Abstract P5-02-05: Variability in ductal carcinoma in situ grading among an international group of pathologists

Abstract

Abstract Introduction: Ductal Carcinoma In Situ (DCIS) is a proliferation of neoplastic cells confined to the ducto-lobular system. Due to uncertainty which DCIS lesions progress to invasive breast cancer women with DCIS receive surgery and radiotherapy. The PRECISION (PREvent ductal carcinoma In Situ Invasive Overtreatment Now) initiative intends to improve the management of DCIS including reduction of unnecessary treatment. Histological high grade is an unfavorable prognostic indicator used to guide treatment decisions, and used as an exclusion criterion for DCIS trials that randomize between active surveillance and standard treatment. Hence, accurate assessment of grade is important in current clinical practice. Pathologists within PRECISION setup this study to compare histological parameters of four retrospective DCIS cohorts from three different countries to investigate interobserver variability particularly in grading of DCIS. Method: Tissue slides from two population-based cohorts (UK Sloane cohort n=110; Dutch DCIS cohort n= 110) and two hospital based cohorts (Duke n=110; MDACC n=95) were included, leading to a total of 425 slides. All slides were reviewed by 9 breast pathologists originating from the UK, US and the Netherlands (NL) using the digital software platform Slidescore. The slides included were a representative selection of histological grades from all cases in these respective studies (see table 1 for distribution). Kappa (κ) values were calculated based on a generalized linear mixed model using the ordinal package in R. Results: The distribution of grade according to the majority opinion and originally assigned during diagnosis is shown in table 1. The cohort from NL demonstrated less grade 1 lesions according to the majority opinion compared to originally assigned grade. Table 2 shows the κ values of all evaluated histological variables. DCIS grade 1,2,3, grade 1/2 vs 3, necrosis and calcification show moderate agreement (κ between 0.4-0.6). Lymphocytic infiltrate, periductal fibrosis and mitoses demonstrate poor agreement (κ between 0.2-0.4). Conclusion: Histological factors including DCIS grade 1/2/3 and DCIS grade 1/2 versus 3 show only moderate agreement in this study using whole slide images. Preliminary analyses show no differences in κ values with the inclusion of country of the pathologist. However, the distribution of grade differs between the grading originally assigned and determined by majority opinion. Since clinical decisions are based on histological variables, it is important to realize that interobserver variability play a part in the diagnostic process. We are currently reviewing con- versus discordant cases and discussing how pathology guidelines can be improved. In addition, we are exploring whether artificial intelligence-based methods could be used to standardize grading of DCIS. Table 1. Distribution of grade of all slides and slides from participating centers. Slides were excluded if the majority opinion was equally divided between two grades or if originally grade 1-2/2-3 was assigned.Origin of tissue slidesGrade 1 (n)Grade 2 (n) Grade 3 (n)Excluded nAll tissue slides (n=425)Distribution according to majority opinion12% (49)40% (163)48% (197)16Distribution originally assigned12% (36)36% (111)53% (165)113Slides from Netherlands Cancer Institute (n=110)Distribution according to majority opinion6% (6)31% (33)64% (68)3Distribution originally assigned17% (19)33% (36)50% (55)0Slides from Sloane (n=110)Distribution according to majority opinion20% (20)51% (52)29% (30)8Distribution originally assigned11% (12)31% (34)58% (64)0Slides from Duke (n=110)Distribution according to majority opinion9% (10)44% (47)47% (50)3Distribution originally assigned5% (5)45% (41)50% (46)18Slides from MDAnderson (n=95)Distribution according to majority opinion14% (13)33% (31)53% (49)2Distribution originally assignedTo be collected Table 2. Kappa values of evaluated histological parametersVariableModel based weighted kappa (κma)95% confidence intervalDCIS grade 1/2/30.500.440.57DCIS grade 1/2 vs 30.520.440.60DCIS grade 1 vs 2/30.450.400.50DCIS grade low vs high0.520.440.59Necrosis absent vs present (manually dichotomized)0.550.510.59Calcification absent vs present0.510.480.55Lymphocytic infiltrate0.470.390.55Periductal fibrosis0.350.300.40Mitoses0.340.250.42 Citation Format: Maartje van Seijen, Katarzyna Jóźwiak, Sarah E Pinder, Allison Hall, Savitri Krishnamurthy, Jeremy SJ Thomas, Laura Collins, Jonathan Bijron, Joost Bart, Danielle Cohen, Wen Ng, Hilary Stobart, Jan Hudecek, Esther H Lips, Jelle Wesseling. Variability in ductal carcinoma in situ grading among an international group of pathologists [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P5-02-05.