Patient-Derived Organoids Predict Treatment Response in Metastatic Colorectal Cancer

Abstract

Purpose: Accurately predicting treatment response in metastatic colorectal cancer (mCRC) is critical to avoid unnecessary toxicity and improve patient outcomes. Patient-derived organoids (PDO) are promising models, but larger prospective studies are needed to confirm their predictive value.

Experimental design: Patients with mCRC underwent a metastatic biopsy for PDO establishment, before starting new systemic treatment. Predictors of PDO establishment were identified. PDOs were incubated with a seven-drug panel, including the patient's treatment, to determine drug sensitivity as measured by CyQUANT cell viability [area under the nonfitted "curve" of the raw viability or growth rate inhibition values (AUC and GRAUC) and concentration that gives half-maximal viability or GR inhibition (IC50 and GR50)]. Patient response was measured by size change of biopsied and all target lesions. The diagnostic performance was evaluated by positive predictive value, negative predictive value, and area under the ROC curve. Additionally, the association between PDO response and survival was assessed.

Results: A total of 232 patients were included, and 205 biopsies were obtained. PDO establishment success increased from 22% to 75%, yielding 52% overall. Male sex, increased lactate dehydrogenase, biopsy in academic hospitals, optimized culture conditions, and experience were related to PDO establishment success. In this interim analysis, focused on oxaliplatin-based doublet chemotherapy, 42 PDOs were screened. PDO drug sensitivity significantly correlated with the response of the biopsied lesion (R = 0.41-0.49, P < 0.011) and all target lesions (R = 0.54-0.60, P < 0.001) for all treatments combined. The 5-fluorouracil and oxaliplatin PDO screens demonstrated high predictive accuracy (positive predictive value: 0.78; negative predictive value: 0.80; area under the ROC curve: 0.78-0.88) and were associated with progression-free survival and overall survival (P = 0.016 and 0.049).

Conclusions: We identified predictors for successful mCRC PDO establishment and validated that PDOs can accurately predict patient outcomes during systemic treatment, specifically with 5-fluorouracil and oxaliplatin.

Figure 1.

Schematic overview of the study. Patients underwent a biopsy of a metastasis for PDO culture and WGS, before starting a new line of systemic treatment. Predictors of successful PDO establishment were identified: male sex, increased LDH, biopsy in academic hospitals, optimized culture conditions, and experience. A total of 42 PDOs were screened for standard-of-care treatments, and the patient received standard systemic treatment. PDO response was correlated with patient response, and the association between PDO response and survival was assessed. (Created in BioRender. Roodhart, J. [2025] https://BioRender.com/by1f1uq.)

Figure 2.

Clinical variables related to PDO establishment success: CEA, LDH, mutational status, sex, primary tumor sidedness, prior chemotherapeutic treatment, metastatic site of the biopsy, WGS success, and aggregated biopsy color (brown, pink, red, and white). CEA, carcinoembryonic antigen; mut, mutant.

Figure 3.

Genomic landscape of 36 tumors and/or PDOs derived from patients with mCRC. Top, microsatellite stability, tumor mutational load, tumor mutational burden, and whether the sequencing was performed on the PDOs or the tumor of origin; bottom, somatic driver mutations. MS, microsatellite status; MSI, microsatellite instability; MSS, microsatellite stability; Seq, sequencing; TML, tumor mutational load.

Figure 4.

Correlations between patient response (best RECIST response, size change of all target lesions, and the biopsied lesion) and PDO response (normalized GR_AUC per treatment in A–C and AUC in D–F). ROC curves showing the performance of the PDO classification based on GR_AUC. The curve plots the true-positive rate (sensitivity) against the false-positive rate (1 − specificity). The AUC indicates the model’s ability to distinguish between responders and nonresponders to 5-FU and oxaliplatin, defined by size change of all target lesions (G) and the biopsied lesion (H). Kaplan–Meier PFS (I) and OS (J) curves of patients stratified by PDO sensitivity to 5-FU and oxaliplatin, based on normalized GR_AUC (cutoff = 0.63). Censored events are indicated by vertical bars on the corresponding curve. The table underneath each plot denotes the numbers at risk. Log-rank test–based P value is shown. SN-38, active metabolite of irinotecan.

Figure 5.

A, Heatmap of the drug sensitivity per PDO for different types of systemic treatment. The columns show the normalized GR_AUC (left) and AUC (right), respectively. Potential other treatment options for intermediate or high 5-FU and oxaliplatin–resistant PDOs, as well as intermediate or high 5-FU and oxaliplatin–sensitive but 5-FU and irinotecan–resistant PDOs, are shown in black bars. PDOs that were not screened are shown in gray. B, Sensitivity (GR_AUC) to doublet and triplet chemotherapy showing increased sensitivity with the addition of irinotecan. C, Sensitivity (normalized GR_AUC) to panitumumab for PDOs categorized according to the tumor’s mutational status and primary tumor sidedness (RAS-mutant, BRAF-mutant, and/or right-sided; RAS/BRAF-WT and left-sided. Boxplots show the minimum, median, maximum, upper and lower quartiles, and individual data points. iri, irinotecan; oxa, oxaliplatin; TT, trifluridine/tipiracil.