PD-L1 expression and survival in triple-negative breast cancer

Programmed death-ligand 1 (PD-L1) positivity in triple-negative breast cancer (TNBC) predicts treatment with anti-PD-L1 immunotherapy and is an independent marker for recurrence-free and metastasis-free survival. Approximately 30-60% of early TNBCs and 30-40% of metastatic TNBCs are PD-L1-positive. PD-L1 expression varies in accordance to the antibody type and method of scoring used. PD-L1 assays are highly discordant, with SP142 cited as less sensitive than 22C3.

We aimed to assess PD-L1 positivity in TNBC across two assays in relation to survival endpoints.

Immunohistochemical staining for PD-L1 was carried out on tissue microarray slides of cases of TNBC diagnosed in GUH from 2000-2017, with one slide-set used for staining with anti-PD-L1 SP142 antibody and the other used for staining with anti-PD-L1 22C3 antibody. All PD-L1-stained cases were scored via digital histological examination by two pathologists. Statistical analyses were performed with SPSS 25 and STATA 14. Univariate survival analyses for survival endpoints were carried out using Kaplan-Meier survival estimate curves and Log rank tests to assess the association between PD-L1 status and survival. Cox regression was performed for both univariate and multivariable survival analysis to calculate hazard ratios and 95% confidence intervals adjusting for clinico-pathological variables, such as age, grade, nodal status, tumour status, and tumour infiltrating lymphocytes (TILs). P-values <0.05 were deemed statistically significant.

Overall, 216 SP142-stained cases and 232 22C3-stained cases were included for analysis. PD-L1 positivity was present in 29% of SP142 cases and 15% of 22C3 cases. Of SP142-positive cases, 29% were also positive with 22C3, and of 22C3-positive cases, 40% were also positive for SP142. There was a statistically significant difference in the distribution of PD-L1 expression as detected by the SP142 and 22C3 assays (X2=15.172, p <0.001). Similar rates of recurrences and metastases occurred in PD-L1-positive and PD-L1-negative cases for both assays, with no significant difference in disease-free or metastasis-free survival. Outcome was assessed for three groups: TILs >20%/PD-L1-negative; TILs <20%/PD-L1-negative; and PD-L1-positive regardless of TILs score. The highest proportion of recurrence, metastasis and TNBC deaths occurred in the TILs <20% PD-L1-negative group, whereas the lowest proportion of events occurred in the TILs >20% PD-L1-negative group. Multivariable analysis with age, grade, tumour stage and nodal status showed TILs 20% cut-point as non-significant, while the combined TILs <20% PD-L1-negative group had significantly worse disease-free, metastasis-free and breast cancer-specific survival (BCSS) versus the combined TILs >20% PD-L1-negative group. Nodal status was a significant independent prognostic indicator across all survival endpoints.

Our results showed PD-L1 positivity in 29% of early TNBC cases stained with SP142 and 15% of cases stained with 22C3, which are lower figures than reported in the literature. We found a significant difference in the distribution of PD-L1 expression between SP142 and 22C3, which reflects the lack of interchangeability between the assays. Survival analysis revealed that PD-L1 status alone was not a significant predictor of disease-free, metastasis-free, or BCSS – for either SP142 or 22C3 – which echoes the limited prognostic role of the variable, however PD-L1 status was found to predict survival when combined with TILs 20% cut-point.