摘要目的:评价降钙素原（procalcitonin, PCT）血清浓度鉴别诊断血流感染（bloodstream infection, BSI）病原学的准确性。方法:收集2015年1月至2020年6月于中日友好医院重症监护室（ICU）诊断为BSI且阳性血培养同时进行PCT检测的患者资料，以血培养当天的参数计算序贯器官衰竭评分（sequential organ failure assessment, SOFA），比较不同病原体感染者之间各指标的差异，并用受试者工作特征（receiver operating characteristic, ROC）曲线分析生物标志物PCT对单一病原体感染的鉴别诊断价值。结果:在1 456例BSI患者中，单一细菌或念珠菌感染分别为1261例（86.6%）和80例（5.5%），混合感染115例（7.9%）；28 d病死率为24.5%（356/1 456），60 d病死率为30.6%（446/1 456）；无论28 d还是60 d病死率，混合感染组均明显高于细菌菌血症组以及念珠菌血症组。PCT在革兰阴性菌（gram-negative bacteria, GNB）菌血症中显著高于革兰阳性菌（gram-positive bacteria, GPB）菌血症和念珠菌血症｛3.4 μg/L [95%可信区间（95% CI）0.7~17.0 μg/L]比1.3 μg/L（95% CI 0.4~7.3 μg/L）；3.4 μg/L（95% CI 0.7~17.0 μg/L）比1.1 μg/L（95% CI 0.4~3.4 μg/L）； P<0.01｝。ROC曲线分析显示：①PCT鉴别诊断单一细菌菌血症与念珠菌血症的最佳截断值为7.25 μg/L，特异性可达90.0%，ROC曲线下面积（area under ROC curve, AUROC）为0.612（95% CI 0.533~0.691）；当PCT>0.51 μg/L，诊断细菌菌血症的敏感性可达73.3%；②PCT鉴别诊断GNB菌血症与念珠菌血症的最佳截断值为7.32μg/L，特异性为90.0%；AUROC为0.695（95% CI 0.614~0.776）；当PCT>0.51 μg/L时，诊断GNB菌血症的敏感性为84.9%；③PCT鉴别诊断GNB菌血症与GPB菌血症的最佳截断值为0.52 μg/L，敏感性为84.9%，AUROC为0.713（95% CI 0.672~0.755）；当PCT>7.36 μg/L，诊断GNB菌血症的特异性可达80.1%。 结论:PCT能够提供BSI患者可能的病原学方面的额外信息，尤其是较高的PCT水平往往提示GNB菌血症可能。

abstractsObjective:To evaluate the accuracy of serum concentration of procalcitonin (PCT) in differential diagnosis of the etiology of bloodstream infections (BSI).Methods:Patients hospitalized in ICU of China-Japan Friendship Hospital from January 2015 to June 2020 with BSI and with PCT test simultaneously when blood drawing for blood culture were enrolled. Sequential Organ Failure Assessment (SOFA) were calculated based on parameters on the day of blood culture. Difference of various indicators among different pathogen infections were compared. Receiver Operating Characteristic (ROC) Curve was used to analyze the value of PCT in differential diagnosis of BSI by different pathogens.Results:Among 1 456 patients with BSI,1 261 (86.6%) patients with monobacterial infection, 80 (5.5%) patients with candidiasis and 115 (7.9%) patients with mixed infection. The 28-day mortality was 24.5% (356/1 456) and the 60-day mortality was 30.6% (446/1456). Mortality of both 28-day and 60-day in the mixed group was significantly higher than that in the bacteriacemia group and candidemia group. PCT levels was significantly higher in patients with bacteremia caused by gram-negative bacteria (GNB) than that in gram-positive bacteria (GPB) infected bacteremia and candidemia {3.4 μg/L[95% confidence interval (95% CI) 0.7-17.0 μg/L] vs 1.3 μg/L (95% CI 0.4-7.3 μg/L); 3.4μg/L (95% CI was 0.7-17.0 μg/L) vs 1.1 μg/L (95% CI was 0.4-3.4 μg/L); P<0.01} . ROC curve analysis showed that: ① the optimal cut-off value of PCT in differential diagnosis of monobacterial bacteremia and candidemia was 7.25 μg/L, with specificity of 90.0% and the area under the ROC curve (AUROC) was 0.612 (95% CI 0.533-0.691). When PCT value was greater than 0.51 μg/L, the sensitivity of diagnostic of bacteremia could reach 73.3%. ② the optimal cut-off value of PCT in differential diagnosis of bacteremia caused by GNB infection and candidemia was 7.32 μg/L, with specificity of 90.0% and AUROC was 0.695 (95% CI 0.614-0.776). When PCT value was greater than 0.51 μg/L, the sensitivity of diagnostic of bacteremia caused by GNB infection was 84.9%.③ the optimal cut-off value of PCT in differential diagnosis of bacteremia caused by GNB and GPB infection was 0.52 μg/L, with sensitivity of 84.9% and AUROC was 0.713 (95% CI 0.672-0.755). When PCT value was greater than 7.36 μg/L, the specificity of diagnostic of bacteremia caused by GNB infection could reach 80.1%. Conclusions:PCT can provide additional information about the possible etiology of patients with BSI, especially as high levels often indicate the possibility of GNB bacteremia.