Introduction

Point-of-care testing is invaluable when submitting samples to a reference laboratory would create an unacceptable delay, as is the case for detection and monitoring of life-threatening electrolyte disturbances.^1,2 In a veterinary primary care or emergency setting, resources to support in-house testing are often limited, and portable instruments that rely on single-use cartridges are popular, as they are perceived as easy to use and maintain.³ Unfortunately, the fixed cost per cartridge means there is no economy of scale, so for hospitals where clinicopathologic tests are performed frequently, larger BGAs with multiuse reagents can improve cost-effectiveness.³

Benchtop instruments typically require more maintenance and quality-assurance protocols than cartridge-based analyzers, increasing demands on the clinical team.^3,4 One manufacturer provides 2 veterinary benchtop BGAs that analyze multiple samples in a single run; the newer of these 2 analyzers has reduced maintenance requirements, owing in part to the sensors for 18 analytes being combined into a single disposable sensor card that is replaced once every 28 to 30 days.⁵ The newer product also uses a multiwavelength optical system to generate cooximetry results without erythrocyte lysis, thus avoiding the cleaning and deproteinizing requirements associated with sample lysis.⁵

Before relying on a new analyzer for assessment of patient samples, it is important to determine agreement for analyte measurements with those obtained by use of a reference method.⁶ Such method agreement studies quantify bias, defined as a systematic error that may occur consistently across the analytic range (ie, constant bias) or may alter with analyte concentration (ie, proportional bias).⁶ Evaluation of bias is particularly important when published decision limits are used to guide diagnostic and therapeutic decisions, as is common in the management of electrolyte disturbances, or when it is desirable to use a common reference interval across multiple analyzers.^1,2,7,8 De novo reference interval generation is a complex and expensive process rarely performed by veterinary practitioners, but when methods agree well, it is likely that a more straightforward transference study will be sufficient to confirm that an already established reference interval can be used for the new analyzer.^9,10,11 At its simplest, a transference study involves comparing results for 20 healthy animals with a reference interval established in another population for the same species or by use of another instrument or method. If ≤ 2 results are outside the established reference interval, the new instrument is considered valid for use in the setting for which it was tested.⁹

The purpose of the study reported here was to assess agreement for measurement of sodium, potassium, and chloride concentrations in canine blood samples between 2 benchtop BGAs and a reference chemistry analyzer. These 3 analytes were chosen for investigation because changes in sodium, potassium, and chloride concentrations help to guide IV fluid therapy in an emergency setting and can be readily measured by both blood gas and reference chemistry analyzers. We aimed to compare the results provided by each device with results provided by a reference laboratory analyzer for analysis of plasma derived from the same samples.

Materials and Methods

Results

The median age of the 40 dogs in the study was 4 years (range, 1 to 14 years). Thirteen dogs were spayed females, 3 were sexually intact females, 21 were neutered males, and 3 were sexually intact males. There were 14 mixed-breed dogs, 6 Golden Retrievers, 4 Labrador Retrievers, 2 Bassett Hounds, 2 American Pit Bull Terriers, and 1 each of the following breeds: American Eskimo Dog, Chihuahua, Dachshund, Fox Terrier, German Shorthaired Pointer, Miniature Schnauzer, Pomeranian, Great Pyrenees, Shetland Sheepdog, Shih Tzu, Standard Poodle, and Weimaraner. For the 23 clinical patients, diagnoses were gastrointestinal disease (n = 6), toxicosis or envenomation (3), bacterial infection or sepsis (2), cardiac disease (2), renal disease (2), trauma or orthopedic injury (4), diabetes mellitus (1), hemangiosarcoma (1), heatstroke (1), and meningoencephalitis of unknown origin (1). All results were within the reportable range for all 3 analyzers.

The results of Passing-Bablok regression analysis indicated that BGA 1 had proportional bias for measurement of chloride concentration and constant positive bias for measurement of chloride and potassium concentrations (Table 1); the regression plots are provided (Supplementary Figure S1, available at: avmajournals.avma.org/doi/suppl/10.2460/ajvr.82.2.105). No constant or proportional biases were detected for BGA 2.

