Glucose Detection and Concentration Estimation in Feline Urine Samples with the Bayer Multistix and Purina Glucotest ⋆

Abstract

The Bayer Multistix are commonly used for detection and estimation of feline glucosuria by veterinarians and cat owners. A newer product, the Purina Glucotest, utilizes the same enzymatic technology for detection of glucose, but has been designed for home use as a litter additive that allows interpretation of glucosuria over an 8-h period. The objectives of this study were to assess the sensitivity, specificity, and accuracy of the Glucotest and Multistix, and to assess the 8-h color stability of the Glucotest. Overall, the Glucotest had greater sensitivity and specificity than the Multistix, and more accurately estimated urine glucose concentration if evaluated at least 30 min after exposure to urine. A significant lack of agreement between the results obtained immediately after exposure to urine vs after 30 min and 8 h contradicts the 8-h color stability claim, but the change in urine glucose concentration estimation over time resulted in improved test accuracy at the 30 and 480 min time points.

Diabetes mellitus (DM) is one of the most commonly diagnosed feline endocrinopathies. The most frequently reported clinical signs and physical examination findings associated with DM, eg, polyuria, polydipsia, polyphagia, weight loss, dehydration, etc, are non-specific and are associated with a number of medical conditions of adult and geriatric cats. For this reason, clinical signs and clinicopathologic data such as the presence of hyperglycemia and glucosuria are needed to confirm the diagnosis. Diagnosis may be complicated by transient hyperglycemia, commonly known as stress hyperglycemia, which may lead to an increased blood glucose concentration. As a consequence, the traditional reliance on blood glucose concentration to diagnose and monitor therapy may be unreliable in some cats. Clinicians are often required to use tests that are not affected by transient or stress-induced hyperglycemia, eg, fructosamine, plasma b-hydroxybutyrate, glycosylated hemoglobin, and persistent glucosuria at home to establish a diagnosis of DM.

Stress-induced hyperglycemia also affects the accuracy of in-hospital monitoring of diabetic cats. Many diabetic cats are stressed as a result of illness, transport, admission to the hospital or venipuncture. An increase in the release of glucocorticoids and epinephrine has been suspected or shown to result from such stressors and may result in insulin resistance and blood glucose curve results that are not representative of actual glycemic control.^1,2 Semi-quantitative measurement of urine glucose concentration can be used as a crude assessment of glycemic control. For this reason, measurement of the urine glucose concentration at home can play an important role in monitoring glycemic control in diabetic cats. However, obtaining urine in a timely manner using a non-invasive technique can be challenging for both veterinarians and pet owners. Use of non-absorbent litter and covering the litter box with plastic wrap are commonly recommended techniques for use both at home and in veterinary facilities, but can be difficult, time consuming, and unsanitary.

The Purina Glucotest Feline Urinary Glucose Detection System was designed to easily and effectively monitor the urine glucose concentration of diabetic cats in their natural environment. It is a litter additive marketed as being able to accurately measure glucosuria. The Bayer Urine Multistix have been used for years by veterinarians and cat owners for detection and quantification of glucosuria. To our knowledge, only one evaluation exists of the ability of the Bayer Multistix 10SG to detect glucose in canine and feline urine samples.³ Although the Multistix were found to be accurate for detecting the presence of glucose, the ability to distinguish between different glucose concentrations was not discussed and the data have not been published in a peer-reviewed source to our knowledge. Furthermore, to our knowledge, the ability of the Glucotest and Multistix to detect and accurately estimate the glucose concentration in glucosuric feline urine samples has not been published. In addition, the color change that occurs after the Glucotest is exposed to urine is claimed to be stable for 8 h. Therefore, the purposes of this study were to assess the sensitivity, specificity, and accuracy of the Purina Glucotest and Bayer Multistix, and to assess the 8-h color stability of the Glucotest.

Materials and methods

Feline urine samples submitted to the Auburn University Clinical Pathology Laboratory for routine urinalysis between October 2005 and March 2006 were used. All samples were centrifuged at 1000 rpm for 6 min and the supernatant stored at −20°C until further evaluation. As only the supernatants were frozen, no cells were present to alter the glucose concentration. Freezing time of the urine samples was not standardized because the samples were submitted over a 6-month period. All urine samples were grouped and handled according to the same protocol after thawing and until the time of evaluation.

