Context

Microsatellite instability (MSI) is a genomic biomarker that plays an important role in the treatment of colorectal tumours (CRC) patients. MSI is predictive of response to immunotherapy across all solid tumors.

‍In 2021, Owkin developed MSIntuit CRC (CE-IVD) for the pre-screening of microsatellite stable patients with colorectal cancer with a deep learning model directly on H&E surgical resection WSI.

‍

Objectives

• Improve MSIntuit CRC algorithm to make it compatible with CRC biopsies
• Conduct a clinical validation with multiple external centers cohorts
• Demonstrate improved accuracy of MSIntuit CRC as a rule-out solution for MSI testing

‍

Methods

An updated model architecture and training cohorts have been leveraged to enhance MSIntuit CRC accuracy.

MSIntuit CRC has been evaluated during a multi-centric clinical validation study on 5 external (1500+ biopsies and 500+ surgical resections) of archived CRC specimens.

‍

Results

MSIntuit CRC shows high levels of diagnostic sensitivity and NPV across both biopsies and surgical resections, on par with gold standard techniques.Additionally, the highly reproducible performances across centers provides confidence in the clinical application.

MSIntuit CRC new model version enables the rule-out of up to 70% of MSS patients. The remaining testing pool is 3x enriched in MSI cases compared to current testing approaches.

‍

Impact

MSIntuit CRC could help decrease workload associated with MMR/MSI testing in clinical routine by 50% for biopsies and 70% for surgical resections.

At laboratory level, this would enable centers to streamline their workflow, liberate auto-stainers and time for tasks with more added value and compensate for pathologist and technician shortage.

At payer level, MSIntuit CRC could be an economic alternative to current MMR/MSI testing techniques enabling cost to decreases between 31% and 57% depending on the specimen analysed and the testing technique used.²

‍

‍

Learn more about MSIntuit CRC

‍

‍

1. ‍Results from MSIntuit CRC v2 performance validation published at USCAP. A rule-out of 50% was obtained on biopsy samples, and 70% on surgical resection samples, 96% sensitivity on surgical resections, 95% sensitivity on biopsies, and 99% NPV for both surgical resections and biopsies, with 10% MSI prevalence.
2. All assumptions are based on US market: CRC incidence of 151.000 patients, IHC cost = 395$, PCR cost = 485$
   Arch Pathol Lab Med (2022) 146 (10): 1194–1210.
   Annals of Oncology 30: 1232–1243, 2019