Department of Psychiatry and Behavioral Sciences, Medical University of South Carolina, Charleston, South Carolina, USA.
Russell H. Morgan Department of Radiology and Radiological Science, The Johns Hopkins University School of Medicine, Baltimore, Maryland, USA.
NMR Biomed. 2023 Jul;36(7):e4907. doi: 10.1002/nbm.4907. Epub 2023 Feb 5.
The present study characterized associations among brain metabolite levels, applying bivariate and multivariate (i.e., factor analysis) statistical methods to total creatine (tCr)-referenced estimates of the major Point RESolved Spectroscopy (PRESS) proton MR spectroscopy ( H-MRS) metabolites (i.e., total NAA/tCr, total choline/tCr, myo-inositol/tCr, glutamate + glutamine/tCr) acquired at 3 T from medial parietal lobe in a large (n = 299), well-characterized international cohort of healthy volunteers. Results supported the hypothesis that H-MRS-measured metabolite estimates are moderately intercorrelated (M = 0.42, SD = 0.11, ps < 0.001), with more than one-half (i.e., 57%) of the total variability in metabolite estimates explained by a single common factor. Older age was significantly associated with lower levels of the identified common metabolite variance (CMV) factor (β = -0.09, p = 0.048), despite not being associated with levels of any individual metabolite. Holding CMV factor levels constant, females had significantly lower levels of total choline (i.e., unique metabolite variance; β = -0.19, p < 0.001), mirroring significant bivariate correlations between sex and total choline reported previously. Supplementary analysis of water-referenced metabolite estimates (i.e., including tCr/water) demonstrated lower, although still substantial, intercorrelations among metabolites, with 37% of total metabolite variance explained by a single common factor. If replicated, these results would suggest that applied H-MRS researchers shift their analytical framework from examining bivariate associations between individual metabolites and specialty-dependent (e.g., clinical, research) variables of interest (e.g., using t-tests) to examining multivariable (i.e., covariate) associations between multiple metabolites and specialty-dependent variables of interest (e.g., using multiple regression).
本研究通过双变量和多变量(即因子分析)统计方法,对主要基于点分辨波谱(PRESS)质子磁共振波谱( H-MRS)的代谢物(即总 NAA/tCr、总胆碱/tCr、肌醇/tCr、谷氨酸+谷氨酰胺/tCr)进行了特征描述,这些代谢物是在 3T 下从国际大样本(n=299)健康志愿者的内侧顶叶获得的。结果支持了这样的假设,即 H-MRS 测量的代谢物估计值中度相关(M=0.42,SD=0.11,p<0.001),超过一半(即 57%)的代谢物估计值总变异性可以用一个单一的共同因素来解释。尽管与任何单个代谢物的水平无关,但年龄越大,所识别的共同代谢物方差(CMV)因子水平越低(β=-0.09,p=0.048)。在保持 CMV 因子水平不变的情况下,女性的总胆碱水平显著降低(即独特的代谢物方差;β=-0.19,p<0.001),这与之前报道的性别与总胆碱之间存在显著的双变量相关性相吻合。对水参照代谢物估计值(即包括 tCr/水)的补充分析表明,代谢物之间的相关性较低,尽管仍然很高,单个共同因子可以解释总代谢物方差的 37%。如果得到复制,这些结果将表明,应用 H-MRS 研究人员将其分析框架从检查个体代谢物与专业相关(例如临床、研究)变量之间的双变量相关性(例如使用 t 检验)转变为检查多个代谢物与专业相关变量之间的多变量(即协变量)相关性(例如使用多元回归)。