47?Pathologists enhance interpretation of automated multiplex immunohistochemistry assays in cancer immunotherapy trials

摘要

Background Multiplex fluorescence immunohistochemistry (mFIHC) enables simultaneous detection of multiple biomarkers on a single tissue section. Spatial patterns and differential expression of immune- and tumor cell biomarkers serve as powerful predictors of immunotherapies. In a recent meta-analyses of 8135 patients treated with PD1/L1 pathway blockers, mFIHC was found to provide highest predictive power (P0.05) amongst commonly utilized biomarker modalities, namely, PD-L1 IHC, Tumor Mutation Burden and Gene Expression Profiling alone. [Lu et al ., JAMA Oncol 2019]. As biomarkers in mFIHC assays are read by computer-aided algorithms, the role of pathologists in the digital workflow has been debated. Utilizing clinical cases representing multiple tumor indications, we illustrate the critical collaboration between pathologists (human intelligence, HI) and computer workflows (artificial intelligence, AI) required for accurate interpretation of mFIHC assays in cancer immunotherapy trials. Methods In our clinical trial laboratory, pathologists are involved in pre-analytical, analytical and post-analytical phases of clinical trial sample testing. In the pre-analytical phase, pathologist(s) perform histological examination of H&E stained tissue sections to annotate and confirm tissue types, diagnosis, tissue integrity and acceptance (including viable tumor component), followed by determination of Region of Interest (ROI) for subsequent analysis by computerized programs. In the analytical phase, pathologists identify specific areas of biological and/or clinical interest within ROI (tumor, non-tumor, invasive margin, and tumor-stromal interphase) in the computer scans, as well as exclude ROI containing necrosis, hemorrhage, blood vessels, and autofluorescence. Those pathologist-selected images are then quantified by digital pathology software such as Automated QUantitative Analyses (AQUA?) technology. Finally, pathologists also provide interpretation and summarize findings relevant to the clinical study during the post-analytical phase. Results Case studies representing distinct malignancies, such as melanoma, non-small cell lung cancer, squamous cell carcinoma of head and neck and diffuse large B-cell lymphoma, illustrating the role of pathologists and especially in rescuing challenging cases and interpreting biomarkers scores from mFIHC assays will be presented. Conclusions With the advancement in technologies to detect increasing number of biomarkers in a single tissue section and accompanied growth of mFIHC assays in immuno-oncology studies, there is a clear transition from conventional pathology (HI) to computer-aided pathology (AI HI) that will ultimately ensure greater accuracy, reproducibility and standardization of clinical trial testing, and enable approval of more effective therapies and better patient care.