Background
Central venous pressure (CVP) is the pressure recorded in the superior vena cava or the right atrium and, to a lesser extent, the left ventricular preload [
1], reflecting venous return and right ventricular function [
2]. Thus, CVP measurements could help in fluid management [
3]. However, the CVP can also be influenced by thoracic, pericardial and abdominal pressures, complicating its understanding. Recently, several studies have found that the CVP cannot predict fluid responsiveness and only higher or lower CVP values may have negative or positive predictive values [
4‐
6]. Therefore, the application of CVP measurement has been questioned. Some researchers have suggested that although there could be some limitations to using the CVP to assist fluid resuscitation, it would be better to fully understand and address these limitations, rather than abandoning them completely [
3,
7]. The CVP levels still provide significant information about patient's cardiocirculatory status.
Many clinical studies have reported that higher CVP values may be associated with patients’ poor outcomes in different conditions, such as patients undergoing cardiac surgery [
11], early Fontan failure [
10], acute kidney injury (AKI) [
8,
9], patients undergoing cardiopulmonary bypass surgery [
9], and critically ill patients [
2]. A controlled lower CVP can also reduce blood loss during hepatectomy [
12]. Meanwhile, CVP measurement has been reported to shorten the time to be medically fit for discharge and benefit the clinical outcomes for patients with acute respiratory distress syndrome or sepsis [
13‐
15]. Some researchers have suggested using the CVP as a stopping rule for fluid infusion [
3,
7]. However, to our knowledge, the value of CVP measurements in patients with coma remains unclear. Fluid administration could increase cardiac output; however, it could also increase the risk of hydrostatic pressure or edema formation. Theoretically, this situation could be more complicated in critically ill comatose patients, and CVP may be beneficial in these conditions.
Therefore, the purposes of this study were to investigate 1) the association between early CVP measurement and patients’ outcomes and 2) the association between initial CVP time and all-cause mortality of intensive care unit (ICU) patients with severe coma.
Methods
Study design and population
Data were collected from patients in the ICU Collaborative Research Database (eICU-CRD) [
16] and the Medical Information Mart for Intensive Care III (MIMIC-III) [
17]. Patient characteristics available in both databases included demographic information, vital signs, laboratory test results, diagnoses, medical histories, treatments and so on. Detailed information on these two databases can be found in PhysioNet [
18].
Patients with severe coma were identified using the Glasgow Coma Scale (GCS). The GCS was first published in 1975 and is the most widely applied outcome measure in clinical settings and researches on brain injury [
19,
20]. Currently, the GCS is recommended by lots of countries to use as a measurement for primary outcome. A severe coma is defined as a score of 3–8 [
21]. We only collected the GCS scores before sedative drug use on the first day for each patient. For patients with multi-ICU admissions, we only include the first admission. Patients aged < 18 years or ≥ 90 years were excluded. An early CVP measurement was defined as a CVP measurement in the first 24 h after first ICU admission. According to early CVP, the entire cohort was divided into CVP and non-CVP groups. The initial CVP time and CVP values were also collected. All comorbidities were identified based on recorded ICD-9 or APACHE components. We only collected the first value within 24 h of ICU admission for vital signs and laboratory test results.
In clinical practice, there may be many reasons for coma development. Due to the nature of retrospective research, we cannot accurately determine the direct cause of coma in patients. Therefore, we collected the primary diagnoses of patients and classified them according to ICD-9 for different diseases, such as circulatory disease and injury disease.
Study outcome
All-cause in-hospital mortality rate was set as the primary outcome. The secondary outcomes included 28-day mortality; 365-day mortality; ICU mortality; MV-free days; vasopressor-free days within 28 days after ICU admission; intravenous fluid (IVF) volumes (ml) of the first, second, and third days after ICU admission; and incidences of AKI within 2 days and 7 days after ICU admission [
22].
Statistics analysis
Continuous variables are illustrated as mean ± standard deviation (SD). Categorical variables are presented as total number and percentage. The Chi-square test was applied to compare proportions, and the T-test or Wilcoxon rank-sum test was used for continuous variables. Early CVP was analyzed as a categorical variable for the primary analysis in the whole cohort, and the CVP value was analyzed as a continuous variable in the CVP group. Multivariable Cox regression was used for all outcomes to adjust for confounders that may affect the outcomes. Variables with a P value of < 0.2 in the univariable analysis were enrolled in the multivariable analysis.
We used several sensitivity analyses to ensure the consistency of the findings, including propensity score matching (PSM), inverse probability treatment weighting (IPTW), and stabilized IPTW (sIPTW). Patients’ propensity score was assessed using a multivariable logistic analysis to minimize the covariate imbalance between the CVP and non-CVP groups. To compare secondary outcomes, the standardized mean differences and statistical significance of parameters between the CVP and non-CVP groups in the PSM cohort were calculated using the Chi-square test or t-test. We also used restricted cubic spline (RCS) curves based on multivariable Cox regression to evaluate the relationship between the initial CVP time or value and the primary outcome. All statistical analyses were performed using R software.
Discussion
The goal of current study was to investigate the application value of CVP in patients with severe coma in the ICU. We found that early CVP was independently associated with lower all-cause in-hospital mortality in severe coma patients (GCS score 3–8). We also found the association of the initial CVP time and in-hospital mortality was presented as a “U-shape” in the RCS curve. As the initial CVP time increases, the risk of all-cause in-hospital mortality increases.
