Evaluation of a cloud-based local-read paradigm for imaging evaluations in oncology clinical trials for lung cancer

BACKGROUND: Although tumor response evaluated with radiological imaging is frequently used as a primary endpoint in clinical trials, it is difficult to obtain precise results because of inter- and intra-observer differences. PURPOSE: To evaluate usefulness of a cloud-based local-read paradigm implementing software solutions that standardize imaging evaluations among international investigator sites for clinical trials of lung cancer. MATERIAL AND METHODS: Two studies were performed: KUMO I and KUMO I Extension. KUMO I was a pilot study aiming at demonstrating the feasibility of cloud implementation and identifying issues regarding variability of evaluations among sites. Chest CT scans at three time-points from baseline to progression, from 10 patients with lung cancer who were treated with EGFR tyrosine kinase inhibitors, were evaluated independently by two oncologists (Japan) and one radiologist (France), through a cloud-based software solution. The KUMO I Extension was performed based on the results of KUMO I. RESULTS: KUMO I showed discordance rates of 40% for target lesion selection, 70% for overall response at the first time-point, and 60% for overall response at the second time-point. Since the main reason for the discordance was differences in the selection of target lesions, KUMO I Extension added a cloud-based quality control service to achieve a consensus on the selection of target lesions, resulting in an improved rate of agreement of response evaluations. CONCLUSION: The study shows the feasibility of imaging evaluations at investigator sites, based on cloud services for clinical studies involving multiple international sites. This system offers a step forward in standardizing evaluations of images among widely dispersed sites.