Introduction
Arterial hypotension is a well-known complication in critically ill patients, occurring in more than half of cases [1–3]. The etiology of this condition is multifactorial and may involve cardiac dysfunction, peripheral vascular abnormalities, or insufficient circulating blood volume [4, 5]. Nevertheless, contemporary clinical guidelines recommend a uniform approach to initiating treatment for arterial hypotension, beginning with an assessment of fluid responsiveness [1, 6].
Currently, numerous methods have been developed for assessing fluid responsiveness [7]. These include technically demanding approaches such as transpulmonary thermodilution to measure cardiac output [8, 9], transesophageal echocardiography [10], and bioimpedance analysis [11, 12], as well as simpler methods like the passive leg-raising test [13–15] and fluid challenge tests [16, 17]. All these techniques aim to address the key question: will fluid therapy lead to an increase in cardiac output and resolution of hypotension?
Among these methods, the assessment of central venous pressure (CVP) has stood out for decades due to its simplicity and accessibility, attracting significant attention from anesthesiologists and intensivists.
CVP was one of the earliest methods used to evaluate volemic status and determine fluid responsiveness [18]. On the one hand, interpreting CVP is intuitively straightforward; on the other hand, this simplicity often leads to erroneous clinical conclusions [19, 20]. A key issue is the confusion between two concepts: volemic status and fluid responsiveness [7]. While these terms are closely related, they are far from equivalent [21]. For instance, patients with hypovolemia may not always exhibit an increase in cardiac output in response to fluid therapy [7].
The current perspective on CVP is polarized [19, 22]. On the one hand, some studies have demonstrated its diagnostic utility in assessing fluid responsiveness [23–26]. On the other hand, recent large-scale studies have criticized CVP, highlighting its low accuracy [27, 28]. Nevertheless, due to its ease of use, low cost, and long-standing application in clinical practice, CVP remains an appealing and widely utilized method for evaluating the response to fluid challenges [29, 30]. Given the ongoing debate surrounding its utility, we decided to conduct a meta-analysis to clarify the role of CVP in this context.
Objective
The aim of our meta-analysis was to evaluate the diagnostic accuracy of central venous pressure (CVP) in determining fluid responsiveness.
Materials and methods
This study was conducted in accordance with Cochrane Collaboration guidelines and adhered to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) standards for systematic reviews and meta-analyses [31]. The study protocol was registered with the International Platform of Registered Systematic Review and Meta-analysis Protocols (INPLASY) under the registration number INPLASY2024100128 (doi:10.37766/inplasy2024.10.0128). The completed PRISMA-NMA checklist is presented in table D1 in the Supplementary Materials (https://doi.org/10.21320/1818-474X-2025-1-32-47).
Search strategy
A systematic search for scientific articles published up to and including March 2024 was conducted independently by two researchers. The search covered databases such as PubMed (Medline) and the Cochrane Central Register of Controlled Trials (CENTRAL) using the following query: (“fluid responsiveness” OR “fluid resuscitation” OR “volume responsiveness” OR “fluid status” OR “volume status” OR “volemic status”) AND (“intensive care” OR “critical care”) AND (“caval” OR “inferior vena cava” OR “IVC” OR “passive leg raising” OR “PLR” OR “fluid challenge” OR “fluid bolus” OR “central venous pressure” OR “CVP” OR “central venous” OR “pulse pressure variation” OR “stroke volume variation”) AND (“AUC” OR “AUROC” OR “ROC”). In addition, backward citation tracking (analyzing references from selected publications) and forward citation tracking (analyzing citations of selected publications) were performed using the Litmaps platform [32]. No language restrictions were applied during the systematic search.
