CO₂ and O₂ removal during continuous veno-venous hemofiltration: a pilot study

Abstract

Background

Carbon dioxide (CO₂) accumulation is a challenging issue in critically ill patients. CO₂ can be eliminated by renal replacement therapy but studies are scarce and clinical relevance is unknown. We prospectively studied CO₂ and O₂ behavior at different sample points of continuous veno-venous hemofiltration (CVVH) and build a model to calculate CO₂ removal bedside.

Methods

In 10 patients receiving standard CVVH under citrate anticoagulation, blood gas analysis was performed at different sample points within the CVVH circuit. Citrate was then replaced by NaCl 0.9% and sampling was repeated. Total CO₂ (tCO₂), CO₂ flow (V̇CO₂) and O₂ flow (V̇O₂) were compared between different sample points. The effect of citrate on transmembrane tCO₂ was evaluated. Wilcoxon matched-pairs signed rank test was performed to evaluate significance of difference between 2 data sets. Friedman test was used when more data sets were compared.

Results

V̇CO₂ in the effluent (26.0 ml/min) correlated significantly with transmembrane V̇CO₂ (24.2 ml/min). This represents 14% of the average expired V̇CO₂ in ventilated patients. Only 1.3 ml/min CO₂ was removed in the de-aeration chamber, suggesting that CO₂ was almost entirely cleared across the membrane filter. tCO₂ values in effluent, before, and after the filter were not statistically different. Transmembrane tCO₂ under citrate or NaCl 0.9% predilution also did not differ significantly. No changes in V̇O₂ were observed throughout the CVVH circuit. Based on recorded data, formulas were constructed that allow bedside evaluation of CVVH-attributable CO₂ removal.

Conclusion

A relevant amount of CO₂ is removed by CVVH and can be quantified by one simple blood gas analysis within the circuit. Future studies should assess the clinical impact of this observation.

Trial registration

The trial was registered at https://clinicaltrials.gov with trial registration number NCT03314363 on October 192,017.

Electronic supplementary material

The online version of this article (10.1186/s12882-019-1378-y) contains supplementary material, which is available to authorized users.

Background

Red blood cells and plasma harbour carbon dioxide (CO₂) in the form of dissolved CO₂, bicarbonate, and carbamino compounds which are in equilibrium with each other [1]. The sum of all components is expressed as total CO₂ (tCO₂).

CO₂ accumulation causes hypercarbia which may be a challenge in intensive care unit (ICU) patients. It has propagated a more extensive use of extracorporeal techniques to enable ultra-protective ventilation in acute respiratory distress syndrome or to avoid intubation in patients with severe exacerbation of chronic obstructive lung disease [2]. Although renal replacement therapy (RRT) is advocated to generate small amounts of CO₂ due to the red blood cell passing through the filter [3], the net effect is a removal of CO₂ in an intermittent hemodialysis (IHD) model with acetate [3, 4] This CO₂ extraction of 41 ml/min seemed to correlate with a deficit of 46 ml/min in expired CO₂ [5]. Different methods have been explored to increase the removal of CO₂ in effluent by increasing pH with THAM or NaOH but it seemed too complex and too dangerous to be used in humans. [6] Extraction reached up to 120 ml/min in an in vitro model of IHD [7]. “CO₂ loss” induced by RRT may become clinically relevant as mean expired CO₂ in ICU patients is 180 ml/min [8]. Continuous RRT (CRRT) is progressively supplanting intermittent dialysis in the ICU. CRRT is hemodynamically well-tolerated, may provide easier control of metabolic alterations and fluid overload, and is associated with less chronic kidney disease in the post-ICU phase [9–11]. The impact of CRRT on CO₂ metabolism is remarkably poorly documented. In addition, trisodium citrate - the preferred anticoagulant for CRRT- acts as a weak acid [12]. This will alter the Henderson-Hasselbalch equation, disrupt the balance between the different CO₂ forms, and thus potentially influence CO₂ extraction during CRRT.

We designed a study to better understand CO₂ and O₂ extraction during CRRT. Based on obtained data, formulas were constructed to assess CRRT-related CO₂ clearance at the bedside.

Methods

A prospective study was performed in critically ill patients undergoing continuous veno-venous hemofiltration (CVVH). The study was approved by the Ethical Committee of the University Hospital Brussels (reference BUN 143201731636) and registered at https://clinicaltrials.gov (reference NCT03314363). Informed consent was obtained from the patient or a legal representative.

