Partial analytical validation of the VetScan cPL rapid test

Abstract Background Serum canine pancreatic lipase immunoreactivity (cPLI) concentrations have become the standard laboratory test used to diagnose canine pancreatitis. Recently, a new point‐of‐care assay for cPLI, the VetScan cPL rapid test (VetScan cPL), has become available, but analytical validation data have not yet been published. Objective This study aimed to perform a partial analytical validation of the VetScan cPL. Methods Leftover serum samples from a diagnostic laboratory were used. Adherence to the manufacturer's guidelines, linearity, repeatability, and reproducibility were evaluated. Results of the VetScan cPL were correlated with the Spec cPL results. Results Observed‐to‐expected ratios for dilutional parallelism ranged from 77.4% to 162.9% (mean 119.3%). Intra‐assay and inter‐assay variabilities ranged from 16.9% to 36.7% (mean 25.1%) and from 14.1% to 51.2% (mean 31.8%), respectively. Adherence to the manufacturer's specification regarding results within ± 60 µg/L of the Spec cPL result was only achieved for 39% of the measurements. The VetScan cPL and Spec cPL correlation showed a Spearman's r of .758 for 29 data pairs. Conclusions Under the conditions of this study, the VetScan cPL did not adhere to the manufacturer's specifications for most measurements. Also, the VetScan cPL showed suboptimal linearity and was not precise. In conclusion, the VetScan cPL failed basic analytical validation.

reasons, the specificity of the serum cPLI concentration was 95.7%. 5 Measuring cPLI concentrations is also highly sensitive for the diagnosis of pancreatitis, though reported sensitivities have depended on the clinical presentation and study design. [6][7][8] Until recently, the only commercially available assays measuring cPLI were the Spec cPL and SNAP cPL (IDEXX Laboratories). The Spec cPL is a laboratory-based ELISA, using a recombinant antigen (recombinant canine pancreatic lipase) and a monoclonal antibody directed against native canine pancreatic lipase. 9 The analytical validation of this assay has been reported in the peer-reviewed literature. 9 The Spec cPL assay was reported to be linear upon dilution with a working range of 30-2000 µg/L. 9 Intra-assay variability for three samples and 12 repeated measurements were reported to be 7.8%, 9.0%, and 11.2%, and the inter-assay variability for three samples and five repeat measurements were reported to be 3.8%, 7.6%, and 5.6%. 9 The SNAP cPL is a point-of-care semi-quantitative device that is easy to perform and results in either a "normal" read (ie, the associated Spec cPL is in the reference interval [RI] of <200 µg/L) or an "abnormal" read (ie, the associated Spec cPL is either suggestive of pancreatitis or in the questionable range). 6,10 Abaxis has recently released a point-of-care test for cPLI that is also based on a rapid assay device that is being read by a reader connected to a smartphone. To the authors' knowledge, no analytical validation data have been provided for this new assay platform either in the peer-reviewed literature or the assay documentation. Thus, the goal of the current study was to perform a partial analytical validation of this new assay platform. This study is especially timely as a recent report suggested that the VetScan cPL correlated well with the Spec cPL and the diagnostic bin of the Spec cPL. 11 However, in that study, the assay was not performed as it would be used in clinical practice, but serum samples were sent to a central research laboratory where all samples were analyzed. 11 Because the comparison of the VetScan cPL with the Spec cPL has been reported previously, this study did not perform a method comparison study. The Spec cPL results are reported as reference points for the VetScan cPL results.

| Samples
All samples used for the partial validation of the VetScan cPL assay were from leftover serum samples that had been submitted to the Gastrointestinal Laboratory at Texas A&M University. All samples had originally been analyzed at the GI Lab, and then the leftover serum samples were frozen at −20°C for up to 4 months. However, all Spec cPL measurements referred to in this manuscript were performed at the same time as the VetScan cPL measurements. Before analyses, all samples were thawed and brought to room temperature, as suggested by the manufacturer.

| The Vue analyzer and VetScan cPL assay
All VetScan cPL measurements were taken using four different Vue Analyzers obtained from Abaxis (Union City, California). The Vue is a point-of-care analyzer that reads proprietary lateral flow devices, such as the VetScan cPL device. The assay is based on lateral flow technology that uses affinity-purified antibodies directed against canine pancreatic lipase, which is bound to colloidal gold particles.
Serum samples are applied to the device, and the pancreatic lipase in the sample binds to antibody-coated gold particles. These complexes are then captured by a secondary antibody on the test strip.
The accumulation of the captured gold particle/enzyme complex causes the color indicator to become visible on the test line. The signal is further amplified by the use of a competitive antibody calibration scheme employed on the control line. The darkness of the lines is quantified by densitometric analysis in the VetScan VUE (information taken from the product insert of the VetScan cPL). The working range of the assay is 50-60 µg/L to <700 to <800 µg/L. The assay marketing material suggests that the assay is quantitative and leads to results that are within a band of ±60 µg/L of the Spec cPL result. However, while most readers would display and report results as ±60 µg/L, some readers report results as ±50 µg/L.

