Study design
Study design was developed using American Society of Veterinary Clinical Pathology (ASVCP) guidelines: allowable total error guidelines for biochemistry [11]. Total allowable error for biochemical analytes was indicated in the guideline. Sample collection and animal use were approved by the institutional animal research ethics committee at the Addis Ababa university college of veterinary medicine and agriculture used in this study [Certificate reference no VM/ERC/09/01/12/2020].
Study animals, Sample size and sampling technique
The World Organisation for Animal Health (OIE) guideline 3.6.6 selection and use of reference samples and panels recommended minimum of 5 samples to prepare serum pool [12]. In addition to compute a statistically valid number of samples as suggested by Bayes Success-Run Theorem for validation studies 95% confidence and 90% reliability used. Therefore n = 28.4. We used 29 samples for the study [13]. Study animals were adult horses recruited by convenient sampling technique at society for protection of animal’s abroad (SPANA-Ethiopia) clinic. Apparently healthy horses from owners who were consent after being informed about the purpose of the study were physically examined and blood sample was collected. Horses with history of medication excluded due to possible impact of drugs on analysis.
Collection and processing of blood samples
Blood samples from study animals were collected by a veterinarian from the jugular vein using standard operating procedure. The blood was allowed to clot at room temperature for between 30 minutes and serum was separated from the red blood cells by centrifugation at 1200xg for 10 minutes at 4 °C. Serum was immediately transferred to polypropylene tubes (Eppendorf Safe-Lock tubes) and stored at -20 °C until measurements. Samples were collected during two weeks in January 2020. Pooled serum samples were created by mixing equal volumes of individual serum then homogenized using an agitator for 10 min at 180 rpm. After homogenization aliquots of homogeneous pool was divided in to twenty portions to avoid effect of repeated thawing and freezing.
Analytical Validation
To examine the accuracy and precision of a commercial clinical chemistry kits (Jourilabs diagnostics reagents and laboratory chemicals) for the quantifying concentration of total cholesterol, urea and total Protein was used and analysis of the parameters were determined by the methods/techniques described as follows: urea by kinetic urease/GLDH (Glutamate dehydrogenase), total protein by biuret and total cholesterol by CHOD-PAP (cholesterol peroxidase4-aminophenazone). The procedure of validation was adopted from Westgard JO method validation protocol. The analytical validation comprises of recovery studies for accuracy and replication experiments for precision [14]. All tests were performed on semi-automated chemistry analyser (AMP clinical Diagnostics, USA)
Replication Experiments
Precision was assessed by evaluating the intra- and inter-assay variability using the pooled serum. Intra-assay variability (repeatability) was determined by measuring total cholesterol, urea and total protein in same sample 20 times sequentially within a single run. Inter-assay variability (reproducibility) was determined by analyzing the same sample in duplicate once on 20 consecutive working days. To avoid effect of repeated thawing and freezing, sample used for the determination of inter-assay variation were a liquated and stored at − 20 °C until use [15].
Recovery Experiments
The Spike and recovery (SAR) assessment is essential for the analysis and accuracy evaluation of the method for particular sample types. Spike and recovery assay is used to determine whether the detection of an analyte is affected by biological sample matrix and differences in the standard curve diluent [16, 17]. Serum samples were spiked with different concentrations of standard Total cholesterol (26 mg/dl; 0.1 ml of 200 mg/dl standard solution was spiked in 1 ml serum) Urea (9.1 mg/dl; 0.1 ml of 100 mg/dl standard solution was spiked in 1 ml serum) and total protein (1.1 mg/dl; 0.1 ml of 12 mg/dl standard solution was spiked in 1 ml serum).
Quality Requirement
Total Allowable Error (TEa)
The analytical performance of the clinical chemistry parameters were assessed by calculating TEobs (%) and σ values. TEobs (%) was determined by the following formula: TEobs (%) = 2 × CV + bias (%). Bias was calculated by the formula: Bias (%) = [(target – measured) ÷ target] × 100%, wherein “target” is the spiked value for each analyte and “measured” is the measured analyte concentration. TEobs (%) was calculated using the inter-assay CV and bias (%). If TEobs (%) is less than TEa (%); the quality assessment passes and no further action needed. Criteria for acceptable performance or total allowable error TEa (%) employed in this study (Total cholesterol: 20%, Urea: 12% and Total Protein: 10%) were adopted from American society of veterinary clinical pathology (ASVCP) guidelines: allowable total error guidelines for biochemistry [11].
Sigma metrics (σ)
Sigmas were calculated using the formula: σ = [TEa (%) – bias (%)] ÷ CV. A method was considered acceptable if TEobs< TEa. Interpretation of the σ values was performed as follows: >2: poor, > 3: marginal, > 4: good, > 5: excellent, and > 6: world class [18, 19].
Quality goal index ratio (QGI)
QGI ratio denotes the relative extent to which both precision and bias meet their respective quality. This was used to analyse the reason for the lower sigma in analytes, i.e., the problem is due to imprecision or inaccuracy or both. The QGI ratio was calculated as, QGI = Bias/1.5 × CV%. The criteria for interpreting QGI of the problem analytes with low sigma performance is as follows: QGI less than 0.8 shows imprecision, QGI falling in the range of 0.8 to1.2 shows both imprecision and inaccuracy and QGI greater 1.2 depicts inaccuracy [20].
Data analysis
Statistical analyses were performed using IBM SPSS 20. Normality distribution of the data was tested using the Kolmogorov–Sminorv test prior to statistical analysis. Data of accuracy from bias and precision from intra- assay and inter-assay CVs were estimated using routine descriptive statistical procedures.