Tumor Lesion Glycolysis and Tumor Lesion Proliferation for Response Prediction and Prognostic Differentiation in Patients With Advanced Non–Small Cell Lung Cancer Treated With Erlotinib

Abstract

The aim was to assess the value of tumor lesion glycolysis (TLG) and tumor lesion proliferation (TLP) determined by FDG and 3'-deoxy-3'-F-fluorothymidine (FLT) PET for response prediction and prognostic differentiation in patients with advanced non-small cell lung cancer (NSCLC) treated with erlotinib. FDG-PET and FLT-PET were performed in 30 patients with untreated Stage IV NSCLC before start of therapy, 1 (early) and 6 (late) weeks after erlotinib treatment. Functional tumor volume parameters including TLG in FDG-PET and TLP in FLT-PET were measured in the sum of up to 5 lesions per scan. Metabolic response was assessed using different cutoff values for percentage changes of TLG and TLP. Absolute baseline and residual levels of TLG and TLP were used for dichotomizing the patients into 2 groups. Kaplan-Meier analysis and the log-rank test were performed to analyze the association with progression-free survival (PFS). Patients with a metabolic response measured by early changes of TLP and late changes of TLG and TLP showed a significantly better PFS than metabolically nonresponding patients. A lower cutoff value of 20% or 30% for definition of metabolic response showed better differentiation between metabolically responding and nonresponding patients in cases where the 45% cutoff value revealed no significant results. Furthermore, patients with lower absolute early and late residual TLG and TLP levels had a significantly prolonged PFS. In contrast, absolute baseline TLG and TLP levels showed no significant association with PFS. In patients with advanced NSCLC, percentage changes of TLG and TLP and absolute residual TLG and TLP levels under erlotinib treatment emerged as strong predictive factors for PFS. Our findings indicate that a cutoff value of 20% or 30% for definition of metabolic response measured by percentage changes of TLG and TLP provides suitable results for response prediction, which should be further validated.