Quantitative glucose concentration was determined in all samples by an automated clinical chemistry analyzer (Hitachi 911, Boehringer Mannheim, IN, USA) with the hexokinase method (Roche/Hitachi reagents, Roche Diagnostics, IN, USA) and was considered to be the true concentration (gold standard). High and low controls were assessed on the clinical chemistry analyzer daily prior to sample analysis to ensure proper analyzer function. The quantification limits for glucose on the Hitachi 911 Chemistry Analyzer is 2–750 mg/dl. Glucose concentrations in excess of 750 mg/dl were diluted 1:10 and remeasured to determine the absolute concentration.

The week of the study, all urine samples were thawed at room temperature. To obtain adequate sample volume and create the number of samples required for evaluation, urine with low glucose concentration as measured on the chemistry analyzer was pooled and divided into aliquots. A stock solution of d-glucose in water (100 g/l [Thermo Fisher Scientific, Waltham, MA, USA]) was added to the aliquots to achieve a final urine glucose concentration of 450 mg/dl (n = 10) or 600 mg/dl (n = 10). The spiked urine samples were then serially diluted in urine with low glucose concentration (<50 mg/dl) to produce standards with approximate glucose concentrations of 50, 113, 150, 225, 300, 450 and 600 mg/dl (n = 70 total), concentrations that correspond to the concentration point estimates on the Glucotest package insert color chart or the midpoint between the point estimates. An additional 18 urine samples with low glucose concentration (<50 mg/dl) and 11 urine samples from glucosuric (>50 mg/dl) cats were evaluated without being altered. Thus, a total of 99 samples were included in all analyses in the study.

The glucose concentration of all samples was measured on the chemistry analyzer the day the samples were thawed. They were then stored at 4°C for 5 days and allowed to reach room temperature before evaluation with the Glucotest and Multistix. Glucose concentration of 15 spiked urine samples (five each of concentration 600, 150 and 50 mg/dl) was measured on the chemistry analyzer the day the samples were made and then 5 days later to ensure that the storage did not affect the glucose concentration.

The glucose concentration of the standards (n = 70) and unaltered samples (n = 29) was measured with the Glucotest and Multistix; both utilize the glucose oxidase method for glucose determination. For the Glucotest, three test pieces were placed in a Petri dish filled with standard clay cat litter to simulate home monitoring by a cat owner. Test pieces in each dish were soaked with a single urine sample, which required 3–5 drops. A trained, blinded observer estimated the glucose concentration by comparing the color chart in the Glucotest package insert (Fig 1A) to the test pieces immediately (t = 0) and at t = 30, 60, 120, 240, 360, 480 and 600 min after exposure to urine. The Multistix strips were used according to the manufacturer's recommendations. A second trained, blinded observer estimated the glucose concentration by comparing the color of the reagent pad to the color blocks on the Multistix bottle label (Fig 1B) 30 s after exposure to a single urine sample. No controls were run for either the Glucotest or the Multistix. No controls exist for the Glucotest; controls are rarely if ever used for the Multistix by cat owners. The packages were newly opened, properly stored and used according to manufacturers' instructions prior to the expiration date.

Fig 1.

(A) Color chart in the Glucotest package insert used to estimate urine glucose concentration. (B) Color blocks on the Multistix bottle label used to estimate urine glucose concentration.

Each color on the Glucotest and Multistix charts corresponds to a single urine glucose concentration point estimate, not a range. The manufacturers do not provide concentration ranges that correspond to each point estimate. In order to allow for assessment of accuracy, we established arbitrary concentration ranges for each point estimate. To give each product the maximal chance to be accurate, each glucose concentration range was bracketed by two point estimates; for example, for the Multistix, an actual glucose concentration between 100 and 249 mg/dl was considered accurate if it was read as either 100 or 250 mg/dl on the dipstick (Table 1).

Table 1.

Established glucose concentration ranges for the Glucotest and Multistix used to assess test accuracy. Each glucose concentration range was bracketed by two point estimates that were considered accurate estimates for absolute glucose concentrations within the range.

