Acid–Base Analysis in the Operating Room: A Bedside Stewart Approach

Clinical acid–base analysis is defined by changes from a “normal point’’: pH, 7.40; Paco₂, 40 mmHg; bicarbonate, 24 mM; and base excess, 0 mM. However, vital to the beside Stewart model, also at this pH 7.40 point are plasma sodium, 140 mM; chloride, 105 mM; lactate, 1.0 mM; and albumin, 42 g/l, the middle of the reference ranges. The bicarbonate-centered approach to acid–base analysis has limited overlap with routinely measured electrolytes, except when estimating the anion gap, traditionally Na–Cl–HCO₃.  The Henderson-Hasselbalch equation can lead to confusion because it is used to represent both (1) the physical chemical equilibria of carbonic acid in solution and (2) physiologic regulation of bicarbonate and carbon dioxide, predominantly through renal and respiratory mechanisms.  Further, the bicarbonate-based approaches to acid–base diagnosis are largely categorical (yes or no) and require accounting for physiologic changes in carbon dioxide using rules of thumb to determine if there is physiologic respiratory compensation (table 1).  The bicarbonate-based approaches do not routinely quantify metabolic derangements and therefore do not routinely estimate the severity of metabolic disorders. While the anion gap helps determine whether there may be pathologic anions during metabolic acidosis (examples 1 and 3 in Boxes 1 and 3), this is usually categorical—raised anion gap acidosis: yes or no?

Base excess  was introduced more than half a century ago to readily separate metabolic acid–base disorders from respiratory disorders and quantifying their severity by modeling changes when blood is equilibrated with an atmosphere with a pCO₂ of 40 mmHg, effectively removing any respiratory change.  Standard base excess (SBE), also known as the base excess of extracellular fluid, is more physiologically sound than the original actual base excess which unfortunately is still often reported. Base excess represents the number of millimoles of strong base (NaOH) for negative base excess or strong acid (HCl) for positive base excess added to a liter of plasma to return pH to 7.40 where the pCO₂ in plasma is 40 mmHg at 37°C. Blood gas machines calculate base excess using the Van Slyke equation. The base excess reference range is approximately –2.5 to +2.5 mM. Although base excess was introduced long ago, many clinicians continue to prefer bicarbonate-based approaches despite the relative simplicity and clinical value of base excess in quantifying, rather than only classifying, metabolic acid–base changes.  A recent excellent review describes clinical use of base excess in detail. Base excess has simpler equations for changes of carbon dioxide physiologic compensation than bicarbonate approaches (table 1).

Peter Stewart never directly addressed base excess but opened his 1978 acid–base publication with, “Acid-base solution chemistry is an emotionally charged area of science, partly because of the frustrations we have all experienced in trying to master it.” Stewart proposed that the hydrogen ion concentration is an important physiologic variable but that is dependent on other factors, and that bicarbonate is unimportant mechanistically and is also dependent on other factors. 

As with both bicarbonate and base excess approaches, the Stewart-derived approaches analyze the partial pressure of carbon dioxide as the central component of the respiratory side of acid–base status. However, in the Stewart approach, the nonrespiratory (metabolic) factors for biologic solutions including plasma and adjoining extracellular fluid are (1) the strong ions that are completely dissociated and (2) weak acids that are only partly dissociated anions and so exist as undissociated acids, anions, and hydrogen ions in equilibrium (HA ↔ A^– + H⁺). The key controlling elements are (1) the difference between the concentrations of the strong positive cations and the strong negative anions—the strong ion difference—with acidosis when the strong ion difference is decreased and alkalosis when the strong ion difference is increased; (2) the total amount of weak acid (HA + A^–) with acidosis from increased total weak acids and alkalosis from decreased total weak acids; and (3) temperature-dependent dissociation constants for the weak acids. The key components of plasma clinical chemistry in health and disease and drivers of acid–base status are the sodium, chloride, and lactate strong ions, and the amino acid weak acids in albumin. Bicarbonate and pH are dependent on these factors and the partial pressure of carbon dioxide, with hydrolysis of water being the primary hydrogen ion source (examples 1 to 4 in Boxes 1 to 4).

The Stewart acid–base approach challenges the mainstream bicarbonate paradigm and is viewed by some as almost heretical.  Unfortunately, limitations of Stewart’s original work include underreferencing, unfamiliar ideas such as strong ion difference, and an intimidating fourth- order polynomial equation derived from six simultaneous equations that combines all the factors to describe acid–base status in a fluid compartment, which is unsuitable for easy clinical application.  In line with concepts proposed George Box, the aim of our group and others has been to develop a simplified Stewart model that, while not fully physiologically correct, is clinically useful.

