Serum fructosamine and HbA1c are 2 glycated proteins that result from the irreversible non-enzymatic binding of glucose to serum proteins or hemoglobin in erythrocytes, respectively. In dogs, SF concentrations reflect the mean BG concentration during the preceding 1 to 2 weeks,1–3 whereas HbA1c values can be considered as an index of the mean plasma glucose concentration over the preceding 2 to 3 months.4
Serum fructosamine concentrations can be measured quickly, accurately, and economically by use of the nitroblue tetrazolium reduction method; however, only a few studies1,3,5 have been conducted to validate this assay for use with canine samples, and the reference interval for SF concentrations differs among laboratories.6 In dogs, HbA1c values have been measured by use of several methods developed for humans, including affinity chromatography,7,8 colorimetric analysis,9 ion-exchange high-performance liquid chromatography,10 immunoturbidimetric assays,11–13 capillary electrophoresis,14 and point-of-care devices that involve the use of boronate-affinity chromatography and immunoassays.15,16 Because HbA1c testing is used infrequently and is widely unavailable,14,15 measurement of HbA1c values is not commonly performed in clinical practice.
In dogs, SF concentrations do not depend only on the BG concentration. They also are affected by several pathological conditions, including hypoproteinemia, hyperlipidemia, azotemia, hypothyroidism, and hyperglobulinemia caused by multiple myeloma.17–19 In human medicine, in contrast to SF concentrations, HbA1c values appear to be less affected by diseases other than DM.20 However, HbA1c values in humans and other animals may decrease in pathological conditions that shorten erythrocyte lifespan or are associated with increased erythrocyte turnover (eg, blood-loss anemia, hemolytic anemia, splenomegaly, and pregnancy).21–23 Falsely increased HbA1c values might occur with diseases that prolong erythrocyte lifespan or are associated with a decrease in erythrocyte turnover (eg, iron deficiency anemia and splenic disorders). Moreover, falsely elevated HbA1c values attributable to interference with the assay have been described in patients with extreme hypertriglyceridemia and hyperbilirubinemia.21–23
Serum fructosamine concentrations are commonly used to monitor long-term control of DM in dogs and cats3; values can be measured during routine evaluation of diabetic dogs to clarify discrepancies between the dogs’ clinical condition and serial BG concentration measurements, thereby allowing clinicians to assess the effectiveness of changes in insulin treatment.24 Because SF concentrations change more rapidly than do HbA1c values in response to alterations in insulin treatment, SF concentration is considered better for assessment of glycemic control in diabetic dogs.24 However, discrepancies between the SF concentration and clinical condition can remain and individual dogs may differ with regard to protein glycation.25
Since 1988, the American Diabetes Association has recommended measurement of HbA1c values for routine monitoring of humans with DM,26 and such measurement is generally accepted as the best method for assessment of glycemic control.27 In situations when HbA1c values may not provide an accurate assessment of glycemic status, SF concentrations may act as a surrogate marker. Some of the potential uses of SF concentrations are in the diagnosis of gestational DM (although not recommended) and monitoring of glycemic status during end-stage renal disease, certain types of anemia, transfusions,23 and recent acute changes in glycemic control (ie, as can occur with glucocorticoids treatment22).
Studies comparing the ability of SF and HbA1c values to reflect glycemic control in dogs with DM are lacking. Therefore, the objectives of the study reported here were to evaluate the performance of 2 assays for the measurement of SF and HbA1c values in dogs, obtain specific reference intervals for both glycated proteins, and compare the usefulness of SF concentrations and HbA1c values for assessment of glycemic control in dogs with DM.
Materials and Methods
Sample
All samples were collected from dogs admitted to the Veterinary Teaching Hospital of the University of Bologna. Blood samples were collected from 40 healthy blood donors and student- or staff-owned dogs; dogs were considered healthy given the lack of abnormalities detected during physical examination and clinicopathologic testing. Blood samples were also collected from 13 client-owned diabetic dogs. Diabetes mellitus was diagnosed on the basis of clinical signs (eg, polyuria and polydipsia, weakness, and weight loss) and an unfed BG concentration > 200 mg/dL combined with glycosuria. Eight dogs had DM that was newly diagnosed, and 5 dogs were receiving insulin treatment; the 8 dogs with newly diagnosed DM were assumed to have extremely high amounts of glycated proteins. In addition, blood samples were obtained from 23 anemic (Hct < 35%) normoglycemic nondiabetic dogs that were included to evaluate the effect of anemia on HbA1c values. Of these, 9 dogs had regenerative anemia (> 71,000 reticulocytes/μL) and 14 had nonregenerative anemia (< 71,000 reticulocytes/μL). All analyses were performed on samples collected for diagnostic or monitoring purposes. Owners provided informed consent for inclusion of their dogs in the study. The study protocol was approved by the Scientific Ethics Committee of the University of Bologna.
Assays
Analytic methods—Serum fructosamine concentrations were assessed with a colorimetric nitroblue tetrazolium reduction method.a Coagulated blood samples were centrifuged for 10 minutes at 3,000 × g, and the serum was collected and stored at −80°C until analysis. The HbA1c values were assessed with an immunoturbidimetric method,b by which total hemoglobin concentration was colorimetrically measured. For this purpose, all EDTA-anticoagulated blood samples were stored at −80°C and thawed (15° to 25°C for approx 1 hour) prior to analysis. An aliquot (10 μL) of each sample was mixed with 1,000 μL of hemolyzing reagent provided by the manufacturer, and hemolysis was allowed to proceed for at least 1 minute. Hemolyzed samples were mixed thoroughly immediately before the assay was conducted. The HbA1c-to-total hemoglobin ratio was expressed as the HbA1c percentage. The following equation, which was provided by the manufacturer, was used for correction of the HbA1c value: ([HbA1c value/total hemoglobin concentration] × 91.5) + 2.15.
