Accurate determination of the degree of glucosuria in dogs and cats may be helpful in assessing the level of diabetic control achieved and in making therapeutic decisions. Urine dipstick tests represent a convenient, easy, and inexpensive method for estimating urine glucose concentration. The accuracy of urine dipsticks for this purpose has not been fully determined, and the best method for reading the results—visual versus automated—is uncertain.
In a previous study,1 moderate agreement (κ = 0.753) was identified between an automated reading of one type of urine dipstick and a visual reading of another type of dipstick for estimation of glucose concentration in canine urine; however, accuracy of the readings was not determined. In another study2 involving automated readings of 2 types of urine dipsticks made by the same manufacturer, good correlations (r = 0.78 and 0.81) with reference standard results were detected for cats, but poor correlations were found for dogs (r = 0.48 for both dipsticks). Differences between these 2 dipsticks in the sensitivity of automated readings for detection of glucosuria in dogs (60% for one and 100% for the other) indicated that large differences in accuracy may exist between various types of dipsticks. Yet another type of urine dipstick was found to have a sensitivity of 93% and specificity of 76% for detection of glucosuria in cats, and 25% of tested urine samples were inaccurately classified in that study.3 Nevertheless, automated (vs visual) readings of urine dipstick results have been suggested to improve precision in small animal hospitals in which multiple people perform this testing.4
Another diagnostic test that could be useful for assessment of glucosuria is determination of the UGCR. For evaluation of proteinuria, the urine protein-to-creatinine ratio is used to avoid the confounding effect of urine concentration and to reflect true changes in protein loss.5 Thus, the UGCR may better reflect loss of glucose in the urine than does urine glucose concentration, although this remains to be established.
The first objective of the study reported here was to assess the accuracy of automated readings of urine dipstick results for estimation of urine glucose concentration in dogs and cats, compared with results of a reference standard (ie, clinical chemistry analyzer). Our second objective was to compare visual readings of dipstick results with automated readings to determine which were more accurate. The final objective was to determine the utility of the UGCR for quantification of glucosuria in dogs and cats.
Materials and Methods
Urine samples
Two sets of urine samples submitted to the Auburn University Clinical Pathology Laboratory for complete urinalysis were used in the study. Those used to assess the accuracy of automated readings of urine dipstick results included 271 canine and 254 feline urine samples submitted from April to June 2007. No exclusion criteria were applied. Samples were included regardless of the patient's breed, sex, and age; method of urine collection (free catch, catheterization, or cystocentesis); reason for hospital admission; current medications; and any abnormalities detected on urinalysis.
Urine samples used to compare visual and automated dipstick readings and for determination of the UGCR included 39 samples from dogs and 25 samples from cats in which glucosuria had been detected on initial analysis as a consequence of natural disease from April to November 2015 and from May 2015 through May 2016, respectively. Again, as for the other set of samples, no exclusion criteria were applied.
Accuracy of automated dipstick readings
To reproduce conditions in a typical veterinary practice or clinical laboratory, all analyses were performed only once. The automated dipstick readings were performed on all samples within 2 hours after urine sample collection. Urine dipsticksa (batches differed throughout the study) were used per the manufacturer's instructions and read by means of an analyzerb in accordance with the manufacturer's instructions. A portion of each urine sample was frozen at −20°C within 2 hours after sample collection for measurement of actual urine glucose concentration. At the time of the study, the Auburn University Clinical Pathology Laboratory underwent regular quality control checks (which included the analyzer) through participation in a multilaboratory quality control program provided by Atlantic College of Veterinary Medicine, PE, Canada.
Urine dipsticks were visually read in accordance with the color chart provided by the dipstick manufacturer, whereby the lowest positive dipstick result (trace) represented a glucose concentration of 100 mg/dL, the second category (1+) a glucose concentration of 250 mg/dL, the third category (2+) a glucose concentration of 500 mg/dL, the fourth category (3+) a glucose concentration of 1,000 mg/dL, and the fifth category (4+) a glucose concentration of ≥ 2,000 mg/mL. The automated dipstick readerb measured the color change on the urine dipstick by spectrophotometry and reported urine glucose concentrations categorically as negative, trace (< 75 mg/dL), 1+ (75 to 375 mg/dL), 2+ (376 to 750 mg/dL), or 3+ (> 750 mg/dL). For statistical purposes, the readings were recorded as 0, 0.5, 1, 2, and 3, respectively.
