Intra- and inter-observer variability in tumor infiltrating lymphocyte scoring in breast cancer core needle biopsy.

Abstract

e12626 Background: Morphological evaluation of tumor lymphocyte infiltration (TIL) in breast cancer (BC) is gaining importance in the clinical setting, as it provides good prognostic information. Most institutions adhere to TIL working group guidelines for evaluating TIL, while having to settle on an acceptable error margin due to its subjective nature, which leads to intra and inter-observer scoring discordance. We aimed to analyze both at our institution, using experienced and inexperienced examiners, with a continuous variable scoring system. Methods: 209 BC core needle biopsy (CNB) samples were stained using hematoxylin-eosin. The percentage of stromal TIL was scored using a numerical variable ranging from 1 to 100. The examination group consisted of two experienced pathologists (pathologist A and B) and one inexperienced examiner - a medical oncology resident with a learning process containing study of the TIL working group analysis criteria and about 100 samples analysed together with an experienced pathologist Pathologist A and the resident analyzed the study samples twice, pathologist B once. Intraclass correlation coefficient (ICC) was used to measure overall intra and inter-observer agreement. Statistical analysis was performed using Google Sheets and Python. Results: 203 CNB samples were analysed (6 were excluded due to inadequate quality or an inconclusive diagnosis). Patients were aged 26 to 79 years (median 49). Sample size ranged from 1 to 16 mm (median 8). The proportion of BC subtypes was: luminal A-like 18%, luminal B-like 39%, HER2+ 10%, luminal B-like HER2+ 12%, TNBC 18%, not defined 2%. The highest proportion of high stromal TIL (≥50%) was seen in HER2+ (30%) and TNBC (22%) subtypes, as observed by pathologist A. We found good intra and inter-observer ICC (Table). Conclusions: Acceptable intra and inter-observer variability was observed in experienced pathologists, suggesting that the proposed methodology could reliably be used in clinical practice, research and trials. Interestingly, variability analysis of scores from a trained non-pathologist has also produced good results. However, it is important to note that inexperienced scoring could be prone to errors, for example counting non-lymphocyte cells as such, not recognizing regions of necrosis or sample damage, or not distinguishing between the tumor and peripheral stroma. Intraclass correlation: two-way random, single score (ICC2); pathologist A and resident (R) first and second analysis (A1, A2, R1, R2); pathologist B analysis (B1). Interpretation of ICC: <0.50 poor; 0.50-0.75 moderate; 0.75-0.90 good (0.75 = minimal acceptable value for a reliable clinical measurement); >0.90 excellent. Clinical trial information: 2018-000547-11. [Table: see text]