Glucotest		Multistix

Point estimate categories (mg/dl)	Acceptable actual urine glucose concentrations (mg/dl)	Point estimate categories (mg/dl)	Acceptable actual urine glucose concentrations (mg/dl)

0/50	0–49	0/100	0–99
50/150	50–149	100/250	100–249
150/300	150–299	250/500	250–499
300/600	≥300	500/1000	500–999
—	—	1000/2000	≥1000

For statistical analysis, the sensitivity and specificity were calculated for the Multistix and Glucotest with the quantitative glucose concentration as the gold standard. A true positive result was defined as a sample positive for glucose using a cutoff value of ≥100 mg/dl and ≥50 mg/dl as measured by the chemistry analyzer for the Multistix and Glucotest, respectively. The manufacturers consider urine glucose concentrations equal to or greater than the chosen cutoff values to be positive and these concentrations are near the sensitivity limit of the tests. A true negative result was defined as a sample negative for glucose based on a cutoff value of <100mg/dl and <50 mg/dl for the Multistix and Glucotest, respectively. Diagnostic sensitivity (%; the ability of the Multistix and Glucotest to identify samples with glucosuria) was calculated as TP/(TP + FN), where TP indicates the number of true positive results and FN is the number of false negative results. Diagnostic specificity (%; the ability of the Multistix and Glucotest to identify samples without glucosuria) was calculated as TN/(TN + FP), where TN indicates the number of true negative results and FP is the number of false positive results.

For each method of urine glucose detection, comparisons were made between the actual glucose concentrations classified in each category by use of Kruskal—Wallis one way analysis of variance on ranks. If a significant difference was found, post hoc comparisons were made by use of Dunn's method. The stability of the Glucotest color change was explored by comparison of the reading at t = 0 and t = 30 min to the readings at each subsequent time point. The agreement between the readings was assessed using McNemar's test for dependent proportions with the null hypothesis of marginal symmetry rejected at P ≤ 0.05. Kappa (K) was used to describe the agreement and the 95% confidence interval of K is reported. Agreement was considered strong when K > 0.8.

To determine whether storage of urine samples for 5 days at 4°C affected the glucose concentration, the glucose concentration in spiked urine samples with approximate concentrations of 600, 150 and 50 mg/dl (n = 5 each) was compared before and after storage using a paired Student's test. For all tests, significance was set at the P < 0.05 level.

Results

Storage of urine for 5 days at 4°C did not significantly change the glucose concentration. Before and after storage, the urine concentrations in the spiked samples of three different concentrations (n = 5 each) were 571 ± 13 and 569 ± 14; 303 ± 11 and 300 ± 11; and 63 ± 5 and 61 ± 4, respectively.

For detection of glucose in all feline urine samples, the Multistix had a sensitivity of 92.9% and a specificity of 75.9% compared with a sensitivity of 90% and specificity 89.5% when only unaltered (ie, nonglucosuric and naturally glucosuric) samples (n = 29) were evaluated. The sensitivity and specificity of the Glucotest were both 100% for detection of glucose in all samples and unaltered samples at the time points evaluated (0, 30, 480 min). In 99 samples evaluated with the Multistix, 24.2% (n = 24) were inaccurately classified, most of which (n = 19, 19.2% of all samples or 79.2% of the inaccurate samples) were overestimations of the glucose concentration. The Glucotest underestimated glucose concentration in one sample (1.0%) at the first time point (t = 0) and the Multistix underestimated five samples (5.2%) (Fig 2).

Fig 2.

Number of samples for each Multistix point estimate group. Accuracy, underestimation, and overestimation are based on the measured glucose concentration and established concentration ranges, which are displayed below the point estimates.

For the Multistix, the actual glucose concentrations classified within a category were compared between categories, eg, the actual concentrations in urine samples classified as negative on the Multistix were compared to the actual concentrations in urine samples classified as 100, 250, 500, 1000 and 2000 mg/dl. A significant difference was detected between the actual glucose concentrations in samples that were classified as negative for glucose (0 mg/dl) and those in samples estimated to be 250, 500, 1000, and 2000 mg/dl. In addition, the actual glucose concentrations in samples classified as 100 mg/dl were significantly different from those in samples classified as 500, 1000, and 2000 mg/dl. A significant difference was not detected between the actual glucose concentrations in samples classified as 250, 500, 1000, and 2000 mg/dl, or between the concentrations in samples classified as any pair of consecutive point estimates, eg, actual urine glucose concentrations classified as 100 mg/dl were not significantly different from those classified as 250 mg/dl (Fig 3).

