Metabolic acidosis in the critically ill: Part 2. Causes and treatment


  • C. G. Morris,

    1. Consultants, Intensive Care Medicine and Anaesthesia, Derby Hospitals Foundation Trust, Derby Royal Infirmary, London Road, Derby DE1 2QY, UK
    Search for more papers by this author
  • J. Low

    1. Consultants, Intensive Care Medicine and Anaesthesia, Derby Hospitals Foundation Trust, Derby Royal Infirmary, London Road, Derby DE1 2QY, UK
    Search for more papers by this author

Craig Morris


The correct identification of the cause, and ideally the individual acid, responsible for metabolic acidosis in the critically ill ensures rational management. In Part 2 of this review, we examine the elevated (corrected) anion gap acidoses (lactic, ketones, uraemic and toxin ingestion) and contrast them with nonelevated conditions (bicarbonate wasting, renal tubular acidoses and iatrogenic hyperchloraemia) using readily available base excess and anion gap techniques. The potentially erroneous interpretation of elevated lactate signifying cell ischaemia is highlighted. We provide diagnostic and therapeutic guidance when faced with a high anion gap acidosis, for example pyroglutamate, in the common clinical scenario ‘I can’t identify the acid – but I know it's there'. The evidence that metabolic acidosis affects outcomes and thus warrants correction is considered and we provide management guidance including extracorporeal removal and fomepizole therapy.

In Part 1 of this review article, we considered the classification and diagnostic approach to metabolic acidosis in the critically ill, including base excess, CO2/HCO3, and anion gap, and proposed albumin-corrected anion gap-based techniques for bedside use in the critically ill. In Part 2 we examine the types of acidosis further, using a (modified) anion gap methodology, and emphasise points of clinical relevance and common pitfalls in practice. It is often unclear whether metabolic acidosis is a ‘primary’ abnormality, i.e. the patient is unwell because they have accumulated H+, or an epiphenomenon reflecting the effects of the underlying process, or the accumulation a toxic aprote anion species. We will consider the impact that metabolic acidosis may have on prognosis, whether its treatment can improve outcome, and propose a diagnostic and management strategy for the clinician faced with a critically ill patient with metabolic acidosis.

Elevated anion gap acidoses

Elevated anion gap acidosis represents the accumulation of an (acidic) anion of which there are three ‘and-a-half’ examples:

  • 1) Lactic acidosis. Lactic acid, like most substances with a pKa of less than 4 (actual pKa 3.78), circulates almost entirely as the freely dissociated aprote anion lactate (i.e. it releases its proton) at physiological pH, strongly favouring the right of the equation below:
    (Hyperlactataemia, lactic acidaemia and acidosis are discussed below.)
  • 2) Ketones. Ketones are also strong acids and may be generated in processes which yield excessive amounts of acetyl CoA, which is unable to proceed to the tricarboxylic acid cycle. The biochemical adage ‘fat burns in the flame of carbohydrate’ refers to the limiting step of oxaloacetate accepting acetyl CoA and forming citrate in the Kreb's cycle. If carbohydrate is limited, oxaloacetate is diverted to gluconeogenesis. Ketoacids may accumulate when the principal metabolic substrate becomes lipid rather than carbohydrate as hepatic lipolysis is less tightly regulated than glycolysis.
  • 3) Toxin ingestion. Examples include ethanol (which will often be associated with lactic and ketoacidosis), methanol, salicylates and ethylene glycol (the actual acid is the metabolite oxalate) and propylene glycol. Less common causes include acetic acid and ascorbic acid (vitamin C). A useful screening test in this setting is to calculate the osmolar gap:

A number of formulae exist to calculate plasma osmolality but one example is 2 × [Na+] + [urea] + [glucose][1, 2]. The effective tonicity neglects the osmotically equilibrated urea. If the gap is > 20 mOsmol.l−1 then it is likely to be significant and should prompt a specific search for one of the anions discussed above [3, 4]. When considering volatile compounds (e.g. ethanol or methanol) the osmolality should be measured by depression of freezing point rather than vapour pressure.

Uraemia is included as the ‘half’ cause of elevated anion gap acidosis. The traditional explanation of uraemic acidosis has been the accumulation of acidic anions that cannot be cleared by a failing kidney [5, 6]. However, it is clear that the situation is more complex and application of physicochemical principles has challenged a number of basic assumptions. During acute renal failure and critical illness, approximately half of patients have a normal anion gap [7]. The use of the strong ion gap (SIG) suggests accumulation of unmeasured anions and significant hyperphosphataemia, in conjunction with pH raising hypoalbuminaemia. Chronic renal failure may not be as strongly associated with unmeasured anions [8] or feature sulphate more prominently [9]. A confounding factor in acute renal failure is that the acidosis may reflect excessive liver urea synthesis which generates H+ rather than failure to clear NH4+[10, 11]:


Hyperlactataemia, lactic acidaemia and acidosis

Hyperlactataemia refers to an elevated plasma concentration of lactate anions. These may have formed when endogenous lactic acid dissociated or come from an exogenous source with an infusion of a lactate-containing solution. In clinical practice lactic acidaemia may be defined as a pH < 7.35 with a lactate concentration > 5 mmol.l−1[12, 13]. Lactic acidaemia typically develops as a result of endogenously produced lactic acid with lactate being measured as the dissociated base. Exogenous lactate, e.g. balanced intravenous fluids or renal replacement fluids, are buffered in solution, typically with the sodium salt [14, 15]. Although these solutions have inherent acidity (narrow strong ion difference) and (racemic) lactate concentrations of 30–40 mmol.l−1, the metabolism of administered lactate, in conjunction with Na+ loading, has an overall alkalinising effect.

Acute renal failure may contribute to lactic acidosis as the kidneys play an under-recognised role in metabolising lactate [16, 17]. Animal work suggests lactate is almost entirely filtered at the glomerulus and undergoes almost 100% re-absorption at the proximal convoluted tubule [18]. Nonetheless, during conditions of hyperlactataemia there is likely to be relatively little clearance in the urine itself [19, 20].

The lactate molecule possesses a chiral centre and may occur in two stereo-isomers, laevo (l) and dextro (d) rotatory. Lactate generated by endogenous metabolism is almost exclusively the l-isomer. Commercial lactate-containing solutions (e.g. Hartmann's solution) and most renal replacement solutions are racemic mixtures. The rate of metabolism of d-lactate, relative to the l-isomer [21–23], is unclear. Several early reports of elevated lactate as a marker of ischaemic bowel actually referred to d-lactate produced by luminal bacteriae [24, 25]. Near-patient blood-gas analysers use a lactate oxidase reaction whilst the laboratory assay uses lactate dehydrogenase assay. The former typically overestimates by 13%, especially in the presence of anaemia [26], and both only detect the l-isomer. Rarely, certain organic acids can cross-react with the lactic oxidase reagents of the near-patient analyser yet the assay is lower (and correct) at the laboratory. This potential ‘lactate gap’ has been described following ethylene glycol toxicity [27, 28].

During critical illness, the source of lactate is often believed to be ischaemic, anaerobically metabolising tissues, amongst which the gut and muscle are popular ‘suspects’. Superficially this is supported by lactate as an adverse prognostic marker [12, 13]. However, lactate metabolism in critical illness is complex and often does not indicate ischaemic tissues [29–31]. The anatomical source of lactate in critical illness is not consistent and may be dependent on the disease process and timing. Evidence for the gut as a source of lactate is limited [32–36] and almost certainly confounded by concurrent use of beta agonists, e.g. adrenaline [37–39]. Muscle appears a net consumer of lactate during endotoxaemia [30, 31]. The lungs appear to be a consistent source of lactate, at least in the setting of acute lung injury [40–43]. There is no simple explanation for the production and metabolism of lactate in critical illness.

Although lactic acidaemia is an adverse prognostic factor during critical illness this may not mean that lactate is an appropriate therapeutic target. ‘Early goal-directed therapy’ in severe sepsis is an example of outcome benefit which arguably targets the patient's primary pathophysiology rather than lactic acidosis per se [44, 45]. Normalisation of high lactate values (endogenous or exogenous infusion) is indicative of better outcomes in sepsis but this may be a surrogate marker for severity of illness [46, 47]. Lactic acidaemia may thus be regarded as a sensitive marker of cell ischaemia and an adverse prognostic marker, but with a poor specificity for the former [48–53].

Lactic acidosis may be subdivided into types A and B. Given the complexities of critical illness, the distinction between the two may be difficult and even artificial. The following list is not exhaustive and alternatives are available elsewhere [54].

Type A lactic acidosis is associated with tissue ischaemia and anaerobic respiration.

Type B lactic acidosis can be further subdivided into

  • B1: an underlying disease process,
  • B2: pharmacological or toxic and
  • B3: metabolic disorder.

One mechanism of elevated lactate is mediated by pyruvate dehydrogenase inhibition, e.g. by endotoxin [56, 57], and the specificity of lactate measurement for ischaemia may be increased by considering a lactate : pyruvate ratio where a ratio < 10 : 1 is unlikely to be consistent with ischaemia [31, 57–59].

Absolute or functional thiamine deficiency (Shoshin beriberi) is associated with impaired cardiac function and/or impaired neurology [60]. One should suspect this among patients with poor nutritional status who are being sugar loaded, enterally or parenterally (see total parenteral nutrition solutions below) and re-feeding syndrome is often apparent [61]. Being water soluble, body stores of thiamine are limited and deficiency can develop rapidly. Given its high degree of safety we would strongly recommend a trial of intravenous thiamine and B vitamin replacement for cases of impaired cardiac function, impaired neurology, lactic acidosis and other ‘unidentified’ high anion gap acidoses in the critically ill [62]. Up to 20% of critically ill patients may be thiamine deficient [63].

