基于脑电特征构建药物过度使用性头痛复发风险预测模型

Development of a prediction model for relapse risk in medication-overuse headache based on electroencephalographic features

摘要:

摘要 目的：基于静息态脑电(electroencephalography, EEG)脑网络特征构建药物过度使用性头痛(medication overuse headache, MOH)复发风险预测模型。 方法：采集MOH患者基线静息态EEG及临床资料，对患者进行长期随访。采用相位锁值(phase locking value, PLV)构建源空间功能连接网络，基于图论分析比较视觉网络和突显网络不同频段模块内总强度差异，并构建Logistic回归模型。 结果：纳入MOH患者211例，经预处理后195例纳入分析，172例完成治疗后首月随访，其中治疗有效者108例，共有52人随访时间满足一年及以上，其中9人失访，复发10人，未复发33人。另外有9人，在不到1年随访时即复发，因此本研究纳入复发组19人，未复发组33人。复发组基线病程、头痛频率及止痛药使用频率均显著高于未复发组。脑电分析显示，复发组视觉网络模块内总强度在delta频段降低而在theta频段升高，突显网络各频段差异无统计学意义。基于theta频段视觉网络模块内总强度构建的模型预测复发的曲线下面积为0.997(95% CI: 0.989–1.000),具有较好的校准度和临床净获益。 结论：theta频段视觉网络模块内总强度可作为预测MOH复发的潜在脑电生理指标。

Abstract:

Abstract Objective To develop a prediction model for relapse risk in medication overuse headache (MOH) based on resting-state electroencephalographic (EEG) brain network features. Methods Baseline resting-state EEG data and clinical information were collected from patients with MOH, followed by long-term follow-up. Source-space functional connectivity networks were constructed using the phase-locking value (PLV). Based on graph theory analysis, differences in the sumstrength within the occipital network and salience network across different frequency bands were compared. Logistic regression models were then established. Results A total of 211 patients with MOH were enrolled. 195 patients were included in the final analysis after preprocessing. 172 patients completed the first-month follow-up after treatment and 108 patients showed effective treatment response. 52 patients had a follow-up duration of at least 1 year, including 9 who were lost to follow-up, 10 who relapsed, and 33 who did not relapse. In addition, 9 patients relapsed within less than 1 year of follow-up. Therefore, 19 patients were included in the relapse group and 33 in the non-relapse group. At baseline, headache history, headache frequency and analgesic use frequency were all significantly higher in the relapse group than in the non-relapse group. EEG analysis showed that the sumstrength within the visual network in the relapse group was decreased in the delta band but increased in the theta band, whereas no statistically significant differences were observed in the salience network across frequency bands. The model based on the sumstrength within the visual network in the theta band achieved an area under the curve (AUC) of 0.997(95% CI: 0.989?1) for predicting relapse and demonstrated good calibration and clinical net benefit. Conclusion The sumstrength within the occipital network in the theta band may serve as a potential objective neuroelectrophysiological biomarker for predicting relapse in MOH patients.