Which Sagittal Assessment System Can Effectively Predict the Risk of Mechanical Complications in the Treatment of Elderly Patients With Adult Degenerative Scoliosis? Roussouly Classi cation or Global Alignment and Proportion (GAP) Score


 Background: In order to achieve the proper sagittal alignment, previous studies have developed different assessment systems for degenerative spinal deformity which could help us in making treatment strategies. The purpose of our study is to evaluate whether Roussouly classification or GAP score is more appropriate in the prediction of mechanical complications in the treatment of ADS.Methods: The ADS patients who received long segmental fusion in the treatment during the period from December 2016 to December 2018 were evaluated in this study. The basic information of the patients and all radiologic measurements, which were included in GAP score and Roussouly classification, were collected for analysis. Patients were divided into two groups according to occurrence or absence of mechanical complications for comparison. The correlation between evaluation systems and mechanical complications could be analyzed in logistic regression model via stepwise backward elimination based on the Wald statistics. ROC curve was used to determine the predictability of the evaluation systems in the occurrence of mechanical complications and calculate their cut-off value. A two-tailed P value < 0.05 was statistically significant for all statistical tests.Results: A total of 80 cases were included in this study. The results of logistic regression showed: GAP score (P = 0.008) and GAP categories (P = 0.007) were positively correlated with Mechanical complications; Roussouly score was negatively correlated with mechanical complications (P=0.034); GAP score was positively correlated with PJK (P = 0.021); Roussouly score was negatively correlated with implant-related complications (P = 0.018); GAP categories were correlated with implant loosening (P = 0.023). Results of ROC showed that GAP score was mostly effective in predicting PJK (AUC = 0.863) and PJF (AUC = 0.724); GAP categories (AUC = 0.561) was more effective than GAP score (AUC = 0.555) in predicting implant-related complications.Conclusions: Roussouly-type matching could not accurately predict the risk of mechanical complications. In contrast, GAP score was mostly effective in predicting PJK and PJF. The GAP score was better than Roussouly classification in predicting mechanical complications.


Introduction
The three-dimensional deformity occurs in patients with adult degenerative scoliosis (ADS). The coronal correction of frontal deformity was the principle concerned in the past; ADS was found to be deeply affected by the rotational thoracolumbar kyphosis which could alter the sagittal pro le [1]. Nowadays, more attention is paid to the sagittal deformity. It was reported that spinal degeneration could decrease lumbar lordosis, increasing thoracic kyphosis, changing the ideal sagittal alignment [2]. In order to achieve the proper sagittal alignment, previous studies have developed different assessment systems for degenerative spinal deformity which could help us in making treatment strategies, such as Scoliosis Research Society (SRS)-Schwab classi cation [3], Roussouly classi cation [4] and Global Alignment and Proportion (GAP) Score [5].
According to the SRS-Schwab classi cation [3 , 6], three targets for corrective surgery realignment were suggested: the pelvic incidence (PI) minus lumbar lordosis (LL) mismatch is less than 10°; pelvic tilt (PT) is less than 20°; sagittal vertical axis (SVA) is less than 4 cm. However, even after matching the targets of Schwab criteria, the mechanical complication rates remain very high [7]. This classi cation is not effective neither in making the treatment strategy nor in predicting clinical outcome, especially when there is no sagittal malalignment.
In Roussouly classi cation, 4 types of spinal alignments were described depending on sacral slope (SS) and the shape of LL. This classi cation was subsequently updated to a modi ed classi cation which included a new type, the anteverted type 3 [8]. This new type was characterized by low-grade PI, SS > 35°a nd low or negative PT [8]. All radiographic factors were compared with ideal spinal alignment to evaluate their deviations from the ideal parameters. In addition, the optimal sagittal alignment was determined on the rate of PI in proportion to these factors. This is because, PI is an unchanged parameter [5]. Roussouly classi cation contribute to the determination of high local stress zones in the whole spine. In this classi cation, the lower the lumbar lordosis or at back, the higher the stress is on the disks; the more the lumbar lordosis increased, the more is the contact force on the posterior column [9]. Roussouly classi cation may help the surgeon predict the best rods bending and the best correction degrees to achieve optimal results. However, degenerative spine modi es the organization of the spinal curves which is responsible for the compensation mechanisms at the spine level or in the pelvis, hips and knees. This can make it di cult to use Roussouly classi cation in degenerative conditions [10].
Apart from Roussouly classi cation to help making surgical strategies, GAP score is an alternative that uses PI-based sagittal parameters to quantify the shape and alignment of the sagittal plane. Both Roussouly classi cation and GAP score share the similar principles to achieve the optimal spinopelvic alignment which includes the restoration of ideal LL, ideal pelvic version, and the ideal lordosis distribution [5]. Planning surgical target in the sagittal plane based on the proportional indices via GAP score can decease the occurrence of mechanical complications [11]. However, no study has compared the effectiveness of these two evaluation systems in predicting mechanical complications after long segmental fusion in the treatment of ADS. Therefore, the purpose of our study is to evaluate whether Roussouly classi cation or GAP score is more appropriate in the prediction of mechanical complications in the treatment of ADS.

