Department of Circulation and Medical Imaging, Norwegian University of Science and Technology (NTNU), 7489, Trondheim, Norway.
NMR Biomed. 2012 Feb;25(2):369-78. doi: 10.1002/nbm.1762. Epub 2011 Aug 8.
This study aimed to evaluate whether MR metabolic profiling can be used for prediction of long-term survival and monitoring of treatment response in locally advanced breast cancer patients during neoadjuvant chemotherapy (NAC).
High resolution magic angle spinning (HR MAS) MR spectra of pre- and post-treatment biopsies from 33 patients were acquired. Tissue concentrations of choline-containing metabolites (tCho), glycine and taurine were assessed using electronic reference to access in vivo concentration (ERETIC) of the signal and receiver operating characteristic (ROC) curves was used to define their potential to predict patient survival and treatment response. The metabolite profiles obtained by HR MAS spectroscopy were related to long-term survival and treatment response by genetic algorithm partial least squares discriminant analysis (GA PLS-DA).
Different pre-treatment MR metabolic profiles characterized by higher levels of tCho and lower levels of lactate were observed in patients with long-term survival (≥5 years, survivors) compared to patients who died of cancer recurrence (<5 years, non-survivors). A significant decrease in glycerophosphocholine (GPC) post-treatment was associated with long-term survival (p = 0.046) and partial response (p = 0.014) to NAC. Long-term survival was best predicted by GPC using ROC analyses (sens. 66.7%, spec. 62.5%), while taurine had the best predictive value of treatment response (sens. 72.7%, spec. 63.2%). GA PLS-DA multivariate classification models successfully discriminated between survivors and non-survivors, resulting in 82.7% and 90.2% cross-validation (CV) classification accuracy, pre- and post-treatment, respectively. Classification of treatment response using GA PLS-DA was not successful for this patient cohort.
Our results demonstrate that HR MAS MR metabolic profiles consisting of important metabolic characteristics of breast cancer tumors could potentially assist the classification and prediction of long-term survival in locally advanced breast cancer patients, in addition to being used as an adjunct for evaluation of treatment response to NAC.
本研究旨在评估磁共振代谢组学是否可用于预测局部晚期乳腺癌患者新辅助化疗（NAC）期间的长期生存和治疗反应监测。
对 33 例患者的治疗前后活检进行高分辨率魔角旋转（HRMAS）MR 光谱采集。使用电子参考以获取信号的体内浓度（ERETIC）评估胆碱代谢物（tCho）、甘氨酸和牛磺酸的组织浓度，并使用受试者工作特征（ROC）曲线来定义它们预测患者生存和治疗反应的潜力。通过 HRMAS 光谱获得的代谢谱通过遗传算法偏最小二乘判别分析（GA PLS-DA）与长期生存和治疗反应相关联。
与癌症复发死亡的患者（<5 年，非幸存者）相比，长期生存（≥5 年，幸存者）患者的不同预处理 MR 代谢谱特征为 tCho 水平较高，乳酸水平较低。NAC 治疗后甘油磷酸胆碱（GPC）水平显著降低与长期生存（p=0.046）和部分缓解（p=0.014）相关。ROC 分析表明，GPC 对长期生存的预测最佳（敏感性 66.7%，特异性 62.5%），而牛磺酸对治疗反应的预测价值最佳（敏感性 72.7%，特异性 63.2%）。GA PLS-DA 多变量分类模型成功区分了幸存者和非幸存者，分别在预处理和治疗后获得了 82.7%和 90.2%的交叉验证（CV）分类准确性。对于该患者队列，使用 GA PLS-DA 对治疗反应进行分类并不成功。
我们的结果表明，HRMAS MR 代谢谱由乳腺癌肿瘤的重要代谢特征组成，可能有助于局部晚期乳腺癌患者的分类和长期生存预测，并且可以作为评估 NAC 治疗反应的辅助手段。
