Respiratory and metabolic acidosis correction with the ADVanced Organ Support system

Background The lung, the kidney, and the liver are major regulators of acid-base balance. Acidosis due to the dysfunction of one or more organs can increase mortality, especially in critically ill patients. Supporting compensation by increasing ventilation or infusing bicarbonate is often ineffective. Therefore, direct removal of acid may represent a novel therapeutic approach. This can be achieved with the ADVanced Organ Support (ADVOS) system, an enhanced renal support therapy based on albumin dialysis. Here, we demonstrate proof of concept for this technology. Methods An ex vivo model of either hypercapnic (i.e., continuous CO2 supply) or lactic acidosis (i.e., lactic acid infusion) using porcine blood was subjected to hemodialysis with ADVOS. A variety of operational parameters including blood and dialysate flows, different dialysate pH settings, and acid and base concentrate compositions were tested. Comparisons with standard continuous veno-venous hemofiltration (CVVH) using high bicarbonate substitution fluid and continuous veno-venous hemodialysis (CVVHD) were also performed. Results Sixty-one milliliters per minute (2.7 mmol/min) of CO2 was removed using a blood flow of 400 ml/min and a dialysate pH of 10 without altering blood pCO2 and HCO3− (36 mmHg and 20 mmol/l, respectively). Up to 142 ml/min (6.3 mmol/min) of CO2 was eliminated if elevated pCO2 (117 mmHg) and HCO3− (63 mmol/l) were allowed. During continuous lactic acid infusion, an acid load of up to 3 mmol/min was compensated. When acidosis was triggered, ADVOS multi normalized pH and bicarbonate levels within 1 h, while neither CVVH nor CVVHD could. The major determinants to correct blood pH were blood flow, dialysate composition, and initial acid-base status. Conclusions In conclusion, ADVOS was able to remove more than 50% of the amount of CO2 typically produced by an adult human. Blood pH was maintained stable within the physiological range through compensation of a metabolic acid load by albumin dialysate. These in vitro results will require confirmation in patients.


Introduction
Lung, kidney, and hepatic dysfunction are common in the critically ill, and acid-base regulation is frequently compromised in these patients. Acidosis is commonly associated with high mortality rates in critically ill and injured patients. Recently, a strong correlation was found between hypercapnic acidosis and increased hospital mortality in mechanically ventilated patients, compared to compensated hypercapnia or normocapnia [1]. Indeed, a delayed pH normalization is associated with increased mortality in the intensive care unit (ICU), reaching 57% in cases of severe metabolic or mixed acidemia [2]. If accompanied by hyperlactatemia, these values could go above 80% [3,4].
While acidosis is well tolerated in healthy humans, acidosis leads to a myriad of physiologic effects that can be deleterious and thus contribute to morbidity and mortality in patients [5]. In fact, an interaction exists between acidosis and inflammation, which is specifically relevant in critically ill patients [6,7]. Acidosis can impair several immune response mechanisms, including lymphocyte cytotoxicity, complement activation, or antibody binding to leukocytes [8]. Since some of these findings have been already reported in patients [7], acid-base imbalances should be considered in the context of a host response to an aggression, and not as an isolated insult. Moreover, pH might also influence normal physiology, among others, modulating oxygen affinity to hemoglobin [9,10], promoting vasoconstriction in the lungs [11], altering potassium and calcium levels, reducing glomerular filtration rate [12], reducing intestinal mobility [13], impairing coagulation [14], or depressing myocardial contractility [9].
The ADVanced Organ Support (ADVOS) system (ADVITOS GmbH, Munich, Germany-previously, Hepa Wash GmbH) is an albumin-based hemodialysis device initially designed to support the liver and kidney of ICU patients. As a hemodialysis system, it removes water-soluble substances, while the albumin dialysate allows to remove protein-bound toxins too [15][16][17]. The ADVOS system consists of three circuits: an extracorporeal blood, a dialysate, and an ADVOS multi circuit with an acidic and an alkaline path (Fig. 1). Briefly, the dialysate's albumin binds protein-bound substances that diffuse from blood into the dialysate through a semi-permeable membrane in the extracorporeal circuit. Differently to conventional single pass albumin dialysis (SPAD), in the ADVOS system, the dialysate is not systematically discarded and recirculates Fig. 1 Schematic representation of ADVOS multi device with a continuous CO 2 supply. Lactic acid was supplied in blood only for experimental settings 2 and 4 (see Table 1 for more details) then through the dialysate circuit into the ADVOS multi circuit. There, dialysate albumin is recycled by systematically modifying the tertiary structure of albumin through temperature and pH changes. This facilitates the release of toxins from albumin (cationic-e.g., copper-and anionic-e.g., bilirubin-substances in the acidic and alkaline paths, respectively) and makes it ready for further binding. pH changes are possible by the addition of an acidic and an alkaline concentrate, whose customizable mixing relation forms a dialysate with a variable composition. This includes modifiable carbonate, sodium, or chloride concentrations that allow to achieve a customizable dialysate with a pH from 7.2 to 10.0. Consequently, this latter feature enables a continuous control and adjustment of dialysate acid-base properties, which in turn corrects deviations in blood pH (e.g., acidosis) by means of altering blood pCO 2 , strong ion difference, or both [18].
In the present study, the ability of ADVOS to eliminate CO 2 and to correct blood pH was investigated in an ex vivo model of either lactic or hypercapnic acidosis, using porcine blood and a continuous supply of lactic acid and/or CO 2 , respectively. A variety of operational parameters, including blood and dialysate flows as well as different dialysate pH settings and acid and base concentrate compositions, were tested. In addition, comparisons with standard continuous veno-venous hemofiltration (CVVH) using high bicarbonate substitution fluid and continuous veno-venous hemodialysis (CVVHD) were performed. Finally, an analysis of the mechanism of action regarding classic and modern acid-base balance approaches is provided.

