Introduction

Chronic obstructive pulmonary disease (COPD) is one of the heaviest burdens placed on medical institutions, not to mention governments, all over the world [1]. Acute exacerbation of COPD (AECOPD) is defined as a progressive deterioration of ventilatory function in patients with COPD, leading to hypoxia and hypercarbia [2]. Arterial blood gas (ABG) is a gold standard to obtain information on the potential of hydrogen (pH), partial pressure of carbon dioxide (PCO2), partial pressure of oxygen (PO2), and bicarbonate (HCO3−) in order to measure the severity AECOPD according to the recommendations of National Institute for Health and Care Excellence (NICE) [3]. However, arterial sampling obtained by puncture biopsy is associated with significant disadvantages, including technical difficulty, pain, infection, thrombosis with distal ischemia, as well as severe subcutaneous hematoma, and repeated arterial puncture biopsy increases the incidence of complications. Alternative solutions for ABG need to be found to lower the cost of assessing and treating AECOPD.

Data from recent studies suggested good agreement between venous blood gas (VBG) and ABG when measuring pH, PCO2, PO2, and HCO3− in patients with diabetic keto-acidosis (DKA), tricyclic overdose, uremia, and emergency department (ED) trauma [4, 5]. Simultaneously, VBG has also been reported in recent studies as an alternative to ABG for monitoring during AECOPD treatment [3, 6,7,8,9,10,11,12,13].

The objective of this study was to perform a meta-analysis of the reported literature entries related to the agreement of parameters pH, PCO2, PO2, and HCO3− between VBG and ABG to determine whether they are comparable, in order to verify the feasibility of replacing ABG with VBG in patients with AECOPD.

Methods

Search strategy

A general search for English-language literature was performed in the following databases: PubMed, Cochrane Library, Google Scholar, and China National Knowledge Infrastructure (CNKI). The date of the last search was February 22 2017. The definite words applied during the search in English included: “venous blood gas,” “arterial blood gas,” “VBG,” “ABG,” “chronic obstructive pulmonary disease,” “chronic obstruction pulmonary disease,” “COPD,” “pH,” “PCO2,” “PO2,” “HCO3−” in the titles or abstracts. The references listed below the articles searched were scanned and selected for inclusion in the meta-analysis.

Inclusion and exclusion criteria

Case-control studies comparing ABG and VBG in patients with COPD in which sample size, study population, criteria for patient selection, and type of blood gas analyzer used for the analysis were included. Searches that reported mean and standard deviation (SD) of paired pH, PCO2, PO2, or HCO3− for peripheral venous and arterial blood samples were included. Case studies and research on animals were excluded. Studies in which venous samples were obtained via central venous access were also excluded.

Data extraction

The basic characteristics of searches, including first author names, publication year, patient cases, and paired variables, i.e., pH, PCO2, PO2, or HCO3− (mean ± SD), were extracted. Full texts and supplementary material from selected literature were reviewed for the extraction of available data. The meta-analysis was carried out according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses report (PRISMA). Data analysis, including study searches, article selection, and data extraction was performed by two independent researchers; any disagreements were discussed and solutions sought.

Statistical analysis

In the present study, the variables derived from the included investigations, which were presented as mean ± SD in both ABG and VBG, were converted into mean differences (mean ± SD) by subtracting venous values from arterial values. For the purposes of pooling the mean differences (MDs), 95% confidence intervals (CIs) were required. In cases in which 95% CIs were not provided in the selected studies, these were calculated from the given MDs. CIs for a mean are a function of the standard error of mean (SEM), which is the SD divided by the square root of the case sample; 95% limits of agreement (LOA) is the range of the mean plus or minus the SD multiplied by 2. Therefore, CIs are necessarily narrower than the LOA. The data obtained from texts directly or calculated with the given variable were combined for MDs.

Tests for ABG and VBG, including pH, PCO2, PO2, and HCO3−, were considered comparable if the average bias was less than the error defined by standard laboratory performance criteria: pH ±0.04, PCO2 ±5 mm Hg, and PO2 ±7 mm Hg [14].

The Higgins I2-test was used as a heterogeneity test to detect heterogeneity of the included data, whereby: I2 = 0 meant no heterogeneity; I2 < 50% meant mild heterogeneity; 50% < I2 < 75% meant moderate heterogeneity; I2 > 50% meant high heterogeneity [15]; and I2 reflects the proportion of the part of heterogeneity among the total variation of effect size. The formula is I2  = 100% × (Q −df) / Q, where Q means statistics and df means degrees of freedom. In the case of I2 < 50%, a fixed effects meta-analysis was used. In the case of I2 ≥ 50%, a random effects meta-analysis was performed. The meta-analysis was performed using the package Metan running in Stata software version 13.0.

Results

Search results

According to the PRISMA diagram shown in Fig. 1, 568 literature entries were obtained from PubMed (n = 447) and other sources (n = 121), including the Cochrane Library, Google Scholar, and CNKI. After the deletion of duplicates, there were 514 remaining results, of which 52 literature entries were selected according to titles. By limiting selection to studies on COPD, seven studies with 1234 patients were included in the final analysis (Ak et al. 2006, McCanny et al. 2012, McKeever et al. 2016, Novovic et al. 2012, Raoufy et al. 2011, Razi et al. 2007, Sadariya et al. 2014).

Fig. 1
figure 1

Study selection procedure shown as a PRISMA flow diagram

Full size image

All studies included were on the comparison between ABG and VBG in patients with COPD. Study characteristics are listed in Table 1.

Table 1 The characteristics of included studies
Full size table

Comparisons

The pooled MD for pH was 0.02 pH units (ABG minus VBG) (95% CI, 0.02–0.03), for PCO2 −2.91 mm Hg (95% CI, −4.23 to −1.59), for PO2 13.13 mm Hg (95% CI, 9.13 to 17.12), and for HCO3− −0.22 mmol/l (95% CI, −0.62 to 0.18) (Figs. 2, 3, 4, and 5).

