Evaluation of two portable meters for determination of blood triglyceride concentration in dogs

Elissa K. Kluger Faculty of Veterinary Science, The University of Sydney, NSW 2006, Australia.

Search for other papers by Elissa K. Kluger in
Current site
Google Scholar
PubMed
Close
 BVSc
,
Navneet K. Dhand Faculty of Veterinary Science, The University of Sydney, NSW 2006, Australia.

Search for other papers by Navneet K. Dhand in
Current site
Google Scholar
PubMed
Close
 BVSc, PhD, MVSc
,
Richard Malik Centre for Veterinary Education, The University of Sydney, NSW 2006, Australia.

Search for other papers by Richard Malik in
Current site
Google Scholar
PubMed
Close
 DVSc, PhD
,
William J. Ilkin Kirrawee Veterinary Hospital, 540 Princes Hwy, Kirrawee, NSW 2232, Australia.

Search for other papers by William J. Ilkin in
Current site
Google Scholar
PubMed
Close
 BVSc
,
David H. Snow Symbion Vetnostics Laboratory, 60 Waterloo Rd, North Ryde, NSW 2113, Australia.

Search for other papers by David H. Snow in
Current site
Google Scholar
PubMed
Close
 DVSc, PhD
, and
Merran Govendir Faculty of Veterinary Science, The University of Sydney, NSW 2006, Australia.

Search for other papers by Merran Govendir in
Current site
Google Scholar
PubMed
Close
 BVSc, PhD

Click on author name to view affiliation information

Abstract

Objective—To evaluate agreement between 2 portable triglyceride meters and a veterinary laboratory for measurement of blood triglyceride concentrations in dogs and evaluate effects of Hct and blood volume analyzed.

Sample Population—97 blood samples collected from 60 dogs.

Procedures—Triglyceride concentrations were measured in blood by use of 2 meters and compared with serum triglyceride concentrations determined by a veterinary laboratory. Within- and between-day precision, accuracy, and effects of blood volume and Hct were analyzed.

Results—Accuracy of both meters varied with triglyceride concentration, although both accurately delineated dogs with triglyceride concentrations < 180 mg/dL versus ≥ 180 mg/dL. One meter had results with excellent overall correlation with results of the standard laboratory method, with a concordance correlation coefficient of 0.94 and mean difference of 20.3 mg/dL. The other meter had a good overall concordance correlation coefficient of 0.86 with a higher absolute mean difference of −27.7 mg/dL. Results were only affected by blood volume; triglyceride concentrations determined via both meters were significantly lower when 7 μL of EDTA-anticoagulated blood was used, compared with larger volumes.

Conclusions and Clinical Relevance—1 meter had greater accuracy in the range of 140 to 400 mg/dL and was therefore well suited to detect hypertriglyceridemia. The other meter was accurate with triglyceride values < 140 mg/dL and yielded results similar to those of the veterinary laboratory in the range of 140 to 400 mg/dL, therefore being suitable for determination of triglyceride concentrations in nonfed dogs and dogs with mildly high concentrations.

Abstract

Objective—To evaluate agreement between 2 portable triglyceride meters and a veterinary laboratory for measurement of blood triglyceride concentrations in dogs and evaluate effects of Hct and blood volume analyzed.

Sample Population—97 blood samples collected from 60 dogs.

Procedures—Triglyceride concentrations were measured in blood by use of 2 meters and compared with serum triglyceride concentrations determined by a veterinary laboratory. Within- and between-day precision, accuracy, and effects of blood volume and Hct were analyzed.

Results—Accuracy of both meters varied with triglyceride concentration, although both accurately delineated dogs with triglyceride concentrations < 180 mg/dL versus ≥ 180 mg/dL. One meter had results with excellent overall correlation with results of the standard laboratory method, with a concordance correlation coefficient of 0.94 and mean difference of 20.3 mg/dL. The other meter had a good overall concordance correlation coefficient of 0.86 with a higher absolute mean difference of −27.7 mg/dL. Results were only affected by blood volume; triglyceride concentrations determined via both meters were significantly lower when 7 μL of EDTA-anticoagulated blood was used, compared with larger volumes.

Conclusions and Clinical Relevance—1 meter had greater accuracy in the range of 140 to 400 mg/dL and was therefore well suited to detect hypertriglyceridemia. The other meter was accurate with triglyceride values < 140 mg/dL and yielded results similar to those of the veterinary laboratory in the range of 140 to 400 mg/dL, therefore being suitable for determination of triglyceride concentrations in nonfed dogs and dogs with mildly high concentrations.