Ex vivo model
For all the experiments, an ex vivo model comprising 5 l fresh porcine blood connected to an extracorporeal support device (either ADVOS multi, CVVH, or CVVHD) was employed. For experiments with ADVOS multi or CVVHD, an additional continuous CO 2 supply was added. Small modifications comprising the administration of different solutions were performed for each specific experiment (Fig. 1).

Blood preparation
Fresh porcine blood was obtained from the local slaughterhouse (Münchner Schlachthof Betriebs GmbH, Munich, Germany) and prepared following a standard operation procedure (SOP) diluting it with modified Ringer's solution (100 mmol/l NaCl, 3 mmol/l KCl, 1 mmol/l MgSO 4 .7H 2 O, 1.75 mmol/l CaCl 2 , 1200 mg/dl glucose) to a hematocrit of 36%, standard electrolyte concentrations, and normal blood gas values. 100,000 IU of heparin (Ratiopharm, Ulm, Germany) prevented coagulation. Blood was maintained at a constant temperature of 37°C using a bath heater while stirring it at 130-180 rpm.

ADVOS multi
For the ADVOS system (ADVITOS GmbH, Munich, Germany), two SURELYZER PES-190 DH dialyzers (Nipro D.Med Germany GmbH, Hamburg, Germany) with blood and dialysate flowing co-currently were employed in the extracorporeal circuit. Blood flow can be adjusted between 100 and 400 ml/min. A dialysate flow of 800 ml/min was used throughout the study, which refers to the amount of fluid per minute being recirculated and detoxified in the ADVOS multi circuit by means of pH adjustments and filtration, instead of being discarded after a single pass. Variations of pH and composition are achieved using acidic and basic concentrates (see below). The concentrate flow (160 or 320 ml/min) determines the quantity of dialysate cleaned by convective transport in the ADVOS multi circuit. Toxins released from albumin or water-soluble toxins are separated from the albumin dialysate through two ELISIO-13H filters (Nipro D.Med Germany GmbH. Hamburg, Germany). Additionally, the concentrate flow refers to the amount of new fresh concentrate (i.e., mix of water and acidic and alkaline concentrates) pumped into the system.

CO 2 administration for ADVOS multi and CVVHD experiments
For this ex vivo model, CO 2 was continuously infused into the blood pool via an additional SURELYZER PES-190 DH dialyzer (Nipro D.Med Germany GmbH, Hamburg, Germany) connected to a CO 2 gas supply (Linde AG, Munich, Germany). Contrary to a standard dialyzer setup, in the CO 2 dialyzer, blood was circulated through the external side to reduce the pressure within the blood circuit. CO 2 was supplied to the inner side of the CO 2 dialyzer in a countercurrent flow via a pressure reducer (FMD 202, Linde AG, Munich, Germany) and a mass flow meter FMA-1618A (OMEGA Engineering, Deckenpfronn, Germany). The bottom outlet of the CO 2 dialyzer was closed to avoid CO 2 losses while the semipermeable membrane allowed the gas to diffuse freely into the blood.

