Ms

DA

ARRA

KAMFRRC

0h

Veterinary Parasitology 198 (2013) 145– 153

Contents lists available at ScienceDirect

Veterinary Parasitology

jo u r nal homep age: www.elsev ier .com/ locate /vetpar

anaging anthelmintic resistance – Parasite fitness, drug usetrategy and the potential for reversion towards susceptibility

ave M. Leathwick ∗

gResearch Grasslands, Private Bag 11008, Palmerston North 4442, New Zealand

a r t i c l e i n f o

rticle history:eceived 3 January 2013eceived in revised form 6 August 2013ccepted 19 August 2013

eywords:nthelmintic resistanceodelling

itnesseversionotationombination

a b s t r a c t

The rotation of different anthelmintic classes, on an approximately annual basis, has beenwidely promoted and adopted as a strategy to delay the development of anthelmintic resis-tance in nematode parasites. Part of the rationale for recommending this practice was theexpectation that resistant genotype worms have a lower ecological fitness than suscepti-ble worms, at least in the early stages of selection, and so reversion towards susceptibilitycould be expected in those years when an alternative class of anthelmintic was used.

The routine use of combination anthelmintics might be expected to negate this opportu-nity for reversion because multiple classes of anthelmintic would be used simultaneously. Asimulation model was used to investigate whether the optimal strategy for use of multipledrug classes (i.e. an annual rotation of two classes of anthelmintic or continuous use of twoclasses in combination) changed with the size of the fitness cost associated with resistance.

Model simulations were run in which the fitness cost associated with each resistancegene was varied from 0% to 15% and the rate at which resistance developed was comparedfor each of the drug-use strategies. Other factors evaluated were the initial frequency ofthe resistance genes and the proportion of the population not exposed to treatment (i.e. inrefugia).

Increasing the proportion of the population in refugia always slowed the development ofresistance, as did using combinations in preference to an annual rotation. As the fitness costassociated with resistance increased, resistance developed more slowly and this was morepronounced when a combination was used compared to a rotation. If the fitness cost wassufficiently high then resistance did not develop (i.e. the resistance gene frequency declinedover time) and this occurred at lower fitness costs when a combination was used. Theresults, therefore, indicate that the optimal drug-use strategy to maximise the benefit of anyfitness cost associated with resistance is the use of combinations of different anthelminticclasses.

Manual calculations confirmed that, within the model, the only resistant genotypes capa-ble of surviving treatment with a combination are those carrying multiple resistance genes.

These individuals are less fit, resulting in the worm population surviving treatment hav-ing a lower overall ecological fitness. This is a previously unreported perspective on theuse of combination anthelmintics and strengthens the argument that any new class ofanthelmintic, for which resistance genes can be expected to be rare, should be brought tomarket in combination.

∗ Tel.: +64 6 3518085; fax: +64 6 3518134.E-mail address: dave.leathwick@agresearch.co.nz

304-4017/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.vetpar.2013.08.022

© 2013 Elsevier B.V. All rights reserved.

1. Introduction

In many countries anthelmintic resistance is a sig-nificant problem amongst nematode parasites of goats

y Parasit

146 D.M. Leathwick / Veterinar

and sheep (Besier, 2007) and is of growing concernamongst parasites of cattle (Sutherland and Leathwick,2011) and horses (Kaplan, 2004). Considerable effort hasbeen invested in understanding the processes involved inthe selection of anthelmintic resistance (Barnes et al., 1995;Leathwick et al., 2009) so that control strategies can bedeveloped which minimise both the impact of parasites onanimal health and welfare and the development of resis-tance.

Because anthelmintics are an integral component ofmost worm control programmes, much attention has beengiven to optimising their use, both in terms of parasite con-trol (Vlassoff and Brunsdon, 1981; Barger, 1999) and themanagement of resistance (Barnes et al., 1995; Leathwickand Hosking, 2009; Dobson et al., 2011). For many yearsrecommendations promoted the use of strategic timing oftreatments (Vlassoff and Brunsdon, 1981; Barger, 1999)combined with rotation of different anthelmintic classeson an approximately annual basis (Donald et al., 1980;Prichard et al., 1980; Waller et al., 1989; Coles and Roush,1992). A perceived benefit of a slow rotation of drugsfrom different classes was that by not exposing individ-ual worms to more than one class of drug, selection forresistance to multiple actives was less likely to occur(Prichard et al., 1980). In addition, if resistant genotypeshave a reduced ecological fitness compared to suscepti-ble genotypes, then some reversion towards susceptibilitywould be expected in those periods when an alternatedrug class was used (Prichard et al., 1980; Roush andMcKenzie, 1987; Coles and Roush, 1992). The annual rota-tion of different anthelmintic classes was widely promoted,and the practice was adopted by many farmers, in NewZealand (Kettle et al., 1983; Lawrence et al., 2007) and inother countries (Waller et al., 1989; Sargison et al., 2007;McMahon et al., 2013).