Drugs. In critical illness, beta agonists, including adrenaline, salbutamol and dobutamine, stimulate glycolysis with production of an excess of pyruvic acid which may not be cleared due to inhibited pyruvate dehydrogenase [55] and which is converted to lactate [64]. Furthermore, stimulation of Na+/K+ ATPase of the muscle cell membrane enhances cellular glycolysis [65, 66]. Experimental blockade of adrenoceptors can prevent this lactic acidosis [67]. A common clinical scenario occurs during resuscitation of patients using volume loading (iatrogenic hyperchloraemia) and adrenaline infusions, where a chloride load and type B lactic acidosis are interpreted as ‘shock’ requiring more fluid and more beta agonist; this leads to a vicious cycle and potentially ‘over-resuscitation’[48]. It is essential to diagnose this syndrome correctly and to adjust management accordingly, as discussed in Figure 1. In clinical practice this phenomenon is less common with noradrenaline and the authors rarely use adrenaline.

Figure 1.

 Erroneous interpretation of hyperchloraemia and lactic acidosis (type B) during resuscitation of septic shock. A patient is admitted in extremis with septic shock via the emergency department. Fluid loading with crystalloid (0.9% saline) and colloid is ongoing and an adrenaline infusion commenced with 50 mg intravenous hydrocortisone to maintain a mean arterial pressure of 65 mmHg. Arterial blood gas and biochemical sampling results are shown. Na+, sodium; K+, potassium; Cl, chloride; Alb, albumin; AG, anion gap (not corrected for albumin); Agcor, anion gap corrected for albumin. Concentrations expressed in mmol.l−1 except albumin g.l−1. The patient's hyperlactataemia is noted, interpreted as tissue ischaemia (type A), and stimulates further fluid loading and the adrenaline infusion is increased. The surgeons are consulted regarding suspected bowel ischaemia as further analysis shows pH 7.10 and lactate 9.7 mmol.l−1. The patient has received 5.0 l of mixed crystalloid and colloid and the adrenaline infusion is at 30 μg.min−1. The plasma chloride, having been 97 mmol.l−1 on admission is now 119 mmol.l−1 and a trial of noradrenaline, to maintain the same mean arterial pressure, is initiated while reducing adrenaline doses. The pH slowly improves and the lactate normalises within 3 h, adrenaline is withdrawn, and laparotomy is not undertaken; community acquired pneumonia and septic shock settles in the intensive care unit. An additional source of diagnostic confusion was the associated hypoalbuminaemia (19 g.l−1), which in conjunction with iatrogenic hyperchloraemia ‘lowers’ the anion gap. This case emphasises the importance of correlating the blood-gas analysis and diagnosis with the clinical presentation, which was strongly suggestive of an initial organic tissue acid (type A lactic acidosis) and which deteriorated when therapeutic interventions were initiated. The admission anion gap of 32 strongly suggested the presence of an acidic anion and could not quantitatively be explained by the [lactate] 4.6 mmol.l−1. This illustrates the presence of multiple classes of acid in the critically ill, and it is highly unlikely that searching for the specific anion with a clinical syndrome of sepsis having been identified would have helped management. This discussion also illustrates the limitations of the type A vs type B lactic acidosis classification in considering a shocked critically ill patient who may have cellular ischaemia, enhanced glycolysis, hepatic and renal impairment and reduced lactate metabolism, and concurrent catecholamine infusion. At 4 h, although the anion gap appeared to have ‘improved’, the patient is arguably sicker and the biochemical components of this calculation, in particular chloride concentrations following volume resuscitation and hypoalbuminaemia, demonstrate its limitations.

The biguanides are the therapy of choice in obese type 2 diabetics, and in the UK only metformin is available [68]. Metformin-associated lactic acidosis (MALA) is typically associated with intentional overdose and/or renal impairment and often associated with minimal sugar abnormality. The mechanism is multifactorial [69] and the current estimated incidence of 5 per 100 000 is likely to rise [70]. Metformin is normally excreted in the urine and can be removed by extracorporeal techniques (see Table 2). Increasingly, antiretroviral therapy [71, 72] and propofol are implicated in type B lactic acidosis. The use of fructose, xylitol, sorbitol and glycerol as glucose alternatives in parenteral nutrition, being largely insulin independent for their metabolism, is uncommon at present [73]. Sorbitol may be included in endoscopic solutions [74].

Ketone species and ketoacidosis

The main pathophysiological ketones are acetoacetate and beta-hydroxybutyrate. These are produced in hepatic mitochondria, acetone being a volatile by-product of acetoacetate. Ketones are strong acids and circulate as free anions, particularly acetoacetate (pKa 3.58, hydroxybutyrate pKa 4.70). Acetoacetate is generated predominately under conditions of adequate oxygen delivery (normal plasma concentrations 20–80 μmol.l−1) and hydroxybutyrate (normal 60–170 μmol.l−1) under anaerobic conditions [75], but only acetoacetate can ultimately enter the tricarboxylic acid cycle.

