Objective To explore the construction of a predictive model for adverse reactions of liver cancer patients with radiotherapy and chemotherapy containing nutritional indicators. Method Prospectively selected 231 patients with liver cancer radiotherapy and chemotherapy admitted to Lianyungang First People's Hospital from January 2019 to March 2021 as the research objects. Randomly selected 70% (162 cases) of cases as the modeling set, and 30% (69 cases) as the test set. Compared the relevant data of the modeling set after radiotherapy and chemotherapy, and apply multi⁃factor Logistic regression to screen the relevant factors, construct the prediction model, use the consistency index (C⁃index) to quantify and calibrate the curve, evaluate the performance of the model, draw the decision curve to analyze and evaluate the clinical net of the nutritional index model. At the same time, external verification of the model is performed on the test set. Result After one course of radiotherapy and chemotherapy, 74 cases (45.68%) in the modeling set had adverse reactions and were classified as the adverse reaction group, and 88 cases (54.32%) without adverse reactions were classified as the non⁃adverse reaction group. In modeling set two groups, tumor diameter, controlling nutritional status (CONUT) score, alpha⁃fetoprotein (AFP), alkaline phosphatase (ALP), γ⁃glutamyl transferase (GGT), des⁃γ⁃carboxy prothrombin (DCP), and prognostic nutritional index (PNI) had statistically significant differences (P<0.05). Multivariate Logistic analysis showed that tumor diameter (OR=1.699, 95%CI=1.117-2.583), CONUT score (OR=2.396, 95%CI=1.205-4.763), AFP (OR=1.068, 95%CI=1.020-1.118), DCP (OR=1.013, 95%CI=1.000-1.025), GGT (OR=1.090, 95%CI=1.037-1.144), ALP (OR=1.013, 95%CI=1.003-1.023) in liver cancer are independent risk factors for adverse reactions and chemotherapy. PNI (OR=0.913, 95%CI=0.875-0.953) is a protective factor for adverse reactions (all P<0.05). The Nomogram model predicts the occurrence of adverse reactions with a C⁃index of 0.867 (95%CI=0.815-0.920). The decision curve shows that when the predicted value of the model with CONUT score and PNI nutrition⁃related index is in the interval (0-0.6), additional clinical benefits can be provided. External validation showed that among the 69 patients in the validation set, 31 (44.93%) had adverse reactions and were classified as the adverse reaction group, and 38 (55.07%) without adverse reactions were classified as the non⁃adverse reaction group. The prediction sensitivity of the model was 90.32% and the specificity was 91.67%. Conclusion The prediction model constructed with CONUT score and PNI nutritional indicators can improve the accuracy of predicting the occurrence of adverse reactions in patients with liver cancer radiotherapy and chemotherapy.