Fig. 2
figure 2

Pooled mean differences for pH (arterial blood gas minus venous blood gas) with 95% confidence intervals (CI). MD mean difference

Full size image
Fig. 3
figure 3

Pooled mean differences for partial pressure of carbon dioxide (PCO2) (arterial blood gas minus venous blood gas) with 95% confidence intervals (CI). MD mean difference

Full size image
Fig. 4
figure 4

Pooled mean differences for partial pressure of oxygen (PO2) (arterial blood gas minus venous blood gas) with 95% confidence intervals (CI). MD mean difference

Full size image
Fig. 5
figure 5

Pooled mean differences for bicarbonate (HCO3−) (arterial blood gas minus venous blood gas) with 95% confidence intervals (CI). MD mean difference

Full size image

Discussion

As a result of the present study, the pooled MD for pH was 0.02 pH units (ABG minus VBG) (95% CI, 0.02−0.03), for PCO2 −2.91 mm Hg (95% CI, −4.23 to −1.59), for PO2 13.13 mm Hg (95% CI, 9.13−17.12), and for HCO3− −0.22 mmol/l (95% CI, −0.62 to 0.18). According to the error defined by standard laboratory performance criteria [14], whereby pH is ±0.04, PCO2 ±5 mm Hg, and PO2 ±7 mm Hg, the data above suggested that they might compare well between ABG and VBG for pH, PCO2, but not for PO2. In addition, HCO3− is a derived value of pH and PCO2 using the Henderson-Hasselbalch equation. Thus, HCO3− of VBG might also compare well to that of ABG.

Data from a prospective trial have revealed that there was a significant correlation between ABG and VBG values for pH, PCO2, and HCO3− (p < 0.001) with the Pearson correlation coefficients, respectively [6], which was further verified by other investigations [3, 8, 9, 12]. Although PO2 showed poor correlation between VBG and ABG in the present analysis [11], it has been reported that, when using VBG values, the designed artificial neural networks (ANN), a non-linear mode, was able to predict all ABG values (even PO2 and O2 saturation) by regression equations such as arterial PO2 = 26.651 + 0.694 × venous PO2; and arterial O2 saturation = 64.724 + 0.297 × venous O2 saturation in AECOPD patients [11]. In the authors’ view, ANNs have the added advantage that they can learn to predict arbitrarily complex nonlinear relationships between independent and dependent variables by including more processing elements in the hidden layer or more hidden layers, which make the ANN a more robust paradigm for application in a real-world setting [16]. In this way, ABG values could be calculated using specific formula from VBG by ANNs, thereby avoiding the puncture biopsy complications associated with ABG.

The ability to detect hypoxemia hypercapnia (arterial PCO2 > 50 mm Hg) in COPD patients is of crucial importance, since it provides the opportunity for earlier intervention to treat respiratory failure [17]. Given the associated relationship between pH and PCO2 and good agreement between VBG and ABG, a trend of peripheral venous pH and HCO3− could potentially be used to follow the trend of respiratory acidosis [18]. Clinically, an unanticipated HCO3− value with acidosis in patients with COPD should prompt the physician to search for other underlying diseases causing metabolic acidosis, such as lactic acidosis from septic shock or uremia [19]. Hence, the authors suggest an algorithm for the management of COPD exacerbations in clinical practice (Fig. 6). However, in terms of this meta-analysis, the majority of suggestions in Fig. 6 were based on data from the studies included, which might be less specific and intended merely as a guide for clinicians.

Fig. 6
figure 6

Suggested algorithm for blood gas analysis during chronic obstructive pulmonary disease (COPD) exacerbation. AECOPD acute exacerbation of chronic obstructive pulmonary disease, ABG arterial blood gas, VBG venous blood gas, pH potential of hydrogen, PCO2 partial pressure of carbon dioxide, HCO3 bicarbonate, SPO2 peripheral capillary oxygen saturation

Full size image

Previous studies, including systematic reviews and meta-analyses, that compared peripheral VBG and ABG analysis in emergency department patients showed that venous and arterial pH agree reasonably well for all values, but the agreement is highest for normal values. In addition, arteriovenous agreement for PCO2 and HCO3− is poor [5]; no sub-group analysis was performed for cases involving patients with, e.g., diabetic keto-acidosis (DKA), tricyclic overdose, or uremia. Another study showed that VBG compares well with ABG analysis for pH estimations in adults, but not for PCO2 or PO2 in emergency department patients [4]. It was revealed that venous PCO2 cannot be used to predict arterial PCO2 by applying a ‘correction factor.’ Another investigation reported that the VBG might be useful in determining whether venous blood has a low amount of CO2 and high O2 to exclude hypercapnia and hypoxia [20]. Likewise, the systematic review of estimates of differences between VBG and ABG was subject to a high degree of heterogeneity, which might lead to a further bias. Whatever the case, pH was assumed to be comparable between VBG and ABG whether or not in AECOPD patients.

One of the most obvious limitations of the present meta-analysis is that all cases in the included literature entries presented with stable hemodynamics and normal blood pressure. According to earlier studies, there is poor agreement for pH, PCO2, and HCO3− in patients with shock or unstable hemoperfusion [21], which gave rise to uncertainties about the correlation in COPD patients with circulatory failure or shock. In addition, the authors tried but failed to specify the relevant coefficient of parameters, including pH, PCO2, or HCO3−, between ABG and VBG, which represents another limitation of this study.

In conclusion, VBG analysis compares well with ABG for pH, PCO2, and HCO3− estimations, but not for PO2 in patients with AECOPD.