Characteristics of the Cerebrospinal Fluid in Septic Patients with Critical Illness Polyneuropathy - A Retrospective Cohort Study

The institutional ethics committee of the General Hospital of the People's Liberation Army approved this single-center retrospective cohort study (S2018–718–02). In total, 175 septic patients were admitted to our Critical Care Medicine department from March 2018 to July 2022, including 94 male and 81 female patients. The process of patient enrollment is shown in Figure 1. The ages of the patients ranged from 18 to 80 years. The inclusion criteria were as follows: (1) a diagnosis of sepsis based on the latest guidelines (sepsis 3.0) [7]; (2) age ≥18 years and ≤80 years; and (3) sepsis caused by abdominal infection, intestinal puncture, obstruction, fistulation or traumatic injury, biliary system infection, pancreatitis infection, or intestinal flora dysfunction. The exclusion criteria were as follows: (1) patients with a history of nervous system diseases, such as Guillain-Barre syndrome, multiple sclerosis, epilepsy, and cerebral infarction; (2) toxic injury, which can impair nerve conduction, including lead, mercury, arsenic, thallium, alcohol, barbital poisoning, bacterial toxin, and animal toxin, such as diphtherin, tetanus, and tetrodotoxin; (3) malnutrition, especially vitamin B deficiency; (4) usage of muscle relaxant; (5) deep sedation; (6) severe limb injury preventing electrophysiology examinations; (7) survival time less than eight days without electrophysiology examination; (8) intracranial infection, such as bacterial meningitis and other intracranial infections; and (9) demyelinating diseases or neuromuscular junction dysfunction.

Fig. 1.

The following demographic and clinical data were collected from each patient on the day of sepsis onset: age; gender; diagnosis; Acute Physiology and Chronic Health Evaluation (APACHE II) score; Sequential Organ Failure Assessment (SOFA) score; blood lactate; blood sugar; albumin; white blood cell (WBC); body mass index (BMI); hemoglobin (Hb), and C responsive protein (CRP). The information on ICU length of stay (LOS), hospital LOS, and mechanical ventilation (MV) were collected from electronic medical records. For each patient, electromyography data were collected on the eighth day of the diagnosis of sepsis. CSF was collected and examined two to four days after sepsis onset. The levels of protein, WBC, interleukin (IL)-1, IL-6, IL-8, and tumor necrosis factor (TNF) α in CSF were measured.

All examinations were performed by a physician expert in electromyography and an assistant using KEY-POINT 7033A (Total functional electromyographic potentiometer, Denmark Dandy Co., Skovlunde,, Denmark). The electromyography examination was carried out on the 8^th day of sepsis onset. CIP was diagnosed according to the definition provided by Bolton (2005) [8]. The CIP diagnosis criteria were as follows: (1) in 2 or more nerves, the wave amplitudes of compound muscle action potential (CMAP) and sensory nerve action potential (SNAP) were at least 80% lower than the normal lower limit in electromyography; and (2) excluding critical illness myopathy (CIM), confirmed by an increase in creatine kinase (CK) and needle electromyography (EMG) showing myogenic abnormality, or other possible diseases except CIP. Median nerves, ulnar nerves, and common peroneal nerves underwent motor and sensory neurophysiological examination. Tibial nerves underwent motor neurophysiological examination. Common peroneal nerves underwent sensory neurophysiological examination. The motor neurophysiological examination included conduction velocity and CMAPamplitude, whereas the sensory neurophysiological examination included conduction velocity and SNAP amplitude.

The R statistical software (version 4.1.2) was used to analyze the data. Normally distributed data were expressed as the mean ±standard deviation (±SD); the differences between groups of normally distributed data were determined using Student's t test. Dichotomous data were expressed as a number (percentage); the differences between groups of dichotomous data were determined using the Chi-square test. The sample size was calculated based on the data of CSF. The values of α and β were set as 0.05 and 0.1, respectively. The power of the test was set as 0.9. A bilateral difference test was conducted. The sample size was not more than 36 for each group after calculation. 59 non-CIP patients and 116 CIP patients were enrolled herein. The sample size of the study met the standard. A multivariate logistical regression model and receiver operating characteristic (ROC) curve were used to analyze the factors associated with CIP. Propensity score matching analysis (PSMA) [9] was conducted to adjust the covariates of SOFA, APACHE II, diabetes history, and lactate. The matched ratio was 1:1 and the method was ‘nearest’. Kaplan-Meier survival curve analysis was used to compare the survival rate within 28 days between the CIP group and the non-CIP group. All results were considered statistically significant at P < 0.05 (two-tailed).