All measurements for both assays were performed by use of an automated chemistry analyzer.c Precision and linearity of both assays were also assessed. A CBC was performed with a flow cytometry-based hematology system.d Reticulocyte counts were performed with the automated analyzer by use of oxazine 750 perchlorate.e
Precision—Intra-assay (within-run) variability for the SF assay was estimated by determining the SF concentration in 10 serum samples (5 samples from healthy dogs with a low SF concentration and 5 samples from dogs with poorly controlled DM with a high SF concentration). Analyses were repeated 10 times within the same assay. To determine intra-assay variability of the HbA1c assay, the same procedures were used for 7 EDTA-anticoagulated blood samples (3 samples harvested from healthy dogs with a low concentration of the analyte and 4 samples from dogs with poorly controlled DM with a high concentration of the analyte).
Interassay (between-run) variability of the SF assay was estimated by determining the SF concentration in 6 serum samples (3 from healthy dogs with a low concentration of the analyte and 3 from dogs with poorly controlled DM with a high concentration of the analyte) twice daily over 5 days. All analyses were performed in duplicate. To determine interassay variability of the HbA1c assay, the same procedures were used for 2 EDTA-anticoagulated blood samples (1 from a healthy dog with a low HbA1c value and 1 from a dog with poorly controlled DM with a high HbA1c value).
Storage stability of the SF and HbA1c assays was evaluated by use of samples from 3 dogs with poorly controlled DM (with a high concentration of the analyte) and 3 healthy dogs (with a low concentration of the analyte) that had been stored at 3 temperatures (−80°C, 4°C, and 15° to 25°C [room temperature]) for 5 consecutive days. On the day of analysis, samples stored at −80°C were first thawed at room temperature for approximately 1 hour, whereas samples stored at 4°C and room temperature were thawed the first day of the analysis and maintained at 4°C and room temperature, respectively, until the end of the study period.
Linearity—Linearity of the SF assay was determined by use of 2 serum samples from 2 dogs with poorly controlled DM with a high SF concentration. Sera were diluted to obtain 75%, 50%, 25%, 12.5%, and 6.25% of the starting analyte concentration. Linearity of the HbA1c assay was assessed by use of 1 EDTA-anticoagulated blood sample from a dog with poorly controlled DM and 1 quality control human sample,f both with high HbA1c values. These samples were used undiluted and also diluted with distilled water to obtain dilutions (canine sample, 75%, 50%, 25%, and 12.5%; human sample, 75%, 62.5%, 50%, 37.5%, and 25%) of the starting analyte concentration. All analyses were performed in duplicate.
Reference intervals —Reference intervals for SF and HbA1c values in dogs were determined by use of serum and EDTA-anticoagulated blood samples obtained from the aforementioned 40 healthy dogs (various breeds and ages and both sexes).
Usefulness of HbA1c and SF values for assessment of glycemic control
Diabetic dogs receiving insulin treatment and that were admitted for routine monitoring at the Veterinary Teaching Hospital of the University of Bologna were enrolled in the study if the medical record, including the medical history, physical examination findings, body weight, BG curve, biochemical profile, and values for glycated proteins (HbA1c and SF), was complete. Dogs that had their insulin dose modified up to 1 week before the glycemic control assessment were also included. Dogs were excluded from the study when the medical records were unavailable or were missing information. Blood samples were obtained from the capillaries of the inner aspect of the pinna at 1, 2, 4, 6, 8, 10, and 12 hours after the morning meal and insulin injection to generate BG curves.
Assessment of glycemic control—To objectively evaluate glycemic control, a clinical scoring method for diabetic dogs was used.28 This clinical score included consideration of the following variables: body weight, presence of polyuria and polydipsia, median BG concentration of the BG curve, nadir BG concentration of the BG curve, and overall evaluation of the BG curve.
For each variable, a score was assigned as follows: 0 = poor, 1 = moderate, and 2 = good. Maintaining or increasing body weight was considered good (score = 2); conversely, a decrease (> 5%) in body weight was considered poor (score = 0). For obese dogs, the amount of weight loss needed to achieve an optimal body condition score was not considered. Such dogs were given a score of 2, even if they were losing weight. Polyuria and polydipsia were scored as absent (good = 2), improved (moderate = 1), or present, unchanged, or worse (poor = 0). A median BG concentration of the BG curve < 230 mg/dL, between 230 and 300 mg/dL, and > 300 mg/dL was considered good (score = 2), moderate (score = 1), and poor (score = 0), respectively. A BG nadir of the BG curve < 180 mg/dL, between 180 and 250 mg/dL, and > 250 mg/dL was considered good (score = 2), moderate (score = 1), and poor (score = 0), respectively. Overall evaluation of the BG curve was considered good (score = 2) when ≥ 50% of BG concentrations were between 80 and 270 mg/dL and poor (score = 0) when < 50% of BG concentrations were between 80 and 270 mg/dL.
Total clinical score (0 to 10) was obtained by summing the scores for each variable. Dogs were grouped on the basis of the total score into a good glycemic control group (score, 7 to 10), moderate glycemic control group (score, 4 to 6), and poor glycemic control group (score, 0 to 3).
Analytic methods —Blood glucose concentrations were measured by use of 2 portable BG meters (glucometer Ag and glucometer Bh).29,30 Detectable BG concentrations for glucometer A ranged from 20 to 600 mg/dL, whereas detectable BG concentrations for glucometer B ranged from 20 to 500 mg/dL. When BG concentrations registered as low and high on the glucometer, values of 20 mg/dL and 600 or 500 mg/dL were arbitrarily assigned. Serum samples and ED-TA-anticoagulated blood samples were obtained on the same days as the BG curves were obtained and evaluated with the SF and HbA1c assays.
Statistical analysis
Statistical analysis was performed with statistical software.i,j Data distributions were assessed graphically and by use of the D'Agostino-Pearson test. Normally distributed data are reported as mean ± SD, and nonnormally distributed data are reported as median and range. Proportions and percentages were used to describe categorical variables. Parametric and non-parametric tests were used to analyze data on the basis of their distribution.
To assess the intra-assay and interassay precision of both assays, the CV for each sample was calculated. The overall CV was calculated as the mean of these CVs. To assess linearity, measured values were plotted against expected values via linear regression analysis. To assess precision and linearity, the mean of 2 determinations was used when duplicate measurements were performed. The reference intervals for SF and HbA1c values were established by use of the robust method.
