Formulae to correct sodium concentrations for serum water fraction in cases of hypo‐ and hyperproteinemia in dogs

Abstract

Background

Biochemistry analyzers, which utilize indirect potentiometry, are used to determine serum electrolyte concentrations in dogs. Artifactual increases or decreases in these electrolyte concentrations can be caused by alterations in the serum water fraction (SWF). Severe hypo‐ and hyperproteinemia cause changes in SWF, which can then result in incorrectly reported serum sodium concentrations.

Objectives

The goals of this study were to determine an average actual SWF (SWF_ACTUAL) in dogs and establish formulae to correct serum sodium concentration measured by indirect potentiometry in hypo‐ and hyperproteinemic patients.

Methods

Serum samples from 115 canine patients were analyzed for electrolytes measured by both indirect and direct potentiometry. Total protein, albumin, triglycerides, and cholesterol were also determined. Each serum sample was then lyophilized to determine SWF_ACTUAL. A canine‐specific formula to estimate SWF (SWF_EST‐CAN) was developed using a multivariable linear model and compared with the human‐estimated formula (SWF_EST‐HUM).

Results

The mean SWF_ACTUAL in this population of dogs was 92.7%, which was significantly different (p < .0001) than the mean (95.1%) calculated using SWF_EST‐HUM. The formula derived from SWF_EST‐CAN recapitulated SWF_ACTUAL more accurately than SWF_EST‐HUM. Based on a slope closer to 1.0, the corrected sodium concentrations calculated using the canine formula correlated marginally better with the serum sodium measured by direct potentiometry than those calculated using the human formula.

Conclusions

Applications of correction formulae are expected to limit the misinterpretation of electrolyte data from indirect potentiometry when altered SWF occurs. A case example of the potential utility of these correction formulae is presented.

1. INTRODUCTION

Veterinary clinicians commonly run serum and plasma electrolyte concentration measurements as part of their diagnostic workup, which is accomplished through either direct or indirect ion‐selective electrode (ISE) potentiometry. Electrolytes, such as ionized sodium, are restricted to the water phase of serum or plasma; thus, the reported sodium concentration is the activity of sodium ions in the water phase. Analyzers in many high‐throughput laboratories use indirect potentiometry ISE, which involves a pre‐analytical dilution step that assumes a constant serum water fraction based on human data (SWF_CON‐HUM) of 93%. ¹ This provides acceptable data for many samples because ionization remains constant under conditions of infinite dilution. ² , ³ However, artifactual fluctuations in electrolyte concentrations can occur under certain circumstances that affect the sample's SWF; these scenarios can then render the correction formula used by the analyzer inaccurate. All electrolytes measured using indirect potentiometry, including sodium, potassium, and chloride, are subject to the same effects. Given the higher concentration of sodium in serum, the relative effect on sodium is greater than that for the other two electrolytes. ³ Of note, direct potentiometry, which is used in blood gas analyzers, does not involve a dilution step and avoids problems inherent in assuming a constant water fraction.

Hyperlipidemic or hyperproteinemic samples are well‐recognized causes of artifactually low serum sodium concentrations in samples measured by indirect potentiometry. ⁴ As mentioned above, the concentration of diluted serum is determined assuming the original sample consists of approximately 93% water. However, excess lipid or protein displaces and, thus, reduces the percent water volume of the serum containing ionized electrolytes. This decrease in SWF may then result in falsely low electrolyte measurements. Severe hypoproteinemia has an opposite effect on measurement of sodium and other electrolytes in that the percent water volume of the sample is increased. As such, instrument calculation based on the assumption of only 93% water volume results in a spuriously high sodium concentration. ⁵ Hypoproteinemic concentrations for which erroneous concentrations can be anticipated, particularly for sodium, are not established in dogs.

There is, therefore, a need to interpret clinical laboratory data with knowledge of the analytical errors that can arise under specific circumstances, such as hypo‐ or hyperproteinemia, to avoid inappropriate initiation of therapy. Our group previously described formulae for calculating corrected sodium concentrations for hypo‐ and hyperproteinemic cats, ⁶ and we aimed to establish similar formulae for dogs in the current study. This process involved: (1) establishing a constant SWF applicable to dogs (SWF_CON‐CAN) by measuring the average actual SWF (SWF_ACTUAL) in canine serum samples; (2) developing a canine‐specific estimated SWF (SWF_EST‐CAN) formula; and (3) using SWF_EST‐CAN to create an updated correction formula to modify falsely reported serum sodium concentrations. Based in part on our finding for cats, our hypothesis was that high or low protein concentrations in dogs would result in falsely reported electrolyte concentrations and that our novel canine formulae would be able to amend these errors more accurately than existing human‐based correction formulae.