Point-of-care triglyceride meters are used in human medicine to monitor fasting and postprandial triglyceride concentrations because high serum triglyceride concentration is a known risk factor for development of cardiovascular disease.1 Fasting hypertriglyceridemia has also become an important component of the metabolic syndrome in humans and in itself may be linked to insulin resistance.2 Individuals at risk, such as those with diabetes or obesity, despite fasting triglyceride concentrations that are not abnormal, can often develop postprandial hypertriglyceridemia.3 Obesity and diabetes mellitus in dogs can also induce similar pre- and postprandial triglyceride changes as those in humans.4

Point-of-care glucose testing is commonly performed in human and veterinary medicine to monitor patients with diabetes. However, POC triglyceride testing has only been validated in human patients.5–9 Veterinary conditions that may benefit from periodic monitoring of nonfed serum triglyceride concentration include familial hypertriglyceridemia in Miniature Schnauzers10 and other idiopathic hyperlipidemias, diabetes mellitus, acute pancreatitis, obesity,4 hyperadrenocorticism,11 and hypothyroidism12 as well as epilepsy in dogs treated with phenobarbital or bromide13,14 and dogs receiving total parenteral nutrition or propofol infusions. Because serial blood samples may be required to monitor postprandial hypertriglyceridemia in a clinical or research setting, triglyceride meters could be used to minimize the total volume of blood required as well as to provide an immediate result.

The objectives of the study reported here were to evaluate agreement between 2 portable triglyceride meters and a veterinary laboratory for measurement of blood triglyceride concentrations in dogs and evaluate effects of Hct values and blood volume analyzed.

Materials and Methods

Animals—Nonfed and postprandial blood samples were collected from a variety of client-owned dogs that were evaluated for a variety of medical reasons. Blood samples were collected at The University of Sydney and Kirrawee Veterinary Hospital. There were 32 female dogs (31 spayed) and 28 male dogs (26 castrated), with ages ranging from 1 to 17 years (mean ± SD, 7.4 ± 3.8 years). A variety of dog breeds were included, although no single breed was overrepresented. Nonfed samples were collected after a 14- to 16-hour period of feed withholding, whereas postprandial samples were collected 2 to 8 hours after the dogs were fed their normal meal or after an oral fat tolerance test. Blood samples were collected as part of a preanesthetic profile or general health profile from healthy dogs or from dogs participating in another study.13 Sixty dogs participated in this study, although blood collection was performed more than once in most dogs because serial triglyceride determinations before and after feeding were performed. All study procedures were approved by The University of Sydney Animal Ethics Committee.

Blood collection and handling—Blood was collected from the external jugular vein or cephalic vein by use of a 22- or 23-gauge needle and syringe. Following blood collection, the needle was removed from the syringe and a drop of venous blood was placed on each triglyceride test strip. The remaining blood was either placed in a serum clot tube only for determination of serum triglyceride concentration at the laboratory or in a serum clot tube and an EDTA-containing tube for determination of serum triglyceride concentration or plasma triglyceride concentration and Hct. Following centrifugation for 10 minutes at 2,500 X g, serum and plasma were separated and triglyceride concentration was determined by the laboratory in both types of samples within 12 hours of collection. Samples were stored at 4°C until processed. An additional smaller study was conducted on 40 serum samples obtained after centrifugation to compare serum triglyceride concentrations determined by use of each meter with the serum triglyceride concentration determined by use of the veterinary laboratory.

Analysis of samples—Two triglyceride meters were designated Aa and Bb and evaluated (Appendix). Meterspecific test strips were used, with a designated area for the application of a drop of blood. Both analyzers initially separate RBCs from the sample; a small amount of plasma is then automatically directed onto a test strip. The reaction principle of the test strip is lipolysis of triglycerides by lipoprotein lipase to produce glycerol and fatty acids. Glycerol is phosphorylated by glycerol kinase, and the resulting glycerol phosphate is oxidized by a GPO-PAP to form dihydroxyacetone phosphate. The presence of oxygen leads to the formation of hydrogen peroxide, which reacts with aminophenazone and chlorophenol to form a blue-gray oxidation product. The color intensity is measured by reflectance photometry. Once completed, the back of the test area on each test strip should change color evenly, provided enough blood is applied.

The detectable triglyceride concentration ranges specified by the manufacturers of each meter were 70 to 600 mg/dL (0.80 to 6.86 mmol/L) and 50 to 500 mg/dL (0.56 to 5.65 mmol/L) for meters A and B, respectively. If triglyceride concentrations were outside these reference ranges, meter A would display either LO or HI, and meter B would display < 50 or > 500. For comparison, serum triglyceride concentrations were also determined by a nationally accredited veterinary laboratory by use of an automated analyzer with a GPO-PAP test kit.c

Precision—To determine within-run precision, 5 to 8 repeated measurements of EDTA-anticoagulated blood known to have a triglyceride concentration that was either within reference range (< 140 mg/dL [≤ 1.60 mmol/L]), mildly increased (compared with reference range; 140 to 400 mg/dL [1.70 to 4.40 mmol/L]), or moderately increased (compared with reference range; > 400 mg/dL [≥ 4.50 mmol/L]) were performed on each meter for 2 dogs with triglyceride concentrations within reference range, 2 dogs with moderately increased triglyceride concentration, and 5 dogs with a mildly increased triglyceride concentration. To determine between-run precision, 2 groups of pooled canine serum (1 group known to have triglyceride concentration that was within reference range and 1 group known to have mildly or moderately increased triglyceride concentration) were measured on each meter for 10 consecutive days.