To compare HbA1c values in anemic versus non-anemic dogs and in dogs with regenerative versus nonregenerative anemia, a t test or Mann-Whitney test was used. Correlations between SF concentration and clinical score, HbA1c value and clinical score, SF concentration and HbA1c value, and HbA1c value and Hct were determined by computation of Spearman rank correlations. The Kruskal-Wallis test was used to compare the SF and HbA1c values among the 3 categories of clinical control (good glycemic control, moderate glycemic control, and poor glycemic control). When a significant difference was detected, the Mann-Whitney U test or unpaired t test was performed.
When the SF and HbA1c results differed significantly between groups, ROC curves were constructed to enable determination of optimal cutoff values with the maximum sum of sensitivity and specificity for SF and HbA1c values for discriminating between the groups. On the basis of the established cutoff values, each glycemic control assessment was categorized as correctly classified or misclassified, with the clinical score considered the reference standard. This type of analysis was performed by first considering all assessments and then separating assessments for dogs receiving insulin treatment for ≤ 3 months from the assessments for dogs receiving insulin treatment for > 3 months as well as separating dogs that had the insulin dose modified during the past 3 months from dogs on a stable insulin dose (not modified during the past 3 months).
The χ2 test was used to compare the percentage of glycemic control assessments that were correctly classified as indicated by the SF and HbA1c values. For all statistical tests, values of P < 0.05 were considered significant.
Results
Evaluation of SF and HbA1c assays
The overall mean intra-assay and interassay CVs for the SF assay were 3.8% and 2.5%, respectively, whereas the overall mean intra-assay and interassay CVs for the HbA1c assay were 1.2% and 1.8%, respectively. The CVs for samples stored frozen, refrigerated, and at room temperature were 1.8%, 1.9% and 2.5%, respectively.
Linear regression analysis of the SF data yielded R2 values of 0.997 and 0.996 for the diluted serum samples from the 2 dogs with poorly controlled DM. Linear regression analysis of the HbA1c data yielded R2 values of 0.993 for the EDTA-anticoagulated blood sample diluted to 12.5% of the starting analyte concentration and R2 values of 0.999 and 0.998 for the human quality control sample diluted to 12.5% and 25% of the starting analyte concentration, respectively.
The computed reference interval for SF concentration was 222 to 382 μmol/L (median, 286 μmol/L; range, 211 to 367 μmol/L). The 95% CI of the lower limit was 209 to 235 μmol/L, and the 95% CI of the upper limit was 357 to 402 μmol/L. The computed reference interval for HbA1c was 1.6% to 4.5% (median, 2.6%; range, 2.0% to 4.7%). The 95% CI of the lower limit was 1.4% to 1.7%, and the 95% CI of the upper limit was 4.8% to 5.2%.
Among dogs with regenerative anemia, 7 had blood loss and 2 had immune-mediated hemolytic anemia. Among dogs with nonregenerative anemia, 7 had chronic kidney disease, 1 had chronic gastrointestinal blood loss, 1 had hypothyroidism, 2 had nonregenerative immune-mediated hemolytic anemia, 2 had neoplastic disease, and 1 had inflammatory disease. The overall mean ± SD Hct of the anemic dogs was 22.2 ± 7.7%, whereas the mean Hct of dogs with regenerative anemia and nonregenerative anemia was 21.1 ± 6.1% and 22.9 ± 8.6%, respectively. The median HbA1c value of anemic dogs was 4.3% (range, 2.6% to 7.5%), which differed significantly (P < 0.001) from the median value of nonanemic dogs (2.6%; range, 2.0% to 4.7%; Figure 1). The median HbA1c value of dogs with regenerative anemia was 4.0% (range, 2.6% to 4.7%), whereas the median value of dogs with nonregenerative anemia was 4.5% (range, 3.1% to 7.5%); these values did not differ significantly (P = 0.058). No significant correlation was detected between the Hct and HbA1c value in all dogs (r = −0.02; P = 0.90), in dogs with regenerative anemia (r = 0.14; P = 0.70), or in dogs with nonregenerative anemia (r = −0.13; P = 0.70).
Box-and-whisker plots of HbA1c values in 40 healthy dogs and 23 anemic dogs (A) and in the anemic dogs grouped by whether the anemia was nonregenerative (n = 14) or regenerative (9; B). The box represents the interquartile (25th to 75th percentile) range, the horizontal line in each box represents the median, and the whiskers represent the range. In panel B, the values do not differ significantly (P = 0.058) between the 2 groups of anemic dogs. *Value differs significantly (P < 0.001) from the value for the healthy dogs.
Citation: American Journal of Veterinary Research 81, 3; 10.2460/ajvr.81.3.233
Box-and-whisker plots of HbA1c values in 40 healthy dogs and 23 anemic dogs (A) and in the anemic dogs grouped by whether the anemia was nonregenerative (n = 14) or regenerative (9; B). The box represents the interquartile (25th to 75th percentile) range, the horizontal line in each box represents the median, and the whiskers represent the range. In panel B, the values do not differ significantly (P = 0.058) between the 2 groups of anemic dogs. *Value differs significantly (P < 0.001) from the value for the healthy dogs.
Citation: American Journal of Veterinary Research 81, 3; 10.2460/ajvr.81.3.233
Box-and-whisker plots of HbA1c values in 40 healthy dogs and 23 anemic dogs (A) and in the anemic dogs grouped by whether the anemia was nonregenerative (n = 14) or regenerative (9; B). The box represents the interquartile (25th to 75th percentile) range, the horizontal line in each box represents the median, and the whiskers represent the range. In panel B, the values do not differ significantly (P = 0.058) between the 2 groups of anemic dogs. *Value differs significantly (P < 0.001) from the value for the healthy dogs.