Dialysate for ADVOS multi
In contrast to other hemodialysis methods, the ADVOS system does not use a fixed dialysate composition. Instead, two concentrates (acid and base) are automatically mixed in a specifically designed reservoir throughout the treatment depending on the desired dialysate pH (range 7.20-10.00). A higher dialysate pH setting means a higher basic/ acidic concentrate ratio and thus higher sodium and lower chloride levels. Two 100 ml bottles of albumin 20% (Human Albumin 200 g/l, Baxter, Vienna, Austria) are added to the above dialysate mix via a specific port in the ADVOS multi device. The dialysate is furthermore supplemented with 40% glucose at an infusion rate of 70 ml/h to maintain glucose levels around 100 mg/dl in blood.

NIKKISO DBB-03
The NIKKISO DBB-03 dialysis system (NIKKISO Europe GmbH, Langenhagen, Germany) consisted of a blood and dialysate circuit working as a single pass process without recirculating the dialysate. The dialysate flow (350 ml/min) determined the amount of dialysate being supplied and discarded. It was equipped with one SUREFLUX-25UX dialyzer (Nipro D.Med Germany GmbH, Hamburg, Germany)

Experimental design
Experiments were divided into three groups. First, blood was titrated with CO 2 or lactic acid to achieve a blood pH range of 7.35-7.45 while being treated with ADVOS multi using different settings (set 1 and 2, respectively). Second, using a fixed CO 2 or lactic acid supply, the performance of ADVOS multi vs. hemodialysis (CVVHD) or hemofiltration (CVVH) was compared (set 3 and set 4, respectively). Third, a hypercapnic acidosis was triggered in blood and further treated with ADVOS multi until recovery of normal blood gas values. Details for each of the experimental sets are summarized in Table 1.

Set 1: Influence of ADVOS multi operational settings on CO 2 removal
In order to determine the influence of blood flow and dialysate composition on CO 2 removal ability of the ADVOS system, different settings were tested during a continuous CO 2 supply (Fig. 1). Briefly, 5 l of blood at physiological levels of pH (7.35-7.45), HCO 3 − (22-28 mmol/l), and pCO 2 (35-45 mmHg) was treated with ADVOS multi at experimental blood flows (Q b ) of 100, 200, or 400 ml/min with co-currently recirculating dialysate at flows of 800 ml/min. At each Q b , dialysate pH was set to 7.5, 8.0, 8.5, and 9.0 using a concentrate flow (Q c ) of 160 ml/min. At the highest Q b of 400 ml/min, additional tests were carried out with a Q c of 320 ml/min. All these experiments were carried out using the concentrates AC and BC-Bic20.
With the intention to test if different bicarbonate concentrations of the dialysate might affect CO 2 removal, additional experiments with BC-Bic0 (without bicarbonate) instead of BC-Bic20 were carried out setting dialysate pH to 10.00.
Prior to blood hemodialysis, every dialyzer was primed with a 0.9% NaCl solution removing air before blood contact. The CO 2 dialyzer was flushed with gas prior to and during NaCl and blood perfusion to create a positive pressure gradient which prevented liquids from entering the capillaries. Every test consisted of a 20-min stabilization period during which CO 2 supply was adjusted such that blood pH remained between 7.35 and 7.45. This was followed by a 1-h treatment phase during which samples from the inlet (pre-dialyzer) and outlet (post-dialyzer) were analyzed by the blood gas analyzer GEM Premier 4000 (Instrumentation Laboratory, Munich, Germany) every 20 min. In addition, pH was measured by an InPro 3253 pH probe inserted into the blood container and M300 displayed it continuously (both Mettler Toledo, Greifersee, Switzerland).
To quantify CO 2 removal in milliliters per minute, post-dialyzer TCO 2 values in millimoles per liter were subtracted from pre-dialyzer values and this difference multiplied by the corresponding blood flow (Q b ) and by the molar volume (V m ) of CO 2 at STP (22.4 ml/mmol) (Eq. 1). The fraction of TCO 2 excreted as dissolved CO 2 or HCO 3 − was calculated likewise [19]. n.a. not applicable (1) CO 2 supply was adjusted such that blood pH remained between 7.35 and 7.45 (2) CO 2 was continuously infused on demand to maintain pCO 2 levels between 35 and 45 mmHg (3) A continuous 2% lactic acid solution was infused such that blood pH remained between 7.35 and 7.45 *The recirculating dialysate flow in the ADVOS system reflects the volume of dialysate that recirculates continuously (not discarded) **The concentrate flow corresponds to the dialysate flow of a conventional single pass dialysis device and reflects the amount of dialysate used and discarded ***Experiments performed with each combination of blood flow, concentrate flow, and dialysate pH The same experimental design as for set 1 was performed to determine the influence of ADVOS parameters, with small modifications. Briefly, a continuous 2% lactic acid solution was used to titrate blood pH to 7.35-7.45. CO 2 , instead, was continuously infused on demand to maintain pCO 2 levels between 35 and 45 mmHg. These experiments were only carried out with Q c of 320 ml/min and with BC-Bic20. For each experiment, the maximal combined acid supply resulting from lactic acid infusion and CO 2 influx supply was calculated from the infusion rate of the pump (Volumat MC Agilia. Fresenius Kabi, Bad Homburg, Germany) and reading of the mass flow meter FMA-1618A (OMEGA Engineering, Deckenpfronn, Germany), respectively.