Quality-control solutions were used for each meter once every 2 weeks and after calibration for each new batch of test strips; meter A had 1 control solution, and meter B had 2. Within-run and between-run precision testing was also performed by the veterinary laboratory on pooled canine serum.

Effect of blood volume or Hct—To determine whether the volume of blood applied to the test strip affected meter accuracy, test strips were covered with 7, 10, 15, 20, and 30 μL of blood with EDTA from samples with triglyceride concentrations that were within reference range or moderately increased. Determinations of triglyceride concentrations for each blood volume were repeated in triplicate. These volumes were chosen because 7 μL approximates 1 drop from a 100-unit (1-mL) insulin syringe with a 30-gauge needle attached, 10 μL is the minimum volume specified for meter A, 15 μL is the minimum volume specified for meter B, 20 μL slightly exceeds the volume required for both meters (approx 1 drop of blood with a 22- or 23-gauge needle), and 30 μL is the approximate volume obtained from a human finger prick15 or 1 drop of blood from a 5-mL syringe. A drop of blood from each size of needle and syringe was weighed, and 1 mL of blood was estimated to weigh 1 g.

To determine the effect of Hct, blood collected in EDTA-containing tubes was used to determine the Hct in 80 samples. For both meters, differences in triglyceride concentrations, compared with the laboratory values, were calculated by comparison of Hct values ≤ 50% or > 50%, a cutoff value used in previous studies.7

Statistical analysis—Statistical analysis was performed with a software packaged and graphs produced with a specific software programe unless otherwise stated. Results are expressed as mean ± SD. Values of P ≤ 0.05 were considered significant. Mean ± SD and the CV were calculated for values obtained via each meter and the laboratory to assess within-run and between-run precision.

A Bland-Altman approach was used to determine agreement between the laboratory and each triglyceride meter.16 Mean difference (bias) in the Bland-Altman analysis represented the systematic error between 2 methods. Lower and upper 95% LOA were calculated as bias ± 2 SD. Because of nonnormal distribution of triglyceride concentration differences and increasing SD with higher mean triglyceride concentrations, the 95% LOA were also estimated after log transformation of data but were back-transformed to the original scale to aid clinical interpretation. In addition to Bland-Altman plots, the Lin CCC was used as a measure of agreement. For those data, the CCC is a superior measure of correlation than Pearson correlation because it takes into account the range of data points above or below the equality line and the difference between the slope of the concordance line and the equality line, whereas Pearson correlation only measures the precision of a linear relationship. A value of CCC > 0.90 was considered to indicate excellent agreement between the 2 methods.17

Triglyceride concentrations were ranked in ascending order and grouped into reference range (< 140 mg/dL), mildly increased (140 to 400 mg/dL), and moderately increased (> 400 mg/dL) concentration ranges, as published for dogs.18 Bland-Altman LOA, bias, and CCC were calculated for each triglyceride range to evaluate agreement between the reference laboratory and each meter within each range.

To determine the ability of meters to correctly classify dogs into clinically important ranges, the numbers of samples classified into reference range, mildly increased, and moderately increased groups by the laboratory and the meters were tabulated and compared by use of kappa and weighted kappa values with software.f Kappa values > 0.81 indicate an almost perfect measure of true agreement across categorical data.19

Receiver operator characteristic curve analyses were used to further evaluate the diagnostic performance of both meters for clinical decision making. This estimated their diagnostic sensitivity and specificity in comparison with the laboratory triglyceride concentrations, by being classified into reference range and high concentrations by use of 180 mg/dL (2.00 mmol/L) as the cutoff limit, a triglyceride concentration slightly greater than the upper triglyceride reference limit.

The laboratory methodology provided measurement across all triglyceride concentrations; however, the meters could only measure from 50 to 500 mg/dL (meter B) and 70 to 600 mg/dL (meter A). Consequently, some results greater and less than these triglyceride ranges were excluded from Bland-Altman, CCC, and linear mixed model statistical analyses.

The effect of Hct on differences between each meter's values and laboratory values was evaluated by fitting a linear mixed model with Hct as a fixed effect and animal as a random effect. In addition to testing Hct as a quantitative variable, 2 categorical groups were formed (≤ 50% and > 50%) to evaluate the effect of low and high Hct values on agreement between each meter and the laboratory. Predicted means from the linear mixed model were compared by use of the least significant difference approach.

