Background: The measurement of plasma unmeasured anions (PUA) is paramount in assessing metabolic acid base disorders. The aim of this study was to compare the accuracy of three methods in diagnosing abnormal PUA values: standard base excess (SBE), the albumin corrected anion gap (AGc), and the Stewart-Figge approach, based on unidentified anions (XAc-). Methods: Acid-base variables were prospectively collected in ICU patients admitted January-September 2008. Whatever the method, PUA values measured two standard deviations above or below the mean of those in control subjects were considered as abnormal. Results: Of the 205 consecutives patients included, 179 had an abnormal PUA value. The accuracy of AGc and XAc- in diagnosing abnormal PUA values was comparable (AUC: 0.89±0.03 and 0.89±0.03, P=0.82) but greater than that of SBE (0.67±0.06, P<0.001 and P<0.001, respectively). Of the high PUA values (n=161), 96% were diagnosed by XAc-, 88% by AGc (P<0.01) and 48% by SBE (P<0.001). Hyperlactataemia (n=111) was diagnosed equally by AGc and XAc-, (81% and 86%, P=0.37) but less by SBE (50%, P<0.001 and P<0.001, respectively). High PUA not associated with hyperlactataemia (n=61) was more frequently diagnosed by XAc -(97%) than by AGc (84%, P=0.03). Conclusions: In ICU patients, AGc and the Stewart-Figge approach should be preferred over SBE for diagnosing abnormal PUA values and predicting hyperlactataemia. The Stewart-Figge approach based on unidentified anions, is the most efficient in diagnosing high PUA values not associated with hyperlactataemia.