INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Blood lactate concentration is a parameter widely used in intensive care units to assess tissue oxygenation. Severe hyperlactatemia during shock is a standard indicator of cellular hypoxia (1), and a blood lactate > 5 mmol/L is usually associated with a poor outcome (2). There is frequently a mild persistent hyperlactatemia during sepsis in hemodynamically stable patients (5, 6). This is widely interpreted, by extrapolation, as evidence of lactate overproduction due to the presence of occult cell hypoxia during sepsis (7). The oxygen debt is generally believed to be due to a defect in oxygen extraction and enhanced oxygen demand (11, 12). However, several lines of evidence refute the existence of occult tissue hypoxia in stable septic patients. The first is that increasing oxygen delivery to hemodynamically stable septic patients improves neither mortality nor morbidity (13). The second is that stable septic patients are frequently not dependent on the oxygen supply (16, 17), even those that have elevated lactate levels (18). Lastly, tissue PO₂ and energy stores seem to be unaffected by sepsis (19). There could be other reasons of lactate overproduction, such as increased glycolysis due to an insulin-like activity of endotoxin, or enhanced catabolism of muscle alanine via alanine aminotransferase (6). The hyperlactatemia that occurs during sepsis could also be evidence of a decrease in blood lactate clearance. This impaired lactate utilization may be caused by altered liver function, which is common during sepsis (20) or due to a metabolic abnormality. Experimental studies have shown that pyruvate dehydrogenase activity is disturbed during sepsis (21) and that hyperlactatemia is more pronounced when acute circulatory failure is triggered by injecting endotoxin (22).

Thus, the significance of mild hyperlactatemia during sepsis remains unclear. Lactate could be overproduced or underutilized because of a metabolic pathway failure. The present study was performed to evaluate the contributions of lactate production and clearance to the concentration of blood lactate during sepsis in hemodynamically stable, critically ill patients.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients

This study was conducted in the intensive care unit of the Saint-Roch University Hospital, Nice, France, from October 1994 to April 1996, and was approved by the local ethics committee. Informed written consent was obtained from each patient or their closest relative. The patients were eligible for study if they satisfied the following criteria: fever > 38.0° C or white cell count > 10,000/mm³ or < 5,000/mm³; presence of a documented deep infection site; presence of at least one of the following conditions: acute lung injury (ratio Pa_O2/FI_O2 < 300), acute renal failure (creatinine clearance < 50 ml/min), heart rate > 120 beats/min, or alteration of mental status in the absence of direct brain injury; blood lactate concentration < 4 mmol/L; no inotropic or vasopressor support.

All patients were invasively monitored with a pulmonary artery catheter (Baxter, Edwards Critical-Care Division, Irvine, CA). Cardiac output was measured by thermodilution with an Explorer laboratory cardiac output computer (Baxter) by injecting 10 ml ice-cold 5% dextrose at the end of expiration. The mean of four consecutive determinations was recorded as effective cardiac output. An arterial catheter was placed in the radial or femoral artery for pressure measurements and blood sampling.

The patients were given a fluid load until the pulmonary artery occlusion pressure was > 12 mm Hg. Patients were excluded from the study if systolic blood pressure was < 90 mm Hg despite a pulmonary artery occlusion pressure  15 mm Hg. The patients were allowed to equilibrate for at least 2 h before two blood lactate measurements were performed 60 min apart. Patients were excluded from the study if the two lactate concentrations differed by > 5%. An arterial blood sample was taken before measuring lactate clearance and production for measuring blood gases (BGS 288; Ciba-Corning, Cergy-Pontoise, France) and liver and renal function tests. The patient's hemodynamic and metabolic parameters were recorded, and the APACHE II score was calculated at entry into the study.

Measurements of Lactate Production and Clearance

Lactate clearance was measured by a method similar to that previously described (23). A precise amount of sodium lactate was infused to create a transient hyperlactatemia, and the resulting change in plasma lactate was analyzed. Sodium L-lactate (1 mmol/kg, 1 M solution; Aguettant, Lyon, France) was infused via a central venous catheter over 15 min (T0 to T15) using a peristaltic pump (IVAC 560; IVAC, Rueil-Malmaison, France). Arterial blood samples were taken for lactate measurement via the arterial catheter 5 min and just before (T5 and T0), during (T5, T10, and T15 min), and after the lactate infusion (T16, T17, T18, T19, T20, T21, T23, T25, T27, T29, T31, T33, T35, T40, T45, and T55 min) into 5-ml tubes containing 12.5 mg sodium fluoride and 10 mg potassium oxalate (Vacutainer-Hemogard; Becton-Dickinson, Meylan, France). The samples were immediately placed on ice and centrifuged at 4° C. Lactate concentration was determined enzymatically (Ektachem; Johnson & Johnson, Les Ulis, France) by spectrophotometric reflectance. The normal lactate concentration range using this analyzer is 0.3 to 1.2 mmol/L and the mean variation of the lactate assay is about 1%, with a standard deviation of nearly 0.04 mmol/L.