CVVH was performed with the Prismaflex®(Lund, Sweden) Baxter® device equipped with a Prismaflex® Baxter®, AN69 surface treated (ST) filter of 1.5 square meter (Meyzieu, France). Prismocitrate®18/0 (Sondalo, Italy) Baxter® was used as predilution and Prismocal® B22 (Sondalo, Italy) Baxter® or NaCl 0.9% as postdilution fluid. Dosing and postdilution fluid use were initiated and adapted according to an implemented standard CVVH protocol.

Inclusion and exclusion criteria are listed in Additional file 1. Blood samples were taken at 5 different sample points (SP) (Fig. 1). SP were located at the distal end of the arterial dialysis catheter lumen (SP1); between citrate predilution infusion port and filter (SP2); directly after the filter (SP3); at the effluent conduit (SP4); proximal to the venous dialysis catheter lumen (SP5). At every SP, pH, HCO₃-, tCO₂, pCO₂, pO₂, hemoglobin (Hb) and Hb saturation were measured using a blood gas analyzer [ABL90 Flex®, Radiometer(Bronshoj, Denmark)]. Subsequently, citrate predilution was stopped and replaced for at least 20 min by NaCl 0.9% at similar flow. Blood gas analysis was repeated according to the same protocol.

Fig. 1.

Schematic representation of CVVH set-up with location of sampling points and ports of fluid infusion. Pre = predilution fluid, Post = postdilution fluid, AC = de-aeration chamber

O₂ content (tO₂) was calculated as Hb x Hb saturation × 1.35 [1]. CO₂ (V̇CO₂) and O₂ flow (V̇O₂) at the specific SP were calculated by multiplying the set fluid flow (Q) on CVVH with respectively tCO₂ and tO₂. Results were adjusted from mmol to ml by using Boyle’s gas law: pV = nRT (p: pressure of the gas, V: volume of gas, n: amount of substance of gas, R: gas constant, T absolute temperature of the gas). The average air pressure recorded by the Belgian national weather institute was applied and ambient temperature was measured. “Transmembrane” (i.e before and after the filter) V̇CO₂ was calculated by subtracting V̇CO₂ at SP3 from V̇CO₂ at SP2. “Transmembrane tCO₂” was calculated in the same way.

tCO₂ of bicarbonate fluid was 22 mmol/l, converted by gas law to ml/l depending on ambient conditions, and calculated in ml/min based on the flow set on CVVH. When bicarbonate was used, the “expected V̇CO₂ at SP5” was calculated by adding the calculated V̇CO₂ of the postdilution fluid to the V̇CO₂ at SP3.

Relevant parameters such as CRRT settings were collected tobe used in a predictive equation.

Statistical analysis

Data were analyzed using Prism Graphpad® version 7(La Jolla, USA). As data sets contained 18 values at the most, normality was not assessed. Data are expressed as mean ± standard deviation. Wilcoxon matched-pairs signed rank test was performed to evaluate significance of difference between 2 data sets. Friedman test was used when more data sets were compared. Differences in measured data, V̇CO₂ and V̇O₂ between SP were evaluated.

V̇CO₂ in the effluent (SP4) was compared with “transmembrane V̇CO₂”. “Expected V̇CO₂ at SP5” was compared with V̇CO₂ at SP5. A subgroup analysis was performed to compare the influence of citrate on CO₂ extraction by the filter by comparing “transmembrane tCO₂” with and without citrate.

Results

Summary of patient characteristics are depicted in Table 1. CVVH settings of patients are provided in Additional file 2. Predilution citrate was not replaced by NaCl 0.9% in 2 patients because pre-existing hypercoagulability could compromise filter function.