Fig 3.

Actual urine glucose concentrations in samples categorized as each Multistix point estimate. The box represents the interquartile (ie, 25th—75th percentile) range, the line within the box represents the median, the upper horizontal bar represents the 10th percentile and lower horizontal bar represents the 90th percentile. The actual urine glucose concentrations in bracketed point estimates were significantly different.

The color stability of the Glucotest was assessed by the agreement between the results at t = 0 or t = 30 and subsequent time points. Moderate agreement was observed between the results at t = 0 and those at 30 min (K = 0.76, P = 0.9), 60 min (K = 0.69, P = 0.29), and 120 min (K = 0.69, P = 0.17). In comparison, stronger agreement was observed between the results at 30 min and those at 60 min (K = 0.93, P = 0.72), 120 min (K = 0.89, P = 0.36), and 240 min (K = 0.86, P = 0.13). Moderate agreement was observed between the results at 30 min and at 360 min (K = 0.84, P = 0.07); though the K value was relatively high for agreement between 30 min and 480 or 600 min (K = 0.84), it was not high enough for agreement in light of the significant marginal symmetry (P = 0.04). Thus, the color present at 30 min had a stronger agreement with the later time points. After discovering a lack of agreement between the initial (t = 0) and 30 min readings and subsequent time points, the accuracy of the Glucotest was evaluated at multiple time points (0, 30, 480 min). The inaccuracy of the test decreased over time from 22.2% at the initial reading to 10.1% at 30 min, and 3.0% at 480 min (Fig 4).

Fig 4.

Number of samples for each point estimate group for the Glucotest at t = 0, 30, 480 min. Accuracy, underestimation, and overestimation are based on the measured glucose concentration and established concentration ranges displayed below the point estimates.

The number of point estimates for which the corresponding actual glucose concentrations were significantly different also increased over time. There was a significant difference between the glucose concentrations in the samples estimated to contain 0, 50, 150, and 300 mg/dl and the samples estimated to contain 600 mg/dl glucose at all times evaluated (0, 30, 480 min). There was also a significant difference between the actual glucose concentrations in samples classified as negative vs 300 mg/dl glucose at all time points evaluated, but a significant difference between the actual concentrations classified as 50 mg/dl vs 300 mg/dl was documented only at 480 min. Additionally, at 480 min, the glucose concentrations classified as negative were significantly different from the actual glucose concentrations classified as 150 mg/dl (Fig 5A and B).

Fig 5.

(A) Actual urine glucose concentrations for each Glucotest point estimate group at t = 0 min. The bracketed point estimates contain significantly different glucose concentrations. (B) Actual urine glucose concentrations for each Glucotest point estimate group at t = 480 min. The bracketed point estimates contain significantly different glucose concentrations.

Discussion

For the study, some samples were obtained by addition of D-glucose to feline urine in predetermined amounts. The addition of small volumes of d-glucose in water to the urine samples is not believed to have affected the results. The lowest dilution was approximately 1:17 and the D-glucose was dissolved in water, the matrix of urine. Further dilution was achieved by adding urine, so the greatest dilution approached 1:200. Thus, any effect on measurements such as the pH and urine specific gravity would likely have been minimal. Furthermore, as neither test is standardized for pH or specific gravity, small changes in either parameter would be clinically insignificant.

Lastly, cat owners do not assess urine pH and urine specific gravity prior to use of either product.