| Linearity
The linearity of the assay was assessed by measuring dilutional parallelism of six canine serum samples of high-quality undiluted and at dilutions of 1:2, 1:4, and 1:8 with a pooled nonlipemic serum sample with an undetectable serum Spec cPL concentration.

| Effect of lipemia
For this experiment, we evaluated 3 naturally hypertriglyceridemic serum samples with serum triglyceride concentrations of 525, 580, and 1319 mg/dL undiluted, and at dilutions of 1:2 and 1:4.

| Reproducibility
Intra-assay variability was tested with three high-quality serum samples, designated Samples 1, 2, and 3, evaluated 10 times on four different VUE analyzers. Measurements were performed on the same analyzer during a single session on a single day. For this study, an intraassay variability of ≤10% was considered acceptable, an intra-assay variability 10% < %CV ≤ 20% was considered poor but acceptable, and an intra-assay variability of >20% was considered unacceptable.

| Repeatability
For the assessment of inter-assay variability, 10 high-quality serum samples were measured eight times on the same analyzer on different days (measurements were performed on consecutive days with no measurements taken on weekend days). Serum sample concentrations spanned the lower third of the Spec cPL assay working range since our previous findings showed that serum samples with a serum Leftover serum sample concentrations were chosen throughout the working range of the VetScan cPL assay. They were then aliquoted, and each aliquot was frozen in a separate sample tube at −20°C.
Samples were removed from the freezer and thawed immediately prior to analysis. For this study, an inter-assay variability of ≤10% was considered acceptable, an inter-assay variability 10% ≤ %CV ≤ 20% was considered poor but acceptable, and an inter-assay variability of > 20% was considered unacceptable.

| Adherence to the manufacturer's specifications
To determine whether the new VetScan cPL adhered to the manufacturer's specification that the results would be within ±60 µg/L of the Spec cPL assay, each nonlipemic measurement was assessed. It should be noted that one VUE analyzer read the result as ±50 µg/L of the Spec cPL assay, but the more conservative criterion of ±60 µg/L was used for these assessments. Spec cPL concentrations were determined and used as the target values. However, only VetScan Vue measurements were used to evaluate this criterion and were measured close to the same time as that of the Spec cPL concentrations.
Thus, for the inter-assay variability determinations where samples were run multiple times on consecutive days, only the initial Spec cPL concentration was used.

| Correlations
While we did not run a correlation study per se, our experiments created a set of data pairs that were then used to assess the correlation. The values for undiluted samples from the dilutional parallelism and lipemia experiments were used. Also, the first data point, each from the intra-and inter-assay variability experiments was used for this analysis, which generated a total of 34 data pairs. After excluding data pairs with results that were outside the working range for either assay (five data pairs had results that were outside the working range of the Vet Scan assay, one data pair from the intra-assay variability study had a VetScan cPL of <60 µg/L and a Spec cPL of 120 µg/L, one data pair from the inter-assay variability study had a VetScan cPL of <50 µg/L and a Spec cPL of 254 µg/L, one data pair from the inter-assay variability study had a VetScan cPL of <60 µg/L and a Spec cPL of 196 µg/L, and two data pairs from the dilutional parallelism study had a VetScan cPL of >800 µg/L and a Spec cPL of 1265 and 1287 µg/L, respectively), a total of 29 data pairs were selected for analysis.

| The statistical methods
For statistical analyses, single results outside the assay working range (eg, <60 µg/L) were transcribed as one unit above or below the working range limit (eg, 59 µg/L). However, sample sets that had results mostly outside of the working range (eg, all 10 samples <60 µg/L) were not used for the statistical analyses. Repeated measure ANOVA was used to assess differences among the different analyzers.
All statistical comparisons were performed using a statistical software package (GraphPad Prism 6.07). The level of statistical significance was set at .05 for all statistical comparisons. Spearman's correlation coefficient (r) was calculated using a statistical software package (GraphPad Prism).

| Linearity
The average observed-to-expected (O/E) ratio for a total of 13 ob-   (Table 2).

| Repeatability
The serum Spec cPL concentrations for the repeatability study had one sample each with Spec cPL concentrations of 120 µg/L (within the RI; Sample 1), 233 µg/L (in the questionable range; Sample 2), and 588 µg/L (in the suggestive for pancreatitis range; Sample 3).
For Sample 1, the %CVs were 19.4, 22.9, 23.0, and 31.5%, for the four different analyzers, which represented an unacceptable %CV for three out of the four analyzers (%CV > 20%, unacceptable).
One out of four analyzers had %CVs that were acceptable but poor (10% > %CV ≤ 20%). Sample 1 had values that ranged from 54 to 188 µg/L with an overall mean of 106 µg/L. Also, 33 Sample 1 measurements (82.5%) met the manufacturer's specifications of being in the range of Spec cPL ± 60 µg/L while seven did not, and all results fell into the RI. The four analyzers produced significantly different results from one another with sample means of 66, 105, 122, and 133 µg/L; P = .0037 (Table 3).