Citation: American Journal of Veterinary Research 81, 3; 10.2460/ajvr.81.3.233
Usefulness of HbA1c and SF values for assessment of glycemic control
A total of 200 assessments of glycemic control in 46 diabetic dogs were performed. These dogs included 24 (52%) spayed females, 12 (26%) sexually intact males, and 10 (22%) neutered males. There were 21 mixed-breed dogs, 6 English Setters, 4 Poodles, 3 Labrador Retrievers, 2 Pomeranians, 2 Yorkshire Terriers, 2 Samoyeds, 2 Maltese, 1 Cocker Spaniel, 1 Yugoslavian Shepherd Dog, 1 Cavalier King Charles Spaniel, and 1 Doberman Pinscher. Mean ± SD age was 10 ± 2.08 years, and median body weight was 10.7 kg (range, 2.8 to 50 kg).
Twenty-eight dogs were treated with lente insulin,k and 18 were treated with neutral protamine Hagedorn insulin.l Nine of 46 dogs had concurrent diseases (6 dogs had hyperadrenocorticism and received trilostane,m and 3 dogs had hypothyroidism and received levothyroxinen). Slightly more than half (n = 106) of the glycemic control assessments were for dogs with newly diagnosed DM that had been receiving insulin treatment for ≤ 3 months, and 94 were for dogs with DM that had been receiving insulin treatment for > 3 months.
More than 1 assessment was performed for 39 dogs; 1 dog had 13 assessments, 1 dog had 11 assessments, 1 dog had 8 assessments, 6 dogs had 7 assessments, 6 dogs had 6 assessments, 6 dogs had 5 assessments, 5 dogs had 4 assessments, 7 dogs had 3 assessments, and 6 dogs had 2 assessments. Glycemic control, as evaluated by use of the clinical score, was classified as good in 90 of 200 (45%) assessments, moderate in 58 of 200 (29%) assessments, and poor in 52 of 200 (26%) assessments. None of the assessments revealed the Somogyi effect (hypoglycemia or rapid decrease of glycemia followed by marked hyperglycemia induced by an excessive dose of insulin).
Total clinical score was significantly (P < 0.001) and inversely correlated with the SF concentration (r = −0.40) and HbA1c value (r = −0.33). In addition, the SF concentration and HbA1c value were significantly correlated with each other (r = 0.48; P < 0.001).
Assessment of differences among the 3 groups of clinical control (good glycemic control, moderate glycemic control, and poor glycemic control groups) revealed that both SF concentrations and HbA1c values were significantly higher in dogs with poor glycemic control than in dogs with moderate glycemic control, and they were also significantly higher in dogs with moderate glycemic control than in dogs with good glycemic control (Figure 2). Because significant differences were identified between the SF concentration and HbA1c value for the dogs with good glycemic control, moderate glycemic control, and poor glycemic control, 2 ROC curve analyses were performed for each glycated protein, first by combining the moderate and good glycemic control groups versus the poor glycemic control group and then by combining the moderate and poor glycemic control groups versus the good glycemic control group.
Box-and-whisker plots of SF concentrations (A) and HbA1c values (B) for 200 assessments of glycemic control in 46 dogs with DM, with assessments grouped by level of control achieved. The definition of clinical control of DM was based on the results of a clinical history obtained from the owner, body weight, and assessment of BG curves. *Value differs significantly (P < 0.001) from the value for the poor glycemic control group. †Value differs significantly (P = 0.03) from the value for the moderate glycemic control group. ‡Value differs significantly (P = 0.01) from the value for the moderate glycemic control group. See Figure 1 for remainder of key.
Citation: American Journal of Veterinary Research 81, 3; 10.2460/ajvr.81.3.233
Box-and-whisker plots of SF concentrations (A) and HbA1c values (B) for 200 assessments of glycemic control in 46 dogs with DM, with assessments grouped by level of control achieved. The definition of clinical control of DM was based on the results of a clinical history obtained from the owner, body weight, and assessment of BG curves. *Value differs significantly (P < 0.001) from the value for the poor glycemic control group. †Value differs significantly (P = 0.03) from the value for the moderate glycemic control group. ‡Value differs significantly (P = 0.01) from the value for the moderate glycemic control group. See Figure 1 for remainder of key.
Citation: American Journal of Veterinary Research 81, 3; 10.2460/ajvr.81.3.233
Box-and-whisker plots of SF concentrations (A) and HbA1c values (B) for 200 assessments of glycemic control in 46 dogs with DM, with assessments grouped by level of control achieved. The definition of clinical control of DM was based on the results of a clinical history obtained from the owner, body weight, and assessment of BG curves. *Value differs significantly (P < 0.001) from the value for the poor glycemic control group. †Value differs significantly (P = 0.03) from the value for the moderate glycemic control group. ‡Value differs significantly (P = 0.01) from the value for the moderate glycemic control group. See Figure 1 for remainder of key.
Citation: American Journal of Veterinary Research 81, 3; 10.2460/ajvr.81.3.233
The ROC curve analyses for the use of SF and HbA1c values to differentiate good glycemic control from moderate or poor glycemic control revealed a significant (P < 0.001) AUC of 0.69 and 0.66, respectively (Figure 3). Use of an SF concentration of < 400 μmol/L to differentiate dogs with good glycemic control from dogs with moderate or poor glycemic control yielded a specificity of 71% and sensitivity of 61%. Use of an HbA1c value of < 5.5% to differentiate dogs with good glycemic control from dogs with moderate or poor glycemic control yielded a specificity of 79% and sensitivity of 41%.
The ROC curves for the use of SF concentration (A and C) and HbA1c value (B and D) in classification of glycemic control (on the basis of clinical scores) as good versus moderate or poor (A and B) and poor versus good or moderate (C and D) in dogs with DM. The AUC was 0.69 (95% CI, 0.61 to 0.76), 0.66 (95% CI, 0.59 to 0.74), 0.75 (95% CI, 0.68 to 0.82), and 0.69 (95% CI, 0.60 to 0.78) for panels A, B, C, and D, respectively.
Citation: American Journal of Veterinary Research 81, 3; 10.2460/ajvr.81.3.233
The ROC curves for the use of SF concentration (A and C) and HbA1c value (B and D) in classification of glycemic control (on the basis of clinical scores) as good versus moderate or poor (A and B) and poor versus good or moderate (C and D) in dogs with DM. The AUC was 0.69 (95% CI, 0.61 to 0.76), 0.66 (95% CI, 0.59 to 0.74), 0.75 (95% CI, 0.68 to 0.82), and 0.69 (95% CI, 0.60 to 0.78) for panels A, B, C, and D, respectively.