2. MATERIALS AND METHODS

Canine patients seen at The Ohio State University's Veterinary Medicine Center (VMC) during the period of May 1, 2011, to September 1, 2012, were selected to be included in the study. As owners for individual patients had consented to “best practice” medical evaluation and/or treatment at this facility, separate approval for the use of live animals by the Institutional Animal Care and Use Committee or other overseeing university/hospital organization was not required at the time these data were collected. The primary inclusion criterion was that the owner had consented to routine serum biochemistry analysis. Measurands included serum electrolytes by indirect potentiometry, total protein, albumin, globulin, triglycerides, and cholesterol using routine methods on the Roche Cobas 6000 chemistry analyzer (Roche, Basel, Switzerland) by the VMC Clinical Laboratories. Upon identification and inclusion of the patient into the study, serum samples were also analyzed using direct potentiometry with a Nova 16 Electrolyte/Chemistry Analyzer (Nova Biomedical, Waltham, MA) within 1 h. Serum samples were assigned into one of three general categories based on previously established reference intervals at the VMC (Table 1): hypoproteinemic, normoproteinemic, and hyperproteinemic samples.

TABLE 1.

Selected serum measurands for a canine patient example and corresponding canine reference intervals for the Cobas 6000 using indirect potentiometry and the Nova 16 using direct potentiometry.

Measurand (unit)	Cobas 6000		Nova 16
	Patient	Reference interval	Patient	Reference interval
Sodium (mEq/L)	149.0	143–153	151.0	143–150
Potassium (mEq/L)	4.58	4.2–5.4	4.89	3.5–4.8
Chloride (mEq/L)	105.9	109–120	113.7	111–119
Bicarbonate (mmol/L)	21.8	16–25	19.4	16.9–24.2
Anion Gap (mEq/L)	26.0	15–25	22.8	N/A
Total Protein (g/dL)	8.4	5.1–7.1	N/A	N/A
Albumin (g/dL)	4.2	2.9–4.2	N/A	N/A
Globulin (g/dL)	4.2	2.2–2.9	N/A	N/A
Triglycerides (mg/dL)	256	25–75	N/A	N/A
Cholesterol (mg/dL)	607	80–315	N/A	N/A

Note: Values in red font are above the reference interval; values in blue font are below the reference interval.

Abbreviation: N/A, not applicable.

The remaining serum sample was capped and refrigerated at 4°C for subsequent batched lyophilization. To lyophilize the samples, 100 μL of serum was transferred into a standard 500 μL microcentrifuge tube and placed into the Virtis FreezeMobile 25EL freeze dryer (SP Scientific, Warminister, PA) within a supporting rack for 48 h. Pre‐ and post‐lyophilization weight (in grams) of the sample were recorded to determine the SWF_ACTUAL. Samples remained in the freeze dryer for an additional 24 h (72 h, total) and were re‐weighed to ensure complete removal of water. No differences were seen between the 48‐ and 72‐h time points.

Estimated SWF based on human data (SWF_EST‐HUM) was calculated using the previously established formula for human serum. ⁷

$$\begin{array}{c}{\mathrm{SWF}}_{\mathrm{EST}-\mathrm{HUM}}=[99.1-\left(0.001\times \text{triglyceride concentration in}\mathrm{mg}/\mathrm{dL}\right)\hfill \\ -\left(0.7\times \text{protein concentration in}\mathrm{g}/\mathrm{dL}\right)]/100\hfill \end{array}$$

A canine‐specific formula to estimate SWF (SWF_EST‐CAN) was then developed with a multivariable linear model using forward model building. Univariable models for continuous variables were tested, which included all biochemical measurands and nominal variables representing proteinemia category. Multiple pairwise comparisons of mean differences among categories of proteinemia were accomplished using the Tukey–Kramer method. Forward, backward, and step‐wise minimum A1Cc linear regression was performed using JMP Software (JMP Pro, version 13: SAS Institute Inc., Cary, NC) with SWF_ACTUAL and indirect potentiometry concentrations. Reported formulas from all methods were the same. For all analyses, a P‐value ≤.05 was considered significant. As SWF_EST‐HUM includes triglycerides, these same methods were also used to force this continuous variable into a model; the resulting canine‐specific formula to estimate SWF using triglycerides was designated (SWF_{EST‐CAN‐TRIG}).