The effect of blood volume on the meter triglyceride concentration was evaluated by use of a similar linear mixed model approach; however, a variable representing the log triglyceride concentration of each meter was used as an outcome and blood volume (7, 10, 15, 20, and 30 μL), meter (A or B), triglyceride range, and an interaction term between triglyceride range and meter were included as fixed effects. Animal was included as a random effect. All interference statistical analyses were conducted by use of software.g

Results

Ninety-seven blood samples were analyzed from 60 dogs to determine agreement between the laboratory (serum) and each triglyceride meter (venous blood with no anticoagulant). Of those triglyceride concentrations determined from serum samples sent to the laboratory, 51 were in the reference range, 28 were mildly increased, and 18 were moderately increased. Of the 51 values within reference range, 20 and 8 values were less than the minimum value detectable by use of meters A and B, respectively. Of the 18 moderately increased triglyceride concentrations, 8 and 11 values were greater than the maximum value detectable by use of meters A and B, respectively. Sixteen samples from meter A (9 measured as LO and 7 measured as HI) and 8 from meter B (2 measured as < 50 mg/dL and 6 measured as > 500 mg/dL) were therefore excluded from the Bland-Altman and CCC analyses.

Within-run and between-run precision—Betweenrun and within-run precision of the 2 meters and the laboratory were determined (Tables 1 and 2). Between-run precision was better in the higher triglyceride range for meter A, with a CV of 4.12%. Precision of meter B was better in the lower triglyceride range (10.7%) but was generally poorer than meter A overall. For within-run precision, the CV for meter A was better at reference range and mildly increased triglyceride ranges. Precision of meter B was better at mild to moderately increased concentrations. Overall, meter A had much better between-run precision than meter B; however, the within-run precision of each was similar. The CV of the laboratory was < 2% within each run and < 3% between runs.

Table 1—

Between-run precision and accuracy of 2 portable meters and a reference laboratory for determination of triglyceride concentrations of various ranges in pooled canine serum samples.

MethodRangeTriglyceride (mg/dL)
MeanSDCV (%)
Meter A
< 140 mg/dL*1118.317.50
140–400 mg/dL32013.194.12
Meter B
< 140 mg/dL71.07.6310.7
140–400 mg/dL208.634.816.7
Laboratory
<140 mg/dL44.50.3562.75
140–400 mg/dL2493.561.40
> 400 mg/dL2,77674.80.82

Reference range for triglyceride concentration.

Mildly increased triglyceride concentration.

Moderately increased triglyceride concentration.

Table 2—

Within-run precision of 2 portable meters and a reference laboratory for determination of mean ± SD triglyceride concentrations (CV [%]) of various ranges in canine venous blood collected in EDTA (portable meters) or pooled canine serum (reference laboratory).

Triglyceride concentration rangeMeter AMeter BLaboratory
< 140 mg/dL*116 ± 8.01 (7.20)90.8 ± 7.12 (7.69)44.5 ± 0.31 (1.50)
140–400 mg/dL235 ± 9.79 (4.16)206 ± 9.79 (4.40)251 ± 2.76 (1.11)
> 400 mg/dL461 ± 49.8 (10.70)415 ± 23.1 (5.51)2777 ± 41.7 (0.95)

See Table 1 for key.

Accuracy and agreement—Results of the Bland-Altman analysis performed by use of original and log-transformed data were determined (Figures 1 and 2). This analysis revealed that the overall mean difference for meter A was closer to zero than was meter B (mean difference, 20.3 and −27.7 mg/dL, respectively [Figure 1; Table 3]). Exponentiation of 95% LOA calculated on the log scale indicated that meter A measurements were from 0.759 to 1.759 times the reference laboratory measurements for 95% of measurements. In comparison, measurements were from 0.491 to 1.774 times those of the reference laboratory for meter B.

Figure 1—
Figure 1—

Scatterplots (A and B) and Bland-Altman plots (C and D) of venous blood triglyceride concentrations in samples (with no anticoagulant) obtained by use of 2 portable triglyceride meters (A and B), compared with serum triglyceride concentrations determined by use of a reference laboratory (Lab).

Citation: American Journal of Veterinary Research 71, 2; 10.2460/ajvr.71.2.203

Figure 2—
Figure 2—

Scatterplots (A and B) and Bland-Altman plots (C and D) of log venous blood triglyceride concentrations (in samples with no anticoagulant) obtained by use of 2 portable triglyceride meters, compared with serum triglyceride concentrations determined by use of a reference laboratory, and Bland-Altman plots obtained by use of the original data with back-transformed LOA (E and F). See Figure 1 for remainder of key.

Citation: American Journal of Veterinary Research 71, 2; 10.2460/ajvr.71.2.203

Table 3—

Evaluation of variables associated with 2 portable meters for measurement of triglyceride concentration in canine venous blood samples (no anticoagulant), compared with a reference laboratory using canine serum from the same clot tubes.