Eligibility criteria and study selection
After automatic removal of duplicates, two researchers independently screened the abstracts of the remaining studies for eligibility based on the inclusion criteria (table 1).
| Titles (PICOS) | Inclusion Criteria |
|---|---|
| Population | Adult patients |
| Index Test (Method of Investigation) | Measurement of central venous pressure (CVP) |
| Comparator ("Gold Standard") | Fluid challenge to assess fluid responsiveness |
| Outcomes | Area under the ROC curve (AUROC) |
| Study Design | Prospective cohort studies |
The final decision regarding inclusion in the meta-analysis was made after a full-text review of the articles. Studies were excluded if they met any of the following criteria: (1) retrospective design; (2) failure to use the appropriate gold standard (fluid challenge test); (3) lack of central venous pressure (CVP) assessment; or (4) reliance on non-cardiac parameters to evaluate outcomes during the fluid challenge test.
Any disagreements were resolved through consultation with the scientific supervisor until a consensus was reached.
Outcome measures and data extraction
Data were extracted by three researchers, each independently reviewing the texts and supplementary materials of all selected studies. The following data were collected: 1) general publication and patient characteristics: first author, patient cohort, sample size, mean age, sex, body mass index (BMI), APACHE II score, baseline CVP, and type of fluid used; 2) index test and gold standard information: type, method, parameter evaluated, and cutoff value; 3) outcome data: area under the ROC curve (AUROC), sensitivity, and specificity. Based on the volume of fluid administered, the type of "gold standard" was categorized as a standard fluid challenge (≥ 5 mL/kg or > 250 mL) and a mini fluid challenge (< 5 mL/kg or ≤ 250 mL). Following independent data extraction, the researchers compared their forms to identify discrepancies and achieve consensus.
For studies where quantitative data were reported as medians with interquartile ranges or as means with standard deviations (SD), established statistical methods were used to compute the mean and 95% confidence intervals (CIs). These approaches were based on the methodologies proposed by Luo et al. [33] and Wan et al. [34], as well as Cochrane guidelines [35]. When AUROC values were not directly available, they were estimated as the arithmetic mean of sensitivity and specificity (AUROC min) [36,37]. To address missing data on SD or 95% CI for AUROC, multiple imputation was applied using sample size, AUROC, and available SD values in IBM SPSS Statistics for Windows, version 27.0 (Armonk, NY, USA: IBM Corp).
The primary endpoint for this meta-analysis was the AUROC for CVP.
Internal validity and risk of bias assessment
The internal validity and risk of bias of the included studies were assessed by two independent investigators using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool [38]. Publication bias and the small-study effect were assessed using Egger's test and funnel plot analysis [39]. The certainty of the evidence was assessed using the GRADE methodology [40].
Statistical analysis
The meta-analysis was performed using specialized software, STATA 17 (StataCorp LLC, Texas, USA). From each study, data were selected based on the outcome with the highest AUROC value. If AUROC values were equal, the results with the highest sum of sensitivity and specificity were chosen. In cases where this parameter also matched, the result with the highest sensitivity was selected. The meta-analysis yielded pooled estimates of AUROC, sensitivity, and specificity for the defined cutoff points.
Heterogeneity among studies was assessed using Cochran’s Q test and the I² statistic. Significant heterogeneity was considered present if p < 0.05 and/or I² > 50 %. A meta-regression model was constructed to evaluate the influence of covariates such as age, sex, and body mass index on AUROC values. The recommended random-effects model (method: REML, Restricted Maximum Likelihood) was used for pooling results and building the meta-regression model [41]. Subgroup analyses were conducted for patient samples stratified by the following parameters: (1) patient cohort, (2) volume of fluid challenge during the test, and (3) spontaneous breathing versus mechanical ventilation. Additionally, a subgroup analysis compared studies published within the last five years to those published earlier (before 2020).
Statistical significance for hypothesis testing was set at a p-value of 0.05.
Results
Characteristics of the Studies
The initial search yielded 286 studies that met the query criteria, and an additional 150 articles were identified through additional records identified from other sources. After reviewing the abstracts, 318 studies were excluded. Subsequently, the full texts and supplementary materials of the remaining 105 articles were assessed, of which 21 were excluded for meeting at least one exclusion criterion. Ultimately, 84 studies met the inclusion criteria. A flowchart illustrating the study selection process is presented in fig. 1.
