Department of Neurology, Tangdu Hospital, The Fourth Military Medical University, Xi'an, China.
Department of Thoracic Surgery, Tangdu Hospital, The Fourth Military Medical University, Xi'an, China.
Ann Clin Transl Neurol. 2023 Apr;10(4):644-655. doi: 10.1002/acn3.51752. Epub 2023 Mar 6.
This study aimed to develop and validate internally a clinical predictive model, for predicting myasthenic crisis within 30 days after thymectomy in patients with myasthenia gravis.
Eligible patients were enrolled between January 2015 and May 2019. The primary outcome measure was postoperative myasthenic crisis (POMC). A predictive model was constructed using logistic regression and presented in a nomogram. The area under the receiver operating characteristic curve (AUC) was calculated to examine the performance. The study population was divided into high- and low-risk groups according to Youden index. Calibration curves with 1000 replications bootstrap resampling were plotted to visualize the calibration of the nomogram. Decision curve analyses (DCA) with 1000 replications bootstrap resampling were performed to evaluate the clinical usefulness of the model.
A total of 445 patients were enrolled. Five variables were screened including thymus imaging, onset age, MGFA classification, preoperative treatment regimen, and surgical approach. The model exhibited moderate discriminative ability with AUC value 0.771. The threshold probability was 0.113, which was used to differentiate between high- and low-risk groups. The sensitivity and specificity were 72.1% and 77.1%, respectively. The high-risk group had an 8.70-fold higher risk of POMC. The calibration plot showed that when the probability was between 0 and 0.5, the deviation calibration curve of the model was consistent with the ideal curve.
This nomogram could assist in identifying patients at higher risk of POMC and determining the optimal surgical time for these patients.
本研究旨在开发和验证一种内部临床预测模型，用于预测重症肌无力患者胸腺切除术后 30 天内发生肌无力危象的风险。
本研究纳入了 2015 年 1 月至 2019 年 5 月间的符合条件的患者。主要结局指标为术后肌无力危象（POMC）。使用逻辑回归构建预测模型，并以列线图形式呈现。计算受试者工作特征曲线下面积（AUC）以评估模型性能。根据约登指数将研究人群分为高风险组和低风险组。绘制 1000 次 bootstrap 重采样校准曲线以直观评估列线图的校准程度。通过 1000 次 bootstrap 重采样进行决策曲线分析（DCA）以评估模型的临床实用性。
共纳入 445 例患者。筛选出 5 个变量，包括胸腺影像学、发病年龄、MGFA 分类、术前治疗方案和手术方式。模型具有中等的区分能力，AUC 值为 0.771。阈值概率为 0.113，用于区分高风险组和低风险组。高风险组发生 POMC 的风险是低风险组的 8.70 倍。校准曲线显示，当概率在 0 到 0.5 之间时，模型的偏差校准曲线与理想曲线一致。
该列线图有助于识别发生 POMC 风险较高的患者，并确定这些患者的最佳手术时机。