Set 3: ADVOS multi vs. CVVHD during continuous maximal CO 2 supply
The maximum CO 2 supply was determined previously in set 1 to be 110 ml/min (Additional file 1: Table S1). Blood was treated for 4 h with either the ADVOS multi or the CVVHD device NIKKISO DBB-03. The settings for each device are detailed in Table 1. Blood gas analysis was performed every 15 min.

Set 4: ADVOS multi vs. CVVH for the treatment of lactic acidosis in vitro
First, 5 l of fresh swine blood was subjected to CVVH with a substitution fluid with 10 mmol/l bicarbonate. Simultaneously, a 2% lactic acid solution was infused in order to reach a pH < 7.15, a pCO 2 between 35 and 45 mmHg, HCO 3 − levels between 12 and 14 mmol/l, and lactate of 5-6 mmol/l, which simulated a severe lactic acidosis. Blood was then treated with either ADVOS multi or CVVH for 1 h. Lactic acid infusion was maintained during the treatment phase. Blood was analyzed as described above.

Set 5: Treatment of hypercapnic acidosis in vitro with ADVOS multi
In this case, a hypercapnic acidosis was triggered first. Briefly, 27 ml/min CO 2 was infused while a blood flow of 100 ml/min and a dialysate pH of 7.8 were set. During this phase, AC was combined with BC-Bic20. Once a pH < 7.15, a pCO 2 > 60 mmHg, and HCO 3 − levels >32 mmol/l were reached, both the settings and the BC were changed, and the treatment phase started. Blood was then treated with ADVOS multi until normal values of pH (7.35-7.45), pCO 2 (35-45 mmHg), and HCO 3 − (22-26 mmol/l) were detected. Blood values were analyzed as described above. CO 2 was continuously supplied with the same flow of 27 ml/min.

Acid-base balance according to Stewart
We obtained blood samples both at the inlet and the outlet of the dialyzers from experimental sets 1 and 2. The analysis of these data provides an understanding of pH variations attending to pCO 2 and SID changes. In order to better understand this physicochemical method proposed by Stewart [20], several authors have tried to calculate specific values for the total concentration and the effective dissociation constant for plasma nonvolatile buffers. Constable suggested that "at normal pH (7.40), a 1-meq/l increase in SID will increase pH by 0.016, a 1-Torr increase in pCO 2 will decrease pH by 0.009, and a 1 g/dl increase in total protein will decrease pH by 0.039" [21]. It is assumed that no variation on total protein content occurs in our setting as it cannot be lost in the dialyzer. Therefore, variations in [A tot ] (i.e., total protein) are not considered within the equation that was employed to predict the resulting outlet pH. Equation 2. Calculated pH in the outlet of the dialyzer based on measured values of pH, SID, and pCO 2 , adapted from [21]:

Statistics
Experiments were performed at different settings between 3 and 6 times each covering a wide range of operational parameters. Student's t test for paired samples was used to compare CO 2 removal and acid supply between different ADVOS settings for experimental sets 1 and 2, respectively. A two-tailed p value lower than 0.05 was considered to indicate statistical significance. For correlations assessment, Pearson's coefficient was employed. Data were documented and analyzed using Microsoft Excel and IBM SPSS 24.0 for Windows®, respectively. Data are presented as mean ± standard deviation (SD)

Influence of ADVOS multi operational settings on CO 2 removal
In this experimental design, where blood was titrated with CO 2 to maintain a blood pH between 7.35 and 7.45, CO 2 removal with ADVOS multi depended on three variables: (1) the amount of CO 2 being supplied, (2) the blood flow, and (3) the dialysate composition (i.e., according to carbonate concentration and dialysate pH setting) ( Fig. 2 and Additional file 1: Table S1). During experiments with BC-Bic20 and Q c of 160 ml/min, higher Q b resulted in higher CO 2 elimination with an average of 77 ± 22 ml/min at Q b 400 ml/min with a dialysate of pH 9.0. With the same dialysate pH, 35 and 19 ml/min of CO 2 were removed at Q b 200 and 100 ml/min, respectively. A lower dialysate pH setting (i.e., lower sodium and higher chloride) resulted in lower CO 2 removal, independently of any other setting. In fact, the dialysate composition (i.e., the presence of albumin and sodium and chloride concentrations) together with the dialysate pH (but not dialysate pH alone) is responsible for blood pH correction. At a Q c of 160 ml/min compared to 320 ml/min, setting dialysate pH to 9.0 resulted in significantly higher blood HCO 3 − (57.0 ± 10.2 vs. 38.5 ± 2.8 mmol/l) and pCO 2 (103 ± 17 vs. 74 ± 4 mmHg), which is correlated with the higher amount of CO 2 removed (Additional file 1: Table S1). Increasing the Q c from 160 ml/min to 320 ml/min means doubling the convective transport, which results in faster removal of substances diffused from blood (e.g., bicarbonate). Therefore, at higher concentrate flows, a more efficient concentration gradient between blood and dialysate is available, resulting, in this case, in lower blood baseline levels of HCO 3 − with a Q c of 320 ml/min in comparison to a Q c of 160 ml/min. Lower blood HCO 3 − can buffer less acid in blood (i.e., less CO 2 can be supplied without altering blood pH), which, due to the lower total CO 2 concentration in blood (i.e., lower pCO 2 at the inlet of the dialyzer), it was translated in a lower CO 2 removal in our experimental setting. Nevertheless, this is an artefactual result caused by the experimental design as not Q c , but Q b , CO 2 supply, and dialysate pH setting affect the CO 2 removal capacity.
In line with these results, using a carbonate-free dialysate BC-Bic0 with dialysate pH set to 10, a maximum removal of 142 ml/min CO 2 was achieved during ADVOS treatment with Q b of 400 ml/min and Q c of 160 ml/min. However, with this setting, pCO 2 and HCO 3 − were extremely above physiological levels (117 ± 5 mmHg and 62.8 ± 3.4 mmol/l, respectively). Thus, blood was then titrated with CO 2 only as long as every blood gas value was maintained within a physiological range (including pCO 2 and HCO 3 − ). This setting allowed a CO 2 removal of 61 ml/min using a Q b of 400 ml/min and a Q c of 320 ml/min. This was only possible with a basic concentrate without carbonate. In fact, similar CO 2 removal rates were achieved with the same flows and with dialysate pH 9 with the BC-Bic20 concentrate (58 ml/min), but blood gas levels were above physiological values in this case ( Fig. 2 and Additional file 1: Table S1). Variations between the inlet (pre) and the outlet (post) of the dialyzers for SID, pCO 2 , HCO 3 − , and pH are shown in Additional file 1: Table S1.