These species are formed in varying proportions according to the prevailing metabolic climate and this is clinically important. The urinary dipstick for ketones utilises a nitroprusside reagent and detects acetoacetate but not beta-hydroxybutyrate. If patients are shocked, the latter may be produced in ratios up to 3 : 1 during diabetic ketoacidosis, i.e. the urine may erroneously appear to contain few ketones and their concentration may paradoxically rise as perfusion and oxidative status improves [76]. Quantitative measurement in plasma avoids this error and some workers suggest using the acetoacetate : beta-hydroxybutyrate ratio to identify cell ischaemia, possibly in conjunction with lactate : pyruvate, as this parameter appears to be dependent on hepatic blood flow and mitochondrial redox status falling from 1.0 : 1.0 in health to 0.2 : 1.0 during septic shock [57].

The renal threshold for ketones is low and there is no active re-uptake mechanism. Once filtered, they effectively ‘pass through’, unlike substances such as glucose and amino acids. In prolonged and/or profound ketosis the loss of ketones may make the anion gap appear paradoxically less than expected, compounded by the body's retention of chloride and bicarbonate.

Normal anion gap acidoses

Due to the terms included in calculating the anion gap it can be seen that in normal anion gap acidosis, hyperchloraemia is typical (hence the alternative titles of high anion gap normochloraemic or normal anion gap hyperchloraemic acidosis). The pathophysiological process commonly includes bicarbonate loss. Causes include:

  • Infusion therapies. Rapid administration of chloride-rich solutions (essentially colloids and normal 0.9% saline containing Cl 154 mEq.l−1) will produce a hyperchloraemic normal anion gap acidosis (see below). Balanced lactate solutions typically contain chloride at a concentration of 110 mEq.l−1.
  • • Alkaline gastrointestinal losses, especially lower intestinal as gastric losses and vomiting, tends to produce a hypochloraemic alkalosis. This may reflect gastric outlet obstruction, fistulae or surgical anastamoses.
  • • Renal tubular acidoses– considered below.
  • Ureteric diversion, e.g. ileal conduit. The release of urea into the ‘gut’ encourages bacterial degradation with release of NH3 and H+ causing acidosis. Metabolic acidosis appears higher with continent reconstructions, and vitamin B12 deficiency is reported [77].

Infusion of chloride ions and ‘hyperchloraemic acidosis

Saline 0.9% has a chloride concentration of 154 mmol.l−1 and most commercially available colloids are suspended in saline. Considering that a normal plasma chloride concentration is 100 mmol.l−1, all these fluids have a relatively high concentration of chloride. Most of these solutions possess an acidic pH of approximately 5.0, which has been ascribed to equilibration with atmospheric CO2[78]. Sodium chloride in water has a pH of 7.0 and the acidifying effect is related to the sodium concentration [79]. In plasma at a pH of 7.40 its behaviour is markedly different [80] tending to act as a potent acidifying agent [14, 15, 81–85]. This may become significant during volume resuscitation of the critically ill and has also been described as a consequence of normovolaemic haemodilution and loading cardiopulmonary bypass circuits. The impact that hyperchloraemic acidosis may have on outcome remains controversial [86–90] but at the very least it can produce diagnostic confusion. The largest evaluation of saline in the critically ill could find no overall evidence of significant differences between 0.9% saline and 4% albumin (suspended in saline) in a protocol guided by blood pressure [91]. The saline group received 1.4 times more fluid volume, and arguably more Cl, yet subsequent analysis suggested a paradoxical, clinically insignificant, increase in [Cl], and lower pH increases in the albumin group [92].

A number of alternatives exist to reduce this chloride load, including consideration of reduced volume colloid, replacing chloride anions with balanced lactate solutions and suspending colloids in such solutions other than 0.9% saline [89, 93]. It is likely that balanced lactate solutions, in providing complex mixes of sodium and lactate and reduced chloride, help correct a strong ion difference (SID) and improve acidosis as it is the relative ratios of (strong) ions rather than their absolute equivalents that determine the overall effects on acid-base status [94–96]. After all, 0.9% sodium chloride, water and 5% dextrose all have an SID of 0.

