Prognostic Value of Early Changes in Metabolic Tumor Heterogeneity for Patients with Locally Advanced Non-Small Cell Lung Cancer
X. Dong,X. Sun,L. Sun,J. Shaffer,X. Zhang,Z. Wang,H. Wan,J. Yu,L. Xing
DOI: https://doi.org/10.1016/j.ijrobp.2015.07.387
2015-01-01
Abstract:To observe the early changes of metabolic tumor heterogeneity and to determine its prognostic value for patients with locally advanced non-small cell lung cancer (NSCLC). From January 2007 to February 2010, 43 patients with NSCLC were included in the study. All patients were evaluated by 18F-fluorodeoxyglucose (18F-FDG) PET/CT before and after receiving radiation with concurrent cisplatin-based chemotherapy (CRT). Tumor response was assessed according to the Response Evaluation Criteria in Solid Tumors (RECIST) at 4 weeks after CRT began. Tumor heterogeneity was determined using 7 textural parameters (coefficient of variation [COV], skewness, kurtosis, contrast, energy, entropy, homogeneity) and the area under the cumulative intensity-volume histogram curve (AUC-IVH). For comparison, conventional parameters (maximum and mean standardized uptake values [SUV], metabolic tumor volume, and total lesion glycolysis [TLG]) were also evaluated. Values on pretreatment PET/CT, as well as changes early in the course of therapy, were compared with tumor response. Overall survival (OS) and disease-free survival (DFS) were estimated by Kaplan-Meier method. Compared with non-responders, RECIST responders showed significantly greater COV, contrast and AUC-IVH temporal changes during CRT (P = .001, .006, .034, respectively). None of the SUV or other textural parameters predicted RECIST response. By Kaplan-Meier analysis, the patients with greater contrast and AUC-IVH changes had significantly higher 5-year OS (P = .024, P = .032) and DFS (P = .014, P = .017) than patients with little changes of metabolic heterogeneity. In multivariate analysis, contrast and AUC-IVH changes were statistically significant independent prognostic factors of OS (HR = 2.72, P < .001; HR = 2.56, P = .008, respectively) and DFS (HR = 2.09, P = .023; HR = 1.85, P = .036, respectively). The remaining parameters during CRT did not show significant associations with OS and DFS. Receiver-operating characteristic curve analysis showed that temporal changes of tumor textural analysis serve as a prognostic factor with higher sensitivity (76%–92%) and specificity (83%–94%) than any baseline imaging parameter. Early in the course of CRT for locally advanced NSCLC, the metabolic tumor heterogeneity characterized by COV, contrast and AUC-IVH may predict tumor response before changes in the SUV; furthermore, contrast and AUC-IVH can provide significant prognostic information; therefore, close monitoring of changes in metabolic tumor heterogeneity is an important step for personalized treatment of locally advanced NSCLC.