Agreement of 2 electrolyte analyzers for identifying electrolyte and acid‐base disorders in sick horses

Abstract Background Use of different analyzers to measure electrolytes in the same horse can lead to different interpretation of acid‐base balance when using the simplified strong ion difference (sSID) approach. Objective Investigate the level of agreement between 2 analyzers in determining electrolytes concentrations, sSID variables, and acid‐base disorders in sick horses. Animals One hundred twenty‐four hospitalized horses. Methods Retrospective study using paired samples. Electrolytes were measured using a Beckman Coulter AU480 Chemistry analyzer (PBMA) and a Nova Biomedical Stat Profile (WBGA), respectively. Calculated sSID variables included strong ion difference, SID4; unmeasured strong ions, USI; and total nonvolatile buffer ion concentration in plasma (Atot). Agreement between analyzers was explored using Passing‐Bablok regression and Bland‐Altman analysis. Kappa (κ) test evaluated the level of agreement between analyzers in detecting acid‐base disorders. Results Methodologic differences were identified in measured Na+ and Cl− and calculated values of SID4 and USI. Mean bias (95% limits of agreement) for Na+, Cl−, SID4, and USI were: −1.2 mmol/L (−9.2 to 6.8), 4.4 mmol/L (−4.4 to 13), −5.4 mmol/L (−13 to 2), and −6.2 mmol/L (−14 to 1.7), respectively. The intraclass correlation coefficient for SID4 and USI was .55 (95%CI: −0.2 to 0.8) and .2 (95%CI: −0.15 to 0.48), respectively. There was a poor agreement between analyzers for detection of SID4 (κ = 0.20, 95%CI, 0.1 to 0.31) or USI abnormalities (κ = −0.04, 95%CI, −0.11 to 0.02). Conclusions and Clinical Importance Differences between analyzer methodology in measuring electrolytes led to a poor agreement between the diagnosis of acid‐base disorders in sick horses when using the sSID approach.

Conclusions and Clinical Importance: Differences between analyzer methodology in measuring electrolytes led to a poor agreement between the diagnosis of acid-base disorders in sick horses when using the sSID approach. ference (sSID) model. 1,2 According to the sSID model, the plasma pH is influenced by 3 independent factors: the p a CO 2 , the strong ion difference (SID), and the total weak acid concentration (A tot − ). 3 In the last decade, the development of point-of-care analyzers has permitted frequent stall-side monitoring of a horse's blood gas and electrolyte concentrations. From these results, veterinarians often calculate the sSID variables to diagnose acid-base disorders, guide medical decision-making, and provide prognosis of sick horses. There are discrepancies in the measured concentrations of strong electrolytes, especially Na + and Cl − , when comparing point-of-care technology and central laboratory analyzers in human 4-7 canine, 8 and equine tertiary institutions. 9,10 Application of the sSID approach depends on the determination of several strong electrolytes, and inconsistencies in the measurement of individual ions can lead to imprecisions in the calculated values. 1,3 Thus, errors in interpretation and diagnosis of acid-base imbalances diminishes the clinical utility of the sSID model. 11,12 Based on the published literature and our experience with several clinical cases where different analyzers were used to measure electrolytes in the same horse and the conclusion regarding acid-base disorders using the sSID were inconsistent, we hypothesized that the calculated sSID differs considerably depending on the analyzer used to measure the strong ions and plasma proteins. The first objective of this study was to determine the level of agreement between 2 different analyzers for determining concentrations of strong electrolytes (Na + , K + , and Cl − ) in sick horses. The second objective was to determine the level of agreement between total solids (TS) values obtained by refractometry and a colorimetric assay for determining total plasma protein (TP) concentrations in sick horses. The third and main objective of this study was to determine the impact of measuring electrolytes and plasma protein using different methodologies on the sSID calculations and, therefore, the diagnosis of acid-base disorders in sick horses.

| Sample collection and measurement techniques
Samples for both WBGA and PBMA were collected simultaneously by venipuncture of the jugular vein. The blood was collected into plastic collection tubes containing lithium-heparin additive. The PBMA and WBGA electrolyte concentrations were measured using a Beckman Coulter AU480 Chemistry analyzer (ion-selective electrode technology based on indirect potentiometry) and Nova Biomedical pHO Ultra Stat Profile (ionselective electrode technology based on direct potentiometry), respectively. The TP was measured with a colorimetric assay based on a modification of the Weichselbaum method when using the PBMA. The WBGA TS were determined using refractometry.

