目的 探讨血糖变异度对危重症患者早期肠内喂养不耐受的预测价值，为精准营养管理提供依据。方法 收集2022年6月—2024年6月入住ICU的283例启动肠内营养患者的临床资料，以是否发生肠内喂养不耐受（enteral feeding intolerance，EFI）分为耐受组及不耐受组，收集患者的一般资料、实验室指标、临床治疗、血糖数值等，同时计算血糖变异度。通过Logistic回归和Cox比例风险评分分析肠内喂养不耐受的独立危险因素；利用受试者操作特征曲线评估血糖变异度对EFI的预测效能。结果 283例患者ICU住院期间喂养耐受156例，不耐受127例，EFI发生率为44.87%，主要临床表现为胃潴留（50.39%）与腹泻（36.22%），发生时间集中在启动后第2~3天。Logistic回归与Cox回归结果显示，营养启动前的血糖标准差（β=0.218，OR=1.244）与急性生理学和慢性健康状况评价Ⅱ（acute physiology and chronic health evaluation Ⅱ，APACHE Ⅱ）评分（β=0.078，OR=1.081）是EFI的独立危险因素。APACHE Ⅱ评分和营养启动前血糖标准差的受试者操作特征曲线下面积（area under the curve，AUC）分别为0.696和0.658，两者联合AUC提升至0.723，准确度由67.137%、63.984%增至70.115%，灵敏度与特异度表现更均衡；营养启动期间血糖变异度指标未显示出显著的预测价值。结论 APACHE Ⅱ评分和营养启动前血糖标准差对早期肠内喂养不耐受的预测具有显著价值，两者联合能显著提升EFI预测的诊断效能。临床上应重视代谢稳定性与疾病严重度的双维度评估，从而优化营养管理策略，降低EFI发生率，改善危重患者的预后。

Objective To explore the value of glycemic variability（GV） in predicting early enteral feeding intolerance（EFI） in critically ill patients，and provide a basis for precision nutrition management. Methods The 283 patients who initiated enteral nutrition in the ICU between June 2022 and June 2024 were enrolled. Patients were divided into a tolerance group and an intolerance group based on the occurrence of enteral feeding intolerance（EFI）. Data collected included general patient characteristics，laboratory parameters，clinical treatments，and blood glucose values，with glycemic variability（GV） calculated. Logistic regression and Cox proportional hazards regression analyses were used to identify independent risk factors for EFI，while receiver operating characteristic （ROC） curves was employed to evaluate the predictive efficacy of GV for EFI. Results Among the 283 patients，156 tolerated enteral feeding，while 127 developed EFI. The incidence of EFI was 44.87%，with primary manifestations of gastric retention（50.39%） and diarrhea（36.22%），mostly occurring between days 2 and 3 after nutrition initiation. Logistic and Cox regression indicated that pre-nutrition glucose standard deviation（GLUsd）（β=0.218，OR=1.244） and APACHE Ⅱ score（β=0.078，OR=1.081） were independent risk factors for EFI. ROC analysis showed that the AUCs for APACHE Ⅱ score and pre-nutrition GLUsd were 0.696 and 0.657，respectively，increasing to 0.723 when combined. Correspondingly，accuracy improved from 67.137% and 63.984% to 70.115%，with more balanced sensitivity and specificity. GV indicators during nutrition initiation did not show significant predictive value. Conclusion APACHE Ⅱ score and pre-nutrition GLUsd are both significant predictors of early enteral feeding intolerance，and their combination further enhances diagnostic efficacy. Clinically，the dual-dimensional assessment of metabolic stability and disease severity should be emphasized prior to nutrition support，in order to optimize nutritional management，reduce the incidence of EFI，and improve outcomes in critically ill patients.