 13 (57%)

.31

67.3  16.79

.03

42.2  27.28

.07

184.5  43.06 285.6  47.50

.03 .01

ASA, American Society of Anesthesiologists classification; M, mean; SD, standard deviation.

multivariable adjustment (p 5 .33). A total of 84 unplanned surgeries were performed because of complications. Univariate analysis for variables associated with postoperative complications is shown in Table 4. Preoperative factors significantly associated with complications were age at implantation (p 5 .008), major curve magnitude (p 5 .03), thoracic height (p 5 .03), maximum thoracic kyphosis (p 5 .07), number of lengthenings (p 5 .04), and spine height (p 5 .01). Sex, etiology, ASA level, construct type (number of rods), and duration of follow-up were not correlated with complication. Sex, etiology, ASA level, and preoperative thoracic and spine height were all significantly associated with age (p 5 .01, .007, .01, !.001, and !.001, respectively). Also, the number of lengthenings, the number of revision Table 5 Multivariable analysis for predictors of complication. Predictor

OR (95% CI)

p

Age 7.6 years or less Preoperative thoracic kyphosis O38  degrees Preoperative major curve O84 degrees

10.1 (2.12, 70.12) 6.0 (1.13, 39.15) 3.6 (0.59, 38.9)

.003 .03 .18a

CI, confidence interval; OR, odds ratio. Adjusting for preoperative curve size decreased the standard error of the model and was therefore included as a predictor of complication. a

348

V.V. Upasani et al. / Spine Deformity 4 (2016) 344e350

Table 6 Algorithm for probability of complication. Under 7.6 years of age at implantation

Preoperative maximum sagittal kyphosis O38 degrees

Preoperative major curve magnitude O84 degrees

Predicted probability of complication (%)

Yes Yes Yes Yes No No No No

Yes Yes No No Yes Yes No No

Yes No Yes No Yes No Yes No

98.3 94.3 90.9 73.6 85.5 62.3 49.7 21.7

surgeries, and the length of follow-up were significantly correlated with age (p ! .001 for all). Preoperative major curve magnitude (p 5 .07) and maximum thoracic kyphosis (p 5 .38), however, were not associated with age. Multivariable regression analysis identified only two significant independent predictors of complication: age of 7.6 years or less at implantation and preoperative thoracic kyphosis of 38 degrees or more, while controlling for preoperative major curve magnitude of 84 degrees or more. Adjusting for preoperative major curve magnitude decreased the standard error of model coefficients and increased the pseudo-R2 of the model to 87%. Adjusted odds ratios with 95% confidence limits are reported in Table 5. An algorithm was developed and cross validated to predict the probability of a subject experiencing at least one complication based on age of 7.6 years or younger, preoperative thoracic kyphosis of 38 degrees or more, and preoperative major curve magnitude of 84 degrees or more (Table 6). For example, a patient who was younger than 7.6 years at implantation with a major curve greater than 84 degrees and thoracic kyphosis greater than 38 degrees had a 98% predicted probability of experiencing at least one 1

Sensitivity

0.8

0.6

Area under the curve = 0.89 95% CI = (0.78-1.00)

0.4

0.2

0

0

0.2

0.4

0.6

0.8

1 - Specificity

Fig. Receiver operating characteristics for prediction algorithm.

1

complication. Conversely, a patient 7.7 years or older, with thoracic kyphosis of 38 degrees or less and a major curve magnitude less than 84 degrees had only a 22% predicted probability of experiencing at least one complication. The algorithm was evaluated using ROC curve analysis (Fig.). The AUC for the predicted probabilities was 0.89, indicating good diagnostic performance of the algorithm. In addition, the AUC for the cross-validated predicted probabilities was 0.81, and the two curves were not found to be significantly different (p 5 .14), thus verifying that the algorithm could be applied to future data in this population with comparable accuracy. Discussion The primary goals of EOS management are preservation of spinal growth as well as chest wall and pulmonary development, stabilization or correction of the spinal deformity, and achieving a stable fusion at completion of treatment. Instrumentation with TGR allows for continued spinal growth, which facilitates pulmonary development and reduces the risk of curve progression. Multiple factors have been proposed to guide initiation of TGR management including age, sex, family history, curve magnitude and location, and ribevertebral angle difference [11,12]. Prior studies suggest that early instrumentation with TGR promotes greater pulmonary development [13]. However, more recent research challenges the pulmonary data used in earlier studies as it was not age- and height-matched to controls [14]. Early insertion is also fraught with multiple complications and generally requires additional surgical procedures [15]. Despite multiple EOS studies assessing the impact of these factors on curve progression, few studies have specifically evaluated the association of preoperative patient factors and postoperative prognosis. Because of the contrasting sequelae of early and late intervention, initiation of TGR instrumentation remains controversial in EOS management. This study was designed to further elucidate preoperative variables associated with postoperative complications for EOS. To our knowledge, this is the largest cohort evaluated with regard to TGR complications in EOS patients that have all completed treatment. Moreover, this is a multicenter study from an international database with extensive quality assurance. Our overall treatment effects for major curve magnitude (27 degree decrease) and sagittal alignment (5 degree decrease) match previous results for TGR in the literature [4,5,7]. Our overall treatment effects for T1eT12 height (6.1 cm increase) and T1eS1 height (10 cm increase) also parallel the results of prior studies by Akbarnia et al. [7] (9.64 cm T1eS1 increase) and Kamachi et al. [4] (6 cm T1eT12 increase, 9 cm T1eS1 increase). Such findings supplement the significance of TGR preserving spinal growth while controlling curve morphology. Our findings expand on prior knowledge that surgical management of EOS is associated with multiple