Inclusion and Exclusion criteria
The ADS patients who received long segmental fusion in the treatment during the period from December 2016 to December 2018 were evaluated in this study. The basic information of the patients, such as gender, age, BMI (body mass index), follow-up time, blood loss, operation time, vertebrae fused, visual analogue scale (VAS), Japanese Orthopaedic Association (JOA), Oswestry Disability Index (ODI) were collected. The inclusion criteria included: age > 60 years at the time of attendance; more than 4 vertebral levels were fused; coronal Cobb angle (CA) ≥ 20°, SVA ≥ 5cm, PT ≥ 25°, and thoracic kyphosis (TK) ≥ 60°; the follow-up time should be more than 2 years. Exclusion criteria included: previous spinal fusion; ADS secondary to syndromic, autoimmune, infectious, tumor, or other pathologic conditions. The written informed consents were signed by all the included cases. The institutional review board approved this study following the declaration of Helsinki principles.

Radiographic Measurements and Scoring
All radiologic measurements included in GAP score and Roussouly classi cation, such as PI, PT, SS, thoracolumbar kyphosis (TLK), TK, LL, L4-S1 lordosis, global tilt (GT), SVA, the number of vertebrae included in the lordosis (NVL), the lumbar sagittal apex (LA) and the in exion point (IP) were recorded at 6 weeks postoperatively (Supplementary le 1). All radiographs were analyzed by validated software (Surgimap, Nemaris Inc., New York, NY). All data were measured separately by two researchers. When discrepancies arose, a consensus would be taken by the coauthors after discussion.
Standard values of these parameters for Roussouly types are shown in Supplementary le 2, as summarized previously [1]. It has been demonstrated that over-correction (a shape resembling a "higher" Roussouly type than ideal shape) in the realignment of ADS was worse than under-correction (realign the spine in a shape resembling a "lower" pro le than ideal shape) or ideal realignment in the treatment [1]. Roussouly modi ers of ADS patients were de ned in this study. The de nition was as follows: modi er "0", patients with ideal pro les; modi er "+", patients with under-corrected pro les; modi er "++", patients with over-corrected pro les. The Roussouly modi ers were then statistically weighted (1 for modi er "0", 2 for modi er "+" and 3 for modi er "++").
The GAP score ranges from 0 to 13 points. It included RPV (relative pelvic version), RLL (relative lumbar lordosis), LDI (lordosis distribution index), RSA (relative spinopelvic alignment) and age [5]. The cut-off values of GAP score are as follows: a GAP score of 0 to 2 indicates a proportioned spinopelvic position; a GAP score of 3 to 6 is de ned as moderately disproportioned; a GAP score more than 6 is de ned as severely disproportioned (Supplementary le 3) [5].

Mechanical Complications
The mechanical complications discussed in this study included: proximal junctional kyphosis/ failure (PJK or PJF), distal junctional kyphosis/ failure (DJK or DJF), and implant-related complications [5]. PJK was de ned as the kyphosis between UIV (upper instrumented vertebra) and UIV + 2 increased ≥ 10° in between early postoperative and follow-up radiographs. PJF was the fracture of UIV or UIV + 1, pullout of instrumentation at UIV, and/or sagittal subluxation. DJK or DJF was the postoperative kyphosis angle between LIV (lower instrumented vertebra) and LIV-1 increased ≥ 10°, and/or pullout of instrumentation at LIV. Implant-related complications were other radiographic implant-related complications such as implant loosening, implant breakage or implant pullout. Patients were divided into two groups according to occurrence or absence of mechanical complications for comparison.