A mathematical model describing the change in lactate concentration was constructed using the Apis software package (INSERM U 278; A. Iliadis, Paris, France) (24). This method assumes that the rate of L-lactate elimination is linear while the lactate concentration remains within a certain range (25). The method is based on analysis of the lactate curve using the least squares method with semi-logarithmic coordinates. The best model was identified by successive approximations using an algorithm of minimizations. This procedure was conducted using one- and two-compartment models for each patient. The model was selected as described below (see DATA ANALYSIS) and used to estimate blood lactate clearance, half-life and distribution volumes of infused lactate. The validity of the mathematical construction was tested by comparing the concentration of L-lactate predicted by the model with the observed results. The areas under the curve formed by plotting the equation of the mathematical model and those formed by the lactate kinetics were analyzed by comparing the lactate clearance obtained from the model with the model-independent lactate clearance (LC_mi) obtained from the formula:

(1)

where D = amount of L-lactate infused and AUC = area under the curve calculated by the trapezoid method when the increase in blood L-lactate concentration is plotted against time from 0 to .

Lactate production was then estimated by assuming that a steady blood lactate concentration means that lactate production equals lactate elimination (23). Lactate production was thus calculated as:

(2)

where blood lactate is the concentration of lactate before the lactate infusion.

Data Analysis

The one- or two-compartment model was selected by comparing their fit qualities (described by the pooled sum of deviations squares) using an F-test. The F value was calculated as: F_calc = [J₁  J₂]/J₂ · [n  p₂]/ [p₂  p₁] (where J and p represent the quality of fit and the number of parameters of each model, and n the number of observations). This F value, which is distributed according to the F-law with p₂  p₁ and n  p₂ degrees of freedom, was then compared with the theoretical F value. The two-compartment model was adopted only in case of statistical significance.

The theoretical relationship between lactate clearance and concentration (for a given production) was made linear by using the reciprocal of the clearance. The relationship linking two continuous variables was tested by simple linear regression using the least squares method. Continuous variables were compared by Student's t test for unpaired data. Values of p < 0.05 were regarded as significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A group of 34 patients, in which the mortality rate was 56%, was included in the study; the demographic data are shown in Table 1. Twelve patients required catecholamines to maintain arterial blood pressure during the 3 d preceding the study (but all these patients were weaned from catecholamines for at least 12 h at the time of the study); 19 patients had sepsis due to a hospital-acquired infection; 30 patients were sedated with a continuous infusion of midazolam and fentanyl and artificially ventilated because of hypoxemia or brain injury or for postoperative care. Lastly, seven patients (No. 25 to 31) were on continuous venovenous hemofiltration (100 ml/min) with dialysis (1,000 ml//h) using bicarbonate-buffered fluids; the ultrafiltration rates were 450 to 900 ml/h (median: 780 ml/h).

The mean increase in blood lactate at the end of sodium L-lactate infusion was 3.5 ± 0.6 mmol/L. The lactate kinetic curve was always best described by a two-compartment model (p values ranging from 0.02 to < 0.001). The mathematical model used was valid as there was a close correlation between the lactate clearance estimated from the model and that calculated using the model-independent method (Figure 1). The lactate kinetics of two patients are shown in Figure 2. The blood lactate concentration, lactate clearance, and lactate production of the patients on continuous hemofiltration and those without it were not significantly different (p = 0.32, 0.12, and 0.65, respectively).

View larger version (17K): [in this window] [in a new window]   Figure 1.   Relationship between lactate clearance estimated from a two-compartment model (model dependent) and lactate clearance estimated by direct sampling (model independent). The excellent correlation shows that the lactate concentrations predicted by the mathematical model are very close to the lactate concentrations measured during and after the lactate infusion.

View larger version (17K): [in this window] [in a new window]   Figure 2.   Change in blood lactate concentration induced by a sodium L-lactate infusion in two patients. The lines indicate the curve calculated from the mathematical model, and the circles the measured lactate concentrations. The closed circles are for Patient No. 28 (basal blood lactate = 0.8 mmol/L, estimated lactate clearance = 1,777 ml/kg/h), and the open circles for Patient No. 33 (basal blood lactate = 3.1 mmol/L, estimated lactate clearance = 384 ml/kg/h).

The patients were first divided into two groups according to their blood lactate: Group 1 included the 20 patients having a blood lactate  1.5 mmol/L and Group 2 the 10 patients with a blood lactate  2 mmol/L. The four remaining patients had a blood lactate of 1.5 to 2.0 mmol/L and were excluded from this first analysis, so as to clearly separate the normolactatemic from the hyperlactatemic patients. The patients in the two groups had similar ages, hemodynamic and metabolic parameters, liver and renal functions, and APACHE II scores (Table 2). The main lactate kinetic parameters of the two groups are shown in Table 3. The estimated lactate clearance was greater in patients with low blood lactate than in those with normal lactate, but lactate production in the two groups was similar. The central distribution volume of infused lactate tended to be slightly greater in the normolactatemic group, whereas the total distribution volumes of the groups were comparable.

