The faecal egg count reduction test: Will identification of larvae to species improve its utility?

Abstract

In the faecal egg count reduction test, visual identification of larvae cultured from faeces enables the egg counts to be apportioned to species/genera, resulting in a more accurate test. However, morphology cannot reliably differentiate some species meaning that, in some cases, efficacy can only be estimated at the genus or species-complex level. We investigated the benefits of identifying larvae to species using DNA to determine how often this would alter the diagnosis of resistance and whether increasing the number of larvae identified would alter the repeatability of an efficacy estimate.

Data on faecal nematode egg counts and the corresponding larval species mixes were acquired from tests conducted on commercial sheep farms. The proportion of each species present in faecal culture was determined using DNA. Efficacy was then compared for individual species and for those genera/species complexes which cannot reliably be differentiated visually. The proportion of each species present was subsequently resampled 10,000 times (repeated random sampling) and efficacy recalculated to produce the median efficacy, along with the 5 % and 95 % simulation percentiles. Subsequently, the number of larvae sampled to determine the species mix in each sample was varied from 50 to 6400 and the process repeated.

Of 152 comparisons of efficacy, 25 % of cases where genus-level identification resulted in a finding of ‘susceptible’ for that category, species-level identification returned at least one diagnosis of ‘resistant’ i.e., genus-level identification resulted in a 25 % false negative diagnosis.

When the number of larvae sampled for species identification was low (<400) variation in efficacy estimates was high, however, as sample size increased the confidence interval around the efficacy estimate decreased.

The results indicate that identifying large numbers of larvae to species using DNA has the potential to increase the accuracy and confidence in efficacy estimates achieved using the faecal egg count reduction test.

Graphical abstract

Highlights

• •Identifying larvae to species using DNA improves the accuracy of the FECRT.
• •Genus-level identification led to a 25 % false negative diagnosis of resistance.
• •Use of DNA methods reliably detects resistance in poorly represented species.
• •Large sample sizes (>500 larvae) reduce uncertainty around efficacy estimates.
• •The nemabiome method enhances confidence and repeatability of FECRT.

1. Introduction

The faecal egg count reduction test (FECRT) has been the most used method of measuring on-farm anthelmintic efficacy and therefore testing for the presence of anthelmintic resistant nematodes, in domestic livestock for several decades (Kaplan and Vidyashankar, 2012; George et al., 2017). In its basic form the method involves counting the number of nematode eggs in subsamples of faeces collected from each of a group of young animals, before and after they are treated with anthelmintic (Coles and Roush, 1992). Calculating the reduction in mean egg count following treatment is a measure of efficacy with a reduction of <95 % being indicative of resistance (Coles and Roush, 1992). This test has several advantages; primarily that it is a direct measure of efficacy (i.e., phenotype) for all anthelmintic classes at manufacturer's recommended dose rates, unlike many in vitro and molecular tests (Kaplan et al., 2023) which must rely on correlating test results back to efficacy in vivo. Further, the FECRT can measure efficacy, and therefore detect anthelmintic resistance (AR), in a wide range of gastrointestinal nematode species, whereas molecular tests can only evaluate the limited number of species and actives for which genetic markers have been determined.

However, the FECRT also has limitations. Even today, anthelmintic efficacy is often measured solely as the reduction in total faecal nematode egg count (FEC) following treatment (McKenna, 1996; McIntyre et al., 2018; Queiroz et al., 2020; Maurizio et al., 2024). However, it has long been recognized, that the reduction in egg counts alone has limited value as it does not allow for a mix of susceptible and resistant parasite species contributing to the egg counts (McIntyre et al., 2018; Queiroz et al., 2020). More than 25 years ago, McKenna (1997) showed that 15/42 (36 %) of tests which returned a diagnosis of susceptible based on a ≥95 % reduction in total egg count, in fact, contained anthelmintic-resistant worm populations once egg counts were partitioned based on the generic composition of faecal cultures i.e., Once egg counts were allocated to species/genera based on the proportions of larvae present in faecal cultures, at least one was reduced by <95 % following treatment. In addition to a diagnosis of resistance, identification of larvae before and after treatment can also contribute additional information of value. For example, McIntyre et al. (2018) found that moderate reductions in total faecal egg count were in fact concealing a highly resistant Teladorsagia circumcincta population once larvae were identified to species, yielding more information than the egg counts alone.