VariableMeter AMeter B
< 140 mg/dL140–400 mg/dL> 400 mg/dLAll< 140 mg/dL140–400 mg/dL> 400 mg/dLAll
Triglyceride (mg/dL; mean ± SD)121 ± 34.7265 ± 163473 ± 88.1219 ± 14185 ± 26.7197 ± 75.7383 ± 46.3156 ± 108
Correct readings (%)42/51 (82%)25/28 (89%)12/18 (67%)79/97 (81%)51/51 (100%)21/28 (79%)12/18 (67%)84/97 (87%)
Lin CCC (95% CI)0.62 (0.46, 0.74)0.80 (0.65, 0.89)0.62 (0.1, 0.87)0.94 (0.90, 0.96)0.83 (0.74, 0.89)0.61 (0.35, 0.78)0.07 (−0.16, 0.28)0.86 (0.81, 0.90)
Mean difference*27.627.2−25.820.31.33−43.9−123−27.7
Regression (r2)0.690.750.400.900.750.500.490.86
Pearson correlation0.830.870.650.950.870.710.200.93

Bland-Altman analysis.

CI = Confidence interval.

Results obtained by use of the original scale with back-transformed LOA were determined (Figure 2). At the mildly increased triglyceride cutoff value (140 mg/dL), the 95% LOA were −60 and 60 mg/dL (meter A) and −88 and 88 mg/dL (meter B). At the moderately increased triglyceride cutoff value (400 mg/dL), the 95% LOA were −165 and 165 mg/dL and −234 and 234 mg/dL for meters A and B, respectively.

Overall concordance was better for meter A (0.94) than for meter B (0.86). The Pearson correlation and coefficient of determination (r2) were 0.95 and 0.90 for meter A, compared with 0.93 and 0.86 for meter B (Table 3). Although these values suggested excellent correlation, these are not the best measures of evaluating agreement.

The use of serum in meter A did not result in significantly different triglyceride concentrations, compared with blood, and did not result in significantly better agreement with the laboratory values (mean difference by use of blood, 26.4 mg/dL; mean difference by use of serum, 24.5 mg/dL). Use of meter B resulted in greater mean differences with the use of serum (19.7 mg/dL), compared with blood (12.9 mg/dL).

Within the 3 designated triglyceride concentration ranges (< 140 mg/dL, 140 to 400 mg/dL, and > 400 mg/dL), the coefficients of determination (r2) for meter A were 0.69, 0.75, and 0.40, respectively, and for meter B were 0.75, 0.50, and 0.49, respectively, compared with the laboratory values (Table 3). Meter B had the best performance at triglyceride concentrations within reference range, with a mean difference close to zero (1.33 mg/dL), although it performed less accurately at mildly increased triglyceride concentrations (−43.9 mg/dL) and even less accurately at moderately increased triglyceride concentrations (−123 mg/dL). The performance of meter A was consistent for all triglyceride concentration ranges (within reference range, 27.6 mg/dL; mildly increased, 27.2 mg/dL; and moderately increased, −25.8 mg/dL).

Clinical decision making—The number and percentage of correct values for each meter in relation to the laboratory values, within each specified triglyceride range, were determined (Table 3). Meter B performed better at lower triglyceride concentrations (< 140 mg/dL), whereas meter A was superior at higher triglyceride concentrations (140 to 400 mg/dL). Both meters were poor at correctly classifying triglyceride values > 400 mg/dL. Kappa and weighted kappa values were 0.742 and 0.800 (meter A) and 0.772 and 0.825 (meter B), which revealed that both meters had almost perfect measure of true agreement across categorical data.

The area under the receiver operator characteristic curve for both triglyceride meters to distinguish between dogs with positive and negative results was excellent (meter A, 0.981; meter B, 0.999). Both meters had a sensitivity of 84% and specificity of 100% at the cutoff value of 180 mg/dL.

Effect of Hct—Mean Hct measured in 80 EDTA blood samples was 48%. No correlation existed between the Hct values and the triglyceride values of either meter A (r2 = −0.08; P = 0.56) or B (r2 = 0.13; P = 0.29) or between Hct values and triglyceride values obtained by the laboratory (r2 = −0.04; P = 0.70). Agreement between the meters and laboratory, as measured by the difference in triglyceride concentration, was not significantly different when Hct values were included as a quantitative variable (P = 0.41 for meter A; P = 0.34 for meter B) or as a categorical variable (P = 0.74 for meter A; P = 0.68 for meter B). For low and high Hct values, the mean differences between meter A and laboratory triglyceride concentrations were 0.138 and 0.305 mg/dL, respectively, and mean differences between meter B and laboratory triglyceride concentrations were −0.411 and −0.228 mg/dL, respectively.