Statistics were performed in GraphPad Prism (GraphPad, La Jolla, CA). As all data passed the D'Agostino and Pearson omnibus normality test, a one‐way ANOVA with Tukey's multiple comparison post‐test was performed to identify statistically significant differences among SWF_ACTUAL, SWF_EST‐CAN, SWF_{EST‐CAN‐TRIG}, and SWF_EST‐HUM groups. Linear regression analysis was used to determine whether there was a significant correlation between serum total protein and the difference between electrolyte concentrations measured by indirect versus direct potentiometry. Linear regression and bias plots (Bland–Altman and Clinical and Laboratory Standards Institute, CLSI) were used to compare corrected sodium concentration measured by indirect potentiometry (using various formulae) to the sodium concentration measured by direct potentiometry.

3. RESULTS

One hundred and fifteen canine patients were enrolled in the study; they were categorized as either hypoprotenemic (n = 49), normoproteinemic (n = 47), or hyperproteinemic (n = 19) (Figure 1). The difference between the serum electrolyte concentration obtained by indirect and direct potentiometry was plotted against the serum total protein concentrations for sodium, potassium, and chloride (Figure 2A–C). The difference in sodium concentration between the two methods was most substantial at total protein concentrations below 4 g/dL. Linear regression analysis for sodium and chloride demonstrated that the slope of the best‐fit line was significantly non‐zero (P < .0001), while the slope of the best‐fit line for potassium differed from zero at P = .0116. These findings highlighted the need for electrolyte correction at these extreme serum protein concentrations, particularly for sodium.

FIGURE 1.

Serum from 115 canine patients. (A) The total protein from these samples was stratified as hypo‐, normo‐, or hyperproteinemic based on established reference intervals. Horizontal black lines represent the mean. Dotted lines indicate the reference interval for total protein. (B) Actual (SWF_ACTUAL) and estimated (SWF_EST‐HUM) serum water fraction was determined for each sample and found to be significantly different (P < .0001) based on a nonparametric paired t‐test. The SWF_EST‐CAN (calculated from a formula derived from SWF_ACTUAL in this population of dogs), and the formula with forced triglycerides (SWF_{EST‐CAN‐TRIG}) are also displayed. Horizontal black lines represent the mean. Horizontal blue lines represent groups that were significantly different from each other (***P < .0001).

FIGURE 2.

Difference between electrolyte concentrations measured by indirect and direct potentiometry over a range of total protein concentrations for sodium (A), potassium (B), and chloride (C) in canine patients. The line of best fit and slope ± standard error are shown in red. Using linear regression analysis, the slope was significantly non‐zero for sodium (P < .0001), potassium (P = .0116), and chloride (P < .0001), demonstrating proportional bias.

The mean SWF_ACTUAL following sample lyophilization was found to be 92.7% (94.1%, 92.4%, and 89.9% for hypo‐, normo‐, and hyperproteinemic samples, respectively; Figure 1) and was then utilized as the constant SWF for canine patients (SWF_CON‐CAN). The mean SWF_EST‐HUM based on the human formula was 95.1% (Figure 1B). Forward step‐wise minimum A1Cc regression analyses (R ² = .66) resulted in the following canine‐specific SWF_EST‐CAN formula, which had a mean of 92.7%:

$$\begin{array}{c}{\mathrm{SWF}}_{\mathrm{EST}-\mathrm{CAN}}=[98.3-\left(0.002\times \text{cholesterol in}\mathrm{mg}/\mathrm{dL}\right)\hfill \\ -\left(0.91\times \text{total protein in}\mathrm{g}/\mathrm{dL}\right)]/100\hfill \end{array}$$

When triglycerides were forced into the SWF_{EST‐CAN‐TRIG} model, there was no overall statistical benefit and cholesterol became insignificant. Of note, this new formula was very similar to that for SWF_EST‐HUM and had a mean of 95.0%.