Statistical analysis leading to confidence intervals for efficacy is recommended for resistance diagnosis conducted on egg counts (Kaplan et al., 2023). However, variance estimates at the genus/species level are seldom available due to the almost universal practice of undertaking faecal cultures on a single pooled sample of faeces. Hence, confidence estimation is not possible at this level. An additional complication associated with efficacy estimates utilizing faecal cultures has been that the standard method for characterising larvae involves a visual identification of 100 infective stage larvae (L3) (van Wyk and Mayhew, 2013; Borkowski et al., 2020), and sometimes fewer (Maurizio et al., 2024). Visual identification of L3 to species is not always possible or accurate due to the overlapping morphological and morphometric traits of some species and genera (Rossanigo and Gruner, 1996; Roeber and Kahn, 2014; Waghorn et al., 2014; Knoll et al., 2021; Maurizio et al., 2024). Because some genera (e.g., Trichostrongylus) frequently occur in species mixes (Waghorn et al., 2014) and the L3 cannot easily be speciated, visual identification generally groups these species into genera. This can result in inaccurate estimates of efficacy and sometimes an inaccurate diagnosis of AR. For example, Waghorn et al. (2014) reported a FECRT with an estimated efficacy against the genus Trichostrongylus of 99 % when the L3 were identified morphologically. However, once the L3 were identified to species using PCR (Bisset et al., 2014) it was found that only 4 % of the pre-treatment larval mix were T. colubriformis, while this species made up 100 % of the larvae recovered post-treatment. Thus a 99 % efficacy against the genus was found to consist of 75 % efficacy against T. colubriformis and 100 % efficacy against the other Trichostrongylus species once L3 were identified to species (Waghorn et al., 2014).

Identification of L3 to species using DNA has been possible for some time (Roeber and Kahn, 2014; Bisset et al., 2014) but has generally been too expensive for routine diagnostic uses. However, recent developments involving deep amplicon sequencing (nemabiome) (Avramenko et al., 2015) have made species identification more efficient with high throughput and, therefore, more affordable. Recent studies have investigated the utility of this method when incorporated into a FECRT (Queiroz et al., 2020).

Here we investigate the use of the nemabiome to identify L3 to species, compared to visual identification of L3 to genera, as part of FECRTs to determine whether the former is likely to influence a diagnosis of resistance and/or the estimated efficacy of treatment. We deliberately focus our attention on those genera and species-complexes which are not readily identified to species on morphology. Because the nemabiome offers an opportunity to greatly increase sample size for larval identification (Queiroz et al., 2020) we also consider the effect of sample size (i.e., the number of L3 speciated) on the degree of uncertainty around an efficacy estimate.

2. Methods

Data on faecal egg counts and the corresponding larval species mixes were acquired from FECRTs conducted on commercial sheep farms as part of routine efficacy testing. The data resulting from each reduction test involving 10–20 lambs treated with the same anthelmintic was then used as the basis for a statistical modelling process (repeated random sampling) where two questions were addressed.

• 1.Does identification of larvae to species influence the diagnosis of resistance (for simplicity defined here as < 95 % efficacy)?
• 2.Does the number of L3 identified influence the degree of uncertainty around the efficacy estimate and, potentially, the interpretation of the results?

Data were acquired from two sources. Waghorn et al. (2014) summarised egg count reductions and species composition pre- and post-treatment with anthelmintic for the genus Trichostrongylus only (identified to species using PCR) from 40 efficacy tests. In addition, FECRTs were conducted specifically for use in this study (Section 2.1.) adding an additional 43 efficacy tests. The latter tests included all species present allowing comparisons for Trichostrongylus spp., Cooperia spp. and the Long-Tailed (LT) species complex (i.e., Oesophagostomum, Chabertia and Bunostomum species).

