Using quantitative pathologic analysis to identify microsatellite stability in digitized colorectal carcinoma images: An effort to screen patients who should be tested for Lynch syndrome.

Le résumé fourni par la source

93 Background: Colorectal cancer (CRC) is the 4th most common cancer among men and women and roughly 2-4% of all CRCs occur in patients with a diagnosis of Lynch syndrome (LS), an autosomal dominant condition found to be the most common cause of hereditary CRC with estimated lifetime risk of developing CRC between 50-80%. LS results from mutations in one or more of 5 genes: MLH1, MSH2, MSH6, PMS2, and EPCAM, which lead to a deficiency in DNA mismatch repair ( dMMR ). Current guidelines suggest testing every CRC for microsatellite instability ( MSI ) or MMR gene expression using immunohistochemistry (IHC) to identify LS-related CRC. QuantCRC is a prognostic AI model that identifies important histological features of CRC and provides a recurrence prediction estimate useful for treatment decisions. It uses computational image analysis to extract quantitative features from digitized H&E stained pathology slides such as tumor grade, tumor-infiltrating lymphocytes, and other morphometric parameters. 4 It has been verified and validated in Aiforia (AI for image analysis) Platform using whole slide images and against reviews by independent pathologists from 8 hospitals. Our study sought to use QuantCRC to identify histological differences between those with MSI (confirmed LS), sporadic MSI , and microsatellite stability (MSS ) CRC to better help clinicians identify which patients would benefit from additional testing. Methods: We conducted a retrospective cohort study of patients with CRC whose pathology was analyzed at Mayo Clinic Arizona. Tumor characteristics were analyzed by QuantCRC and Receiver Operating Characteristic (ROC) curve analysis using Area Under the Curve (AUC) was used to see if QuantCRC could distinguish CRCs with MSI or LS. Results: Of the total 5132 patients with CRC, 1101 (21.5%) samples were MSI/dMMR with 468 (42.5%) of those having confirmed LS. The study had roughly equal parts men and there were more women with non-LS MSI (59%) than with LS or microsatellite stability. The QuantCRC model’s ROC analysis revealed an AUC of 0.83 for predicting MSI/dMMR among all patients with a sensitivity of 0.75 and a specificity of 0.79. AUC for predicting LS among all patients was 0.76 and of predicting LS among MSI/dMMR was 0.64. Conclusions: The QuantCRC model was useful in predicting MSI/dMMR status among all patients with CRC. It can be a helpful clinical tool when assessing which patients may benefit from additional genetic or IHC testing, especially in lower resource areas where molecular testing may be difficult to access or expensive relative to an H&E slide.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.