$$\begin{array}{c}{\mathrm{SWF}}_{\mathrm{EST}-\mathrm{CAN}-\text{TRIG}}=[99-\left(0.001\times \text{triglycerides in}\mathrm{mg}/\mathrm{dL}\right)\hfill \\ -\left(0.7\times \text{total protein in}\mathrm{g}/\mathrm{dL}\right)]/100\hfill \end{array}$$

These new formulae were plotted against the measured SWF_ACTUAL and data generated by the human formula, SWF_EST‐HUM (Figure 3). SWF_EST‐CAN was not statistically different from SWF_ACTUAL, which was expected since the SWF_ACTUAL values were used to generate the canine‐specific formula. Unsurprisingly, given the similarities of the formulae, SWF_EST‐HUM and SWF_{EST‐CAN‐TRIG} were also not significantly different from one another. In contrast, SWF_ACTUAL and SWF_EST‐CAN were each significantly different (P  < .0001) from both SWF_EST‐HUM and SWF_{EST‐CAN‐TRIG} (Figure 1B).

FIGURE 3.

Data from a canine‐specific formula developed using forward model building. Serum water fraction estimated using the canine‐specific formula (SWF_EST‐CAN, A) or the human formula (SWF_EST‐HUM, B) are plotted against the actual (measured) serum water fraction (SWF_ACTUAL) for 115 canine samples.

The corrected serum sodium concentration using the canine formula with the SWF_CON‐CAN = 92.7% (Figure 4A–C) and the human formula with the SWF_CON‐HUM = 93% (Figure 4D–F) were compared with the serum sodium measured by direct potentiometry using regression analysis, Bland–Altman bias plots, and CLSI bias plots. The canine method produced results closer to the sodium concentration as measured by direct potentiometry, with the slope for the canine formula closer to 1.0; however, the R ² values for the canine and human sodium correction formulae were similar (.83 and .84, respectively). The standard deviation of both bias plots included zero using both methods (canine and human formulae), suggesting that they are clinically equivalent.

FIGURE 4.

Corrected sodium concentration is compared with the sodium concentration measured by direct potentiometry using the canine formula and the canine SWF_CON‐CAN = 92.7% (A–C), the human formula and the human SWF_CON‐HUM = 93% (D–F), or the canine formula with forced triglycerides and the canine SWF_CON‐CAN (G–I) via Passing–Bablock regression analysis (A, D, G), Bland–Altman bias plots (B, E, H), and CLSI bias plots (C, F, I). The line of best fit, slope, and goodness of fit statistic (R ²) for linear regression are shown in red for A, D, and G; lines of equality are shown in gray dashes. The mean and standard deviation of the bias are show in red and the mean ± 2 times the standard deviation are plotted in black dotted lines for B, C, E, F, H, and I.

The corrected serum sodium concentration, using the canine formula with forced triglycerides (SWF_{EST‐CAN‐TRIG}) and the SWF_CON‐CAN = 92.7%, was compared with serum sodium measured by direct potentiometry (Figure 4G–I). Interestingly, the formula with forced triglycerides is nearly identical to the human formula and, therefore, has a slope and R ² close to the human sodium correction formulae (.93 and .84, respectively). Again, the standard deviation of both types of bias plots included zero.

4. DISCUSSION

While indirect potentiometry may be the only technique available to some general practitioners to measure serum sodium concentration, it is worthwhile emphasizing that this method is susceptible to analytical errors related to SWF. The following correction formula is available based on human data. ⁷

$$\text{Corrected}\left[\mathrm{Na}\right]=\left(\text{indirect}\left[\mathrm{Na}\right]\times 0.93\right)/{\mathrm{SWF}}_{\text{ACTUAL}}$$

This formula assumes a constant serum water fraction for human patients (SWF_CON‐HUM) of 93%, which is programmed into the calculations of electrolytes by commercial chemistry analyzers as part of the instrument's software package. In the canine patients collected at the VMC, the mean SWF_ACTUAL was found to be 92.7%. Just as with the feline study, this SWF_CON‐CAN (based on SWF_ACTUAL) characterizes a mix of clinically healthy and ill animals; however, it is likely still representative of the target population for the application of these formulae. ⁶

Using a species‐specific formula to estimate SWF (SWF_EST‐CAN) is more practical for a clinical setting rather than measuring SWF_ACTUAL in each patient. As such, this study evaluated the equation for the estimated SWF previously established for human patients (SWF_EST‐HUM). ⁷ It was found that the mean canine SWF_ACTUAL (determined by gravimetric methods) and mean SWF_EST‐HUM (determined by the human formula) were significantly different at 92.7% and 95.1%, respectively.

