January 2006

Section 8 Critical Care

There are three widely accepted methods of describing and classifying acid-base abnormalities. Essentially, they differ from one another only with respect to assessment of the metabolic component of the abnormality; all three treat PCO₂ as an independent variable. The first method quantifies the metabolic component by using the bicarbonate ion (HCO₃^-) concentration (in the context of PCO₂); the second, by using the standard base excess (SBE); and the third, by using the strong ion difference (SID). In practice, these three methods yield virtually identical results when employed to quantify the acid-base status of a given blood sample.^1–4 Thus, the only significant distinctions between the methods are conceptual ones, related to how each one approaches the understanding of the mechanism of the disorder.^5–7 In this chapter, we emphasize the physicochemical determinants of pH in the blood and the tissues; however, it is a simple matter to convert from one approach to the other if desired.

There are three mathematically independent determinants of blood pH: (1) the SID, defined as the difference in concentration between strong cations (e.g., sodium [Na⁺] and potassium [K⁺]) and strong anions (e.g., chloride [Cl^-] and lactate); (2) the total concentration of weak acids (A_tot), mainly consisting of albumin and phosphate; and (3) PCO₂. These three variables, and only these three, can independently affect plasma pH. The H⁺ and HCO₃^- concentrations are dependent variables whose values in plasma are determined by the SID, A_tot, and PCO₂. Changes in the plasma H⁺ concentration occur as a result of changes in the dissociation of water and A_tot, brought about by the electrochemical forces generated by changes in the SID and PCO₂. The SBE is mathematically equivalent to the difference between the current SID and the SID required to restore the pH to 7.4, given a PCO₂ of 40 mm Hg and the prevailing A_tot. Thus, an SBE of -10 mEq/L means that the SID is 10 mEq less than the value required to achieve a pH of 7.4.

The essential element of this physicochemical approach is the emphasis on independent and dependent variables. Only changes in the independent variables can bring about changes in the dependent variables. That is, movement of H⁺ or HCO₃^- cannot affect plasma H⁺ or HCO₃^- concentrations unless changes in the SID, A_tot or PCO₂ also occur. Several reviews of this approach are available in the literature.^3–9 In what follows, we discuss the clinical application of this approach to the diagnosis and treatment of individual acid-base disorders.

Once identified, an acid-base disorder can be classified according to a simple set of rules [see Table 1]. A disorder that does not fit well into the broad categories established by these rules can be considered a mixed (or complex) disorder. Some of the basic categories can be further divided into various subcategories (see below), but before the issue of classification is addressed in detail, three general caveats must be considered.

First, interpretation of arterial blood gas values and blood chemistries depends on the reliability of the data. Advances in clinical chemistry have improved the sensitivity of instruments used to measure electrolyte concentrations (e.g. ion-specific electrodes) and have greatly enhanced the speed and ease of analysis. Inevitably, however, prolonged exposure to the atmosphere results in a lowering of the PCO₂, and over time, there may be ongoing cellular metabolism. Accordingly, prompt measurement is always advisable. Even with prompt measurement, laboratory errors may occur, and information may be incorrectly reported. Samples drawn from indwelling lines may be diluted by fluid or drug infusions (a notorious source of error). When the situation is confusing, it is usually best to repeat the measurement.

Second, interpretation of arterial blood gas values may be problematic in patients with severe hypothermia (e.g., trauma patients undergoing damage-control interventions, who often are severely hypothermic and sometimes experience severe acidosis), in that the findings may not reflect the actual blood gas values present. Because blood samples are 'normalized' to a temperature of 37° C before undergoing analysis, the results obtained in samples from a patient whose body temperature is significantly lower than 37° C may not be sufficiently accurate. To obviate this potential problem, the results may have to be adjusted to take the patient's actual temperature into account. At present, however, such temperature correction is not routinely done, and there has been some controversy regarding whether it has real clinical value.^10,11

Third, whereas the aforementioned four indicators are useful for identifying an acid-base disorder, the absence of all four does not suffice to exclude a mixed acid-base disorder (i.e., alkalosis plus acidosis) in which the two components are completely matched. Fortunately, such conditions are rare. In addition, apart from distinguishing a respiratory acid-base disorder from a metabolic acid-base disorder, the four indicators and the rules previously mentioned [see Table 1] provide no information on the mechanism of an acid-base disorder.

Traditional approaches to the question of renal acid handling have focused on H⁺ excretion, emphasizing the importance of ammonia (NH₃) and its add-on cation, ammonium (NH₄⁺). However, H⁺ excretion per se is irrelevant, in that water provides an essentially infinite source of free H⁺. Indeed, the kidney does not excrete H⁺ to any greater degree in the form of NH₄⁺ than in the form of H₂O. The purpose of renal ammoniagenesis is to allow the excretion of Cl^- without Na⁺ or K⁺. This purpose is achieved by supplying a weak cation (NH₄⁺) that is 'coexcreted' with Cl^-. The mechanisms of renal tubular acidosis are currently being reinterpreted by some authors in the light of a growing body of evidence showing that abnormal chloride conductance, rather than H⁺ or HCO₃ handling per se, is responsible for these disorders.³

Kidney-Liver Interaction

The importance of NH₄⁺ to systemic acid-base balance, then, rests not on its carriage of H⁺ or its direct action in the plasma (normal plasma NH₄⁺ concentration < 0.01 mEq/L) but on its coexcretion with Cl^-. Of course, production of NH₄⁺ is not restricted to the kidney. Hepatic ammoniagenesis (as well as glutaminogenesis) is also important for systemic acid-base balance, and as expected, it is tightly controlled by mechanisms sensitive to plasma pH.¹² Indeed, this reinterpretation of the role of NH₄⁺ in acid-base balance is supported by the evidence that hepatic glutaminogenesis is stimulated by acidosis.¹³ Metabolism of nitrogen by the liver can yield urea, glutamine, or NH₄⁺. Normally, the liver releases only a very small amount of NH₄⁺, incorporating most of its nitrogen into either urea or glutamine. Hepatocytes have enzymes to enable them to produce either of these end products, and both allow regulation of plasma NH₄⁺ at suitably low levels. At the level of the kidneys, however, the production of urea or glutamine has significantly different effects, in that the kidneys use glutamine to generate NH₄⁺ and facilitate the excretion of Cl^-. Thus, production of glutamine by the liver can be seen as having an alkalinizing effect on plasma pH because of the way in which the kidneys use this substance.

Further support for this scenario comes from the discovery that hepatocytes are anatomically organized according to their enzymatic content.¹⁴ Hepatocytes with a propensity to produce urea are positioned closer to the portal venule; those with a propensity to produce glutamine are positioned farther downstream. The upstream (urea-producing) hepatocytes have the first chance at the NH₄⁺ delivered. However, acidosis inhibits ureagenesis, thereby leaving more NH₄⁺ available for the downstream (glutamine-producing) hepatocytes. The leftover NH₄⁺ is thus, in a sense, packaged as glutamine for export to the kidney, where it is used to facilitate Cl^- excretion.