On the basis of regression equations, the predicted percentage differences from reference analyzer values were small for both BGAs at selected theoretical concentrations within, below, and above the reference chemistry analyzer and BGA 1 reference intervals for each electrolyte (Table 2). The observed median measurement differences between each BGA and the reference analyzer were smaller than the ASVCP TE_a for all 3 electrolytes, meeting our requirement for good agreement (Table 3). However, this difference exceeded the TE_a for ≥ 1 sample/analyte/BGA (range, 1 to 9).

Discussion

The present study found no significant proportional or constant biases for measurements of chloride, potassium, or sodium concentrations in canine blood samples with BGA 2, compared with measurements of these electrolyte concentrations in plasma from the same samples by use of a reference chemistry analyzer. In contrast, BGA 1, an earlier model of benchtop analyzer produced by the same manufacturer, had significant constant and proportional biases for measurement of chloride concentration and significant constant bias for measurement of potassium concentration. However, for both analyzers, predicted differences between the BGA and reference chemistry results as calculated from the regression equation were small and unlikely to be clinically relevant. This suggested that BGA 2 is an acceptable alternative to BGA 1 for diagnosis and monitoring of electrolyte disorders in dogs.

Good agreement between each BGA and the reference analyzer for measurement of electrolyte concentrations was expected, as all 3 devices used the direct potentiometry method for these measurements. In previous studies^13,14,15 that found poor agreement between BGAs and a reference method for electrolyte measurements, investigators generally compared point-of-care analyzers that use direct potentiometry with reference analyzers that use an indirect method. Although there may be differences in the performance of, for example, the specific electrodes used in different instruments that employ direct potentiometry, it is predictable that there will be greater differences between direct and indirect potentiometry methods.¹⁶ Indirect potentiometry, which involves sample dilution, is susceptible to interferences from nonaqueous plasma components such as proteins and lipids.¹⁶ Nonaqueous components do not influence direct potentiometry; therefore, conditions such as hyperproteinemia will not cause pseudohyponatremia results in samples analyzed by any of the devices used in the present study.¹⁶ It is important to note that the number of samples with marked electrolyte imbalances in the present study was small. Results predicted on the basis of the regression analysis should therefore be interpreted with caution when these were extrapolated beyond the range of values found for the study sample, as for assessment of predicted differences at our selected low potassium and high potassium and sodium concentrations. Future studies that focus on samples with markedly abnormal concentrations of these electrolytes would provide a more accurate assessment of systematic error at the extremes of the analytic ranges.

It should also be emphasized that the present study was focused on agreement between results generated by the BGAs and reference analyzer. Our regression analysis results therefore provide estimates of bias (also termed systematic error).⁶ Evaluation of bias is important for determining whether reference intervals, clinical decision limits, and critical values can be transferred between instruments.^7,9 However, random error, sometimes termed imprecision, also contributes to analytic error.¹² For each electrolyte measured with the 2 BGAs, there was a small number of samples for which observed differences between the BGA and reference analyzer exceeded the ASVCP TE_a guidelines.¹² Considering that imprecision creates sporadic, unpredictable errors, it is likely to have contributed to these occasional larger errors. Preanalytic problems, such as delayed plasma separation, sample contamination (eg, from K₂EDTA), or interferences (eg, hemolysis in certain breeds), could also have contributed to sporadic errors, although our sample-handling protocol was designed to minimize such effects.¹⁷ Ideally, sample analysis would have been repeated to verify the results, but because the BGAs required whole blood and the reference chemistry analyzer required plasma, this was not feasible.

When discussing TE_a, it is important to consider that in a veterinary context, these are guidelines rather than regulatory requirements. The ASVCP guidelines were developed by an expert panel of veterinary clinical pathologists and clinicians, and the TE_as reflect the maximum error considered clinically acceptable in most veterinary settings.¹² However, it is recognized that quality goals may need to be modified in some situations.¹⁸ Point-of-care testing is often employed in a triage setting to allow rapid identification of abnormalities that need to be rapidly addressed. Quickly recognizing electrolyte disturbances before initiating fluid therapy may be more clinically important than measuring concentrations to the level of accuracy achievable in a reference laboratory. For many clinicians who use point-of-care devices, some relaxation of TE_a targets may be acceptable, particularly if abnormal results can alter plans for resuscitative treatment or trigger confirmatory testing with a reference method.