Unlike the human formula, the canine formula obtained by regression analysis does not include triglycerides. This result might be considered somewhat unexpected given the known effect that triglycerides as large, hydrophobic molecules have on displacing water in serum. ⁴ However, similar to our findings for cats, triglycerides added no statistical benefit to our model, even when forced into the regression analysis. Interestingly, when triglycerides were forced into the model, the resulting formula was nearly identical to the human formula. One possible explanation is that disorders of increased serum triglyceride concentrations are comparatively rare in dogs compared with humans. Furthermore, it is common practice in our laboratory for personnel to clarify grossly lipemic serum via high‐speed centrifugation prior to running it on the biochemistry analyzer in order to avoid artifacts caused by lipemia. Technicians in our laboratory estimate that this occurs in approximately 1%–2% of all biochemistry samples, and, indeed, this occurred in 2 out of the 115 samples in our study. Importantly, the same (occasionally clarified) serum samples were used for indirect potentiometry, direct potentiometry, and lyophilization in this study, so there is less concern that excess triglycerides were affecting the measured SWF_ACTUAL but not the other values measured. Finally, the difference might also relate to the differences between dogs and humans in lipid subsets (HDL, LDL, and VLDL). Clinically healthy dogs tend to have lower TG (OSU RI: 25–75 mg/dL), while humans can normally be up to 150 mg/dL. Conversely, dogs tend to have higher cholesterols (OSU RI: 80–315 mg/dL), while humans usually have <200 mg/dL. Therefore, under normolipemic conditions, trigylcerides might contribute less in dogs, while cholesterol might contribute more to the lipid space. The absence of triglycerides in the canine formula is of benefit, given that cholesterol is included in most standard veterinary biochemistry panels, while triglycerides are not.

One surprising finding was that the goodness of fit statistic (R ²) for corrected sodium was .84 (Figure 4D) for the human‐based calculations versus .83 for the canine formulae (Figure 4A), even though the SWF_EST‐CAN resulted in values closer to the SWF_ACTUAL. In addition, the Bland–Altman and CLSI bias plots demonstrate similar bias and standard deviation across these methods. Of note, the canine formulae produced a slope closer to equality (closer to 1.0) in regression analysis and is considered of equal utility in predicting the true sodium concentration, given its reliance on cholesterol versus triglycerides. Of importance, should triglycerides be available, the SWF_EST‐HUM or SWF_{EST‐CAN‐TRIG} formula with forced triglycerides can be used since they are essentially statistically equivalent. The options available for correcting serum sodium in cases of altered SWF are summarized in Figure 5. This figure also aims to emphasize that measuring electrolyte values via direct potentiometry is preferred in such cases whenever possible.

FIGURE 5.

Summary of the options available for correcting serum sodium in cases of altered serum water fraction in dogs. Options are listed in order of preference, as determined by this study and due to the practical limitations of measuring SWF_ACTUAL in a clinical setting.

It is interesting to note that the species‐specific formulae for SWF_EST‐CAN did not confer as much advantage for the corrected sodium concentration in dogs as it did for cats. ⁶ This could be due to generally increased variability in the size and breed of dogs relative to cats, which could subsequently impact SWF. Alternatively, canine physiology may simply be more closely related to human physiology in this regard, as evidenced by the similarity between the average human SWF and the average canine SWF determined in this study. Although the mean SWF_ACTUAL for dogs (92.7%) was very similar to the human value (93%) in this study, our findings suggest that relative impacts on SWF from various serum constituents (triglycerides, cholesterol, and total protein) are different between humans and dogs. Indeed, the canine correction formula uses a larger multiplier for total protein (0.91 for dogs instead of 0.7 for humans); this implies that, in a fixed volume of serum, total protein appears to contribute relatively more to sample volume in dogs than in humans. The reason for this is unknown but presumed to be related to a difference in the serum proteome of dogs versus humans, particularly the physical and chemical properties (eg, size and hydrophobicity) of their respective constituents.

