Brain Network Changes Accompanying and Predicting Responses to Pharmacotherapy in OCD

摘要 强迫症和强迫症影响了大约30%的人口；当它们变得严重时，会导致诊断为 强迫症(OCD)，每40个人中就有一个人患有强迫症。现有的治疗方法，包括选择性5-羟色胺再摄取抑制剂(SSRI)抗抑郁剂的药物治疗和专门的心理治疗，对许多人都有好处，但个性化的反应是不同的和不可预测的。迫切需要了解治疗变化的大脑机制，并可能指导新干预措施的发展。最终，预测谁将对特定治疗有反应的能力将是一项重大的理论和临床进步，将加快有效治疗的部署，从而大大降低发病率。早期使用灌注成像的研究已经暗示，基线神经标记物可以预测药物治疗的反应。然而，这些研究没有利用现代的以网络为中心的分析方法，也没有产生机械性的见解或临床实用。 神经精神障碍被认为是由功能改变的大脑网络引起的。静止态泛函 连通性磁共振成像(rs-fcMRI)已成为描述人类功能网络结构的有力工具。 我们建议使用RS-fcMRI，使用人类连接组计划开创的最先进的方法，来映射功能神经网络和强迫症治疗反应之间的关系。具体来说，我们的目标是 表征映射到治疗相关变化和预测反应的RS-fcMRI连通性简档。这个 我们的试点数据支持了该项目的可行性。我们专注于用氟西汀进行一线SSRI药物治疗，作为一个容易处理的第一步；未来的研究将纳入其他治疗方式，包括心理治疗。 我们提出了一种创新的临床设计，将治疗与时间效应分开，这是 治疗机理的研究。80名未服用药物的强迫症患者将以1：1的随机比例接受氟西汀治疗，要么立即开始，要么在为期6周的安慰剂先导期结束后开始接受治疗。强迫症受试者将在基线以及6、12和18周接受成像。所有受试者将汇集在一起，以确定症状改善的相关因素。即时治疗组和延迟治疗组将进行对比，以将治疗引起的神经变化与治疗接触(即安慰剂)的非特异性影响分开。在药物治疗之前，将对40名匹配的对照组进行一次扫描，并与强迫症受试者进行基线比较，以表征未用药状态下的连接性变化。神经成像数据将使用全脑通用线性模型(GLMS)进行分析，包括组间和纵向效应，以分离时间效应、药物暴露本身的影响以及临床改善的相关性。将使用基于GLM的回归和最近优化的个人分类器，对基线成像数据进行检查以预测治疗反应，对75%的样本进行训练，然后在其余25%的样本上进行测试。 这项研究将产生丰富的多模式神经成像数据集，阐明强迫症症状的神经相关性 和治疗反应。如果成功，我们将确定新疗法的网络目标，并迈出重要一步 为精神病学的精确医学服务开发预测性措施的目标。

ABSTRACT Obsessions and compulsions affect ~30% of the population; when they become severe they lead to a diagnosis of obsessive-compulsive disorder (OCD), which affects one person in 40. Available treatments, including pharmacotherapy with the selective serotonin reuptake inhibitor (SSRI) antidepressants and specialized psychotherapy, are of benefit to many, but individualized response is heterogeneous and unpredictable. Understanding the brain mechanisms of therapeutic change is urgently needed and may guide the development of new interventions. Ultimately, the ability to predict who will respond to a particular treatment would be a major theoretical and clinical advance, would accelerate deployment of effective treatment, and would thereby greatly reduce morbidity. Early studies using perfusion imaging have hinted that baseline neural markers can predict response to pharmacotherapy. However, these studies have not harnessed modern network-focused analytic methods and have not yielded mechanistic insight or clinical utility. Neuropsychiatric disorders are hypothesized to derive from altered functional brain networks. Resting-state functional connectivity MRI (rs-fcMRI) has emerged as a powerful tool to characterize functional network architecture in humans. We propose to use rs-fcMRI, employing state-of-the-art methodologies pioneered by the Human Connectome Project, to map the relationship between functional neural networks and treatment response in OCD. Specifically, we aim to characterize rs-fcMRI connectivity profiles that map onto treatment-associated changes and that predict response. The feasibility of this project is supported by our pilot data. We focus on first-line SSRI pharmacotherapy with fluoxetine as a tractable first step; future studies will incorporate other treatment modalities, including psychotherapy. We propose an innovative clinical design that dissociates treatment from time effects, which is a major challenge in studies of treatment mechanism. 80 medication-free OCD subjects will be randomized 1:1 to receive fluoxetine treatment starting either immediately or after a 6-week placebo lead-in phase. OCD subjects will undergo imaging at baseline and at 6, 12 and 18 weeks. All subjects will be pooled to identify correlates of symptom improvement. The immediate and delayed treatment groups will be contrasted to dissociate treatment-induced neural changes from the non-specific effects of therapeutic contact (i.e. placebo). 40 matched controls will be scanned once and compared with OCD subjects at baseline, prior to pharmacotherapy, to characterize connectivity alterations in the unmedicated state. Neuroimaging data will be analyzed using whole-brain general linear models (GLMs), including between-group and longitudinal effects to isolate effects of time, effects of drug exposure per se, and correlates of clinical improvement. Baseline imaging data will be examined for treatment response prediction, using both a GLM-based regression and via a recently optimized individual classifier, trained on 75% of the sample and then tested on the remaining 25%. This study will yield a rich multi-modal neuroimaging dataset elucidating the neural correlates of OCD symptomatology and of treatment response. If successful, we will identify network targets for novel treatments and take a major step towards the goal of developing predictive measures in the service of precision medicine in psychiatry.