Table 1.

patient characteristics

Patients (n)	10
Age (years)	68.7 ± 11.3
Gender (male/female)	8/2
BMI (kg/m2)	29.8 ± 7.3
APACHE II	27.1 ± 9.0
Reason for admission:
Medical	8
Surgical	2
Receiving controlled or assisted ventilation at day of study (n)	9
Mean CVVH settings in all series of blood gas analysis (ml/h)
Bloodflow	9000 ± 0
Predilution	1750 ± 447
Postdilution	444 ± 170
Effluent flow	2380 ± 175

Data are presented as means ± standard deviation

Comparison of V̇CO₂

V̇CO_{2 (SP1)} was higher than V̇CO_{2 (SP5)} and V̇CO_{2 (SP2)} [111.3 ± 8.1 ml/min vs. respectively 87.4 ± 14.6 ml/min (p < 0.01) and 110.5 ± 9.6 ml/min (p = 0.03)]. V̇CO₂ dropped significantly between SP2 and SP3 [from 110.5 ± 9.6 ml/min to 84.5 ± 6.5 ml/min (p < 0.01)]. V̇CO₂ at SP4 (26.0 ± 5.8 ml/min) and transmembrane V̇CO₂ at SP4 (24.2 ± 2.6 ml/min) were not statistically different (p = 0.39).

Results for NaCl 0.9% postdilution were plotted in Fig. 2. Patients receiving postdilution NaCl 0.9% exhibited higher V̇CO_{2 (SP1)} than V̇CO_{2 (SP5)} [109.8 ± 7.1 ml/min vs. 81.7 ± 5.8 ml/min (p < 0.01)] and a 1.3 ml/min difference between V̇CO_{2 (SP3)} and V̇CO_{2 (SP5)} [83.0 ± 4.9 ml/min vs. 81.7 ± 5.8 ml/min (p = 0.01)].

Fig. 2.

Evolution of CO₂ flow in the extracorporeal blood circuit during NaCl 0.9% postdilution. Sample point 4 was not included as it is not situated in the extracorporeal blood circuit and is not suited to represent evolution of V̇CO₂ in the blood. When difference between data was statistical significant different, this was marked with an asterisk

No statistical analysis was performed when bicarbonate was used as postdilution fluid as it only consisted of 3 data sets. A 21.7 ml/min difference was noted between V̇CO₂ at SP5 (116.1 ± 9.7 ml/min) and “expected V̇CO₂ at SP5” (94.4 ± 9.9 ml/min) (Fig. 3).

Fig. 3.

Expected CO₂ flow in sample place 5 versus CO2 flow at sample place 5 during bicarbonate containing Prismocal B22® postdilution. When difference between data was statistical significant different, this was marked with an asterisk

Comparison of tCO₂

Results are given in Fig. 4. tCO₂ at SP2 (25.5 ± 2.8 mmol/l), SP3 (25.0 ± 2.6 mmol/l) and SP4 (25.1 ± 2.6 mmol/l) were not statistically different (p = 0.51). tCO₂ decreased significantly between SP1 and SP2 from 30.6 ± 2.3 mmol/l to 25.5 ± 2.8 mmol/l (p < 0.01). At all SP, tCO₂ consisted of CO₂ in gas form (pCO₂) and HCO₃- (Fig. 5).

Fig. 4.

Evolution of tCO₂ at different sample points in all series of blood gas analysis. Sample point 5 was not included as it is influenced by bicarbonate Prismocal B22 postdilution fluid. When difference between data was statistical significant different, this was marked with an asterisk

Fig. 5.

Distribution of pCO₂ and HCO₃- at different sample points in all series of blood gas analysis. Sample point 5 was not included as it is influenced by bicarbonate Prismocal B22 postdilution fluid

Effect of citrate vs. no-citrate predilution on transmembrane tCO₂

Patients in whom citrate could not be withdrawn were excluded from analysis. ΔtCO₂ between SP2 and SP3 was not different in the citrate vs no-citrate group (p = 0.21) (Fig. 6).

Fig. 6.

Effect of citrate on transmembrane tCO₂

Comparison of V̇O₂

V̇O₂ at SP1, SP2, SP3 and SP5 was respectively 10.6 ± 3.7 ml/min,10.9 ± 3.9 ml/min, 10.3 ± 3.8 ml/min, and 10.9 ± 3.7 ml/min. V̇O₂ at SP4 was 0 ml/min as effluent contains no Hb. V̇O_{2 (SP1)} and V̇O_{2 (SP5)} were not different (p = 0.33) (Fig. 7).

Fig. 7.