The relationships between the main parameters were studied in a second analysis that included all the patients. This confirmed the first analysis, showing a close relationship between the blood lactate concentration and the reciprocal of lactate clearance (r ² = 0.73, p < 0.001), whereas the correlation between the blood lactate concentration and production was not significant (r ² = 0.034, p = 0.29; Figure 3). The correlations linking lactate clearance with prothrombin time (p = 0.13), total bilirubin (p = 0.37), transaminases (p = 0.49 and 0.41), renal creating clearance (p = 0.36), and blood pH (p = 0.17) were all nonsignificant.

View larger version (18K): [in this window] [in a new window]   Figure 3.   (Upper panel ) Relationship between blood lactate concentration and lactate production in septic patients. (Lower panel ) Relationship between blood lactate production and the reciprocal of lactate clearance in the same patients. The open circles represent the patients on continuous hemofiltration.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The normal blood lactate concentration is < 1.5 mmol/L. Lactate is the end product of anaerobic glycolysis and is produced by a shift in pyruvate metabolism, depending on the redox state of the cytosol (25, 26). The persistence of severe lactic acidosis during shock is classically a marker of tissue hypoxia (1). By extrapolation, an increase in blood lactate concentration, even a mild one (i.e., 2 to 4 mmol/L), has been widely used to confirm data indicating the presence of tissue hypoxia in septic patients (7). Our study was designed to determine whether blood lactate is an indicator of anaerobic metabolism (i.e., lactate production) or only a marker of lactate utilization (i.e., lactate clearance) in hemodynamically stable septic patients. The results show that the persistence of a small increase in blood lactate concentration during sepsis is a sign of altered lactate clearance and not evidence of lactate overproduction. These results raise doubts about the reliability of mild hyperlactatemia as an indicator of the intensity of anaerobic metabolism in septic patients.

  The methods used to measure lactate clearance and production assume that these variables are not affected by a transient increase in lactate produced by an intravenous bolus injection and that the resulting decrease in lactate is linear. This assumption is reasonable, as the two-compartment mathematical model describing the change in blood lactate gave data very close to that obtained by sampling. This would not be so if lactate production and clearance had changed during the infusion, or if the rate of lactate loss had been not linear. Seven patients were on continuous venovenous hemofiltration with dialysis during the study period, using bicarbonate-buffered and lactate-free fluids. Their blood lactate concentration, lactate clearance, and lactate production were not significantly different from those without continuous hemofiltration. We have previously used the same fluids and rates of hemofiltration/dialysis to show that this renal replacement technique accounts for < 3% of the total plasma lactate clearance (27). It is therefore unlikely that hemofiltration could significantly affect the results. The lactate production was slightly greater than that reported for healthy subjects using several techniques (23, 28- 30). It is probably the consequence of the effect of sepsis on both the enhanced glycolysis rate (31) and the pyruvate dehydrogenase inactivation.

  Lactate is mainly cleared by the liver, but the kidneys and skeletal muscles are also involved (25, 26). It may be cleared by oxidation via the Krebs cycle, or by gluconeogenesis via the Cori cycle (32). Blood pH is also an important determinant of the lactate metabolism, as alkalosis enhances lactate production (25) and alters lactate elimination (33). We found no significant relationship between lactate clearance and the main liver and renal function tests or with blood pH. The initial distribution volume of the infused lactate tended to be slightly smaller in the hyperlactatemic patients. If we assume that lactate metabolism takes place only in the central volume of distribution, i.e., the initial one, this could be evidence for a lower metabolic capacity for lactate in the hyperlactatemic group than in the normolactatemic patients. These results tend to show that the defect in lactate clearance during sepsis is more complex than a simple isolated organ dysfunction. We did not evaluate the hepatic circulation, and the decrease in lactate clearance could be the consequence of a liver hypoperfusion. However, the liver cytolysis of the patients with normal and mild hyperlactatemia were similar and lactate clearance was not correlated with the transaminases. Experimental data also indicate that hepatic lactate uptake is only modestly reduced when hepatic blood flow is severely decreased (34). Other studies show that sepsis per se alters lactate metabolism. Hurtado and coworkers (22) found that a decrease in cardiac output caused by injecting endotoxin led to a more marked hyperlactatemia than when a similar decrease in cardiac output was induced mechanically, despite similar tissue oxygen parameters. Others have pointed out that there is a defect in pyruvate dehydrogenase activity during sepsis (21). This enzyme, which normally allows pyruvate to enter the Krebs cycle and is thus an important step in lactate clearance, can be shifted by sepsis from its active form to a phosphorylated inactive form. Similarly, animal studies have indicated that the increase in blood lactate caused by sepsis can be reversed by dichloroacetate, a drug that stimulates pyruvate dehydrogenase activity (35, 36).

  We conclude that a persistent mild hyperlactatemia occurring in hemodynamically stable septic patients should not be considered to be a reliable indicator of anaerobic metabolism but rather as a defect of lactate utilization. The disturbances of lactate metabolism that occur during sepsis are probably more complex than an isolated defect of cellular oxygenation.