Recently there has been increased interest in the use ofcombinations of different anthelmintic classes, both as ameans of maintaining control in the presence of resistantnematodes (Leathwick et al., 2009), and as a means of slow-ing the development of resistance (Smith, 1990; Barneset al., 1995; Dobson et al., 2011; Leathwick, 2012). Combi-nation products are readily available and used extensivelyin some countries, but not in others (Bartram et al., 2012). Anumber of modelling studies have indicated that combin-ing anthelmintics from different classes, especially wherethese still have high efficacy, is likely to result in a signif-icant delaying of resistance development (Dobson et al.,2011; Leathwick, 2012; Learmount et al., 2012). However,use of combinations will naturally reduce the opportu-nity for rotation, potentially interfering with any reversiontowards susceptibility, in the presence of a fitness costassociated with resistance. To date, no modelling studieshave fully investigated the interactions between drug-usestrategies (i.e., rotation of single active products versesuse of multiple actives in combination) and any potentialreductions in fitness associated with resistance genes innematodes.

The purpose of this study was to address this issueby investigating the potential influence of resistance-associated fitness costs, of varying magnitudes, on the beststrategy for use of more than one anthelmintic class.

ology 198 (2013) 145– 153

2. Methods

A previously developed simulation model was used toinvestigate the influence of three variables and two drug-use strategies on the rate at which anthelmintic resistancedeveloped. The variables investigated were the magnitudeof the fitness cost associated with resistance, the initialfrequency of resistance genes in the population, and theproportion of the worm population which escaped expo-sure to each treatment (i.e. those in refugia). The impactof these variables under either an annual rotation of twodrugs or continuous use of a combination of two drugs wasassessed.

2.1. The model

The model has been described previously (Leathwicket al., 1992, 1995) and used to investigate both parasitedynamics and the development of resistance under dif-ferent management practices (Leathwick and Sutherland,2002; Leathwick et al., 2008; Leathwick and Hosking, 2009;Leathwick, 2012). Briefly, the model is generic in that it sim-ulates a mixed nematode infection, rather than infectionby individual species (Leathwick et al., 1992), and incor-porates both lamb and adult ewe flocks. The developmentand survival of free-living nematode stages on pasture isdescribed for each of a suite of paddocks. When a pad-dock is being grazed by ewes or lambs it receives inputs ofnematode eggs, based on the faecal egg output of an indi-vidual multiplied by the stocking rate, and infective stagelarvae are removed with ingested herbage. The numberof ingested larvae determines subsequent worm burdenand faecal nematode egg count (FEC) under the influenceof anti-parasite immunity which develops in response toboth ingestion of infective larvae and the presence of adultworms.

Importantly, outputs from this model have previouslybeen evaluated and found to be consistent with the resultsof large-scale field trials with respect to both the devel-opment of anthelmintic resistance (Leathwick et al., 2006,2012) and nematode epidemiology (Leathwick et al., 2008).

2.2. Genetics of resistance

For the purpose of this study the model was used toconsider two anthelmintic drugs which had independentmodes of action and mechanisms of resistance, and forwhich the anthelmintic effect was additive (Bartram et al.,2012). Resistance to each drug was assumed to be theresult of a single mutation resulting in three genotypes,a homozygous susceptible (SS), a heterozygote (RS) and ahomozygous resistant (RR), giving a total of nine genotypesoverall (Table 1). It was assumed that resistance was par-tially recessive under treatment with either drug, leadingin each case to mortalities under treatment of 99.9, 50.0 and

5.0% for the SS, RS and RR genotypes, respectively. The ini-tial frequency of resistance genes (q) was varied across setsof simulations in order to assess the effect of this variableon model outcomes.

D.M. Leathwick / Veterinary Parasitology 198 (2013) 145– 153 147

Table 1The fitness values allocated to different worm genotypes in the model under the different assumed fitness costs associated with being resistant.

Genotype Fitness relative to the SS SS genotype

Drug 1 Drug 2 Fit = 0 Fit = 1 Fit = 2 Fit = 5 Fit = 10 Fit = 15

SS SS 1.0 1.0 1.0 1.0 1.0 1.0SS RS 1.0 0.99 0.98 0.95 0.9 0.85SS RR 1.0 0.98 0.96 0.9 0.8 0.7RS SS 1.0 0.99 0.98 0.95 0.9 0.85RS RS 1.0 0.98 0.96 0.9 0.8 0.7RS RR 1.0 0.97 0.94 0.85 0.7 0.55

2

agnmwtewgvag

2

Zapwactnat1swl

ooabF1Ctrtae

RR SS 1.0 0.98

RR RS 1.0 0.97

RR RR 1.0 0.96

.3. Fitness costs

It was assumed that each R gene carried a fitness dis-dvantage of 0, 1, 2, 5, 10 or 15% compared to the SSenotype, and that the disadvantages were additive as theumber of R genes increased (Table 1). This was imple-ented as a reduction in the fecundity of adult femaleorms. For example, if the fitness disadvantage was 5%,

hen an SS SR genotype worm would produce 5% fewerggs than an SS SS genotype worm, while an SS RR genotypeorm would produce 10% fewer. Thus, the more resistance

enes an individual carried, the greater its ability to sur-ive anthelmintic treatment but the lower its fitness in thebsence of anthelmintic treatment. The complete array ofenotype fitness values used is presented in Table 1.

