Histopathological grading of breast ductal carcinoma In Situ: validation of a web-based survey through intra-observer reproducibility analysis

Abstract

BackgroundHistopathological grading diagnosis of ductal carcinoma in situ (DCIS) of the breast may be very difficult even for experts, and it is important for therapeutic decisions. The challenge may be due to the inaccurate and/or subjective application of the diagnosis criteria. The aim of this study was to investigate the intra-observer agreement between a traditional method and a developed web-based questionnaire for scoring breast DCIS.MethodsA cross-sectional study was carried out to evaluate the diagnostic agreement of an electronic questionnaire and its point scoring system with the subjective reading of digital images for 3 different DCIS grading systems: Holland, Van Nuys and modified Black nuclear grade system. Three pathologists analyzed the same set of digitized images from 43 DCIS cases using two different web-based programs. In the first phase, they accessed a website with a newly created questionnaire and scoring system developed to allow the determination of the histological grade of the cases. After at least 6 months, the pathologists read again the same images, but without the help of the questionnaire, indicating subjectively the diagnoses. The intra-observer agreement analysis was employed to validate this innovative web-based survey.ResultsOverall, diagnostic reproducibility was similar for all histologic grading classification systems, with kappa values of 0.57 ± 0.10, 0.67 ± 0.09 and 0.67 ± 0.09 for Holland, Van Nuys classification and modified Black nuclear grade system respectively. Only two 2-step diagnostic disagreements were found, one for Holland and another for Van Nuys. Both cases were superestimated by the web-based survey.ConclusionThe diagnostic agreement between the web-based questionnaire and a traditional method, both using digital images, is moderate to good for Holland, Van Nuys and modified Black nuclear grade system. The use of a scoring point system does not appear to pose a major risk of presenting large (2-step) diagnostic disagreements. These findings indicate that the use of this point scoring system in this web-based survey to grade objectively DCIS lesions is a useful diagnostic tool.

Grading Ductal Carcinoma In Situ (DCIS) of the Breast – What’s Wrong with It?

“…Several studies have looked at the consistency of grading or prognostically classifying DCIS, and generally found that these DCIS classifications were moderately reproducible, while a few others suggested worse or somewhat better agreement (Supplementary Table 2 and Fig. 4 representing it) [6,19,20,[30][31][32][33][34][35][36][37][38][39]. This degree of interobserver variability is also reflected by a recent analysis of nearly 5000 DCIS diagnosed during a period of 4 years in the Fig.…”

“…Nuclear grade 2 tiered (Cyto)nuclear grade 3 tiered WHO 2012 [6] Bethwaite [31] Agreement [39] Sneige [32] slight/missing (kappa 0-0.2) Sloane [33] fair (kappa 0.21-0.4 Sloane [33] moderate (kapa 0.41-0.6) Wells [34] substantial (kappa 0.61-0.8) Douglas-Jones [35] almost perfect (kappa >0.8) Gomes [36] empty cells Gomes [36] Schuh [37] van Bockstal [38] van Bockstal [38] Normal cells for comparative size [20] RBC RBC G1 [5,21,22,23,25 ] RBC/ductal epithelial cell nucleus G2 RBC/ductal epithelial cell nucleus G3 [3,4] RBC/ductal epithelial cell nucleus [24] No comparative size LORIS inclusion [18] Fig. 3 Comparative nuclear size described in some publications concerning the grading of DCIS.…”

“…Different agreement levels reached in reproducibility studies on DCIS grade[6,19,20,[31][32][33][34][35][36][37][38][39] …”

Impact of Preoperative Breast MR Imaging and MR-guided Surgery on Diagnosis and Surgical Outcome of Women with Invasive Breast Cancer with and without DCIS Component

“…and stained with four different colors to allow a three-dimensional orientation. Nuclear grade was categorized according to the Van Nuys classification system (28). When the differential diagnosis was between ductal hyperplasia and low-grade DCIS, additional immunohistochemical staining (cytokeratin 5/6 and p63) was performed.…”

An interactive learning framework for scalable classification of pathology images

“…In cancers, the shapes and types of cells present carry important diagnostic and prognostic information that are used to classify the tumor and assess how advanced a patient’s disease is. Manual evaluation of tissue histology by trained professionals is highly subjective, being prone to both considerable intra-observer and inter-observer variations [2]. The development of more objective quantitative metrics for evaluating pathology images remains a significant barrier in effectively using this resource in research and clinical care.…”