Analytical validation of a novel agglutination immunoassay for the quantification of cystatin B in canine and feline urine

Abstract

Urinary cystatin B (uCysB) is a biomarker of kidney injury in dogs and cats. A high-throughput agglutination immunoassay (Idexx Laboratories) was developed for widespread commercial availability of uCysB testing in a reference laboratory setting. We evaluated immunoassay performance and included analyses of precision, accuracy, linearity, interference, analytical specificity, lot-to-lot variation, and stability. CVs from precision studies on the range of 50–500 ng/mL were 0.38–2.53% (canine) and 0.44–3.5% (feline) for within-run precision, and 1.49–5.09% (canine) and 0.65–5.05% (feline) for between-run precision. Accuracy was measured by recovery percentage and was 89–101% (canine) and 91–112% (feline). Amoxicillin, ciprofloxacin, low concentrations of doxycycline, bilirubin, glucose, ketones, RBCs, hemoglobin, cloudiness, lipids, protein, and pH did not affect results. Urinary cystatin A did not cross-react with the uCysB immunoassay. Results of lot-to-lot linear regressions were 0.90–1.07 (slopes) and 0.97–1.00 (coefficient of determination). One or more freeze-thaw cycles and storage at 30°C impacted the immunoassay stability of canine samples but not feline samples under the same conditions. Our results validate this novel agglutination immunoassay for accurate and precise measurement of uCysB in canine and feline urine samples. For optimal immunoassay performance, samples should be kept at 4°C for a maximum of 1 wk. Our uCysB immunoassay is a useful and practical tool to be used in assessing kidney injury in canine and feline patients in the clinical setting.

Cystatin B (CysB) is an 11-kD cysteine protease inhibitor found ubiquitously in many cell types. Due to its intracellular location, CysB is found in only trace amounts in the serum and urine of healthy people.^1,14 In people, detection of CysB in urine was first shown to be a biomarker of urothelial carcinoma, with elevated urinary concentrations correlated with increasing tumor grade, stage, and shorter time to disease recurrence and progression.^7,12 In dogs and cats, detection of urinary CysB (uCysB) has emerged as a promising biomarker of active kidney injury due to presumptive tubular epithelial cell rupture and release of CysB into the urine.^3,10,23 Dogs with renal tubular injury due to envenomation by the European adder had elevated uCysB concentrations compared to control groups of healthy dogs,^9,11 and uCysB concentrations were correlated with snakebite severity score. ¹¹ Clinical utility of CysB in detecting kidney injury in dogs with acute canine monocytic ehrlichiosis, ¹³ undergoing cardiac surgery with cardiopulmonary bypass, ²¹ and with spontaneous acute kidney injury ⁸ has also been established. Percent homology (Clustal Omega ²⁰ ) with human CysB was 81.6% for dogs and 79.6% for cats.

Detection of uCysB in the setting of canine chronic kidney disease (CKD) has been shown to have clinical utility in predicting disease progression. In a study of 20 dogs with International Renal Interest Society stage 1 CKD, uCysB differentiated stable (24 ng/mL) from progressive (212 ng/mL) disease and clinically healthy dogs (17 ng/mL). ¹⁹ In cats, uCysB has been reported to be higher in cats with acute kidney injury (AKI; 1,050 ng/mL) compared with cats with CKD (112 ng/mL) or controls (22 ng/mL). ³ Diagnostic guidelines from Idexx Laboratories are that uCysB <100 ng/mL suggests a decreased potential of kidney injury, and that uCysB ≥100 ng/mL suggests an increased risk of kidney injury (https://www.idexx.com). The availability of uCysB in a commercial laboratory setting enables exploration of the utility of this biomarker in additional clinical contexts. Our goal was to provide analytical validation of a novel agglutination immunoassay for the quantification of uCysB in canine and feline urine.

Materials and methods

Sample collection

Feline and canine urine samples were obtained from Idexx reference laboratories, shipped on cold packs, and stored at 4°C unless otherwise noted.

Measurement of cystatin B

Our direct (non-competitive) immunoassay is a particle-assisted turbidimetric assay designed to quantitatively determine CysB concentrations in feline and canine urine samples measured on Beckman AU analyzers, and it was developed and manufactured by Idexx Laboratories. When antibody-coated latex particles (Bangs) are incubated at 37°C in quartz cuvettes in buffer in the presence of measurand, agglutination progresses. The Beckman instrument measures light scattering over time of incubation, and the difference between the end-point and the time of particle addition is the analyzer output.

The immunoassay protocol includes 2 controls (100, 400 ng/mL) and 6 calibrator solutions (0, 50, 125, 250, 350, 550 ng/mL). The calibrator solution concentrations were defined to cover the dynamic range of 50–500 ng/mL. The calibration and control solutions are a protein-containing buffer spiked with recombinant CysB. Each lot of the recombinant CysB is analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and size exclusion chromatography to establish purity >95%, and its concentration is determined (Nanodrop spectrophotometer; ThermoFisher) to measure absorbance at 280 nm in triplicate.