V.V. Upasani et al. / Spine Deformity 4 (2016) 344e350

complications. Bess et al. found a high overall rate of complications (58%) in 81 of 140 patients treated with single and dual TGR [8]. They also identified age at initiation of surgery and the number of procedures as significant predictors of postoperative complications. Although they found more unplanned surgeries for implant complications in single versus dual TGR, they similarly found no significant differences in overall complication rate between single and dual TGR constructs. While we found that age was an independent predictor of complications, we did not assess number of lengthenings as an independent predictor because it is not a known objective preoperative patient factor and it did not correlate with complication rates. In another retrospective multicenter study examining dual growing rod management of EOS, Watanabe et al. also found a high complication rate for instrumentation (57%) [16]. Their findings showed that thoracic curve magnitude and degree of kyphosis were significantly correlated with the complication rate. Although they did not find that age at index surgery correlated with complication rate, they admittedly had patients who had not yet completed treatment. Additionally, thoracic hyperkyphosis greater than 40 degrees was recently shown to be associated with implant complications, especially in syndromic patients [17]. Our data re-demonstrate age, kyphosis, and curve size as preoperative factors associated with postoperative complications. The patient’s diagnosis was not associated with complications in this cohort. Our sample size was likely not large enough to demonstrate a significant difference between the idiopathic patients and the other types. For example, we performed a power analysis and found that we would need 76 patients in each group to find a statistically higher rate of complications in the neuromuscular patients. Magnetically controlled growing rods (MCGRs) are a newer distraction-based technique for EOS management that offer the potential advantage of nonoperative lengthenings in an outpatient clinic setting rather than repeated surgical lengthenings. Short-term data for patients currently undergoing MCGR treatment suggest similar initial treatment effects for curve correction (42%-58%), T1eT12 height increase (3 cm), and T1eS1 height increase (2.3-5.4 cm) compared with TGR [18-21]. Early pulmonary studies suggest that MCGR significantly increases forced vital capacity [22]. Although initial MCGR implantation is more costly, surgical savings over time may prove it to be more cost efficient compared to TGR methods [23]. Long-term studies are forthcoming, and more data are needed in regard to overall treatment effects and complications of MCGR instrumentation. Ultimately, time will tell whether MCGRs offer a significant advantage over TGR for EOS management. Our results provide a valuable preoperative tool for predicting postoperative complication rates. Single and dual GR constructs showed similar complication rates, suggesting that construct type was not correlated with the complication rate. We found the three most important

349

preoperative predictors for postoperative complications were age, maximum kyphosis, and curve magnitude. These variables can be combined to preoperatively predict complication rates with reasonable strength based on our ROC analysis (AUC 0.89). The data suggest that in certain cases, it may be advisable to operate earlier on curves of small magnitude rather than wait until the patient is older but has a more significant curve. For example, based on our algorithm, a 6-year-old patient with a curve size of 80 degrees would have a 74% probability of having a complication; waiting until the same patient is 8 years old but has a larger curve more than 85 degrees would increase the probably to 91%, thus suggesting earlier intervention may help prevent complications under certain circumstances. Together, these findings may aid in surgical planning and shared decision-making discussions with families. There are a number of limitations associated with our study. We did not match our patients to a comparison or control group without surgical intervention to evaluate clinical differences in complication rates. It would be very difficult if not impossible to execute such a study, which would ideally require controls to be matched according to age, gender, height, and curve magnitude. We included all types of complications, whereas only clinically significant complication analyses might arguably provide more applicable results for surgical management. The finding that an age of 7.6 years or older significantly reducing complication rates may not have high clinical applicability as most surgeons do not use TGR implants over age eight. Regardless, it still provides an objective data point for prediction. Finally, our data analysis did not include a direct evaluation of pulmonary function or quality of life [24]. Thus, we cannot suggest our markers be used as surgical indications. Rather, more robust studies combining outcome data with complication rate analysis are needed to improve surgical management. Ideally our data may be combined with future research to develop an evidencebased algorithm to aid in surgical decision making for instrumentation in EOS.