Citation: American Journal of Veterinary Research 81, 3; 10.2460/ajvr.81.3.233
The ROC curves for the use of SF concentration (A and C) and HbA1c value (B and D) in classification of glycemic control (on the basis of clinical scores) as good versus moderate or poor (A and B) and poor versus good or moderate (C and D) in dogs with DM. The AUC was 0.69 (95% CI, 0.61 to 0.76), 0.66 (95% CI, 0.59 to 0.74), 0.75 (95% CI, 0.68 to 0.82), and 0.69 (95% CI, 0.60 to 0.78) for panels A, B, C, and D, respectively.
Citation: American Journal of Veterinary Research 81, 3; 10.2460/ajvr.81.3.233
The ROC curve analyses for the use of SF concentrations and HbA1c values to differentiate poor glycemic control from moderate or good glycemic control revealed a significant (P < 0.001) AUC of 0.75 and 0.69, respectively. Use of an SF concentration of > 500 μmol/L to differentiate dogs with poor glycemic control from dogs with moderate or good glycemic control yielded a specificity of 85% and sensitivity of 45%. Use of an HbA1c value > 6.8% to differentiate dogs with poor glycemic control from dogs with moderate or good glycemic control yielded a specificity of 79% and sensitivity of 52%.
The classification of each glycemic control assessment was determined on the basis of clinical score and by use of established cutoff values for glycated proteins (Table 1). The number of misclassified assessments was determined on the basis of the clinical score and use of established cutoff values for glycated proteins. There were 36 of 200 (18%) assessments misclassified by use of cutoff values for SF concentrations, 41 of 200 (21%) assessments misclassified by use of cutoff values for HbA1c values, and 65 of 200 (33%) assessments misclassified by use of cutoff values for both SF concentrations and HbA1c values. The classification when considering the duration of insulin treatment (≤ 3 months or > 3 months) as well as the time from the last adjustment of the insulin dose was also determined (Table 2).
Results for 200 assessments of glycemic control in 46 dogs with DM classified on the basis of the level of control achieved (based on clinical score) and cutoff values for SF and HbA1c values as determined via ROC curve analysis.
| Analyte | Good | Moderate | Poor |
|---|---|---|---|
| SF (μmol/L) | |||
| < 400 | 55 | 24 | 8 |
| 400–500 | 25 | 21 | 21 |
| > 500 | 10 | 13 | 23 |
| HbA1c (%) | |||
| < 5.5 | 38 | 16 | 8 |
| 5.5–6.8 | 40 | 25 | 19 |
| > 6.8 | 12 | 17 | 25 |
Number (%) of assessments of glycemic control (n = 200) in 46 dogs with DM that were correctly classified or misclassified as indicated by SF and HbA1c values, grouped by various conditions of insulin treatment.
| SF | HbA1c | |||
|---|---|---|---|---|
| Insulin treatment | Correctly classified | Misclassified | Correctly classified | Misclassified |
| Receiving ≤ 3 mo (n = 106) | 50 (47) | 56 (53) | 47 (44) | 59 (56) |
| Receiving > 3 mo (n = 94) | 49 (52) | 45 (48) | 42 (44) | 52 (55) |
| Dose adjusted within past 3 mo (n = 168) | 81 (48) | 87 (52) | 78 (46) | 90 (54) |
| Dose stable > 3 mo (n = 32) | 18 (56) | 14 (44) | 16 (50) | 16 (50) |
Finally, the SF concentration could be used to correctly classify glycemic control in 99 of 200 (50%) assessments, and the HbA1c value could be used to correctly classify glycemic control in 88 of 200 (44%) assessments; these percentages did not differ significantly (P = 0.22).
Discussion
The colorimetric and immunoturbidimetric assays used in the present study for the quantification of SF and HbA1c values, respectively, in dogs were precise and accurate. The colorimetric nitroblue tetrazolium reduction method is extensively used in veterinary laboratories for measurement of SF concentrations; however, the analytic performance of this assay for samples from dogs has been evaluated in only a few studies.1,5 In the present study, the overall intra-assay CV was slightly higher than that reported for one study1 and similar to that reported for another study.5 In contrast to results for those studies, the study reported here revealed higher, but still acceptable, variability in SF concentrations. The interassay precision in the present study compared favorably with previously reported data.1,5 Linearity was deemed good, although the expected SF concentration was generally overestimated in all diluted samples, which was similar to results reported for another study.1 The reference interval for SF concentrations obtained in the study reported here was comparable to that reported previously when the same analytical method was used.3,5,31
The human and canine β-chains of hemoglobin are identical in the sequences of the last 5 amino acids32; the immunoturbidimetric method used in the present study was designed for humans and involved monoclonal antibodies directed against this region of the molecule; therefore, it could also be used for dogs. Compared with other studies11–13 in which the same analytic method was used, lower intra-assay and interassay CVs were obtained in the present study. Linearity was good, although the actual HbA1c value was generally underestimated in diluted samples, especially for values of HbA1c < 0.3 mg/dL. However, such low concentrations are not expected in canine samples, particularly those from diabetic dogs, so the clinical importance of this underestimation would be negligible. Compared with data reported in another study,13 results of the present study provided slightly better linearity.
The reference interval for HbA1c values obtained in the present study (1.6% to 4.5%) partially overlapped with results obtained in another study11 by use of the same method, even though the upper limit was higher. However, the inclusion criteria for healthy dogs in that other study11 were not reported, and a different immunoturbidimetric assay was used. In 2 other studies12,13 in which the immunoturbidimetric assay was used, the reference interval was narrower and the lower limit was higher, compared with results obtained in the study reported here. The reference interval for HbA1c values obtained in other studies by use of different analytic assays was higher33 or similar8,21,34 to results obtained in the present study. Because of this high variability, a reference interval should be determined at each laboratory.
