Objective To explore the clinical value of albumin-globulin ratio(AGR)combined with prostate-specific antigen(PSA)and Gleason score in predicting biochemical recurrence(BCR)after radical resection in patients with localized prostate cancer, and to provide reference for early identification of BCR high-risk patients and individualized treatment. Methods The clinical data of 114 patients with localized prostate cancer treated during Sep.2019 and Sep.2023 were retrospectively analyzed.According to postoperative BCR status, patients were divided into the BCR group(n=43)and non-BCR group(n=71).Preoperative clinical and pathological data were compared between the two groups.Univariate and multivariate Cox regression analyses were used to assess the correlation between these indicators and BCR.The predictive performance of the combined model was evaluated using receiver operating characteristic(ROC)curve.The Kaplan-Meier method was used to compare biochemical recurrence-free survival(BFS)among patients with different AGR, PSA levels and Gleason scores.Results Compared with the non-BCR group, the BCR group had significantly higher PSA level and ISUP grading, but significantly lower AGR level(P＜0.001).Multivariate Cox regression analyses showed that AGR was negatively correlated with BCR(HR=0.031, 95%CI:0.008-0.119, P＜0.001), PSA was positively correlated with BCR(HR=1.009, 95%CI:1.002-1.016, P=0.017), and Gleason score was positively correlated with BCR(HR=1.551, 95%CI:1.199-2.005, P＜0.001).The area under the ROC curve(AUC)for the combined prediction model(AGR + PSA + Gleason score)was 0.886, with a sensitivity of 90.7% and a specificity of 78.9%.Kaplan-Meier analyses showed that patients in the low AGR group, high PSA group, and high Gleason score group had significantly lower BFS than those in the high AGR group, low PSA group and low Gleason score group, respectively(all P＜0.0001).Conclusion The combination of AGR, PSA and Gleason score is closely associated with postoperative BCR in patients with localized prostate cancer.The combined prediction model demonstrates high predictive efficacy and can provide an important reference for the early identification of high-risk BCR patients and for guiding individualized postoperative treatment.