Gastrointestinal Tract

The GI tract is an underappreciated component of acid-base balance. In different regions along its length, the GI tract handles strong ions quite differently. In the stomach, Cl^- is pumped out of the plasma and into the lumen, thereby reducing the SID of the gastric juice and thus the pH as well. On the plasma side, the SID is increased by the loss of Cl^-, and the pH rises, producing the so-called alkaline tide that occurs at the beginning of a meal, when gastric acid secretion is maximal.¹⁵

In the duodenum, Cl^- is reabsorbed and the plasma pH restored. Normally, only slight changes in plasma pH are evident because Cl^- is returned to the circulation almost as soon as it is removed. If, however, gastric secretions are removed from the patient, whether by catheter suctioning or vomiting, Cl^- will be progressively lost and the SID will steadily increase. It is important to remember that it is the loss of Cl^-, not of H⁺, that determines the plasma pH. Although H⁺ is lost as HCl, it is also lost with every molecule of water removed from the body. When Cl^-, a strong anion, is lost without the corresponding loss of a strong cation, the SID is increased, and therefore, the plasma H⁺ concentration is decreased. When H⁺ is lost as water rather than as HCl, the SID does not change, and thus, the plasma H⁺ concentration does not change either.

The pancreas secretes fluid into the small intestine that possesses an SID much higher than the plasma SID and is very low in Cl^-. Thus, the plasma perfusing the pancreas has its SID decreased, a phenomenon that peaks about 1 hour after a meal and helps counteract the alkaline tide. If large amounts of pancreatic fluid are lost (e.g., as a consequence of surgical drainage), the resulting decrease in the plasma SID will lead to acidosis.

In the large intestine, the fluid also has a high SID, because most of the Cl^- was removed in the small intestine and the remaining electrolytes consist mostly of Na⁺ and K⁺. Normally, the body reabsorbs much of the water and electrolytes from this fluid, but when severe diarrhea occurs, large amounts of cations may be lost. If this cation loss persists, the plasma SID will decrease and acidosis will result.

In addition to the acid-base effects of abnormal loss of strong ions from the GI lumen, the small intestine, in particular, may contribute strong ions to the plasma. This contribution is most apparent when mesenteric blood flow is compromised and lactate is produced, sometimes in large quantities. Although global hypoperfusion may compromise the mesentery, the intestine does not appear to be a source of lactic acid in patients resuscitated from a septic state [see Metabolic Acidosis, Positive-Anion Gap Acidosis, Lactic Acidosis, below].¹⁶ Moreover, whether the GI tract is capable of regulating strong ion uptake in a compensatory fashion has not been well studied. There is some evidence that the gut may modulate systemic acidosis in experimental endotoxemia by removing anions from the plasma¹⁷; however, the full capacity of the gut to affect acid-base balance remains to be determined.

Traditionally, metabolic acidoses are categorized according to the presence or absence of unmeasured anions. These unmeasured anions are routinely detected by examining the plasma electrolytes and calculating the anion gap (AG) (see below). The differential diagnosis for a positive-AG acidosis includes various common and rare causes [see Table 2]. Generally speaking, non-AG acidoses can be divided into three types: renal, gastrointestinal, and iatrogenic [see Table 3]. In the ICU, the most common types of metabolic acidosis are lactic acidosis, ketoacidosis, iatrogenic acidosis, and acidosis secondary to toxins.

Even extreme acidosis appears to be well tolerated by healthy persons, particularly when the duration of the acidosis is short. For example, healthy individuals may achieve an arterial pH lower than 7.15 and a lactate concentration higher than 20 mEq/L during maximal exercise, with no lasting effects.¹⁸ Over the long term, however, even mild acidemia (pH < 7.35) may lead to metabolic bone disease and protein catabolism. Furthermore, critically ill patients may not be able to tolerate even brief episodes of acidemia.¹⁹ There do appear to be significant differences between metabolic and respiratory acidosis with respect to patient outcome, and these differences suggest that the underlying disorder may be more important than the absolute degree of acidemia.²⁰

If prudence dictates that symptomatic therapy is to be provided, the likely duration of the disorder should be taken into account. When the disorder is expected to be a short-lived one (e.g., diabetic ketoacidosis), maximizing respiratory compensation is usually the safest approach. Once the disorder resolves, ventilation can be quickly reduced to normal levels, and there will be no lingering effects from therapy. If the SID is increased (e.g., by administering NaHCO₃), there is a risk of alkalosis when the underlying disorder resolves. When the disorder is likely to be a more chronic one (e.g., renal failure), therapy aimed at restoring the SID to normal is indicated. In all cases, the therapeutic target can be accurately determined from the SBE. As noted (see above), the SBE corresponds to the amount by which the current SID differs from the SID necessary to restore the pH to 7.4, given a PCO₂ of 40 mm Hg. Thus, if the SID is 30 mEq/L and the SBE is -10 mEq/L, the target SID is 40 mEq/L. Accordingly, the plasma Na⁺ concentration would have to increase by 10 mEq/L for NaHCO₃ administration to correct the acidosis completely.

It should be noted that the target SID is the SID at the equilibrium point between the SID, PCO₂, and A_tot and that it may not be equal to 40 mEq/L, as in the example given. By convention, PCO₂ is set at 40 mm Hg, but the SBE is not corrected for abnormalities in A_tot. In many hypoalbuminemic patients, A_tot is lower than normal, and thus, the SID at the equilibrium point will be less than 40 mEq/L. Also, it is rare that the choice would be made to correct the acid-base abnormality completely. Therefore, the target SID should be used as a reference value, but in most cases, partial correction is all that is required.

If increasing the plasma Na⁺ concentration is inadvisable for other reasons (e.g., hypernatremia), NaHCO₃ administration is inadvisable. It is noteworthy that NaHCO₃ administration has not been shown to improve outcome in patients with lactic acidosis.²¹ In addition, NaHCO₃ administration is associated with certain disadvantages. Large (hypertonic) doses, if given rapidly, may actually reduce blood pressure²² and may cause sudden, severe increases in P_aCO₂.²³ Accordingly, it is important to assess the patient's ventilatory status before NaHCO₃ is administered, particularly if the patient is not on a ventilator. NaHCO₃ infusion also affects serum K⁺ and Ca²⁺ concentrations, which must be monitored closely.

To avoid some of the disadvantages of NaHCO₃ therapy, alternative therapies for metabolic acidosis have been developed. Carbicarb is an equimolar mixture of sodium carbonate (Na₂CO₃) and NaHCO₃.²⁴ Like NaHCO₃, carbicarb works by increasing the plasma Na⁺ concentration, except that it does not raise the PCO₂. Results with carbicarb in animal studies have been mixed,²⁵ and experience in humans is extremely limited.

