Predicting the risk of severe adverse reactions to chemotherapy is of great clinical significance for proper selection of effective and safe treatment for elderly cancer patients. The present study aimed to verify and compare the value of two evaluation models of chemotherapy risk prediction for elderly cancer patients through prospective analysis. The two evaluation models assessed were the Chemotherapy Risk Assessment Scale for High-Age Patients (CRASH) and Cancer Aging Research Group (CARG) toxicity scores. Elderly patients aged >= 70 with cancer were recruited at two participating hospitals in China and completed an assessment prior to starting chemotherapy. CRASH and CARG toxicity scores of each participant were calculated. Chemotherapy-related toxicity was recorded through each cycle of chemotherapy. A total of 106 participants were recruited between September 2015 and August 2018. The CRASH and CARG toxicity scores were positively correlated (r=0.689; P<0.01). Of the participants, 54 (50.9%) participants underwent a grade 3-5 chemotherapy-related toxicity and 21 (19.8%) experienced grade 3-5 nonhematological toxicity in the process of treatment. CRASH and CARG toxicity scores predicted severe chemotherapy-related toxicity and had a high discriminatory value based on receiver operating characteristic curve analysis (area under the curve of 0.772 and 0.760, respectively; P<0.001). The results of the present study indicate that the CRASH and CARG toxicity scores are helpful tools for the prediction of severe chemotherapy-related toxicity, and are recommended for routine oncology practice.