Single-dimension CT measurements with RECIST 1.1 to evaluate liver metastases in GIST patients on imatinib (Abstract)

10091 Background: Tumor response assessment to therapy is crucial in oncology, both in daily practice and research.We analyzed the morphology of liver metastases (LM) in gastrointestinal stromal tumors (GIST). Aims were to 1) determine whether uni-dimensional measurements of tumor lesions (according to RECIST 1.1) are accurate reflections of their volumes and 2) estimate eventual bias in volume change calculations, as introduced by using only the longest diameter. Methods: 78 GIST patients with LM were evaluated at baseline (n=139 lesions), 3 (n=129), 6 (n=128), and 12 months (n=108) after imatinib therapy. LM were segmented semi-automatically on standard CT scans. All measurements at baseline were obtained by 2 independent investigators and agreement between observers was assessed with Bland-Altman plots. Values for segmented volume (Vs; mL), maximum transaxial diameter (MTD; mm), and elongation value (EV, defined as 1 - [width/length], where EV of 1 = fully ellipsoidal and 0 = fully spherical) were reported for each lesion at several time-points. A calculated volume (Vc) of each LM was also generated using the MTD, with the hypothesis that metastases were fully spherical. Results: At baseline, 44% (61/139) of the LM was spherical and 56% (78/139) was ellipsoidal. Their EV ranged from 0.08 to 0.94 (mean, 0.53±0.17). During treatment, overall 42% of the lesions kept their morphology (spherical or ellipsoidal) and the other half changed from spherical to ellipsoidal or vice versa. At baseline, there was strong inter-observer agreement in the determination of Vs, with mean inter-observer variability of 1.5% (95% CI, -14.5% to 17.5%). The difference between Vs and Vc was highly significant (P<0.0001). Vc overestimates the actual volume of 35% (95% CI, -26% to 95%). Conclusions: LM are not always spherical and can change their morphology during imatinib therapy. Therefore, a lesion's single largest trans-axial dimension, as used in RECIST 1.1, is not an accurate surrogate of its actual volume. Further studies are warranted to fully understand the clinical impact of these findings and to assess whether we should apply different criteria for response assessment to therapies in solid tumors.