脑脊液乳酸含量联合细胞学指标对神经外科手术后细菌性脑膜炎的早期诊断价值

Diagnosis of post-neurosurgical bacterial meningitis by the models composed of lactate levels and cytological parameters in cerebrospinal fluid

目的探索脑脊液乳酸含量(cerebrospinal fluid lactate,CSF-Lac)联合细胞学指标所建之模型对于神经外科手术后的细菌性脑膜炎(post-neurosurgical bacterial meningitis,PNBM)早期诊断的价值。方法选取2019年1—12月于神经外科有创操作术后连续转入首都医科大学附属北京天坛医院重症医学科的成年患者(≥18岁)组成研究队列;采用巢式病例对照的研究方法,并按相关标准从中分别筛选出61例PNBM患者(病例组)及神经外科有创操作术后无菌性脑膜炎的患者(对照组)122例;对比两组患者的相关脑脊液指标,包括:脑脊液的白细胞计数(cerebrospinal fluid white blood cell,CSF-WBC)、多核细胞比例(cerebrospinal fluid neutrophil percentage,CSF-Neu%)、蛋白含量(cerebrospinal fluid protein,CSF-Pro)、葡萄糖糖含量(cerebrospinal fluid glucose,CSF-Glu)与CSF-Lac。采用logistic回归的方法将上述脑脊液指标拟合成相应的多参数及双参数诊断模型,同时通过受试者工作特征曲线分析各单项指标及模型在诊断PNBM时的效能,利用Z检验分别比较各个诊断模型之间的AUC值。结果CSF-WBC、CSF-Neu%、CSF-Lac、CSF-Pro与CSF-Glu独立诊断PNBM的AUC值分别为0.72、0.72、0.79、0.66、0.67,多参数模型“0.0002×CSF-WBC+0.0410×CSF-Neu%-0.2436×CSF-Glu+0.1941×CSF-Lac-4.6996”的AUC值则可提高至0.86,且灵敏度、特异度均≥0.80(以0.30为最佳临界值),CSF-Lac+CSF-WBC或CSF-Neu%拟合而成的双参数模型“0.0002×CSF-WBC+0.1869×CSF-Lac-2.3252与0.0410×CSF-Neu%+0.2439×CSF-Lac-5.3468”的AUC值则分别为0.84和0.83,它们的灵敏度、特异度亦均不小于0.80(分别以0.31和0.30为最佳临界值);Z检验的结果则提示上述多参数与两个双参数模型之间两两比较AUC值的差异无显著性,而两个双参数模型的AUC值比较差异无显著性(Z=1.58,P=0.22)。结论脑脊液的双参数诊断模型与多参数模型在PNBM早期诊断的效能方面是相近的,可以考虑其在临床的进一步应用。

Objective To explore the values of the models built of cerebrospinal fluid lactate(CSF-Lac)and its cytological parameters in the early diagnosis of post-neurosurgical bacterial meningitis(PNBM).Method The patients aged≥18 who had underwent invasive neurosurgical procedures and were ransferred to the department of intensive care of Beijing Tiantan Hospital from January to December,2019 were consecutively screened.We conducted a nested case-control study,and those suffering from PNBM(case group,n=61)and post-neurosurgical aseptic meningitis(control group,n=122)were selected from the above cohort.All the diagnosis of PNBM were based on relevant guidelines.Next,the counts of CSF leukocyte(CSF-WBC),and the levels of neutrophil percentages(CSF-Neu%),CSF-Lac,CSF protein(CSF-Pro)and CSF glucose(CSF-Glu)were compared between the 2 groups,respectively.Multivariate logistic regressions were then applied to establish the models formed of all the above CSF parameters,and the receiver operating characteristic(ROC)curves were drawn for each CSF parameter and those clinical models to evaluate their diagnostic power.Also,the areas under the ROC curves(AUC)for those clinical models were compared with each other using Z tests.Result The AUC values for CSF-WBC,CSF-Neu%,CSFLac,CSF-Pro and CSF-Glu were 0.72,0.72,0.79,0.66 and 0.67 respectively,while the model“0.0002×CSF-WBC+0.0410×CSF-Neu%-0.2436×CSF-Glu+0.1941×CSF-Lac-4.6996”carried an AUC level of 0.86,with its sensitivity and specificity higher than 0.80(using the optimal cutoff at 0.30).The respective AUC levels for the models“0.0002×CSF-WBC+0.1869×CSF-Lac-2.3252 and 0.0410×CSFNeu%+0.2439×CSF-Lac-5.3468”were 0.84 and 0.83,and either of their sensitivities and specificities were less than 0.80(at the optimal cutoff values of 0.31 and 0.30,respectively).The results of Z tests show no significant difference for the pairwise comparisons of the above multivariate model and those two-parameter models,and no markable difference were observed between both the two-parameter models(Z=1.58,P=0.22),either.Conclusion The two-parameter models composed of“CSF-Lac plus CSF-WBC or CSF-Neu%”have quite similar diagnostic power with the multivariate model of CSF parameters,and should be considered for clinical usage.