THAM (tris-hydroxymethyl aminomethane) is a synthetic buffer that consumes CO₂ and readily penetrates cells.²⁶ It is a weak base (pK = 7.9) and, as such, is unlike other plasma constituents. The major advantage of THAM is that it does not alter the SID, which means that there is no need to be concerned about having to increase the plasma Na⁺ concentration to achieve a therapeutic effect. Accordingly, THAM is often used in situations where NaHCO₃ cannot be used because of hypernatremia. Although THAM has been available since the 1960s, there is surprisingly little information available regarding its efficacy in humans with acid-base disorders. In small uncontrolled studies, THAM appears to be capable of reversing metabolic acidosis secondary to ketoacidosis or renal failure without causing obvious toxicity²⁷; however, adverse reactions have been reported, including hypoglycemia, respiratory depression, and even fatal hepatic necrosis, when concentrations exceeding 0.3 mol/L are used. In Europe, a mixture of THAM, acetate, NaHCO₃, and disodium phosphate is available. This mixture, known as tribonate (Tribonat; Pharmacia & Upjohn, Solna, Sweden), seems to have fewer side effects than THAM alone does, but as with THAM, experience with its use in humans is still quite limited.

Anion Gap

Determination of anion gap The AG has been used by clinicians for more than 30 years and has evolved into a major tool for evaluating acid-base disorders.²⁸ It is calculated—or, rather, estimated—from the difference between the routinely measured concentrations of serum cations (Na⁺ and K⁺) and the routinely measured concentrations of anions (Cl^- and HCO₃^-). Normally, albumin accounts for the bulk of this difference, with phosphate playing a lesser role. Sulfate and lactate also contribute a small amount to the gap (normally, < 2 mEq/L); however, there are also unmeasured cations (e.g., Ca²⁺ and Mg²⁺), which tend to offset the effects of sulfate and lactate except when the concentration of either one is abnormally increased. Plasma proteins other than albumin can be either positively or negatively charged, but in the aggregate, they tend to be electrically neutral,²⁹ except in rare cases of abnormal paraproteins (as in multiple myeloma). In practice, the AG is calculated as follows:

AG = (Na⁺+ K⁺) - (Cl^-+ HCO₃^-)

Because of its low extracellular concentration, K⁺ is often omitted from the calculation. In most laboratories, normal values fall into the range of 12 ± 4 mEq/L (if K⁺ is considered) or 8 ± 4 mEq/L (if K⁺ is not considered). In the past few years, the introduction of more accurate methods of measuring Cl^- concentration has led to a general lowering of the normal AG range.^30,31 Because of the various measurement techniques employed at various institutions, however, each institution is expected to report its own normal AG values.

Clinical utility of anion gap The primary value of the AG is that it quickly and easily limits the differential diagnosis in a patient with metabolic acidosis. When the AG is increased, the explanation is almost invariably one of the following five disorders: ketosis, lactic acidosis, poisoning, renal failure, and sepsis.³²

In addition to these disorders, however, there are several conditions that can alter the accuracy of AG estimation and are particularly frequent in critical illness.^33,34 Dehydration increases the concentrations of all of the ions. Severe hypoalbuminemia lowers the AG, with each 1 g/dl decline in the serum albumin reducing the apparent AG by 2.5 to 3 mEq/L; accordingly, some recommend adjusting the AG for the prevailing albumin concentration.³⁵ Alkalosis (respiratory or metabolic) is associated with an increase of as much as 3 to 10 mEq/L in the apparent AG as a consequence of enhanced lactate production (from stimulated phosphofructokinase enzymatic activity), reduction in the concentration of ionized weak acids (A^-)(as opposed to A_tot, the total concentration of weak acids), and, possibly, the additional effect of dehydration (which, as noted, has its own impact on AG calculation). A low Mg²⁺ concentration with associated low K⁺ and Ca²⁺ concentrations is a known cause of an increased AG, as is the administration of sodium salts of poorly reabsorbable anions (e.g., b-lactam antibiotics).³⁶ Certain parenteral nutrition formulations (e.g., those containing acetate) may increase the AG. In rare cases, citrate may have the same effect in the setting of multiple blood transfusions, particularly if massive doses of banked blood are used (as during liver transplantation).³⁷ None of these rare causes, however, will increase the AG significantly,³⁸ and they usually are easily identified.

In the past few years, some additional causes of an increased AG have been reported. The nonketotic hyperosmolar state of diabetes has been associated with an increased AG that remains unexplained.³⁹ Unmeasured anions have been reported in the blood of patients with sepsis,^40,41 patients with liver disease,^42,43 and experimental animals that received endotoxin.⁴⁴ These anions may be the source of much of the unexplained acidosis seen in patients with critical illness.⁴⁵

The accepted clinical utility of the AG notwithstanding, doubt has been cast on its diagnostic value in certain situations.^33,41 Some investigators have found routine reliance on the AG to be 'fraught with numerous pitfalls.'³³ The primary problem with the AG is its reliance on the use of a supposedly normal range produced by albumin and, to a lesser extent, phosphate. Concentrations of albumin and phosphate may be grossly abnormal in patients with critical illness, and these abnormalities may change the normal AG range in this setting. Moreover, because these anions are not strong anions, their charge is altered by changes in pH. These concerns have led some authors to advocate adjusting the normal AG range on the basis of the patient's albumin³⁵ or even phosphate⁶ concentration. Each 1 g/dl of albumin carries a charge of 2.8 mEq/L at a pH of 7.4 (2.3 mEq/L, pH = 7.0; 3.0 mEq/L, pH = 7.6), and each 1 mg/dl of phosphate carries a charge of 0.59 mEq/L at a pH of 7.4 (0.55 mEq/L, pH = 7.0; 0.61 mEq/L, pH = 7.6). Thus, the normal AG for a given patient can be conveniently estimated as follows⁶:

Normal AG = 2(albumin [g/dl]) + 0.5(phosphate [mg/dl])

or, in international units,

Normal AG = 0.2(albumin [g/L]) + 1.5(phosphate [mmol/L])

In one study, when this formula for calculating a patient-specific normal AG range was used to determine the presence of unmeasured anions in the blood of critically ill patients, its accuracy was 96%, compared with an accuracy of 33% with the routine AG (normal range = 12 mEq/L).⁶ This technique should be employed only when the pH is less than 7.35; even in this situation, it is only accurate within 5 mEq/L. When more accuracy is needed, a slightly more complicated method of estimating unmeasured anions is required.^42,46

Strong anion gap Another alternative to relying on the traditional AG is to use a parameter derived from the SID. By definition, the SID must be equal and opposite to the sum of the negative charges contributed by A^- and total CO₂. This latter value (A^-+ total CO₂) has been termed the effective SID (SIDe).²⁹ The apparent SID (SIDa) is obtained by measuring concentrations of each individual ion. The SIDa and the SIDe should both equal the true SID. If the SIDa differs from the SIDe, unmeasured ions must be present. If the SIDa is greater than the SIDe, these unmeasured ions are anions; if the SIDa is less than the SIDe, they are cations. The difference between the SIDa and the SIDe has been termed the strong ion gap (SIG) to distinguish it from the AG.⁴² Unlike the AG, the SIG is normally 0 and is not affected by changes in the pH or the albumin concentration.

