Three-dimensional analysis of advanced stage NSCLC: prognostic significance of volumetric measurements

Purpose: To evaluate the reproducibility of the three-dimensional (3-D) analysis of advanced stage non-small cell lung cancer (NSCLC) using a semi- automated computer software and to determine the value of volumetric CT measurements in predicting prognosis. Methods and Materials: CT scans of 43 measurable NSCLCs were included in this study. Four observers independently measured the greatest diameter of each lesion by using the electronic caliper on baseline and follow-up (post first cycle chemotherapy) CT scans. In the same session, a semi-automated three- dimensional lung cancer segmentation software commercially available, was applied to assess both unidimensional and volumetric measurements of the same lesions. The Bland and Altman test evaluated the interobserver variability for each measuring system; in order to further compare reproducibility between diameters and volumetric measurements, the latter were transformed into equivalent diameters by assuming the tumor as having a spherical-like shape. The value of each CT measurement change between baseline and follow-up in predicting progression-free survival (PFS) was analysed by logistic regression and Cox proportional hazards regression model. Results: Volumetric measurements were the most reproducible measurements. The only significant indicators of lower PFS were the volume change for two observers (greatest HR, 1.57; 95% CI: 1.04-2.38), and the maximal diameter obtained through the software for one observer (HR, 3.58; 95% CI: 1.18-10.9). Conclusion: Our preliminary data suggests that the software-based assessment of the lesion size seems more reproducible than the manual one and of more prognostic value.