.4. Management and refugia

A management scenario typical of sheep farming in Newealand was used as the basis for all simulations. Ewesnd their lambs were set-stocked over all paddocks fromre-lambing until weaning. Lambing occurred in August,eaning in November when lambs were 12 weeks of age,

nd lamb numbers were progressively reduced thereafteronsistent with a farmer selling a proportion of his lambso slaughter as they reached target weights. Ewes wereot treated with anthelmintic at any time. Commencingt weaning, all lambs received a programme of six preven-ive treatments at 28 day intervals (Vlassoff and Brunsdon,981) which is a widely used practice on New Zealandheep farms (Lawrence et al., 2007). After weaning lambsere rotationally grazed around all paddocks and were fol-

owed by a mob of ewes.In order to simulate different levels of refugia with-

ut significantly altering the grazing management, the FECf the ewes was varied in the model both before andfter weaning. Thus three refugia scenarios were createdy altering the size of the post-parturient rise in eweEC from a maximum value of 250 epg (low refugia) to000 epg (medium refugia) and 1750 epg (high refugia).oncurrently, ewe faecal egg count post-weaning was seto constant values of 25 epg (low refugia), 100 epg (medium

efugia) and 175 epg (high refugia). This effectively variedhe inputs of nematode eggs which had not been exposed tonthelmintic treatment without altering any other param-ters which may have affected outputs. The parasitological

0.96 0.9 0.8 0.70.94 0.85 0.7 0.550.92 0.8 0.6 0.4

consequences of these manipulations on model output areoutlined in Fig. 1.

2.5. Simulations and outputs

Initially, simulations were run which compared all theassumed fitness costs (i.e. 0, 1, 2, 5, 10 and 15%) togetherwith the three levels of refugia (low, medium and high) ateach of three initial gene frequencies (0.01, 0.05 and 0.1).These comparisons were repeated for the two drug-usestrategies i.e. an annual rotation of two drugs or the contin-uous use of two drugs in combination. In order to simplifythe presentation of outputs all variables associated withthe two anthelmintic actives were set to the same valuesfor any given simulation. For example, if the starting resis-tance gene frequency for drug 1 was set at 0.1 then the samevalue was used for drug 2. In all cases the output used tocompare simulations was the resistant gene frequency fordrug 1 after the model was run for 40 years (i.e. only theq-value for drug 1 in year 40 is presented).

Subsequently, in order to more fully explore the influ-ence of the initial q-values on model output, a second seriesof simulations was run in which only the medium refugiascenario was used. For these simulations the final q-valuefor drug 1 was plotted against the initial q-value and arange of assumed fitness costs were compared. Becausethe parameter values were set the same for both drugs, theresulting final q-values for drugs 1 and 2 were identical forthe combination and very similar for the rotation. Becausedrug 1 was always used in year 1 of the annual rotationit was appropriate to compare this with the combinationafter 40 years – effectively drug 2 had only been exposedto selection within the rotation for 39 years.

2.6. Calculation of average post-treatment fitness

In order to better understand the mechanism respon-sible for some of the models outputs, the average fitnessof the worm population surviving different drench treat-ments was calculated manually, as;

y =(

9∑Fi · fi · si

)/

(9∑

fi · si

)

i=1 i=1

where Fi = the relative fitness of genotype i, fi = thefrequency of genotype i at the time of anthelmintictreatment assuming Hardy–Weinberg equilibrium, and

148 D.M. Leathwick / Veterinary Parasitology 198 (2013) 145– 153

(a)

(b)

(c)

050

100150200250300350400450500

0 4 8 12 16 20 24 28 32 36 40 44 48 52

Epg

Weeks post-lambing

high refugiamedium refugialow refugia

0200400600800

100012001400160018002000

0 4 8 12 16 20 24 28 32 36 40 44 48 52

Epg

Weeks post-lambing

0

200

400

600

800

1000

1200

0 4 8 12 16 20 24 28 32 36 40 44 48 52

L3/k

g

Weeks post-lambing

Fig. 1. Model outputs for faecal nematode egg count of lambs (a) and adult ewes (b) and number of infective stage larvae on herbage (c) when egg outputsof the ewes were varied (see text) to create high, medium and low refugia scenarios.

D.M. Leathwick / Veterinary Parasit

Table 2The calculated average fitness of those worms surviving treatment witheither a single active anthelmintic or a combination of two actives, whenthe initial q-value is 0.1, the fitness cost associated with a resistance geneis either 0, 1, 2, 5 or 10%, and the RS genotype is either equal in fitnessto the SS (SS = RS > RR), the RR (SS > RS = RR), or intermediate between thetwo (SS > RS > RR).