Positive-Anion Gap Acidosis

Lactic acidosis In many forms of critical illness, lactate is the most important cause of metabolic acidosis.⁴⁷ Lactate concentrations have been shown to correlate with outcomes in patients with hemorrhagic⁴⁸ and septic shock.⁴⁹ Traditionally, lactic acid has been viewed as the predominant source of the metabolic acidosis that occurs in sepsis.⁵⁰ In this view, lactic acid is released primarily from the musculature and the gut as a consequence of tissue hypoxia, and the amount of lactate produced is believed to correlate with the total oxygen debt, the magnitude of hypoperfusion, and the severity of shock.⁴⁷ This view has been challenged by the observation that during sepsis, even in profound shock, resting muscle does not produce lactate. Indeed, various studies have shown that the musculature may actually consume lactate during endotoxemia.^16,51,52

The data on lactate release by the gut are less clear. There is little question that the gut can release lactate if it is underperfused. It appears, however, that if the gut is adequately perfused, it does not release lactate during sepsis. Under such conditions, the mesentery either is neutral with respect to lactate release or takes up lactate.^16,51 Perfusion is likely to be a major determinant of mesenteric lactate metabolism. In a canine model of sepsis induced by infusion of endotoxin, production of lactate by the gut could not be demonstrated when flow was maintained with dopexamine.⁵²

Both animal studies and human studies have shown that the lung may be a prominent source of lactate in the setting of acute lung injury.^16,53,54 These studies do not address the underlying pathophysiologic mechanisms of hyperlactatemia in sepsis, but they do suggest that the conventional wisdom regarding lactate as evidence of tissue dysoxia is, at best, an oversimplification. Indeed, many investigators have begun to offer alternative explanations for the development of hyperlactatemia in this setting [see Table 4].^54–58 One proposed mechanism is metabolic dysfunction from mitochondrial enzymatic derangements, which can and do lead to lactic acidosis. In particular, pyruvate dehydrogenase (PDH), the enzyme responsible for moving pyruvate into the Krebs cycle, is inhibited by endotoxin.⁵⁹ Current data, however, suggest that increased aerobic metabolism may be more important than metabolic defects or anaerobic metabolism. In a 1996 study, production of glucose and pyruvate and oxidation were increased in patients with sepsis.⁶⁰ Furthermore, when PDH was stimulated by dichloroacetate, there was an additional increase in oxygen consumption but a decrease in glucose and pyruvate production. These results suggest that hyperlactatemia in sepsis occurs as a consequence of increased aerobic metabolism rather than of tissue hypoxia or PDH inhibition.

Such findings are consistent with the known metabolic effects of lactate production on cellular bioenergetics.⁶¹ Lactate production alters cytosolic, and hence mitochondrial, redox states, so that the increased ratio of reduced nicotinamide adenine dinucleotide to nicotinamide adenine dinucleotide (NADH/NAD) supports oxidative phosphorylation as the dominant source of ATP production. Finally, the use of catecholamines, especially epinephrine, also results in lactic acidosis, presumably by stimulating cellular metabolism (e.g., increasing hepatic glycolysis), and may be a common source of lactic acidosis in the ICU.^62,63 It is noteworthy that this phenomenon does not appear to occur with either dobutamine or norepinephrine⁶⁴ and does not appear to be related to decreased tissue perfusion.

Although the source and interpretation of lactic acidosis in critically ill patients remain controversial, there is no question about the ability of lactate accumulation to produce acidemia. Lactate is a strong ion because at a pH within the physiologic range, it is almost completely dissociated. (The pK of lactate is 3.9; at a pH of 7.4, 3,162 ions are dissociated for every one ion that is not.) Because lactate is rapidly produced and disposed of by the body, it functions as one of the most dynamic components of the SID. Therefore, a rise in the concentration of lactic acid can produce significant acidemia. Just as often, however, critically ill patients have a degree of hyperlactatemia that far exceeds the degree of acidosis observed. In fact, hyperlactatemia may exist without any metabolic acidosis at all. This is not because acid generation is separate from lactate production (e.g., through 'unreversed ATP hydrolysis'), as some have suggested.⁶⁴ Phosphate is a weak acid and does not contribute substantially to metabolic acidosis, even under extreme circumstances. Furthermore, the H⁺ concentration is determined not by how much H⁺ is produced or removed from the plasma but by changes in the dissociation of water and weak acids. Virtually anywhere in the body, the pH is higher than 6.0, and lactate behaves as a strong ion. Generation of lactate reduces the SID and results in an increased H⁺ concentration; however, the plasma lactate concentration may also be increased without an accompanying increase in the H⁺ concentration.

There are two possible explanations for these observations. First, if lactate is added to the plasma, not as lactic acid but rather as the salt of a strong acid (e.g., sodium lactate), the SID will not change significantly, because a strong cation (Na⁺) is being added along with a strong anion. Indeed, as lactate is metabolized and removed, the remaining Na⁺ will increase the SID, resulting in metabolic alkalosis. Hence, it would be possible to give enough lactate to increase the plasma lactate concentration without increasing the H⁺ concentration. However, given that normal metabolism results in the turnover of approximately 1,500 to 4,500 mmol of lactic acid each day, rapid infusion of a very large amount of lactate would be required to bring about an appreciable increase in the plasma lactate concentration. For example, the use of lactate-based hemofiltration fluid may result in hyperlactatemia with an increased plasma HCO₃^- concentration and an elevated pH.

A more important mechanism whereby hyperlactatemia can exist without acidemia (or with less acidemia than expected) involves correction of the SID by the elimination of another strong anion from the plasma. In a study of sustained lactic acidosis induced by lactic acid infusion, Cl^- was found to move out of the plasma space, thereby normalizing the pH.⁶⁵ Under these conditions, hyperlactatemia may persist, but compensatory mechanisms may normalize the base excess and thus restore the SID.

Traditionally, lactic acidosis has been subdivided into type A, in which the mechanism is tissue hypoxia, and type B, in which there is no hypoxia.⁶⁶ This distinction may, however, be an artificial one. Disorders such as sepsis may be associated with lactic acidosis through a variety of mechanisms [see Table 4], some conventionally labeled type A and others type B. A potentially useful method of distinguishing between anaerobically produced lactate and lactate from other sources is to measure the serum pyruvate concentration. The normal lactate-to-pyruvate ratio is 10:1,⁶⁷ with ratios greater than 25:1 considered to be evidence of anaerobic metabolism.⁵⁸ This approach makes biochemical sense because pyruvate is shunted into lactate during anaerobic metabolism, dramatically increasing the lactate-to-pyruvate ratio. However, the precise test characteristics, including normal ranges and sensitivity and specificity data, have not yet been defined for patients. Accordingly, this method remains investigational.