To perform the immunoassay, the calibrators, controls, and urine samples are run undiluted. The 6 calibration solutions are run weekly before the controls, and their optical densities fit with a spline trend line. The equation of the spline trend line is the calibration curve used to interpolate uCysB concentrations from the optical density of each control and patient sample. The controls provide daily internal QC and are run at the start of every shift, and before the patient urine sample runs; the control solutions must fall within the QC limits.

Analytical validation methods

Methods and requirements used to perform the analytical validation testing were based on industry standard guidance documents published by the United States Food and Drug Administration (FDA), ²² Clinical and Laboratory Standards Institute (CLSI),^5,6 and the American Society for Veterinary Clinical Pathology. ² Data analysis was performed using Excel 2007 (Office 365, v.2406; Microsoft) and JMP 17 software (v.17; SAS Institute). Each sample was tested in singlicate unless specified otherwise.

The immunoassay performance was evaluated for precision, accuracy, linearity, interference, analytical specificity, lot-to-lot variation, and stability analysis (Suppl. Fig. 1). Determination of the clinical sensitivity, clinical specificity, and positive and negative predictive value experiments was out of the scope of our validation study.

The QC limits were generated as 3 SDs from 620 and 640 replicates of 2 QC fluids (100 and 400 ng/mL of CysB), run across 15 locations and over 5 d with material from reagent lot 1. The total error observed (TEo) for each of the concentrations tested on the QC validation study was within the pre-defined total error allowable (TEa) CI. The TEa used for this immunoassay was defined internally and based on results of uCysB biological variation in healthy dogs and cats, clinical study data, and expert opinion. Based on the nature of the biomarker and its clinical utility, the derived TEa was marker concentration-range dependent and set to ensure high fidelity of results relative to clinical interpretations and to ensure that no >5% of results fall on the incorrect side (false-negative or false-positive) of 100 ng/mL due to analytical variation.

Precision and stability experiments used samples or pools with known native uCysB concentration. Accuracy, linearity, interference, analytical specificity, and lot-to-lot variation experiments used samples, pools, and aliquots spiked with different levels of recombinant CysB. Gravimetric quantification defined the CysB concentration of spiked samples, pools, and aliquots, and it was used as the reference method for accuracy, pH interference, and analytical specificity experiments. For simplicity, concentrations obtained by gravimetric quantification are referred to as spiked CysB concentrations. The gravimetric quantification was chosen as the reference method because CysB is a new biomarker, and no concentrated standard or validated assay was available for use as a reference method. Moreover, the actual amount used for spiking cannot serve as a reference method, as accurately weighing such small quantities presents significant challenges.

Within- and between-run precision testing was performed using native canine and feline urine pools prepared by screening urine samples and pooling according to their uCysB concentration to create 3 (75, 250, 450 ng/mL) or 4 (50, 100, 250, 450 ng/mL) native pools across the reportable range (50–500 ng/mL). Three reagent lots and 7 native pools per species were used for the precision analysis. Native pools 1–3 were evaluated with reagent lot 1, and native pools 4–7 were evaluated with reagent lots 2 and 3 (Suppl. Fig. 1). For within-run testing, 20 replicates were analyzed per reagent lot in a single day. For concentrations <100 ng/mL, the CV requirement was 15%, and for concentrations ≥100 ng/mL, the CV requirement was 10%. These requirements were defined based on U.S. FDA industry standard guidance ²² and well-accepted industry practices.

The within-run precision was defined by the CV of the uCysB concentration in the 20 replicates. For between-run testing, we evaluated pool replicates (5–23 replicates/d) across 5 or 10 d and 2 runs/d. Due to the unbalanced number of replicates per day, we considered only the results of the first 6 replicates per pool/d (3 results from the morning run and 3 results from the afternoon run, when available), and the remaining replicates were censored to perform the statistical analysis. The between-run precision was defined as the uCysB CV, considering each replicate as a data point. Reagent lot 1 was evaluated over 20 runs and 59 replicates (10 d, 2 runs/d, 19 runs with 3 replicates, and 1 run with 2 replicates), and reagent lots 2 and 3 were evaluated over 10 runs and 27 replicates (5 d, 2 runs/d, 7 runs with 3 replicates, and 3 runs with 2 replicates).

Accuracy was evaluated by recovery percentage (recovery %) using native urine pools of canine (15 pools with lot 1; 9 pools on lots 2 and 3) and feline (15 pools on lot 1; 10 pools on lots 2 and 3) samples. Each pool was divided into 4 aliquots and spiked with different levels of recombinant CysB. The different levels of CysB were designed to cover the immunoassay dynamic range (50–500 ng/mL) and considering that 100 ng/mL is the threshold for clinical decision-making. Spiked amounts were targeted at 50, 100, 250, and 500 ng/mL. The spiked pool’s final concentration did not reach the target amount due to technical limitations on weighing, such as the small amounts of CysB during the spiking procedure. The immunoassay concentration of the 4 replicates per spiked CysB level per reagent lot were averaged into a single value. The recovery % per spiked level per reagent lot was calculated as follows:

$$\mathrm{recovery}\%=\frac{\begin{array}{l}\mathrm{immunoassay}\mathrm{concentrations}\\ \left(\mathrm{turbidimetric}\right)\end{array}}{\begin{array}{l}\mathrm{spiked}\mathrm{concentration}\\ \left(\mathrm{measured}\mathrm{by}\mathrm{gravimetry}\right)\end{array}}\times 100\%$$

The TEo was calculated as 2CV + bias using the maximum CV and bias observed on concentrations within the immunoassay dynamic range (50–500 ng/mL) during the precision (between-run) and accuracy analyses. We used the maximum value because the CysB concentration of each experiment did not align 100%. We excluded the SD and CV of concentrations that were out of the immunoassay reportable range (50–500 ng/mL). This value was then compared to the smaller TEa (i.e., 100 ng/mL), the suggested threshold for clinical decision-making.