The HbA1c values were significantly higher in anemic dogs than in healthy dogs, and no correlation was found between Hct and HbA1c values. These results were consistent with those in another study,21 in which HbA1c values of anemic dogs were significantly higher than those of the control group and there was no correlation between Hct and HbA1c values. In contrast, investigators in another study8 found significantly lower HbA1c values in anemic dogs than in healthy control dogs and a positive correlation between Hct and HbA1c values. Because glycation of hemoglobin occurs only as the erythrocytes circulate in the plasma, hemoglobin in older erythrocytes is more glycosylated, whereas in younger erythrocytes such as reticulocytes, hemoglobin is less glycosylated.27 In humans, any condition that shortens erythrocyte survival time or decreases mean erythrocyte age (eg, acute blood loss or hemolytic anemia) falsely lowers HbA1c test results regardless of the assay method.35 In contrast, iron-deficiency anemia reportedly increases test values,36 probably because the mean age of circulating erythrocytes is higher. In the present study, the median HbA1c value of dogs with regenerative anemia was lower than that of dogs with nonregenerative anemia, although these values did not differ significantly. The lack of a significant difference could have been attributable to the low number of dogs included in these 2 groups. However, it seems that dogs with nonregenerative anemia generally had higher HbA1c values. Hence, interpretation of HbA1c values in dogs should always take into consideration the Hct value and, when that value is low, the type of anemia.
Clinical score was significantly and inversely correlated with the SF concentration and HbA1c value; however, the correlation was weak to moderate for both but higher for the SF concentration than for the HbA1c value. This result suggested that the SF concentration might reflect glycemic control better than does the HbA1c value. In the study reported here, the clinical score was used as a reference method to classify glycemic control; however, most variables used in the clinical score reflected glycemic control for that particular day (BG curve) or a few days before measurement of blood glycated protein concentrations (clinical signs observed by the owner). For these reasons, a variable that is more indicative of recent glycemic control, such as the SF concentration, was more likely to be correlated with criteria used to classify glycemic control in our study than were glycosylated protein concentrations that reflected glycemic control over a longer period (HbA1c values). Moreover, the dose of insulin was often changed a few weeks before glycemic control was reassessed. Therefore, this brief period could have allowed alteration of glycemic control but not changes in HbA1c values because 2 weeks are required for HbA1c to respond to increases in BG concentrations.37 Potentially, the HbA1c value is a better indicator of glycemic status in dogs that are not receiving changes in insulin treatment. In humans, self-monitoring and the daily adjustment of insulin dose makes the use of a variable that reflects long-term glucose regulation more reliable. Some authors24 claim that a variable reflective of short-term glucose control (eg, SF concentration) would be better for monitoring the long-term diabetic stability of dogs with DM. In the present study, SF concentrations and HbA1c values were poorly correlated. Another study38 revealed a higher degree of correlation between SF and HbA1c values in dogs. The reason for the poor correlation in the present study is unknown but may be have been attributable to the increased sensitivity of SF concentrations, compared with that of HbA1c values, to temporary fluctuations in BG concentrations.39
Although the median SF concentration and HbA1c value differed significantly among the 3 clinical control groups in the study reported here, marked overlap was evident in values of the 2 glycated proteins among groups. Such overlap may have been the result of differences in the duration of good, moderate, or poor glycemic control before SF concentrations and HbA1c values were measured. Previous studies3,8 of the use of glycated proteins to monitor metabolic control in dogs with DM have shown that SF and HbA1c values differ significantly according to the degree of clinical control achieved; however, for both glycated proteins, there is substantial overlap among groups. In addition, in one of those studies,3 metabolic control was defined only on the basis of the fasting glucose concentration, whereas in the other study,8 the dog's clinical condition was also considered. Detection of glycated protein values within the reference interval in dogs with poorly controlled DM and high glycated protein values in dogs with well-controlled DM may suggest a delay between the change in mean BG concentration and the corresponding change in glycated protein values. Moreover, a recent study40 of the in vitro effect of hyperglycemia on plasma protein concentrations in dogs revealed a lower predisposition to glycation in samples from dogs versus humans. This might lead to an underestimation of the severity of hyperglycemia in dogs with DM that could explain the low accuracy of SF concentrations for the classification of glycemic control. On the other hand, any condition (ie, hypothyroidism) that leads to a decrease in protein turnover might cause prolonged exposure between glucose and plasma proteins that results in an increase in SF concentration and misclassification of well-controlled patients. Furthermore, dogs with recently adjusted insulin doses may have had an improvement in glycemic control and clinical signs that was not yet reflected by a decrease in the SF concentration.
Some authors have suggested the use of glycated proteins for monitoring diabetic dogs to clarify discrepancies between the medical history, physical examination findings, and serial BG concentrations, and they propose an HbA1c value < 6.5%8 or < 6.0%24 and an SF concentration < 450 μmol/L24 to identify dogs receiving insulin treatment that are well controlled. Results of the study reported here indicated that both glycated proteins had poor ability to differentiate among the 3 groups of clinical control as defined on the basis of the clinical score. In 99 of 200 (50%) assessments, glycemic control was misclassified by use of SF concentrations. In particular, 77 (39%) assessments were misclassified by use of at least one of the glycated proteins, whereas 65 (33%) assessments were misclassified by use of both the glycated proteins. Hence, the cutoff values had limited clinical usefulness, and use of the glycated proteins alone to determine glycemic control cannot be recommended.
For the same reasons, in dogs with DM in which repeated BG testing is not a practical option, assessment of glycemic stability by measurement of the glycated protein values over a period of weeks to months might not be appropriate. Because HbA1c has a long half-life (2 to 3 months), and 94 glycemic control assessments in the present study were for diabetic dogs receiving insulin treatment for ≤ 3 months, each assessment was further classified by separating the dogs receiving insulin treatment for > 3 months from dogs receiving insulin treatment for ≤ 3 months as well as by separating dogs that had an insulin dose adjustment within the past 3 months from dogs that had received a stable insulin dose for at least 3 months. However, the percentage of dogs correctly classified by use of the 2 glycated proteins was similar in these subgroups.