2.1. Faecal egg count reduction tests

Between 2010 and 2012, Waghorn et al. (2014) processed faecal samples from 70 on-farm FECRT conducted by veterinarians for their farmer clients. The farms were predominantly mixed sheep and beef operations, and they were located across all regions of the country. On the day of treatment, 10–12 lambs were allocated to one of four treatment groups, a sample of faeces was collected from each and each lamb was dosed to individual liveweight with either albendazole, levamisole or ivermectin at the manufacturer's recommended rate, or left untreated as a control. All lambs were returned to pasture. Seven to 10 days after treatment a second faecal sample was collected from each lamb.

Faecal samples were couriered overnight to the laboratory where egg counts were determined using a modified McMaster technique (Lyndal-Murphy, 1993) with a sensitivity of 1 egg equating to 50 eggs per g (epg) of faeces. For the post-treatment samples only, a further 5g of faeces from each animal was pooled within treatment groups, mixed with vermiculite and water to achieve a uniform consistency, and incubated at 22 °C for 14 days. This temperature and incubation period was previously demonstrated in our laboratory to result in a more uniform percentage development of eggs to L3 across a range of nematode species compared to the higher temperatures (e.g., 27 °C) often used in diagnostic labs. Developed larvae were extracted by baermannisation into 10 ml of tap water (Hendrix, 1998).

In 2023-24, FECRT were conducted on nine commercial sheep farms (listed anonymously in Table 1) as part of their routine anthelmintic resistance testing programme, following a protocol similar to the guidelines in Kaplan et al. (2023). On each farm, a mob of lambs (N = 50–100) were set aside for the tests and monitored regularly for FEC until the mean count of 10 samples exceeded 500 epg. The number of treatment groups ranged from four to seven with 12–20 lambs allocated to each treatment. The anthelmintic treatments tested on the different farms were not always the same and were a compromise between the use of single actives to profile the resistance status on the farm and the use of combination products which the veterinarian and/or farmer wanted to evaluate for the purpose of controlling parasites (Table 1). The number of treatments was limited by the number of lambs available.

Table 1.

– Treatments used in faecal egg count reduction tests on 9 different farms where ‘Y’ indicates that treatment was tested on that farm.

N°	Treatment	Farm
		M	Sh	Sm	A	H	W	O	F	T
1	Oxfendazolea	Y	Y	Y	Y	Y	Y	Y	Y	Y
2	Levamisoleb	Y	Y	Y	Y	Y	Y	Y	Y	Y
3	Abamectinc	Y	Y	Y	Y	Y	Y	Y	Y	Y
4	Oxfendazole + levamisole + abamectind	Y	Y	Y	Y	Y	Y	Y	Y	Y
5	Eprinomectin + levamisole (spot on)e	Y			Y	Y
6	Monepantel + abamectinf	Y			Y
7	Derquantel + abamecting	Y			Y

^aBomatak C (Elanco Animal Health NZ) – 4.75 mg/kg.

^bNilverm (Coopers Animal Health NZ) 7.5 mg/kg.

^cGenesis (Merial/Boehringer Ingelheim) 0.2 mg/kg.

^dMatrix (Boehringer Ingelheim NZ) – 0.2, 4.75 & 7.5 mg/kg, respectively.

^eScorpius Elite (Donaghys NZ) 6.0 & 10.0 mg/kg, respectively.

^fZolvix Plus (Elanco Animal Health NZ) – 3.75 & 0.2 mg/kg, respectively.

^gStartect (Zoetis NZ) – 2.0 & 0.2 mg/kg, respectively.