The linearity of the immunoassay across the reportable range (50–500 ng/mL) for the feline and canine immunoassay was evaluated using serial cross-dilutions of a spiked pool (5,000 ng CysB/mL) by a native pool (30 ng CysB/mL) resulting in 5 canine pool dilutions (54, 133, 232, 330, 429 ng/mL) and 4 feline pool dilutions (70, 208, 380, 552 ng/mL). Four replicates of each dilution were evaluated using a single reagent lot and averaged. The average of the CysB concentration obtained per dilution was plotted against the original CysB dilution concentration. The least squares regression line was used to calculate the slope and coefficient of determination (R²) between the observed (averaged) CysB value and the expected value according to the dilution protocol. According to industry guidelines for the linearity test,^2,5,6,22 the slope must be between 0.8 and 1.2, and R² >0.9.

Commonly occurring sample matrix components (interferents) were investigated to understand their interfering effects. Urine sample pools that did not have detectable uCysB concentrations were spiked with the interferents (antibiotics, bilirubin, glucose, RBCs, hemoglobin, ketones) or were native samples containing the interferent (cloudiness, protein, lipids, pH) collected at Idexx reference laboratories. Pools of each sample containing the interferent were aliquoted and spiked with 3–6 levels of recombinant CysB. Four or more replicates were tested per level of CysB for each interferent type, without (control) and with the interferent. The recovery % was calculated relative to the control: aliquots without the interferent (antibiotics, bilirubin, glucose, ketones, RBCs, hemoglobin, protein), before being spun (cloudiness, lipids), or the spiked level [i.e., the concentration obtained by the gravimetric method (pH)]. For protein, the control was a native pool that did not have detectable uCysB, which was spiked with the same amount of CysB as the sample with the interferent. The recovery % across replicates was averaged (x̄ recovery % per CysB concentration).

Lot-to-lot variation was measured by testing 105 feline and 199 canine urine samples on each of 3 unique reagent lots. Each sample was run on all reagent lots in singlicate. The uCysB concentrations of each reagent lot were plotted against other reagent lots (e.g., lot 2 vs. lot 1, lot 3 vs. lot 1, lot 3 vs. lot 2). A least squares regression was used to determine slope and intercept. R² was used to understand the magnitude of variability between lots. To pass the test, the slope must fall between 0.8 and 1.2, and R² >0.9.^2,5,6,22

The analytical specificity was evaluated by calculating the percentage of cross-reactivity. This was determined by comparing the uCysB concentration before and after spiking 2,000 ng/mL cystatin A (CysA) in a urine sample. The concentration of CysA in the spiked samples was purposefully higher to understand the effects on analytical specificity in the face of a large concentration of competitor molecules. The percentage of cross-reactivity was calculated as follows:

$$\begin{array}{l}\%\mathrm{cross}-\mathrm{reactivity}\\ =\left(\frac{\mathrm{CysB}\mathrm{concentration}\mathrm{of}\mathrm{sample}\mathrm{after}\mathrm{spiking}\mathrm{with}\mathrm{CysA}}{\mathrm{CysB}\mathrm{concentration}\mathrm{of}\mathrm{sample}\mathrm{beforespiking}\mathrm{with}\mathrm{CysA}}\right)\\ \times 100\%\end{array}$$

For freeze-thaw stability evaluation, uCysB measurement in 80 feline and 73 canine urine samples was studied before and after up to 3 cycles of freezing (–80^oC) and thawing. Samples with native uCysB concentrations in the dynamic range of 50–500 ng/mL were divided into 6 aliquots. The aliquots were then subjected to 0 (control), 1, 2, or 3 freeze-thaw cycles (2 aliquots/cycle) by transferring the tubes from a −80°C freezer to 4°C for 6 h and returning to the freezer overnight. After the last freeze cycle, all samples were stored at 4°C (maximum 1 wk) and tested in duplicate with the agglutination immunoassay. The percent bias of the mean result of each freeze-thaw challenged sample was calculated relative to its paired unchallenged control (4°C storage only with no freeze-thaw cycle) as follows:

$$\%\mathrm{bias}=\frac{\begin{array}{l}\mathrm{concentration}\mathrm{with}\mathrm{freeze-thaw}\\ -\mathrm{concentration}\mathrm{with}\mathrm{no}\mathrm{freeze-thaw}\end{array}}{\mathrm{concentration}\mathrm{with}\mathrm{no}\mathrm{freeze}-\mathrm{thaw}}\times 100\%$$

For storage stability analysis, 10 canine and 10 feline urine sample pools, each made with either 1 or 2 samples, were evaluated for measurand integrity over time at 4°C and 30°C storage. Each pool was divided into 9 aliquots; 4 aliquots were stored at 4°C, 4 aliquots were stored at 30°C, and 1 aliquot was stored at −80°C to serve as a baseline control. Aliquots stored at 4°C and 30°C were tested for uCysB on days 1, 2, 3, and 7. A reverse keep strategy was employed for testing. Once each aliquot was exposed to the desired storage level at 4°C and 30°C, the temperature-stressed aliquots were transferred to −80°C. On day 7, the samples were thawed together and tested in a single event, in duplicate. The average uCysB concentration of each temperature-stressed aliquot was compared to the average concentration of the −80°C baseline control (i.e., recovery % relative to −80°C).