Statistical Analysis
The statistical analysis was performed using the SPSS 17.0 (SPSS Inc, Richmond, CA, USA). Continuous variables were reported as mean ± standard deviations. Kolmogorov-Smirnov test was performed to the normal distribution of the data. Normally distributed values were analyzed with independent Student t test. Skewed values were analyzed with Kruskal-Walllist test. Categorical variables were reported as the number of cases and compared using Pearson's Chi-square test. The correlation between evaluation systems and mechanical complications could be found by odds ratio (OR) and 95% con dence interval (CI) in logistic regression model via stepwise backward elimination based on the Wald statistics. ROC (receiver operating characteristic) curve was used to determine the predictability of the evaluation systems in the occurrence of mechanical complications and calculate their cut-off value. A two-tailed P value < 0.05 was statistically signi cant for all statistical tests.

Demographics
A total of 80 cases were included in this study ( Table 1). The mean age is 76.5 ± 2.5 years old. The mean follow-up is 19.3 ± 6.2 months. Implant-related complication (42.5%) has the highest incidence in mechanical complications (51.3%). The most common implant-related complication is implant loosening (37.5%). The postoperative radiographic parameters and clinical scoring systems were signi cantly improved compared with the preoperative data ( Table 2).

Comparison of parameters in Roussouly classi cation
More cases without mechanical complications were Roussouly-type 1 compared to those with mechanical complications (P = 0.035). Compared to cases with mechanical complications, there was more patients without mechanical complications matching ideal LA (P < 0.001). There were more patients who matched Roussouly-type in no mechanical complication group compared with that in mechanical complication groups (P = 0.048). The Roussouly score in no mechanical complication group was more than that in mechanical group (P = 0.032) ( Table 3).

Comparison of parameters in GAP score
The GAP score in mechanical group was higher than that in no mechanical complication (P = 0.005). The Post-RPV score (P = 0.003) and Post-GT (P = 0.007) in mechanical group were signi cantly higher than those in no mechanical group. The Post-RPV (P = 0.019) and Post-RLL (P = 0.006) in mechanical group were signi cantly lower than those in no mechanical group. The number of patients with moderately disproportioned GAP score in no mechanical group was more than that in mechanical group (P = 0.010). There were more patients with severely disproportioned GAP score in mechanical group compared with those in no mechanical group (P = 0.003) ( Table 4).

Correlations between evaluation systems and mechanical complications
The results of logistic regression showed: GAP score (P = 0.008) and GAP categories (P = 0.007) were positively correlated with Mechanical complications; Roussouly score was negatively correlated with mechanical complications (P = 0.034); GAP score was positively correlated with PJK (P = 0.021);

ROC of evaluation systems in predicting mechanical complications
Results of ROC showed that GAP score was more effective in prediction mechanical complications than Roussouly classi cation (Fig. 1). GAP score (Cut-off value = 10) was mostly effective in predicting PJK (AUC = 0.863) and PJF (AUC = 0.724). GAP categories (AUC = 0.561, Cut-off value = Severely disproportioned) was more effective than GAP score (AUC = 0.555, Cut-off value = 5) in predicting implant-related complications; In other respects, however, the GAP score is superior to the GAP categories (Table 6).

