Anti-depressant response in neurobiologically defined psychiatric veteran groups

面对战斗的士兵面临着潜在的重大影响的高风险， 战斗压力对抗压力会导致一些有影响力的情绪和认知 创伤后应激障碍（PTSD），严重抑郁症 (MDD)广泛性焦虑症（GAD）和轻度创伤性脑损伤（mTBI）。而 已经尝试将特定的神经生物学模式与特定的DSM相匹配， 识别障碍，追求已满足有限的成功。临床上，不同的DSM诊断是 通常采用相似的治疗方法，反应率相似（约33-50%）。临床 群体被确定为我的行为集（DSM障碍），研究的目的是找到 神经生物学基础的这些行为定义的群体。我们的目标是找出 在潜在神经生物学背景下的神经生物学组， 提高对医学试验的反应率的关键在于相关特征的聚类。但由于 神经生物学变量之间的复杂关系是简单的线性关系， 风险评分不合适。在这里，我们提出了一种新的方法，我们定义 神经生物学上不同的亚组-基于最可行的，最强大的，最有可能的， 与治疗结果有关-在这些退伍军人与战斗有关的精神痛苦。 我们选择了一组脑成像、分子生物学和生理学标记， 这些措施不会受到当前临床模型的影响。然后我们将试图确定 从这个基于模型的层次聚类方法的强大的子组。接下来，我们对比一下 神经生物学定义的组与传统组或一般反应。最后，为了帮助 为了更好地了解现有数据并为未来的研究提供反馈，我们将在监督下运行一个 机器学习（随机森林），以确定最佳变量和预测组 治疗反应。我们选择单独测试舍曲林，因为这是最常见的 在圣地亚哥弗吉尼亚州精神卫生诊所（FDA）， MDD和PTSD）。基于模型的聚类方法允许我们查看 感兴趣的变量之间的非线性关系，并促进尝试更好地将临床 研究和临床实践，以最好地造福我们的退伍军人群体。

The soldiers that face combat are at high risk for the potentially significant repercussions of combat stress. Combat stress can lead to a number of impactful emotional and cognitive conditions, most notably Posttraumatic Stress Disorder (PTSD), Major Depressive Disorder (MDD), Generalized Anxiety Disorder (GAD), and mild Traumatic Brain Injury (mTBI). While there have been attempts to match a specific neurobiological pattern to a specific DSM identified disorder, the pursuit has met limited success. Clinically, different DSM diagnoses are often approached with similar treatments with similar response rates (~33-50%). Clinically groups are identified my sets behaviors (DSM disorders) and research has aimed to find the neurobiological underpinnings of these behavior defined groups. Our aim is to instead identify neurobiological groups in the context of underlying neurobiological and with the long term goal of improving response rates to medical trials be clustering of relevant features. However, due to the complex relationship between the neurobiological variables a simple linear relationship or risk score is not appropriate. Here we present a novel approach in which we define neurobiologically distinct subgroups — based on the most feasible, most robust, and most likely to relate to treatment outcomes — in these Veterans with combat related psychiatric distress. We have selected a set of brain imaging, molecular biology, and physiological markers such that measures will not be influenced by current clinical models. We will then seek to determine robust subgroups from this model-based hierarchical clustering approach. Next, we contrast our neurobiologically defined groups with traditional groups or general response. Finally, to help best understand the available data and feed forward for future studies, we will run a supervised machine learning (random forest) to determine the optimal variables and groups to predict treatment response. We have opted to solely test sertraline, as this is the most commonly prescribed medication in this population at the San Diego VA mental health clinics (FDA approved for MDD and PTSD). The model-based clustering approach allows us to look at the non-linear relationship between variables of interest and foster an attempt to better link clinical research and clinical practice to best benefit our Veteran population.