Effect of blood volume—Ninety EDTA blood samples were analyzed to determine the accuracy of measurements with respect to volume of blood applied to each test strip. Blood volumes of 10 to 30 μL were adequate for both meters. At blood volumes < 10 μL, both took longer to produce a readout, and on occasion, meter B failed to produce a readout and meter A displayed LO. Both meters had significantly lower triglyceride concentration readings when 7 μL of blood was applied, compared with all other blood volumes applied. There were only minor differences between triglyceride readings at the other volumes, with no significant difference between meters for each blood volume category. Both meters yielded consistent results at each blood volume except meter A at higher triglyceride concentrations; when 7 μL of blood was applied, the triglyceride concentration varied widely from 291 to 542 mg/dL. There was a large discrepancy between the 2 meters at higher triglyceride concentrations; meter B consistently read 178 mg/dL lower than meter A, regardless of the blood volume applied.

Discussion

Point-of-care testing in veterinary and human medicine is commonly performed to monitor blood glucose concentrations in diabetic patients. Given the increased prevalence of cardiovascular disease, obesity, and type 2 diabetes mellitus in humans,20,21 regular monitoring of triglyceride concentration and other lipid analytes is now encouraged. With an increasing prevalence of diabetes mellitus and obesity in small animals and the importance of pre- and postprandial triglyceride concentrations with regard to diseases such as pancreatitis becoming more apparent, regular monitoring of triglyceride concentrations in dogs at risk has an emerging place in veterinary practice.

In comparison with veterinary laboratory values, both meters tested in the present study gave reliable values of triglyceride concentration from 40 to 400 mg/dL and can therefore be used clinically to detect hypertriglyceridemia. Meter B provided an accurate triglyceride reading within the reference range, proving more useful to determine nonfed triglyceride concentrations in clinically normal dogs. In contrast, meter A appeared more accurate in the mildly increased triglyceride concentration range, making it more suitable for quantifying increased serum triglyceride. In the present study, the meters had weighted kappa values of 0.80 (meter A) and 0.82 (meter B) for comparison of reference range, mildly increased, and moderately increased triglyceride concentrations. This indicated that both meters had almost perfect measure of true agreement across categorical data.

Scatterplots and the Bland-Altman plot based on original data (Figure 1) revealed that both assumptions required for the Bland-Altman approach were not met. Therefore, Bland-Altman plots were created by use of log-transformed triglyceride concentrations, which approximately satisfied both the assumptions (Figure 2). Although the plots created on the basis of logarithmically transformed data were technically correct, the interpretation of log triglyceride concentrations, particularly the 95% limits of agreement, is not intuitive. Therefore, Bland-Altman plots were also presented on the original scale with back-transformed 95% LOA to aid clinical interpretation by use of the approach recently described by Euser et al.22 As is evident from those graphs, the LOA increased with the mean, suggesting increasing discrepancies between the meter and laboratory measurements, as triglyceride concentration increased.

The Bland-Altman approach also assumes that all observations are independent. In the study reported here, multiple samples were obtained from some dogs and triglyceride concentrations of such repeated samples were not independent but clustered. Clustering of the observations from individual dogs was accounted for by adopting the approach suggested by Bland and Altman.23

The upper fasting triglyceride reference limit in humans (< 200 mg/dL)24 exceeds that of nonfed dogs (140 mg/dL).13 Because both triglyceride meters were designed for use in humans, many triglyceride readings in dogs were less than the lower analytic limits of the meters. For laboratory triglyceride concentrations lower than the analytic limits of the meters (< 50 mg/dL for meter B and < 70 mg/dL for meter A), triglyceride values obtained by use of both meters were correctly categorized as being within the reference range.

Intraindividual variation in serum triglyceride concentrations is attributable to both biological variation and analytic imprecision. Intraindividual variation can reach 23% in humans25 and up to 18% in dogs,26 with up to 90% accounted for by biological variation. The recommended total error in triglyceride concentration when comparing methods of determination should be ≤ 15% with an analytic CV for accuracy not more than ± 5% and precision ≤ 5%.27 In a study6 in humans, meter A did not achieve the ≤ 15% target total error because of differences in the concentration range of < 180 mg/dL. In the present study, the 5% CV target was achieved for between-run and within-run precision by use of pooled canine serum in the mildly increased range with meter A. Meter B did not attain this between-run precision target, with CVs of 10.7% (within reference range triglyceride concentration) and 16.7% (mildly increased triglyceride concentration). Meter B had better within-run precision for mildly and moderately increased triglyceride concentrations, with CVs of approximately 5%.

Both meters are designed to use capillary blood, and although venous blood can be used, it is known to result in slightly higher triglyceride readings.28 Although the National Cholesterol Education Program guidelines for triglyceride reference ranges are based on serum, not plasma,27 many POC lipid studies5–9 in humans still compare lithium heparin-anticoagulated or EDTA-anticoagulated capillary blood analyzed by use of the meters with plasma analyzed at a reference laboratory. From those POC studies, although the correlation between both meters and the reference laboratory was > 0.90, it still appeared that meter A produced higher mean triglyceride concentration readings, compared with the laboratory, and that meter B recorded lower triglyceride concentration readings. These findings are consistent with results from the present study. Apart from the limitations of both triglyceride meters, measurements in venous blood correlated well with the serum triglyceride concentrations measured by the laboratory in this study.

