Determination of Base Excess by Blood Gas Analyzers: Assessment of Errors

A clinically useful way to reflect the nonrespiratory portion of acid-base balance is the calculation of base excess (BE). However, this calculation varies among contemporary analyzers and depends on pH and pCO2. Thus, base excess may not truly reflect the acid-base status. The goal of the present study was to assess such errors, including the impact of the oxygen saturation.


The study included seven blood gas analyzers (AVL Compact 2, Ciba-Corning 860, IL 1620, Nova Stat Profile 9, Nova Stat Profile Ultra, Radiometer ABL 500 and Radiometer ABL 510). The algorithms integral to the analyzers were compared with the equations suggested in literature. Subsequently, arterial and venous blood samples were obtained from twenty patients undergoing cardiac surgery and measured on one analyzer per manufacturer (n = 700). Additionally, arterial and venous blood samples obtained from healthy volunteers were analyzed with and without equilibration using four different gas mixtures (low / high O2, CO2, and N2 concentrations) (n=415). The results obtained by the calculations provided by the manufacturers and integral to the devices were compared to results from a unified equation, which included compensation for incomplete oxygen saturation (Haldane effect). Data processing and statistical analysis were performed automatically using two standard personal computer systems in parallel.


Analysis of the formulas showed, that there is a linear relationship between the partial pressure of carbon dioxide (pCO2) and both the base excess of whole blood (BEbl) and the base excess of extracellular fluid (BEecf). The relationship between pH and both BEbl and BEecf has an exponential and a linear component. Radiometer, however, uses an algorithm that was designed to correspond precisely with the readings from the traditional acid-base nomogram of Siggaard-Andersen. The formulas for calculating BEecf were all variations of formulas for BEbl, using a constant and low concentration of hemoglobin.  At normal pCO2 and normal or elevated pH, BEbl roughly equals BEecf. At elevated pCO2, the calculation of BEecf yields higher results in comparison to BEbl; at reduced pCO2, the results are also comparably lower. At moderately pathologic values, typical deviations are in the range of 2 mmol/L.

The formulas for calculating BEecf used in the different analyzers diverge by up to 0,7 mmol/L at the standard values pH = 7.4 (cH+ = 40 nmol/L) and pCO2 = 40 mmHg (5.33 kPa). At metabolic alkalosis (pH = 7,65 und pCO2 = 50 mmHg), however, the differences between formulas amount to nearly 5 mmol/L.

Standard deviation at repeated measurement of BEecf was between 0,21 and 0,53 mmol/L for six of the seven analyzers, but 1,79 mmol/L for the remaining one (Nova Stat Profile 9). The coefficients of variation for pH were between 0,07 % and 0,15 %, those for pCO2 between 1,3 % and 6,8 %.

Systematic differences of up to 4,24 mmol/L were found between the average BEecf-values (65 parallel measurements) when the formulas specific to each analyzer were used. Unified calculation of BEecf reduced this systematic difference to 3,07 mmol/L.

BEecf of the venous samples was on average 0,6 mmol/L (n = 468, SD = 0,82, 95%-CI: 0,53 to 0,68 mmol/L) higher than BEecf of the arterial samples, according to the respective algorithms. This difference was reduced to 0,1 mmol/L by unified calculation with correction for the Haldane effect.

Equilibration of venous blood effected a fall in BEecf by 2,2 mmol/L in average (95%-CI 1,7 to 2,7 mmol/L) when BE was corrected for the Haldane effect. Significant differences between the different gas mixtures could not be determined. If the Haldane effect was not considered, differences between the test gases of up to 3,3 mmol/L were found. A significant change in BE after storage of the samples on ice water during 6 ½ hours did not occur.


The calculations integral to the devices were more (Ciba-Corning, Radiometer, Instrumentation Laboratory, Radiometer, Nova Biomedical) or less (AVL) consistent with those provided by the handbooks. Random deviations of the base excess from other measurements on the same sample occurred occasionally and covered a range of 6.3 to 0.7 mmol/L depending on the device. The deviations encountered were less pronounced (p < 0,05) with the Radiometer devices. The clinical challenge is that these errors cannot be related to the time the measurement was obtained nor is it attributable to specific conditions. The deviations occurred instantaneously and in a non-reproducible fashion.

Concentration of hemoglobin was determined by three of the analyzers (Nova Stat Profile 9, Nova Stat Profile Ultra and Radiometer ABL 510). The coefficients of variation of the measurements on patients’ blood (22.1, 9.5, and 14.5 % respectively) were acceptable for the calculation of base excess; for detection of blood losses, however, they would be insufficient. cHb-measurements on native and tonometered blood of the healthy volunteers showed considerably lower coefficients of variation (Nova Stat Profile Ultra: 2.3 %; Radiometer ABL 510: 1.7 %). Measurement of pH also yielded higher coefficients of variation on all analyzers for the samples from patients (0,11 %) in comparison to the samples from healthy volunteers (0,08 %). This means that measurements for the evaluation of analyzing equipment should be performed using blood of humans in various clinical conditions.

The coefficients of variation of pCO2 was increased on all analyzers (from 1.6 to 2.3 % on average during a period of 101 days). This effect was particularly marked for the AVL analyzer (from 1.3 to 2.6 %). However, the two machines that were replaced (Radiometer, Nova Biomedical) determined base excess with more precision than their predecessors. This may be interpreted as the result of aging processes. The coefficients of variation of the pH increased less and not on all analyzers (from 0.10 to 0.12 %).

The maximum difference of BEecf between the results of the three analyzers that were not replaced was 1.0 mmol/L (mean difference of 116 measurements of each analyzer) at the beginning of the measurements. Unified calculation of BEecf reduced this difference to 0,5 mmol/L. About 100 days later, this difference had increased to 4.6 mmol/L using the specific algorithms of the analyzers and 3.7 mmol/L using the common formula. This was mainly the consequence of divergence in pH-measurement and may be the result of insufficient stability of the buffers used for calibration of the pH-electrode.

It could be shown that compensation of BEecf for the Haldane effect can avoid errors of up to 3 mmol/L. In the unified formula that had been originally used for this study, the Haldane effect was quantified as 0,02 mmol H+ per g of desoxyhemoglobin. The results of this study suggest that a value of 0,03 mmol H+ per gram is more adequate. Quantification of the Haldane effect independent of the BE-formula was not possible in this study, however, since the BE-difference between venous and arterial samples is also depending on the particular formula without compensation for the Haldane effect. Additionally, the difference is markedly smaller when calculating BEbl than for BEecf since an increase in pCO2 (in venous blood) causes BEbl to increase less than BEecf. This might be one reason why the correction for the Haldane effect has not been generally accepted for the calculation of BEbl yet; another reason might be that in some critical health conditions venous BE has been observed to be lower than arterial BE.

Because of the relevant inter-analyzer differences regarding base excess measurement, an improvement of quality control in blood gas analysis seems necessary. Measurements on the analyzers used in this study with buffer solutions especially designed to present a constant BE-value (a combination of HEPES buffers with physiological pK’) had discouraging results because they did not correlate well with measurements on samples of whole blood. Therefore, quality control by systematic measurements of fresh blood on different analyzers and continuous control of reference intervals still seems to be more useful to detect systematic errors. For these measures of quality control, the correction of BE for the Haldane effect also has the advantage, that both arterial and venous blood samples of human beings without disturbance of the acid-base equilibrium have an expected average BE of 0 mmol/L.

In conclusion, the current findings indicate that a potentially true or untrue laboratory result for the base excess may complement the clinical picture, thus reinforcing the concept of clinical diagnoses and reliable quality control.