Abstract: Background　Diabetic ketoacidosis（DKA） is a serious complication of diabetes that can lead to death.Many studies have shown that the most common inducement of DKA is bacterial infection.Objective　This study aims to assess the predictive value of neutrophil / lymphocyte ratio（NLR） and procalcitonin（PCT） in DKA with bacterial infection.Methods　Totally 127 cases of DKA admitted to Department of Endocrinology of Bazhong Central Hospital from January 2015 to January 2020 were retrospectively enrolled，including 64 with bacterial infection and 63 without.Results of bacterial culture and laboratory test were analyzed.NLR，PCT and the combination of the two in the diagnosis of DKA with bacterial infection were evaluated by receiver operating characteristic （ROC） curve analysis.Results　The bacterial culture test showed that the prevalence of Staphylococcus aureus infections，Escherichia coli infections，Klebsiella pneumoniae infections，Pseudomonas aeruginosa infections，Haemophilus influenzae infections，and other bacterial infection was 18.8%（12/64），15.6%（10/64），12.5%（8/64），4.7%（3/64），and 32.8%（21/64），respectively.Those with bacterial infection had higher average levels of fasting blood glucose，NLR and PCT but lower average levels of arterial blood pH and serum HCO3- than those without（P<0.05）.In predicting DKA with bacterial infection，the AUC of NLR was 0.707〔95%CI（0.617，0.797）〕when the optimal cutoff value was selected as 12.94 μg/L，with 59.4% sensitivity and 76.2% specificity.And that of PCT was 0.715〔95%CI（0.624，0.805）〕when the optimal cutoff value was selected as 2.57 μg/L，with 46.9% sensitivity and 93.7% specificity.The AUC of NLR with PCT developed based on the equation of binary logistic regression ｛PRE（P）：Ln〔P/（1-P）〕=-1.235+0.046×NLR+0.335×PCT｝was 0.761〔95%CI（0.680，0.843）〕 when the optimal cutoff value was selected as 0.458，with 60.9% sensitivity and 82.5% specificity.Conclusion　Either NLR or PCT can be used as a clinical predictor for DKA with bacterial infection，and the combination of the two is more valuable.

Key words: Diabetic ketoacidosis, Bacterial infections, Neutrophil/lymphocyte ratio, Procalcitonin, Diagnosis, Forecasting

摘要： 背景　糖尿病酮症酸中毒（DKA）是一种可导致死亡的糖尿病急性并发症，国内外研究显示其最常见的诱因为感染。目的　探讨中性粒细胞/淋巴细胞比值（NLR）及降钙素原（PCT）对DKA患者合并细菌感染的预测价值。方法　回顾性分析2015-01-01至2020-01-01巴中市中心医院内分泌科收治的127例DKA住院患者的临床资料，其中合并细菌感染患者64例（DKA合并细菌感染组）、未合并细菌感染患者63例（DKA无细菌感染组）。分析DKA合并细菌感染组患者的细菌培养结果。比较两组患者实验室检查结果，绘制NLR、PCT及二者联合预测DKA合并细菌感染的受试者工作特征（ROC）曲线。结果　64例DKA合并细菌感染患者中，金黄色葡萄球菌感染12例（18.8%），大肠埃希菌感染10例（15.6%），肺炎克雷伯菌感染10例（15.6%），铜绿假单胞菌感染8例（12.5%），流感嗜血杆菌感染3例（4.7%），其他类型细菌感染21例（32.8%）。DKA合并细菌感染组患者血糖、NLR、PCT高于DKA无细菌感染组，动脉血pH值、血清HCO3-低于DKA无细菌感染组（P<0.05）。NLR、PCT单独预测DKA患者合并细菌感染的ROC曲线下面积（AUC）分别为0.707〔95%CI（0.617，0.797）〕、0.715〔95%CI（0.624，0.805）〕，最佳截断值分别为12.94、2.57 μg/L，灵敏度分别为59.4%、46.9%，特异度分别为76.2%、93.7%。采用二元Logistic回归分析建立NLR联合PCT检测的回归方程PRE（P）：Ln〔P/（1-P）〕=-1.235+0.046×NLR+0.335×PCT，使用该模型预测DKA患者合并细菌感染的AUC为0.761〔95%CI（0.680，0.843）〕，最佳截断值为0.458，灵敏度为60.9%，特异度为82.5%。结论　NLR、PCT均可作为单独预测DKA患者合并细菌感染的临床指标，且二者联合对DKA患者合并细菌感染的预测效能更高。

关键词: 糖尿病酮症酸中毒, 细菌感染, 中性粒细胞/淋巴细胞比值, 降钙素原, 诊断, 预测