On day zero, faecal samples were collected rectally from all lambs, before they were randomised into treatment groups, weighed, and treated with their allocated anthelmintic, administered to their individual liveweight, following the manufacturer's recommendation. All treatments were administered orally with a syringe except for one treatment on Farms M, A and H which was a recently released topical (spot-on) formulation of eprinomectin + levamisole, which was included at the farmers' request.

Following treatment, the lambs were mobbed up and returned to pasture. Between 8 and 13 days after treatment faecal samples were again collected from the same lambs, before they were treated with a combination of abamectin and either monepantel or derquantel (treatments 6 or 7 - Table 1) and returned to the farm. Faecal egg counts were determined using a modified McMaster method in which each egg counted equated to either 15 or 30 epg and, as above, pooled faecal samples were cultured to produce L3 for identification.

2.2. Identification of infective stage larvae

In Waghorn et al. (2014) larvae were individually picked from a mixed culture as being visually Trichostrongylus-type and identified using multiplex PCR and the primers sets described by Bisset et al. (2014) with a minimum of 32 L3 identified in each test.

In the more recent tests, the relative abundance of nematode species in each sample was determined using ITS-2 rDNA nemabiome metabarcoding (Avramenko et al., 2015) on approximately 2000 L3 (estimated by aliquot). Amplification of the ITS-2 marker using NC1 and NC2 primers and metabarcoding library preparation was conducted as previously described, with details available at www.nemabiome.ca/sequencing.html (Avramenko et al., 2015).

Bioinformatic analysis of samples used the Mothur bioinformatic tool version 1.36.1 (Schloss et al., 2009) as previously described (Avramenko et al., 2015). After removing reads <200 bp or >450 bp, paired-end reads were assembled into single contigs aligned to a bespoke ITS-2 database (Nematode ITS2 database version 1.3; https://www.nemabiome.ca/its2-database) and assigned to reference sequences using the k-nearest-neighbour method with k = 3. The percentage species composition of samples was calculated by dividing the total reads assigned to each species by the total number of reads per sample to obtain the relative percentage of each species and then multiplied by a species-specific correction factor to account for species-specific biases in the assay (Avramenko et al., 2015; Redman et al., 2019). Species which returned a prevalence below 0.05 % were removed from the data set, and the percentages recalculated, on the basis that they would relate to <1 L3 per sample, and thus likely due to contamination or misidentification of reads (Avramenko et al., 2015). Detailed ITS-2 rDNA nemabiome sequencing and bioinformatic analysis information are available at https://www.nemabiome.ca/analysis.html.

2.3. Modelling

A field test was considered to be a single evaluation of efficacy, i.e., one group of 10–20 lambs treated with the same anthelmintic on a single farm. Faecal egg counts for each animal, both pre- and post-treatment were available, along with the proportions of all nematode species present from each group of lambs both pre- and post-treatment. This was the key data used in the modelling as it constitutes ‘real’ proportions of different species mixes found on commercial farms, along with corresponding egg counts. These data were used in the modelling to address both questions and the proportions of different species measured using DNA methods were taken as the ‘true’ population.

Synthetic data (see below) consisted of the same variables, but these were manufactured to create specific situations for a structured comparison where each variable could be altered in turn to determine its impact on the outcomes of interest.

To address question 1, for every test 100 L3 (consistent with numbers normally used for visual differentiation) were sampled at random (with replacement) from the larval population based on the measured proportions i.e., the proportions of different species measured using DNA methods defined the ‘true’ population.

