内镜腰椎间盘切除术后复发的因素及预测模型△

Factors and a predicting model of recurrent lumbar disc herniation after percutaneous endoscopic lumbar discectomy

摘要:

[目的]探讨经皮内镜腰椎间盘切除术（percutaneous lumbar endoscopic discectomy, PELD）后复发性椎间盘突出（re- current lumbar disc herniation, rLDH） 的影响因素并建立预测模型。[方法] 回顾性分析 2017 年 1 月—2020 年 1 月作者采用 PELD 治疗的腰椎间盘突出症的 286 患者的临床资料。根据术后 3 年内是否出现复发分为复发组和未复发组，采用单因素比较和多因素逻辑回归分析，筛选出复发的相关因素，并建立数学预测模型，采用受试者工作特征（ROC）曲线等分析以评估模型的临床价值。[结果]286 例患者中，44 例确诊为复发，占 15.4%；242 例未复发，占 84.6%。单因素比较表明，复发组的 BMI [(25.8±3.0) vs (24.2±3.3), P=0.004] 和病程 [(17.9±18.3) 个月 vs (10.7±16.8) 个月, P=0.01] 均显著大于未复发组；复发组的术前影像 Modic 改变显著多于未复发组 [无/有, (30/14) vs (206/36), P=0.006]；复发组术前影像测量椎间 ROM 显著大于未复发组 [(9.3±3.4)° vs (7.1±2.8)°, P＜0.001]；复发组经椎间孔入路手术比率显著大于未复发组 [TF/IL, (29/15) vs (114/128), P=0.022]。逻辑回归表明，BMI (OR=1.154, 95%CI 1.031~1.291, P=0.013）、病程（OR=1.023, 95%CI 1.005~1.042, P=0.013）、Modic 改变（OR=3.143, 95%CI 1.369~ 7.070, P=0.007）、椎间 ROM （OR=1.264, 95%CI 1.126~1.419, P＜0.001） 和手术入路 （椎间孔/椎板间）（OR=2.104, 95%CI 1.007~ 4.396, P=0.048）是复发的独立危险因素。按逻辑回归得出预测模型，其预测值 ROC 分析的曲线下面积为（AUC）为 0.787（95% CI 0.721~0.853）；模型校准曲线与实际曲线一致性较好；决策曲线分析表明，风险阈值为 10%~50%时，该模型可产生较大净获益。[结论]本研究表明 BMI、病程、术前 Modic 改变、术前椎间 ROM 和手术入路是 PELD 术后 rLDH 的危险因素。本研究得出预测 rLDH 模型可能帮助临床医生判断术后复发风险。

Abstract:

[Objective] To explore the factors related to recurrent lumbar disc herniation (rLDH) after percutaneous endoscopic lumbar discectomy (PELD) and establish a predicting model. [Methods] A retrospective study was conducted on 286 patients who underwent PELD for lumbar disc herniation in our hospital from January 2017 to January 2020. Based on whether rLDH happened within 3 years after the primary PELD, the patients were fell into the recurrence group and non-recurrence group. Univariate comparison and multiple logistic regression analysis were performed to search the factors related to the recurrence, establish a mathematical predicting model and draw a Nomogram figure. Then, receiver operating characteristic (ROC) curve, and relative analysis were used to evaluate the clinical significance of this model. [Results] Among the 286 patients, 44 patients were diagnosed of rLDH, accounting for 15.4%, while the remaining 242 patients were confirmed as the non-rLDH, accounting for 84.6%. Regarding univariate comparison, the rLDH group was significantly greater than the nonrLDH group in terms of BMI [(25.8±3.0) vs (24.2±3.3), P=0.004] and course of disease [(17.9±18.3) months vs (10.7±16.8) months, P= 0.009]. In addition, the rLDH group had significantly more Modic change in preoperative images than the non-rLDH group[no/yes, (30/14) vs (206/36), P=0.006], the former got significantly greater range of motion (ROM) in the affected segment measured on preoperative radiographs than the latter [(9.3±3.4)° vs (7.1±2.8)°, P＜0.001]. Moreover, the rLDH group had significantly higher ratio the transforaminal approach than the non-rLDH group [transforaminal/interlaminar, (29/15) vs (118/124), P=0.036]. As results of logistic regression, the BMI (OR=1.154, 95% CI 1.031~1.291, P=0.013), disease course (OR=1.023, 95%CI 1.005~1.042, P=0.013), Modic changes (OR=3.143, 95%CI 1.369~7.070, P= 0.007), intervertebral ROM (OR=1.264, 95% CI 1.126~1.419, P＜0.001) and surgical approach ratio (foraminal/interlaminar) (OR=2.104, 95%CI 1.007~4.396, P=0.048) were independent risk factors for recurrence. The predicting model obtained by logistic regression got area under the curve (AUC) of 0.787 (95%CI 0.721~0.853) by ROC analysis, is in good agreement with the actual curve, and a large net benefit with risk threshold between 10%~50% by decision curve analysis (DCA). [Conclusion] In this study, the BMI, disease course, preoperative Modic changes, preoperative intervertebral ROM and surgical approach were risk factors for rLDH after PELD, while this predicting mode might be helpful for clinicians to determine the risk of recurrence after surgery.