Renal tubular acidoses (RTA)

Renal tubular acidosis may be defined as an intrinsic deficiency of the kidney to clear the body’s acid load in the presence of an adequate glomerular filtration rate. If we accept acute tubular necrosis (ATN) as the most common model of intrinsic renal failure in the critically ill [97, 98] then it is apparent that tubular dysfunction and thus at least an element of RTA will frequently occur at the same time. The similarities that such functional disorders may share with ‘textbook’ chronic RTAs are limited where a hyperchloraemic normal anion gap acidosis is expected. Furthermore, in critical illness RTA will rarely be the only cause of acidosis. Once identified RTA may be further subdivided into:

Distal: classical type 1 involving the collecting ducts, which are relatively impermeable to H+, allowing excretion of highly acidic urine (pH 5.0). Whereas intercalated cells actively secrete H+, principal cells exchange sodium for H+. A major limitation to the quantitative molar urinary clearance of acid is the buffering capacity, where H+ combines with ammonia to form NH4+. The remaining H+ is buffered by titratable acid, principally phosphate and citrate. This distal excretion of acid is enhanced by aldosterone and hypokalaemia, and often both together in the presence of hypovolaemia, with an associated metabolic alkalosis. From a therapeutic perspective, aggressive correction of hypokalaemia can aid renal acid clearance.

Proximal: classical type 2 constitutes an HCO3 resorption defect with inadequate proximal acidification, often with associated polyuria, natriuresis and secondary kaliuresis. In isolated proximal disease urinary acidification is preserved, even in the face of systemic acidosis, but HCO3 losses persist. This condition can be therapeutically created with acetazolamide during therapy of metabolic alkalosis.

Type 3: 1 + 2 combined.

Type 4 is an absolute or functional renin/mineralocorticoid deficiency where potassium retention results in impaired Na+/H+ and K+ exchange mechanisms [99]. Adequate tubular Na+ is required for H+ exchange and natriuresis must be present for diagnosis (> 40 mmol.l−1). Outside critical care, this is classically seen in diabetics with adrenal failure, with prominent hyperkalaemia and cardiac events. In the critically ill this may be seen in association with renal disease, e.g. obstruction, toxins such as cyclosporin and distal nephron acting diuretics, e.g. spironolactone or amiloride. Trimethoprim is responsible for an amiloride-like blockade of the Na+ channels producing potassium retention (hyperkalaemia) and a voltage-dependent type 4 RTA, and this may be seen during therapy of ventilator-associated or Pneumocystis carinii pneumonia with cotrimoxazole [100–102].

It is beyond the scope of this article to consider the many inherited and systemic diseases that may produce RTA. In the presence of systemic metabolic acidosis the urine, if produced, should be maximally acidified, i.e. laboratory analysis pH 5.0–5.5 (typical maximal urinary alkalinisation pH is 8–8.5). If not, this is suggestive of a urinary acidification (distal) defect; a definitive diagnosis is obtained by assay of urinary NH4+ (not readily available) or, as a surrogate, an adapted urinary anion gap or SID [103]. However, excreted H+ is buffered in urine. The absolute pH may thus be misleading as a quantitative guide to acid clearance, e.g. acidic urine may reflect an inability to produce renal NH3 and consequently the quantitative acid excretion is low [104]. If administration of 1.5−−1 HCO3 fails to correct an acidosis (equivalent to the typical titratable class III acid load) then it is likely an element of proximal RTA is present, especially if HCO3 is cleared rapidly in the urine and alkalinises it. However, in the critically ill this bicarbonate dose will also be titrated against other classes of acid in the plasma, e.g. lactate, making this test less informative.

I can't identify the acid – but I know it's there!

This scenario will most commonly arise due to an elevated anion gap in the presence of a metabolic acidosis which is not due to the three (and-a-half) causes of lactate, ketones, ingestion or renal failure. We routinely measure only a few acids but within the realms of biochemistry there are literally hundreds that exist and which could cause an elevated anion gap acidosis. This can therefore produce the clinical scenario ‘I can’t identify the acid – but I know it's there! It is common that several acids accumulate when patients become critically ill, e.g. both lactate and ketones, so judgement is required in determining the relative importance of these acids [105]. Acidic anions identified during chronic renal failure include sulphate, hydroxypropionate, hippurate, oxalate, glutamate, aspartate and funapropionate, but their presence during acute renal failure and critical illness is largely speculative [106–108]. Before embarking on an extensive (and expensive) search for non-routine acids it is worth considering whether this will usefully alter management. We feel it is of more benefit to review the clinical context, administer parenteral B vitamins in unexplained acidosis and review prescribed and non-prescribed medications meticulously.

Concerns regarding the ability of the anion gap to identify unidentified anions have increasingly made a physicochemical approach more popular [109, 110]. However, bedside use of standard blood gas results supplemented with anion gap corrected for albumin (see Part 1 in series) offers comparable diagnostic accuracy for critically ill patients [111].

Toxins and drugs including methanol, salicylates or ethylene glycol and their metabolites, e.g. oxalate

Certain drug infusions, e.g. nimodipine (20% ethanol), lorazepam and etomidate (in propylene glycol solvent, the latter rarely used by infusion in the UK) or methylene blue, may also produce acidosis either directly or through organ damage such as renal failure [112–115]. The administration of paraldehyde (seizures) and ingestion of formaldehyde can produce high anion gap acidoses. Calculation of the osmolar gap is usually helpful (see Part 1).