Absolute urine glucose concentrations were spectrophotometrically measured within 30 days after sample collection by use of a clinical chemistry analyzerc and an enzymatic hexokinase oxidase reaction.d This analyzer was considered the reference standard, and results are hereafter referred to as actual glucose concentrations. In judging whether a dipstick reading was accurate, the intervals were bisected to assign corresponding actual concentrations. The readings of 0, 0.5, 1, 2, and 3 were judged as accurate if the actual glucose concentrations were ≤ 75 mg/dL, 76 to 175 mg/dL, 176 to 375 mg/dL, 376 to 750 mg/dL, and > 750 mg/dL, respectively. All urine samples with glucose concentrations ≥ 76 mg/dL were deemed positive for glucosuria for all techniques.3
Comparison of visual versus automated dipstick readings
Urine samples from glucosuric dogs and cats were frozen at −20°C until use; on the day of use, samples were thawed and brought to room temperature (approx 23°C). Urine dipsticksa were used in accordance with the manufacturer's instructions. The dipstick was dipped into the urine and withdrawn, and excess urine was removed by tapping on the counter. Results were read visually by the same trained investigator (HPL), who was unaware of the patient's status (ie, visual reading), and with the automated reader by a laboratory technician, who was also unaware of the patient's status (ie, automated reading), as described previously.
UGCR
The same urine samples as used for comparison of the visual and automated readings were used for determination of UGCRs. On the same day that the visual and automated readings were obtained, after the samples had reached room temperature, urine glucose concentration was measured with the clinical chemistry analyzer as previously described. The urine creatinine concentration was determined by use of the same analyzer via the Jaffe method with picric acid.d The UGCR was then calculated by dividing the urine glucose concentration by the urine creatinine concentration.
Statistical analysis
The sensitivity and specificity of the urine dipsticks for detection of glucosuria (ie, urine glucose concentration ≥ 76 mg/dL) were calculated by use of standard formulae. The Spearman rank order correlation test was used to evaluate the relationship between urine glucose concentrations as measured by automated dipstick reading (reading of 0, 0.5, 1, 2, or 3), visual dipstick reading (reading of 0, 0.5, 1, 2, or 3), and clinical chemistry analyzer (actual glucose concentration). Correlations were defined as excellent if the value of ρ was 0.93 to 0.99, good if 0.80 to 0.92, fair if 0.59 to 0.79, and poor if < 0.596 To assess agreement between methods, Cohen κ and weighted κ values were calculated; for both types of κ values, almost perfect agreement was defined as a κ value of 0.81 to 1, substantial agreement a value of 0.61 to 0.80, moderate agreement a value of 0.41 to 0.60, fair agreement a value of 0.21 to 0.40, and slight agreement a value ≤ 0.20.
For samples from naturally glucosuric dogs and cats, the χ2 test was used to determine whether proportions of results classified as an underestimate, an overestimate, or accurate per reference standard results differed between automated and visual readings. An underestimate was defined as a dipstick reading below the actual glucose concentration range deemed to be accurate. An overestimate was defined as a dipstick reading above the actual glucose concentration range deemed to be accurate. All analyses were performed with statistical software.e,f Values of P < 0.05 were considered significant.
Results
Accuracy of the automated dipstick reader
Dogs—For dogs, 163 of the 271 (60.1%) urine samples were accurately classified by the automated dipstick reader and placed in a color change category that reflected the actual urine glucose concentration as measured by the reference standard (Table 1). The glucose concentration was underestimated by the dipstick, compared with the actual urine glucose concentration, in 103 (38.0%) samples and was overestimated in 5 (1.8%) samples. Sensitivity of the automated reader for detection of glucosuria in dogs was 23% (95% CI, 15% to 30%), and specificity was 99% (95% CI, 97% to 100%).
Number (%) of canine urine samples with various glucose concentrations as determined by automated reading of urine dipstick results and actual glucose concentration as measured by a clinical chemistry analyzer (reference standard).
Actual glucose concentration (mg/dL) | Negative | Trace | 1+ | 2+ | 3+ |
---|---|---|---|---|---|
≤ 75 (n = 144) | 142 (99) | 2 (1) | 0 | 0 | 0 |
76–175 (n = 89) | 86 (97) | 1 (1) | 2 (2) | 0 | 0 |
176–375 (n = 15) | 11 (73) | 2 (13) | 1 (7) | 0 | 1 (7) |
376–750 (n = 5) | 1 (20) | 0 | 3 (60) | 1 (20) | 0 |
> 750 (n = 18) | 0 | 0 | 0 | 0 | 18 (100) |
Readings for which the automated reader and quantitative analyzer agreed are indicated in bold font. The automated dipstick reader used in the study assessed a color on the urine dipstick by spectrophotometry and reported urine glucose concentrations categorically as negative, trace (< 75 mg/dL), 1+ (76 to 375 mg/dL), 2+ (376 to 750 mg/dL), or 3+ (> 750 mg/dL).