Treatment of lactic acidosis continues to be subject to debate. At present, the only noncontroversial approach is to treat the underlying cause; however, this approach assumes that the underlying cause can be identified with a significant degree of certainty, which is not always the case. The assumption that hypoperfusion is always the most likely cause has been seriously challenged, especially in well-resuscitated patients (see above). Thus, therapy aimed at increasing oxygen delivery may not be effective. Indeed, if epinephrine is used, lactic acidosis may worsen.

Administration of NaHCO₃ to treat lactic acidosis remains unproven.²¹ In perhaps the most widely quoted study on this topic, hypoxic lactic acidosis was induced in anesthetized dogs by ventilating them with gas containing very little oxygen.⁶⁸ These animals were then assigned to treatment with NaHCO₃ or placebo, and surprisingly, the group receiving NaHCO₃ actually had higher plasma concentrations of both lactate and H⁺ than the control group did. Furthermore, the NaHCO₃-treated animals exhibited decreases in cardiac output and blood pressure that were not seen in the control group. One possible explanation for these findings is that the HCO₃^- was converted to CO₂, and this conversion raised the PCO₂ not only in the blood but also inside the cells of these animals with a fixed minute ventilation; the resulting intracellular acidosis might have been detrimental to myocardial function. This hypothesis has not, however, been supported by subsequent experimental studies, which have not documented paradoxical intracellular acidosis or even detrimental hemodynamic effects after NaHCO₃ treatment of hypoxic lactic acidosis.⁶⁹ Furthermore, it is not clear how this type of hypoxic lactic acidosis, induced in well-perfused animals, relates to the clinical conditions in which lactic acidosis occurs. The results of clinical studies have been mixed, but overall, they do not support the use of NaHCO₃ therapy for lactic acidosis.²¹

Ketoacidosis Another common cause of a metabolic acidosis with a positive AG is excessive production of ketone bodies, including acetone, acetoacetate, and b-hydroxybutyrate. Both acetoacetate and b-hydroxybutyrate are strong anions (pK 3.8 and 4.8, respectively).⁷⁰ Thus, their presence, like the presence of lactate, decreases the SID and increases the H⁺ concentration.

Ketones are formed through beta oxidation of fatty acids, a process that is inhibited by insulin. In insulin-deficient states (e.g., diabetes), ketone formation may quickly get out of control. The reason is that severely elevated blood glucose concentrations produce an osmotic diuresis that may lead to volume contraction. This state is associated with elevated cortisol and catecholamine secretion, which further stimulates free fatty acid production.⁷¹ In addition, an increased glucagon level relative to the insulin level leads to a decreased malonyl coenzyme A level and an increased carnitine palmityl acyl transferase level—a combination that increases ketogenesis.

Ketoacidosis may be classified as either diabetic ketoacidosis (DKA) or alcoholic ketoacidosis (AKA). The diagnosis is established by measuring serum ketone levels. It must be kept in mind, however, that the nitroprusside reaction measures only acetone and acetoacetate, not b-hydroxybutyrate. Thus, the measured ketosis is dependent on the ratio of acetoacetate to b-hydroxybutyrate. This ratio is low when lactic acidosis coexists with ketoacidosis because the reduced redox state characteristic of lactic acidosis favors production of b-hydroxybutyrate.⁷² In such circumstances, therefore, the apparent degree of ketosis is small relative to the degree of acidosis and the elevation of the AG. There is also a risk of confusion during treatment of ketoacidosis, in that ketone levels, as measured by the nitroprusside reaction, sometimes rise even though the acidosis is resolving. This occurs because the nitroprusside reaction does not detect b-hydroxybutyrate, and as b-hydroxybutyrate is cleared, ketosis persists despite improvement in acid-base balance. Furthermore, conversion of b-hydroxybutyrate to acetoacetate may cause an apparent increase in ketone levels—again, because the nitroprusside reaction detects the rising levels of acetoacetate but misses the falling levels of b-hydroxybutyrate. Hence, it is better to monitor the success of therapy by measuring the pH and the AG than by assaying serum ketones.

Treatment of DKA includes administration of insulin and large amounts of fluid (0.9% saline is usually recommended); potassium replacement is often required as well. Fluid resuscitation reverses the hormonal stimuli for ketone body formation, and insulin allows the metabolism of ketones and glucose. Administration of NaHCO₃ may produce a more rapid rise in the pH by increasing the SID, but there is little evidence that this result is desirable. Furthermore, to the extent that the SID is increased by increasing the plasma Na⁺ concentration, the SID will be too high once the ketosis is cleared, thus resulting in a so-called overshoot alkalosis. In any case, such measures are rarely necessary and should probably be avoided except in extreme cases.⁷³

A more common problem in the treatment of DKA is the persistence of acidemia after the ketosis has resolved. This hyperchloremic metabolic acidosis occurs as Cl^- replaces ketoacids, thus maintaining a decreased SID and pH. There appear to be two reasons for this phenomenon. First, exogenous Cl^- is often provided in the form of 0.9% saline, which, if given in large enough quantities, will result in a so-called dilutional acidosis (see below). Second, some degree of increased Cl^- reabsorption apparently occurs as ketones are excreted in the urine. It has also been suggested that the increased tubular Na⁺ load produces electrical-chemical forces that favor Cl^- reabsorption.⁷⁴

AKA is usually less severe than DKA. Treatment consists of administration of fluids and (in contrast to treatment of DKA) glucose rather than insulin.⁷⁵ Insulin is contraindicated in AKA patients because it may cause precipitous hypoglycemia.⁷⁶ Thiamine must also be given to keep from precipitating Wernicke encephalopathy.

Acidosis secondary to renal failure Although renal failure may produce a hyperchloremic metabolic acidosis, especially when it is chronic, the buildup of sulfates and other acids frequently increases the AG; however, the increase usually is not large.⁷⁷ Similarly, uncomplicated renal failure rarely produces severe acidosis, except when it is accompanied by high rates of acid generation (e.g., from hypermetabolism).⁷⁸ In all cases, the SID is decreased and is expected to remain so unless some therapy is provided. Hemodialysis permits the removal of sulfate and other ions and allows the restoration of normal Na⁺ and Cl^- balance, thus returning the SID to a normal (or near-normal) value. However, those patients who do not yet require dialysis and those who are between treatments often require some other therapy aimed at increasing the SID. NaHCO₃ may be used for this purpose, provided that the plasma Na⁺ concentration is not already elevated.

Acidosis secondary to toxin ingestion Metabolic acidosis with an increased AG is a major feature of various types of intoxication [see Table 2]. Generally, it is more important to recognize these conditions and provide specific therapy for them than it is to treat the acid-base imbalances that they produce.

