Ethics approval
The present study followed the indications concerning the animal's welfare under the supervision of a veterinarian in conformity with the European Council guideline 86/609/EEC for the use of animals for experimental and other scientific purposes and, fecal samples do not require ethical approval. Study design, analysis, and reporting of methods and results are presented in accordance with the ARRIVE guidelines.
Samples
In 2018 and 2019, with previous permission of owners, fecal samples from horses naturally infected with mixed species of parasites from 49 farms, were collected and analyzed for the detection of strongyle eggs, regardless of sex, age, or breeding status. Prior permission was obtained from the owners for individual rectal collection of feces in latex gloves, recorded and refrigerated at 4°C for transport to the Veterinary Parasitology Laboratory of University of Caldas, Colombia.
Methods for fecal egg counts
Fecal samples were processed following the methodology previously described by Noel et al. [12] and Cringoli et al. [39] for the McMaster (MM) and mini-FLOTAC (mF) tests, respectively, using either glucose (350 g) and sodium chloride (175 g) with a specific gravity of 1.25 as flotation solution. Each fecal sample was previously homogenized, and 15 grams of fecal subsamples were separated and weighed on a laboratory balance (0.01 g accuracy) for further analysis as described below [40].
For McMaster's test, four grams of feces measured in an analytical balance were diluted and gently mixed with 56 ml of flotation solution (MM56) in a plastic cup, a detection limit of 50 (i.e., multiplication factor). The solution was then filtered through a sieve of ~1000 µm diameter, and the filtered solution was used to fill the two counting chambers of a McMaster slide (Bioplast® Brazil) with the aid of a transfer pipette, which was previously shaken 8 times and, the grids of both chambers were read after a resting period of 10 minutes. The total volume examined was 0.3 ml. The same procedure was performed for MM45, with a stool solution ratio of 5:45 (detection limit of 33.33).
For the Mini-FLOTAC test (University of Naples, Naples, Italy), four grams of feces were added to the Fill-FLOTAC (mF56) and homogenized (8 times with the plunger) with 56 ml of the flotation solution (detection limit of 7.5). In this case, filtering of the material was performed through the Fill-FLOTAC mesh (250 μm diameter). Each mF counting chamber was filled with 1 ml (the total volume examined was 2 ml), and reading was performed after 10 min of rest. The same procedure was performed for mF45, with a stool solution ratio of 5:45 (detection limit of 5).
Data Plotting
Raw egg count (REC) data were recorded for each FECT in an Excel spreadsheet and subjected to analysis [5]. Strongyle egg levels (SEL) expressed as EPG (eggs per gram of feces) were obtained according to the corresponding multiplication factor as negative (eggs not seen), 1 - 400 EPG, 401 - 600 EPG, 601 - 1200 EPG, 1201 - 2400 EPG and > 2401 EPG, categories that responded to normality analysis as mentioned below.
The normality distribution of the data was determined by the Kolgomorov-Smirnov test (p> 0.05), the original REC values were logarithmically transformed (log (REC + 15)). Subsequently, the differences between the compared methods were established and, the normality of the differences (log) was verified by a histogram, box plot, and statistical determination of the mean and median [7, 41, 42].
Statistical Analysis
The study was designed based on recommendations recently published by Ghafar et al. [1], following stage I (analytical performance) and stage III (comparison), and the STARD 2015 guidelines for the reporting of diagnostic accuracy studies [43].
The statistical software JAMOVI® version 2.3, modules blandr and SimplyAgree (Jamovi Computer Software; https://www.jamovi.org; 2022), and the software MedCalc® version 19.5.2 (MedCalc Software Ltd. Ostend, Belgium; http://www.medcalc.org; 2020) was used for the analysis of intra-observer variation "repeatability" (Study I) and interchangeably evaluation or "concordance" (Study II) by B-A test [34].
The dispersion of the data was visualized by plotting graph B-A, the differences between the paired values obtained by the methods (X - Y) located on the Y axis "bias", against the logarithmic means of the values obtained by the same methods ((X + Y)/2), on the X axis, understanding as agreement, the value of the bias; and precision, interpreted as the dispersion of the bias [44, 45].
The following indicators were determined: bias, LoA limits (limits of agreement), CV (coefficient of variation) and CR (coefficient of repeatability), CCC (concordance correlation coefficient), and the intraclass correlation coefficient (ICC); the latter, interpreted at 95% CI interpreted as ≤0.5 "poor", 0.5 - 0.75 "moderate", 0.75 - 0.9 "good" and ≥ 0.9 "excellent" reliability index [46].
Study I: Intra-observer variation or repeatability
130 fecal samples (from eleven farms) from naturally infected equines were analyzed with duplicate readings by MM56, mF56, MM45, and mF45 methods, resulting in a total of 1040 observations; with a power (1 - ß) 0.90 for a significance level α = 0.05 and minimally interesting effect size δ = 0.5, obtained by Jamovi® software (Jpower module). In this phase of the study, the ICC was estimated based on a single rater/measurement, consistency, and 2-way mixed-effects model [47].
Study II: Interchangeably evaluation or concordance
For this phase of the study, a total of 1050 fecal samples from 41 farms were analyzed by a single reading by the MM56, mF56, and mF45 methods for each sample (3150 replicates), with a power (1 - ß) 0.95, significance level α = 0.05 and minimally interesting effect size δ = 0.064; obtained by Jamovi® software (Jpower module). Due to insufficient paired data, the MM45 technique was not considered at this phase. The intraclass correlation coefficient (ICC) was estimated in this phase of the study, based on a mean-rating (k = 2), absolute-agreement, and a 2-way mixed-effects model [47].
Normality, homogeneity of variance (heteroscedasticity), and proportionality of the residuals to bias were plotted and used as precision indicators [47]. Finally, the diagnostic test was evaluated by calculating sensitivity, specificity, positive predictive value, and negative predictive value, using the Jamovi® software (Meddecide module). The MM56 test was considered the reference test, after analysis of the dichotomous results (positive and negative) and compared with the results obtained by mF45 (test under evaluation).