| Data collection
Institution electronic medical record systems were reviewed for data on horse signalment, presenting complaint, final diagnosis, outcome.
The following data were extracted from the VBGA and electrolyte analyses: pH, P v CO 2 (mm Hg),

| Calculations
The sSID variables were calculated as strong ion difference using 4 elec- The SID 4

| Definitions
The acid-base disorders were defined when the following variables were outside of the following reference ranges: SID 4 (38 to 46 mmol/ L), A tot (12 to 16 mmol/L), and USI (−2 to 2 mmol/L). The SID 4 acidosis was defined as a SID 4   and acid-base disorders was the same between the 2 analyzers. Significance was defined as a P-value <.05.
Intraclass correlation coefficient (ICC) was also used to determine the reliability between the SID 4 and USI calculations from the 2 analyzers and was interpreted as: ICC ≤ .5 = poor indicator of reliability; .5 < ICC ≤ .75 = moderate reliability; .75 < ICC ≤ .9 = good reliability; and >.9 = excellent reliability. 18 3 | RESULTS
There were no differences in the proportion of horses detected with Na + or K + abnormalities with each analyzer. The results of Passing-Bablok regression indicated significant methodologic differences and proportional error in measuring Na + , K + , and Cl − . The Bland-Altman plot indicated that the mean bias (95% limits of agreement) for Na + , Cl − , between the 2 analyzers were −1.2 mmol/L (−9.2 to 6.8) and 4.4 mmol/L (−4.4 to 13), respectively (Table 3 and Figure 1).

| Total solids and total protein measurements
The median (25%-75% IQR) values for TS and TP are presented in Table 1. The proportion of horses detected with TS and TP abnormalities is presented in Table 2. The measurement of TS and TP was significantly different between the refractometer and plasma biochemistry analysis (P < .01) ( Table 1). Methodologic differences, T A B L E 1 Median (25%-75% IQR) values for whole blood and plasma electrolytes, total solids (TS), and total plasma protein (TP) and calculated simplified strong ion difference obtained from 124 sick horses using both a whole blood gas analyzers (WBGA) and plasma biochemistry multianalyzer (PBMA) T A B L E 2 Prevalence of electrolyte and total protein abnormalities of 124 sick horses detected using the simplified strong ion difference approach after determining electrolytes using a whole blood gas analyzers (WBGA) and plasma biochemistry multianalyzer (PBMA) Abbreviations: A tot , the total nonvolatile buffer ion concentration in plasma; SID 4 , strong ion difference; TP, total plasma protein; TS, total solids; USI, unmeasured strong ions.
F I G U R E 1 Bland-Altman agreement plot between whole blood gas analyzers (WBGA) and plasma biochemistry multianalyzer (PBMA) for measurement of Na + , K + and Cl − and calculation of the strong ion difference (SID 4 ) and unmeasured strong ions (USI) but not proportional error in measuring TS and TP using the refractometer or chemical analyzer, were identified (Table 3 and Figure 2).

| sSID calculation and acid-base disorder diagnosis
Data for SID 4 , USI, and A tot were available for 124 horses. The median (25%-75% IQR) values for sSID variables are presented in Table 1. The measurements of SID 4 and USI were significantly different between the 2 analyzers (P < .001) ( Table 1). The measurement of A tot was also significantly different between the 2 different techniques (P = .009) The proportion of horses diagnosed with SID 4 , USI, and A tot acid-base disorders using the measured electrolytes from each machine is displayed in Table 4. The proportion of horses diagnosed with SID 4 acidosis was significantly greater when using the values of Na + , K + , and Cl − from the WBGA (85%) analyzer compared to the PBMA (44%). The proportion of horses diagnosed with USI acidosis was greater when using the values of strong electrolytes obtained from the PBMA (47%) than from the WBGA (2%) analyzer. The number of horses in which electroneutrality appeared to be violated (USI alkalosis) was higher when USI was calculated using the Na + , K + , and Cl − values from WBGA (50%) than the PBMA (2%) analyzer (Table 4). The proportion of horses diagnosed with A tot acidosis was greater when using the values of TP obtained from the PBMA (89%) than from the TS (63%) obtained with refractometer.
The Passing-Bablok regression showed significant methodologic differences, but not proportional error, in the calculated values of SID 4 , USI, and A tot (Table 3 and Figures 2 and 3). The Bland-Altman plot indicated that the mean bias (95% limits of agreement) for SID 4

