Plasma metabolomics profile (PMP) in COPD has been associated with clinical characteristics, but PMP’s relationship to survival has not been reported. We determined PMP differences between patients with COPD who died an average of 2 years after enrollment (Non-survivors, NS) compared to those who survived (S) and also with age matched controls (C). We studied prospectively 90 patients with severe COPD and 30 controls. NS were divided in discovery and validation cohorts (30 patients each) and the results compared to the PMP of 30 S and C. All participants completed lung function tests, dyspnea scores, quality of life, exercise capacity, BODE index, and plasma metabolomics by liquid and gas chromatography / mass spectometry (LC/MS, LC/MS2, GC/MS). Statistically, we used Random Forest Analysis (RFA) and Support Vector Machine (SVM) to determine metabolites that differentiated the 3 groups and compared the ability of metabolites vs. clinical characteristics to classify patients into survivors and non-survivors. There were 79 metabolites statistically different between S and NS [p < 0.05 and false discovery rate (q value) < 0.1]. RFA and SVM classification of COPD survivors and non-survivors had a predicted accuracy of 74 and 85% respectively. Elevation of tricyclic acid cycle intermediates branched amino acids depletion and increase in lactate, fructose and xylonate showed the most relevant differences between S vs. NS suggesting alteration in mitochondrial oxidative energy generation. PMP had similar predictive power for risk of death as information provided by clinical characteristics. A plasma metabolomic profile characterized by an oxidative energy production difference between survivors and non-survivors was observed in COPD patients 2 years before death.

中文翻译：

---

血浆代谢组学和COPD患者生存差异的临床预测因素

COPD中的血浆代谢组学谱（PMP）与临床特征有关，但尚未报道PMP与生存的关系。我们确定了入组后平均死亡2年的COPD患者（非幸存者，NS）与存活的患者（S）以及年龄匹配的对照组（C）之间的PMP差异。我们对90名重度COPD患者和30名对照进行了前瞻性研究。将NS分为发现和验证队列（每组30例），并将结果与​​30 S和C的PMP进行比较。所有参与者均通过以下方式完成了肺功能测试，呼吸困难评分，生活质量，运动能力，BODE指数和血浆代谢组学液相色谱和气相色谱/质谱法（LC / MS，LC / MS2，GC / MS）。从统计上讲 我们使用随机森林分析（RFA）和支持向量机（SVM）来确定区分3组的代谢物，并比较代谢物的能力与临床特征，以将患者分为幸存者和非幸存者。S和NS之间有79种代谢物具有统计学差异[p <0.05和错误发现率（q值）<0.1]。对COPD幸存者和非幸存者的RFA和SVM分类分别具有74％和85％的预测准确性。三环酸循环中间体的升高使分支氨基酸的消耗减少，乳酸，果糖和木糖酸盐的增加显示出S与NS之间最相关的差异，表明线粒体氧化能的产生发生了变化。PMP对死亡风险的预测能力与临床特征所提供的信息相似。