Fig. 1. PRISMA — flow-chart for study selection Note: CVP — central venous pressure
This meta-analysis evaluated data from 3.729 patients (4.472 observations), with a mean age ranging from 44 to 75.3 years Among the 84 studies included in the systematic review: two studies involved patients undergoing intraoperative one-lung ventilation [42, 43], thirty-one studies focused on patients in intensive care units (ICUs), including 15 in cardiac ICUs [24, 44–73], fifteen studies involved patients undergoing cardiac surgery [74–88], fourteen studies included patients undergoing non-cardiac surgery [89–102], the remaining studies assessed patients with shock in various ICU settings [25, 26, 103–122]. In 55 studies, colloid solutions were used for fluid challenge tests, crystalloids in 22, mixed solutions in 5, and either crystalloids or colloids in one study [107]. One study used autotransfusion [57]. The characteristics of the studies included in the meta-analysis are summarized in table D3.
Meta-analysis
The pooled AUROC was 0.6 (95% CI [0.57; 0.62]), with individual study results ranging from 0.4 to 0.89 (table 2, fig. 2). Statistical heterogeneity (I²) was 81.09% (p-value < 0.001). The result remained robust in sensitivity analysis using the leave-one-out method (fig. D1). The pooled sensitivity and specificity were 61% and 65%, respectively (table 2).
| Subgroups | N | AUROC (95% CI) | p-value (AUROC) | I2, % | p-value (I2) | p-value (the Egger's test) | Se | Sp |
|---|---|---|---|---|---|---|---|---|
| All studies | 84 | 0.596 (0.575–0.618) | < 0.001 | 81.09 | < 0.001 | 0.696 | 0.61 | 0.65 |
| Patient Cohorts | ||||||||
| Cardiac ICU | 15 | 0.600 (0.521–0.679) | 0.013 | 63.57 | < 0.001 | 0.835 | 0.72 | 0.62 |
| Cardiac Surgery | 15 | 0.583 (0.531–0.636) | 0.002 | 70.58 | < 0.001 | 0.525 | 0.56 | 0.57 |
| Non-Cardiac ICU | 16 | 0.588 (0.549–0.628) | < 0.001 | 69.99 | < 0.001 | 0.07 | 0.63 | 0.66 |
| Non-Cardiac Surgery | 14 | 0.571 (0.530–0.611) | 0.001 | 73.38 | 0.015 | 0.578 | 0.57 | 0.71 |
| One-Lung Ventilation | 2 | 0.574 (0.401–0.746) | 0.402 | 0 | 0.792 | NA | 0.70 | 0.65 |
| Patients with Shock | 22 | 0.623 (0.583–0.663) | < 0.001 | 68.92 | < 0.001 | 0.613 | 0.59 | 0.71 |
| Type of Gold Standard | ||||||||
| Fluid Challenge Test | 64 | 0.580 (0.558-0.603) | < 0.001 | 69.68 | < 0.001 | 0.929 | 0.64 | 0.59 |
| Mini Fluid Challenge Test | 14 | 0.609 (0.560–0.657) | < 0.001 | 51.8 | 0.01 | 0.057 | 0.50 | 0.78 |
| Unclassified | 6 | 0.719 (0.620–0.818) | < 0.001 | 71.94 | 0.011 | 0.805 | 0.61 | 0.87 |
| Type of Breathing | ||||||||
| Mixed | 5 | 0.564 (0.489–0.639) | 0.095 | 34.17 | 0.235 | 0.157 | 0.78 | 0.42 |
| Mechanical Ventilation | 76 | 0.597 (0.574–0.619) | < 0.001 | 76.76 | < 0.001 | 0.699 | 0.59 | 0.67 |
| Spontaneous Breathing | 2 | 0.560 (0.436–0.684) | 0.344 | 71.37 | 0.062 | NA | 0.82 | 0.47 |
| Year of Publication | ||||||||
| Before 2020 | 75 | 0.602 (0.579–0.625) | < 0.001 | 81.59 | < 0.001 | 0.608 | 0.61 | 0.65 |
| 2020-2024 | 9 | 0.554 (0.492–0.617) | 0.090 | 64.89 | 0.019 | 0.534 | 0.62 | 0.64 |
Fig. 2. Forest plot presenting mean AUROC with 95% СI. Reference line: AUROC = 0.5
Subgroup Analysis
The pooled AUROC based on patient cohorts ranged from 0.57 to 0.62 for patients with shock (fig. 3, table 2).