Influence of ADVOS multi operational settings on blood pH during continuous acid load
Experiments with ADVOS multi employing different operational settings showed that higher blood flows and dialysate pH settings were able to allow higher acid loads. The influence of each parameter is shown in Fig. 3. These results show a low level of CO 2 removal needed (< 1 mmol/min or 22.5 ml/min) to maintain pCO 2 stable (35-45 mmHg) in those cases where CO 2 elimination is not required.
Variations between the inlet (pre) and the outlet (post) of the dialyzers for SID, pCO 2 , HCO 3 − , and pH are shown in Additional file 1: Table S2. and not only blood pH. Mean ± SD.1, p < 0.05 for CO 2 removal between consecutive dialysate pH settings among the same blood flow; 2, p < 0.05 for CO 2 removal between consecutive blood flows among the same dialysate pH setting; 3, p < 0.05 for CO 2 removal between consecutive concentrate flows among the same blood flow and dialysate pH setting

ADVOS multi vs. CVVHD during continuous maximal CO 2 supply
Using a carbonate-free dialysate with a pH of 10, ADVOS multi removed 79 ± 1 ml/min TCO 2 on average during 4 h of continuous supply of 110 ml/min CO 2 . Blood pCO 2 (66 ± 9 mmHg) and HCO 3 − (33.1 ± 0.1 mmol/l) levels were stable throughout the experiments. The ADVOS device was able to maintain pH stable and within the physiological range (Fig. 4) while post-dialyzer pH remained always below 8. In contrast, despite identical CO 2 influx, blood pH dropped already 15 min after circulating blood through the NIKKISO renal dialysis device reaching a constant pH of around 6.60 after 1 h. Post-dialyzer pH remained below 7 during the experiment.

Treatment of metabolic acidosis in vitro: ADVOS multi vs. CVVH
Once metabolic acidosis was triggered (minute 40, Fig. 5), bicarbonate therapy during CVVH was able to normalize HCO 3 − levels. However, as expected, due to the lack of ventilation, pCO 2 was correspondingly elevated (> 90 mmHg) resulting in an even lower pH (< 7.00) after the treatment. Conversely, ADVOS multi normalized pH and bicarbonate levels in less than 1 h. Due to the high CO 2 removal ability of the ADVOS system when dialysate pH is set to 9.0, even 15 ml/min of CO 2 was additionally provided during the treatment with ADVOS multi to maintain blood pH between 7.35 and 7.45.

Treatment of hypercapnic acidosis in vitro with ADVOS multi
Using a basic concentrate without Na 2 CO 3 (BC-Bic 0), ADVOS multi was able to restore a hypercapnic acidosis in vitro in less than 30 min using a Q b of 200 ml/min, a Q c of 160 ml/min, and dialysate pH set to 9.0 during a continuous supply of 27 ml/min of CO 2 . After changing the settings (dialysate pH 7.8 vs. 9.0) and the basic concentrate (BC-Bic 20 vs. BC-Bic 0), values of pH (7.12 vs. 7.35), pCO 2 (99 vs. 40 mmHg), and HCO 3 − (32.7 vs. 22.6 mmol/l) returned to physiological standards (Fig. 6). Fig. 3 Total acid load (CO 2 + lactic acid) with different operational settings during ADVOS multi treatments. CO 2 and lactic acid were supplied to maintain pCO 2 and blood pH between 35-45 mmHg and 7.35-7.45, respectively. A supply of 1 mmol/l of CO 2 corresponds to 22.5 ml/min in normal conditions. 1, p < 0.05 for CO 2 removal between consecutive dialysate pH settings among the same blood flow; 2, p < 0.05 for CO 2 removal between consecutive blood flows among the same dialysate pH setting.

Acid-base balance according to Stewart
As shown in Additional file 1: Tables S1 and S2, the higher the pCO 2 reduction, the higher the pH increase that can be achieved in blood, describing a direct correlation for more than 200 blood samples in both experimental sets 1 and 2 (Additional file 2: Figure S1 r 2 = 0.812 and Additional file 2: Figure S2 r 2 = 0.935, respectively).
To note, to increase the dialysate pH, the ratio "basic concentrate/acidic concentrate" increases. Attending to the composition of these solutions, an increase in the basic/ acidic concentrate ratio results in a dialysate with higher sodium and lower chloride concentrations. Indeed, an increase in SID was observed for set 1 (Additional file 1: Table S1), but not for set 2 due to the lactate addition (Additional file 1: Table S2).
Finally, we were able to predict pH variations solely by measuring SID and pCO 2 changes (Additional file 2: Figures S1 and S2; Additional file 3: Figures S3 and S4; Additional file 4: Figures S5 and S6). The calculated values correlated perfectly with measured values (r 2 = 0.98). Variations in total protein are not expected since no albumin loss occurs in the dialyzer.

