Abstract P3-01-18: Clinical validation of image-based AI predictive biomarkers for precision neoadjuvant triple-negative breast cancer treatment (PEAR-TNBC): Interim results from Cohort A

Background: Pathological complete response (pCR) to neoadjuvant therapy strongly predicts the risk of recurrence and death in triple-negative breast cancer (TNBC). There are multiple neoadjuvant regimens available for TNBC but no biomarkers to match patients to treatments. Pear Bio has developed a functional precision medicine test coupling an organ-on-a-chip with a computer vision (CV) pipeline through which an individual patient’s response to different treatments can be monitored simultaneously ex vivo using time-lapse 3D microscopy. The results are available within 1 week of the biopsy being received. PEAR-TNBC (NCT05435352) is an ongoing prospective, two-cohort, observational trial to assess the accuracy of the Pear Bio test in predicting pCR to neoadjuvant treatment in TNBC patients. Here we report the interim results from Cohort A of the trial. Methods: To be eligible, patients need to have ≥10mm tumor, stage I-III, TNBC, planned for neoadjuvant treatment (chosen by the treating oncologist) followed by surgery and be willing to undergo a study-specific core needle biopsy. To be evaluable, the tumor sample needs to be successfully cultured in the 3D ex vivo assay and the patient has to complete at least 4 cycles of neoadjuvant chemotherapy. The primary endpoint is the accuracy of the Pear Bio test in predicting pCR (ypT0/Tis ypN0), defined as specificity, with secondary outcomes as sensitivity, and positive and negative predictive values. The Pear Bio test is run in parallel with the patient’s treatment and the treating oncologist is blinded to the results. We used the change in dead cell count between Day 0 and Day 3 as the metric to predict pCR, optimised the cut-off by plotting a Receiver Operating Characteristic (ROC) curve, and calculated the area under curve (AUC) for the ROC curve, as well as other measures of classification performance. The treatment responses in the test to different regimens were ranked and compared to the response seen in the patient with the regimen chosen by the treating oncologist. Results: Cohort A recruitment began on 27th May 2022 and completed on 4th July 2024, having recruited 34 patients from five UK sites. Four patients had to be excluded (one bacterial contamination, one declined NAC, two yielded too few cells) and 18 were still awaiting clinical outcomes. In the 12 evaluable patients, median age was 53 (IQR: 42 – 64) years and median tumor size was 27mm (IQR: 20 – 34). One patient was node positive; seven patients were Stage 1 – 2B, one stage 3A and four undetermined staging. Seven patients received the KEYNOTE-522 regimen, four received AC-CarboTaxol and one EC-CarboTaxol. The median number of cycles of neoadjuvant therapy received was 8.5. Nine patients achieved a pCR at surgery. The AUC for predicting pCR was 0.81 (p=0.046), with the Pear Bio test correctly predicting outcomes in 10 of the 12 patients. For the three patients who did not achieve a pCR, the assay suggested that other regimens may have been more effective. Discussion: We have presented interim results of a novel functional precision medicine assay to predict pCR in patients with TNBC receiving neoadjuvant therapy. We have shown the successful culture of tumor cells and demonstrated intra- and inter-patient variation in response to different therapies. The assay shows promise in predicting pCR, although the number of patients analyzed is small currently. The trial data continues to mature, and Cohort B is now recruiting. Future trials will run the Pear Bio test prior to therapy selection, guiding the choice of regimen to assess the test’s ability to increase pCR rates or de-escalate treatment safely. Citation Format: Peter Hall, Matthew Williams, Eleonora Peerani, Elli Tham, Francesco Iori, George de Fraine, Kerrie Loughrey, Andreas Kaffa, Thomas Richardson, Carolina Liberal, Angeliki Velentza-Almpani, Demi Wiskerke, Farah Sangkolah, Aston Crawley, Jay Kearney, Edgar Molina, Nourdine Bah, Marios Tasoulis, Cliona Kirwan, Susan Cleator, Steve Chan, Duleek Ranatunga. Clinical validation of image-based AI predictive biomarkers for precision neoadjuvant triple-negative breast cancer treatment (PEAR-TNBC): Interim results from Cohort A [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2024; 2024 Dec 10-13; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(12 Suppl):Abstract nr P3-01-18.