It is also intriguing that the absolute value of the bias between corrected sodium and direct ISE sodium was greater for the canine formula than either the human formula or the canine formula with forced triglycerides, despite that the canine formula recapitulated serum water fraction better than the other two. One possible explanation is that the “true” sodium measurement is innately higher on the Cobas than the Nova instrument, as demonstrated by the bias in their respective reference intervals (both established in healthy dogs)—upper end of 153 mEq/L on the Cobas and 150 mEq/L on the Nova. Thus, the apparent “negative bias” between the methods could be real in the sense that there are instrument‐related differences in true measured sodium. That said, our study found these correction formulae essentially statistically equivalent, and the greater negative bias in the canine formulae could also be considered random variation.

Limitations to this study include the lack of a separate cohort to validate the formulae and the relatively small number of dogs that were classified as hyperproteinemic (n = 19). Future studies should address these concerns.

4.1. Case example

To demonstrate the potential utility of SWF_EST‐CAN, an example is provided of a hyperproteinemic and hyperlipidemic canine patient (Tables 1 and 2). In this case, the SWF_ACTUAL was 88.3% following complete lyophilization. This was notably lower than the average SWF of 92.7%, attributable to elevated serum proteins and lipids. As such, a corrected sodium concentration using the mean SWF_ACTUAL as SWF_CON‐CAN in the numerator and the SWF_ACTUAL for this dog in the denominator is found below.

$$\begin{array}{c}\text{Corrected}\left[\mathrm{Na}\right]=\left(\text{indirect}\left[\mathrm{Na}\right]\times 0.927\right)/{\mathrm{SWF}}_{\text{ACTUAL}}\hfill \\ =\left(149.0\times 0.927\right)/0.883=156.4\hfill \end{array}$$

TABLE 2.

Comparison of methods to estimate serum water fraction (SWF), and correct serum sodium concentration based upon these estimates, for a canine patient example.

SWF (%)		[Na] (mEq/L)		Reference interval (mEq/L)
		Measured [Na] by Nova 16 (direct potentiometry)	151.0	143–150
		Measured [Na] by Cobas 6000 (indirect potentiometry)	149.0	143–153
SWFACTUAL	88.3	Corrected [Na] using SWFACTUAL and SWFCON‐CAN	156.4	143–153
SWFEST‐CAN	89.4	Corrected [Na] using SWFEST‐CAN and SWFCON‐CAN	154.5	143–153
SWFEST‐HUM	93.0	Corrected [Na] using SWFEST‐HUM and SWFCON‐HUM	149.0	143–153
SWFEST‐ CAN‐TRIG	92.9	Corrected [Na] using SWFEST‐ CAN‐TRIG and SWFCON‐CAN	148.7	143–153

Note: This patient's SWF_ACTUAL, and sodium concentrations measured by both direct and indirect potentiometry, are provided for reference. Values in red are above the reference interval.

If the canine‐specific correction formula determined in this study is applied to this patient, the SWF_EST‐CAN is calculated as 89.4% (below).

$$\begin{array}{cc}{\mathrm{SWF}}_{\mathrm{EST}-\mathrm{CAN}}=\hfill & \left[98.3-\left(0.002\times \text{cholesterol concentration in}\mathrm{mg}/\mathrm{dL}\right)\right]\hfill \\ \hfill & -\left(0.91\times \text{protein concentration in}\mathrm{g}/\mathrm{dL}\right)/100\hfill \\ \hfill & =\left[98.3-\left(0.002\times 607\right)-\left(0.91\times 8.4\right)\right]/100=0.894\hfill \end{array}$$

The advantage of the species‐specific formula is evident when comparing the canine‐specific SWF_EST‐CAN (0.894, above) to the SWF_EST‐HUM (0.930, below) using the human‐specific formula, as the SWF_ACTUAL was measured at 88.3%.

$$\begin{array}{cc}{\mathrm{SWF}}_{\mathrm{EST}-\mathrm{HUM}}=\hfill & \left[99.1-\left(0.001\times \text{triglyceride concentration in}\mathrm{mg}/\mathrm{dL}\right)\right.\hfill \\ \hfill & \left.–\left(0.7\times \text{protein concentration in}\mathrm{g}/\mathrm{dL}\right)\right]/100\hfill \\ \hfill & =\left[99.1-\left(0.001\times 256\right)-\left(0.7\times 8.4\right)\right]/100=0.930\hfill \end{array}$$