Initial R genefrequency

Fitnesscost (%)

Mean fitness relative to SS SS

Single drug Combination

SS > RS > RR0.1 0 1 10.1 1 0.9878 0.97970.1 2 0.9757 0.95950.1 5 0.9393 0.89870.1 10 0.8787 0.7974

SS > RS = RR0.1 0 1 10.1 1 0.9889 0.98150.1 2 0.9777 0.96300.1 5 0.9442 0.90750.1 10 0.8885 0.8151

SS = RS ≥ RR0.1 0 1 10.1 1 0.9990 0.9982

safatasbfimp

2

chnRTradtwcembSpono

0.1 2 0.9980 0.99650.1 5 0.9951 0.99120.1 10 0.9902 0.9823

i = the proportion of genotype i individuals surviving thenthelmintic treatment. This was set up in a spread sheetormat (Microsoft Excel) so that for any defined q-valuend fitness cost the proportion of individuals of each geno-ype surviving a treatment (with both a single active and

combination), and the average fitness of the populationurviving treatment, was calculated. By changing two num-ers in the spread sheet (i.e. the initial q-value and thetness cost), the calculation corresponding to any of theodel simulations could be reproduced. Examples of out-

ut are presented in Table 2.

.7. Fitness cost in relation to genotype

There are several different ways in which a fitness costould potentially be linked to resistant genotypes. Here, itas been assumed that each resistance gene carries a fit-ess cost and that these are additive (i.e. SS is more fit thanS which is more fit than RR; represented as SS > RS > RR).his might reasonably approximate the situation whereesistance is intermediate in dominance under treatmentnd the SS, RS and RR genotypes are all phenotypicallyifferent in the presence of drug. Another possibility ishat the SS and RS are equal in terms of ecological fitness,hich would be consistent with the resistance gene being

ompletely recessive i.e. the resistance mechanism is notxpressed in the heterozygote. Alternatively, the RS and RRay be equal, consistent with the resistance mechanism

eing fully dominant. These options can be represented asS = RS > RR or SS > RS = RR. To ensure that these alternative

ossibilities would not negate the effects seen in modelutput, manual calculations of average post-treatment fit-ess were carried out for all three options, under a rangef initial q-values and fitness costs.

ology 198 (2013) 145– 153 149

3. Results

Increasing FEC of the ewes slowed the development ofresistance regardless of the fitness cost of resistance, theinitial q-value or the drug-use strategy (Fig. 2). Thus, inall cases the larger the proportion of the adult, reproduc-tive, worm population which was not exposed to treatment(i.e. those in refugia) the slower the development of resis-tance.

Furthermore, resistance always developed more slowlywhen a combination was used compared to an annualrotation (Figs. 2 and 3) and this occurred regardless ofany fitness cost associated with resistance. If resistantgenotypes were less fit than susceptibles then resistancedeveloped more slowly, and the greater the fitness cost themore resistance development was slowed (Figs. 2 and 3).Indeed if the fitness cost was sufficiently large thenresistance did not develop at all, with the q-value declin-ing over time despite routine anthelmintic treatments(Figs. 2 and 3). This occurred under both drug-usestrategies, although it was more pronounced when a com-bination was used than under an annual rotation i.e. whena combination was used resistance failed to develop if thefitness cost was 10% or 15% and in some simulations whenthe fitness cost was 5% (Figs. 2 and 3). If the fitness costfor each resistance gene was 5%, resistance only devel-oped under use of a combination if the initial q-value wasgreater than about 0.17. In contrast, under an annual rota-tion, resistance always developed unless the fitness cost perresistance gene was 15% and there was at least a mediumlevel of refugia i.e. under low refugia resistance alwaysdeveloped even when the fitness cost per resistance genewas 15%.

Manual calculations were used to elucidate the mecha-nism underpinning the difference between the drug-usestrategies. These calculations indicated that, except forthe case where there was no fitness cost associated withresistance, the average fitness of those parasites survivingtreatment with a single active was always higher than thatfollowing use of a combination (some examples are pre-sented in Table 2). This reflects the greater efficacy of thecombination against genotypes carrying low numbers ofresistance genes, and which were therefore, more fit. Forexample, using the values outlined above, efficacy of a sin-gle active against an SS SR genotype would be 50% for oneof the drugs, but the combination would remove 99.95%of these individuals every treatment. These calculationsdemonstrate that the combination would not only leavefewer resistant individuals surviving treatment but thesewould, on average, be less fit than the resistant individualssurviving treatment with a single drug.

What is more, this outcome occurred regardless ofwhether the RS genotype was equal in fitness to the SS (i.e.SS = RS > RR), was equal in fitness to the RR (i.e. SS > RS = RR)or was intermediate in fitness (i.e. SS > RS > RR) (Table 2).Although the magnitude of differences varied betweenthese scenarios (Table 2), in all cases the average fitness

of those worms which survived treatment remained lowerafter use of a combination than a single active, indicatingthat the overall results of the modelling are unlikely to beinfluenced by the assumption of additive fitness costs.

