Reproducibility and predictive value of scoring stromal tumour infiltrating lymphocytes in triple-negative breast cancer: a multi-institutional study

Several studies have demonstrated a prognostic role for stromal tumour infiltrating lymphocytes (sTILs) in triple-negative breast cancer (TNBC). The reproducibility of scoring sTILs is variable with potentially excellent concordance being achievable using a software tool. We examined agreement between breast pathologists across Europe scoring sTILs on H&E-stained sections without software, an approach that is easily applied in clinical practice. The association between sTILs and response to anthracycline-taxane NACT was also examined. Pathologists from the European Working Group for Breast Screening Pathology scored sTILs in 84 slides from 75 TNBCs using the immune-oncology biomarker working group guidance in two circulations. There were 16 participants in the first and 19 in the second circulation. Moderate agreement was achieved for absolute sTILs scores (intraclass correlation coefficient (ICC) = 0.683, 95% CI 0.601-0.767, p-value < 0.001). Agreement was less when a 25% threshold was used (ICC 0.509, 95% CI 0.416-0.614, p-value < 0.001) and for lymphocyte predominant breast cancer (LPBC) (ICC 0.504, 95% CI 0.412-0.610, p-value < 0.001). Intra-observer agreement was strong for absolute sTIL values (Spearman rho = 0.727); fair for sTILs >= 25% (kappa = 0.53) and for LPBC (kappa = 0.49), but poor for sTILs as 10% increments (kappa = 0.24). Increasing sTILs was significantly associated with an increased likelihood of a pathological complete response (pCR) on multivariable analysis. Increasing sTILs in TNBCs improves the likelihood of a pCR. However, inter-observer agreement is such that H&E-based assessment is not sufficiently reproducible for clinical application. Other methodologies should be explored, but may be at the cost of ease of application.