The data presented by Waghorn et al. (2014) only included proportions of Trichostrongylus spp., and so to maintain some parity between the data sets it was assumed that for these data Trichostrongylus spp. represented 50 % of the total larvae present (this was consistent with the 53 % mean value in the rest of the (all species) data). The species which could not be separated visually (i.e., Trichostrongylus spp., Cooperia spp. and the LT complex) were then summed to give genus/complex counts to represent normal outputs from a ‘visual’ identification. For example, the proportion of L3 of T. colubriformis, T. vitrinus and T. axei were summed to give the proportion of Trichostrongylus in the sample for use in the calculations of efficacy. By calculating the ‘visual’ identification in this way, rather than using an actual visual identification from the original data set, potential sampling errors in the comparison of visual (genus) and DNA identified (species) samples were eliminated. Identification error was not considered in this study. The mean FEC (pre- and post-treatment) was then apportioned to either species or genus and the percentage reduction for each of the ‘genus’ and ‘species’ samples was calculated and stored. The random sampling of 100 L3, and the calculation of both 'species' and ‘genus’ reductions, was repeated 10,000 times for each test. The median efficacy (along with the 5 % and 95 % simulation percentiles) was then calculated for each species vs genus comparison for each test.

Initially, to investigate the influence of sample size, (Question 2), artificial scenarios (synthetic data) were used. Egg counts and post-treatment proportions were constructed to produce efficacies of 50 %, 70 % and 90 % and the proportion of the test species present pre-treatment in an infinite larval population was set at 30 %, 20 %, 10 % or 2 %. Ten thousand random samples varying in size from 50 to 6400 were then taken for each combination of variables and the efficacy calculated for each. The median efficacy (along with the 5 % and 95 % simulation percentiles) was recorded and stored. Random number seeds were set to ensure the simulation outputs were reproducible.

Subsequently, a similar approach was used utilizing the field-derived FECRT data. Random samples (N = 50–6400) of L3 were taken from an infinite larval population defined by the species proportions measured from each test. Using mean pre- and post-treatment FECs (constant for each test) the percentage reduction for species of interest was calculated 10,000 times to produce a median and the 5 % and 95 % percentiles, for each sample size within each test.

3. Results

3.1. Species present and summary test results

Waghorn et al. (2014) visually identified 8 main genera of larvae as present in the untreated lambs i.e., Cooperia, Haemonchus, Nematodirus, Teladorsagia, Trichostrongylus, Oesophagostomum, Strongyloides and Chabertia. The focus of their study was the Trichostrongylus species, which made up, on average, 47 % of all the L3 identified, and they present data from only the 40 tests where all three species of Trichostrongylus were present in the control group. Using visual identification of L3 to genus the FECRT resulted in a <95 % reduction in FEC for Trichostrongylus in 11, 1 and 6 tests for albendazole, ivermectin and levamisole, respectively (Waghorn et al., 2014). Note that these results were not used in the current modelling study.

In the second set of tests, nemabiome identified 8–9 species as being present on the different farms in proportions above the 0.05 % threshold. The dominant species (i.e., in the highest proportions and on most farms) were Haemonchus contortus, T. circumcincta, Cooperia curticei, Oesophagostomum venulosum, T. axei, T. colubriformis, T. vitrinus and Chabertia ovina. After partitioning egg counts based on a visual identification of L3 in the cultures, simultaneous resistance to albendazole, levamisole and abamectin was present in T. circumcincta on 7 farms and in Trichostrongylus spp. on 8 farms. One farm exhibited resistance to only abamectin in T. circumcincta and one to albendazole in Oesophagostomum. Again, these results were not used in the modelling.

3.2. Does identification to species influence test result?

Results were available for 83 efficacy tests on-farm, which resulted in 152 comparisons of efficacy based on a ‘genus-level’ identification and multiple ‘species-level’ identifications, (there were 332 efficacy calculations at the species level). Of the 152 sets of genus vs species level comparisons, 83 involved the genus Trichostrongylus (i,e., the genus was present in all 83 tests), 36 the genus Cooperia, and 33 included the LT complex. There were four possible outcomes for each comparison, i.e., both methods could result in a diagnosis of ‘resistant’, or ‘susceptible’, or one could result in a diagnosis of ‘resistant’ while the other produced a diagnosis of ‘susceptible’. These are represented for Trichostrongylus by the four quadrants in Fig. 1A. It should also be noted that in this figure each ‘visual’ estimate is compared to 3 species –level estimates because there is only a single diagnosis for the former but multiple for the latter. Thus, most of the data points in the upper left category represent T. axei and T. vitrinus, which were almost always susceptible, but the genus was sometimes ‘resistant’ due to the presence of T. colubriformis. Hence, although the visual ID sometimes resulted in a ‘resistant’ diagnosis, two of the three species present were in fact susceptible and so they appear in the upper left quadrant. Fig. 1B enlarges the upper right portion of Fig. 1A, giving a clearer view of the four quadrants.