Acidosis secondary to rhabdomyolysis The extensive muscle tissue breakdown associated with myonecrosis may also be a source of positive-AG metabolic acidosis. In this situation, the acidosis results from accumulation of organic acids. The myoglobinuria associated with the disorder may also induce renal failure. In most cases, the diagnosis is a clinical one and can be facilitated by measuring creatinine kinase or aldolase levels. Early identification and aggressive resuscitation may prevent the onset of renal failure and improve the prognosis.⁷⁹

Acidosis of unknown origin Several causes of an increased AG have been reported that have yet to be elucidated. An unexplained AG in the nonketotic hyperosmolar state of diabetes has been reported.³⁹ In addition, even when very careful measurement techniques have been employed, unmeasured anions have been reported in the blood of patients with sepsis,^40,41 patients with liver disease,⁴² and animals to which endotoxin had been administered.⁴³ Furthermore, unknown cations also appear in the blood of some critically ill patients.⁴¹ The significance of these findings remains to be determined.

Prognostic significance of positive-AG metabolic acidosis Several studies have examined whether the presence of unmeasured anions in the blood is associated with particular outcomes in critically ill patients. Two such studies focused on trauma patients. In one, the investigators examined 2,152 sets of laboratory data from 427 trauma patients and found that the SIG altered the acid-base disorder diagnosis in 28% of the datasets.⁸⁰ Simultaneous measurements of blood gas, serum electrolyte, albumin, and lactate values were used to calculate the base deficit, the AG, and the SIG. Unmeasured anions (defined by the presence of an elevated SIG) were present in 92% of patients (mean SIG, 5.9 ± 3.3); hyperlactatemia and hyperchloremia occurred in only 18% and 21% of patients, respectively. The arterial SBE at ICU admission was poorly predictive of hospital survival, and its predictive ability was only slightly improved by controlling for unmeasured ions. In this dataset, survivors could not be differentiated from nonsurvivors in the group as a whole on the basis of the SIG. However, in the subgroup of patients whose lactate level was normal at admission, there was a significant difference in the SIG between survivors and nonsurvivors, though no such differences were noted in the conventional measures (i.e., SBE and AG).

The poor predictive ability of the SBE, the AG, and even the SIG has been confirmed by studies of general ICU patients. In one study, analysis of data from 300 adult ICU patients demonstrated statistically significant but weak correlations between these measures and hospital mortality.⁸¹ In another study, however, pretreatment SIG was found to be a very strong predictor of outcome in 282 patients who had sustained major vascular injury.⁸² All but one of the nonsurvivors had an initial emergency department (ED) pH of 7.26 or lower, an SBE of -7.3 mEq/L or lower, a lactate concentration of 5 mmol/L or higher, and an SIG of 5 mEq/L or higher. All of the acid-base descriptors were strongly associated with outcome, but the SIG was the one that discriminated most strongly. The investigators concluded that initial ED acid-base variables, especially SIG, could distinguish survivors of major vascular injury from nonsurvivors.

Even though the uncorrected AG and the SBE correlate poorly with the arterial lactate concentration in trauma patients,⁸³ several investigators have proposed that these parameters be used as surrogate measures of the severity of shock or lack of resuscitation. Various studies have shown that the SBE is a poor predictor of lactic acidosis and mortality both in medical patients and in surgical or trauma patients and that it cannot be substituted for direct measurement of the serum lactate concentration.^33,84,85 Some investigators, however, have found that the SBE can be used as a marker of injury severity and mortality and as a predictor of transfusion requirements.^86–88 Unfortunately, the SBE can determine only the degree of acid-base derangement, never the cause. In many critically injured patients, abnormalities in body water content, electrolyte levels, and albumin concentration limit any potential correlation between SBE and lactate concentration, even when other sources of acid are absent.

Several reports in the trauma literature have focused on the prognostic value of persistently elevated lactic acid levels during the first 24 to 48 hours after injury. In one study, involving 76 patients with multiple injuries who were admitted directly to the ICU from the operating room or the ED, serum lactate levels and oxygen transport were measured at ICU admission and at 8, 16, 24, 36, and 48 hours.⁸⁹ In those patients whose lactate levels returned to normal within 24 hours, the survival rate was 100%, and in those whose lactate levels returned to normal between 24 and 48 hours, the survival rate was 75%. However, in those whose lactate levels did not return to normal by 48 hours, the survival rate was only 14%. Thus, the rate of normalization of the serum lactate level is an important prognostic factor for survival in a severely injured patient.

Non-Anion Gap (Hyperchloremic) Acidoses

Hyperchloremic metabolic acidosis occurs as a result of either an increase in the level of Cl^- relative to the levels of strong cations (especially Na⁺) or a loss of cations with retention of Cl^-. The various causes of such an acidosis [see Table 3] can be distinguished on the basis of the history and the measured Cl^- concentration in the urine. When acidosis occurs, the kidney normally responds by increasing Cl^- excretion; the absence of this response identifies the kidney as the source of the problem. Extrarenal hyperchloremic acidoses occur because of exogenous Cl^- loads (iatrogenic acidosis) or because of loss of cations from the lower GI tract without proportional loss of Cl^- (gastrointestinal acidosis).

Renal tubular acidosis Most cases of renal tubular acidosis (RTA) can be correctly diagnosed by determining urine and plasma electrolyte levels and pH and calculating the SIDa in the urine [see Table 3].⁹⁰ However, caution must be exercised when the plasma pH is greater than 7.35, because urine Cl^- excretion may be turned off. In such circumstances, it may be necessary to infuse sodium sulfate or furosemide. These agents stimulate excretion of Cl^- and K⁺ and may be used to unmask the defect and to probe K⁺ secretory capacity.

Establishing the mechanisms of RTA has proved difficult. It is likely that much of the difficulty results from the attempt to understand the physiology from the perspective of regulation of H⁺ and HCO₃^- concentrations. As noted, however (see above), this approach is simply inconsistent with the principles of physical chemistry. The kidney does not excrete H⁺ to any greater extent as NH₄⁺ than it does as H₂O. The purpose of renal ammoniagenesis is to allow the excretion of Cl^-, which balances the charge of NH₄⁺. In all types of RTA, the defect is the inability to excrete Cl^- in proportion to excretion of Na⁺, though the precise reasons for this inability vary by RTA type. Treatment is largely dependent on whether the kidney will respond to mineralocorticoid replacement or whether there is Na⁺ loss that can be counteracted by administering NaHCO₃.

Classic distal (type I) RTA responds to NaHCO₃ replacement; generally, the required dosage is in the range of 50 to 100 mEq/day. K⁺ defects are also common in this type of RTA, and thus, K⁺ replacement is also required. A variant of the classic distal RTA is a hyperkalemic form, which is actually more common than the classic type. The central defect in this variant form appears to be impaired Na⁺ transport in the cortical collecting duct. Patients with this condition also respond to NaHCO₃ replacement.

