荟萃分析血清可溶性白细胞分化抗原14亚型(sCD14-ST，presepsin)检测对新生儿败血症(NS)的诊断价值。

采用Cochrane系统评价方法，应用计算机检索PubMed、Web of Science等英文数据库，以及中国知网、万方数据知识服务平台等中文数据库中，有关血清presepsin检测对NS诊断价值的前瞻性或回顾性病例对照研究文献。检索时间设定为1990年1月至2020年9月。按照本研究设定的检索策略进行文献检索、筛选、提取，采用诊断准确性研究质量评价工具(QUADAS-2)进行文献质量评价，采用Meta-Disc 1.4及Stata 14.0软件进行Meta分析。血清presepsin检测诊断NS的主要结局指标为合并敏感度、特异度、阳性似然比(PLR)、阴性似然比(NLR)、诊断比值比(DOR)，汇总受试者工作特征(SROC)曲线及其曲线下面积(AUC)。采用血清presepsin检测诊断NS的SROC曲线分布特征及其敏感度对数与(1-特异度)对数的Spearman相关系数，评价阈值效应；采用Cochrane-Q检验分析纳入研究文献的异质性，并采用亚组分析进一步探讨该异质性来源。绘制Deek′s漏斗图，分析文献发表偏倚。

对最终纳入的15篇文献中血清presepsin检测诊断NS价值研究结果如下。①共计纳入6个国家的1 394例新生儿，研究组为NS患儿(n＝812)，对照组为非NS患儿(n＝582)。②本研究纳入文献偏倚风险均在可接受范围，NS诊断标准适用性高，符合QUADAS-2质量评价标准。③绘制纳入15篇文献中，血清presepsin检测诊断NS的SROC曲线结果显示，SROC曲线未呈典型"肩臂状"分布，诊断NS敏感度对数与(1-特异度)对数的Spearman相关系数为-0.020(P＝0.945)，表明研究间不存在阈值效应。血清presepsin检测诊断NS敏感度、特异度、PLR、NLR、DOR，均存在高度异质性(I²＝81.0%、75.8%、67.2%、76.3%、68.7%，P均<0.001)，提示纳入研究文献间存在非阈值效应引起的异质性，故采用随机效应模型进行Meta分析。血清presepsin检测诊断NS的合并敏感度为89%(95%CI：87%~91%，P<0.001)，合并特异度为90%(95%CI：88%~93%，P<0.001)，合并PLR为8.39(95%CI：5.18~13.60，P<0.001)，合并NLR为0.13(95%CI：0.09~0.21，P<0.001)，合并DOR为98.83(95%CI：43.21~226.07，P<0.001)，SROC-AUC为0.96(95%CI：0.95~0.98)。亚组分析结果显示，纳入研究文献异质性来源，可能为血清presepsin检测方法[酶联免疫吸附测定(ELISA)：血清presepsin检测诊断NS的PLR为12.10(6.31~23.17)，I²＝42.9%、P＝0.105，DOR为270.76(84.62~866.32)，I²＝40.5%、P＝0.121；化学发光酶联免疫分析(CLEIA)：血清presepsin检测诊断NS的PLR为5.80(3.27~10.28)，I²＝65.9%、P＝0.005，DOR为45.47(17.5~118.15)，I²＝69.0%、P＝0.002]。④Deek′s漏斗图显示，纳入15篇文献基本对称分布于回归线2侧，并且差异无统计学意义(P＝0.640)。

本研究纳入15篇文献的血清presepsin检测对NS患儿具有较高诊断价值，使用ELISA法检测血清presepsin诊断NS的结果更为准确、可信。受纳入文献的研究对象和研究质量限制，上述结论尚待更多高质量研究予以证实。

To explore diagnostic values of serum soluble CD14 subtype (sCD14-ST, presepsin) in neonatal sepsis (NS) by Meta-analysis.

Prospective or retrospective case-control studies literature on diagnostic value of serum presepsin in NS in English databases such as PubMed and Web of Science, as well as Chinese databases such as CNKI and Wanfang database were searched by computer and Cochrane system evaluation method. Literature retrieval time was set from January 1990 to September 2020. According to retrieval strategies set in this study, researchers searched and screened literature and extracted data, and evaluated quality of literature by Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2), and then used Meta-Disc 1.4 and Stata 14.0 software for Meta-analysis. The main outcome indicators of serum presepsin in diagnosis of NS were combined sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR), summary receiver operating characteristics (SROC) curve and area under curve (AUC). Threshold effect was evaluated by distribution characteristics of SROC curve and Spearman correlation coefficient between sensitivity logarithm and (1-specificity) logarithm of serum presepsin in diagnosis of NS. Heterogeneity between studies was analyzed by Cochrane-Q test, and the source of heterogeneity was further explored by subgroup analysis. Publication bias was analyzed by Deek′s funnel plot.

A total of 15 pieces of literature were finally included, and results of study on value of serum presepsin in diagnosis of NS were as follows. ①A total of 1 394 newborns from 6 countries were included with 812 cases of NS children in study group and 582 cases of non-NS children in control group. ②The 15 pieces of literature met the criteria of diagnostic accuracy quality evaluation of QUADAS-2 with acceptable bias risk and applicable NS diagnostic standard. ③Results of SROC curve of serum presepsin in diagnosis of NS in 15 pieces of included literature showed that SROC curve did not show a typical " shoulder-arm" distribution, and Spearman correlation coefficient between sensitivity logarithm and (1-specificity) logarithm of serum presepsin in diagnosis of NS was -0.020 (P=0.945), indicating that there was no threshold effect among 15 pieces of literature. The sensitivity, specificity, PLR, NLR and DOR value of serum presepsin in diagnosis of NS were highly heterogeneous (I²=81.0%, 75.8%, 67.2%, 76.3%, 68.7%; all P<0.001), suggesting that there was heterogeneity caused by non-threshold effect among 15 pieces of literature. So, random effect model was used for Meta-analysis. Combined sensitivity of serum presepsin in diagnosis of NS was 89% (95%CI: 87%-91%, P<0.001), combined specificity was 90% (95%CI: 88%-93%, P<0.001), combined PLR was 8.39 (95%CI: 5.18-13.60, P<0.001), combined NLR was 0.13 (95%CI: 0.09-0.21, P<0.001), and combined DOR value was 98.83 (95%CI: 43.21-226.07, P<0.001), SROC-AUC was 0.96 (95%CI: 0.95-0.98). Subgroup analysis showed that non-threshold effect heterogeneity among literature may be caused by serum presepsin detection method [enzyme-linked immunosorbent assay (ELISA) method: PLR of serum presepsin in diagnosis of NS was 12.10 (6.31-23.17), I²=42.9%, P=0.105, DOR value was 270.76 (84.62-866.32), I²=40.5%, P=0.121; chemiluminescence enzyme immunoassay (CLEIA) method: PLR was 5.80 (3.27-10.28), I²=65.9%, P=0.005, DOR value was 45.47 (17.5-118.15), I²=69.0%, P=0.002]. ④Deek′s funnel plot showed that 15 pieces of literature were basically symmetrically distributed on 2 sides of the regression line, and the difference was not statistically significant (P=0.640).

Diagnostic value of serum presepsin in NS is relatively high in 15 pieces of literature, and detection of serum presepsin by ELISA is more accurate and reliable in diagnosis of NS. Limited by the subjects and quality of included literature, the above conclusions need to be confirmed by more high-quality studies.