Acid–Base Changes Caused by 5% Albumin versus  6% Hydroxyethyl Starch Solution in Patients Undergoing Acute Normovolemic Hemodilution: A Randomized Prospective Study

We studied 20 female patients without apparent cardiac, pulmonary, or renal disease (American Society of Anesthesiologists class I or II) scheduled for radical hysterectomy because of cancer of the cervix. An expected mean surgical blood loss of about 1,500 ml (about 35% of the patient’s blood volume) and a reported transfusion rate between 22 and 92% were the reasons for using preoperative ANH as a blood-saving method. 4–6Before surgery, written informed consent was obtained from each patient, and the protocol was approved by the ethics committee of our institution. The patients were randomly assigned to receive either 6% hydroxyethyl starch solution (HES group: n = 10) or 5% albumin solution (HA group: n = 10). The HES solution contained 154 mm of sodium and 154 mm of chloride (pH 5.44; Fresenius AG, Bad Homburg, Germany), and the HA solution contained 158 mm of sodium, 150 mm of chloride, 4 mm of octanoate, and 4 mm of N¹-acetyltryptophanate (pH 6.86; Centeon Pharma GmbH, Marburg, Germany). Neither solution contained bicarbonate or citrate. These solutions are commonly used in Europe; however, their electrolyte compositions may differ from those of colloids used in the United States.

General anesthesia was induced using intravenous thiopental, sufentanil, and cisatracurium and was maintained with isoflurane 0.4–1.5 vol% in an oxygen–nitrous oxide mixture at a 1:1 ratio; additional doses of sufentanil and cisatracurium were administered as appropriate. Radial arterial and central venous catheters were inserted. Mechanical ventilation was performed to maintain arterial oxygen tension (Pao²) at 250–300 mmHg and arterial carbon dioxide tension (Paco²) as close as possible to 40 mmHg. Intraoperative monitoring consisted of end-tidal Paco², electrocardiography, central venous pressure, arterial blood pressure, pulse oximetry, and esophageal temperature. During hemodilution and surgery, the patient’s temperature was kept constant by warming infusion fluids and using a warming blanket.

All measurements were performed before starting surgery during periods of stable anesthesia and hemodynamics. Before ANH, no intravenous infusions were given, and during ANH, no crystalloids were administered. Simultaneous measurements of hematocrit (centrifugation [12,000/min; 4 min] without correction for plasma trapping), PV, and red cell volume were carried out (see Determination of Plasma Volume and Determination of Red Cell Volume). At the same time, arterial blood samples were analyzed for Pao², pH, and Paco²using standard electrodes; concentrations of serum sodium (Na⁺), potassium (K⁺), chloride (Cl⁻), and ionized calcium (Ca²⁺) using ion-elective electrodes; and lactate (Lac⁻) using an enzymatic method (quantification of H²O²), all integrated in a Radiometer analyzer (Radiometer ABL 620 GL, Radiometer Co., Copenhagen, Denmark). Additionally, serum phosphate (PO⁴), determined by ultraviolet photometry of a phosphomolybdate complex,serum total protein concentration, determined using the Biuret method, and albumin concentration, determined using colorimetry of bromocresol complex, were measured from the same blood samples. The amount of weak plasma acid was calculated with the aid of a computer program developed by Figge et al.  3and was termed [Prot⁻]. Standard serum bicarbonate concentration [Bic] and standard base excess [BE] were determined with a blood gas analyzer that uses the Henderson-Hasselbalch equation and the Siggaard-Andersen nomogram. 7 

After baseline measurements, blood was removed at a rate of about 60 ml/min (following gravity) and was simultaneously replaced with HA or HES at nearly the same rate. A previous study in our laboratory had shown normovolemia after ANH with a 15% surplus of HA in relation to the amount of blood removed. 8Consequently, we planned to infuse 15% more colloid in relation to blood removed. Initially, approximately 500 ml/m²of blood were removed. The hemodilution bags were weighed on a precision scale so that the volume of blood withdrawn could be evaluated immediately. For fine tuning, frequent determinations of hematocrit were performed at the end of ANH to reach a target value of 22%. The hemodilution procedure took about 20 min. After completion of ANH and a steady-state interval of 20 min without any further infusions and before starting surgery, measurements of all variables mentioned previously were performed in the same manner as before ANH.