Fig. 1.

Scatter plot A) of median efficacy for the genus Trichostrongylus, against median efficacy for T. axei (circle), T. colubriformis (square) and T. vitrinus (triangle) when the species are considered separately, where plot B is an enlargement of the upper right region of plot A) with 5 % and 95 % percentiles included. Note that percentiles for variance in species-level efficacy have not been plotted, to minimize visual clutter.

For visual identification, 95 efficacy calculations were ≥95 % and were therefore ‘susceptible’ whilst 57 resulted in a diagnosis of ‘resistant’ (Table 2). However, when species were identified by DNA, 24 of the 95 ‘susceptible’ diagnoses were found to contain at least one resistant species. Thus, for the categories of Cooperia, LT and Trichostrongylus, 25 % of cases where genus-level identification resulted in a finding of ‘susceptible’ for that category, species-level identification returned at least one diagnosis of ‘resistant’. This was largely driven by the genus Trichostrongylus where 20/36 (56 %) cases returned a diagnosis of ‘susceptible’ for the genus but a diagnosis of ‘resistant’ (<95 % efficacy) for at least one species.

Table 2.

– The number of tests which resulted in a diagnosis of susceptibility or resistance (<95 % efficacy) when the larvae were identified visually to the species-complex (i.e., Trichostrongylus spp., Cooperia spp., or Long-tailed) or by DNA to species.

Diagnosis	Visual	DNA
Susceptible	95	71
Resistant	57	81

3.3. The number of larvae identified influences the test result

In the constructed (synthetic) data sets there was an interaction between the efficacy of treatment, the proportional representation of species and the number of larvae used to estimate the proportions (Fig. 2). If efficacy of treatment was low (50–70 %) the test diagnosis was always ‘resistant’ even though the level of uncertainty around the efficacy estimates increased as the proportion of species decreased (compare 30 % with 2 % in Fig. 2). However, at an efficacy of 90 %, if the number of larvae counted was low (50–200) there was often uncertainty around the diagnosis i.e. a proportion of tests returned an efficacy >95 %. This uncertainty decreased as sample size increased until >95 % of tests resulted in a diagnosis of ‘resistant’. For the examples in Fig. 2 this threshold for confident diagnosis was between 200 and 400 larvae sampled for species identification.

Fig. 2.

– The effect of number of larvae identified to species on the variation in efficacy estimates (shaded areas = 5 % and 95 % simulation percentiles) in the constructed (synthetic) data sets where the proportion of the test species in the population were set to A) 30 %, B) 20 %, C) 10 % and D) 2 %, and treatment efficacy was fixed at either 50, 70 or 90 % (triangle, square, circle).

Similar results were seen when using examples from the on-farm trials (Fig. 3, Fig. 4). These examples were chosen as they demonstrate the same patterns as the synthetic data but using data from real farms. When efficacy was low (0–60 %) the diagnosis was almost always ‘resistant’ but note some uncertainty for one species in Fig. 3 B at lower sample sizes (species representation pre-treatment was 2.6 %) where median efficacy was 60 % but some results were >95 %. In the other tests (Fig. 3, Fig. 4) where species representation was low (1.1–2.2 %) and efficacy was high, there was always uncertainty of diagnosis until the sample sizes increased (usually to >500).

Fig. 3.