150 D.M. Leathwick / Veterinary Parasitology 198 (2013) 145– 153

a) b)

c) d)

e) f)

0

0.05

0.1

0.15

0.2

0.25

0.3

0 1 2 5 10 15

Fina

l q v

alue

Fitness disadvantage ( %)

Low refugia

Medium refugia

High refugia

0

0.05

0.1

0.15

0.2

0.25

0.3

0 1 2 5 10 15

Fina

l q v

alue

Fitness disadvantage ( %)

0

0.2

0.4

0.6

0.8

1

0 1 2 5 10 15

Fina

l q v

alue

Fitness disadvantage ( %)

0

0.2

0.4

0.6

0.8

1

0 1 2 5 10 15

Fina

l q v

alue

Fitness disadvantage ( %)

00.10.20.30.40.50.60.70.80.9

0 1 2 5 10 15

Fina

l q v

alue

Fitness disadvantage ( %)

00.10.20.30.40.50.60.70.80.9

0 1 2 5 10 15

Fina

l q v

alue

Fitness disadvantage ( %)

lection, (b, d anline reflrds susc

Fig. 2. Resistance gene frequency (q-value) for drug 1 after 40 years of setwo drugs were used in an annual rotation (a, c and e) or in combination0.1 (e and f). The horizontal line is the initial q-value so values above this

while those below the line reflect an overall decrease (i.e. reversion towa

4. Discussion

Numerous modelling studies have indicated that inmost situations the development of anthelmintic resis-tance will be delayed by the use of combinations comparedto any single drug-use strategy (Smith, 1990; Barnes et al.,1995; Dobson et al., 2011; Leathwick, 2012; Learmountet al., 2012) and recently this conclusion has been validatedin the field (Leathwick et al., 2012). The mechanism bywhich combinations are expected to delay the develop-

ment of resistance involves an increased efficacy againstresistant genotypes, principally those only carrying genesfor resistance to one of the actives in the combination.The result is fewer resistant survivors of treatment and

under a range of fitness disadvantages associated with resistance, whend f) and where the initial q-value was 0.01 (a and b), 0.05 (c and d) and

ect an overall increase in gene frequency (i.e. development of resistance)eptibility).

a consequently greater dilution of those survivors bysusceptible genotypes in ‘refugia’ (Smith, 1990).

However, none of these earlier modelling studies fullyinvestigated the potential influence of different fitnesscosts associated with resistance mechanisms on the devel-opment of resistance under different drug use strategies.The results of the current study suggest that if there is afitness cost associated with resistance the use of combina-tions is likely to be even more advantageous, compared toan annual rotation, than when there is no fitness cost.

The basis for this is that where each resistance mech-anism (i.e. to each anthelmintic class) has an associatedfitness cost, the use of combinations is likely to reducethe average fitness of those worms surviving treatment,

D.M. Leathwick / Veterinary Parasit

a)

b)

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 0.05 0.1 0.15 0.2 0.25

Fina

l q-v

alue

Ini�al q-value

fit=0fit=2fit=5fit=10fit=151:1 line

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0 0.05 0.1 0.15 0.2 0.25

Fina

l q-v

alue

Ini�al q-value

Fig. 3. Resistance gene frequency (q) for drug 1 after 40 years of selection,under a range of different levels of fitness costs (%) associated with resis-tance (fit), when two drugs were used in combination (a) or in an annualrotation (b) plotted against the gene frequency at the start of each sim-ulation. Note that the 1:1 line represents the situation where q remainstfr

bwtt

roEcaMdpta(tgti

he same and so points above this line reflect an overall increase in generequency (i.e. the development of resistance) while those below this lineeflect an overall decrease (i.e. reversion towards susceptibility).

ecause the only individuals which can survive are thosehich possess multiple resistance genes and are therefore

he least fit. This is a previously unreported perspective onhe use of combination anthelmintics.

A comprehensive review of selection for anthelminticesistance and the relationship with ecological fitnessf resistant genotypes can be found in Prichard (1990).ssentially, experiments to demonstrate a fitness cost asso-iated with anthelmintic resistance have provided variablend sometimes contradictory results (Donald et al., 1980;aclean et al., 1987; Waller et al., 1989). However, it is

ifficult to interpret these findings as in many cases thearasites compared have been isolated independently fromhe field and so may differ in aspects of their epidemiologys a result of adaptation to host and microclimatic factorsKelly et al., 1978). Despite this, it is reasonable to assume

hat when an anthelmintic is first introduced, resistanceenes, should they exist in a population, will be rare or elsehe anthelmintic would not be effective. Individuals carry-ng these genes are unlikely to be more fit, in the absence

