205 Reproducibility study on invasiveness of pulmonary AdC with a modified classification the hematoxylin-eosin (H&E) stained slides and selected for the study 1 FFPE block containing the largest amount of invasive tumor from each case. Using the selected block, serial 4 µm sections were cut and stained with H&E, Elastin Van Gieson and immunohistochemically for Cytokeratin 7 (CK7) (clone OVTL12/30 (Agilent/Dako (Glostrup, Denmark, catno. M701801 performed in a Roche/Ventana benchmark Ultra [Roche, Basel Switzerland]), at UMC-VUmc. Cases with excessive delay in tissue fixation were excluded from the Amsterdam and Milan series (n=17 and n=5, respectively)330. From the remaining cases of Amsterdam (n=55) and Milan (n=75), 40 samples from each laboratory were selected after randomization. After digitizing 10 cases were omitted due to poor digitization with focus issue or fixation problems. All slides were sent to Tsukuba, Japan (MN, YM) and scanned [with 40x objective] as described before317. Subsequently, files were transferred to a server of UKNEQAS and the pathogate.net platform was used to organize the patient cases and served as the interface for data collection (A.S.). Morphologic evaluation All cases were initially read twice. For both readings the H&E image was supplied, and elastin and CK7 stain were optionally available for all cases. The data requested were: i) whether invasion was present (yes/no/not sure), ii) percentage of growth patterns, iii) total size, iv) invasive size, determined in two ways: A) by calculation according to the WHO (maximum diameter multiplied by invasive proportion) (according to WHO19 page 67) and B) by direct measurement using a digital ruler available in the image viewer, where the location and the length of the line was also recorded. Up to 3 lines designating separate invasive areas could be recorded. If >1 line was designated, the sum of all lines was used in further analysis. The first round of case scoring was performed according to current histopathological practice, as governed by the WHO pattern recognition criteria. After the first reading a tutorial was provided (see Supplementary file), which explained the distinctions between iatrogenic and biological collapse, and the morphological differences between them. For the second reading the morphological features of iatrogenic collapsed AIS were taken into account when determining non-invasive areas. Heatmaps were created on the platform in the following way. The drawn line(s) for total tumor and invasive size were interpreted as diameters of a circle. For each position in the image, the number of overlapping circles was calculated, and the numbers were normalized. Finally, an arbitrary digital color scale was added for visual purpose. The color-range from low to highest number of invasion scores for a pixel was blue, green, yellow orange, red. This implies that an overlay with regional preference for invasion is obtained for those pathologists who assigned invasive line(s). Sample size consideration Assuming the fraction of cases that could be downgraded to non-invasive would be around 15-20%320 with a 95% confidence interval between 10-30% the minimum sample size required was 70 cases. For this study 80 cases were digitized and 70 cases read. 15
RkJQdWJsaXNoZXIy MTk4NDMw