Because of the standardization of advances in technologies, such as MVs, in ICU management, more and more comatose patients are surviving life-threatening diseases. However, the high mortality rate and cost for this type of patient are still among the most severe clinical and social problems [
23,
24]. Any efforts that could increase the chances of reasonable functional outcomes in patients with coma are of utmost clinical and ethical significance [
25].
This is the first study to investigate the association between the early CVP and clinical outcomes in patients with severe coma. The current study was a post hoc analysis. The data included in this study were from the eICU and MIMIC-III databases, which contain ICU patients across more than 200 different hospitals in the USA, with varying CVP practices. Our analysis leverages the availability of time-stamped demographic information, comorbidities, vital signs, laboratory results, etc., to investigate whether CVP could benefit patients with severe coma. The results of the study were highly consistent across different sensitivity analyses.
Conflicting evidence on the application of the CVP has been reported in previous studies. Eskesen et al. found that the application value of the CVP measurement is low in a re-analysis of 1148 patient data sets [
5]. They reported that only in certain extreme cases, such as higher or lower, does CVP measurement demonstrate some of its value in liquid management. Marik et al. conducted a meta-analysis and reported that only 57% of patients were fluid responders, and the correlation between CVP level and stroke volume index was not strong; thus, the use of the CVP for fluid resuscitation should be abandoned in widespread practice due to insufficient evidence. However, many studies have reported the association between the CVP levels and patients’ clinical outcomes with various conditions. For example, Chen et al. found in a meta-analysis that a higher CVP value was associated with higher mortality (odds ratio [OR] 1.65) and a greater risk of AKI (OR, 2.09). Quail et al. found that a higher CVP is associated with a higher risk of early Fontan failure after total cavopulmonary connections [
10]. Liu et al. also found that a controlled lower CVP could reduce blood loss during hepatectomy and recommended the promotion of CVP use in clinical settings [
12].
Despite the conflicting evidence reported and the value of the CVP being questioned, some clinicians still recommend the use of the CVP in clinical settings [
3,
7]. One recommendation is to use CVP as a stop symbol in liquid management. Weil and Henning first proposed this approach and suggested that fluid administration could be stop as the CVP increased by 5 cmH2O or more [
26]. However, Hamzaoui et al. reported that changes in the stroke volume could not be reflected by the CVP levels. Therefore, the CVP levels could not represent as a criterion for predicting fluid responsiveness [
6].
Whatever the approach to CVP application, some evidence also indicates that CVP use is associated with a better clinical outcome in some patients [
14,
15]. Chen et al. reported the association between the CVP using and lower 28-day mortality in septic patients [
14]. Tang et al. reported that the CVP could improve the clinical outcomes of patients with acute respiratory distress syndrome [
15]. However, both studies were from a single-center database; thus, the generalization of the results could be hindered. In these multicenter studies, we found that severe comas patients with CVP suing have a lower in-hospital mortality than those without CVP. Moreover, we found that with the delay in the initial CVP time, the risk of in-hospital mortality is increasing. In the secondary outcomes, we found that CVP was also associated with ICU mortality, 28-day mortality, 365-day mortality, and more days without vasopressor drugs. We believe that these results confirm the application value of CVP in comatose patients.
The detailed influence of the CVP on therapeutic interventions was difficult to explore due to the retrospective design. Previous studies reported that the trigger caused by CVP, such as fluid therapy, could lower lactate levels of patients, thus could improve outcomes [
14,
15]. Semler et al. found that different fluid management based on initial CVP values could influence outcomes of acute respiratory distress syndrome patients, and conservative fluid administration could lower the mortality of patients with a low initial CVP [
27]. Wang et al. found that patients who have a peak CVP value more than 12 mmHg also have worse organ function, higher SOFA score and longer hospital stay [
2]. These results indicate that CVP could positively impact the latter management of patients. In the current study, we found that the IVF was significantly different between patients with and -out CVP. Generally, IVF volumes in patients with CVP were higher on day 1 and lower on day 3 compared to those in the non-CVP group. Most patients (62.1%) with CVP had an initial CVP value of ≤ 12 mmH2O. It is difficult to investigate whether the difference in the IVF on the first day was influenced by the CVP. However, relatively free fluid management may be safe when the levels maintaining as low as possible [
3]. Fluid management is a double-edged sword. Correction of intravascular hypovolemia is critical for preventing and managing AKI, but excessive fluid administration could also increases the risk of AKI [
28,
29]. In the current study, we also discovered that the 2-day and 7-day AKI was higher in patients with CVP, which is different from that in previous studies. The causal relationship between the increased risk of AKI and the CVP requires further research and evaluation.
Limitations
Our study had some limitations. First, although the GCS has been used for more than 40 years and is the most widespread measurement in brain injury studies [
20], using the GCS score of 3–8 to define severe coma in patients in the ICU could still introduce patient selection bias. Meanwhile, all data in this study were extracted using a structured query language, which may also cause misclassification of patients. Second, considering that the significant covariate imbalance between the CVP and non-CVP groups may influence the explanation of the results, we used a PSM approach to minimize the differences and performed several sensitivity analyses. However, covariate imbalance was still observed for some variables. Third, although we included patients from two large databases from more than 200 hospitals to improve the generalization of the results, the retrospective design may still introduce some analysis bias. Fourth, because CVP values are generally measured and read manually, measurement and reading errors cannot be avoided. Finally, CVP-related management changes are complex in clinical practice. The nature of post hoc analysis of the current study should not be ignored. Although we found a robust association between CVP and the primary outcome, the causal relationship of this condition remains unclear and should be investigated in further studies.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.