Use of quantitative pathology to improve grading and predict prognosis in tumours of the gastrointestinal tract

Abstract

Cancer represents a formidable health burden and was the second leading cause of death globally in 2018. In Norway, almost 35000 new cancer cases were reported in 2019. For colon cancer, the incidence and mortality rates in Norway are among the highest in the world. Furthermore, the tumour-node-metastasis (TNM) system used today is not optimal for selecting which patients should receive adjuvant therapy or not.

With the implementation of digital pathology in different pathology departments, there will be better opportunities for digital image analysis, a tool aimed at giving a more reproducible and objective diagnosis than subjective evaluation in a microscope. In digital image analysis, a computer programme is used for the quantification of different biomarkers. This can improve cancer diagnostics because several biases in manual evaluation can be reduced or avoided. One of the challenges in pathology is intra-and inter-observer variability of prognostic and predictive biomarkers. This especially applies for gastroenteropancreatic neuroendocrine neoplasms (GEP-NENs), in which the proliferation marker Ki67 is important for grading (1–3), prognosis and treatment of patients. Several studies have shown interand intra-observer variations in the manual evaluation of Ki67 positivity, which can be improved with digital image analysis. This is important because the interpretation of the immunohistochemical staining of different biomarkers, such as Ki67, often influences patient prognosis and treatment.

The immune system, especially the number of T-cells in and around the tumour, has been investigated as a promising biomarker for predicting prognosis and survival in colorectal cancer (CRC). The immune system is closely linked to microsatellite instability (MSI) in CRC, and MSI-high CRC has been shown to respond well to immune therapy. A TNM-immune is suggested based on scoring of the number of T-cells in the tumour centre and the invasive margin using digital image analysis.

In this study, we explored the correlation between T-cells in presurgical blood samples and T-cells in the invasive margins and the tumour centres in CRC with digital image analysis in a feasibility study and found a correlation. Furthermore, we used digital image analysis to calculate the immune score in colon cancer patients based on immunohistochemical (IHC) staining of cluster of differentiation (CD)3+ and CD8+ T-cells in invasive margins and tumour centres in a prospective cohort. This immune score corresponded strongly with known clinicopathological features, such as stage and MSI status.

Also, we evaluated digital image analysis as an objective assessment tool for two different proliferation markers in GEP-NENs: Ki67 and Phosphohistone 3 (PHH3). We compared manual (visual) evaluation of Ki67 from pathology reports with digital image analysis of Ki67 and found excellent agreement, but there is a tendency to upgrade cases from grade 1 to grade 2 with digital image analysis. For the digital image analysis of PHH3, the measurements were more diverging.

The data presented show the use of digital image analysis in two settings: developing an immune score as a prognostic marker in colon cancer and providing an objective and reproducible evaluation of proliferation in neuroendocrine neoplasms. With the transition to digital pathology, digital image analysis can be implemented in daily diagnostics. This implementation requires more research for the validation of the different methods. With time, digital image analysis is expected to be utilized for tasks performed by pathologists today.