Immediately before each dye injection, a calibration curve was constructed using the patient’s blood with two known indocyanine green concentrations (Paesel, Frankfurt, Germany; 1.25 and 2.5 μg/ml of whole blood). At the same time, blood samples were taken to determine hematocrit (centrifugation of blood samples without correction for plasma trapping; variation coefficient < 2%) and serum total protein (Biuret method; variation coefficient < 2%). Next, 0.25 mg/kg of body weight indocyanine green was injected into the central venous catheter as a bolus (within 5 s). From 2–5 min after injection, blood was continuously withdrawn from the arterial catheter at a rate of 15 ml/min (consequently, without a relevant impact on PV) through a cuvette attached to a densitometer by means of a calibrated pump in a closed system. The pump, the cuvettes, and the densitometer were constructed by one of the authors (H.B.). Optical density of the blood (corrected for blank) was read in the densitometer at 800 and 900 nm. The density of blood at injection time was derived by monoexponential extrapolation of the density curve between 2–5 min back to zero time (the time of injection). This value, when put in the calibration curve, gives CBo, the theoretical whole-blood concentration of the dye at injection time, which is the initial volume of distribution.

The theoretical plasma concentration of the dye at injection time (CPo) was calculated as: CPo = CBo/(1 − Hct). PV was calculated as: PV = D/CPo, where D is the amount of dye injected.

This method is referred to as the “whole-blood method” for PV determination, methodologic aspects of which have been published previously. 9The whole-blood method provides reproducible results in the operating room within 10 min (mean difference and variation coefficient between double measurements: 0.3 and 6.2%, respectively). 10 

This method, using autologous red cells stained with sodium fluorescein and determination by flow cytometry, was developed in our laboratory and has been published in detail elsewhere. 11,12Briefly, per measurement, 20 ml of the patient’s blood were labeled with 50 mg of sodium fluorescein and injected into a central vein. After injection, samples were drawn from the arterial catheter at 4, 6, and 8 min, stored on ice, and analyzed by flow cytometry (Becton Dickinson, Heidelberg, Germany).

Red cell volume (in milliliters) was calculated according to the following equation: RCV = (RCi × Vi × Hctp)/(RCp × FRCf), where RCi indicates number of injected red cells per milliliter of tagged cell suspension, Vi indicates volume of injected cell suspension in milliliters, Hctp indicates hematocrit of the patient’s blood, RCp indicates number of red cells per milliliter in the patient’s blood, and FRCf indicates fraction of fluorescent red cells determined by flow cytometry.

FRCf was taken as the mean value of determinations of samples drawn at 4, 6, and 8 min after injection, counting in triplicate the number of fluorescent red cells per 50,000 cells with the aid of the flow cytometer. RCi and RCp were obtained using a cell counter (Coulter Electronics, Miami, FL). In our laboratory, mean difference and variation coefficient for red cell volume double measurements with sodium fluorescein were 0.6 and 3.1%, respectively. 11 

Before and after ANH, for each sample, a “bedside” strong ion difference (SID) was calculated as follows:

The anion gap was calculated as follows:

The ECV before ANH was calculated as follows:

ECV after ANH was derived as follows:

The amount of plasma removed was determined by weighing the empty and full hemodilution bags and measuring hematocrit of blood contained in the respective bags:

where 110 g is the sum of weight (40 g) and of CPDA1 fluid content (70 g) of empty ANH bags. The specific gravity of blood was taken to be 1.05. 13 

The calculated amount of bicarbonate or chloride in ECV before ANH was obtained as follows:

The amount of bicarbonate or chloride removed during ANH was derived as follows:

This equation involves the premise that the concentration of the respective electrolyte in plasma removed was the mean of the respective electrolyte plasma concentration before and after ANH.

The amount of chloride supplied was calculated as follows:

Because both colloids used were free from bicarbonate, no bicarbonate was supplied during ANH.

The predicted amount of bicarbonate in ECV after ANH was calculated as follows:

The predicted amount of chloride in ECV after ANH was calculated as follows:

The predicted concentration of bicarbonate or chloride in ECV after ANH was calculated as follows:

The calculated amount of bicarbonate or chloride in ECV after ANH was derived as follows:

The predicted change in [Bic] during ANH according to the Stewart approach 1was calculated as follows:

This equation demands that Paco²remains constant.