Results

Within-run CV was 0.38–3.63% for dogs and 0.44–7.14% for cats (Table 1). Between-run CV was 1.49–5.09% for dogs and 0.65–5.05% for cats (Table 1). The number of runs achieved for reagent lots 1, 2, and 3 (n = 59, 27, and 27, respectively) was not the same due to protocol adjustments or because some replicates failed to obtain results.

Table 1.

Precision of the agglutination immunoassay evaluated by urinary cystatin B means (x- uCysB) and CV% within (A) and between (B) runs. Results were summarized per species (canine or feline), reagent lots (1–3), and pools (1–7).

Reagent lot (native pool)	Days	Canine			Feline
		Replicates	x- uCysB, ng/mL	CV%	Replicates	x- uCysB, ng/mL	CV%
A) Within run
1 (1)	1	20	132	1.41	20	84.1	3.50
1 (2)	1	20	287	0.97	20	266	0.63
1 (3)	1	20	480	1.78	20	460	0.54
2 (4)	1	20	44.4	3.63	20	49.7	4.32
2 (5)	1	20	79.7	2.53	20	102	1.74
2 (6)	1	20	198	0.93	20	198	1.41
2 (7)	1	20	391	0.51	20	431	0.44
3 (4)	1	20	44.0	3.40	20	45.7	7.14
3 (5)	1	20	80.3	2.17	20	93.4	2.42
3 (6)	1	20	197	0.68	20	180	0.99
3 (7)	1	20	395	0.38	20	407	0.82
B) Between run
1 (1)	10	59	130	1.96	60	84.8	2.09
1 (2)	10	59	281	1.93	60	261	1.40
1 (3)	10	59	466	2.43	60	449	1.92
2 (4)	5	27	43.7	2.60	27	50.7	4.04
2 (5)	5	27	78.5	3.60	27	103	2.47
2 (6)	5	27	197	2.64	27	200	2.31
2 (7)	5	27	388	1.49	27	429	0.65
3 (4)	5	27	46.0	5.09	27	45.0	5.05
3 (5)	5	27	81.9	2.14	27	96.1	1.92
3 (6)	5	27	204	3.10	27	191	2.48
3 (7)	5	27	406	1.89	27	415	1.52

For accuracy, the recovery % was 89–101% for dogs and 91–112% for cats (Table 2). After reagent lot 1 was tested, the number of urine pools was reduced to 9 for dogs and 10 for cats due to resource constraints (Table 2). The maximum CV and absolute proportional bias observed on concentrations between 50 and 500 ng/mL were 5.09% and 11.4% for dogs and 5.05% and 12.5% for cats, respectively; consequently, the maximum TEo was 21.6% and 22.6%. If we focus on concentrations between 70 and 130 ng/mL (100 ± 30%), the maximum CV and proportional bias observed were 3.60% and 9.41% for dogs and 2.47% and 8.91% for cats, respectively; consequently, the maximum TEo is equal to 16.3% (dogs) and 13.9% (cats). The R² and slopes meet internal requirements for linearity across the reportable range in dogs and cats (Fig. 1).

Table 2.

Reagent lot, number of canine (A) and feline (B) pools (n pools) aliquoted and spiked with 4 different levels of cystatin B (x- spiked CysB), replicates per pool aliquot, x- CysB obtained by the agglutination immunoassay, recovery percentage (Recovery %), and proportional bias of the CysB immunoassay used for accuracy analysis.

Reagent lot (n pools)	x- spiked CysB, ng/mL	Replicates per pool aliquot	x- immunoassay CysB, ng/mL	Recovery %	Proportional bias, %
A) Canine
1 (15)	85.1	4	77.3	91	−9
	130	4	119	91	−9
	261	4	236	91	−10
	464	4	411	89	−11
2 (9)	52.6	4	49.0	93	−8
	101	4	97.7	97	−3
	227	4	230	101	0
	427	4	433	101	1
3 (9)	52.6	4	48.9	93	−8
	101	4	95.9	95	−5
	229	4	219	96	−4
	427	4	411	96	−4
B) Feline
1 (15)	55.1	4	56.0	101	2
	101	4	94.0	93	−7
	226	4	209	92	−8
	428	4	390	91	−9
2 (10)	48.3	4	54.2	112	13
	100	4	99.3	99	−2
	249	4	258	104	4
	501	4	544	108	8
3 (10)	48.3	4	50.3	104	4
	100	4	91.4	91	−9
	249	4	233	94	−6
	501	4*	526	105	5

Recovery % and proportional bias % columns have been rounded.

