AIM: To investigate the effect of magnesium level on the arteriovenous fistula (AVF) dysfunction in patients on maintenance hemodialysis (MHD). METHODS: We selected patients who underwent AVF surgery at The Second Affiliated Hospital of Nantong University from May 2011 to May 2022 and received MHD regularly for over 3 months. Patients were divided into dysfunction and non-dysfunction groups based on their AVF function, with follow-up until 30 November 2022. Retrospective data collection included pre-dialysis general data and clinical laboratory indicators. The magnesium cut-off for AVF dysfunction prediction was determined using the receiver operating characteristic (ROC) curve, and patients were categorized into high and low magnesium groups. AVF survival rates were compared using Kaplan-Meier methods, and the risk of AVF dysfunction and independent risk factors were analyzed with logistic and Cox regressions. RESULTS: In a study of 263 hemodialysis patients with a median age of 61 years, including 164 males (62.4%), 95 developed AVF dysfunction over a median follow-up of 32 months. Two groups of MHD patients were classified based on their AVF function: 95 in the dysfunction group and 168 in the non-dysfunction group. The subjects in the dysfunction group were older than those in the non-dysfunction group. Compared with the non-dysfunction group, the dysfunction group suffered significant reduction in magnesium and creatinine levels (p < 0.05), and significant increase in calcium and hemoglobin levels (p < 0.05). The ROC curve results showed that magnesium = 0.88 mmol/L was the best critical point for predicting AVF dysfunction in MHD patients, with a sensitivity of 68.42% and a specificity of 77.38%. The results of Kaplan–Meier survival analysis showed that the AVF dysfunction in the low magnesium group was significantly higher than that in the high magnesium group (log-rank χ2 = 68.678, p < 0.001). Logistic analysis showed that the low magnesium group was 9.223 times more likely to experience AVF dysfunction than the high magnesium group after adjusting for multiple confounding factors (odds ratio [OR] = 9.223, 95% confidence interval [CI], 4.876–17.445; p < 0.001). After adjusting for multiple confounding factors, multivariate Cox regression analysis suggested that advanced age, low serum magnesium, high serum calcium and high hemoglobin were independent risk factors for AVF failure in MHD patients. The risk in the low serum magnesium group was 4.534 times higher than that in the high serum magnesium group (hazard ratio [HR] = 4.534, 95% CI, 2.633–7.808; p < 0.001). CONCLUSIONS: Low serum magnesium is an independent risk factor for AVF dysfunction and can be used as a predictor of AVF dysfunction.