References [1] El-Hawary R, Akbarnia B. Early onset scoliosisdtime for consensus. Spine Deform 2015;3:105e6. [2] Lloyd-Roberts GC, Pilcher MF. Structural idiopathic scoliosis in infancy: a study of natural history of 100 patients. J Bone Joint Surg Br 1965;47:520e3. [3] Phillips JH, Knapp Jr DR, Herra-Soto J. Mortality and morbidity in early-onset scoliosis surgery. Spine 2013;38:324e7. [4] Skaggs DL, Akbarnia BA, Flynn J, et al. Classification of treatment of early-onset scoliosis. Paper presented at: 2nd International Congress of Early Onset Scoliosis and Growing Spine; November 7e8, 2008; Montreal, Quebec, Canada. [5] Kamachi S, Demirkiran G, Ismavilov V, et al. The effect of dual growing rod instrumentation on the apical vertebral rotation in early-onset idiopathic scoliosis. J Pediatr Orthop 2014;34:607e12. [6] Thompson GH, Akbarnia BA, Campbell RM. Growing rod techniques in early-onset scoliosis. J Pediatr Orthop 2007;27:354e61.

350

V.V. Upasani et al. / Spine Deformity 4 (2016) 344e350

[7] Mineiro J, Weinstein SL. Subcutaneous rodding for progressive spinal curvatures: early results. J Pediatr Orthop 2002;22:290e5. [8] Akbarnia BA, Marks DS, Boachie-Adjei O, et al. Dual growing rod technique for the treatment of progressive early-onset scoliosis: a multicenter study. Spine 2005;30:S46e57. [9] Bess S, Akbarnia BA, Thompson GH, et al. Complications of growing-rod treatment for early-onset scoliosis: analysis of one hundred and forty patients. J Bone Joint Surg Am 2010;92:2533e43. [10] Karol LA. Commentary on an article by S. Bess, et al.: Complications of growing-rod treatment for Early-Onset Scoliosis. Analysis of one hundred and forty patients. J Bone Joint Surg Am 2010;92:e27. [11] Gillingham BL, Fan RA, Akbarnia BA. Early onset idiopathic scoliosis. J Am Acad Orthop Surg 2006;14:101e12. [12] Mehta MH. The rib-vertebra angle in the early diagnosis between resolving and progressive infantile scoliosis. J Bone Joint Surg Br 1972;54:230e43. [13] Jiang Y, Zhao Y, Wang YP, et al. Lung function after growing rod surgery for progressive early-onset scoliosis: a preliminary study. Chin Med J 2011;124:3858e63. [14] Newton PO, Faro FD, Gollogly S, et al. Results of preoperative pulmonary function testing of adolescents with idiopathic scoliosis. J Bone Joint Surg 2005;87:1937e46. [15] Sankar WN, Skaggs DL, Yazici M, et al. Lengthening of dual growing rods and the law of diminishing returns. Spine 2011;36:806e9. [16] Watanabe K, Uno K, Suzuki T, et al. Risk factors for complications associated with growing-rod surgery for early-onset scoliosis. Spine 2013;38:E464e8.

[17] Schroerlucke SR, Akbarnia BA, Pawelek JB; Growing Spine Study Group. How does thoracic kyphosis affect patient outcomes in growing rod surgery? Spine 2012;37:1303e9. [18] Dannawi Z, Altaf F, Harshavardana NS, et al. Early results of a remotely-operated magnetic growing rod in early-onset scoliosis. Bone Joint J 2013;95B:75e80. [19] Akbarnia BA, Cheung K, Noordeen H, et al. Next generation of growth-sparing techniques: preliminary clinical results of a magnetically controlled growing rod in 14 patients with early-onset scoliosis. Spine 2013;38:665e70. [20] Cheung KM, Cheung JP, Samartzis D, et al. Magnetically controlled growing rods for severe spinal curvature in young children: a prospective case series. Lancet 2012;379:1967e74. [21] Hickey BA, Towriss C, Baxter G, et al. Early experience of MAGEC magnetic growing rods in the treatment of early onset scoliosis. Eur Spine J 2014;23(suppl 1):S61e5. [22] Yoon WW, Sedra F, Shah S, et al. Improvement of pulmonary function in children with early-onset scoliosis using magnetic growth rods. Spine 2014;39:1196e202. [23] Rolton D, Richards J, Nnadi C. Magnetic controlled growth rods versus conventional growing rod systems in the treatment of early onset scoliosis: a cost comparison. Eur Spine J 2015;24: 1457e61. [24] Corona J, Matsumoto H, Roye DP, Vitale MG. Measuring quality of life in children with early onset scoliosis: development and initial validation of the early onset scoliosis questionnaire. J Pediatr Orthop 2011;31:180e5.