^*One pool had only 3 replicates.

Figure 1.

Linearity of the urinary cystatin B (uCysB) agglutination immunoassay evaluated using cross-diluted pools from A) 5 dogs and B) 4 cats, across the reportable range (50–500 ng/mL). The intercept, slope, and coefficient of determination (R²) were 9.65, 1.1, and 0.99 for dogs, and 3.98, 0.94, and 1 for cats, respectively.

Amoxicillin and ciprofloxacin in the urine did not impact the immunoassay performance (recovery % of 99–103%) in dogs and cats (Table 3). Recovery % in the presence of doxycycline was 65.6–101%, with better immunoassay performance (recovery % of 81.7–101%) when doxycycline was <0.5 mg/mL. Non-antibiotic interferents (Tables 4–6) did not impact the immunoassay performance (recovery % of 80–120%). R² and slopes from the lot-to-lot variation analysis for dogs and cats were within the internal requirement of R² >0.9 and slope of 0.8–1.2 (Table 7; Fig. 2). For analytical specificity of the immunoassay, CysA did not cross-react in the CysB immunoassay, as the concentrations of CysB obtained on the sample before and after spiking with CysA were both zero.

Table 3.

Cystatin B mean (x- CysB), SD, CV%, recovery percentage (Recovery %), and number of replicates (n) obtained on antibiotic interferent analysis of the agglutination immunoassay using canine (A) and feline (B) urine pools spiked with antibiotics and different levels of recombinant CysB.

Antibiotic	Controls (pools without antibiotic)				Test (pools spiked with antibiotic)
	x- CysB, ng/mL	SD	CV%	n	x- CysB, ng/mL	SD	CV%	n	Recovery %
A) Canine
Doxycycline
0.1 mg/mL	78.6	0.2	0.33	4	74.9	0.3	0.49	4	95
	93.7	0.9	0.96	7	85.2	0.4	0.54	4	91
	178	1.0	0.60	4	173	0.5	0.31	4	98
	223	2.0	0.92	4	211	0.8	0.40	4	94
	344	2.4	0.70	4	342	3.6	1.07	4	99
	467	3.3	0.71	6	471	1.6	0.35	4	101
0.2 mg/mL	78.6	0.2	0.33	4	70.0	1.3	1.94	4	89
	93.7	0.9	0.96	7	76.6	1.2	1.67	4	82
	178	1.0	0.60	4	172	1.1	0.68	4	97
	223	2.0	0.92	4	203	3.1	1.55	4	91
	344	2.4	0.70	4	342	2.5	0.75	4	99
	467	3.3	0.71	6	472	3.4	0.74	4	101
0.3 mg/mL	78.6	0.2	0.33	4	71.4	2.1	2.98	4	91
	178	1.0	0.60	4	169	1.0	0.64	4	95
	344	2.4	0.70	4	334	0.5	0.18	4	97
0.4 mg/mL	78.6	0.2	0.33	4	71.4	0.6	0.97	4	91
	178	1.0	0.60	4	167	0.9	0.56	4	94
	344	2.4	0.70	4	332	3.5	1.05	4	97
0.5 mg/mL	78.6	0.2	0.33	4	69.2	0.3	0.43	4	88
	93.7	0.9	0.96	7	61.5	2.7	4.52	4	66
	178	1.0	0.60	4	172	1.3	0.77	4	97
	223	2.0	0.92	4	181	18.7	10.4	4	81
	344	2.4	0.70	4	335	1.5	0.47	4	97
	467	3.3	0.71	6	470	2.6	0.56	4	101
1.0 mg/mL	93.7	0.9	0.96	7	68.8	2.1	3.15	4	73
	223	2.0	0.92	4	198	1.2	0.62	4	89
	467	3.3	0.71	6	470	4.7	1.01	4	101
1.5 mg/mL	93.7	0.9	0.96	7	73.4	4.7	6.52	4	78
	223	2.0	0.92	4	196	4.1	2.12	4	88
	467	3.3	0.71	6	467	3.5	0.76	4	100
Amoxicillin
200 ug/mL	99.9	2.4	2.48	4	99.4	1.3	1.36	4	99
	224	1.4	0.64	4	223	3.4	1.53	4	100
	410	5.3	1.31	4	412	4.1	1.01	4	100
Ciprofloxacin
260 ug/mL	243	1.8	0.76	4	243	3.2	1.33	4	100
	102	1.3	1.32	4	104	1.4	1.38	4	102
	429	3.5	0.83	4	426	4.7	1.12	4	99
B) Feline
Amoxicillin
722 ug/mL	89.5	0.5	0.66	4	89.0	1.6	1.88	4	100
	200	0.9	0.47	4	198	3.0	1.53	4	99
	355	0.8	0.23	4	353	5.7	1.62	4	100
Ciprofloxacin
260 ug/dL	119	0.9	0.83	4	117	1.3	1.14	4	99
	256	2.2	0.86	4	257	2.6	1.03	4	101
	471	5.5	1.18	4	469	3.2	0.69	4	100

Recovery % column has been rounded.

Table 4.

