Background: The I-SPY 2 TRIAL is a multi-site response adaptive clinical trial evaluating novel drug combinations for neoadjuvant treatment of breast cancer. Serial measurement of functional tumor volume (FTV) by MRI during treatment is used to assess response. Under FDA IDE approval, FTV plays an integral role in adjusting patient randomization and evaluating treatment efficacy. Standardized image acquisition methods are used across multiple MRI system platforms, and MRI image quality is reviewed in the Imaging Core Lab (ICL) for protocol adherence. Sites communicate with the ICL through a centralized email account, and standardized forms and procedures are used to submit MRI studies. As a result, the I-SPY 2 TRIAL consistently reports a high level of data quality and data acceptance for FTV measurements. We present an overview of MRI operational performance and share lessons learned about maintaining high quality MRI data in a multi-site clinical trial.

Methods: Over the 10-year course of the I-SPY 2 TRIAL, workflow has been improved to optimize communication between ICL and sites and to accurately record details about the MRI. A standardized imaging acquisition protocol is distributed to all sites, and new sites submit two protocol adherent test cases for review at site initiation. A scan verification form (SVF) is required for each MRI study completed at sites to document critical information about the study. Sites submit studies using TRIAD image transfer and deidentification software (American College of Radiology), and data is archived and processed at the ICL. All MRI studies are reviewed by the ICL for protocol adherence as soon as they are submitted, and feedback is provided to sites. Image quality factors including motion, fat suppression, and signal-to-noise ratio are qualitatively assessed. The ICL communicates with sites through centralized emails, regular Coordinator Calls, and Imaging Working Group meetings to discuss emerging issues and offer ongoing training. The ICL also contributes to revisions of the study protocol and manual of operating procedures.

Results: As of June 2020, 3020 patients had been registered in I-SPY 2, 1741 patients randomized to treatment with one of 18 experimental drugs or standard therapy (controls), and a total of 7527 MRI studies were performed. FTV could be calculated for 97% (7317/7527) of studies. Of the 7317 studies where FTV could be calculated, relatively minor issues with image quality or imaging protocol adherence were documented for 28% (2030/7317) of studies. These issues included motion artifacts (32%, 659/2030), off-protocol scan duration (21%, 433/2030), off-protocol contrast injection rate (14%, 281/2030), and off-protocol imaging field of view (9%, 191/2030).

Conclusion: Operational standardization, clear communication with sites, and streamlined workflow yield high quality MRI data across multiple sites and scanner vendors. As a result, FTV is a robust biomarker of response to treatment, and is being used to predict patient response and guide treatment planning. We are actively investigating strategies that will improve FTV accuracy for predicting response and informing guidelines for treatment de-escalation. Later this year, an imaging phantom will be distributed to a subset of I-SPY 2 sites, allowing for quantitative assessment of image quality and precise scanner calibration. This will allow the ICL to maintain high image quality for all sites and will provide the foundation for testing a variety of imaging biomarkers in the I-SPY 2 TRIAL.

‍