ology 198 (2013) 145– 153 151

of anthelmintic selection, than individuals carrying equiv-alent genes for susceptibility or else resistant genes wouldbe more common in the population, and their rarity there-fore suggests some degree of fitness disadvantage (Martin,1987; Roush and McKenzie, 1987; Prichard, 1990). It hasbeen proposed that in a fully susceptible (naive) pop-ulation susceptible genotypes have a fitness advantageover those carrying resistance mechanisms (Martin et al.,1988; Prichard, 1990). Following the introduction of a newanthelmintic class the resistant genotypes are selected forbut, in the absence of the drug, they retain a lower biolog-ical fitness. However, as selection continues the resistancegenes re-associate (co-adapt) with other fitness character-istics and so the fitness differential between resistant andsusceptible genotypes becomes smaller over time. Hence,once the resistance genes have become common in thepopulation, and resistance is detectable based on efficacy oftreatment, then any fitness disadvantage may be reducedor negligible (Martin et al., 1988; Prichard, 1990). This veryreasonable hypothesis would be consistent with the initialrarity of resistance genes but the variability of demonstrat-ing a fitness cost in resistant isolates (Donald et al., 1980;Waller et al., 1989).

Support for this concept was given by several studies(Kelly et al., 1978; Maingi et al., 1990) in which strains ofHaemonchus contortus showing moderate levels of resis-tance to benzimidazole anthelmintics were subjected tofurther selection in the laboratory. In both cases, furtherselection resulted not only in increasing levels of resis-tance, but also to increasing ecological fitness, indicatingthat fitness of the parasites can indeed increase with on-going selection for resistance (Kelly et al., 1978; Maingiet al., 1990). While, it is likely that this co-adaptation ofresistance and fitness traits occurred more rapidly underlaboratory selection than is likely in the field, because of thevery low refugia of susceptibility (i.e. the opportunity forresistance and fitness genes to align was high) these stud-ies do demonstrate that co-adaptation can occur. It alsofollows that, in the field, refugia will be important in therate of co-adaptation of resistance and fitness-modifyinggenes, in that co-adaptation is likely to occur more rapidlyunder low-refugia conditions. This again emphasises theimportance of maintaining a sizeable pool of unselectedgenotypes in order to slow the development of resistance.

A fitness cost associated with resistance is therefore areasonable assumption, even if it only occurs in the earlystages of selection (i.e. when resistance genes are still rare)(Martin et al., 1988). In some studies, the magnitude ofthe fitness costs associated with resistance has been quitelarge. For example, a benzimidazole resistant isolate ofTrichostrongylus colubriformis showed a 49% reduction inestablishment of infective larvae in gerbils, compared toa susceptible isolate (Maclean et al., 1987), while a benz-imidazole resistant H. contortus showed a 44% reductionin establishment rate and a 45% lower egg viability than asusceptible isolate (Maingi et al., 1990). In another study,the proportion of eggs from a resistant isolate which

developed into infective stage larvae was fewer than halfthe number from a susceptible isolate (Scott and Armour,1991). It must be recognised, however, that while some ofthese differences are large, they are differences in just one

y Parasit

152 D.M. Leathwick / Veterinar

variable. Many variables contribute to the overall fitnessof an organism, which is best represented as the BasicReproductive Rate i.e. the number of eggs produced by asingle female parasite that themselves survive to repro-duce in the absence of density-dependence (Anderson andMay, 1982; Maingi et al., 1990). Hence, extrapolating fromdifferences in a single variable to an overall fitness costmay be misleading. However, given that some of thesemeasured differences are considerably larger than thefitness costs assumed in the current study, the values usedhere are probably not unreasonable.

If each resistance mechanism has an associated fitnesscost, then worms carrying multiple resistance mechanismsare likely to be even less fit. In the current study, all sim-ulations assumed that the fitness costs associated withbeing simultaneously resistant to multiple anthelminticsare additive. In all cases the model outputs indicate thatreduced fitness associated with resistance mechanisms islikely to slow the development of resistance, and in somecases may prevent resistance from developing at all. Thiswill undoubtedly be a balance between the magnitude ofthe fitness cost, and the intensity of the selection pressurefor resistance. This can be seen in the model outputs wherethe development of resistance is slowed by both the fitnesscost of being resistant and the proportion of the popula-tion which is not exposed to treatment (which influencesselection pressure) (Fig. 2).

The issue of drug-use strategy has risen in importancewith the recent release of two new classes of anthelminticsfor use in sheep. One of these new actives, monepantel,has been released as a single active product (Kaminskyet al., 2008) while the other, derquantel, is only availablein combination with abamectin (Little et al., 2011). Pre-sumably, genes for resistance to both these new activesare currently rare and so if there is any fitness cost asso-ciated with resistance to them it will be at its maximumprior to any significant selection. Clearly, the optimum timeto take advantage of any such cost is immediately after anew active comes onto the market and before resistancehas the opportunity to develop. The results of the cur-rent study indicate that the optimum approach to usingany new actives sustainably is to formulate them as com-binations. Not only are combinations likely to minimisethe survival of resistant genotypes thereby slowing thedevelopment of resistance (Barnes et al., 1995; Leathwick,2012; Learmount et al., 2012) but, based on the cur-rent analysis, the use of combinations will also maximisethe benefit of any fitness disadvantage associated withresistance.