Ethanol is capable of producing a variety of metabolic acidoses both directly and indirectly. Acute and chronic ingestion is associated with ketoacidosis, lactic acidosis, aldehyde accumulation and hepatic derangement. The benefits of compound B vitamins are re-emphasised.

Total parenteral nutrition (TPN) solutions

TPN solutions are acidic with a low SID and low SIG, and are described as producing acidosis with high anion gap, directly and indirectly [116, 117]. By containing high phosphate loads (lipid content) and sulphate loads (amino acids) TPN can be a direct cause of acidosis, especially in the presence of reduced renal clearance.

Propofol infusion (syndrome)

Propofol infusion (syndrome) has been reviewed elsewhere [118] and comprises lipaemia, cardiac failure, and profound lactic acidosis typically in the setting of catecholamine infusions and multiple organ failure or shock [119, 120] and is not restricted to children. Clinicians should also be aware of the lipid load of concurrent propofol and TPN and restrict lipid infusions to < 1.5−−1.

Hepatic failure

Hepatic failure is associated with accumulation of a multitude of acids including organic acids (lactate and ketones) and ammonium, often compounded by renal and multiple organ failure. Identification of individual compounds is unlikely to be helpful or radically alter management. Partial support of the excretory functions of the liver (and kidney) is possible using albumin-based extracorporeal techniques.

Pyroglutamic (5-oxoproline) acid

Pyroglutamic (5-oxoproline) acid is associated with paracetamol, flucloxacillin and vigabatrin therapy and is formed in conditions of reduced hepatic glutathione synthase such as sepsis or hepatic/renal dysfunction [121–124]. The acidosis is temporally related to the drug administration and diagnosis suggested by improvement with removal of the offending agent. N-acetyl cysteine may aid resolution. This acid may be detected by urinary organic acid screen or plasma assay, although this is not routinely available. Although paracetamol has an excellent safety profile in health, reports exist of hepatic necrosis at otherwise safe doses [125, 126].

Fatty acid metabolites

Fatty acid oxidation disorders are rare metabolic syndromes associated with lipid metabolism disorders and varying degrees of high anion gap acidosis and hypoglycaemia; presentation is rare but recognised in adults. Although ketones or lactate may also be elevated they do not account for the grossly elevated anion gap. The acidosis is due to (dicarboxylic) acidic derivatives of long chain fatty acids which have been unable to enter the mitochondrion. There may be B-vitamin responsiveness and again we strongly recommend that intravenous compound B vitamins be administered for all causes of unexplained metabolic acidosis in the critically ill. A number of syndromes are described including multiple acyl CoA deficiency (MADD) [127] and there may be improvement following dextrose loading and l-carnitine, which facilitates long chain fatty acid entry into the mitochondria [128, 129].

Metabolic disorders of amino acids and the urea cycle

Metabolic disorders of amino acids and the urea cycle can occasionally present outside of childhood, often following a dietary protein challenge. A useful screen is plasma ammonia level. Encephalopathy is usually present and plasma glutamine levels are often raised [130–132]. Moderately elevated plasma ammonia levels are common during critical illness associated with hepatic and renal dysfunction.

Enhanced cell breakdown

Enhanced cell breakdown, e.g. rhabdomyolysis, status epilepticus, tumour lysis syndrome, malignant hyperpyrexia or heat stroke produce multifactorial metabolic acidaemia (including lactic acidosis and renal failure), but the release of cellular components in large amounts is frequently associated with acidaemia. Uric acid (pKa 5.8), a metabolite of deoxyribose nucleic acid metabolism, is not a prominent cause of acidosis in plasma but does dissociate effectively in urine.

Organic solvents

Organic solvents may be suggested by occupational exposure or suicidal intent and include toluene (metabolite hippurate) in addition to the alcohols discussed above.

Does metabolic acidosis or acidaemia affect clinical outcomes?

In deciding whether metabolic acidosis is harmful, a number of hypotheses present themselves:

  • 1Is an elevated [H+] directly damaging?
  • 2Is the associated aprote acidic anion harmful?
  • 3Is a combination of 1 and 2 harmful?
  • 4Is it the primary abnormality, e.g. sepsis, which is harmful and the acid-base changes largely secondary phenomena [133]?

The controversies surrounding the significance of hyperchloraemic metabolic acidosis illustrate the notorious difficulties in distinguishing association and causation [86–97]. One possible level of distinction is metabolic acidosis due to tissue or organic acids (e.g. lactate, overlapping with raised anion gap) and the inorganic acidoses, including hyperchloraemia [133]. The former are generally considered to reflect serious pathology with potential adverse outcomes [88].

