Abstract P5-02-10: In-silico approaches that detect immune contexture to trastuzumab response in neo-adjuvant studies

Abstract Introduction: Computational approaches have aided in estimating cellular composition of the tumour microenvironment. The evaluation of immune composition in tumours before treatment may predict pathologic complete response (pCR). The aim of the study was to perform a meta-analysis of HER2-positive breast cancer subjects who received neoadjuvant trastuzumab to detect associations between immune cells measured by CIBERSORT and ESTIMATE and pCR. Methods: PubMed was used to identify transcriptomic data of HER2-positive breast cancer patients who received neoadjuvant trastuzumab. Baseline data from eight neoadjuvant studies (N=338) was downloaded from GEO. Data from each study was background corrected and quantile normalised using ‘limma’ or ‘oligo’ packages in R. Immune profiles per sample was generated using computational softwares CIBERSORT and ESTIMATE, and were then linked to pCR status. Correlations between immune contexture and pCR for each study were interpreted using statistical testing. Meta-analysis by a logistic regression model was conducted on studies which passed assumptions to identify CIBERSORT immune subsets robust to pCR. Results: CIBERSORT results showed that three studies had reduced T follicular helper cells (Tfh) (Brodsky p=0.38, CHER-LOB p=0.17, TransNOAH p=0.25) and two studies had reduced plasma cells (CHER-LOB p=0.15, Brodsky p=0.38) in the pCR group, but was not significant after multiple correction. ESTIMATE analysis showed that data from two studies had elevated immune infiltration in pCR (Brodsky p=0.19, CHER-LOB p=0.10) but was not significant. A meta-analysis of pooled data from four studies (TRIO-US B07, 03-311, TransNOAH, CHER-LOB) showed that low Tfh (p=0.053, OR=0.04, CI [0.0012-0.99]) and high memory B-cells (p=0.008, OR=2126.9, CI [8.12-7.65 × 10+5]) prior to trastuzumab treatment may be associated with a better chance of achieving pCR. Conclusion: Results from our meta-analysis proposed that memory B- and T follicular helper subsets may predict a role in achieving pCR. Incorporating studies with larger sample cohorts such as the CALGB-40601 (N=265) study can achieve statistical power of this analysis. Citation Format: Dalal AlSultan, Alex J. Eustace, Stephen F. Madden, John Crown. In-silico approaches that detect immune contexture to trastuzumab response in neo-adjuvant studies [abstract]. In: Proceedings of the 2022 San Antonio Breast Cancer Symposium; 2022 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2023;83(5 Suppl):Abstract nr P5-02-10.