In total, 175 septic patients were included, of which 116 patients (66.3%) were diagnosed as CIP. The differences in age, gender, history of diabetes, cause of sepsis, hemoglobin, WBC, blood sugar, and albumin were not significant between CIP patients and non-CIP patients. Significant differences were recorded in lactate levels (P < 0.01), the SOFA score (P < 0.01), and the APACHE II score (P < 0.01). We also found that the prognosis of CIP patients was worse than that of non-CIP patients. The CIP patients had a longer ICU LOS (P = 0.01), a higher ratio of MV assistance (P = 0.01), and a higher mortality in 28 days (P = 0.02) (Table 1).

Nerve injury was symmetric in CIP patients. Left limbs were selected to perform electromyography. We examined the median nerves, ulnar nerves, tibial nerves, peroneal nerves, and sural nerves in all 175 selected patients. In the upper limb, 163 nerves were diagnosed as abnormal, while in the lower limb, 171 nerves were diagnosed as abnormal (Table 2).

Significant differences were recorded in the lactate level (P < 0.01), SOFA score (P < 0.01), and APACHE II score (P < 0.01) between CIP patients and non-CIP patients (Table 3). PSMA was performed to adjust the influence of lactate, SOFA, and APACHE II. Patients were matched with a 1:1 ratio. Based on PSMA, 59 patients were selected in the CIP group and 59 patients were selected in the non-CIP group. The distribution of propensity scores was shown in Supplementary File 1. Significant differences in lactate, SOFA, and APACHE II were removed by PSMA (P > 0.05). Significant differences in CSF protein, WBC, IL-1, IL-6, IL-8, and TNFα were also recorded between CIP patients and non-CIP patients (Table 3).

Within 28 days after the onset of sepsis, 29 patients (25.0%) died in the CIP group and six patients (10.2%) died in the non-CIP group; the 28-day mortality in the CIP group was higher than that in the non-CIP group (P = 0.02) (Table 1). The percentage of patients receiving mechanical ventilation was higher in the CIP group than in the non-CIP group. Patients in the CIP group had longer ICU LOS (P = 0.01). The difference in hospital LOS between the CIP group and the non-CIP group was not statistically significant (P = 0.08).

To show the difference in survival, the Kaplan-Meier survival curves were plotted for the 28-day survival rate of CIP patients and non-CIP patients (Figure 2). Because the patients who lived less than 8 days were excluded, the observation window was from day 8 to day 28. Our results showed a significant difference in the 28-day mortality between groups (P = 0.02) before PSMA, but not after PSMA (P = 0.4).

Fig. 2.

Conditional logistical regression was used to investigate the best predictor for septic CIP. The results showed that the WBC (P = 0.006), IL-1 (P = 0.009, and TNF-a (P = 0.002) levels in CSF were significantly related to septic CIP. However, no differences in lactate (P = 0.26), SOFA (0.36), and age (P = 0.78) were found (Table 4).

The results of the logistical regression analysis showed that the level of WBC, IL-1, and TNFα in CSF was significantly related to the occurrence of CIP (Table 4). Thus, the correlation between these factors and the prognosis of CIP was analyzed by receiver operator characteristic curve (ROC) analysis. The results showed that these indicators could effectively predict CIP (Figure 3 and Table 5). The combined indicator of WBC, IL-1, and TNFα had the highest AUC of 0.926. Details on the sensitivity, specificity, and cut-off value are provided in Table 5.

Fig. 3.

First, this was a single-center, retrospective study; thus, patient selection bias might exist. However, we had rigid inclusion and exclusion criteria for enrolling participants. These steps helped achieve adequate consistency and decreased the heterogeneity of the study. Additionally, PSMA was used to increase the comparability between the CIP and non-CIP groups, which efficiently decreased the bias of a retrospective cohort study. Second, the sample size was relatively small, which probably affected the strength of the statistical analysis. This was probably why we obtained a negative result and found no difference in the 28-day mortality between the patients in the CIP and non-CIP groups in the survival analysis after PSMA. The sample size of this study met the standard, as determined by statistical calculations. Third, data on pathological biopsy were not used in this study. Nerve biopsy is the gold standard for the diagnosis of CIP. Thus, some non-CIP patients in this study may have been misdiagnosed as CIP. Nerve biopsy is an invasive operation and is difficult to perform. Thus, more time is required to collect the data on enough participants for statistical analysis. To resolve this problem, we recommend a multicenter, prospective study in the future.