Acknowledgements

Stewart Bisset, Ian Sutherland, Richard Scott and twoanonymous referees made helpful comments on a draftmanuscript. This work was supported by funding fromAgResearch.

References

Anderson, R.M., May, R.M., 1982. Population dynamics of human helminthinfections: control by chemotherapy. Nature 297, 557–563.

ology 198 (2013) 145– 153

Barger, I.A., 1999. The role of epidemiological knowledge and grazing man-agement for helminth control in small ruminants. Int. J. Parasitol. 29,41–47.

Barnes, E.H., Dobson, R.J., Barger, I.A., 1995. Worm control andanthelmintic resistance: adventures with a model. Parasitol. Today11, 56–63.

Bartram, D.J., Leathwick, D.M., Taylor, M.A., Geurden, T., Maeder, S.J., 2012.The role of combination anthelmintic formulations in the sustainablecontrol of sheep nematodes. Vet. Parasitol. 186, 151–158.

Besier, R.B., 2007. New anthelmintics for livestock: the time is right. TrendsParasitol. 23, 21–24.

Coles, G.C., Roush, R.T., 1992. Slowing the spread of anthelmintic resistantnematodes of sheep and goats in the United Kingdom. Vet. Rec. 130,505–510.

Dobson, R.J., Hosking, B.C., Besier, R.B., Love, S., Larsen, J.A.W., Rolfe, P.F.,Bailey, J.N., 2011. Minimising the development of anthelmintic resis-tance, and optimising the use of the novel anthelmintic monepantel,for the sustainable control of nematode parasites in Australian sheepgrazing systems. Aus. Vet. J. 89, 160–166.

Donald, A.D., Waller, P.J., Dobson, R.J., Axelsen, A., 1980. The effect of selec-tion with levamisole on benzimidazole resistance in Ostertagia spp. ofsheep. Int. J. Parasitol. 10, 381–389.

Kaminsky, R., Ducray, P., Jung, M., Clover, R., Rufener, L., Bouvier, J.,Schorderet Weber, S., Wenger, A., Wieland-Berghausen, S., Goebel,T., Gauvry, N., Pautrat, F., Skripsky, T., Froelich, O., Komoin-Oka, C.,Westlund, B., Sluder, A., Mäser, P., 2008. A new class of anthelminticseffective against drug-resistant nematodes. Nature 452, 176–180.

Kaplan, R.M., 2004. Drug resistance in nematodes of veterinary impor-tance: a status report. Trends Parasitol. 20, 477–481.

Kelly, J.D., Whitlock, H.V., Thompson, H.G., Hall, C.A., Martin, I.C.A., Le Jam-bre, L.F., 1978. Physiological characteristics of free-living and parasiticstages of strains of Haemonchus contortus, susceptible or resistant tobenzimidazole anthelmintics. Res. Vet. Sci. 25, 376–385.

Kettle, P.R., Vlassoff, A., Reid, T.C., Hotton, C.T., 1983. A survey of nematodecontrol measures used by milking goat farmers and of anthelminticresistance on their farms. N. Z. Vet. J. 31, 139–143.

Lawrence, K.E., Leathwick, D.M., Rhodes, A.P., Jackson, R., Heuer, C., Pom-roy, W.E., West, D.M., Waghorn, T.S., Moffat, J.R., 2007. Management ofgastrointestinal nematode parasites on sheep farms in New Zealand.N. Z. Vet. J. 55, 228–234.

Learmount, J., Taylor, M.A., Bartram, D.J., 2012. A computer sim-ulation study to evaluate resistance development with aderquantel–abamectin combination on UK sheep farms. Vet.Parasitol. 187, 244–253.

Leathwick, D.M., 2012. Modelling the benefits of releasing a new class ofanthelmintic in combination. Vet. Parasitol. 186, 93–100.

Leathwick, D.M., Hosking, B.C., 2009. Managing anthelmintic resistance:modelling strategic use of a new anthelmintic class to slow thedevelopment of resistance to existing classes. N. Z. Vet. J. 57,203–207.

Leathwick, D.M., Sutherland, I.A., 2002. Heads or Tails – which drench doI choose? In: Proc. 32nd Seminar, The Soc. of Sheep and Beef CattleVet., N. Z. Vet. Assoc, pp. 115–127.

Leathwick, D.M., Barlow, N.D., Vlassoff, A., 1992. A model for nematodiasisin New Zealand lambs. Int. J. Parasitol. 22, 789–799.

Leathwick, D.M., Vlassoff, A., Barlow, N.D., 1995. A model for nematodi-asis in New Zealand lambs: The influence of drenching and grazingmanagement on the development of anthelmintic resistance. Int. J.Parasitol. 25, 1479–1491.

