A multigene model for response stratification to neoadjuvant chemotherapy in triple negative breast cancer

TL;DR

A 31-gene model predicts which triple negative breast cancer patients will respond well to chemotherapy, potentially avoiding unnecessary treatments.

Contribution

A 31-gene transcriptomic model was developed and validated to predict response to neoadjuvant chemotherapy in TNBC patients.

Findings

A transcriptomic model correctly clustered nearly all patients into good or poor responder categories.

External validation showed 85% accuracy in predicting good responders but only 58% accuracy for poor responders.

The model may help identify patients who can avoid additional therapies after achieving a complete response.

Abstract

Around half of triple negative breast cancer (TNBC) patients achieve a pathological complete response (pCR) based on neoadjuvant chemotherapy (NAC), which is associated with a good outcome. Conversely, in patients with a poor response to NAC, there is a clear need to administer more effective therapeutic strategies. Accurate prediction of tumor response could enable the implementation of more personalized and effective treatment strategies.

In this retrospective multicenter study, formalin-fixed paraffin-embedded tissues of pre-NAC needle biopsies from TNBC patients treated between 2013 and 2022 were analyzed. Clinical, pathological, and transcriptomic data were combined in a prediction model, using a leave-one-out design, to predict the response to NAC, followed by external validation in an independent dataset.

In total, 204 patients were included, comprising 87 good responders and 117 poor responders. A transcriptomic based prediction model showed that all samples but one clustered correctly in the good or the poor responder category. External validation showed an accuracy of 85% in predicting a good response to NAC, using a 31-gene signature. On the other hand, prediction of having a non-pCR was not substantial in this external cohort, since only 58% were predicted correctly.

This study suggests that a 31-gene prediction model may help identify TNBC patients who are likely to achieve a pCR following NAC alone. These patients may not require therapeutic intensification, such as addition of immunotherapy, thereby minimizing exposure to unnecessary treatment-related toxicity and reducing associated healthcare costs. Nonetheless, further optimization and prospective validation are needed prior to moving towards clinical implementation.

•204 TNBC biopsies studies, comparing pCR and very poor response (RCB3-like).•A transcriptomic model clustered nearly all patients as good or poor responders.•External validation: 85% accuracy in good, 58% in poor, due to low RCB3-like share.•Model may help identify pCR patients who can avoid extra therapy.

204 TNBC biopsies studies, comparing pCR and very poor response (RCB3-like).

A transcriptomic model clustered nearly all patients as good or poor responders.

External validation: 85% accuracy in good, 58% in poor, due to low RCB3-like share.

Model may help identify pCR patients who can avoid extra therapy.

Figures