GIST Trial With RECIST 1.1 and Choi Criteria

Challenge

In gastrointestinal stromal tumors treated with targeted therapy, responding lesions often change density and become cystic or hemorrhagic without shrinking, and may even enlarge. Size-based RECIST 1.1 assessment alone can therefore misclassify responders as stable or progressive disease. The sponsor of this multi-year trial of a novel GIST therapeutic required both RECIST 1.1 and Choi criteria applied in parallel, which demands reproducible Hounsfield-unit density measurements that are highly sensitive to contrast phase and acquisition technique.

GIST lesions are frequently large, heterogeneous and located in the stomach, small bowel, mesentery and peritoneum, where bowel motion and variable distension complicate measurement. Consistent contrast timing across 38 sites, and consistent placement of density regions of interest across readers and timepoints, were prerequisites for a credible Choi analysis.

Intrinsic Imaging’s Solution

Intrinsic Imaging developed an imaging charter specifying portal venous phase contrast-enhanced CT with defined bolus timing, slice thickness and reconstruction kernels so that density measurements would be comparable across sites and timepoints. Site qualification included review of test submissions for contrast phase, and sites received targeted feedback where phase timing fell outside the charter window.

Oncology radiologists experienced in sarcoma trials performed blinded double reads applying RECIST 1.1 and Choi criteria at every timepoint. The charter defined how density regions of interest would be placed and propagated across timepoints, how cystic and hemorrhagic change would be recorded, and how intra-tumoral nodules suggesting progression within an otherwise responding lesion would be captured. Adjudication was performed on discordant response categories under either criterion.

Image quality control and processing were centralized, with density measurements captured in the reading platform rather than transcribed, and the dual-criteria dataset was delivered in a structure that allowed the sponsor to analyze RECIST 1.1 and Choi outcomes separately and in comparison.

Intrinsic Imaging’s Results

• Dual RECIST 1.1 and Choi assessment across approximately 1,900 timepoints from 38 sites with blinded double reads and adjudication
• Contrast-phase compliance above 95% after site qualification and feedback, supporting reproducible Hounsfield-unit density measurement
• Roughly one in five patients classified as responders under Choi criteria while stable under RECIST 1.1, captured in a dataset that let the sponsor characterize the difference
• Density measurements captured directly in the reading environment with a complete audit trail of region-of-interest placement and propagation