To the Editor:—
The problem of method-related errors, arising from different measuring technologies on calculation of the anion gap and the strong ion difference, was clearly demonstrated in patients in a recent study by Morimatsu et al. 1Mean values differed significantly, in some cases leading to a diverging assessment of the acid-base or electrolyte status. To a clinician, however, this is not acceptable.
All quantities, including the base excess (BE), used for diagnosis of nonrespiratory acid-base disturbances (e.g. , metabolic, renal, or intestinal) are calculated quantities. Hence, each quantity must be assessed for reliability and associated diagnostic interpretation (normal range and normal values: mean ± SD; pathologic range). This greatly depends on the inaccuracy and imprecision of each of the measured primary values used in the particular calculation and on how these are propagated. BE is usually obtained with arterial blood from measured pH, partial pressure of carbon dioxide (Pco2), and hemoglobin-concentration in a blood gas analyzer, the strong ion difference from measured plasma electrolyte concentrations (e.g. , Na, K, Cl, lactate).
For clinical practice, we propose the whole blood BE, which should be preferably used depending on the following considerations:
Inaccuracy of calculated BE: If correctly calculated, BE can be obtained from measured pH, Pco2, oxygen saturation, and total hemoglobin in any blood sample (venous or arterial). Over the whole range (−30 to +30 mmol/l), mean inaccuracy is less than 1 mmol/l. 2
Normal values for BE and variability: Normal values for BE in arterialized capillary blood of men (n = 20) are −0.1 ± 1.2 mmol/l, and of women (n = 20), −1.0 ± 1.1 mmol/l. 3Variability of BE in healthy individuals is very low and could be reproduced if calculated also from measurement in venous blood. Typical results (mean ± SD) as obtained with blood from the vena cubitalis (50 healthy volunteers: colleagues and medical students) are shown in Table 1. Values for pH, Pco2, partial pressure of oxygen, and oxygen saturation were measured in a blood gas analyzer (AVL OMNI 9; Roche Diagnostics, Mannheim, Germany), calibrated with two precision buffers for pH (Radiometer phosphate buffer S 1500; S 1510), and the mean calculated BE was −0.1 ± 1 mmol/l. It should be noted that the SD of BE is still low, even though measured values used in the calculation varied largely (Table 1).
Tolerable values of inaccuracy: Representative normal values and maximal inaccuracy of the measured acid-base and electrolyte variables in blood, as tolerated by the recommendations of the German Medical Association (Bundesärztekammer 2002, Köln, Germany) for the electrolyte concentrations in the plasma are Na 142 ± 2.8 mmol/l (2.0%); K 4.5 ± 0.2 mmol/l (3.7%); Cl 103 ± 4.1 mmol/l (4%); and lactate 1.5 ± 0.1 mmol/l (6%). Using these figures, the normal value of the strong ion difference (Na + K − Cl − lactate) is calculated as 42 ± 5.0 mmol/l, with high inaccuracy from propagation of errors. Hence, reliability is strongly reduced, dominated by the largest errors (Na:± 2.8 mmol/l; Cl:± 4.1 mmol/l).
In comparison, normal BE is 0 ± 2.2 mmol/l when calculated from normal values of pH 7.40 ± 0.02, Pco240 ± 1.6 mmHg (4%), hemoglobin 15 ± 0.3 g/dl (2%), and full oxygen saturation ∼ 100%. Reliability of BE is mainly affected by the inaccuracy of measurement for pH and Pco2, whereas that for hemoglobin is negligible.
BE as a prognostic factor: The diagnostic use and prognostic value of BE is well documented. Among 10 clinical hemodynamic and 20 blood laboratory parameters tested, change in BE proved to be the best predictor of blood volume changes in a canine hemorrhagic shock model. 4In critically ill patients BE has been established as an independent predictor of mortality 5and endpoint of resuscitation. On the basis of 8,200 polytrauma patients, statistically selected out of 15,200 from four clinical studies, mortality increased significantly with a decrease in BE, e.g. , approximately by 25% if the assessed BE was −6 mmol/l within the first 24 h after admission (see figure 4 in reference 6). 6
Conclusions: Because mortality may increase considerably (∼ 8%) when BE is decreased by only 2 mmol/l, the clinical requirements for reliability (inaccuracy, imprecision) should be high (≤ 2 mmol/l). By using present standard measuring technology, this is met only for the BE and not for the strong ion difference or lactate alone. The latter should be used only as secondary parameters in combination with the BE for a differential diagnosis, e.g. , nonrespiratory acidosis correlated to lactate, hyperchloremia, or anion gap. 7