Discussion
Previous studies showed that the postoperative complication rates (8.4-42%), revision rates (9-17.6%) in ADS were still high, which could increase after long-term follow-up [14 , 15]. Increased junctional stress concentration ADS to the collapse of the implant or vertebra; this could cause mechanical complications such as PJK, distal junction kyphosis (DJK), pseudoarthrosis, rod breakage or vertebral fracture [5 , 9 , 11 , 16]. The patients included in this study were all elderly, and therefore had poor tolerance to spinal stress concentration. The most common mechanical complication in this study was screw loosening. This is due to a decrease in bone density in older patients. Therefore, the stress concentration on the contact surface between screw and bone structure can be alleviated by screw loosening [16].
Several parameters are reported to be associated with mechanical complications, including TK, SVA, and SS [16][17][18][19]. The PI-LL in SRS-Schwab classi cation could be used to quantify spinopelvic harmony, which is also thought to be an important vital factor for predicting mechanical complication and is usually used to predict better postoperative health-related quality of life [20 , 21]. However, this kind of classi cation is based on linear absolute numerical parameters, which could not include the whole spectrum of PI. The realignment targets in SRS-Schwab classi cation could be misleading when the PI values were near the ends of normality (too high or too low) [12]. Therefore, previous studies used the degree of compensation and present PI to estimate the ideal spinal alignment according to Roussouly-types [22 , 23].
Roussouly de ned four basic shapes of normal sagittal spine alignment in healthy population based on SS [4]. However, lumbar degeneration and thoracolumbar coronal deformity could modify lumbar lordosis, which could consequently in uence SS [2]. Therefore, SS becomes an inadequate parameter to classify sagittal types in pathologic patients. In addition, the Roussouly classi cation relies on PI which is considered not to vary with age, pathology, or compensation [24]. However, Roussouly classi cation is mainly based on the classi cation of normal spine; most of the studies related to the compensatory mechanism of spinal degeneration were cross-sectional studies [12 , 22 , 25]. In this study, more cases without mechanical complications were Roussouly-type 1 compared to those with mechanical complications. This is because Roussouly-type 1 is a combination of long kyphosis and short lordosis at the lower arc of the spine; the in exion point, which always represent the region with highest junctional stress concentration, has already been xed in the central structure of the long-segment internal xation system [26]. Our study showed: there were more patients who matched Roussouly-type in no mechanical complication group compared with that in mechanical complication groups; compared to cases with mechanical complications, there was more patients without mechanical complications matching ideal LA. These results suggested that the difference in Roussouly type matching between the two groups was mainly due to the ideal LA matching rather than the ideal IP matching. Changing the original IP of the spine can easily lead to overcorrection of spinal deformities, thus increasing the stress on the internal xation system and then the risk of mechanical complications. Therefore, it is more important to adjust LA of ADS patients during surgery. Our study showed: there was no signi cant correlation between Roussouly-type matching and mechanical complications; the ROC analysis implied that Roussouly-type matching could not accurately predict the risk of mechanical complications. Roussouly-type only morphologically described the sagittal characteristics of ADS patients, which lacked three-dimensional analysis and quantitative indicators of the spinal deformity in ADS patients. Therefore, the usage of Roussouly classi cation in the realignment of ADS is di cult.
In contrast, the GAP score quanti es the imaging parameters as well as the age of the patients, thus intuitively predicting the risk of postoperative mechanical complications in ADS patients [1]. However, there is no study comparing Roussouly classi cation with GAP score in their effectiveness of predicting the risk of mechanical complications following ADS surgery. In our study, GAP was better than Roussouly classi cation in predicting mechanical complications; GAP score was mostly effective in predicting PJK and PJF; however, the prediction accuracy of GAP for implant breakage and DJK or DJF is low. This is because implant breakage is closely related to the material properties of the internal xation system itself, the living habits of patients and the overall structure of the internal xation. The occurrence of DJK is affected by many factors, such as the distal xation method, the severity of ADS and the levels of internal xation; these factors are not fully re ected in GAP score, so the accuracy of prediction is also low [22].
There are some limitations in this study. First, because older patients are more sensitive to spinal sagittal orthodontics, the patients included in this study were older than those in previous studies.This may make the results of this study different from those of previous studies. However, the age span of patients included in this study was small, so the conclusion of this study is more accurate when applied to elderly patients. Second, this study only analyzed the parameters involved in Roussouly classi cation and GAP score, while did not discuss the conditions of paraspinal muscles and lower limb compensations. This will prevent the results of this study from explaining all the causes of postoperative mechanical complications. For example, if the paravertebral muscles of the thoracic vertebrae are weak and the PT is large (pelvic compensation is poor), the compensatory capability of the patient without lower limb compensation will be poor; then the sagittal imbalance of the spine will develop quickly. Despite the de ciencies mentioned above, this study compared the Roussouly classi cation with the GAP score through careful statistical analysis, and the results of this study were still of high reference value. A new classi cation method, considering the compensation of patient-speci c spinal alignment and spinal balance, is still needed.

Conclusion
In matching Roussouly-type during surgical treatment for elderly ADS patients, changing the original IP of the spine can easily lead to overcorrection of spinal deformities, thus increasing the stress on the internal xation system and then the risk of mechanical complications. It is more important to adjust LA of ADS patients during surgery. Roussouly-type matching could not accurately predict the risk of mechanical complications. In contrast, GAP score was mostly effective in predicting PJK and PJF. The GAP score was better than Roussouly classi cation in predicting mechanical complications. Supplementary Files