Materials and Methods

Dogs and samples—Dogs were of different sexes, breeds, and ages and were evaluated at private clinics in southern Spain for routine checkups between February and July of 2008. The dogs had no remarkable findings via physical, clinical (with the exception of obesity), hematologic, or serum biochemical evaluations. Assessment of the nutritional condition of dogs was performed by use of a 5-point BCS following a weight guide chart based on the characteristics of ribs, tail base, side view, and overhead view. Dogs were classified as follows: 1/5, very thin; 2/5, thin; 3/5, ideal weight; 4/5, overweight; and 5/5, obese.¹³ Serum samples with expected high and low leptin concentrations were used for analytic validation of the assays. Low leptin concentration specimens were obtained from 7 thin or ideal-weight dogs (BCS, 2/5 or 3/5), and high leptin concentration specimens were obtained from 7 overweight or obese dogs (BCS, 4/5 or 5/5). None of the samples were of sufficient volume for all analyses required for complete analytic validation of all 5 ELISAs, so samples with expected low leptin concentrations were pooled and samples with expected high leptin concentrations were pooled.

After food was withheld overnight for at least 12 hours, blood samples were collected in the morning via saphenous venipuncture with a 21-gauge needle and placed into 5-mL tubes containing a clotting accelerator.^a No later than 15 minutes after collection, samples were centrifuged at 2,000 × g for 10 minutes at 20°C to obtain serum, and serum was stored in aliquots in plastic vials at −20°C until analysis. Duration of storage was < 6 months. On the day of analysis, samples were thawed (1 hour at 20°C) and thoroughly vortexed prior to leptin measurement.

Leptin analysis—Five commercially available ELISAs developed for serum leptin determinations in dogs (A^b and B^c), humans (C^d and D^e), and mice or rats (E^f) were evaluated for leptin determination in the canine serum samples (Appendix). All spectrophotometric measurements were performed in a microtiter plate reader.^g

Analytic validation—To evaluate the existence of reactivity between assay antibodies and leptin in canine sera, 3 pools prepared from overweight-obese dogs' serum samples were analyzed in duplicate by use of all 5 ELISAs. Because purified canine leptin was not commercially available, it could not be used for test reactivity, so it was supposed that a positive analytic signal corresponded to leptin present in the canine sera. The ELISAs that yielded a positive signal were further tested for precision, accuracy, linearity under dilution, spiking recovery, and limit of quantification.

For evaluation of precision, intra-assay CV was calculated after analysis of 2 serum pools with high and low leptin concentrations, 5 times in a single assay run. Interassay CV was determined by analysis of the same serum pools in 5 runs performed on different days.

Accuracy was indirectly evaluated by investigation of linearity under dilution and spiking recovery because neither a certified species-specific reference material nor a reference method was available.¹⁴ To assess linearity under dilution, 2 canine serum pools with high leptin concentrations were serially diluted. In assay A, pools were diluted with the corresponding sample diluent provided with the assay, and in assay B, saline solution (NaCl, 9 g/L) was used because no diluent was provided by the manufacturer.

To evaluate the ability of the assays to recover the amount of analyte added to baseline serum samples, a spiking recovery was performed. Because purified canine leptin was not commercially available, this test was based on the addition of a calibrator.¹⁵ For this purpose, 2 samples were divided into 4 aliquots and each aliquot was mixed with an equal amount of calibrator (calibrator included in each assay kit [A or B] was used for the corresponding assay) at 3 concentrations or dilution buffer. Test recovery (percentage) was calculated for each dilution for comparison of expected versus measured leptin concentrations.

Limit of quantification was calculated as the amount of analyte that could be measured with an intra-assay variation < 15%.^16,17 Six dilutions (100%, 75%, 50%, 37.5%, 18.75%, and 9.38%) of 2 serum pools (1 for each assay tested) were prepared and analyzed 5 times. The CV of each dilution was calculated and plotted as a function of leptin concentration.

All pools used for repetitive analysis were frozen in aliquots, and only vials needed for each run were used to avoid possible changes caused by repetitive thawing and freezing.

