Author of

• 10946

  +

  P1.19 - Poster Session 1 - Imaging (ID 179)

  • Event: WCLC 2013
  • Type: Poster Session
  • Track: Imaging, Staging & Screening
  • Presentations: 1
  • Moderators:
  • Coordinates: 10/28/2013, 09:30 - 16:30, Exhibit Hall, Ground Level

  • 10950

    +

    P1.19-004 - Evaluation of a cloud-based solution for local imaging evaluations in clinical trials for lung cancer (ID 1362)

    09:30 - 16:30  |  Author(s): N. Kobayashi
    • Abstract

    Background
    Background: Clinical trials for determination of anti-cancer effects of chemotherapeutic agents are key for developing new strategies for cancer treatment. In most trials, primary endpoints are provided by imaging evaluations. Regulatory authorities have been recommending an Independent Central Review (ICR) with several readers to mitigate potential biases resulting from variance between investigator sites in clinical trials. Based on recent publications promoting site -based evaluation as imaging endpoint, they currently investigate an alternative to ICR. The goal of this study is to evaluate a cloud-based paradigm implementing software solutions and services that standardize the imaging evaluations among international investigator sites.

    Methods
    Methods: Ten lung cancer patients, who received epidermal growth factor receptor-tyrosine kinase inhibitor and for which chest CT scans were available at three time points from baseline to progression, were retrospectively selected. CT scans were evaluated according to the RECIST 1.1 criteria by two oncologists (Saga University) and one radiologist (Nice University Hospital) independently, through a cloud-based software solution named LMS (Lesion Management Solutions, MEDIAN Technologies), which offers reviewing tools and lesion segmentation algorithms. Such system was hosted on the data center (Canon IT Solutions, Japan) and used by readers and data managers (Canon and MEDIAN Technologies) for de-identification, quality control and centralization of the images and their evaluations. The study compared response evaluations between readers and analyzed the reasons for discordances.

    Results
    Results: Readers with different medical training and education, working at distant locations were able to reliably perform radiological evaluations using the same cloud-based system. Between the oncologists and the radiologist, a discordance rate of 35 % (14/40 evaluations) was observed when considering RECIST overall response (CR, PR, SD, PD) at all time points. Precisely, 6 and 8 evaluations were discordant at time point 1 and 2, respectively. . Half of discordances (7/14) were explained by a difference in the selection of target lesions. There were 3 /14 discordances due to a difference in lesion segmentation. The different segmentations occurred when the lesions were adjacent to mediastinum or pleura, where limits of lesions are not very contrasted.

    Conclusion
    Conclusion: The study shows the feasibility of imaging evaluation based on cloud services for clinical studies involving multiple international sites. Centralization of data made possible the on-going monitoring of evaluations through specialized services reducing variability among sites. Analysis of discordances between readers identified areas of improvement for cloud-based services such as a consensus process for target selection at baseline and the development of segmentation tools to standardize management of measurable lesions.