Background
When long-term balance studies in humans demonstrated that sodium could accumulate without weight gain or hypernatremia, this challenged the generally accepted model on sodium homeostasis [
1]. This model states that changes in sodium homeostasis can primarily be explained by a two-compartment model with an intracellular (ICV) and extracellular volume (ECV), where key ions are completely dissolved—i.e., osmotically active. An extra compartment that stores sodium nonosmotically without causing an expansion of the ECV has been proposed by Titze et al. [
2]. In both animal and human studies, they found that sodium is stored nonosmotically in the skin [
2,
3]. Nonosmotic sodium storage is presumably facilitated by large strongly negatively charged polymers such as glycosaminoglycans [
4,
5]. The accumulation of chloride in the skin has been suggested in animal models [
6,
7], but has not been as extensively studied as sodium storage.
Patients admitted to the intensive care unit (ICU) typically receive large amounts of sodium and chloride during their ICU treatment [
8]. Both hypernatremia and hyperchloremia are a frequent complication in critically ill patients and are associated with adverse outcome [
9‐
12]. The infusion of high amounts of chloride is also recognized as cause of hyperchloremic acidosis [
11,
12]. Improved understanding of sodium chloride homeostasis in this patient group is therefore of utmost importance. To our knowledge, no studies have tried to measure missing sodium as evidence of stored sodium in ICU patients. Likewise, a potentially similar phenomenon for chloride has not been studied yet.
The objective of our study was therefore to estimate sodium and chloride that might ‘disappear’ in balance studies in ICU patients. Since random and systematic errors as well as different assumptions on the size of the ECV and perspiration strongly affect the calculated sodium or chloride deficit, five scenarios were tested in which the assumed sizes of the ECV or perspiration were varied.
Discussion
This is the first balance study that aimed to estimate missing extracellular sodium (MES) in ICU patients and missing extracellular chloride (MEC) in any patient group. Although we found considerable variations in estimated MES and MEC according to the various scenarios, the results suggest a considerable MES and a somewhat lower MEC (Fig.
3).
To calculate MES and MEC, we used one default and four more extreme scenarios, which we believe cover the scope of published sizes of the ECV and rates of perspiration. An ECV of 20% of body weight is a conservative choice in patients arriving at the ICU after major surgery [
16], while 60% is an extreme estimate [
17]. Regarding perspiration, defining the extremes was more difficult, but nearly all sources assume a perspiration ≥ 400 ml/day for both the skin and the respiratory tract without fever [
14]. Our estimate of 5 mL/kg/day probably is thus the lower limit, while 20 mL/kg/day is a large estimate. We believe that the true value of both MES and MEC should be somewhere in between the four more extreme scenarios as depicted in Fig.
3. MES and MEC were mainly influenced by perspiration and MEC also somewhat by the ECV. This underscores that both the size of the ECV and insensible perspiration are important determinants in the estimated size of MES and MEC. In sex-specific models (Additional file
1: Fig. S1), males showed slightly higher MES and MEC compared with females.
More sodium and chloride disappeared from the balances during the first two days of ICU admission than in the subsequent days (Fig.
1). Resuscitation fluids, often high in sodium and chloride content, are frequently administered during surgery and in the early postoperative period. Whereas the recommended limits for dietary sodium intake are 2.3 g/day [
18], our patients received an average of 8.3 g (i.e., 360 mmol) sodium per day, with positive sodium balances but stable sodium concentrations. This resulted in a MES of 296 mmol after four ICU days. When this MES is expressed in terms of NaCl 0.9% infusion, 1.9 L of this fluids sodium went missing in our patients.
Nonosmotic sodium storage has been studied in several non-critically ill patient groups. In healthy individuals, it has recently been observed that half of an acute intravenous hypertonic saline load of 201 mmol appears to be briefly stored nonosmotically [
19], possibly in interaction with the endothelial glycocalyx. Sodium storage has been reported to increase with advancing age, to be greater in men and patients with hypertension, hyperaldosteronism, end-stage kidney disease and infection [
20‐
22]. Tissue sodium levels are variable and may be altered by dialysis and diuretic treatment [
23,
24]. However, the precise clinical significance of nonosmotic sodium storage has not been defined yet. The existence of nonosmotic sodium storage has not been examined in critically ill patients. Nonosmotic sodium storage could also be relevant in ICU-acquired hypernatremia (IAH) [
25] and could explain the relatively long duration of IAH once it develops, although sodium balances were not performed in this study. It is believed that the electrical binding capacity of various tissues for sodium is altered during inflammation [
26], which may interact with the development of IAH in critically ill patients. Irrespective of a potential relation between IAH and sodium storage, a strategy in which infusion fluids with lower sodium chloride content are used to reduce IAH is probably desirable [
27]. We reported earlier [
10] that changes in bulk intravenous fluid constitution paralleled changes in the incidence of ICU-acquired hypernatremia. Recently, it was elegantly shown that maintenance fluid therapy constitutes a higher sodium, chloride and water burden than acute resuscitation fluid administration [
28].
