扩散加权成像预测原发性中枢神经系统淋巴瘤基于大剂量甲氨蝶呤的化疗疗效

Predictive Evaluation of Pretherapeutic DWI in Primary Central Nervous System Lymphoma for High-Dose MTX-Based Chemotherapy

目的探讨原发性中枢神经系统淋巴瘤(PCNSL)治疗前基于扩散加权成像(DWI)计算得出的表观扩散系数(ADC)值对于实施大剂量甲氨蝶呤(MTX)化疗疗效的预测价值。方法搜集28例经病理证实并且实施基于大剂量MTX化疗的PCNSL患者,所有患者化疗前均行DWI检查并记录强化全瘤灶的最小ADC值、最大ADC值及平均ADC值。化疗4个循环后进行初始疗效评价,分为完全缓解组和部分缓解组。应用独立样本t检验分析化疗前两组之间各ADC参数值、Ki-67标记指数、年龄、体能状态评分的差异。利用受试者工作特性曲线(ROC)分析ADC值对PCNSL化疗初始疗效的预测价值,获得最佳诊断阈值、敏感度、特异度、准确率、阳性预测值、阴性预测值等指标。结果完全缓解组、部分缓解组分别为18例、10例。两组之间的最小ADC值、平均ADC值、体能状态评分具有统计学差异,最大ADC值、年龄、Ki-67指数无统计学差异。ROC分析显示肿瘤最小ADC值曲线下面积最大(0.975),最小ADC值大于最佳阈值474×10^(-6)mm^2/s时,预测完全缓解的敏感度为94.44%,特异度为90%,准确率为92.86%,阳性预测值为94.44%,阴性预测值为90%。结论 ADC值可作为预测PCNSL化疗疗效的参量值,其中最小ADC值的预测效能最好。

Objective To investigate the predictive value of pretherapeutic apparent diffusion coefficient （apparent diffusion coefficient, ADC） value derived from diffusion-weighted imaging （DWI） in primary central nervous system lymphoma （PCNSL） based on high-dose methotrexate （HDMTX） chemotherapy. Methods Retrospective observations of pretherapeutic DWI were studied in 28 patients with PCNSL and received HDMTX-based combination chemotherapy, which were diagnosed by pathologic analysis. The mean （ADCmean ）, maximum （ADCmax）, and minimum ADC （ADCmin） values of the enhancing tumor volume were measured. According to initial response assessed after 4 cycles of chemotherapy, the patients were classified to complete remission （CR） group and partial response （PR） group. Differences of ADC values, Ki-67 label index, age and Karnofsky Performance Status （KPS） were statistically inspected by means of independent sample t-test between CR and PR groups. Then, predictive value of ADC values were evaluated by receiver operating characteristic（ROC） curve, in order to obtain threshold, sensitivity, specificity, accuracy, positive predictive value, and negative predictive value. Results 28 cases included 18 cases of CR and 10 cases of PR. By independent sample t-test method, a statistical difference on KPS, ADCmean, and ADCmin were demonstrated between CR and PR. However, ROC curve analysis revealed the use of ADCmin （0.975） outperformed the use of ADCmean （0. 822） and ADCm= （0. 703 ） notably, then the best prediction threshold of minimum ADC value was 474 × 10^-6mm2/s. Based on this threshold, sensitivity was 94.44% , specificity of 90%, accuracy of 92.86%, positive predictive value of 94.44% and negative predictive value was 90 % in predicting the efficacy. Conclusion ADC measurements derived from DWI may be good predictive factors for response to HD-MTX on PCNSL patients, especially ADCmin.