Metabolic acidosis of hypoxic states or anaerobic exercise has been traditionally explained by lactate production. Nevertheless, there is biochemical evidence that lactate production does not cause acidosis, but retards its development 12. During anaerobic metabolism, protons derived from ATP hydrolysis that cannot be reutilized in oxidative phosphorylation might be the actual explanation for metabolic acidosis 12. Nevertheless, there is some evidence showing that aerobic lactate production (that is, during catecholamine administration) is associated with metabolic acidosis 34.

Whichever mechanism produces acidosis, increased lactate production coincides with cellular acidosis, and remains a good indirect marker for cell metabolic conditions that induce metabolic acidosis. Many studies have consequently established the use of blood-lactate levels as a diagnostic, therapeutic, and prognostic marker of tissue hypoxia in circulatory shock 5. In addition, lactic acidosis is the most frequent cause of metabolic acidosis 6 and one of the most common metabolic abnormalities in critically ill patients 5. Moreover, Gunnerson and colleagues demonstrated a higher mortality in critically ill patients with lactic acidosis than in patients with hyperchloremic acidosis 7. For a correct diagnostic and prognostic evaluation of critically ill patients, therefore, severe metabolic acid–base disorders such as lactic acidosis must be identified.

Lactic acidosis is primarily suspected because of the presence of metabolic acidosis. Nevertheless, [HCO₃^-] and base excess ([BE]) levels might be normal despite the presence of hyperlactatemia, as a result of simultaneous alkalinizing processes. Accordingly, Fencl and colleagues showed that, in 152 critically ill patients, Stewart's approach could detect metabolic acidosis in some patients with normal [HCO₃^-] and [BE] levels 8. In those patients, the metabolic acidosis with a low strong-ion difference ([SID]) was counterbalanced by alkalinizing processes 8.

Although the lack of correlation of hyperlactatemia with pH, [HCO₃^-], and [BE] values has been previously reported, these reports have not used a systematic approach to understand the underlying metabolic acid–base disorders 910111213. The objective of the present investigation was to study a large series of critically ill patients with high lactate levels and to quantitatively analyze the presence of alkalinizing processes that might neutralize the decrease of [BE], and thus occult metabolic disorders. Our hypothesis was that the metabolic acidosis associated with hyperlactatemia could be frequently hidden by the effect of alkalinizing processes that neutralize [BE].

Severe hyperlactatemia was present in 168 of the patients (11%). One-hundred and thirty-four (80%) patients had low [BE] values while 34 (20%) did not.

Clinic, epidemiologic, and outcome data are presented in Table 1. Both groups had similar values of the Acute Physiologic and Chronic Health Evaluation II score, the predicted and actual mortality, the Sepsis-related Organ Failure Assessment score, and the McCabe score. Patients with low [BE] were more frequently associated with shock and surgical admission. Chronic obstructive pulmonary disease and medical admission were more commonly found in patients with normal [BE].

Patients with low [BE] received more saline solution before ICU admission (1,000 (500 to 2,000) versus 0 (0 to 500) ml, P = 0.0004). There were no differences in the volume of Ringer-lactate solution received (0 (0 to 1,000) versus 0 (0 to 0) ml, P = 0.18). Twenty-one percent of the patients in each group received diuretics before ICU admission (P = 0.97).

Levels of lactate were slightly higher in patients with low [BE] (6.4 ± 2.4 mmol/l versus 5.6 ± 2.1 mmol/l, P = 0.08). According to the study design, the pH, [HCO₃^-], and [SID] levels were lower in patients with low [BE] (Figures 1 and 2). Patients with normal [BE] had lower [Cl^-]_corrected (Figure 2) and higher Δ[AG]_corrected - [HCO₃^-] and Δ[AG]_corrected - Δ[BE] values (5 ± 6 mmol/l versus 1 ± 4 mmol/l and 3 ± 6 mmol/l versus 4 ± 4 mmol/l, respectively; P < 0.0001 for both). These patients also had levels of [AG]_corrected and [SIG]_corrected that were slightly lower (21 ± 5 mmol/l versus 23 ± 5 mmol/l, P = 0.07 and 9 ± 5 mmol/l versus 11 ± 5 mmol/l, P < 0.05, respectively). The [albumin] and [A_tot^-] values were lower in patients with low [BE] (Figure 2).