The body has an astonishing capacity to tolerate relatively severe levels of acidosis. From a pH of 7.4 (40 nmol.l−1) to a severe acidosis of pH 6.7 (equivalent to > 175 nmol.l−1) the H+ concentration increases by a factor of four and this could not be tolerated with other ions such as potassium. However, acidosis is associated with many adverse haemodynamic, respiratory, cerebral and metabolic adverse effects which are reviewed elsewhere [134]. Conversely, in animal models acidosis can protect myocardial and hepatic cells during hypoxia [135, 136] and elevated [H+] is well tolerated, often for protracted periods, from a variety of causes including permissive hypercapnia (a respiratory acidosis) in managing acute respiratory distress syndrome [137–139]. However, there is an association between persistent acidosis and increased mortality in the critically ill receiving high volume haemofiltration, even if urea clearance is achieved [140].

Although we measure [H+] in (arterial) plasma, this may not accurately reflect the intracellular status, and in health a typical intracellular pH is relatively acidotic, pH 7.0–7.3, depending on the cell type. The use of gastric tonometry and pHi are examples where arterial plasma sampling may not reflect intracellular processes [141, 142] and during cardiac arrest venous or pulmonary artery sampling may be the most appropriate [143]. If the plasma metabolic acidosis is ‘contained’ in the extracellular space then it may have little role in pathophysiology [144].

It is possible that the anion species, rather than [H+], is harmful, although distinguishing this from the disease process is difficult. Attempts to identify otherwise undetected anions have suggested undetected anions have no apparent impact on outcome [108, 145, 146], although following trauma or major vascular injury this may not be the case [147–150]. Interrogation of arterial (lactate) and splanchnic (mucosal Pco2) compartments may improve the predictive ability [151, 152], although no perfect marker of cellular dysoxia currently exists [153]. Chloride ions can adversely affect the course of experimental sepsis [90, 93]. However, experimental toxicity of anions, e.g. lactate [154], is difficult to reconcile with significant, if brief, elevations associated with anaerobic exercise [155] or accumulation of lactate during, for example, renal replacement therapy. Similarly, the SAFE study failed to demonstrate a difference in outcome between groups of critically ill patients despite differing administration of chloride (and albumin) anions [91, 92]. Direct comparison of the effects of respiratory, lactic and inorganic acidoses on rabbit myocardial cells suggests differing effects at similar pH values [156]. In addition, certain toxins are also acidic, but their toxicity is not related to acidity per se, and the pathophysiological process is associated with a metabolic acidosis, e.g. salicylates.

There are potentially beneficial effects of acidosis which could make it protective during times of physiological challenge. For example the leftward shift of the oxyhaemoglobin curve and cerebral and coronary vasodilation may confer benefit. Furthermore, there are many harmful effects of alkalosis; outside the setting of resuscitation alkalosis is the most common acid-base disturbance in the critically ill. Metabolic alkalosis exerts a number of detrimental effects including hypoventilation, enzyme system changes and opposing the effects of acidosis above; metabolic alkalosis is associated with increased morbidity and mortality among the critically ill [157].

Therefore, although we exert enormous efforts in the diagnosis of metabolic acidosis it remains largely unclear which of the four hypotheses above is correct, or the relative importance of each. Multiple interventions for syndromes such as diabetic ketoacidosis are clearly required; however, the role of acidosis correction per se as distinct from treating the pathophysiological process will be considered below.

How and when should one treat metabolic acidosis?

A flowchart is provided to illustrate some of the key components in diagnosing and managing metabolic acidosis (Fig. 2).

Figure 2.

 Suggested approach to diagnosis and management of metabolic acidosis in the critically ill.

A number of points are emphasised:

  • ‘Adequate’ diagnosis. Where a specific acid has not been identified, or multiple acids exist, e.g. acute renal failure, an accurate diagnosis of a clinical syndrome must be attempted at the very least. Any investigations and laboratory work up should complement the clinical context, not vice versa.
  • • Supportive care and treat primary pathology. The importance of basic physiological stabilisation and organ support cannot be overemphasised. Most acidoses will resolve if the primary condition is identified, treated and the patient supported. This would include multi-organ resuscitation with fluid therapy, ventilation, antimicrobials and nutritional support. A more fundamental approach, while treating the underlying cause, is to remove the class of acid responsible. Respiratory acidosis may be supported by assisted ventilation and renal failure by renal replacement therapy and, most recently, hepatic failure supported by extracorporeal technologies. In practice the support of the critically ill patient involves support of multiple organs and often concurrent therapies.
  • • Review all drugs. It is essential to be mindful of prescribed and non-prescribed substances as causes of acidosis which may have a specific therapeutic option.

If elevated [H+] underlies the mechanism of worse outcomes then it should be possible to reduce [H+] and improve outcome. It is difficult to identify a target pH or [H+] associated with improved outcomes, although there is general consensus that pH > 7.20–7.25 is desirable [134, 139]. Evidence that acidosis correction improves outcome is not consistent, although this is very difficult to establish in isolation without treating the underlying condition [158]. Strategies aimed at the primary cause of acidosis, e.g. early goal-directed therapy in sepsis, generally improve outcome [44, 45].