Lithium heparin-anticoagulated or EDTA-anticoagulated blood is recommended for use in both meters. Because EDTA can cause an artifactual decrease in triglyceride concentrations, it may be of benefit to multiply values obtained with EDTA-anticoagulated blood by 1.03, which corrects for the dilutional effect of this anticoagulant.27,29 We found no significant difference between results obtained with lithium heparin-anticoagulated blood or EDTA anti-coagulated whole blood, compared with values obtained with serum at a reference laboratory, for triglyceride concentration within reference range or in the mildly increased range (results not shown). We also found no significant difference in triglyceride concentrations between blood and serum analyzed by use of either meter or between serum values measured by the meters and the laboratory.

Studies on the effects of blood volume analyzed have not been published for either meter. Among all blood volumes used in this study (7 to 30 μL), volume only was important at low and high triglyceride concentrations. Blood volumes > 10 μL appeared sufficient for both meters, and volumes > 20 μL were preferred. A single blood drop obtained from a 100-unit insulin syringe (approx 7 μL) consistently resulted in significantly lower readings, more noticeably at higher triglyceride concentrations. Both meters are designed to be used with a single drop of blood directly from a finger prick (approx 30 μL) or from a capillary pipette. The capillary pipette specific for meter B contained 15 μL, but no specific delivery device was supplied with meter A. The processing time of each meter is dependent on the blood triglyceride concentration; blood drop volumes < 10 μL also delayed processing time, at times resulting in meter B failing to register a value and meter A displaying a reading of LO, regardless of the triglyceride concentration. Although EDTA-anticoagulated blood was used for the blood volume study, it is likely blood with no anticoagulant would yield similar results.

Both meters evaluated in the present study provided a useful tool to classify dogs as having triglyceride concentrations that were within reference range or mildly increased. Both machines were easy to use and only required a small volume of blood. Meter A appeared most accurate with mildly increased triglyceride concentrations, and meter B was more accurate within the reference range for dogs, although typically, results obtained with both meters deviated more from the laboratory values for higher triglyceride concentrations. The meters provide an alternative method for rapid blood triglyceride concentration screening in-house, when multiple samples are required, such as in an oral fat tolerance test or in animals from which only a small amount of blood can be obtained. Further studies are recommended to determine the usefulness of both meters in clinical cases, in which decision making based on results is essential.

ABBREVIATIONS

CCC

Concordance correlation coefficient

CV

Coefficient of variation

GPO-PAP

Glycerol-3-phosphate oxidase–p-amino phenazone

LOA

Limits of agreement

POC

Point of care

a.

Accutrend GCT, Roche Diagnostics GmbH, Mannheim, Germany.

b.

PTS CardioChek, Polymer Technology Systems Inc, Indianapolis, Ind.

c.

GPO-PAP test kit Roche Diagnostics GmbH, Mannheim, Germany.

d.

MedCalc Software, Mariakerke, Belgium.

e.

GraphPad Prism, version 5.00, GraphPad Software Inc, La Jolla, Calif.

f.

GenStat, version 11.1.0.1575, VSN International Ltd, Hemel Hempstead, Hertfordshire, England.

g.

SAS/STAT Software, SAS Institute Inc, Cary, NC.

References

  • 1.

    Austin MA, Hokanson JE, Edwards KL. Hypertriglyceridemia as a cardiovascular risk factor. Am J Cardiol 1998;81:7B12B.

  • 2.

    Moro E, Gallina P, Pais M, et al. Hypertriglyceridemia is associated with increased insulin resistance in subjects with normal glucose tolerance: evaluation in a large cohort of subjects assessed with the 1999 World Health Organization criteria for the classification of diabetes. Metabolism 2003;52:616619.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 3.

    Paglialunga S, Cianflone K. Regulation of postprandial lipemia: an update on current trends. Appl Physiol Nutr Metab 2007;32:6175.

  • 4.

    Jeusette IC, Lhoest ET, Istasse LP, et al. Influence of obesity on plasma lipid and lipoprotein concentrations in dogs. Am J Vet Res 2005;66:8186.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 5.

    Moses RG, Calvert D, Storlien LH. Evaluation of the Accutrend GCT with respect to triglyceride monitoring. Diabetes Care 1996;19:13051306.

  • 6.

    Luley C, Ronquist G, Reuter W, et al. Point-of-care testing of triglycerides: evaluation of the Accutrend triglycerides system. Clin Chem 2000;46:287291.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 7.