Discussion
In the present study, the ability of an albumin hemodialysis system (ADVOS multi) to correct hypercapnic and lactic acidosis in vitro has been demonstrated. Different settings can be varied in this device (blood flow, concentrate flow, carbonate content, and dialysate pH), allowing different rates of CO 2 removal or acid load. This is only possible due to the presence of albumin in the dialysate, which permits to alter the composition of the dialysate (including the strong ion difference and the CO 2 content). A concentrate gradient between blood and dialysate for electrolytes or bicarbonate is then possible, allowing the correction of acidosis from hypercapnic or metabolic origin.

Lung support through CO 2 removal
The lung removes CO 2 directly, thanks to the fast transformation of H + + HCO 3 − through carbonic anhydrase into CO 2 gas and water. However, in blood, CO 2 is mainly found as HCO 3 − . ADVOS reduces pCO 2 through the removal of HCO 3 − by forming a concentration gradient between blood and dialysate. Additionally, the high dialysate pH helps to reduce the H + concentration in blood and thus increase the pH. This is possible due to the presence of albumin in the dialysate. Albumin increases the buffer capacity via a protonation of its imidazole side chain [22], which contains several buffering residues of histidine [23,24]. Preliminary in-house studies indicate that a dialysate of pH 9 containing two 100 ml bottles of albumin 20% increased the buffer capacity by 35% compared to the same dialysate without albumin (Additional file 5: Figure S7). Bearing this in mind, first, a higher blood flow may account for a higher HCO 3 − concentration gradient between blood and dialysate (i.e., 400 ml/min). Second, a higher dialysate pH (with a higher SID) allows a higher decrease in H + concentration (i.e., ADVOS and CVVH. A metabolic acidosis was triggered in blood reaching baseline values before treatment of pH < 7.2, HCO 3 − < 14 mmol/l, and pCO 2 of 45 mmHg (preparation phase). Then, for 1 h, either a conventional hemofiltration using a commercially available substitution fluid with 35 mmol/l bicarbonate or a treatment with ADVOS multi with a dialysate pH of 9.00 was performed. Lactic acid was continuously supplied to maintain lactate levels over 5 mmol/l dialysate pH 10). Third, lower (or none) dialysate carbonate levels permit a more effective convective transport in the ADVOS multi circuit (Fig. 1). Finally, this convective transport will be faster insofar a higher concentrate flow is set (i.e., 320 ml/min) ( Fig. 2 and Additional file 1: Table S1). Even at lower blood and concentrate flows (200 and 160 ml/min, respectively) and with dialysate pH set to 9, the use of a dialysate without carbonate allowed a correction of hypercapnic acidosis in vitro in less than 1 h (Fig. 6).

