Unmeasured anions in critically ill patients: can they predict mortality?

Crit Care Med. 2003 Aug;31(8):2131-6. doi: 10.1097/01.CCM.0000079819.27515.8E.


Objective: To determine whether base excess, base excess caused by unmeasured anions, and anion gap can predict lactate in adult critically ill patients, and also to determine whether acid-base variables can predict mortality in these patients.

Design: Retrospective study.

Setting: Adult intensive care unit of tertiary hospital.

Patients: Three hundred adult critically ill patients admitted to the intensive care unit.

Interventions: Retrieval of admission biochemical data from computerized records, quantitative biophysical analysis of data with the Stewart-Figge methodology, and statistical analysis.

Measurements and main results: We measured plasma Na+, K+, Mg2+, Cl-, HCO3-, phosphate, ionized Ca2+, albumin, lactate, and arterial pH and Paco2. All three variables (base excess, base excess caused by unmeasured anions, anion gap) were significantly correlated with lactate (r2 =.21, p <.0001; r2 =.30, p <.0001; and r2 =.31. p <.0001, respectively). Logistic regression analysis showed that the area under the receiver operating characteristic (AUROC) curves had moderate to high accuracy for the prediction of a lactate concentration >5 mmol/L: AUROC curves, 0.86 (95% confidence interval [CI], 0.78-0.94), 0.86 (95% CI, 0.78-0.93), and 0.85 (95% CI, 0.77-0.92), respectively. Logistic regression analysis showed that hospital mortality rate correlated significantly with Acute Physiology and Chronic Health Evaluation (APACHE) II score, anion gap corrected (anion gap corrected by albumin), age, lactate, anion gap, chloride, base excess caused by unmeasured anions, strong ion gap, sodium, bicarbonate, strong ion difference effective, and base excess. However, except for APACHE II score, AUROC curves for mortality prediction were relatively small: 0.78 (95% CI, 0.72-0.84) for APACHE II, 0.66 (95% CI, 0.59-0.73) for lactate, 0.64 (95% CI, 0.57-0.71) for base excess caused by unmeasured anions, and 0.63 (95% CI, 0.56-0.70) for strong ion gap.

Conclusions: Base excess, base excess caused by unmeasured anions, and anion gap are good predictors of hyperlactatemia (>5 mmol/L). Acid-base variables and, specifically, "unmeasured anions" (anion gap, anion gap corrected, base excess caused by unmeasured anions, strong ion gap), irrespective of the methods used to calculate them, are not accurate predictors of hospital mortality rate in critically ill patients.

Publication types

  • Evaluation Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Acid-Base Equilibrium
  • Acidosis, Lactic / etiology*
  • Acidosis, Lactic / mortality*
  • Acidosis, Lactic / pathology
  • Aged
  • Bicarbonates / blood
  • Blood Gas Analysis / methods*
  • Calcium / blood
  • Carbon Dioxide / blood
  • Chlorides / blood
  • Data Interpretation, Statistical
  • Female
  • Hospital Mortality
  • Humans
  • Hydrogen-Ion Concentration
  • Intensive Care Units
  • Lactic Acid / blood
  • Magnesium / blood
  • Male
  • Medical Records
  • Middle Aged
  • Phosphates / blood
  • Potassium / blood
  • Predictive Value of Tests
  • Retrospective Studies
  • Sensitivity and Specificity
  • Serum Albumin
  • Sodium / blood
  • Victoria / epidemiology


  • Bicarbonates
  • Chlorides
  • Phosphates
  • Serum Albumin
  • Carbon Dioxide
  • Lactic Acid
  • Sodium
  • Magnesium
  • Potassium
  • Calcium