CT-based whole-body tumor volumetry versus RECIST 1.1: Feasibility and implications for inter-reader variability

To investigate whether volumetric measurements of the whole-body tumor volume (WBTV) are feasible and whether they improve inter-reader variability in patients in whom conventional RECIST 1.1 assessment yielded discordant results.

50 patients (29 male, 21 female, mean age 60.9 ± 12.3 years) with metastases of solid tumors in whom three readers had selected different sets of target lesions and subsequently reached different results for response assessment (progressive vs. non-progressive disease) when using RECIST 1.1 were included. In a second read, all readers performed volumetric measurements of the WBTV on neck/chest/abdomen/pelvis CTs and measured the time needed for these measurements. Cohen's kappa and Fleiss kappa statistics were used to compare the intra- and inter-reader agreement for response assessment.

In 8/50 patients (16 %), the WBTV was too extensive for volumetric measurements and these patients were therefore excluded. In the remaining 42 patients, WBTV measurements required a mean time of 18 min and 9 s. Readers assigned the same response categorizations based on unidimensional RECIST measurements and WBTV measurements in 15/42 patients (33 %), 24/42 patients (57 %) and 30/42 patients (71 %) for reader 1,2 and 3 respectively. When performing response assessment based on WBTV measurements, the three readers agreed in 40/42 patients (95 %) regarding the distinction progressive vs. non-progressive disease, resulting in a near-perfect agreement on a patient-based level (Fleiss' κ = 0.921, 0.95-CI:0.746-1.095).

WBTV measurements yielded an almost perfect inter-reader agreement in a cohort of patients, in which three readers reached discordant response assessment results when following conventional RECIST 1.1 guidelines. This supports the hypothesis, that a limited subset of metastases may not be sufficient to accurately assess response-to-treatment.