As all measured and calculated data described previously were normally distributed (tested by Kolmogorov-Smirnov tests), these are presented as mean values with standard deviations. For demographic data, Student t  tests for unpaired data were performed. Student t  tests for paired or unpaired sampleswere performed as appropriate to compare intragroup and intergroup differences before and after ANH. A value of P < 0.05 for intergroup differences before ANH, a value of P < 0.002 for intergroup differences after ANH, and intragroup differences (α-adjustment according to Bonferroni: 0.05/25) concerning acid–base variables were considered to be statistically significant.

Patient characteristics are presented in table 1. There were no significant differences in weight, height, body surface area, and the time of infusion. Because of slightly higher preoperative body surface area and hematocrit values in the HA group, the amount of blood removed and amount of colloid infused was significantly higher than in the HES group. As shown in table 2, there were no differences in blood volume, PV, red cell volume, or hematocrit before and after ANH between both groups. Total protein and plasma albumin concentration decreased during ANH with HES, whereas total protein remained unchanged and plasma albumin concentration significantly increased during ANH with HA. The only intergroup difference was in serum albumin concentration; the mean value after ANH in the HES group was exactly half that in the HA group.

In table 3, pH, Paco², [Bic], [BE], [Lac⁻], [Na⁺], [Cl⁻], [K⁺], SID, [Prot⁻], [PO⁴], [Ca²⁺], and the anion gap at the measuring points before and after ANH are shown. At baseline, no intergroup differences in these variables were found. Both groups had a significant decrease in pH from about 7.39 to 7.34, and Paco²did not show major deviations from 40 mmHg in either group. In the HES group, [Bic] and [BE] significantly decreased from 23.6 to 21.6 mm and −0.9 mm to −3.4 mm, respectively, after ANH. In the HA group, [Bic] significantly decreased from 23.6 to 21.0 mm and [BE] significantly decreased from −1.1 mm to −4.0 mm after ANH. [Lac⁻] did not change in either groups. There were no relevant changes in [Na⁺] and [K⁺]; however, [Cl⁻] significantly increased in both groups during ANH. This increase was larger after application of HES (+6 mm) than after HA (+3 mm). SID decreased by a mean of 5.3 mm in the HES group and did not change significantly in the HA group (−1.6 mm). As a result of dilution with an albumin-free solution, [Prot⁻] decreased by about 5 mm in the HES group in contrast to the HA group, in which [Prot⁻] increased slightly but significantly during ANH. [Ca²⁺] and [PO⁴] levels remained fairly constant and did not show any relevant difference within or between groups. The anion gap significantly decreased in the HES group, whereas it did not significantly change in the HA group. After ANH, significant intergroup differences were found for SID, [Prot⁻], and the anion gap.

Table 4shows the bicarbonate and chloride balance for ECV, provided that ECV is the distribution space of the respective electrolytes; calculated and predicted values in table 4are based on mean values in table 1 and 3; for calculation, see Methods, equations 3–12. Table 4shows that in both groups the predicted mean amounts of bicarbonate and chloride in ECV after ANH (based on a balance of removed and supplied electrolytes) differed only slightly from the respective calculated mean values (based on measured [Bic] and [Cl⁻] after ANH). Additionally, in both groups the mean values of predicted [Bic] and [Cl⁻] after ANH were only slightly different from the respective measured mean values (table 3). This may indicate that the measured changes in [Bic] and pH in both groups are simply caused by the dilution of ECV with bicarbonate-free solutions. Table 5presents a comparison between measured changes in [Bic] and predicted changes in [Bic], which were derived from the changes in SID and [Prot⁻] during ANH according to the Stewart approach (see equation 13. 1The measured decrease in [Bic] during ANH was well explained in the HA group, whereas the Stewart equations 1did not explain the measured decrease in [Bic] in the HES group.