    Panz VR, Raal FJ, Paiker J, et al. Performance of the Cardio-Chek PA and Cholestech LDX point-of-care analysers compared to clinical diagnostic laboratory methods for the measurement of lipids. Cardiovasc J S Afr 2005;16:112117.

    • Search Google Scholar
    • Export Citation
  • 8.

    Dale RA, Jensen LH, Krantz MJ. Comparison of two point-of-care lipid analyzers for use in global cardiovascular risk assessments. Ann Pharmacother 2008;42:633639.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 9.

    Mendez-Gonzalez J, Bonet-Marques R, Ordonez-Llanos J. Lipid profile obtained in ambulatory subjects by three point-of-care devices. Comparison with reference methods. Point Care J Near-Patient Test Technol 2008;7:132.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 10.

    Xenoulis PG, Suchodolski JS, Levinski MD, et al. Investigation of hypertriglyceridemia in healthy Miniature Schnauzers. J Vet Intern Med 2007;21:12241230.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 11.

    Feldman E. Hyperadrenocorticism. In: Ettinger S, Feldman E, eds. Textbook of veterinary internal medicine. Philadelphia: WB Saunders Co, 1995;15381578.

    • Search Google Scholar
    • Export Citation
  • 12.

    Chastain C, Panciera D. Hypothyroid diseases. In: Ettinger S, Feldman E, eds. Textbook of veterinary internal medicine. Philadelphia: WB Saunders Co, 1994;14871501.

    • Search Google Scholar
    • Export Citation
  • 13.

    Kluger EK, Malik R, Ilkin WJ, et al. Serum triglyceride concentration in dogs with epilepsy treated with phenobarbital or with phenobarbital and bromide. J Am Vet Med Assoc 2008;233:12701277.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 14.

    Gaskill CL, Cribb AE. Pancreatitis associated with potassium bromide/phenobarbital combination therapy in epileptic dogs. Can Vet J 2000;41:555558.

    • Search Google Scholar
    • Export Citation
  • 15.

    Hoogtanders K, van der Heijden J, Christiaans M, et al. Dried blood spot measurement of tacrolimus is promising for patient monitoring. Transplantation 2007;83:237238.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 16.

    Bland JM, Altmann DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307310.

  • 17.

    Lin L-K. A concordance correlation coefficient to evaluate reproducibility. Biometrics 1989;45:255268.

  • 18.

    Whitney MS. Evaluation of hyperlipidemias in dogs and cats. Semin Vet Med Surg (Small Anim) 1992;7:292300.

  • 19.

    Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159174.

  • 20.

    Hossain P, Kawar B, El Nahas M. Obesity and diabetes in the developing world—a growing challenge. N Engl J Med 2007;356:213215.

  • 21.

    Dunstan DW, Zimmet PZ, Welborn TA, et al. The rising prevalence of diabetes and impaired glucose tolerance: The Australian Diabetes, Obesity and Lifestyle Study. Diabetes Care 2002;25:829834.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 22.

    Euser AM, Dekker FW, le Cessie S. A practical approach to Bland-Altman plots and variation coefficients for log transformed variables. J Clin Epidemiol 2008;61:978982.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 23.

    Bland JM, Altman DG. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stats 2007;17:571582.

  • 24.

    Stein EA, Myers GL. National Cholesterol Education Program recommendations for triglyceride measurement: executive summary. Clin Chem 1995;41:14211426.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 25.

    Smith SJ, Cooper GR, Myers GL, et al. Biological variability in concentrations of serum lipids: sources of variation amoung results from published studies and composite predicted values. Clin Chem 1993;39:10121022.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 26.

    Jackson KP, Davies DT. Recognising the importance of temporal changes in the concentrations of plasma constituents in healthy dogs to aid veterinary and toxicology interpretations of biochemistry results. In: Ubaldi A, ed. State of art in animal clinical biochemistry. Parma, Italy: Novastampa, 1992;301.

    • Search Google Scholar
    • Export Citation
  • 27.

    Rifai N, Dufour D, Cooper G. Preanalytical variation in lipid, lipoprotein and apolipoprotein testing. In: Rifai N, Warnick G, eds. Handbook of lipoprotein testing. 2nd ed. Washington, DC: American Association for Clinical Chemistry Press, 2000;161187.

    • Search Google Scholar
    • Export Citation
  • 28.

    Kupke IR, Zeugner S, Gottschalk A, et al. Differences in lipid and lipoprotein concentrations of capillary and venous blood samples. Clin Chim Acta 1979;97:279283.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 29.

    Beheshti I, Wessels LM, Eckfeldt JH. EDTA-plasma vs serum differences in cholesterol, high-densitylipoprotein cholesterol, and triglyceride as measured by several methods. Clin Chem 1994;40:20882092.

    • Crossref
    • Search Google Scholar
    • Export Citation
All Time Past Year Past 30 Days
Abstract Views 35 0 0
Full Text Views 397 263 25
PDF Downloads 83 46 10
Advertisement