Abstract Background: Metastatic breast cancer (mBC) is the leading cause of cancer mortality in women, yet treatment selection in later lines remains empirical. Patient-derived xenografts (PDX) and organoids (PDO) have been shown to predict patient’s drug response, however their clinical utility for real-time clinical decision-making is limited by lengthy turnaround times and a low tumor take for luminal breast cancers, and rapid functional assays are urgently needed. We developed a microfluidic tumor-on-chip (ToC) for rapid exvivo chemosensitivity profiling to demonstrate the potential of ToC using cancer cells from BC PDX and a fresh patient's tumor sample. Methods: After tumor dissociation, tumor cells were isolated and embedded in 3D in collagen type I inside microfluidic device chips (AIMBiotech). Viability of tumor cells was assessed using a live/dead assay after several days of drug exposure (carboplatin, paclitaxel, 5-fluorouracil, and trastuzumab-deruxtecan if applicable) and compared to tumor growth measured invivo in the corresponding PDX model using a caliper. A panel of 6 PDX models was selected to reflect distinct BC subtypes and chemosensitivity profiles: 4 TNBC, 1 HER2 3+ and 1 ER+. To mimic clinical biopsies, we engineered a miniaturized chip by casting PDMS in 3D-printed silanized molds optimized for low cell input, and evaluated its performance using simulated core needle biopsies (18G) on PDX tumors and a patient sample. Results: We first investigated the response of ToC derived from fresh PDX samples. Differential drug responses were observed after 4 days of on-chip culture, suggesting that this duration is required to discriminate between sensitive and resistant models. The #152 PDX model (TNBC) showed a strong invivo response to carboplatin and paclitaxel, mirrored by a significant drop in cell viability on the ToC at D4. In contrast, neither mouse nor ToC responded to 5-FU. We performed live imaging to validate our endpoint measurement method, and comparable patterns of drug sensitivity were obtained. Overall, when a drug was identified as effective in the PDX model, we observed ToC sensitivity at D4 in 78 % of cases (in 7 cases out of 9, sensitivity was correctly detected on ToC in 6 PDX models using 4 drugs). Conversely, when the drug was resistant in PDX, we confirmed resistance in 100% of cases (11/11 using 3 drugs). Altogether, these findings highlight the correlation between the two models and demonstrate the ability of ToC platform to deliver rapid functional readouts within a clinically actionable timeframe of 4 days. Next, we successfully generated a functional ToC model using a fresh human primary tumor of TNBC; exposure to paclitaxel suggested tumor sensitivity. We finally developed an innovative microfluidic device compatible with patient biopsy samples. From three 18G core needle biopsies per tumor, we consistently isolated 100,000 to 200,000 viable tumor cells, allowing the generation of up to 12 individualized ToC units and demonstrating the feasibility of establishing ToC from limited biopsy material. Conclusions: Microfluidic ToC replicates PDX drug responses within 4 days using minimal tissue, a timeframe compatible with clinical decision-making. Optimization for cell use and ease of automation supports future clinical-scale implementation and high-throughput workflows, addressing a key gap in functional precision oncology. Prospective studies are warranted to validate ToC-guided therapy in BC. Citation Format: C. Helal, N. Schintu, J. Jin, L. Pinon, E. Montaudon, L. Sourd, H. Derrien, M. Nurmik, M. Simon, M. Parrini, L. Cabel, S. Descroix. Tumor-on-chip as a personalized platform for rapid drug testing in breast cancer abstract. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS4-04-06.
Helal et al. (Tue,) studied this question.