| Adherence to the manufacturer's specifications
A total of 252 measurements were assessed for adherence to manufacturer's specifications, which would require the results of the VetScan cPL to be within a range of the Spec cPL result ±60 µg/L.
Adherence to this specification was only achieved for 99 measurements (39%) but was not achieved for the majority of the measurements (153 samples or 61%).
TA B L E 3 Repeatability (intra-assay variability) of three serum samples from dogs analyzed on four different analyzers performed 10 times in one run. None of the variabilities was less than 10%, which is a general target value for repeatability, and only three of the 12 variabilities were within the "poor but acceptable" range of 10%-20% TA B L E 4 Reproducibility (inter-assay variability) of 10 samples analyzed on four different analyzers (two samples were analyzed on all four analyzers, and eight samples were analyzed on one analyzer). None of the %CVs were less than 10%, which is a general target value for reproducibility. Only three of the 14 variabilities were within the "poor but acceptable" range of 10%-20%. The samples that are shown in red are also depicted in Figure 1 Sample#

| Correlations
Our studies created several data pairs that could be used for rudimentary correlation assessments. The Spearman r for 29 data sets

| D ISCUSS I ON
In this study, we found that the Abaxis VetScan cPL assay, for measurement of serum cPLI concentrations, showed poor linearity, repeatability, and reproducibility when tested on a single occasion as is usually performed by veterinarians in general practice. As mentioned previously, the aim of this study was not to compare the diagnostic specificity and/or sensitivity of the VetScan cPL assay to that of the Spec cPL but rather the partial analytical validation of the VetScan cPL using the Spec cPL as a reference point since this assay has been analytically validated. 9 Of the other three cPLI assays described in the literature, two are no longer available, and the third one failed analytical validation. 9,[12][13][14] The VetScan cPL rapid test showed limited linearity. This was especially significant since the working range of the assay is relatively narrow (50 or 60 µg/L to 700 or 800 µg/L depending on the VUE analyzer used) when compared with the Spec cPL (30-2,000 µg/L).
Therefore, to be able to monitor disease progression using serum cPLI concentrations, dilutions would be required for many samples. 15 It should be noted that Abaxis does not currently recommend a protocol for the dilution of samples with high results; thus, based on the narrow assay working ranges, a sample dilution protocol should be developed that could improve the linearity of the assay. Evaluation of assay linearity was complicated by the fact that two of the six undiluted serum samples read outside the working range of the assay.
The standard method to test the influence of lipemia on serum biochemical assays is the addition of various amounts of Intralipid to various serum samples. 16 to assess renal function), but would be considered poor for trying to correlate two assays that measure the same analyte. As the calculated bias of 40.66% µg/L between the two assays was apparently random with no proportional bias, it would not be possible to correct the results with a constant or proportional adjustment factor.
One limitation of this study is that all VetScan cPL results were only obtained with single measurements. As indicated in the introduction of this manuscript, validation studies for clinical pathology analytes often use duplicate or even triplicate measurements.
However, the aim of this study was to describe analytical validation parameters under veterinary practice conditions, where a veterinarian would only run a patient sample on VetScan cPL once. Further studies using duplicate and/or triplicate measurements could provide further analytical validation data for this assay.
In summary, the Abaxis VetScan cPL assay for measurement of serum cPLI concentrations shows poor linearity, repeatability, and F I G U R E 3 A Bland-Altman plot shows a mean bias of 40.7 µg/L (solid line) and 95% limits of agreement from −239.4 to 311.7 µg/L (dashed lines). Most variation between the results of the two different assays was random as demonstrated by the wide limits of agreement; and therefore, correction with a constant or proportional adjustment factor would not improve performance reproducibility when tested on a single occasion as usually performed by veterinarians in general practice. Intra-and especially inter-assay variability were poor, and most sample results were identified in more than one diagnostic bin upon repeat analysis. The manufacturer's specification, suggesting that the VetScan cPL provides results within ±60 µg/L of the Spec cPL result was not confirmed by our studies as we often identified large differences with the serum Spec cPL concentrations. Further validation of this assay is, therefore, needed before this assay's clinical usefulness can be appropriately studied.

D I SCLOS U R E S
All authors are employed by or affiliated with the Gastrointestinal Laboratory at Texas A&M University, which performs cPLI testing on a fee-for-service basis. Dr Steiner also serves as a paid consultant for IDEXX Laboratories, the manufacturer of Spec cPL and SNAP cPL.