Comparison of assays with positive reactivity and ability of assays to differentiate between thin–ideal weight and overweight-obese dogs—Samples from 45 client-owned dogs were included for method comparison and to determine whether leptin ELISAs were able to differentiate between thin–ideal weight and overweight-obese dogs. Thirty-seven mixed-breed dogs were overweight or obese (20 females [3 neutered] and 17 males [4 neutered]; BCS, 4/5 or 5/5; body weight range, 3.0 to 48.5 kg; age range, 1.5 to 11 years), and 8 were thin–ideal weight, healthy mixed-breed dogs (4 females [1 neutered] and 4 males; BCS, 2/5 or 3/5; body weight range, 5 to 14.1 kg; age range, 1.0 to 6.0 years).

Statistical analysis—Arithmetic means, medians, and intra- and interassay CVs were calculated by use of routine descriptive statistical procedures and software.^h Comparison of leptin concentrations between thin–ideal weight and overweight-obese dogs was determined by use of the Mann-Whitney test because data did not pass the Kolmogorov-Smirnov normality test. Visual inspection of results of linear regression analysis comparing measured and expected concentrations of leptin was used to evaluate the linearity under dilution. Bland-Altman plots were used to compare the results obtained with the assays that had reactivity with canine leptin. The 95% limits of agreement were calculated as mean difference ± 1.96 × SD of the difference. For all tests, a value of P < 0.05 was considered significant.

Results

Assay characteristics—Mean results obtained from 3 canine serum pools from overweight-obese dogs that were analyzed to evaluate the existence of reactivity between assay antibodies and leptin, along with mean values of blank samples from all kits evaluated, were determined (Table 1). Because a positive result that could be differentiated from that obtained with the blank was not detected by use of any of the human or mouse or rat leptin ELISAs, the rest of the study was focused on evaluation and validation of assays A and B.

Assays A and B had intra-assay CVs < 6.1% and 14.0% and interassay CVs < 7.6% and 13.7%, respectively (Table 2). Linearity under dilution of 2 canine serum pools for assays A and B was determined (Figures 1 and 2). In assay A, dilution of pools with different leptin concentrations resulted in linear regression equations with correlation coefficients of 0.99 and 0.99; in assay B, these correlation coefficients were 0.97 and 0.95. Recovery between observed and expected leptin concentrations ranged from 77% to 101% for assay A and from 67% to 125% for assay B (Table 3). Limits of quantification were 3.6 μg/L for assay A and 11.6 μg/L for assay B (Figure 3).

View Full Size

Linearity under dilution of 2 canine serum pools (circles and squares) with different leptin concentrations measured by use of an ELISA (assay A).

Citation: American Journal of Veterinary Research 72, 2; 10.2460/ajvr.72.2.169

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Linearity under dilution of 2 canine serum pools (circles and squares) with different leptin concentrations measured by use of an ELISA (assay B).

Citation: American Journal of Veterinary Research 72, 2; 10.2460/ajvr.72.2.169

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Limits of quantification for 2 ELISAs used to measure canine leptin concentrations. Different dilutions of 2 serum pools were assayed with assay A and B, and the CVs were plotted as a function of leptin concentration. Dotted line indicates 15% CV which is considered the acceptable limit of quantification.

Citation: American Journal of Veterinary Research 72, 2; 10.2460/ajvr.72.2.169

• Download Figure
• Download figure as PowerPoint slide

Comparison of canine leptin ELISAs and ability of the assays to differentiate between thin–ideal weight and overweight-obese dogs—Seven of 8 serum samples collected from thin–ideal weight dogs did not yield a result when analyzed with assay A (leptin concentrations of these samples and all samples with results less than the limit of quantification were set to 0 μg/L for further statistical analysis; however, values are reported as less than or equal to the limit of quantification). For the 45 clinical samples, ranges of leptin concentrations obtained with assays A and B were ≤ 3.6 to 45.9 μg/L and ≤ 11.6 to 103.8 μg/L, respectively. A Bland-Altman plot of comparison between the methods revealed a proportional error indicated by differences in leptin concentrations increasing relative to the increase in mean value (Figure 4).

View Full Size

Bland-Altman plot illustrating the 95% limits of agreement for leptin concentrations determined by use of 2 ELISAs (A and B). Upper and lower horizontal dashed lines represent 95% limits of agreement, and the middle horizontal dotted-dashed line represents mean difference (bias).