Taken together, findings of the present study confirmed that glycated protein concentrations in blood or serum should not be used alone to assess control of DM in dogs and raised questions as to whether these values have any clinical utility in the monitoring of DM. Although trends in HbA1c and SF values in individual patients during ongoing treatment were not evaluated in the present study, several studies8,28,41,42 have shown that SF and HbA1c values decrease substantially in dogs with DM during insulin treatment. Moreover, humans differ in their ability to glycate hemoglobin.43 Therefore, monitoring individual dogs on the basis of their previous values rather than by use of population-derived reference intervals might be more appropriate. Establishing a baseline serum or blood glycated protein concentration for an individual patient and then monitoring patterns rather than absolute changes associated with alterations in control of DM might be the most appropriate use of glycated protein measurements in diabetic dogs.
The present study had some limitations. The definition of clinical control was based on the results of a clinical history obtained from the owner, body weight, and assessment of BG curves. Although owners of the diabetic dogs had been instructed to evaluate the clinical signs of their dogs, they could have underestimated or overestimated some signs or failed to recognize the signs of DM. Furthermore, all BG curves were performed in a hospital setting, which might have affected their accuracy. Another limitation was that some dogs may have had concurrent diseases (eg, hypothyroidism, hyperlipidemia, hypoalbuminemia, anemia, or hyperadrenocorticism) that could have influenced the SF and HbA1c values as well as interpretation of the clinical signs (eg, presence of polyuria and polydipsia in dogs with hyperadrenocorticism). However, these dogs were included in the study to mirror the population of dogs with DM, despite evidence of concurrent disease that may have affected the obtained SF and HbA1c values. Moreover, as mentioned previously, because the dose of insulin had often been changed shortly before glycemic control assessment, the dogs with a dose adjustment possibly had a change in clinical score that was not yet reflected by the glycated protein values.
The assays used in the present study were reliable methods with good analytic performance for measurement of SF concentrations and HbA1c values in dogs. In particular, HbA1c appeared to be quite stable when blood samples were stored under various storage conditions and could represent an alternative to measurement of SF concentrations in diabetic dogs. Use of SF concentrations and HbA1c values yielded results that suggested a similar and poor ability to classify glycemic control; thus, these analytes should not be used alone in the assessment of glycemic control or adjustment of insulin dose.
Acknowledgments
No extrainstitutional funding or support was obtained for this study. The authors declare that there were no conflicts of interest.
Presented in part as an abstract at the 28th European College of Veterinary Internal Medicine-Companion Animal Congress, Rotterdam, Netherlands, September 2018.
ABBREVIATIONS
| AUC | Area under the curve |
| BG | Blood glucose |
| CI | Confidence interval |
| CV | Coefficient of variation |
| DM | Diabetes mellitus |
| HbA1c | Glycated hemoglobin |
| ROC | Receiver operating characteristic |
| SF | Serum fructosamine |
Footnotes
Fructosamine 17350H, Sentinel Diagnostic, Milano, Italy.
HbA1c, B00389, Beckman Coulter Inc, Brea, Calif.
AU 480, Beckman Coulter Inc, Brea, Calif.
ADVIA 2120, Siemens Healthcare Diagnostics, Tarrytown, NY.
autoRetic, Siemens Healthcare Diagnostics, Tarrytown, NY.
CQ2 extendSURE level 2, B12396, Canterbury Scientific Ltd, Christchurch, New Zealand.
Gluco Calea, WellionVet, Med Trust, Marz, Austria.
Optium Xceed, Abbott Laboratories, Witney, England.
Prism, version 7.0d, GraphPad Software Inc, San Diego, Calif.
Microsoft Excel, version 2016, Microsoft Corp, Redmond, Wash.
Caninsulin, MSD, Boxmeer, Netherlands.
Humulin I, Eli Lilly Italia S.p.A., Sesto Fiorentino, Italy.
Vetoryl, Dechra Pharmaceuticals, Northwich, England.
Canitroid, Dechra Pharmaceuticals, Northwich, England.
References
1. Jensen AL. Serum fructosamine in canine diabetes mellitus. An initial study. Vet Res Commun 1992;16:1–9.
2. Kawamoto M, Kaneko JJ, Heusner AA, et al. Relation of fructosamine to serum protein, albumin and glucose concentrations in healthy and diabetic dogs. Am J Vet Res 1992;53:851–855.
3. Reusch CE, Liehs MR, Hoyer M, et al. Fructosamine. A new parameter for diagnosis and metabolic control in diabetic dogs and cats. J Vet Intern Med 1993;7:177–182.
4. Mahaffey EA, Buonanno AM, Cornelius LM. Glycosylated albumin and serum protein in diabetic dogs. Am J Vet Res 1984;45:2126–2128.
5. Coppo JA, Coppo NB. Serum fructosamine: a reference interval for a heterogeneous canine population. Vet Res Commun 1997;21:471–476.
6. Reusch CE. Diabetic monitoring. In: Bonagura JD, Twedt DC, eds. Kirk's current veterinary therapy XV. St Louis: Elsevier Saunders, 2014;193–199.
7. Smith JE, Wood PA, Moore K. Evaluation of a colorimetric method for canine glycosylated hemoglobin. Am J Vet Res 1982;43:700–701.
8. Elliott DA, Nelson RW, Feldman EC, et al. Glycosylated hemoglobin concentrations in the blood of healthy dogs and dogs with naturally developing diabetes mellitus, pancreatic beta-cell neoplasia, hyperadrenocorticism, and anemia. J Am Vet Med Assoc 1997;211:723–727.
9. Mahaffey EA, Cornelius LM. Evaluation of a commercial kit for measurement of glycosylated hemoglobin in canine blood. Vet Clin Pathol 1981;10:21–24.
10. Hasegawa S, Sako T, Takemura N, et al. Glycated hemoglobin fractions in normal and diabetic dogs measured by high-performance liquid chromatography. J Vet Med Sci 1991;53:65–68.