With regard to chloride, which also disappeared in our balance calculations, both sodium and chloride storage may affect changes in blood pressure [
6,
29].
As we cannot explain MES and MEC by the conventional two-compartment model where sodium is extracellular and potassium intracellular, a specific storage compartment may be the buffer of these sodium and chloride loads. The key alternative to nonosmotic storage is loss of sodium and chloride to the ICV. This effect has been demonstrated by healthy persons who sustained muscular injury [
30]. Critical illness is often accompanied with critical illness myopathy, and loss of sodium and chloride to the ICV might then also be conceivable [
31]. As we reported in an earlier study [
8], our patients displayed a
negative potassium balance of 101 mmol, which is another argument for possible intracellular uptake of sodium in exchange for potassium release. Moreover, we also observed a negative electrolyte-free water (EFW) balance in these patients. Together, this suggests that no ICV expansion occurred [
8]. Therefore, we assumed that all fluids administered (including EFW) remained in the ECV. However, if part of the EFW would enter the ICV, this would result in lower increases and thus even higher MES and MEC estimates.
The presence of nonosmotic storage could be verified through direct tissue analysis or via specialized MRI [
2,
3,
20]. Sodium changes in the tissues of ICU patients resulting from MES could be imaged via
23Na MRI. To our knowledge,
35Cl MRI has not yet been used to study MEC, but it is a promising and intriguing technique to identify the anatomical spaces where salt is stored [
32,
33]. Importantly, this technique should be able to differentiate between the two main explanations for missing sodium and chloride: nonosmotic storage or intracellular uptake.
Our study has a number of limitations. Due to its retrospective design, we could not control for many variations in standard care. We had to make several assumptions, as, for example, for the insensible perspiration or the size of the ECV. However, we believe that the extreme scenarios on perspiration and ECV in our sensitivity analyses covered all realistic scenarios. We did not account for fecal losses, as we could not retrieve this information. Since we observed early postoperative patients, fecal production was absent or very low and, moreover, loss of sodium and chloride through the gut is usually very limited [
34]. We did not measure weight changes as this is not routine procedure at our unit. Daily weight measurements could be added in the future to further validate our results.
On the relatively short term,
differences in body weight measured in kg as measured in ICU patients will be less accurate than fluid balances measured in mL. Therefore, we only used initial recorded weight to estimate the ECV. Fluid balances in critically ill patients often have a poor correlation with changes in body weight [
35,
36]. Especially cumulative fluid balance is prone to errors, as measuring errors get cumulated [
36], which we accounted for in our error estimates. It must be noted that body weight measurement also has multiple possible errors, which could be the explanation of the lack of association between fluid balance and differences in body weight [
36]. However, we believe that due to the short time this study covers and the meticulous recalculation of the fluid balance, including gastric retention, drain fluids and insensible perspiration, we have minimized errors as far as realistically possible.
Insensible perspiration remains very challenging to measure. As MES and MEC were most influenced by insensible perspiration, the lack of direct measurement of perspiration is an important limitation of our study. We tried, however, to maximize the chance to include the true value as much as possible with our five different scenarios. Direct measurement of (in)sensible perspiration would make estimated of MES and MEC more accurate (Fig.
3). Unfortunately, we are not aware of reliable tools to measure (in)sensible perspiration.
In this first observational balance study, we selected our patients based on complete balance data, which could have induced selection bias. The Androque-Madias [
37] and Nguyen-Kurtz [
38] formulas are frequently used when estimating the plasma sodium level after a saline infusion in dysnatremic ICU patients [
17,
19]. However, we choose not to use these formulas in our study, as they do not account for excretion of sodium or chloride or they use empirically derived constants which were not suitable for using in our model. However, predictions on the size of the ECV from both formulas fall within the four scenarios.
In conclusion, our detailed sodium and chloride balances in ICU patients after cardiothoracic surgery suggest a loss of osmotically active sodium and chloride from the ECV. The estimates depend considerably on the scenarios used. Whether these ions are nonosmotically stored or transferred to the intracellular space needs further study.