The effect of number of larvae identified to species on the variation in efficacy estimates (shaded areas = 5 % and 95 % simulation percentiles) from simulations using real world parameters for faecal egg counts and the proportions of T. axei (circle), T. colubriformis (square) and T. vitrinus (triangle) in the population. A, B, C and D = different farms (Table 1) and anthelmintics. The proportions of T. axei on Farms A, B and D prior to treatment were 2.2 %, 2.6 % and 1.1 %, respectively; of T. colubriformis on Farms A, B and C prior to treatment were 49 %, 56 % and 18 %, respectively; and T. vitrinus on Farm C was 0.3 % prior to treatment.

Fig. 4.

– The effect of number of larvae identified to species on the variation in efficacy estimates (shaded areas = 5 % and 95 % simulation percentiles) from simulations using real world parameters for faecal egg counts and the proportions of Teladorsagia circumcincta in the population. A, B, C and D = different farms (Table 1) and anthelmintics. The proportions of T. circumcincta in the populations prior to treatment were A, 3 %, B, 9 %, C, 4 % and D, 25 %.

Also, as the number of larvae counted increased the variance of the efficacy estimate became smaller (Fig. 3, Fig. 4). Even if this didn't change the diagnosis of resistance, when the number of larvae counted was low, efficacy estimates could vary widely between samples (Fig. 3, Fig. 4). The decline in the variance of the estimates as the sample sizes increase shows that sampling error in larval identification is contributing substantially to the uncertainty around the efficacy estimate.

4. Discussion

For decades, the faecal egg count reduction test has been the primary method for testing on-farm anthelmintic efficacy, and hence, the detection of anthelmintic resistance. Initial recommendations on the method were made (Coles and Roush, 1992; McKenna, 1996) but improvements have continued to be made culminating in the most recent set of guidelines published in 2023 (Kaplan et al., 2023). The benefits of partitioning FECs to the genera of nematodes present in pre- and post-treatment samples has long been recognized (McKenna, 1997) but there has still been an issue with this in that larvae of some nematode species cannot readily be distinguished visually (Rossanigo and Gruner, 1996; Roeber and Kahn, 2014; Waghorn et al., 2014; Knoll et al., 2021; Maurizio et al., 2024). The results of this study indicate that even after apportioning FEC to genera using visual identification of larvae, the FECRT is likely to be underestimating the presence of resistant populations. In this study these ‘false negative’ test results occurred in 25 % of the tests which involved species complexes that cannot be separated visually. It can be concluded that incorporation of methods capable of identifying larvae to species will improve the ability of the FECRT to detect anthelmintic resistance.

Further, the results from this study may be underestimating the uncertainty of efficacy associated with visual identification of larvae. The approach used here effectively assumes that visual identification of L3 to species/genus is 100 % accurate i.e., by summing the proportions of related species identified using DNA. This approach was necessary to avoid uncontrolled sampling errors affecting the comparison of the two methods. However, visual identification of genera is not always reliable due to overlapping morphological and morphometric traits (Rossanigo and Gruner, 1996; Roeber and Kahn, 2014; Waghorn et al., 2014; Knoll et al., 2021; Maurizio et al., 2024). For example, separating Teladorsagia and Trichostrongylus spp. visually can be difficult due to overlap in length which often results in incorrect classification of some larvae (Waghorn et al., 2014). Hence, the differences in diagnostic potential between the two methods are likely greater than indicated by the current comparisons.

A further source of error associated with visual identification of L3 comes from the published criteria used to make the differentiations. Roeber and Kahn (2014) identified L3 of Trichostrongylus and Teladorsagia sourced from common pooled cultures using different sets of published criteria and found that the resulting proportions differed between criteria. The use of DNA-based methods should remove any variations associated with method or observer bias, resulting in more consistent results (Roeber and Kahn, 2014).

