新型腰椎前凸计算回归方程对成人脊柱畸形矫形术后机械性并发症的预测作用

Predictive effect of a novel regression model calculating lumbar lordosis on the mechanical complications after corrective surgery of adult spinal deformity

中文关键词:  成人脊柱畸形  机械性并发症  骨盆入射角  胸椎后凸  腰椎前凸  腰椎顶点

英文关键词:Adult spinal deformity  Mechanical complication  Pelvic incidence  Thoracic kyphosis  Lumbar lordosis  Lumbar apex

中文摘要:

  【摘要】 目的：探讨由骨盆入射角（pelvic incidence，PI）和胸椎后凸（thoracic kyphosis，TK）计算腰椎前凸的新型回归方程对成人脊柱畸形（adult spinal deformity，ASD）矫形术后机械性并发症的预测作用。方法：回顾性分析2011年1月~2019年3月于我科接受后路矫形的ASD患者的临床资料。术后随访至少2年，收集患者的人口学和术前、术后即刻及末次随访时的影像学资料。根据Pan等提出的新型腰椎前凸计算回归方程由术后PI和TK计算腰椎前凸（lumbar lordosis，LL）、腰椎顶点（lumbar apex，LLA）和拐点（inflection point，IP）的理论值，其与术后即刻实际测量参数的差值定义为ΔLL、ΔLLA和ΔIP。根据术后是否发生机械性并发症将患者分为有机械性并发症组和无机械性并发症组，应用单因素和多因素二元Logistic回归分析对两组患者的临床资料进行比较，确定机械性并发症的独立危险因素；绘制受试者工作特征（receiver operating characteristic, ROC）曲线确定危险因素预测机械性并发症的最佳临界值。结果：共有92例患者纳入研究，男15例，女77例，随访时间2.00～4.60年（2.93±0.63年），其中41例术后出现机械性并发症。单因素分析显示有机械性并发症组患者具有较大的体重指数（24.32±2.09kg/m2 vs 23.34±2.26kg/m2）、术前整体倾斜角（global tilt，GT）（34.26°±18.24° vs 25.68°±15.61°）、术前骨盆倾斜角（27.70°±12.61° vs 22.47°±11.99°）、术后GT（23.29°±12.21° vs 18.35°±10.50°）、术后ΔLL（12.19°±6.39° vs 6.49°±5.49°）和ΔLLA（0.74±0.48 vs 0.45±0.28），且差异均有统计学意义（P＜0.05）。Logistic回归分析显示术后较大的ΔLL、ΔLLA是机械性并发症的独立危险因素（ΔLL OR=1.251， 95%CI =1.043～1.270，P=0.005；ΔLLA OR=4.357，95%CI=1.484～19.334，P=0.01）。根据两者的偏回归系数，定义联合指数＝ΔLL+ΔLLA×1.678/0.140。ROC曲线显示ΔLL、ΔLLA和联合指数的最佳临界值分别为9.17°、0.71和12.93，三者的曲线下面积分别为0.777、0.681和0.809。结论：ASD矫形时LL需要与 PI 和 TK匹配；术后LL、LLA与新型回归方程计算的理论值的差值（ΔLL、ΔLLA）可用于有效预测机械性并发症的发生。

英文摘要:

  【Abstract】 Objectives: To explore the predictive effect of a novel regression model calculating lumbar lordosis based on pelvic incidence(PI) and thoracic kyphosis(TK) on the mechanical complications after corrective surgery of adult spinal deformity(ASD). Methods: The clinical data of ASD patients undergoing posterior corrective surgery from January 2011 to March 2019 with follow-up for at least 2 years were retrospectively analyzed. Demographic and radiographic(preoperative, immediately postoperative and final follow-up) information were collected. According to the regression model proposed by Pan et al, the deviations of the actual lumbar lordosis(LL), lumbar lordosis apex(LLA) and inflection point(IP) immediately after surgery from the theoretical values derived from postoperative PI and TK were defined as ΔLL, ΔLLA and ΔIP. The patients were divided into two groups based on the presence or absence of mechanical complications. Univariate and multivariate analysis were performed to identify independent risk factors for mechanical complications. Finally, the receiver operating characteristics(ROC) curves were plotted to determine the best cut-off values for the independent risk factors. Results: A total of 92 patients were included, including 77 females and 15 males. The average follow-up time was 2.00-4.60 years(averaged 2.93±0.63 years), amd mechanical complications occurred in 41 patients. Univariate analysis showed that BMI(24.32±2.09kg/m2 vs 23.34±2.26kg/m2), preoperative global tilt(34.26°±18.24° vs 25.68°±15.61°), preoperative pelvic tilt(27.70°±12.61° vs 22.47°±11.99°), and postoperative global tilt(23.29°±12.21° vs 18.35°±10.50°) were significantly higher in the mechanical complications group. Besides, greater postoperative ΔLL(12.19°±6.39° vs 5.49°±5.49°) and ΔLLA(0.74±0.48 vs 0.45±0.28) were significantly associated with mechanical complications(P＜0.05). Logistic regression analysis showed that greater postoperative ΔLL(OR=1.251, 95%CI=1.043-1.270, P=0.005) and ΔLLA(OR=4.357, 95%CI=1.484-19.334, P=0.01) were independent risk factors. The combined index=ΔLL+ΔLLA×1.678/0.140 was defined according to their partial regression coefficients to comprehensively evaluate the predictive effect of ΔLL and ΔLLA on mechanical complications. The area under the ROC curves established using ΔLL, ΔLLA and combined index were 0.777, 0.681 and 0.809. Moreover, the optimal cut-off values of ΔLL, ΔLLA and combined index were 9.17°, 0.71 and 12.93, respectively. Conclusions: Lumbar lordosis should be matched for both PI and TK during ASD corrective surgery. The differences between postoperative LL, LLA and the theoretical values calculated by the novel regression model can be used to effectively predict the occurrence of mechanical complications.