| DISCUSSION
This study revealed significant differences in measuring concentrations of Na + and Cl − using a whole blood gas and electrolyte analyzer compared with a plasma biochemistry multianalyzer. These differences affected the calculation of SID 4 and USI. Similarly, determination of the TS and TP using 2 different methodologies affected the calculation of the A tot . The impact of the different methodologies on the sSID variables resulted in poor overall agreement between the analyzers and techniques to diagnose acid-base disorders in sick horses.
The reasons for the differences in the measured electrolyte concentrations between analyzers are discussed elsewhere. 11,19,20 Briefly, time between sample collection and processing by each analyzer is often different. Samples analyzed with the WBGA are generally processed within minutes after blood withdrawal, while samples analyzed with PBMA require additional time to be processed which could explain the differences in K + , but not in Na + or Cl − concentrations. 21 The type of preferred sample (whole blood vs plasma) used in each analyzer could have also affected the concentration of strong ions in the samples. However, a previous study measured plasma electrolyte F I G U R E 3 Clinical interpretation of agreement between whole blood gas analyzers (WBGA) and plasma biochemistry multianalyzer (PBMA) when assessing strong ion difference (SID 4 ), unmeasured strong ions (USI); and the total nonvolatile buffer ion concentration in plasma (A tot ) measured as total solids or totals plasma proteins. The blue shaded area indicates the area of agreement between the 2 analyzers using established clinical thresholds concentration using a point-of-care blood gas analyzer and a central laboratory chemistry analyzer and lower plasma Na + and higher Cl − concentrations were measured with the point-of-care analyzer, 22 as was the case when determining strong ions concentrations in whole blood in this study. This suggests that, at least in part, the differences in electrode activity can impact the electrolyte concentration variability. 10,19,20 This finding is expected as the methodology differences are commonly encountered between blood gas analyzers and diagnostic laboratory equipment.
The variability in the SID 4 and USI results between 2 analyzers can be explained by the accumulation of errors in each electrolyte measurement. [10][11][12] The variability in the A tot is explained by the differences in the measurement of TS and TP using either the refractometer or the chemistry analyzer. In our study, the mean differences for Na + was −1.2 and 4.4 mmol/L for Cl − . Although these differences appeared to be small, the 95% limits of agreement of these differences extended from −9.2 to +6.8 and from −4.4 to +13 mmol/L, respectively. In human medicine, there are wide limits of agreement for the SID m (−3.4 to +9.5 mmol/L and −5 to +4.7) when electrolytes are measured using point-of-care blood gas analyzer and was compared with central laboratory biochemistry multianalyzer. 11,12 In horses, there is a large limit of agreement (−3.6 to +11.5) for SID 3 (measured as SID 3 = Na + + K + − Cl − ) when the electrolyte concentrations were determined using a blood gas analyzer and automated multianalyzer system. 10 These broad limits of agreement have a compounding effect on the calculation of SID 4 (ie, 95% limits of agreement of −13 to +2 in this study) and, therefore, in the USI, as changes in SID m will produce a change of similar magnitude in the USI. 2,3,11,12 In addition, the limits of agreement of A tot also exert effect in the calculation of the USI. There is a wide range of USI (measured as SIG) in healthy foals, 22 35 and calves with diarrhea. 30 In horses, the concentration of USI (measured as strong ion gap, SIG) was greater in nonsurviving than surviving hospitalized foals. 22 This is of importance as differences in the USI values can lead clinicians to make different conclusions about the diagnosis and prognosis of the patient depending on the analyzer or techniques used for measurement of electrolytes and plasma proteins. Findings from this study suggest that clinicians should be aware of the methods and assays used by their laboratories to ensure that they are similar to those used in studies from which clinical guidelines and recommendations are provided. 11,12 This study has several limitations. The most notable is its retrospective design that prevents the determination of preanalytic effects including the exact time from sample collection to processing, sample quality and the collected blood volume or the volume of plasma used.
All these variables could have an impact on the differences in electrolyte concentrations detected in this study. Additionally, this study only investigated the SID 4 and USI in a large sample of sick horses from a single teaching hospital. Therefore, our results cannot be extrapolated to a different population. However, the cases admitted to our teaching hospital include a variety of sick horses similar to those admitted to other tertiary referral veterinary hospitals. This view is supported by the wide distribution of the values of the sSID component variables reported here. An additional limitation was that this study only compared 2 analyzers for measuring electrolytes and 2 techniques for determination of plasma proteins, limiting our conclusions to these technologies.

Sébastien Buczinski serves as Consulting Editor for Experimental
Design and Statistics for the Journal of Veterinary Internal Medicine.
He was not involved in review of this manuscript.

OFF-LABEL ANTIMICROBIAL DECLARATION
Authors declare no off-label use of antimicrobials.