The proportion of eggs which develop successfully to L3 in faecal cultures is not the same for all nematode species, meaning that representation in a faecal culture does not necessarily equate to representation in the faecal egg count (Berrie et al., 1988; Dobson et al., 1992). For each species, development will undoubtedly be influenced by the optimal temperature and moisture requirements of that species (Leathwick, 2013) relative to the culture conditions (McKenna, 1998). However, faecal cultures and larval differentiation do indicate the species contributing to the egg counts, and with some bias, their relative contributions (Dobson et al., 1992). In the current study, this bias should be less significant because efficacy was calculated for each species independently, using the egg counts and the proportion of each species present in the cultures before and after treatment i.e., the comparisons were made within each species (a single source worm population) and so assuming that the culture conditions used were the same pre- and post-treatment, the proportion of eggs of each species developing to L3 should be the same or similar. Hence, any bias in development between the different species should have minimal impact on the efficacy calculations.

The primary purpose of a FECRT is to determine the presence of anthelmintic resistance, and the results show that speciating L3 will improve the test's ability to reliably achieve this. However, even when resistance has been diagnosed the estimated efficacy value is often of interest to the farmer/veterinarian and an interpretation is often put on this value. For example, an efficacy value of 90 % might be regarded differently by a farmer or veterinarian than an estimate of 10 %. The former may be interpreted as ‘mild’ resistance, with some ability to continue controlling infection using anthelmintics while the latter might be regarded as ‘severe’ necessitating no further use of the implicated active (McIntyre et al., 2018). Thus, the repeatability of an efficacy estimate is of some interest even if it doesn't change the diagnosis of resistance. The results show that if the number of L3 identified is low (50–100) an efficacy estimate of 10 % or 90 % could occur within the same test purely due to sampling error. Perhaps not surprisingly, the uncertainty around an efficacy estimate declines as the number of larvae identified to species increases. Thus, a significant advantage from use of nemabiome in FECRT would be the opportunity for large sample sizes, at little to no extra cost, which would greatly increase the confidence around efficacy estimates.

Practically, this is potentially useful when dealing with species which are poorly represented in the pre-treatment faecal cultures. It has been shown that accurately determining a classification of resistance is difficult when the observed efficacy is close to the threshold of 95 % reduction (Levecke et al., 2012, 2018; Denwood et al., 2023). In New Zealand there has long been a guideline that efficacy estimates for any species with <50 epg representation (i.e., FEC x proportion of species/genus present) prior to treatment in a test, is unreliable and should not be considered (McKenna, 1996, 1997). This has been problematic in New Zealand, especially for T. circumcincta, which whilst being the first species to show widespread resistance to anthelmintics, is also a species which is often poorly represented in faecal cultures (i.e., <5 %). The current analysis suggests that if methods such as nemabiome are incorporated into routine FECRT with speciation of large numbers of L3, then the 50 epg threshold for inclusion may be unnecessary, and evaluation of resistance in poorly represented species will become more reliable. This aspect of use of methods such as nemabiome warrants further investigation, especially as the benefit of large sampling size leading to tighter confidence intervals around efficacy estimates should apply to all nematode species, including those readily identified visually such as H. contortus. Based on the current results, if greater than 500–600 L3 are identified, estimating efficacies for all species will be improved, and therefore more useful. Given that 2000–3000 L3 can easily be included in nemabiome assays such an outcome seems very achievable.

In summary, this work indicates that routine adoption of species identification, using methods such as nemabiome, will improve the utility of the FECRT in two aspects. Firstly, false negative results in species complexes which are not readily discernible visually will decrease and detection of resistance earlier in the development phase is likely. Secondly, the ability to greatly increase sample size for speciating larvae will reduce levels of uncertainty around both efficacy estimates and diagnosis of resistance. This is likely to improve the test's ability to detect resistance in species which are poorly represented in larval cultures.

CRediT authorship contribution statement

Dave Leathwick: Writing – review & editing, Writing – original draft, Methodology. Peter Green: Methodology, Formal analysis, Data curation. Charlotte Bouchet: Methodology, Investigation. Alex Chambers: Writing – review & editing, Methodology, Investigation. Tania Waghorn: Writing – review & editing, Methodology, Investigation. Christian Sauermann: Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.