Buffer therapy

The use of buffers in the critically ill is common and largely lacks consensus on indications and possible benefits [159]. The decision to use buffers, most commonly sodium bicarbonate, is guided by limited evidence, especially at the extremes of physiology and the possible indications have been reviewed elsewhere [134, 160, 161]. The ‘bicarbonate space’ concept suggests equivalence with 50% of the body weight in mild acidosis, but also that this space is heavily influenced by the initial [HCO3] and thus the effects of therapy influenced by metabolic/respiratory interactions [162]. Administration to ‘well’ patients in the peri-operative period corrects acidosis but with little evidence of benefit or harm [163]. Bicarbonate is administered in a variety of doses, including fixed 50 mmol [143], titrated to the base deficit multiplied by ‘bicarbonate space’ or to a plasma bicarbonate level or pH [134]. The end-point for buffer therapy is not clear but in targeting pH > 7.20–7.25 this corresponds to an approximate [HCO3 ] of 10 mmol.l−1.

Injudicious administration of (hypertonic) sodium bicarbonate may be associated with rebound alkalosis, cardiac failure, hypernatraemia and extracellular cation depletion. Whether bicarbonate therapy can precipitate intracellular acidosis is complex and related to local buffering capacities [164, 165] and of questionable significance especially if administered as an infusion [166]. Although there is little evidence of efficacy of sodium bicarbonate during metabolic (lactic) acidosis [167, 168] there are compelling basic science arguments to consider its use [169, 170]. Some workers advocate sodium bicarbonate in (prolonged) cardiac arrest [171], rhabdomyolysis and tricyclic antidepressant overdose, although it remains unclear if the latter reflects acid-base modulaton or sodium loading [172]. Although alkaline diuresis is described during certain intoxications, e.g. methanol or during rhabdomyolysis, it is a challenging and potentially hazardous technique and arguably inferior to extracorporeal elimination. The use of bicarbonate to manage hyperkalaemic emergencies has largely fallen from favour due to slow onset time and superior alternatives [173].

Alternative buffering techniques include administering organic anion salts which act as an endogenous source of bicarbonate, e.g. lactate, citrate or acetate as their sodium salts. These compounds, essentially being class 2 organic acids, require hepatic metabolism to be processed and may prove toxic in excess, especially acetate. Sodium lactate is the only agent used commonly. Pyruvate is theoretically attractive but has limited stability in solution [174]. Carbicarb (a mixture of sodium bicarbonate and carbonate) theoretically has improved buffering capacity and less carbon dioxide production, although the clinical benefits remain unproven [175]. In a hypoxic canine model carbicarb produced positive inotropic effects [176]. Tromethamine (Tris hydroxymethyl aminomethane, THAM) is a sodium-free buffer that can combine with both respiratory carbon dioxide and conventional acid solutions. Although these properties are extremely attractive, there is little clinical evidence of superiority and there are concerns regarding adverse effects (hyperkalaemia, hypoglycaemia and apnoea) [177, 178].

Extracorporeal therapy

Certain toxins or acidic metabolites thereof may be removed by extracorporeal technologies (Table 1) [179]. Appropriate substances are of low molecular weight and thus readily permeable across synthetic filters and water soluble (i.e. small volume of distribution). Limited protein binding (or at least saturable protein binding following overdose) aids elimination. The extracorporeal technique (charcoal haemoperfusion, haemodialysis filtration or dia-filtration) for optimal removal has yet to be defined, but continuous techniques have significantly lower clearance rates relative to intermittent techniques [180, 181]. Whereas haemoperfusion is a less commonly employed technique, combinations of dialysis and haemoperfusion may allow the fastest removal [182–185].

Table 1.   Acidic toxins which may be eliminated by extracorporeal technology.
  • *

    Requires albumin-based techniques.

Methanol, ethanol, ethylene and propylene glycol and other industrial alcohols
Ketones, aldehydes (paraldehyde and formaldehyde)
Carbamazepine, phenytoin, valproate
Theophylline, aminophylline
Flecainide, lidocaine
Tricyclic antidepressants (controversial)
Paraquat (controversial), amatoxins* (controversial)
Toluene (hippuric acid)

4-Methylpyrazole (fomepizole)

Previous therapies for ethylene glycol and methanol poisoning have included alkalinisation, forced diuresis, ethanol infusion and extracorporeal removal. In the UK the competitive alcohol dehydrogenase antagonist fomepizole is available for use in ethylene glycol poisoning [186] and methanol poisoning on a named patient basis [187]. This agent may still be required in conjunction with extracorporeal removal as methanol redistribution postdialysis may occur [188, 189].


Dr Charlie McAllister (Craigavon) who helped me (CM) to think beyond the textbooks and put Hartmann's into a blood gas analyser and Dr John Trinder (Dundonald) for precision in all aspects of patient care. Dr David Staples (Derby) for assistance and suggestions in preparing the manuscript.