The main finding of this study was a slight but significant metabolic acidosis after ANH with both HA and HES solutions. In the current investigation, the double label measurements of blood volume showed that ANH was isovolemic in both groups. The acidosis in both groups seems to be without major clinical relevance. However, the generated data could constitute a further step toward a better understanding of acid–base equilibrium. What are the mechanisms behind this acidosis? The acidosis definitely was of metabolic origin, because Paco²was kept almost constant and [Bic] and [BE] decreased significantly, but it was not a lactic acidosis. The term “dilutional acidosis” might be appropriate, because in both groups, about 35% of blood volume was replaced by intravenous infusions free of bicarbonate and ECV increased by about 600–800 ml as a result of ANH. Originally proposed by Peters and van Slyke, 2this term was generally used in some older publications, 14–16recent case reports, 17–20and several letters to the editor 21–25in the context of diluting not only PV but the entire ECV with different electrolyte solutions that were free of bicarbonate. This would result in a dilution of serum bicarbonate, with resulting metabolic acidosis. Table 4shows a balance for extracellular bicarbonate that was based on this “dilution hypothesis.” In both groups, about 20 mmol of bicarbonate were removed without substitution (see equation 7. This should result in a predicted decrease in the amount of extracellular bicarbonate (equation 9). The mean calculated amount of extracellular bicarbonate after ANH in both groups, which was based on measured [Bic] after ANH (equation 12), was only slightly different from the respective predicted value, which was based on balance calculations (equation 9). Figure 1shows this comparison by means of a regression analysis and a Bland-Altman plot. One patient in the HA group was marked separately with an asterisk, because with 110 kg of body weight, the respective ECV could not be quantified with sufficient precision. This patient was not included in the calculations of linear regression and mean difference. The correlation coefficient of 0.99 and the slope of 1.03 indicate the excellent correlation between predicted and calculated amount of bicarbonate in ECV after ANH. Figure 2presents a comparison between predicted [Bic] after ANH (see equation 10and the respective measured [Bic]. In this case also, there is good correlation and a regression slope close to 1. The fact that, according to figures 1 and 2, this simple mathematical balance of extracellular bicarbonate gives such precise bicarbonate concentrations after ANH almost in every patient in each group leads to the conclusion that changes in bicarbonate in the course of ANH are caused by the dilution of ECV with bicarbonate-free solutions. The balance for chloride of ECV, which is also presented in table 2, shows an almost similar result. As a result of the large chloride concentrations in the colloid solutions used, the net chloride balance during ANH in both groups was more than 100 mmol (equations 7 and 8). The predicted mean increase in the amount of extracellular chloride (equation 10) as well as the predicted mean increase in [Cl⁻] (equation 11) were also in close agreement with the respective mean calculated or measured values. These results support the conclusion that changes in serum anions and cations in the course of ANH are simply caused by the dilution of ECV. The “dilution hypothesis” takes electrolyte concentrations in PV and ECV (which should be similar), such as bicarbonate or chloride, and the amount of water in ECV into account. However, the effect of changes in plasma albumin concentration on changes in acid–base equilibrium is not represented in this theory.

There is an alternative to explain the metabolic acidosis observed. The Stewart approach, which is discussed in detail elsewhere, 26defines Paco², SID, and the sum of all anionic charges of weak plasma acids (called A^totby Stewart and calculated as the product of total protein and the empirically derived factor 2.43 according to van Slyke et al.  27) as independent pH-regulating variables, whereas pH and [Bic] are dependent variables. 1Apart from a few studies in exercise medicine, 28,29experimental medicine, 30and critical care medicine, 31–35The Stewart approach has not been commonly applied in the past decade. The Stewart approach postulates that a decrease in SID or an increase in [Prot⁻] will result in a decrease in pH on condition of constant Paco²(and vice versa; see equation 13. 1By defining [Bic] as being exclusively dependent on SID and [Prot⁻] during metabolic acid–base changes, the traditional “dilution hypothesis” is poorly considered by the Stewart approach. 1A previous investigation by our laboratory demonstrated an ongoing hyperchloremic metabolic acidosis after rapid intraoperative infusion of 0.9% saline in contrast to an infusion of lactated Ringer’s solution in patients undergoing gynecologic surgery. 36In both groups, the different changes in SID and [Prot⁻] were in good agreement with the predicted changes in [Bic] according to the Stewart approach. 1However, in this study, each infusion regimen involved a similar intraoperative decrease in total protein, so that the impact of hypoproteinemia on acid–base balance could not be quantified adequately. 37Recent investigations gave a new quantification of the negative charges of serum proteins and therefore their quality as anions. 3,38,39Figge et al.  3demonstrated that nonalbumin proteins may be neglected, because they do not bear negative charges and consequently do not act as anions. 39Calculations of [A^tot] or [Prot⁻] using the van Slyke factor therefore will result in an overestimation of the negative charges of plasma proteins. In the current study, [Prot⁻] was calculated using a computer program developed by Figge et al.  3 

