Unmeasured anions identified by the Fencl-Stewart method predict mortalitybetter than base excess, anion gap, and lactate in patients in the pediatric intensive care unit

Objectives: This study was undertaken to compare three methods for the iden tification of unmeasured anions in pediatric patients with critical illness . We compared the base excess (BE) and anion gap (AG) methods with the less commonly used Fencl-Stewart strong ion method of calculating BE caused by unmeasured anions (BEua). We measured the relationship of unmeasured anions identified by the three methods to serum lactate concentrations and to mor tality. Design: Retrospective cohort study. Setting: Tertiary care pediatric intensive care unit in an academic pediatr ic hospital. Patients: The study population included 255 patients in the pediatric inten sive care unit who had simultaneous measurements of arterial blood gases, e lectrolytes, and albumin during the period of July 1995 to December 1996. S ixty-six of the 255 patients had a simultaneous measurement of serum lactat e. Measurements and Main Results: The BEua was calculated using the Fencl-Stew art method. The AG was defined as (sodium plus potassium) - (chloride plus total carbon dioxide). BE was calculated from the standard bicarbonate, whi ch is derived from the Henderson-Hasselbalch equation and reported on the b lood gas analysis. A BE or BEua value of less than or equal to-5 mEq/L or a n AG greater than or equal to 17 mEq/L was defined as a clinically signific ant presence of unmeasured anions, A lactate level of greater than or equal to 45 mg/dL was defined as being abnormally elevated for this study. The p resence of unmeasured anions identified by significantly abnormal BEua was poorly identified by BE or AG. Of the 255 patients included in the study, 6 7 (26%) bad a different interpretation of acid base balance when the Fend m ethod was used compared with when BE and AG were used. Plasma lactate conce ntration correlated better with BEua (r(2) = .55; p = .0001) than with AG ( r(2) = .41; p = .0005) or BE (r(2) = .27; p = .025). Mortality was more str ongly related to BEua less than or equal to-5 mEq/L (relative risk of death = 10.25; p = .002) than to lactate greater than or equal to 45 mg/dL (rela tive risk of death = 2.35; p = .04). In logistic regression analysis, morta lity was more strongly associated with BEua (area under the receiver operat ing characteristic curve = 0.79; p = .0002) than lactate (receiver operatin g characteristic curve area = 0.63; p = .05). BE (receiver operating charac teristic curve area = 0.53; p = .32), or AG (receiver operating characteris tic curve area = 0.64; p = .08) in this patient sample. Conclusions: Critically ill patients with normal BE and normal AG frequentl y have elevated unmeasured anions detectable by BEua. The Fencl-Stewart met hod is better than BE and similar to AG in identifying patients with high l actate levels. Elevated unmeasured anions identified by the Fencl-Stewart m ethod were more strongly associated with mortality than with BE, AG, or lac tate in this patient sample.