Proximal (type II) RTA is characterized by defects in the reabsorption of both Na⁺ and K⁺. It is an uncommon disorder and usually occurs as part of Fanconi syndrome, in which reabsorption of glucose, phosphate, urate, and amino acids is also impaired. Treatment of type II RTA with NaHCO₃ is ineffective; increased ion delivery merely results in increased excretion. Thiazide diuretics have been used to treat this disorder, with varying degrees of success.

Type IV RTA is caused by aldosterone deficiency or resistance. It is diagnosed on the basis of the high serum K⁺ and the low urine pH (< 5.5). The most effective treatment usually involves removal of the cause (most commonly a drug, such as a nonsteroidal anti-inflammatory agent, heparin, or a potassium-sparing diuretic). Occasionally, mineralocorticoid replacement is required.

Gastrointestinal acidosis Fluid secreted into the gut lumen contains more Na⁺ than Cl^-; the proportions are similar to those seen in plasma. Massive loss of this fluid, particularly if lost volume is replaced with fluid containing equal amounts of Na⁺ and Cl^-, will result in a decreased plasma Na⁺ concentration relative to the Cl^- concentration and a reduced SID. Such a scenario can be prevented by using solutions such as lactated Ringer solution (LRS) instead of water or saline. LRS has a more physiologic SID than water or saline and therefore does not produce acidosis except in rare circumstances [see Positive-Anion Gap Acidosis, Lactic Acidosis, above].

Iatrogenic acidosis Two of the most common causes of a hyperchloremic metabolic acidosis are iatrogenic, and both involve administration of Cl^-. One of these potential causes is parenteral nutrition. Modern parenteral nutrition formulas contain weak anions (e.g., acetate) in addition to Cl^-, and the proportions of these anions can be adjusted according to the acid-base status of the patient. If sufficient amounts of weak anions are not provided, the plasma Cl^- concentration will increase, reducing the SID and causing acidosis.

The other potential cause is fluid resuscitation with saline, which can give rise to a so-called dilutional acidosis (a problem first described more than 40 years ago).^91,92 Some authors have argued that dilutional acidosis is, at most, a minor issue.⁹³ This argument is based on studies showing that in healthy animals, large doses of NaCl produce only a minor hyperchloremic acidosis.⁹⁴ These studies have been interpreted as indicating that dilutional acidosis occurs only in extreme cases and even then is mild. However, this line of reasoning cannot be applied to critically ill patients, for two reasons. First, it is common for patients with sepsis or trauma to require large-volume resuscitation; sometimes, such patients receive crystalloid infusions equivalent to 5 to 10 times their plasma volumes in a single day. Second, critically ill patients frequently are not in a state of normal acid-base balance to begin with. Often, they have lactic acidosis or renal insufficiency. Furthermore, critically ill patients may not be able to compensate for acid-base imbalance normally (e.g., by increasing ventilation), and they may have abnormal buffer capacity as a result of hypoalbuminemia. In ICU and surgical patients,^95–97 as well as in animals with experimentally induced sepsis,⁹⁸ saline-induced acidosis does occur and can produce significant acidemia.

The reason why administration of saline causes acidosis is that solutions containing equal amounts of Na⁺ and Cl^- affect plasma concentrations of Na⁺ and Cl^- differently. The normal Na⁺ concentration is 35 to 45 mEq/L higher than the normal Cl^- concentration. Thus, adding (for example) 154 mEq/L of each ion in 0.9% saline will result in a greater relative increase in the Cl^- concentration than in the Na⁺ concentration. So much is clear; what may be less clear is why critically ill patients are more susceptible to this disorder than healthy persons are.

It appears that many critically ill patients have a significantly lower SID than healthy persons do, even when these patients have no evidence of a metabolic acid-base derangement.⁹⁹ This finding is not surprising, in that the positive charge of the SID is balanced by the negative charges of A^- and total CO₂. Because many critically ill patients are hypoalbuminemic, A^- tends to be reduced. Because the body maintains PCO₂ for other reasons, a reduction in A^- leads to a reduction in SID so that a normal pH can be maintained. Thus, a typical ICU patient may have an SID of 30 mEq/L, rather than 40 to 42 mEq/L. If a metabolic acidosis (e.g., lactic acidosis) then develops in this patient, the SID will decrease further. If this patient is subsequently resuscitated with large volumes of 0.9% saline, a significant metabolic acidosis will result.

The clinical implication for management of ICU patients is that if large volumes of fluid are to be given for resuscitation, fluids that are more physiologic than saline should be used. One alternative is LRS, which has a more physiologic ratio of Na⁺ content to Cl^- content and thus has an SID that is closer to normal (roughly 28 mEq/L, compared with an SID of 0 mEq/L for saline). Of course, the assumption here is that the lactate in LRS is metabolized, which, as noted (see above), is almost always the case. Volume resuscitation also reduces the weak acid concentration, thereby moderating the acidosis. One ex vivo study concluded that administration of a solution with an SID of approximately 24 mEq/L will have a neutral effect on the pH as blood is progressively diluted.¹⁰⁰

Unexplained hyperchloremic acidosis Critically ill patients sometimes manifest hyperchloremic metabolic acidosis for reasons that cannot be determined. Often, other coexisting types of metabolic acidosis are present, making the precise diagnosis difficult. For example, some patients with lactic acidosis have a greater degree of acidosis than can be explained by the increase in the lactate concentration,⁴⁰ and some patients with sepsis and acidosis have normal lactate levels.¹⁰¹ In many instances, the presence of unexplained anions is the cause,^40–42 but in other cases, there is a hyperchloremic acidosis. Saline resuscitation may be responsible for much of this acidosis (see above), but experimental evidence from endotoxemic animals suggests that as much as a third of the acidosis cannot be explained in terms of current knowledge.⁹⁸

One potential explanation for unexplained hyperchloremic acidosis is partial loss of the Donnan equilibrium between plasma and interstitial fluid. The severe capillary leakage that accompanies this loss of equilibrium results in loss of albumin from the vascular space, which means that another ion must move into this space to maintain the charge balance between the two compartments. If Cl^- moves into the plasma space to restore the charge balance, a strong anion is replacing a weak anion, and a hyperchloremic metabolic acidosis results. This hypothesis appears reasonable but, at present, remains unproven.

Metabolic alkalosis occurs as a result of an increased SID or a decreased A_tot, secondary either to loss of anions (e.g., Cl^- from the stomach and albumin from the plasma) or increases in cations (rare). Metabolic alkaloses can be divided into those in which Cl^- losses are temporary and can be effectively replaced (chloride-responsive alkaloses) and those in which hormonal mechanisms produce ongoing losses that, at best, can be only temporarily offset by Cl^- administration (chloride-resistant alkaloses) [see Table 5]. Like hyperchloremic acidosis, metabolic alkalosis can be confirmed by measuring the urine Cl^- concentration.

