BACKGROUND The Stewart approach theorizes that plasma pH depends on P(aCO₂), the strong ion difference, and the plasma total concentration of non-volatile weak acids (A(tot)). The conventional approach measures standardized base excess, bicarbonate (HCO₃⁻), and the anion gap. OBJECTIVE To describe acid-base disorders with the Stewart approach and the conventional approach in patients with chronic respiratory failure. METHODS This was an observational prospective study in a medical intensive care unit and a pneumology ward of a university hospital. There were 128 patients included in the study, of which 14 had more than one admission, resulting in 145 admissions. These were allocated to 4 groups: stable respiratory condition and elevated HCO₃⁻ (Group 1, n = 23), stable respiratory condition and non-elevated HCO₃⁻ (Group 2, n = 41), unstable respiratory condition and elevated HCO₃⁻ (Group 3, n = 44), and unstable respiratory condition and non-elevated HCO₃⁻ (Group 4, n = 37). Elevated HCO₃⁻ was defined as ≥ 3 standard deviations higher than the mean value we found in 8 healthy volunteers. Measurements were taken on admission. RESULTS In groups 1, 2, 3, and 4, the respective mean ± SD values were: HCO₃⁻ 33 ± 3 mM, 26 ± 3 mM, 37 ± 4 mM, and 27 ± 3 mM (P < .001); strong ion difference 45 ± 3 mM, 38 ± 4 mM, 46 ± 4 mM, and 36 ± 4 mM (P < .001); and A(tot) 12 ± 1 mM, 12 ± 1 mM, 10 ± 1 mM, 10 ± 2 mM (P < .001). Non-respiratory disorders related to high strong ion difference were observed in 12% of patients with elevated HCO₃⁻, and in none of those with non-elevated HCO₃⁻ (P = .003). Non-respiratory disorders related to low strong ion difference were observed in 9% of patients with non-elevated HCO₃⁻, and in none of those with elevated HCO₃⁻ (P = .02). Hypoalbuminemia was common, especially in unstable patients (group 3, 66%; group 4, 65%). Normal standardized base excess (16%), HCO₃⁻ (28%), and anion gap (30%) values were common. The Stewart approach detected high effective strong ion difference in 13% of normal standardized base excess, and in 20% of normal anion gap corrected for albuminemia, and low effective strong ion difference in 22% of non-elevated HCO₃⁻. CONCLUSIONS In patients with chronic respiratory failure the acid-base pattern is complex, metabolic alkalosis is present in some patients with elevated HCO₃⁻, and metabolic acidosis is present in some with non-elevated HCO₃⁻. The diagnostic performance of the Stewart approach was better than that of the conventional approach, even when corrected anion gap was taken into account.