Cystatin B mean (x- CysB), SD, CV%, number of replicates (n), and recovery percentage (Recovery %) obtained on non-antibiotic interferent analysis using the agglutination immunoassay on canine (A) and feline (B) urine pools spiked with different levels of interferents and recombinant CysB.

Antibiotic	Controls (pools without interferent)				Test (pools spiked with interferent)
	x- CysB, ng/mL	SD	CV%	n	x- CysB, ng/mL	SD	CV%	n	Recovery %
A) Canine
Bilirubin
15 mg/dL	85.0	0.8	1.01	4	86.5	1.2	1.42	4	102
	213	2.3	1.08	4	214	0.1	0.04	4	100
	398	3.6	0.90	4	401	3.1	0.80	4	101
30 mg/dL	85.0	0.8	1.01	4	89.9	1.0	1.12	4	106
	213	2.3	1.08	4	218	2.1	0.99	4	102
	398	3.6	0.90	4	408	3.0	0.75	4	102
Glucose
200 mg/dL	85.2	0.7	0.89	8	83.1	1.4	1.73	4	98
	230	2.4	1.05	8	224	1.9	0.88	4	97
	436	3.0	0.71	8	428	2.3	0.55	4	98
1,000 mg/dL	85.2	0.7	0.89	8	80.0	0.4	0.60	4	94
	230	2.4	1.05	8	215	2.0	0.94	4	94
	436	3.0	0.71	8	417	3.1	0.75	4	96
Ketones
150 mg/dL	85.2	0.7	0.89	8	82.0	0.9	1.11	4	96
	230	2.4	1.05	8	220	2.3	1.06	4	96
	436	3.0	0.71	8	420	3.9	0.94	4	96
Erythrocytes
250 cell/dL	107	1.3	1.30	4	106	1.0	0.99	4	99
	244	3.6	1.50	4	242	3.6	1.52	4	99
	444	4.0	0.90	4	443	4.6	1.04	4	100
Hemoglobin
20,000 lysed RBCs	99.4	1.3	1.34	4	108	1.4	1.37	4	109
	234	3.0	1.30	4	243	2.1	0.90	4	104
	429	2.5	0.59	4	437	5.7	1.32	4	102
B) Feline
Bilirubin
15 mg/dL	108	1.1	1.09	4	110	1.0	0.92	4	102
	237	2.5	1.08	4	238	3.4	1.44	4	100
	432	2.8	0.65	4	433	4.7	1.10	4	100
30 mg/dL	108	1.1	1.09	4	113	1.9	1.70	4	104
	237	2.5	1.08	4	245	1.9	0.77	4	103
	432	2.8	0.65	4	444	8.3	1.87	4	103
Glucose
200 mg/dL	134	2.4	1.80	8	131	1.0	0.82	4	98
	308	2.3	0.77	8	305	1.9	0.65	4	99
	492	4.5	0.92	8	490	5.5	1.13	4	100
1,000 mg/dL	134	2.4	1.80	8	127	1.4	1.12	4	95
	308	2.3	0.77	8	295	2.5	0.85	4	96
	492	4.5	0.92	8	468	5.4	1.17	4	95
Ketones/acetone
150 mg/dL	134	2.4	1.80	4	131	2.5	1.93	4	98
	308	2.3	0.77	4	294	1.0	0.37	4	95
	492	4.5	0.92	4	475	7.3	1.54	4	96
Erythrocytes
250 intact cells/dL	115	1.4	1.24	4	114	0.1	0.13	4	100
	253	1.7	0.70	4	254	2.1	0.83	4	100
	465	2.5	0.54	4	464	4.1	0.90	4	100
Hemoglobin
20,000 lysed RBCs	112	1.5	1.41	4	120	0.2	0.24	4	108
	248	2.2	0.92	4	258	2.3	0.90	4	104
	451	3.0	0.69	4	466	3.0	0.64	4	104

Recovery % column has been rounded.

Table 5.

Cystatin B mean (x- CysB), SD, CV%, number of replicates (n), and recovery percentage (Recovery %) obtained on non-antibiotic interferent analysis of the agglutination immunoassay using canine (A) and feline (B) urine pools spiked with different levels of CysB. For cloudiness and lipids, the aliquots were run before (controls) and after spun; for protein, the controls were samples without the presence of protein.

Interferent	Controls (pools without interferent)				Test (pools spiked with interferent)
	x- CysB, ng/mL	SD	CV%	n	x- CysB, ng/mL	SD	CV%	n	Recovery %
A) Canine
Cloudiness	116	1.5	1.31	4	112	2.3	2.09	4	97
	223	0.9	0.42	4	222	2.7	1.26	4	99
	415	2.7	0.66	4	411	5.0	1.23	4	99
Lipids	117	1.3	1.15	4	115	0.7	0.63	4	98
	258	3.5	1.37	4	254	3.1	1.25	4	99
	478	5.5	1.15	4	474	6.4	1.35	4	99
Protein 3+	130	1.4	1.07	4	129	1.8	1.40	4	99
	282	1.7	0.60	4	279	0.2	0.07	4	99
	432	3.8	0.88	4	418	4.1	0.98	4	97
B) Feline
Cloudiness	119	0.9	0.81	4	118	1.1	0.96	4	100
	261	3.1	1.21	4	264	2.7	1.05	4	101
	483	2.9	0.60	4	481	3.1	0.66	4	100
Lipids	129	1.4	1.11	4	129	2.2	1.74	4	100
	258	3.3	1.29	4	258	3.5	1.38	4	100
	445	4.5	1.02	4	449	3.7	0.84	4	101
Protein 3+	105	1.0	0.95	10	106	1.3	1.23	10	100
	246	1.5	0.61	10	225	1.3	0.58	10	91
	376	2.2	0.59	10	345	4.5	1.31	10	92

Recovery % column has been rounded.