Evolution of O₂ flow in the extracorporeal blood circuit in all series of blood gas analysis. Sample point 4 was not included as it is not situated in the extracorporeal blood circuit and is not suited to represent evolution of V̇O₂ in the blood

Development of formulas

The above findings allow to propose following formulas:

$$\dot{\mathbf{V}}\mathbf{C}{\mathbf{O}}_{\mathbf{2}\left(\mathbf{\text{effluent}}\right)}\mathbf{=}\dot{\mathbf{V}}\mathbf{C}{\mathbf{O}}_{\mathbf{2}\left(\mathbf{SP}\mathbf{4}\right)}\mathbf{=}{\mathbf{Q}}_{\mathbf{SP}\mathbf{4}}\mathbf{x}{\left[\mathbf{tC}{\mathbf{O}}_{\mathbf{2}}\right]}_{\mathbf{SP}\mathbf{4}}$$

As tCO₂ is similar at SP2, SP3 and SP4, the equation becomes:

$$\dot{\mathbf{V}}\mathbf{C}{\mathbf{O}}_{\mathbf{2}\left(\mathbf{\text{effluent}}\right)}\mathbf{=}{\mathbf{Q}}_{\mathbf{SP}\mathbf{4}}\mathbf{x}{\left[\mathbf{tC}{\mathbf{O}}_{\mathbf{2}}\right]}_{\mathbf{SP}\mathbf{3}}\mathbf{=}{\mathbf{Q}}_{\mathbf{SP}\mathbf{4}}\mathbf{x}{\left[\mathbf{tC}{\mathbf{O}}_{\mathbf{2}}\right]}_{\mathbf{SP}\mathbf{2}}\left[{}^{\mathbf{\ast }}\right]$$

By assuming that

$$\dot{\mathrm{V}}{\mathrm{CO}}_{2\left(\mathrm{SP}1\right)}\approx \dot{\mathrm{V}}{\mathrm{CO}}_{2\left(\mathrm{SP}2\right)}<=>{\mathrm{Q}}_{\mathrm{SP}1}\mathrm{x}{\left[{\mathrm{tCO}}_2\right]}_{\mathrm{SP}1}\approx {\mathrm{Q}}_{\mathrm{SP}2}\mathrm{x}{\left[{\mathrm{tCO}}_2\right]}_{\mathrm{SP}2}<=>{\left[{\mathrm{tCO}}_2\right]}_{\mathrm{SP}2}\approx {\mathrm{Q}}_{\mathrm{SP}1}\mathrm{x}{\left[{\mathrm{tCO}}_2\right]}_{\mathrm{SP}1}/{\mathrm{Q}}_{\mathrm{SP}2}$$

When [tCO₂]_SP2 is substituted in the above formula [*], it becomes

$$\dot{\mathbf{V}}\mathbf{C}{\mathbf{O}}_{\mathbf{2}\left(\mathbf{\text{effluent}}\right)}\mathbf{\approx }{\mathbf{Q}}_{\mathbf{SP}\mathbf{4}}\mathbf{x}{\mathbf{Q}}_{\mathbf{SP}\mathbf{1}}\mathbf{x}{\left[\mathbf{tC}{\mathbf{O}}_{\mathbf{2}}\right]}_{\mathbf{SP}\mathbf{1}}/{\mathbf{Q}}_{\mathbf{SP}\mathbf{2}}$$

Discussion

We present the first study that prospectively evaluated CO₂ and O₂ behavior in patients undergoing CVVH. The main finding was that a substantial amount of 26.0 ml/min CO₂ was removed in the effluent. This represents approximately 14% of the average expired V̇CO₂ measured in ICU patients [8] and thus could be clinically relevant. Furthermore, CO₂ removal during CVVH was found to be 80% lower than previously observed in an in vitro hemodialysis model. This is explained by the almost threefold higher blood flow rate used in this model as compared to our CVVH setting [7]. V̇CO₂ before and after predilution (Δ V̇CO₂ between SP1 and SP2) was statistically different, probably because the set CVVH fluid flow at these SP did not correspond with real fluid flow [9]. Blood analysis also depended on “snapshot” sampling which might not exactly reflect average flow. However, this difference did not seem clinically relevant compared with average expired V̇CO₂ (< 1%). CO₂ flow was then divided between the effluent and the blood running to the de-aeration chamber (SP3 and SP4). The CO₂ flow in the effluent correlated with V̇CO₂ loss in the blood after passing the filter [$\dot{\mathrm{V}}{\mathrm{CO}}_{2\left(\mathrm{SP}2\right)}-\dot{\mathrm{V}}{\mathrm{CO}}_{2\left(\mathrm{SP}3\right)}=\dot{\mathrm{V}}{\mathrm{CO}}_{2\left(\mathrm{SP}4\right)}$].

