Some colorectal cancer patients harboring FGFR (fibroblast growth factor receptor) genetic alterations, such as copy number gain, mutation, and/or mRNA overexpression, were selected for enrollment in several recent clinical trials of FGFR inhibitor, because these genetic alterations were preclinically reported to be associated with FGFR inhibitor sensitivity as well as poor prognosis, invasiveness, and/or metastatic potential. However, few enrolled patients were responsive to FGFR inhibitors. Thus, practical strategies are eagerly awaited that can stratify patients for the subset that potentially responds to FGFR inhibitor chemotherapy. In the present study, we evaluated the sensitivity to FGFR inhibitor erdafitinib on 25 patient-derived tumor-initiating cell (TIC) spheroid lines carrying wild-type RAS and RAF genes, both in vitro and in vivo. Then, we assessed possible correlations between the sensitivity and the genetic/genomic data of the spheroid lines tested. Upon their exposure to erdafitinib, seven lines (7/25, 28%) responded significantly. Normal colonic epithelial stem cells were unaffected by the inhibitors. Moreover, the combination of erdafitinib with EGFR inhibitor erlotinib showed stronger growth inhibition than either drug alone, as efficacy was observed in 21 lines (84%) including 14 (56%) that were insensitive to erdafitinib alone. The in vitro erdafitinib response was accurately reflected on mouse xenografts of TIC spheroid lines. However, we found little correlation between their genetic/genomic alterations of TIC spheroids and the sensitivity to the FGFR inhibitor. Accordingly, we propose that direct testing of the patient-derived spheroids in vitro is one of the most reliable personalized methods in FGFR-inhibitor therapy of colorectal cancer patients.