This study shows that while the majority of postoperative patients in our population exhibit a neutral pH with pCO2 and BE in reference range, underlying metabolic acid-base disturbances are frequent when applying the Stewart approach and the albumin corrected AG. The two most prominent disturbances according to Stewart are the acidifying effect of a lowered SIDa caused by hyperchloremia and the alkalizing effect of hypoalbuminemia. The concurrent presence of these two metabolic abnormalities can balance each other out and thus result in an apparently normal acid base state as defined by the traditional criteria, however, classifying these cases as normal seems inappropriate given the underlying abnormalities. This result matches the findings of a study on metabolic disturbances in 137 apparent acid-base neutral ICU patients (Moviat et al. 2013).
In our study the incidence and severity of hypoalbuminemia is constant across the spectrum of pH, with an albumin level of 30 g/L amounting to an alkalizing effect on the BE of approximately +3 mmol/L (Story 2016). Acidifying metabolic factors oppose the alkalizing effect of hypoalbuminemia with the incidence and magnitude of these effects (low SIDa, increased SIG and corrected AG) increasing as pH shifts from alkalemia to neutral to acidemia. As witnessed in our study, patients presenting with normal pH and pCO2 can thus have either no or opposing metabolic changes, either way resulting in a normal BE.
Previous studies addressing postoperative acid-base balance and outcome have focused primarily on metabolic acidosis while defining this as a decrease in BE, with decreased BE due to hyperlactatemia associated with increased mortality (Silva et al. 2016; Rutherford et al. 1992). In our study we used LOS as outcome measure as well as the occurrence of major morbid events (e.g. AKI, unplanned ICU admission, mortality). BE and SIDa performed equally in predicting these adverse outcomes. However, changes in BE were not associated with an increase in LOS whereas a decrease in SIDa and increase in corrected AG were, even within pH neutral patients. Aside from providing insight in the mechanisms involved in acid-base balance, this provides further argument as to evaluate the independent metabolic variables defined by Stewart (i.e. SIDa, unmeasured strong anions and albumin) instead of only the dependent variables pH and BE when confronted with ABG results. Whether one uses the AG or SIG as to evaluate the presence of unmeasured strong anions seems a matter of personal preference as these parameters correlate very well (Moviat et al. 2003). Disturbances in SIDa and SIG or AG are most likely a marker of underlying illness rather than the cause of increased LOS by itself and should prompt the clinician to further investigate.
The association of hypoalbuminemia with increased LOS is not surprising as hypoalbuminemia is commonly recognized as a marker for severity of illness and is associated with an unfavorable prognosis (Gibbs et al. 1999; Herrmann et al. 1992). Among its causes are impaired synthesis due to malnutrition, chronic inflammation or critical illness and alterations in distribution due to major surgery or (hemo)dilution (Gatta et al. 2012), with the latter being a likely cause in our population as evidenced by a significantly greater amount of surgical blood loss and subsequently intravenous fluids infused in those with hypoalbuminemia compared to those without.
The observed hyperchloremia causing the low SIDa is most likely iatrogenic in nature due to the administration of 0.9% NaCl solution in the intra- and postoperative period as hyperchloremic patients received a greater amount of 0.9% NaCl, although an adaptive renal response to the alkalizing effect of hypoalbuminemia could be an additional explanation. Whether low SIDa is only a marker of associated hypoalbuminemia, thereby explaining the increase in LOS, or whether there is an additive effect on outcome is unclear. Excess chloride has been associated with increased renal vasoconstriction, decreased urine output and increased mortality and length of stay after noncardiac surgery (Reid et al. 2003; Wilkes et al. 2001; McCluskey et al. 2013). However, in meta-analysis, when comparing the use of balanced versus non-balanced intravenous solutions in the perioperative period, although the incidence of hyperchloremia was significantly higher in those receiving unbalanced solutions, no difference was found in mortality or major morbidity (Bampoe et al. 2017). Although the consequences of hyperchloremia may be subtle, it seems reasonable to avoid any iatrogenic hemostatic imbalance by reducing the use of 0.9% NaCl solution.
Some limitations of this study should be noted. The incidence of hypoalbuminemia is high in our population due to selection bias and can not be extrapolated to the complete postsurgical HDU population, since only those cases in which the attending physician chose to measure albumin were included. Furthermore, while the crucial determinants of SIDa (sodium, chloride, lactate) were measured by the same blood gas analyzer in one sample, analysis of albumin and other electrolytes were performed on a separate sample in our clinical laboratory. Due to the retrospective nature of the study the exact time difference between samples is unknown, but should be less than 10 minutes in the majority of cases, as blood sampling is part of the nursing routines in the morning. This and the fact that magnesium, calcium and phosphate were not measured and therefore imputed as normal in a minority of patients could have resulted in inaccuracies in the Stewart calculations. Due to its reliance on multiple measurements, the Stewart approach has been said to be vulnerable to inaccuracies, which could provide an alternative explanation for the low SIDa found in our study, especially in those cases in which no overt hyperchloremia or hyperlactatemia was present (Nguyen et al. 2009). However, from clinical experience we know, that calcium and magnesium in most patients show a tendency to be low-normal or below normal. This implies that the results of our calculations of SIDa in cases with missing data might be less pronounced than they would be in reality. Furthermore, magnesium and calcium are considered of minor importance in SIDa variation and are often left out or replaced with normal values in simplified Stewart-based calculations (Story 2016; Agrafiotis et al. 2018).
In order to be able to correct metabolic disturbances one first needs to identify its existence. In routine clinical care ABG results are primarily analyzed by observation of pH, BE/HCO3- and pCO2, while the Stewart approach is often regarded as overly complex and offering no advantage over this method (Emmett 2016). However, our study shows that augmenting the traditional approach with the SIDa leeds to increased identification of clinically relevant metabolic disturbances, also in patients in which all traditional parameters are within reference range. These disturbances could otherwise easily be missed. Furthermore, a simplified version of SIDa can be easily calculated at the bedside using only the information provided by the blood gas analyzer.
In conclusion, when applying the Stewart approach and albumin corrected AG metabolic disturbances hidden beneath the surface in the patient with seemingly neutral acid-base balance based on pH and BE are frequently detected, can be interpreted within the clinical context and are associated with increased LOS.