Colorectal cancer (CRC) shows highly ineffective therapeutic management. An urgent unmet need is the random assignment to adjuvant chemotherapy of high-risk stage II and stage III CRC patients without any predictive factor of efficacy. In the field of drug discovery, a critical step is the preclinical evaluation of drug cytotoxicity, efficacy, and efficiency. We proposed a patient-derived 3D preclinical model for drug evaluation that could mimic in vitro the patient’s disease. Surgically resected CRC tissue and adjacent healthy colon mucosa were decellularized by a detergent-enzymatic treatment. Scaffolds were recellularized with HT29 and HCT116 cells. Qualitative and quantitative characterization of matched recellularized samples were evaluated through histology, immunofluorescences, scanning electron microscopy, and DNA amount quantification. A chemosensitivity test was performed using an increasing concentration of 5-fluorouracil (5FU). In vivo studies were carried out using zebrafish (Danio rerio) animal model. Permeability test and drug absorption were also determined. The decellularization protocol allowed the preservation of the original structure and ultrastructure. Five days after recellularization with HT29 and HCT116 cell lines, the 3D CRC model exhibited reduced sensitivity to 5FU treatments compared with conventional 2D cultures. Calculated the half maximal inhibitory concentration (IC50) for HT29 treated with 5FU resulted in 11.5 µM in 3D and 1.3 µM in 2D, and for HCT116, 9.87 µM in 3D and 1.7 µM in 2D. In xenograft experiments, HT29 extravasation was detected after 4 days post-injection, and we obtained a 5FU IC50 fully comparable to that observed in the 3D CRC model. Using confocal microscopy, we demonstrated that the drug diffused through the repopulated 3D CRC scaffolds and co-localized with the cell nuclei. The bioengineered CRC 3D model could be a reliable preclinical patient-specific platform to bridge the gap between in vitro and in vivo drug testing assays and provide effective cancer treatment.