Assessing primary respiratory changes or expected respiratory compensation is applied in similar ways across bicarbonate-centered, base excess, and Stewart approaches (table 1, examples 1 to 4 in Boxes 1 to 4). Early approaches to simplifying the Stewart acid–base approach combined base excess as an overall metabolic acid–base metric with changes in sodium and chloride as the principal plasma strong ions and albumin as the principal plasma nonvolatile weak acid. They examined sodium and chloride effects on base excess in two separate equations and referred to changes in free water indicated by changes in sodium concentration and chloride corrected for measured sodium relative to 140 mM. This has problems: (1) the concept of free water excess or deficit is not particularly useful or familiar in clinical anesthesia (2) the importance of the difference between sodium and chloride concentrations is hidden; (3) this approach requires two calculations that most would require a calculator to perform; (4) even recent versions do not asses the effect of lactate if measured and (5) the calculated acid–base effect of changes in albumin using the formula [albumin] × (0.123 × pH – 0.631) which is unnecessarily complex for clinical use and can be simplified to 0.25 × [42 – measured albumin, g/l] for a reasonable approximation. 

Residual base excess effects are attributed to other ions, usually anions.  The anion gap when corrected for albumin is a closely related chemical construct (examples 1 and 3 in Boxes 1 and 3). Other ions are common among critically ill and high-risk surgical patients. In the past, these other ions included lactate.  With advances in point-of-care blood gas analyzers, plasma lactate measurements are now often routinely available, along with sodium and chloride. There is good evidence that increasing plasma lactate concentration is associated with increasing mortality among patients admitted to ICU, including those admitted from the operating room.  Quantitatively assessing extent of any lactic acidosis from venous or arterial blood gasses should be routine for perioperative patients.  Beyond lactate, the effects and importance of other ions are an area of scientific uncertainty, with some evidence for associations with mortality. 

Our group started with the previous approaches combining base excess and elements of the Stewart approach to create a simplified Stewart approach to analyzing metabolic acid–base changes. Our aim was to create a simple pragmatic bedside approach particularly for managing high-risk or critically ill perioperative patients, and those receiving large volumes of intravenous fluid therapy (examples 1 and 3 in Boxes 1 and 3). Like others, we used SBE  as the overall metric for metabolic acid–base status and then analyzed the quantitative effects on base excess of important routinely measured independent elements controlling the metabolic (nonrespiratory) acid–base status in plasma: the strong ion difference (sodium−chloride–lactate) and total weak acids (albumin).  The residual base excess effects from other ions included ions that could be measured, such as ketones or phosphate, or other strong ions or weak acids ions not routinely measured in clinical chemistry. The associated assumed set points for the strong ions and weak acids are commonly used medians of the reference range for each of these elements (sodium, chloride, lactate, and albumin) but may vary with analyzers and populations.

Therefore, the acid–base starting “normal point” is as follows: pH 7.40, pCO₂ 40 mmHg, bicarbonate 24 mM, base excess 0 mM, sodium 140 mM, chloride 105 mM, lactate 1.0 mM, and albumin 42 g/l. Acid–base analysis centers on changes in these variables from this ideal point.

In summary, the most recent version of our approach is

This assumes that lactate is measured; otherwise, the older version can be used.

After we updated our simplified Stewart approach to quantitively analyzing acid–base disorders to routinely include lactate and changes in chloride reference ranges,^20,21  John Friesen, M.D., from University of Manitoba, Winnipeg, Canada, reported³¹ turning this simplified Stewart approach into a web application (https://www.abgst.altervista.org/) to further aid bedside use of this approach. The help section of the web application also references important acid–base equations.

As a further contribution to a simple bedside approach, an anesthesiologist from India (Anitha Nileshwar, M.B.B.S., M.D., Department of Anesthesiology, Kasturba Medical College, Manipal, India, personal communication, November 2020, email) devised the mnemonic SALT for the elements of the bedside Stewart approach:

S = Sodium − chloride base excess effect

A = Albumin base excess effect

L = Lactate base excess effect

T = Trash ions (oTher ions) base excess effect

Therefore, base excess = S + A + L + T, an easy-to- remember summary of plasma metabolic acid–base status.

Lactate and other ions fill a widened anion gap and acid–base changes related to these ions are associated with worse outcome in critically ill patients than similar acid–base changes associated with sodium−chloride and albumin. When we first described our version of a simplified Stewart approach lactate was not a routine clinical measurement. In the subsequent update, we modified the simplified Stewart approach to include lactate.  However, it is likely that there is a large array of other ions (examples 1, 2, and 4 in Boxes 1, 2, and 4) from organ dysfunction, cellular injury, disturbed metabolism, and exogenous poisoning. A new source of other ions is euglycemic ketoacidosis associated with patients with diabetes taking sodium glucose on cotransporter 2 inhibitors (SGLTIs) and is an increasingly important differential diagnosis now that these drugs have indications beyond diabetes care. 

Albumin lies at an intersection of the Stewart and bicarbonate approaches. Many use the anion gap (Na–Cl–HCO₃) to qualitatively detect other anions during metabolic acidosis.  Like the Stewart approach, the anion gap relies on the principle of plasma electroneutrality: sum of cations = sum of anions. For the last 30 yr, many have recommended correcting the anion gap for decreased albumin when analyzing metabolic acidosis (examples 1 and 3 in Boxes 1 and 3).