Table 6.

Mean (x-) spiked cystatin B (CysB), x- immunoassay CysB, SD, CV%, recovery percentage (Recovery %), and number of replicates (n) obtained on non-antibiotic interferent analysis of an agglutination immunoassay using canine (A) and feline (B) urine pools spiked with recombinant CysB.

Interferent	Control	Test
	x- spiked CysB, ng/mL	x- immunoassay CysB, ng/mL	SD	CV%	Recovery %	n
A) Canine
pH 5.1	105	106	0.8	0.78	101	4
	235	199	2.2	1.11	85	4
	444	377	2.0	0.53	85	4
pH 6.5	109	126	0.7	0.58	115	4
	242	253	1.0	0.42	105	4
	386	397	1.5	0.39	103	4
pH 7.5	96.4	88.2	0.4	0.52	92	4
	243	236	0.3	0.14	97	4
	398	411	1.9	0.47	103	4
pH 9.5	99.9	94.3	0.5	0.56	94	4
	239	214	0.4	0.19	89	4
	384	350	0.8	0.23	91	4
B) Feline
pH 6	92.5	87.9	1.0	1.15	95	4
	249	232	1.2	0.54	93	4
	387	360	1.4	0.39	93	4
pH 6.5	104	95.4	0.9	1.02	92	4
	244	219	1.3	0.63	90	4
	385	348	1.6	0.49	91	4
pH 8	105	103	1.9	1.85	98	4
	242	228	2.2	0.99	94	4
	388	365	0.7	0.22	94	4
pH 8.5	102	114	2.2	1.95	112	4
	238	234	1.8	0.79	98	4
	377	362	1.6	0.45	96	4

Recovery % column has been rounded.

Table 7.

Intercept, slope, and coefficient of determination (R²) obtained from the reagent lot regression analysis (lot 1 vs. lot 2, lot 1 vs. lot 3, lot 2 vs. lot 3) using data from 199 dog samples and 105 cat samples.

Species	Regression analysis	Intercept	Slope	R 2
Canine	lot 1 vs. lot 2	−0.15	0.90	0.99
	lot 1 vs. lot 3	−2.94	0.96	0.99
	lot 2 vs. lot 3	−2.80	1.07	1.00
Feline	lot 1 vs. lot 2	1.59	0.99	0.99
	lot 1 vs. lot 3	1.10	0.96	0.97
	lot 2 vs. lot 3	−0.12	0.96	0.97

Figure 2.

Lot-to-lot variation among 3 reagent lots (lot 1 vs. lot 2, lot 1 vs. lot 3, lot 2 vs. lot 3) demonstrated using 199 dog samples and 105 cat samples. Dogs: left side (A–C); cats: right side (D–F).

At least 10% of canine samples (9 of 73 dogs) exposed to one or more freeze-thaw cycles resulted in bias outside the ±10% interval (Fig. 3). However, the median and interquartile range (IQR, between brackets) of the % bias for 0, 1, 2, and 3 freeze-thaw cycles for dogs was 0.75 (0.58–1.11), 1.61 (0.61–3.85), 1.90 (0.88–4.41), and 2.09 (0.93–4.15) respectively, which was within the ±10% bias interval. In cats, the biases from samples with freeze-thaw cycles were all within the ±10% interval. The median and IQR (between brackets) of the percent bias for 0, 1, 2, and 3 freeze-thaw cycles for cats was 0.40 (0.17–0.70), 1.37 (0.65–2.58), 1.21 (0.52–2.33), and 1.46 (0.72–2.65), respectively.

Figure 3.

Cystatin B stability in urine samples determined by bias of urinary cystatin B measurement per number of freeze-thaw cycles for A) 73 dogs and B) 46 cats, 2 replicates per cycle for each. Each graph represents one sample from one animal.

The recovery % for dog and cat samples stored at 4°C and cat samples stored at 30°C passed internal requirements (recovery % of 90–110%) and were 102–103% (dogs at 4°C), 99–101% (cats at 4°C), and 97–99% (cats at 30°C), for samples stored for 1, 2, 3, 4, and 7 d (Fig. 4; Table 7). The storage of dog samples at 30°C did not meet the recovery % requirement and was 89–106%, in which 5 of 10 samples had a recovery % <90% at least one time (samples 1, 2, 5, 6, 9; Fig. 4). Dog samples 2 and 5 did not meet the recovery % requirement in 3 of the 4 times (days 4–7) evaluated.

Figure 4.