Cats—For cats, 234 of 254 (92.1%) samples were read accurately by the automated dipstick reader. The glucose concentration was underestimated by the dipstick, compared with the actual urine glucose concentration, in 15 (5.9%) samples and was overestimated in 5 (2.0%) samples (Table 2). Sensitivity of the automated reader for detection of glucosuria in cats was 68% (95% CI, 51% to 84%), and specificity was 98% (95% CI, 96% to 100%).
Number (%) of feline urine samples with various glucose concentrations as determined by automated reading of urine dipstick results and actual glucose concentration as measured by a clinical chemistry analyzer (reference standard).
Actual glucose concentration (mg/dL) | Negative | Trace | 1+ | 2+ | 3+ |
---|---|---|---|---|---|
≤ 75 (n = 223) | 219 (98) | 2 (9) | 2 (9) | 0 | 0 |
76–175 (n = 10) | 8 (80) | 2 (20) | 0 | 0 | 0 |
176–375 (n = 7) | 2 (29) | 2 (29) | 2 (29) | 1 (14) | 0 |
376–750 (n = 3) | 0 | 1 (33) | 2 (66) | 0 | 0 |
> 750 (n = 11) | 0 | 0 | 0 | 0 | 11 (100) |
See Table 1 for key.
Comparison of visual and automated dipstick readings
Dogs—For dogs, all correlations between visual and automated readings and between automated readings and actual glucose concentration were good; the correlation between visual readings and actual glucose concentrations was fair (Table 3). All such correlations were significant (P < 0.001). The visual reading was higher than the automated reading for 30 of 39 (77%) samples, lower than the automated reading for 2 of 39 (5%) samples, and the same as the automated reading for 7 of 39 (18%) samples. The κ and weighted κ values for agreement between the visual and automated reading classifications were 0.03 and 0.38, respectively.
Spearman rank correlations (ρ) between results of various techniques for measurement of urine glucose concentration in canine (n = 39) and feline (25) urine samples.
Technique | Automated dipstick reading | UGCR | Actual glucose concentration |
---|---|---|---|
Visual dipstick reading | |||
Dog | 0.82 | 0.72 | 0.76 |
Cat | 0.82 | 0.87 | 0.83 |
Automated dipstick reading | |||
Dog | — | 0.80 | 0.83 |
Cat | — | 0.82 | 0.84 |
UGCR | |||
Dog | — | — | 0.91 |
Cat | — | — | 0.85 |
— = Not applicable.
Correlations were defined as excellent if 0.93 to 0.99, good if 0.80 to 0.92, fair if 0.59 to 0.79, and poor if < 0.59. All correlations were significant (P < 0.001).
Compared with actual urine glucose concentrations, automated dipstick readings most commonly underestimated the urine glucose concentration, compared with the actual concentration, and visual readings more often were accurate (Table 4). Distributions of results categorized as underestimated, overestimated, and accurate differed significantly between automated and visual reading classifications. The κ and weighted κ values for agreement between the automated reading classifications and chemistry analyzer results were 0.17 and 0.55, respectively. Respective values for agreement between the visual reading classifications and the chemistry analyzer results were 0.30 and 0.55, respectively. Incomplete distinction of categories existed between visual dipstick readings and UGCRs (Figure 1).
Number (%) of urine samples for which automated and visual dipstick readings underestimated, overestimated, or accurately reflected the urine glucose concentrations measured with a reference method.
Dog (n = 39) | Cat (n = 25) | |||
---|---|---|---|---|
Result | Automated | Visual* | Automated | Visual* |
Underestimate | 22 (56) | 5 (13) | 14 (56) | 5 (20) |
Overestimate | 6 (16) | 14 (36) | 4 (16) | 7 (28) |
Accurate | 11 (28) | 20 (51) | 7 (28) | 13 (52) |
Data distributions for the 3 result categories differed significantly (P < 0.001 for both dogs and cats) between visual and automated readings.