Table 5illustrates changes in [Bic], SID, and [Prot⁻] during ANH in both groups. In the HA group, [Prot⁻] increased and SID decreased slightly by about 1.5 mm each, so that the measured decrease in [Bic] by a mean of 2.5 mm was in good agreement with the predicted decrease in [Bic] according to the Stewart equations. 1In contrast, both SID and [Prot⁻] decreased by about 5 mm in the HES group, so that the measured decrease in [Bic] by a mean of 2.5 mm could not be explained adequately by the model of Figge et al.  3This observation seems to be important, and the reasons behind it should be clarified. The mathematical model by Figge et al.  3is based on electrical neutrality, the dissociation equilibrium for water and the carbonic acid system, and the dissociation equilibrium for all ionizable groups on the albumin molecule. It solves [Prot⁻] as a function of SID, Paco², plasma albumin concentration, and [PO⁴]. In this study, the latter three values are measured variables; however, the calculation of SID was different from that done by Figge et al.  3In the current study, SID was calculated as: SID =[Na⁺]+[K⁺]−[Cl⁻]−[Lac⁻] (see equation 1. All components of this equation are variables measured by the Radiometer analyzer; thus, SID can be calculated quickly in the operating room. Figge et al.  3calculated an apparent SID using the following equation: SIDapp =[Na⁺]+[K⁺]+[Mg²⁺]+[Ca²⁺]−[Cl⁻]. They demonstrated an excellent correlation between SIDapp and an effective SID, which was predicted by pH, Paco², plasma albumin concentration, and [PO⁴] according to the Stewart approach. 1,Figure 3shows the comparison between our “bedside” SID and the effective SID predicted by the computer program. Because there is a good correlation between these two SID values, we conclude that the apparent SID by Figge et al.  3should also be in good correlation with our “bedside” SID. The computer program by Figge et al.  3also gives a predicted pH as a function of SID, Paco², plasma albumin concentration, and [PO⁴]. Figure 4illustrates that the preoperatively measured pH values were in good agreement with the predicted pH values obtained using the Figge et al.  3model (slope 1.03). However, in both groups, the changes in pH during ANH were not predicted precisely, as shown in figure 5by the respective slopes of the regression lines of 0.80 in the HES group and 1.27 in the HA group. This indicates that the computer program did not predict the pH values after ANH with satisfying precision. The reason for this observation is unclear. One explanation could be that after ANH, [Prot⁻] was not estimated correctly by the computer program of Figge et al.  3Especially concerning the HA group, the dissociation of some amino acids of the albumin molecule, and therefore anionic charges of albumin, possibly could have been altered after ANH as a result of a disturbed chemical equilibrium in serum. 3,38,39Some mechanisms could be the presence of different anionic charges on the synthetic albumin molecule or changes in binding ions. However, the discrepancy between the predicted constancy in [Bic] during ANH in the HES group according to the Stewart approach 1and the measured decrease in [Bic] (table 5) leads to the assumption that, in this group, the Stewart approach fails.

The main link between the Stewart approach 1and the “dilution hypothesis” is found in the changes in SID. SID depends on the dilution of the respective cations and anions in ECV. However, if hydroxyethyl starch is used as a volume substitute, a second compartment, namely PV, must be considered. The large decrease in [Prot⁻] after having infused an artificial colloid that is free of albumin should be independent of and disproportional to the dilution of ECV. This is different from in vitro  experiments with plasma samples. In this setting, only one distribution space (the plasma sample in vitro ), which determines the concentrations of both the proteins and the electrolytes, must be considered. It has already been demonstrated that the Stewart equations work well under these conditions. 3,38,39However, it seems possible that the different distribution spaces of colloids and electrolytes in vivo  are not considered sufficiently by the Stewart approach. 1Further systematic analysis of this topic will be valuable.

After having replaced 35% of blood volume during ANH with HA or HES solutions, a slight metabolic acidosis was observed. The reason for this acidosis in both groups was the dilution of bicarbonate in ECV and a concomitant hyperchloremia. Some indications were found that the impact of the artificial decrease in [Prot⁻] on pH due to the infusion of an albumin-free colloid could not be quantified adequately by the Stewart equations 1and the Figge et al.  3model. Acidosis in the course of ANH possibly could be avoided by a more “physiologic” composition of electrolytes in colloid solutions, such as that in lactate-buffered solutions, which unfortunately are not yet available in Europe. 40Further investigations, perhaps with these “new” colloid solutions, could provide more information regarding the impact of isolated hypoproteinemia on acid–base equilibrium.

The authors thank Vladimir Fencl, M.D., Respiratory Division, Brigham and Women’s Hospital, Boston, Massachusetts, for allowing them to work with his computer program.