Urinary cystatin B recovery percentage (Recovery %) at 4°C and 30°C relative to −80°C per time (1, 2, 3, and 7 d) observed in native urine pools from A) 10 dogs and B) 10 cats. The open circles are results at 4°C, and the closed circles are results at 30°C.

Discussion

Our particle-assisted, turbidimetric, quantitative uCysB immunoassay provided accurate, precise, and reproducible measurements specific to CysB in canine and feline urine samples that were robust to interference from sample matrix components, cloudiness, and differences in pH. Furthermore, we performed our validation studies using high-throughput Beckman AU analyzers that can streamline laboratory workflows. Thus, our immunoassay is a useful and practical tool to be used in assessing kidney injury in canine and feline patients in the clinical setting. Our study also provided valuable guidance on sample handling prior to testing with the immunoassay.

The CVs obtained on precision analysis were <10% for within-run precision analysis, demonstrating that the immunoassay produces consistent results within a run. The between-run CV was generally slightly higher than the within-run CV. However, all between-run CVs were <25% of TEa (not shown), as recommended by ASVCP guidelines. ² TEo values calculated based on the maximum CV and bias observed were all within the TEa interval (not shown) and <30%. The TEo on concentrations of ~100 ng/mL was <20%.

High doxycycline concentrations (0.5–1.5 mg/mL) decreased CysB concentration in canine urine (% recovery, 65.6–101%), with no clear relationship between the doxycycline dose and the level of bias. This suggests that doxycycline at concentrations >0.5 mg/mL may impact immunoassay performance, potentially affecting clinical interpretation in dogs. Ciprofloxacin, amoxicillin, and all other non-antibiotic interferents did not affect immunoassay results, confirming that our assay can be used in different clinical situations. The lot-to-lot variation analysis (slopes of 0.96–1.07) showed that the immunoassay’s performance remains consistent when different reagent lots are used, supporting the validation of the immunoassay for diagnostic utility.

The bias % fell outside the acceptable ±10% bias limits for 9 of 73 dog samples following freeze-thaw cycling. This may be due to patient-specific sample matrix effects, particularly because 5 of the 9 samples had bias with one or more freeze-thaw cycles. In contrast, most of the canine samples (64) and all feline samples (46) did not have bias exceeding ±10% at one or more times. The recovery % for storage of samples at 4° or 30°C was inconsistent across species. While feline samples were not observed to be impacted by the tested temperature of storage, 50% of the canine samples stored at 30°C had a recovery % outside the acceptable interval (90–110%). Further information on patient health and sample collection conditions could help explain the reason for this inconsistency. However, the evidence collected and available during our study was insufficient to identify the definitive cause for this observation. Replication of the freeze-thaw and storage stability experiment might be valuable and helpful in clarifying whether the outlier values obtained on freeze-thaw and storage stability analysis for dog samples were due to the complex nature of urine, patient or sample conditions, or an immunoassay limitation. For optimal immunoassay performance, clinical samples should be submitted fresh for testing, with a target temperature of 4°C during transport from time of draw through time of testing, and freeze-thaw cycling should be avoided where practicable.

In the assessment of kidney injury, uCysB should be considered as a screening test for animals prone to developing AKI. Due to the multiple limitations of the functional markers and their low sensitivity for the diagnosis of AKI in its early phases, the diagnosis of AKI is often delayed in dogs ¹⁸ and cats, ⁴ which negatively affects the outcome of these patients. Hence, the most important consideration would be to identify any clinically relevant increase in uCysB that would alert the clinician to the injury rather than the extent of the increase. Additional information regarding the magnitude of the increase, which might reflect the extent of the damage to the renal parenchyma, may be helpful for prognostic purposes. We have provided validation results in the range of 50–500 ng/mL, yet this does not negatively affect the role for uCysB as a screening test for AKI, especially considering that extension of the immunoassay range is feasible through sample dilution. Further studies are needed to explore the clinical utility of uCysB in a broad range of clinical contexts. The primary goal of our validation study was to demonstrate that the measurement is accurate.

A potential limitation of our validation study is that it was performed in a controlled research laboratory environment rather than in the conditions of a routine commercial laboratory environment in which it would be used. However, in addition to being consistent with the rigor recommended by U.S. FDA guidelines, multiple study arms investigated immunoassay accuracy in a variety of ways that were meant to simulate conditions that would be encountered in the field. The comprehensiveness of our validation study provides confidence in generalization to the veterinary clinical environment.

A second potential limitation is that the cross-reactivity with cystatin C (CysC) was not assessed. This is because percent identity of CysC with CysB is only 22%. Cross-reactivity was instead tested for CysA, which has 50% identity to CysB,^15–17 and no cross-reactivity was observed. Further testing to rule out any potential cross-reactivity with CysC could be pursued in the future if warranted.

A third limitation is related to the stability analysis. The design of our study included one freeze-thaw cycle, and we found that a small number of dog samples were sensitive to freeze-thaw cycles. The freeze-thaw cycle might have caused the observed stability outlier values; however, we could not confirm whether the stability outliers were due to the freeze-thaw cycle or another reason. A fourth limitation arises from the current lack of a published consensus TEa for uCysB, which was addressed by the internal derivation of TEa for development purposes.