Radiobiological Modeling of Tumor Control Probability Using FDG PET Imaging Results Combined with Tumor Volume
M Guerrero1*, S Tan2, W Lu3, (1) University of Maryland School of Medicine, Baltimore, MD, (2) Huazhong University of Science & Technology , Wuhan, ,(3) University of Maryland School of Medicine, Baltimore, MDTH-A-WAB-8 Thursday 8:00AM - 9:55AM Room: Wabash Ballroom
Purpose: The purpose of this work is to investigate radiobiological modeling based on SUVmax values from FDG-PET imaging for lung cancer patients with PET imaging before and/or after chemoradiation, combined with initial tumor volume, to determine their status as surgical candidates.
Methods:Twenty-six lung cancer patients treated with chemoradiation and surgery had available FDG-PET SUVmax values before treatment (SUVbe) and twenty-one patients had SUVmax values available after chemoradiation (SUVaf). The initial tumor volumes (Vo) and pathological response status were known for all patients. Mean SUVbe and SUVaf as well as mean values of Vo*SUVbe and Vo*SUVaf were computed for complete responders (CR) and non-complete-responders (NCR). P-values were calculated to establish statistical significance and a sigmoid function with two parameters was fitted using the maximum likelihood method to calculate the tumor recurrence probability (TRP) as a function of SUVmax or Vo*SUVmax.
Results:Mean SUVaf was 4.2+/- 0.8 for CR patients and 4.24 +/-1.3 for NCR (p-value 0.5). Mean Vo*SUVaf (normalized to overall average) was 0.76 +/- 0.2 for CR and 1.69+/- 0.5 for NCR (p-value=0.09). If tumor smaller than 10cc where excluded the mean Vo*SUVaf became 0.81 +/- 0.2 for CR and 2.40 +/-0.9 for NCR (p-value=0.04). The TRP model was fitted for the reduced set of patients and the normalized value of Vo*SUVaf that gives a TRP of 0.5 was 1.78 (90%C.I. 1-4.5) and the normalized slope was 0.41 (90%C.I.0.02-0.8). Similar results were obtained for Vo*SUVbe.
Conclusion:Radiobiological modeling based on before and after SUVmax values is hampered by the existence of false negatives (patients with low SUVmax that are NCR). This problem can be mitigated by using tumor volume information and confining the analysis to initial tumor volumes larger than 10cc. In that case, tumor recurrence probability curves can be constructed as a function of Vo*SUVmax.