Cats—All correlations between visual and automated readings, between automated readings and actual glucose concentrations, and between visual readings and actual glucose concentrations were good (Table 3) and significant (P < 0.001). For 15 of 25 (60%) samples, the visual reading was higher than the automated reading; the readings were the same for 5 of 25 (20%) samples, and the visual reading was lower than the automated reading for 5 of 25 (20%) samples. The κ and weighted κ values for agreement between the visual and automated readings were 0.16 and 0.54, respectively.
Compared with actual urine glucose concentrations, automated readings most commonly underestimated the urine glucose concentration, and visual readings more often were accurate (Table 4). Distributions of results categorized as underestimated, overestimated, and accurate differed significantly between automated and visual readings. The κ and weighted κ values for agreement between automated readings and chemistry analyzer results were 0.21 and 0.53, respectively. Respective κ and weighted κ values for agreement between visual readings and chemistry analyzer results were 0.39 and 0.66. Again, incomplete distinction of categories existed between visual dipstick readings and both actual glucose concentrations and UGCRs (Figure 1).
Discussion
Assessment of glucosuria can be helpful in the management of diabetic cats and dogs.7,8 If glucosuria is persistently absent, veterinarians may recommend an insulin dose reduction or additional monitoring, such as a blood glucose curve. However, automated readings of the urine dipsticks used in the present study lacked accuracy for assessment of glucosuria in dogs and cats. Nevertheless, the high specificity of automated dipstick readings for detection of glucosuria in both species suggested that a positive reading would very likely signify the presence of glucose.
The correlation between visual and automated dipstick readings was good, similar to the correlation found with a different automated dipstick analyzer in a previous study.1 The correlations between actual urine glucose concentrations (as measured with the reference standard) and both visual and automated dipstick readings were higher than those found with use of a different type of dipstick and analyzer.2 Nevertheless, agreement between techniques ranged from slight to moderate; only a single weighted κ value (agreement between visual dipstick readings and chemistry analyzer readings for feline urine samples) was categorized as substantial. If knowledge of actual urine glucose concentration is necessary, consideration should be given to measurement on a biochemical analyzer typically used for blood.3
Contrary to findings of the present study, automated urine dipstick readings for dogs and cats in a previous study2 were more accurate than visual readings. The reason for the difference between the studies remains unclear but may be attributable to differences in technologies of the different automated readers that were used. The dipstick color changes were assessed via spectrophotometry9 by the automated reader used in the present study, whereas color changes were assessed via dual wavelength reflectance10 by the reader in the other study. In addition, having the same trained observer perform all visual readings in the present study likely resulted in increased precision4 and may have improved accuracy.
The accuracy of the automatic dipstick readings in the study reported here was much higher for feline urine than for canine urine. Again, the reason for the difference between species remains unclear. One possibility is that dog urine contains substances that interfere with automated readings. Our findings were similar to those of a previous study2 that showed urine dipstick readings are more accurate for cats than for dogs. Interestingly, the sensitivity of automated readings of 2 types of dipsticks for detection of glucosuria in dogs differed in that study, even though the dipsticks had the same manufacturer, at 60% and 100%. The glucose pad on both types of dipsticks detected glucose via a glucose oxidase reaction; therefore, the apparent difference between the 2 types may have been artifactual given that only 5 glucose-containing canine urine samples were included in the study.2
We investigated the UGCR as a potential means to more clearly define the degree of glucosuria in dogs and cats, which could potentially be helpful in assessing diabetic control. We expected that the effect of overall urine concentration on glucose readings would be nullified when the UGCR was used instead. Therefore, we used the UGCR to determine whether there would be less overlap between categories as determined by dipstick readings; if so, the UGCR could potentially distinguish between degrees of glucosuria when urine glucose concentration could not, and further study would be warranted. Although the UGCR had good correlation with results of other techniques, incomplete distinction of categories was still observed between readings; for example, the ranges of UGCR for samples classified as 3+ and 4+ were quite similar. Consequently, the UGCR did not appear to provide any additional information to aid assessment of glucosuria in dogs and cats.