The SWF_EST‐CAN and SWF_EST‐HUM were then used to determine the corrected sodium concentration.

$$\begin{array}{c}\text{Corrected}\left[\mathrm{Na}\right]=\left(\text{indirect}\left[\mathrm{Na}\right]\times 0.927\right)/{\mathrm{SWF}}_{\mathrm{EST}-\mathrm{CAN}}\hfill \\ =\left(149.0\times 0.927\right)/0.894=154.5\hfill \end{array}$$

$$\begin{array}{c}\text{Corrected}\left[\mathrm{Na}\right]=\left(\text{indirect}\left[\mathrm{Na}\right]\times 0.93\right)/{\mathrm{SWF}}_{\mathrm{EST}-\mathrm{HUM}}\hfill \\ =\left(149.0\times 0.93\right)/0.930=149.0\hfill \end{array}$$

Based on the indirect measurement alone, the clinician may not have initiated the appropriate therapy for hypernatremia. This case illustrates that significant hyperproteinemia and hyperlipidemia decreased serum water fraction and thereby falsely lowered indirect ISE serum sodium concentration to within the reference interval when, in fact, this patient was hypernatremic based on direct ISE (Table 1). Of note, the trend for potassium and chloride were similar to that of sodium, in that both were artifactually decreased when measured by indirect potentiometry due to the decreased SWF present in this sample. The canine sodium correction formula appropriately corrected this patient into the hypernatremic range, while the human correction formula and the canine formula with forced trigylcerides did not (Table 2). Thus, this particular case demonstrates the advantages of the canine‐specific formula over relying on the formula developed in humans, even though they were found to be statistically equivalent in this study.

Spurious increases in indirect ISE‐measured sodium secondary to hypoproteinemia can result in pseudohypernatremia in a patient with a relatively normal serum sodium concentration or it could mask a true hyponatremia. As such, there is the potential for unwarranted treatment of patients with spurious hyponatremia secondary to hyperproteinemia. Conversely, spurious decreases in indirect ISE‐measured sodium secondary to hyperproteinemia can result in pseudohyponatremia in a patient with a relatively normal serum sodium concentration or it could mask a true hypernatremia. Thus, ignoring the effect of alterations to serum water fraction or reliance on human‐specific formulae alone to make corrections may adversely affect institution of appropriate patient care.

With regard to practical application of these formulae, an important next step will be to establish abnormal total protein cutoff values above/below which SWF estimation and electrolyte correction formulas should be used (or direct potentiometry should be pursued). These cutoff values may or may not be the same as total protein reference intervals, but it is the authors' view that use of the relevant total protein reference intervals is a good starting point for this determination. Thus, laboratories could consider automated programming to include corrected electrolyte values (in addition to reporting the original electrolyte values by indirect potentiometry) for cases of hypo‐ or hyperproteinemia in dogs. An automatic comment directing the clinician to confirm these corrected sodium values using direct potentiometry in these cases is also prudent.

5. CONCLUSION

In conclusion, this study demonstrated an average serum water fraction of 92.7% for canine patients. Hypo‐ and hyperproteinemia caused falsely increased and decreased sodium concentrations, respectively, when measured by indirect potentiometry. Canine‐specific formulae were established for estimating serum water fraction (SWF_EST‐CAN) based on measured total protein and cholesterol and for correcting serum sodium based on this calculated SWF_EST‐CAN in place of SWF_ACTUAL. Application of these formulae is expected to limit the misinterpretation of electrolyte data from indirect potentiometry in the case of significantly altered serum water fraction as seen with dysproteinemia or hyperlipidemia. Future experiments should include establishment of standard serum protein cutoff concentrations, beyond use of the relevant reference interval, for which action must be taken to correct for falsely reported serum electrolyte concentrations.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest relevant to this work.

Evans SJM, Allen B, Mollenkopf D, et al. Formulae to correct sodium concentrations for serum water fraction in cases of hypo‐ and hyperproteinemia in dogs. Vet Clin Pathol. 2025;54:35‐42. doi: 10.1111/vcp.13410