Since the normal reference intervals for plasma [Cl^-]_corrected are 103 to 111 mmol/l in our laboratory, most of the patients in the normal [BE] group had absolute hypochloremia (65% of the patients). Conversely, most of the patients in the low [BE] group had normal [Cl^-]_corrected levels (90% of the patients).

The normal [BE], [HCO₃^-], and [SID] levels were almost completely explained by the alkalinizing effect of hypochloremia. A quantitative analysis shows that the differences in mean [BE], [HCO₃^-], and [SID] values between both groups were about 8 mmol/l (lower in the low [BE] group) while the difference in the [Cl^-] level was 7 mmol/l.

The main finding of the present study was that 20% of the patients with severe hyperlactatemia showed normal pH, [HCO₃^-], [BE], and [SID] levels because of the simultaneous presence of hypochloremic metabolic alkalosis.

Madias reported previously that in lactic acidosis the increase in the [AG] might be occasionally greater than the decrease in the corresponding [HCO₃^-] 26. This finding indicates the diagnosis of a mixed metabolic disorder 27. The actual incidence of mixed metabolic disorder in patients with severe hyperlactatemia has not been previously described. To our knowledge, the present study is the first to systematically address this issue and quantify the underlying metabolic acid–base.

Acid–base disorders might be characterized by different methods. First, by a traditional approach in which the metabolic component of acid–base physiology is assessed by analysis of [HCO₃^-] levels 27. The evaluation of the metabolic component might be further completed by the inclusion of [BE] 28. Despite considerable argument about which of these parameters is better 29303132, both are usually employed in clinical practice and their calculations are included in all blood-gas analyzers. The [AG] constitutes an additional diagnostic contribution 23, although hypoalbuminemia might decrease the usefulness of this parameter. For this reason, many researchers have recommended adjusting the [AG] for the albumin level 253334353637.

An alternative approach is the application of basic physicochemical principles of aqueous solutions to blood. Stewart identified variables that primarily and independently of one another determine the pH 38: PCO₂, the [SID] (that is, the difference between the sums of all the strong cations and all the strong anions), and the [A_tot^-]. Using this approach, Fencl and colleagues have shown that the traditional analysis frequently failed to identify severe disturbances such as metabolic acidosis 8. In that study, a low [SID] was undetected through changes in [BE] because the low [SID] acidosis was masked by the alkalinizing effect of hypoalbuminemia present in all patients.

As we previously shown 14, however, the combined use of [HCO₃^-], [BE], and [AG]_corrected allowed the same acid–base diagnosis. Accordingly, in the group with normal [BE], normal values of pH, [HCO₃^-], and [BE] matched with a normal [SID]. The diagnosis of mixed metabolic acidosis and alkalosis was performed by the presence of positive Δ[AG]_corrected - Δ[HCO₃^-] and Δ[AG]_corrected - Δ[BE] levels 27 in the traditional analysis, and by increased [SIG] and low [Cl^-]_corrected values in Stewart's approach 8. Our data reinforce the concept that acid–base analysis only based on pH, [HCO₃^-], and [BE] values might be frequently misleading. An adequate diagnosis should rely on a more comprehensive approach that might include the use of [AG]_corrected or chloride levels.

Different to previous studies in which hypoalbuminemia was the confounding factor in the interpretation of acid–base data 8, the alkalinizing factor in the present study was hypochloremia. Albuminemia and the [A_tot^-] value were lower in patients with low [BE] – which might be due to the presence of shock, a condition that increases extravascular albumin losses 39. McAuliffe and colleagues described 'primary hypoproteinemic alkalosis' in hypoalbuminemic ICU patients with positive [BE] and elevated [HCO₃^-] levels 40. Nevertheless, the actual role of hypoalbuminemia to produce metabolic alkalosis has been recently challenged 1440. We previously could only detect one patient fulfilling the criteria of primary hypoproteinemic alkalosis among 700 hypoalbuminemic patients 14. Wilkes showed that the loss of weak acid secondary to hypoproteinemia is compensated by a renal-mediated increase in [Cl^-], so the [SID] decreases without changes in pH 41.