A limitation of the study reported here was the lack of exclusion of urine samples from dogs or cats on the basis of diagnostic testing (eg, contrast imaging) that they may have undergone or medications (eg, certain antimicrobials) they were receiving at the time of sample collection, which might have confounded the dipstick readings. However, because the number of positive results was low in our study sample, the likelihood of interference from contrast agents or drug administration appeared low as well. The urine collection method, overall concentration, and color were not controlled for in the data analysis, and the influence of these variables on automated urine dipstick readings remains unknown. The urine collection method could impact readings because home-caught urine may be placed in containers exposed to substances that affect the dipstick reaction, such as bleach, hydrogen peroxide, or ascorbic acid.11 Urine concentration and color could also affect results of the automated reader by influencing the color of the test pad. Nevertheless, although extremes in urine color may potentially influence results, the presence of blood in urine has little effect on glucose concentration readings for canine and feline urine.12 In addition, for the comparisons of automated and visual dipstick readings, only glucosuric samples were used. Therefore, no conclusions could be made regarding how the 2 techniques compare for use with glucose-free urine samples, which constitute a large percentage of urine samples submitted to laboratories.
In conclusion, automated dipstick readings had low sensitivity for detection of glucosuria in cats and particularly dogs. Overall accuracy of dipstick readings for estimation of urine glucose concentration with the automated reader also was low (60.1%) for dogs and much higher (92.1%) for cats. On the basis of these findings, we recommend that negative test strip results be confirmed by some other method when glucosuria is expected. Although good correlation was identified between automated and visual dipstick readings for naturally glucosuric dogs and cats, visual readings were more accurate than the automated readings. No apparent advantage was found for measurement of UGCR.
Acknowledgments
No third-party funding or support was received in connection with this study or the writing or publication of the manuscript. The authors declare that there were no conflicts of interest.
Presented in part in abstract form at the 2016 American College of Veterinary Internal Medicine Forum, Denver, June 2016.
ABBREVIATIONS
UGCR | Urine glucose-to-creatinine ratio |
Footnotes
Multistix 10SG, Bayer Diagnostics, Whippany, NJ.
Clinitek 50, Bayer Diagnostics, Whippany, NJ.
Hitachi 911, Boehringer Mannheim Corp, Indianapolis, Ind.
Roche Diagnostics, Indianapolis, Ind.
Systat Software Inc, San Jose, Calif.
Prism 8, GraphPad Software, San Diego, Calif.
References
1. Bauer N, Rettig S, Moritz A. Evaluation of the Clinitek status automated dipstick analysis device for semiquantitative testing of canine urine. Res Vet Sci 2008;85:467–472.
2. Defontis M, Bauer N, Failing K, et al. Automated and visual analysis of commercial urinary dipsticks in dogs, cats and cattle. Res Vet Sci 2013;94:440–445.
3. Fletcher JM, Behrend EN, Welles EG, et al. Glucose detection and concentration estimation in feline urine samples with the Bayer Multistix and Purina Glucotest. J Feline Med Surg 2011;13:705–711.
4. Ferreira MDF, Garcia Arce M, Handel IG, et al. Urine dipstick precision with standard visual and automated methods within a small animal teaching hospital. Vet Rec 2018;183:415.
5. Chew DJ, DiBartola SP, Schenck PA. Diseases of the glomerulus. In: Chew DJ, DiBartola SP, Schenck PA, eds. Canine and feline nephrology and urology. 2nd ed. St Louis: Elsevier/Saunders, 2011;218–239.
6. Papasouliotis K, Cue S, Crawford E, et al. Comparison of white blood cell differential percentages determined by the in-house LaserCyte hematology analyzer and a manual method. Vet Clin Pathol 2006;35:295–302.
7. Nelson RW. Canine diabetes mellitus. In: Feldman EC, Nelson RW, Reusch C, et al, eds. Canine and feline endocrinology. 4th ed. St Louis: Elsevier, 2015;213–257.
8. Reusch CE. Feline diabetes mellitus. In: Feldman EC, Nelson RW, Reusch C, et al, eds. Canine and feline endocrinology. 4th ed. St Louis: Elsevier, 2015;258–314.
9. Welles EG, Whatley EM, Hall AS, et al. Comparison of Multistix PRO dipsticks with other biochemical assays for determining urine protein (UP), urine creatinine (UC), and UP:UC ratio in dogs and cats. Vet Clin Pathol 2006;35:31–36.
10. Arkray Inc. Aution Eleven AE-4020. Available at: www.arkray.eu/english/products/laboratory/analyzers/ae-4020.html. Accessed Apr 1, 2019.
11. Stockham SL, Scott MA. Urinary system. In: Fundamentals of veterinary clinical pathology. 2nd ed. Ames, Iowa: Blackwell Publishing, 2008;415–494.
12. Vientós-Plotts AI, Behrend EN, Welles EG, et al. Effect of blood contamination on results of dipstick evaluation and urine protein-to-urine creatinine ratio for urine samples from dogs and cats. Am J Vet Res 2018;79:525–531.