In the subgroup meta-analysis accounting for the volume of fluid challenge (type of gold standard), the pooled AUROC ranged from 0.58 to 0.72, with a value of 0.58 for the standard fluid challenge test. The AUROC for mechanically ventilated patients was 0.60 (95% CI [0.57; 0.62]). For patients with spontaneous breathing, no statistically significant result was obtained (fig. 3, table 2).
Fig. 3. Forest plot presenting subgroup analysis results. Reference line: AUROC = 0.5
Meta-regression
Univariate meta-regression analysis did not reveal a statistically significant effect of age, sex, or BMI on the AUROC value (table D4).
Risk of bias and GRADE assesment
No publication bias was detected in any of the analyses (p-value from Egger's test > 0.05; table 2). Funnel plots are provided in the Supplementary Materials (fig. D2–D6).
Risk of bias assessment for the 84 studies included in the meta-analysis identified 26 studies with low risk, 48 with moderate risk, 9 with high risk, and 1 study with very high risk (table D5).
Using the GRADE approach, one conclusion was formulated: central venous pressure demonstrates low diagnostic accuracy for determining fluid responsiveness (quality of evidence: moderate).
Discussion
Key findings
The results of this meta-analysis, which included 84 studies encompassing 3,729 patients, indicate that central venous pressure has low diagnostic accuracy for predicting fluid responsiveness (AUROC 0.6) with a moderate quality of evidence.
Subgroup analyses revealed that the diagnostic accuracy of central venous pressure ranged from 0.55 to 0.71, further supporting its low diagnostic precision.
Univariate meta-regression analysis showed no statistically significant impact of age, sex, or BMI on the diagnostic accuracy of the method.
Relationship with previous studies
Despite the large number of studies and several meta-analyses, the applicability of central venous pressure (CVP) as an indicator of fluid responsiveness remains unresolved.
In a 2008 meta-analysis, Marik P.E. and colleagues, based on 10 studies involving 356 patients, demonstrated that CVP is not a reliable indicator of fluid responsiveness (AUROC = 0.56; 95% CI [0.51; 0.61]) [27]. Five years later, the authors confirmed their findings in a larger meta-analysis of 33 studies (AUROC = 0.56; 95% CI [0.52; 0.60]) [28]. However, a 2024 meta-analysis reported significantly higher diagnostic accuracy for CVP (AUROC = 0.77; 95% CI [0.69; 0.87]) [23]. Notably, this recent meta-analysis included only 12 studies with 429 patients. Our meta-analysis, encompassing 84 studies and 3,729 patients, aligns with the findings of Marik P.E. and colleagues, confirming the low diagnostic accuracy of CVP. Furthermore, our results are consistent with the study by Chaves RCF and colleagues, which reported low sensitivity and specificity of the method: 61% and 69%, respectively [23].
In terms of patient stratification, our findings are also in agreement with prior research. Marik P.E. and colleagues evaluated the diagnostic accuracy of CVP in subgroups of patients in operating rooms (AUROC = 0.56; 95% CI [0.54; 0.58]) and ICUs (AUROC = 0.56; 95% CI [0.52; 0.60]) [28]. They also analyzed cardiac (AUROC = 0.56; 95% CI [0.51; 0.61]) and non-cardiac (AUROC = 0.56; 95% CI [0.54; 0.58]) patients [28]. These results correspond with our findings, where the AUROC for these subgroups ranged from 0.57 to 0.60. However, our result for mechanically ventilated patients (AUROC = 0.60; 95% CI [0.57; 0.62]) contrasts with that of Chaves RCF and colleagues (AUROC = 0.77; 95% CI [0.69; 0.87]). This discrepancy likely reflects the greater comprehensiveness of our analysis, which included more studies (84 vs. 12) and patients (3,729 vs. 429). Additionally, we conducted subgroup analyses by fluid challenge volume and year of publication for the first time, demonstrating the robustness of our findings across time and varying gold standard.
