Computer-aided volumetry analysis in assessing pulmonary chemotherapy response in advanced NSCLC comparing with RECIST criteria.

Abstract

10593 Background: Assessing of therapeutic tumor response in clinical trial by Response Evaluation Criteria in Solid Tumors (RECIST) was suggested to carry high variability. Computer-aided volumetry analysis for pulmonary nodules has demonstrated better sensitivity and reproducibility than RECIST criteria, however mostly on phantom nodules. There are no studies testing volumetry analysis on disease response in a clinical trial. Methods: We identified 28 patients with valid pulmonary lesions from a phase II clinical trial using new chemotherapy regimen in patients with stage IIIB/IV Non-small cell lung cancer. Up to five largest pulmonary lesions on pre-treatment CT and CT after 2 cycles of chemotherapy were analyzed using computer-aided volumetry software (LMS-lung/track, Median Technologies, Valbonne, France). Disease response on pulmonary lesions from volumetry analysis and from RECIST were compared. Progression free survival was correlated to the disease response analyzed by volumetry software and RECIST. Results: Total of 91 pulmonary lesions were analyzed on 56 CT scans. The patients with disease progression, stable disease or partial response determined by volumetry were 6 (21.43%), 13 (46.43%) and 9 (32.14%) respectively. The pulmonary progression, stable disease and partial response by RECIST were 1 (3.57%), 23 (82.14%) and 4 (14.29%). The difference was statistically significant by Bhapkar’s test of marginal homogeneity (p-value 0.0046). Among stable disease categorized by RECIST, 5 cases were determined to be progressive disease by volumetry, 7 cases were diverged into partial response category. From Wald Test of Cox regression model, the volumetry analysis is better at predicting progression-free survival (p-value 0.3650 for RECIST, p-value 0.0688 for volumetry). Conclusions: Volumetry analysis may identify disease progression or response earlier than RECIST criteria. It also predicts PFS better compare to RECIST.