Kidney support through HCO 3 − generation
The renal compensatory mechanism during a respiratory acidosis tries to increase the acid excretion into urine and the HCO 3 − resorption into blood [25]. When using a bicarbonate containing dialysate with high pH, this mechanism is mimicked by ADVOS. This can be explained following the CO 2 equilibrium in Eq. 3. Carbonic acid, or CO 2 in its gas form, is converted to HCO 3 − and H + . The reduction of H + concentration in blood forces the equilibrium to the right of the equation, increasing HCO 3 − even as it is removed from blood, and consequently, CO 2 is transferred down its concentration gradient from the intracellular space (i.e., correcting intracellular acidosis) into the blood and into the dialysate in the form of HCO 3 − . This means that in the absence of adequate ventilation, as simulated in our ex vivo model through elevated CO 2 supply without additional oxygenation, ADVOS multi could "imitate" the renal compensatory mechanism for acidemia control [26].
In the case of ADVOS, the higher the dialysate pH setting, the higher the H + concentration reduction is achieved (Fig. 2). This results into HCO 3 − generation, which helps to correct HCO 3 − levels during metabolic acidosis (Fig. 5). This is only possible if a concomitant pCO 2 reduction is achieved, which does not occur during conventional renal replacement therapy or bicarbonate infusion. In fact, in the absence of adequate ventilation (i.e., CO 2 removal), a metabolic acidosis can turn into a hypercapnic acidosis (Fig. 5), which cannot be corrected with conventional CVVHD (Fig. 4). The Stewart model to explain acid-base balance with the ADVOS system We analyzed if the observed changes could also be explained by the mathematical model proposed by Stewart [20] and its revision by others [27][28][29], who showed that three independent variables are responsible for determining the pH in plasma: PaCO 2 , plasma weak acids (i.e., phosphate and albumin), and the SID as the difference between fully dissociated plasma anions and cations. Therefore, neither the H + movement nor the buffering effect of HCO 3 − is necessary or sufficient to explain acid-base regulation. In any case, our data can be also explained using this model, as shown in Additional files. Taking into account the conclusions obtained by Constable [21], blood pH could be predicted solely by changes in total protein, pCO 2 , and SID. Applying this to our data demonstrated a perfect correlation between the measured and the calculated pH in the outlet of the dialyzer (Additional file 2: Figures S1 and S2; Additional file 3: Figures S3  and S4; Additional file 4: Figures S5 and S6). Indeed, SID variations were correlated with dialysate pH variations, specifically at high values of 10.0. To reach such a dialysate pH, a higher rate of basic/acidic concentrate is needed, which provides higher Na + and lower Cl − , and can therefore result into SID reductions in blood through the dialyzer. Moreover, the presence of albumin as a weak acid facilitates this process.
Although SID and pCO 2 are considered independent variables by Stewart, it has been suggested that an interdependency between both values might exist [30]. Langer et al. observed that the greater the variation in pCO 2 , the greater the reduction in plasma SID. We might reach the same conclusion when, as this group did, quartiles of pCO 2 variations are analyzed and plotted against the corresponding SID variations (r 2 = 0.991; Additional file 6: Figure S8). However, if raw data are drawn, no correlation is observed (r 2 = 0.190; Additional file 7: Figure S9).

Rationale for multi organ support with ADVOS during acidosis
Using either the classical or the modern approach, this work should serve as a proof of concept of the ability of the ADVOS therapy to correct acid-base disturbances. As described above, the lungs (i.e., CO 2 ) and kidneys (i.e., NH 4 + , HCO 3 − , for the classic approach or Na + , Cl − for the Stewart model) are usually defined to be responsible for acid-base control. However, the liver plays also an important role (i.e., metabolism of organic acid anions like citrate and certain amino acids) [31,32] and can also be supported by ADVOS. Indeed, acidemia and metabolic acidosis are associated with poor outcome in cirrhosis patients, as demonstrated by Drolz and colleagues in a cohort of 178 critically ill patients with liver cirrhosis and acute on chronic liver failure [33].Therefore, attention should not only be paid to a specific organ. In addition, the majority of the cases of acidosis reflect a mixed nature, involving both a metabolic and a respiratory component [2]. In view of this, a multiple organ approach seems to be needed while facing acidosis, where the variety of adjustable parameters of the ADVOS multi might play an important role.

Limitations and justification of the ex vivo model
Although our results are encouraging, our work is limited by its in vitro nature, the selection of parameters, and the number of experiments performed. Nevertheless, so as to serve as a proof of concept, this experimental setting is adequate based on the following: (1) the different parameters analyzed and varied (i.e., pCO 2 , HCO 3 − , lactate) to resemble different types of acidosis, (2) the possibility to control the concrete amount of acid load (i.e., CO 2 and/or lactic acid) being supplied, and (3) the analysis of inlet and outlet measurements to describe the course of blood values along the dialyzer. These encouraging results need now to be confirmed in the clinical setting.

Conclusions
In conclusion, the ADVOS albumin hemodialysis system was able to remove 61 ml/min while maintaining blood gas values in the physiological range or up to 142 ml/min CO 2 in hypercapnic conditions at low blood flow without the need of a gas phase. Blood pH was maintained stable within the physiological range of 7.35-7.45 by the albumin-containing dialysate. Moreover, during continuous lactic acid addition, up to 3 mmol/min of acid load was compensated. The major determinants to stabilize blood pH were blood flow, dialysate composition, and blood bicarbonate levels. The mechanism of action of ADVOS multi can be explained using either the classical acid-base balance model or the newer Stewart approach. This feature in combination with the previously demonstrated ability to eliminate water-soluble and protein-bound toxins may be of valuable help in the management of critically ill patients with multiple organ failure.