11. Marca MC, Loste A. Glycosylated haemoglobin assay of canine blood samples. J Small Anim Pract 2000;41:189–192.
12. Lien YH, Huang HP. Glycosylated hemoglobin concentrations in dogs with hyperadrenocorticism and/or diabetes mellitus compared to clinically healthy dogs. J Vet Clin Sci 2009;2:37–41.
13. Goemans AF, Spence SJ, Ramsey JK. Validation and determination of a reference interval for canine HbA1c using an immunoturbidimetric assay. Vet Clin Pathol 2017;46:227–237.
14. Oikonomidis IL, Tsouloufi TK, Soubasis N, et al. Validation, reference intervals and overlap performance of a new commercially available automated capillary electrophoresis assay for the determination of the major fraction of glycated haemoglobin (HbA1c) in dogs. Vet J 2018;234:48–54.
15. Davison LJ, Podd SL, Ristic JME, et al. Evaluation of two point-of-care analysers for measurement of fructosamine or haemoglobin A1c in dogs. J Small Anim Pract 2002;43:526–532.
16. Catchpole B, Mountford S, Barabas S, et al. Evaluation of a disposable device for the measurement of haemoglobin A1c in dogs. Vet Rec 2008;162:47–49.
17. Reusch CE, Haberer B. Evaluation of fructosamine in dogs and cats with hypo- or hyperproteinaemia, azotemia, hyperlipidaemia and hyperbilirubinaemia. Vet Rec 2001;148:370–376.
18. Reusch CE, Gerber B, Boretti FS. Serum fructosamine concentrations in dogs with hypothyroidism. Vet Res Commun 2002;26:531–536.
19. Zeugswetter F, Kleiter M, Wolfesberger B, et al. Elevated fructosamine concentrations caused by IgA paraproteinemia in two dogs. J Vet Sci 2010;11:359–361.
20. Little RR, Rohlfing CL. The long and winding road to optimal HbA1c measurement. Clin Chim Acta 2013;418:63–71.
21. Haberer B, Reusch CE. Glycated haemoglobin in various pathological conditions: investigations based on a new, fully automated method. J Small Anim Pract 1998;39:510–517.
22. Radin MS. Pitfalls in hemoglobin A1c measurement: when results may be misleading. J Gen Intern Med 2014;29:388–394.
23. Welsh KJ, Kirkman MS, Sacks DB. Role of glycated proteins in the diagnosis and management of diabetes: research gaps and future directions. Diabetes Care 2016;39:1299–1306.
24. Nelson RW. Canine diabetes mellitus. In: Feldman EC, Nelson RW, Reusch CE, et al, eds. Canine and feline endocrinology. 4th ed. St Louis: Elsevier Saunders, 2015;213–253.
25. Fracassi F. Canine diabetes mellitus. In: Ettinger SJ, Feldman EC, Cäté E, eds. Textbook of veterinary internal medicine. 8th ed. St Louis: Elsevier Saunders, 2017;1767–1781.
26. American Diabetes Association. Standards of medical care for patients with diabetes. Diabetes Care 1989;12:365–368.
27. Saudek CD, Brick JC. The clinical use of hemoglobin A1c. J Diabetes Sci Technol 2009;3:629–634.
28. Fracassi F, Linari G, Del Baldo F, et al. Comparison of lente insulin and NPH insulin therapy for the treatment of newly diagnosed diabetic dogs: a randomised study. Vet Rec 2018;183:262.
29. Del Baldo F, Malerba E, Corradini S, et al. Evaluation of one portable blood glucose meter and one portable glucoseketones meter in dogs. Research Communications of the 27th ECVIM-CA Congress. Malta, September 14 to 16, 2017. J Vet Intern Med 2017;32:525–609.
30. Fracassi F, Hadar GS, Pietra M, et al. Assessment of two portable blood glucose meters for use in cats and dogs. J Vet Clin Sci 2009;2:108–121.
31. Jensen AL, Aaes H. Reference interval and critical difference for canine serum fructosamine concentration. Vet Res Commun 1992;16:317–325.
32. Jensen AL. Glycated blood proteins in canine diabetes mellitus. Vet Rec 1995;137:401–405.
33. Easley JR. Glycosylated hemoglobin in dogs: precision, stability, and diagnostic utility. Vet Clin Pathol 1986;15:12–15.
34. Hooghuis H, Rodriguez M, Castano M. Ion-exchange microchromatography and thiobarbituric acid colorimetry for the measurement of canine glycated hemoglobins. Vet Clin Pathol 1994;23:110–116.
35. Goldstein DE, Little RR, Lorenz RA, et al. Tests of glycemia in diabetes. Diabetes Care 1995;18:896–909.
36. Tarim O, Kucukerdogan A, Gunay U, et al. Effects of iron deficiency anemia on hemoglobin A1c in type 1 diabetes mellitus. Pediatr Int 1999;41:357–362.
37. Wood PA, Smith JE. Elevation rate of glycosylated hemoglobins in dogs after induction of experimental diabetes mellitus. Metabolism 1982;31:906–909.
38. Loste A, Marca MC. Fructosamine and glycated hemoglobin in the assessment of glycaemic control in dogs. Vet Res 2001;32:55–62.
39. Akol KG, Waddle JR, Wilding PJ. Glycosylated hemoglobin and fructosamine in diabetic and nondiabetic cats. J Am Anim Hosp Assoc 1992;28:227–231.
40. Rivera-Velez SM, Hwang J, Navas J, et al. Identification of differences in the formation of plasma glycated proteins between dogs and humans under diabetes-like glucose concentration conditions. Int J Biol Macromol 2019;123:1197–1203.
41. Maggiore AD, Nelson RW, Dennis J, et al. Efficacy of protamine-zinc recombinant human insulin for controlling hyperglycemia in dogs with diabetes mellitus. J Vet Intern Med 2012;26:109–115.
42. Fracassi F, Corradini S, Hafner M, et al. Detemir insulin for the treatment of diabetes mellitus in dogs. J Am Vet Med Assoc 2015;247:73–78.
43. Hudson PR, Child DF, Jones H, et al. Differences in rates of glycation (glycation index) may significantly affect individual HbA1c results in type 1 diabetes. Ann Clin Biochem 1999;36:451–459.