This meta-analysis also uniquely performed a meta-regression to evaluate the impact of covariates on AUROC. Basic variables like age, sex, and BMI were shown to have no effect on the diagnostic accuracy of CVP. This finding supports the uniform interpretation of CVP regardless of these patient characteristics.
Significance of study findings
The significance of the findings from this meta-analysis is supported by the following points.
Firstly, we demonstrated the low diagnostic accuracy of central venous pressure (CVP) based on 4.472 observations from 3.729 patients (moderate quality of evidence), making this meta-analysis the largest conducted on this topic to date.
Secondly, we conducted a subgroup analysis to evaluate the performance of CVP assessment in cardiac and general ICU patients, as well as during various surgical procedures. In all these scenarios, the AUROC consistently indicated low diagnostic accuracy. Furthermore, we assessed CVP in a cohort of mechanically ventilated patients, again demonstrating its insufficient accuracy. Additionally, we showed that the temporal factor does not influence the quality of the method, as findings from studies published in the past five years align with earlier research.
Thirdly, we performed a meta-regression for the first time to investigate the impact of covariates on the diagnostic accuracy of CVP. Our analysis provided evidence that patient age, sex, and BMI do not affect the accuracy of the method.
Despite the robust evidence presented, the diagnostic accuracy of extreme CVP values remains an open question. In the studies analyzed, CVP cutoff points were consistently within the normal range. The low quality of the method within this range does not imply ineffectiveness of extreme pressure values in detecting fluid responsiveness. Unfortunately, addressing this question is beyond the scope of the current meta-analysis.
Strengths and limitations of meta-analysis
This meta-analysis has several strengths.
Firstly, it incorporates data from over 4.000 CVP measurements, allowing for a precise pooled AUROC estimate with narrow confidence intervals. Additionally, the robustness of the findings was demonstrated using a leave-one-out sensitivity analysis. Secondly, strict inclusion criteria were applied, focusing on the standardization of the gold standard for assessing fluid responsiveness. This approach reduced the variability associated with the lack of consensus, thereby improving the reliability of the results. Thirdly, the majority of studies included in the meta-analysis had a low or moderate risk of bias, ensuring confidence in the absence of result distortion. Furthermore, Egger's test confirmed the absence of publication bias. Fourthly, the use of meta-regression enabled an evaluation of the influence of covariates on the meta-analysis outcomes.
However, this meta-analysis is not without limitations. The primary limitation is the statistical heterogeneity observed in the results. Nonetheless, the application of the GRADE approach accounted for this when determining the quality of the evidence. Additionally, the inability to perform a trial sequential analysis (TSA) to evaluate AUROC due to the lack of a suitable mathematical framework in current medical statistics is another constraint.
Future studies and prospects
The findings highlight the need for further clinical research on this topic in two key directions:
Firstly, studies are needed to investigate the diagnostic accuracy of extreme CVP values. The low accuracy of the method within the normal range does not rule out its potential utility in detecting fluid responsiveness at extremely low or high pressures in the right atrium.
Secondly, research should focus on evaluating alternative methods for assessing fluid responsiveness that are as simple and accessible as CVP but offer greater diagnostic accuracy.
Conclusions
CVP demonstrates low diagnostic accuracy in assessing fluid responsiveness (moderate quality of evidence), which precludes its recommendation for routine clinical practice.
Disclosure. The authors declare no competing interests.
Author contribution. All authors according to the ICMJE criteria participated in the development of the concept of the article, obtaining and analyzing factual data, writing and editing the text of the article, checking and approving the text of the article.
Ethics approval. Not requred.
Funding source. This study was not supported by any external sources of funding.
Data Availability Statement. The data that support the findings of this study are available from the corresponding author upon reasonable request.