Leathwick, D.M., Hosking, B.C., Bisset, S.A., McKay, C.H., 2009. Managinganthelmintic resistance: is it feasible in New Zealand to delay theemergence of resistance to a new anthelmintic class? N. Z. Vet. J. 57,181–192.

Leathwick, D.M., Waghorn, T.S., Miller, C.M., Candy, P.M., Oliver,A-M.B., 2012. Managing anthelmintic resistance – use of a com-bination anthelmintic and leaving some lambs untreated to slowthe development of resistance to ivermectin. Vet. Parasitol. 187,285–294.

Leathwick, D.M., Miller, C.M., Atkinson, D.S., Haack, N.A., Waghorn, T.S.,Oliver, A.-M., 2008. Managing anthelmintic resistance: untreatedadult ewes as a source of unselected parasites, and their role in reduc-ing parasite populations. N. Z. Vet. J. 56, 184–195.

Leathwick, D.M., Miller, C.M., Atkinson, D.S., Haack, N.A., Alexander, R.A.,Oliver, A.-M., Waghorn, T.S., Potter, J.F., Sutherland, I.A., 2006. Drench-

ing adult ewes: implications of anthelmintic treatments pre- andpost-lambing on the development of anthelmintic resistance. N. Z.Vet. J. 54, 297–304.

Little, P.R., Hodge, A., Maeder, S.J., Wirtherle, N.C., Nicholas, D.R., Cox,G.G., Conder, G.A., 2011. Efficacy of a combined oral formulation of

Parasit

M

M

M

M

M

P

P

D.M. Leathwick / Veterinary

derquantel–abamectin against the adult and larval stages of nema-todes in sheep, including anthelmintic resistant strains. Vet. Parasitol.181, 180–193.

aclean, J.M., Lewis, D., Holmes, P.H., 1987. The pathogenesis ofbenzimidazole-resistant and benzimidazole-susceptible strains ofTrichostrongylus colubriformis in the Mongolian gerbil (Merionesunguiculatus). J. Helminthol. 61, 179–189.

aingi, N., Scott, M.E., Prichard, R.K., 1990. Effect of selection pressure forthiabendazole resistance on fitness of Haemonchus contortus in sheep.Parasitology 100, 327–335.

artin, P.J., 1987. Development and control of resistance to anthelmintics.Int. J. Parasitol. 17, 493–501.

artin, P.J., Anderson, N., Brown, T.H., Miller, D.W., 1988. Changes in resis-tance of Ostertagia spp. to thiabendazole following natural selectionor treatment with levamisole. Int. J. Parasitol. 18, 333–340.

cMahon, C., Barley, J.P., Edgar, H.W.J., Ellison, S.E., Hanna, R.E.B., Mal-one, F.E., Brennan, G.P., Fairweather, I., 2013. Anthelmintic resistancein Northern Ireland (II): variations in nematode control practicesbetween lowland and upland sheep flocks. Vet. Parasitol. 192,173–182.

richard, R.K., 1990. Anthelmintic resistance in nematodes: extent, recent

understanding and future directions for control and research. Int. J.Parasitol. 20, 515–523.

richard, R.K., Hall, C.A., Kelly, J.D., Martin, I.C.A., Donald, A.D., 1980. Theproblem of anthelmintic resistance in nematodes. Aust. Vet. J. 56,239–251.

ology 198 (2013) 145– 153 153

Roush, R.T., McKenzie, J.A., 1987. Ecological genetics of insec-ticide and acaricide resistance. Ann. Rev. Entomol. 32,361–380.

Sargison, N.D., Jackson, F., Bartley, D.J., Wilson, D.J., Stenhouse, L.J., Penny,C.D., 2007. Observations on the emergence of multiple anthelminticresistance in sheep flocks in the south-east of Scotland. Vet. Parasitol.145, 65–76.

Scott, E.W., Armour, J., 1991. Effect of development of resis-tance to benzimidazoles, salicylanilides and ivermectin on thepathogenicity and survival of Haemonchus contortus. Vet. Rec. 128,346–349.

Smith, G., 1990. A mathematical model for the evolution of anthelminticresistance in a direct life cycle nematode parasite. Int. J. Parasitol. 20,913–921.

Sutherland, I.A., Leathwick, D.M., 2011. Anthelmintic resistance in nema-tode parasites of cattle – a global issue? Trends Parasitol. 27,176–181.

Vlassoff, A., Brunsdon, R.V., 1981. Control of gastro-intestinalnematodes: advantages of a preventive anthelmintic drench-ing programme for lambs on pasture. N. Z. J. Exp. Agric. 9,221–225.

Waller, P.J., Donald, A.D., Dobson, R.J., Lacey, E., Hennessey, D.R., Allerton,G.R., Prichard, R.K., 1989. Changes in anthelmintic resistance statusof Haemonchus contortus and Trichostrongylus colubriformis exposedto different anthelmintic selection pressures in grazing sheep. Int. J.Parasitol. 19, 99–110.