No abstract

Breast cancers are complex ecosystems of malignant cells and the tumour microenvironment1. The composition of these tumour ecosystems and interactions within them contribute to responses to cytotoxic therapy2. Efforts to build response predictors have not incorporated this knowledge. We collected clinical, digital pathology, genomic and transcriptomic profiles of pre-treatment biopsies of breast tumours from 168 patients treated with chemotherapy with or without HER2 (encoded by ERBB2)-targeted therapy before surgery. Pathology end points (complete response or residual disease) at surgery3 were then correlated with multi-omic features in these diagnostic biopsies. Here we show that response to treatment is modulated by the pre-treated tumour ecosystem, and its multi-omics landscape can be integrated in predictive models using machine learning. The degree of residual disease following therapy is monotonically associated with pre-therapy features, including tumour mutational and copy number landscapes, tumour proliferation, immune infiltration and T cell dysfunction and exclusion. Combining these features into a multi-omic machine learning model predicted a pathological complete response in an external validation cohort (75 patients) with an area under the curve of 0.87. In conclusion, response to therapy is determined by the baseline characteristics of the totality of the tumour ecosystem captured through data integration and machine learning. This approach could be used to develop predictors for other cancers.

No abstract

Background: Pathologic complete response (pCR) is a proposed surrogate endpoint for predicting long-term clinical benefit on endpoints such as disease-free survival (DFS), event-free survival (EFS), or overall survival (OS). A meta-analysis is needed to establish the magnitude of pCR improvement on a trial level that results in improved DFS, EFS, or OS. Methods: We identified 12 neoadjuvant randomized trials (N = 13,125) with pCR clearly defined and long-term follow-up available for EFS and OS. Trials included AGO 1 (n = 668), ECTO (n = 1355), EORTC 10994/BIG 1–00 (n = 1856), GeparDuo (n = 907), GeparQuattro (n = 1495), GeparTrio (n = 2072), GeparTrio-Pilot (n = 285), NOAH (n = 234), NSABP B18 (n = 760), and NSABP B27 (n = 2411), PREPARE (n = 733), and TECHNO (n = 217). Key objectives of the meta-analysis were to determine: (1) the relationship of pCR to EFS and OS, (2) the definition of pCR that correlates best with long-term outcome, (3) the breast cancer subtypes in which pCR is best correlated with long-term outcome and (4) the magnitude of pCR effect needed to improve EFS and OS. We compared three pCR definitions: absence of invasive cancer and in situ cancer in the breast and axillary nodes (ypT0ypN0), absence of invasive cancer in the breast and axillary nodes with DCIS allowed (ypT0/isypN0), and absence of invasive cancer in the breast with DCIS allowed irrespective of nodal involvement (ypT0/is). Results: Overall 13%, 18% and 22% of patients achieved a pCR defined as ypT0ypN0, ypT0/isypN0, and ypT0/is, respectively. Eradication of tumor from both the breast and lymph nodes (ypT0ypN0 or ypT0/isypN0) was better associated with improved EFS and OS compared to eradication of tumor from the breast alone (ypT0/is). Patients who achieved a pCR (ypT0/isypN0) had an improved EFS (HR = 0.48) and OS (HR = 0.36) compared to those who did not. pCR was uncommon in patients with low-grade hormone receptor-positive (HR+) tumors (7%) and more common in the following tumor subtypes: high-grade HR+ (16%), triple negative (34%), HR+/HER2+ (30%), and hormone receptor-negative (HR−)/HER2+ (50%). Patients with more aggressive tumor subtypes who achieved pCR had greater EFS compared to patients who did not achieve pCR as follows: HR+ high grade (HR = 0.27), HR+/HER2+ (HR = 0.58), HR−/HER2+ (HR = 0.25), and triple negative (HR = 0.24). A trial level analysis on the relationship between pCR effect size and EFS did not show a correlation. Conclusions: Individual patients who attain a pCR, defined as either ypT0ypN0 or ypT0/isypN0, have a more favorable long-term outcome. The data show comparable EFS or OS regardless of the presence or absence of DCIS. For consistency, a standard pCR definition (ypT0ypN0 or ypT0/isypN0) should be used in future trials. Impact of pCR effect is limited to patients with HR+/grade 3, HR−/HER2−, and HER2+ tumors. This meta-analysis did not establish the magnitude of increase in pCR rate needed to predict the superiority of one regimen over another in terms of EFS or OS. This may be due to low pCR rates and the heterogeneity of the patient population included in this meta-analysis. The absolute magnitude of improvement in pCR rate needed to impact long-term outcome may be greater than the observed difference in these trials and may vary according to breast cancer subtype. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr S1-11.

No abstract