Abstract P2-07-12: Triple negative breast cancer subtypes and early dynamics of the 27-gene IO score predict pCR in the NeoTRIPaPDL1 trial

Abstract Background A post-hoc of NeoTRIP trial showed that 27-gene IO score assessed on baseline samples is predictive of increased pathological complete response (pCR) with the addition of atezolizumab to carboplatin/nab-paclitaxel, whereas the LAR subtype has the lowest rate of pCR with and without atezolizumab (Bianchini G ESMO 2021). We evaluated 27-gene IO score and TNBC subtypes on biopsies collected during treatment, assessed biomarker dynamics, and studied the association with pCR. Methods In NeoTRIP, patients randomly received 8 cycles of nab-paclitaxel/carbo alone (CT) or with atezolizumab (CT/A). 258 patients were evaluable for pCR (Per-Protocol Population). We assessed IO score as binary and continuous variable, and the five 101-gene TNBC types (BL1, BL2, LAR, M, and MSL; Ring et al 2016) by RNA-seq on biopsies at baseline and day 1 of second treatment cycle (d1c2) (n: baseline 242/258, 94%; d2c2 161/258, 62%; paired 152/258, 59%). Forty-four paired samples were excluded due to lack of tumor cells at d1c2. PD-L1 (Ventana SP142) and sTILs data were available. We evaluated the association with pCR of biomarkers assessed at d1c2 and their dynamics from baseline. Results Frequency of TNBC types at d1c2 showed minor differences between arms (p = 0.055). TNBC type frequencies were 22.9% BL1, 11.4% BL2, 22.9% LAR, 21.4% M, and 21.4% MSL in the CT/A arm and 43.8% BL1, 6.2% BL2, 11.2% LAR, 21.2% M, and 17.5% MSL in the CT arm. Individual TNBC type changes from baseline to d1c2 were observed, but overall, it was not significant. Frequency of IO positive score at d1c2 was similar in CT and CT/A arm (p = 0.75). Only in CT/A, an increase from baseline to d1c2 was observed (30.9% to 49.3%, p = 0.04).Overall, TNBC types at d1c2 were predictive of pCR (p = 0.00002). Compared to BL1, LAR and M were associated with lower pCR rate in CT (OR = 0.09, 95% CI = 0.01-0.83, p = 0.034 for LAR; OR = 0.16, 95% CI = 0.04-0.66, p = 0.011 for M) and CT/A arm (OR = 0.05, 95% CI = 0.01-0.49, p = 0.010 for LAR; OR = 0.28, 95% CI = 0.06-1.28, p = 0.102). pCR rate in LAR was 11.1% and 6.2% in CT and CT/A arm, respectively. TNBC types were predictive of pCR independently of PD-L1 and sTILs.Continuous IO score at d1c2 was predictive of pCR in both CT/A (p = 0.004) and CT arms (p = 0.009). The binary IO score was significantly associated to higher pCR rate in CT/A arm only (OR = 5.42, 95% CI = 1.95-15.07, p = 0.001). A strong predictive value of the highest quartile of IO score compared to the lowest was observed in CT/A (OR = 14.73, 95% CI = 2.97-73.21, p = 0.001) and CT (OR = 4.38, 95% CI = 1.21-15.81, p = 0.024) arms. pCR rates for the highest and lowest quartiles were 72.2% vs 15.0% in CT/A and 65.2% vs 30.0% in CT arm. In CT/A binary IO score at d1c2 retained significance after adjustment for baseline PD-L1 and sTILs (p = 0.036).Combining baseline and d1c2 IO score, only d1c2 assessment was informative in CT arm. In CT/A arm, both biomarkers were informative, with assessment at d1c2 being more informative than baseline IO score when continuous scores were considered. Baseline binary IO score (OR = 25.0, 95% CI = 3.31-188.9, p = 0.002) and ΔIO score (d1c2-baseline) (OR = 11.3, 95% CI = 1.07-120.1, p = 0.044) retained significance. The combination of baseline and d1c2 binary IO score defined four groups with different likelihood of pCR: 73.7% vs 15.2% in positive/positive and negative/negative groups, respectively (OR = 15.68, 95% CI = 3.88-63.32, p = 0.0001). Conclusions Dynamic of IO score early on treatment was linked to likelihood of pCR independently of baseline biomarkers and may be an early surrogate of treatment benefit especially in atezolizumab arm. LAR and M are associated with lower pCR rate, suggesting that different therapeutic strategies may be beneficial. Combining baseline and on-treatment biomarkers can be more informative than baseline only of the complex tumor/immune co-evolution dynamic and of clinical outcome. Citation Format: Matteo Dugo, Chiun-Sheng Huang, Daniel Egle, Begoña Bermejo, Claudio Zamagni, Robert S. Seitz, Tyler J. Nielsen, Marc Thill, Antonio Anton, Stefania Russo, Eva Maria Ciruelos, Brock L. Schweitzer, Douglas T. Ross, Barbara Galbardi, Richard Greil, Vladimir Semiglazov, Balázs Gyorffy, Marco Colleoni, Catherine Kelly, Gabriella Mariani, Lucia Del Mastro, Pinuccia Valagussa, Giuseppe Viale, Maurizio Callari, Luca Gianni, Giampaolo Bianchini. Triple negative breast cancer subtypes and early dynamics of the 27-gene IO score predict pCR in the NeoTRIPaPDL1 trial [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P2-07-12.