Despite the numerous papers published regarding accuracy of other means of diagnosing and monitoring DM, especially blood glucose concentrations (for example,^4–10), urine glucose measurement has not been evaluated similarly. The Bayer Multistix dipsticks are commonly used in veterinary hospitals and clinics for routine urinalysis. Many facilities use the Clinitek 100 Analyzer (Bayer Corporation, Diagnostics Division, Tarrytown, NY, USA) to avoid variability in results caused by visual inspection and the inability of people to discern minor differences in color. The analyzer was not used in this study so as to mimic home use by a cat owner and because the Glucotest is not designed for automated interpretation. The Purina Glucotest detects and quantifies glucose using the same type of reagent pad and enzymatic reaction as the Multistix, but unlike the Multistix, which the manufacturer recommends be interpreted 30 s after exposure to a urine sample, the manufacturer does not provide a precise time period for evaluation. The recommendation for the Glucotest is to evaluate the pads within 8 h of exposure to urine. In the current study, the initial reading (t = 0) of the Glucotest was performed shortly after exposure to urine and corresponded to the recommendation for the Multistix.

The study was not designed to evaluate either test with urine glucose concentrations greater than 1000 mg/dl due to the 600 mg/dl point estimate maximum for the Glucotest. Despite the fact that only one sample had a glucose concentration greater than 1000 mg/dl (ie, 2487 mg/dl), the Multistix classified 22 samples in either the 1000 or 2000 mg/dl category. At high urinary glucose concentrations, the Multistix tended toward overestimations (Fig 2). Evaluation of urine samples containing large amounts of glucose (>1000 mg/dl) may have been useful for assessing the ability of the Multistix to accurately classify higher concentrations.

Assessing the accuracy of semi-quantitative tests is difficult. To do so in the current study, glucose concentration ranges for the point estimates were established in such a manner as to maximize the number of readings defined as accurate. Despite this, approximately 25% of the samples tested by the Multistix were inaccurately classified. The majority (73.7%) of the samples were overestimated by one category, but it is important to note that 66.7% (16/24) of the overestimated samples were classified in the 1000 and 2000 mg/dl categories, which has potential clinical consequences. At least one urine sample from all concentration ranges except the 100–249 mg/dl was classified in the 1000 or 2000 mg/dl point estimates and would have been considered markedly glucosuric. The overall inaccuracy of the Multistix is concerning, but of greatest importance is the overestimation of negative values that may result in an inappropriate increase in the insulin dose for such a patient if these results were used to adjust treatment. Increasing the insulin dose based on glucosuria is not ideal, but is an option when monetary concerns or the cat's personality (eg, aggression, stress) does not allow successful completion of a serial blood glucose curve. A published protocol for cats receiving insulin glargine utilizes urine glucose concentration to make insulin dosage adjustments and has been used successfully to manage feline diabetics.¹¹ The success of the protocol and safety of the patient depends on the ability of the test strip used to accurately distinguish negative or low urine glucose concentration from marked glucosuria. The overestimation of the Multistix would be problematic and potentially harmful when using this protocol.

It is also not uncommon for veterinarians to recommend an insulin dose reduction based on the persistent absence of glucosuria. For this reason, the 5% of samples in which underestimation occurred using the Multistix compared with the single sample that was underestimated only at the initial reading with the Glucotest is relevant. Although the Multistix appeared able to differentiate lower glucose concentrations (0 and 100 mg/dl) from higher glucose concentrations (500, 1000, 2000 mg/dl), the inaccuracies and discrepancies that occurred at all point estimates and the lack of a significant difference between concentrations for other point estimates make interpretation of these results difficult. Wider and overlapping actual glucose concentration ranges for the higher point estimates (ie, 250, 500, 1000, 2000 mg/dl) resulted in a lack of significant difference between the concentrations. The Multistix glucose concentrations may have been more accurate if an analyzer was used, but it would negate the purpose of the study, which was to compare two tests as used by owners.

The Glucotest initially inaccurately classified a similar percentage of samples as the Multistix, but the percentage decreased to 10% after just 30 min and 3% if the test was evaluated at 480 min. As the Glucotest is designed to be placed in the litter box and eliminate the need to catch and immediately test a urine sample, cat owners should be instructed to evaluate the reagent pads 30 min to 8 h after the cat uses the litter box. Ironically, the lack of agreement between the initial (t = 0) reading and all subsequent time points disproves the manufacturer's color stability claim. Though the most accurate results were obtained by the Glucotest 8 h after exposure to urine, the agreement between the 30 min results and 8 h results makes evaluation after 30 min a reasonable option for owners who cannot or do not want to wait 8 h. The urine samples being at room temperature may have affected the initial readings (t = 0) because the Glucotest is designed to test urine at or near body temperature. Increased urine temperature would be expected to increase the glucose concentration estimate at the time of initial exposure by affecting the glucose oxidase reaction. If this had occurred, a greater discrepancy between the actual and estimated glucose concentrations would be expected because of the total number of inaccuracies (n = 22) at t = 0, the majority (n = 21) were initially overestimated.