The presence of hypochloremic metabolic alkalosis in patients with normal [BE] can be related to the high number of patients with chronic obstructive pulmonary disease. In these patients, hypochloremia is the consequence of an appropriate kidney response to chronic respiratory acidosis 42. Despite the prior administration of diuretics being similar both groups, in some patients the diuretics might have contributed to the development of hypochloremia.

Although most of the patients in the normal [BE] group had absolute hypochloremia and most of the patients in the low [BE] group had normal [Cl^-] levels, a mechanism other than hypochloremic alkalosis might have contributed to the differences in [Cl^-] between both groups. Since shock was more frequently present in patients with low [BE], these patients received more aggressive fluid resuscitation before ICU admission. Consequently, a subtle component of hyperchloremic metabolic acidosis might be present in this group 43.

In a study of patients admitted to the ICU after cardiac arrest, lactic acidosis was the most frequent disorder. Its effects on the pH, however, were attenuated by the presence of a metabolic alkalosis caused by hypochloremia and hypoalbuminemia. The authors explained these findings by extravascular passage of albumin, and pre-existing disease 44.

The presence of metabolic acidosis in critical patients has prognostic implications. Gunnerson and colleagues recently demonstrated that patients with metabolic acidosis ([BE] < -2 mmol/l) had a higher mortality rate than those without this disorder (45% versus 25%). Similarly, patients with lactic acidosis had higher mortality than those with hyperchloremic acidosis (56% versus 29%) 7. Nevertheless, it is not clear whether differences in outcome are dependent on the process that produces metabolic acidosis or on the acidosis itself.

Although our two groups of patients had quite different pH values, their mortality was not different. A possible explanation for this observation might be that the severity of critical illness, as evaluated by the Acute Physiologic and Chronic Health Evaluation II and Sepsis-related Organ Failure Assessment scores, was similar in both groups. Recent data showed that metabolic acid–base variables had a poor discriminating ability for predicting mortality in a general ICU population. Areas under receiver operating characteristic curves for acid–base parameters were significantly lower than that of the Sepsis-related Organ Failure Assessment score 14. Nevertheless, a higher number of patients are required to confirm that the presence of acidemia itself does not worsen the outcome.

The present study has some limitations. The study is observational, aimed at describing the incidence of severe hyperlactatemia with normal [BE] and quantifying its underlying acid–base alterations. Nevertheless, patients were only evaluated on ICU admission. Serial measurements might have allowed a more comprehensive understanding of the acid–base disorders and allowed better insights into the mechanisms of acid–base disorders.

Our results suggest that 20% of critically ill patients have severe hyperlactatemia with normal pH, [HCO₃^-], and [BE] levels because of a concomitant presence of hypochloremic alkalosis. Both the conventional and Stewart approaches allow the identification of this mixed metabolic disorder. The results also suggest the evaluation of plasma [Cl^-] and Δ[AG]_corrected - Δ[HCO₃^-] levels should always be considered for a correct diagnosis of acid–base disorder.

• Twenty percent of critically ill patients have severe hyperlactatemia with normal pH, [HCO₃^-], and [BE] levels because of a concomitant presence of hypochloremic alkalosis.

• As previously shown, both the conventional and Stewart approaches allow the correct identification of mixed metabolic acidosis and alkalosis.

• The evaluation of plasma chloride and the difference between the changes in the anion gap and bicarbonate should always be considered for a correct diagnosis of acid–base disorders.

[AG] = anion gap; [A_tot^-] = total concentration of plasma nonvolatile buffers; [BE] = base excess; [HCO₃^-] = bicarbonate concentration; PCO₂ = partial pressure of carbon dioxide; [P_i] = inorganic phosphate concentration; [SID] = strong-ion difference; ICU = intensive care unit; [SIG] = strong-ion gap.

The authors declare that they have no competing interests.

GT and MCP mainly contributed to the conception and design of the study. GT, MCP, FDM and DOK performed acquisition of data, and contributed to the analysis and interpretation of data. AD drafted the manuscript and performed the statistical analysis. All authors read and approved the final manuscript.