Chloride-Responsive Alkalosis

Chloride-responsive metabolic alkalosis usually occurs as a result of loss of Cl^- from the stomach (e.g., through vomiting or gastric drainage). Treatment consists of replacing the lost Cl^-, either slowly (with NaCl) or relatively rapidly (with KCl or even HCl). Because chloride-responsive alkalosis is usually accompanied by volume depletion, the most common therapeutic choice is to give saline along with KCl. Dehydration stimulates aldosterone secretion, which results in reabsorption of Na⁺ and loss of K⁺. Saline is effective even though it contains Na⁺ because the administration of equal amounts of Na⁺ and Cl^- yields a larger relative increase in the Cl^- concentration than in the Na⁺ concentration (see above). In rare circumstances, when neither K⁺ loss nor volume depletion is a problem, it may be desirable to replace Cl^- by giving HCl.

Diuresis and other forms of volume contraction cause metabolic alkalosis mainly by stimulating aldosterone secretion; however, diuretics also directly stimulate excretion of K⁺ and Cl^-, further complicating the problem and inducing metabolic alkalosis more rapidly.

Chloride-Resistant Alkalosis

Chloride-resistant alkalosis [see Table 5] is characterized by an increased urine Cl^- concentration (> 20 mmol/L) and ongoing Cl^- loss that cannot be abolished by Cl^- replacement. Most commonly, the proximate cause is increased mineralocorticoid activity. Treatment involves identification and correction of the underlying disorder.

Alkalosis from Other Causes

In rare situations, an increased SID—and therefore metabolic alkalosis—occurs secondary to cation administration rather than to anion depletion. Examples include milk-alkali syndrome and intravenous administration of strong cations without strong anions. The latter occurs with massive blood transfusion because Na⁺ is given with citrate (a weak anion) rather than with Cl^-. Similar results ensue when parenteral nutrition formulations contain too much acetate and not enough Cl^- to balance the Na⁺ load.

CO₂ is produced by the body at a rate of 220 ml/min, which equates to production of 15 mol/L of carbonic acid each day.¹⁰² By way of comparison, total daily production of all the nonrespiratory acids managed by the kidney and the gut amounts to less than 500 mmol/L. Pulmonary ventilation is adjusted by the respiratory center in response to P_aCO₂, pH, and PO₂, as well as in response to exercise, anxiety, wakefulness, and other signals. Normal P_aCO₂ (40 mm Hg) is attained by precisely matching alveolar ventilation to metabolic CO₂ production. P_aCO₂ changes in predictable ways as a compensatory ventilatory response to the altered arterial pH produced by metabolic acidosis or alkalosis [see Table 1].

One option for treatment of hypercapnia is noninvasive ventilation with a bilevel positive airway pressure (BiPAP) system. This technique may be useful in the management of some patients, particularly those whose sensorium is not impaired.¹⁰³ Rapid infusion of NaHCO₃ in patients with respiratory acidosis may induce acute respiratory failure if alveolar ventilation is not increased to account for the increased CO₂. Thus, if NaHCO₃ is to be given, it must be administered slowly, with alveolar ventilation adjusted appropriately. Furthermore, it must be remembered that NaHCO₃ works by increasing the plasma Na⁺ concentration; if this effect is not possible or not desirable, NaHCO₃ should not be given.

Occasionally, it is useful to reduce CO₂ production. This can be accomplished by reducing the amount of carbohydrates supplied in feedings (in patients requiring nutritional support), controlling body temperature (in febrile patients), or providing sedation (in anxious or combative patients). In addition, treatment of shivering in the postoperative period can reduce CO₂ production. Rarely, however, can hypercapnia be controlled with these CO₂-reducing techniques alone.

Permissive hypercapnia. In the past few years, there has been considerable interest in ventilator-associated lung injury. Over distention of alveoli can result in tissue injury and microvascular permeability, which lead to interstitial and alveolar edema. In animal studies, prolonged use of elevated airway pressures and increased lung volumes resulted in increased pathologic pulmonary changes and decreased survival when compared with ventilatory strategies employing lower pressures and volumes.^104,105 In a large multicenter clinical trial, simply lowering the tidal volume on the ventilator from 12 ml/kg to 6 ml/kg in patients with acute lung injury resulted in a 9% absolute reduction in mortality risk.¹⁰⁶ Although the protocol followed in this trial did not advocate a reduced minute ventilation and hence an elevated P_aCO₂, this approach, often referred to as permissive hypercapnia or controlled hypoventilation, has become increasingly popular. Uncontrolled studies suggest that permissive hypercapnia may reduce mortality in patients with severe ARDS.²⁰ This strategy is not, however, without risks. Sedation is mandatory, and neuromuscular blocking agents are frequently required. Intracranial pressure rises, as does transpulmonary pressure; consequently, this technique is unusable in patients with brain injury or right ventricular dysfunction. There is controversy regarding how low the pH can be allowed to fall. Some authors have reported good results with pH values of 7.0 or even lower,²⁰ but most have advocated more modest pH reductions (i.e., ³ 7.25).

Respiratory alkalosis may be the most frequently encountered acid-base disorder. It occurs in residents of high-altitude locales and in persons with any of a wide range of pathologic conditions, the most important of which are salicylate intoxication, early sepsis, hepatic failure, and hypoxic respiratory disorders. Respiratory alkalosis also occurs in association with pregnancy and with pain or anxiety. Hypocapnia appears to be a particularly strong negative prognostic indicator in patients with critical illness.¹⁰⁷ Like acute respiratory acidosis, acute respiratory alkalosis results in a small change in the HCO₃^- concentration, as dictated by the Henderson-Hasselbalch equation. If hypocapnia persists, the SID begins to decrease as a consequence of renal Cl^- reabsorption. After 2 to 3 days, the SID assumes a new and lower steady state.¹⁰⁸

Severe alkalemia is unusual in respiratory alkalosis. Management therefore is typically directed toward the underlying cause. In general, these mild acid-base changes are clinically important more for what they can alert the clinician to, in terms of underlying disease, than for any direct threat they pose to the patient. In rare cases, respiratory depression with narcotics is necessary.

Pseudorespiratory Alkalosis

The presence of arterial hypocapnia in patients experiencing profound circulatory shock has been termed pseudorespiratory alkalosis.¹⁰⁹ This condition occurs when alveolar ventilation is supported but the circulation is grossly inadequate. In such circumstances, the mixed venous PCO₂ is significantly elevated, but the P_aCO₂ is normal or even decreased as a consequence of reduced CO₂ delivery to the lung and increased pulmonary transit time. Overall CO₂ clearance is therefore markedly decreased, and profound tissue acidosis—usually both metabolic and respiratory—ensues. The metabolic component of the acidosis comes from tissue hypoperfusion and hyperlactatemia. Arterial oxygen saturation may also appear adequate despite tissue hypoxemia. Pseudorespiratory alkalemia is rapidly fatal un less the patient's systemic hemodynamic status can be normalized.