Citation: American Journal of Veterinary Research 72, 2; 10.2460/ajvr.72.2.169

• Download Figure
• Download figure as PowerPoint slide

When assay A was used, samples collected from overweight-obese dogs (median, 7.2 μg/L; 10th to 90th percentile range, 4.2 to 25.0 μg/L) had significantly (P < 0.001) higher leptin concentrations than those collected from thin–ideal weight dogs (median, ≤ 3.6 μg/L; 10th to 90th percentile range, ≤ 3.6 to 5.4 μg/L). In contrast, no significant (P = 0.89) difference was observed in leptin concentration when serum samples from overweight-obese (median, 28.1 μg/L; 10th to 90th percentile range, 13.7 to 92.3 μg/L) and thin–ideal weight dogs (median, 27.6 μg/L; 10th to 90th percentile range, 20.8 to 81.0 μg/L) were analyzed with assay B.

Discussion

Leptin concentration is high in obese dogs and can be used in the study of obesity and obesity-associated diseases in this species.^5,6 However, leptin concentration is still not widely used in canine practice, perhaps because, to the authors' knowledge, no validation studies for use in dogs have been published for any leptin ELISA methods that are presently commercially available. Validation studies are needed to ensure that analytic methods can successfully measure the analyte and provide reproducible and accurate results that are useful to discriminate between healthy and diseased animals.¹⁸

In the study in dogs reported here, we tested 5 ELISA kits for leptin measurement: 2 for use in dogs, 2 for use in humans, and 1 for use in mice or rats. Despite an 82% and 79% homology in amino acid sequence between canine and human or mouse leptin, respectively, only canine-specific ELISAs (A and B) yielded a positive reaction with canine serum. These results could suggest that antibodies of the different kits recognize leptin epitopes located in regions with a high heterogeneity in the protein sequence between species.¹⁹ It must be pointed out that canine serum with potentially high leptin concentrations was used instead of purified canine leptin to evaluate the presence of reactivity between assay antibodies and leptin; therefore, it was supposed that the analytic signal obtained did indicate the presence of leptin because antibodies of the positive reacting kits were canine leptin specific.

The CVs reported by assay A manufacturers (5.9% and 6.7% at mean concentrations of 3.1 and 20.8 μg/L, respectively) were slightly higher than those observed in the present study. Manufacturers of assay B do not provide such data with the ELISA kit. However, obtained results indicated that assay A was more precise in measuring leptin in canine serum samples than assay B, and assay B imprecision was near 15%, which is considered the maximum acceptable value for analytic methods.²⁰

The accuracy of A and B ELISAs was indirectly evaluated by use of linearity under dilution and spiking recovery.^14,15 Regression analysis revealed that both assays yielded linear results when measuring leptin in different dilutions of canine serum samples. The recovery study revealed that assay A slightly underestimated serum leptin concentrations because most recovery values were < 100%. The low recovery found at the expected value of 2.9 μg/L could be attributable to the fact that it was less than the limit of quantification of the assay (3.6 μg/L). Leptin values obtained with assay B were highly under- or overestimated (reaching up to 33% of the expected value), compared with expected values.

When assay A was used, overweight-obese dogs had higher plasma leptin concentrations, compared with thin–ideal weight dogs, which was similar to previous reports.^5–7 In the present study, the median leptin concentration in overweight-obese dogs (7.2 μg/L) obtained with assay A was lower than reported by Ishioka et al^6,7 in dogs with a BCS of 5/5 (12.3 ± 1.5 μg/L and 12.8 ± 0.8 μg/L, respectively). However, no comparison could be made in thin–ideal weight dogs between assay A and reports from Ishioka et al^6,7 because median values obtained in our assay were less than the limit of quantification of the method.

No significant difference in leptin concentrations was detected between overweight-obese and thin–ideal weight dogs when analyzed with assay B. Factors that might explain these results were the high imprecision and poor performance in recovery studies. In addition, this assay's high limit of quantification reflected a limited ability to measure low concentrations of leptin.

Of the 5 leptin ELISAs evaluated in the present study, 1 canine-specific ELISA had adequate analytic characteristics of precision, linearity, recovery, and ability to discriminate between high and low leptin concentrations corresponding to overweight-obese and thin–ideal weight dogs, respectively. Although these results should be interpreted with caution because no purified canine leptin was available for cross-reactivity studies and variations in reactivity may be obtained with different batches of antibodies used in the canine kit, this assay could be used for canine leptin determination. Nevertheless, it would be desirable to decrease its limit of quantification to permit better detection of low concentrations of leptin. In addition, results of this study stress the importance of performing validation studies on every analytic kit before routine use in a veterinary laboratory.