A Novel Method Of Characterizing Response Heterogeneity Using Fdg-Pet

Existing guidelines for response evaluation using FDG-PET are based on changes in maximum or peak activity and do not address the possibility of heterogeneous responses in patients with multiple malignant lesions. We have developed a rapid algorithm for visualizing and quantifying PET response heterogeneity. Here we present our initial results using this approach for a cohort of stage III non-small cell lung cancer (NSCLC) patients treated with definitive chemoradiation therapy. Using an institutional database, we identified patients who were treated with definitive chemoradiation therapy for stage III NSCLC between 2007 and 2010 and underwent pre- and post-treatment PET. Post-treatment scans were performed a median of 72 days after radiation therapy completion (range 34 to 113). Patients with progressive disease on post-treatment PET were excluded. For each patient, pre- and post-treatment PET scans were coregistered using a commercially available software package. All visible hypermetabolic lesions were delineated on each scan using a semiautomatic gradient-based contouring tool. Maximum standardized uptake value (SUV), metabolic tumor volume (MTV), and total glycolytic activity (TGA, equal to MTV x mean SUV) for each lesion and scan were tabulated. For each patient, an intrapatient waterfall plot was generated to depict the range of responses observed for each metric across all lesions. Patients with post-treatment maximum SUV < 3.0 were deemed complete responders. A novel heterogeneity index was defined for each remaining patient as the standard deviation of the percent change in TGA observed in individual lesions. Based on visual assessment of each intrapatient waterfall plot, a heterogeneity index cutoff to separate heterogeneous partial responders from homogeneous partial responders was selected. Thirteen patients met all inclusion criteria, and within those patients there were 13 primary tumors and 41 hypermetabolic lymph nodes. 5 of the included patients had complete responses. For the remaining eight patients, heterogeneity index values ranged from 2% to 65%. A cutoff of 20% effectively separated three heterogeneous partial responders from five homogenous partial responders. We have developed a rapid workflow to visualize and quantify intrapatient PET response heterogeneity. Larger studies to explore the implications of response heterogeneity in NSCLC and other diseases are planned.