In patients receiving postdilution NaCl 0.9%, 1.3 ml/min of CO₂ was removed in the de-aeration chamber. This is a very small quantity compared to the average V̇CO₂ in ICU patients [8]. Thus, CO₂ removal was almost entirely determined by transmembrane filtering and measurable in the effluent. However, when postdilution bicarbonate was used, the expected V̇CO₂ did not correspond with the calculated V̇CO₂ in the blood before it re-entered the body. Several assumptions may explain this observation. First, measurements may be incorrect when CO₂ fails to enter red blood cells after being infused in the postdilution fluid into the extracorporeal circuit. Second, tCO₂ was calculated and not measured. Formulas for these calculations may not be applicable in a non-physiological state of bicarbonate-induced blood alkalinization. Studies measuring blood tCO₂ are needed to elucidate this problem.

As suggested by in vitro hemodialysis, CO₂ is removed in the effluent in gas form and as HCO₃- [7]. The CO₂ concentration or tCO₂ is the driving force for this removal as it remains constant in effluent and in the blood passing through the filter [tCO_2(SP2) = tCO_2(SP3) = tCO_2(SP4)]. By adding predilution fluid, tCO₂ decreased between SP1 and SP2.

Citrate anticoagulation did not influence tCO₂ extraction. Only the short term effect of citrate upon CO₂ removal was evaluated as an influencer of acid-base homeostasis. Over a longer time period, citrate could possibly affect CO₂ clearance because it preserves membrane porosity better than heparin. tCO₂ in blood passing through the CVVH circuit decreased as it was diluted by bicarbonate-free solutions. CVVH had no impact on V̇O₂ because values remained constant at the different SP.

Based on previous findings, different formulas were constructed to calculate CO₂ removal by CVVH in a clinical setting with the use of only one blood gas analysis in the extracorporeal circuit at a preexisting sample point. As these are the first data that were acquired in a CVVH setting, formulas could not be compared to data from other articles [7]. Further studies need to confirm our findings.

Several limitations of our study must be emphasized. First, despite the high number of analyses per patient, the sample size remains small and future studies in more patients are needed to confirm our results. Second, assumptions were made based on “snapshot” blood gas analysis. Continuous monitoring would be more precise. Third, fluid flows as set on CVVH may not correlate with real flow [13]. In-circuit flow measurements may be a better option. Finally, it remains to be determined whether the removed CO₂ influences expired CO₂.

Conclusion

A significant amount of CO₂, both as gas and bicarbonate and measurable in the effluent, is removed during CVVH under citrate anticoagulation. Pre-filter tCO₂ is the major determinant for CO₂ removal. Citrate does not influence CO₂ elimination. To a certain extent, bicarbonate fluids influence blood gases but data are too limited to permit relevant conclusions. Oxygen flow is not influenced by CVVH. CO₂ removal by CVVH in bicarbonate-free conditions can be calculated by multiplying effluent or blood flow with CO2 content at a preexisting sample point. Their clinical relevance requires confirmation.

Abbreviations

CO₂

Carbon dioxide

CRRT

Continuous renal replacement therapy

CVVH

Continuous veno-venous hemofiltration

Hb

Hemoglobin

ICU

Intensive care unit

IHD

Intermittent hemodialysis

n

Amount of substance of gas

O₂

Oxygen

p

Pressure of the gas

Q

Fluid flow

R

Gas constant

RRT

Renal replacement therapy

SP

Sample point

T

Absolute temperature of the gas

tCO₂

Total CO₂

tO₂

Oxygen content

V

Volume of gas

V̇CO₂

CO₂ flow

V̇O₂

O₂ flow

Authors’ contributions

JJ and EDW were responsible for the study design and background search. JJ, EDW and JD collected and assembled the data. JJ, HS, AD, JD, MD, OC, KL, TO, PH, MM and EDW participated in analysis and interpretation of data and manuscript writing. JJ, HS, AD, JD, MD, OC, KL, TO, PH, MM and EDW approved the final version of the manuscript.

Ethics approval and consent to participate

The study was approved by the Ethical Committee of the University Hospital Brussels (reference BUN 143201731636). Written informed consent for participation was obtained from the patient or a legal representative.

Consent for publication

Not applicable.

Competing interests

Dr. Jonckheer and Prof. Dr. De Waele have received a grant from Baxter as a replacement fee and for logistic support.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.