The high negative predictive value (100%) and positive predictive value (100%) of the Glucotest make it a good home screening test for ruling out DM in nondiabetic hyperglycemic cats and for monitoring noninsulin dependent diabetic cats during remission. Nevertheless, the lack of false negative results is also ideal for assessment of glycemic control in diabetic cats. While in this study of 99 samples Glucotest had positive and negative predictive values of 100%, when evaluated over thousands of samples, these will undoubtedly be slightly lower. Because of this, the reader is cautioned not to overlook the possibility that for an individual cat, there is a small possibility of the result being falsely positive or falsely negative. In addition, it is possible that certain exogenous factors exist that would lower the predictive values. For example, cephalexin and enrofloxacin in canine urine can cause false positive and negative results on the Multistix,¹² and the methodology of the reagent pads is the same, ie, a glucose oxidase reaction. However, the point remains that when using the same urine samples, the Glucotest accuracy was greater. In addition, the significant difference between the glucose concentrations estimated to be 0 and 50 mg/dl and 300 and 600 mg/dl allows the severity of the glucosuria to be estimated and provides a crude assessment of glycemic control.

In conclusion, the Glucotest allows for more accurate assessment of the urine glucose concentration in cats and, therefore, may be preferred to the use of the Multistix in this species. In addition, the accuracy improves if the Glucotest is interpreted at least 30 min after exposure to urine. The Glucotest, if used as a litter additive, is also a more convenient and less tedious method of quantifying urine glucose concentration in cats than are the Multistix.

References

Carlstead

Brown

Strawn

. Behavioural and physiological correlates of stress in laboratory cats. Appl Anim Behav Sci 1993; 38: 143–58.

Rand

Kinnaird

Baglioni

Blackshaw

Priest

. Acute stress hyperglycemia in cats is associated with struggling and increased concentrations of lactate and norepinephrine. J Vet Intern Med 2002; 16: 123–32.

Osborne

Stevens

. Biochemical analysis of urine: indications, methods, interpretation. In: Osborne

Stevens

, eds. Urinalysis: a clinical guide to compassionate patient care. Shawnee Mission, KS: Bayer Corporation, 1999: 86–124.

Ristic

JME

Herrtage

Walti-Lauger

SMM

. Evaluation of a continuous glucose monitoring system in cats with diabetes mellitus. J Feline Med Surg 2005; 7: 153–62.

Casella

Wess

Reusch

. Measurement of capillary blood glucose concentrations by pet owners: a new tool in the management of diabetes mellitus. J Am Anim Hosp Assoc 2002; 38: 239–45.

Fleeman

Rand

. Evaluation of day-to-day variability of serial blood glucose concentration curves in diabetic dogs. J Am Vet Med Assoc 2003; 222: 317–21.

Alt

Kley

Haessig

Reusch

. Day-to-day variability of blood glucose concentration curves generated at home in cats with diabetes mellitus. J Am Vet Med Assoc 2007; 230: 1011–7.

Cohn

McCaw

Tate

Johnson

. Assessment of five portable blood glucose meters, a point-of-care analyzer, and color test strips for measuring blood glucose concentration in dogs. J Am Vet Med Assoc 1999; 216: 198–209.

Wess

Reusch

. Capillary blood sampling from the ear of dogs and cats and use of portable meters to measure glucose concentration. J Small Anim Pract 2000; 41: 60–6.

10.

Davison

Slater

Herrtage

. Evaluation of continuous glucose monitoring system in diabetic dogs. J Small Anim Pract 2003; 44: 435–42.

11.

Rand

Marshall

. Diabetes mellitus in cats. Vet Clin North Am Small Anim Pract 2005; 35: 211–24.

12.

Rees

Boothe

. Evaluation of the effect of cephalexin and enrofloxacin on clinical laboratory measurements of urine glucose in dogs. J Am Vet Med Assoc 2004; 224: 1455–8.