In the presence of hypoalbuminemia, the correction calculation for the anion gap is the same as estimating for the albumin effect on base excess: 0.25 × (42 – albumin, g/l).  A less apparent point is that due to effects on both base excess and the anion gap, hypoalbuminemia can mask acidosis  from causes including relative hyperchloremia, increased lactate, and other ions (example 2 in Box 2). For any patient with hypoalbuminemia, the severity of acidifying strong ion and weak acids changes is masked. 

A decrease in preoperative plasma albumin concentration to 32 g/l will increase base excess and decrease the anion gap by 2.5 mM, and severe hypoalbuminemia (22 g/l) will have a 5-mM change in base excess and anion gap. Ideally plasma albumin should be measured before major surgery and for high-risk patients, and repeated after large volumes of intravenous fluids or blood products (examples 1 and 3 in Boxes 1 and 3). For anesthesiologists, knowing preoperative plasma albumin concentration has added value in assessing patient risk due to the strong association between hypoalbuminemia (less than 35 g/l) and postoperative complications and mortality due to both malnutrition and chronic disease.

Unfortunately, point-of-care blood gas machines do not routinely have albumin assays. All approaches to assessing metabolic acidosis status ideally need measured plasma albumin concentration.  Without a known plasma albumin concentration, it is possible to still calculate the sodium−chloride and lactate effects on base excess, which provide important information. This also allows calculating other ions but without correcting for albumin effects. In sicker patients, preoperative albumin concentrations are likely to be close to or less than 35 g/l. In two large Australian and New Zealand studies, we found the median albumin concentrations for patients undergoing major surgery were 36 g/l for American Society of Anesthesiologists (Schaumburg, Illinois) Physical Status III patients and 32 g/l for American Society of Anesthesiologists Physical Status IV patients. These translate to base excess effects of +1.5 and +2.5 mM, respectively. In the absence of intraoperative albumin therapy, plasma albumin concentrations after major surgery may be lower still, around 30 g/l  a base excess effect of +3 mM. Therefore, in the absence of a known albumin concentration, the other-ion effect on base excess will often be between 1 to 4 mM more negative than calculated if not corrected for changes in albumin (examples 1, 2, and 4 in Boxes 1, 2, and 4). In contrast, therapy with 5% albumin can cause acidosis aggravated by chloride in carrier solutions (example 3 in Box 3).

The Stewart approach dramatically simplifies understanding intravenous fluid therapy.  A Stewart explanation for acidosis after saline therapy is that as saline infusion continues, plasma chloride concentration increases faster than sodium concentration, decreasing the sodium−chloride difference, the primary component of the plasma strong ion difference, and is acidifying.  The key point is the relative difference between the sodium and chloride concentrations. A patient can have hyponatremia but chloride in the reference range, a relative hyperchloremic metabolic acidosis. Infused lactated Ringer’s solution (Hartmann’s solution) produces less metabolic acidosis than similar volumes of saline because when lactate strong anions are removed from plasma, the plasma strong ion difference is increased, which is alkalinizing. This is a much simpler explanation than complex descriptions of hepatic bicarbonate production.  Solutions such a PlasmaLyte (Baxter Healthcare, Australia)  also have rapidly cleared acetate and gluconate strong anions, again widening the strong ion difference. An alternative explanation for postinfusion acidosis uses bicarbonate dilution as the mechanism but is limited in quantifying effects.

Using Stewart principles, sodium bicarbonate is sodium ions with carbon dioxide.  Sodium bicarbonate therapy is alkalinizing because it is an infusion of sodium strong cations without a strong anion. Using the concentrated (1,000 mM, 8.4%) preparations will predominantly increase plasma sodium relative to chloride and increase the strong ion difference, which is alkalizing. Conversely, the more dilute 1.26% (150 mM) solutions available in some countries will require 6.7 times the volume for the same amount of sodium bicarbonate and will largely act by decreasing plasma chloride concentration rather than increasing sodium again increasing the strong ion difference (example 3 in Box 3).

Bedside Stewart does not definitively describe metabolic acid–base changes, but is meant to be a useful tool in clinical practice in the operating room and other critical care settings. The main points in using Bedside Stewart in anesthesia care are as follows: (1) the partial pressure of carbon dioxide remains the key respiratory metric; (2) use base excess; (3) consider the sodium−chloride difference; (4) adjust for albumin; (5) measure lactate whenever possible and include in assessing risk; and (6) look for trash (other) ions. Bedside Stewart allows us to quantitatively assess overall metabolic acid–base status and have quantitative insight into the underlying causes for changes. Nileshwar’s SALT mnemonic and Friesen’s web application can enhance using this approach. I hope that readers will try this bedside approach and compare it to their current practice for patients requiring arterial or venous blood gas analysis. Give it a test drive.