Use of MammaTyper® in predicting response to neoadjuvant therapy in HER2-positive breast cancer

Goals: Neoadjuvant therapy (NAT) has become the standard of care for most HER2-positive early breast cancer (BC). However, 25 to 50% of patients fail to achieve a pathological complete response (pCR). Our study aims to test the CE/IVD MammaTyper® kit (Cerca Biotech) as a predictor of response in this setting. Methods: One-hundred and sixty-one HER2-positive/3+ IHC-score invasive BC patients were enrolled. All patients underwent trastuzu mab-based NAT combined with a taxane backbone. The study was approved by the local Ethical Committee. Three patients were excluded due to insufficient amount of RNA, along with 4 patients treated with dual anti-HER2 blockade. Therefore, a total of 154 FFPE preoperatory core-biopsies were tested. Of these, 79 were hormone positive (HR+) and 75 hormone-negative (HR-). Ninety-one subjects achieved a pCR while 63 attained a pathological partial response (pPR). MammaTyper®, a molecular diagnostic RT-qPCR test, in vitro was used to assess the relative mRNA expression levels of the ERBB2, ESR1, PGR and MKI67 genes. A machine-learning, Python-based Decision Tree Algorithm was used to predict pCR from the ΔΔCq values of the four genes alongside tumor size and nodal status (MTClin). Samples were stratified according to hormone receptor (ER and/or PgR) status from MammaTyper®. Focusing on a balance of interpretability and generalizability, we tuned key hyperparameters and used GridSearchCV with a 5-fold cross-validation. Analytical accuracy was evaluated in terms of sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). The study was funded by: PNRR - M4.C2 - Tuscany Health Ecosystem (THE) - Spoke n.6, CUP I53C22000780001 awarded to Cristian Scatena and Antonio Giuseppe Naccarato; PNRR - M4C2-I1.3 Project PE00000019 “HEAL ITALIA,” CUP E63C22002080006 awarded to Federica Martorana and Paolo Vigneri. Results: Of the Decision Trees generated, two were selected showing high specificity and sensitivity, and plausible biomarker hierarchy. In detail, the selected tree for HR+ tumors had a sensitivity of 90.2%, a specificity of 86.8%, a PPV of 88.1% and a NPV of 89.2%. The decision tree for HR- tumors had a sensitivity of 96%, a specificity of 88%, a PPV of 94.1% and a NPV of 91.7%. Conclusions: MTClin may discriminate patients with HER2-positive BC who will achieve pCR from those who will not, representing a powerful decision tool in terms of escalation/de-escalation of treatment approaches.