8
Preventive Veterinary Medicine 97 (2010) 220–227 Contents lists available at ScienceDirect Preventive Veterinary Medicine journal homepage: www.elsevier.com/locate/prevetmed Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya B. Bett a,b,, T.F. Randolph a , P. Irungu a,c , S.O. Nyamwaro b , P. Kitala d , J. Gathuma d , D. Grace a , G. Vale e , J. Hargrove e , J. McDermott a a International Livestock Research Institute, P.O. Box 30709-00100, Nairobi, Kenya b Trypanosomiasis Research Centre, Kenya Agricultural Research Institute, P.O. Box 362-00902, Kikuyu, Kenya c Department of Agricultural Economics, University of Nairobi, P.O. Box 29053-00625, Nairobi, Kenya d Department of Public Health Pharmacology and Toxicology, University of Nairobi, P.O. Box 29053-00625, Nairobi, Kenya e South African Centre for Epidemiological Modelling and Analysis (SACEMA), Private Bag X1, Matieland, Stellenbosch University 7602, South Africa article info Article history: Received 2 December 2008 Received in revised form 31 August 2010 Accepted 2 September 2010 Keywords: Effectiveness Tsetse Repellent-technology Trypanosomosis Cattle Kenya abstract We conducted a field trial among Maasai cattle-keepers in Nkuruman and Nkineji areas of Kenya to evaluate the effectiveness of a synthetic tsetse-repellent technology developed for the control of trypanosomosis in cattle. The technology was a repellent (2-methoxy 4-methylphenol) emitted from dispensers attached to collars worn by cattle. Treatment was allocated at the herd level to ensure adequate protection of all the animals in a herd, with measurements of effectiveness conducted at the individual-animal level. The trial began in April 2005 and ran for 16 months including a baseline phase of 4 months. We recruited 12 herds in each area using a restricted random-sampling technique and dis- tributed them equally into intervention (repellent) and control groups. Sample size was determined using a formal power calculation. Effectiveness or minimal worthwhile dif- ference was defined as a 50% reduction in the incidence of trypanosome infection in the treated versus control group (effectiveness below which the technology was considered by experts as not viable compared to existing control techniques). All the animals in the recruited herds were screened monthly (buffy-coat technique) for trypanosome infections. The analysis followed the principle of intention-to-treat by which subjects are analysed according to their initial treatment assignment, regardless of the mechanical performance of the device. Crude and adjusted effects of the technology were 23% (p < 0.001) and 18% (p = 0.08) reduction in the infection incidence in the treatment compared to the control groups, respectively. The impact of the technology estimated in this study did not achieve the threshold of 50% reduction in the trypanosome infection incidence set a priori to indicate effectiveness (p < 0.001). We therefore concluded that the prototype repellent technology package was not sufficiently effective in reducing trypanosome infection incidence under natural tsetse challenge to merit commercial development. © 2010 Elsevier B.V. All rights reserved. Corresponding author at: International Livestock Research Institute, P.O. Box 30709-00100, Nairobi, Kenya. Tel.: +254 020 4223000. E-mail address: [email protected] (B. Bett). 1. Introduction Tsetse-transmitted animal trypanosomosis, the most serious cattle disease of sub-Saharan Africa, constrains live- stock production in much of arid and semi-arid Kenya. It is estimated that 25% of the country (including 60% of the rangelands with pasture suitable for raising cattle) is 0167-5877/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.prevetmed.2010.09.001

Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya

  • Upload
    b-bett

  • View
    220

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya

Fc

BDa

b

c

d

e

a

ARRA

KETRTCK

P

0d

Preventive Veterinary Medicine 97 (2010) 220–227

Contents lists available at ScienceDirect

Preventive Veterinary Medicine

journa l homepage: www.e lsev ier .com/ locate /prevetmed

ield trial of a synthetic tsetse-repellent technology developed for theontrol of bovine trypanosomosis in Kenya

. Betta,b,∗, T.F. Randolpha, P. Irungua,c, S.O. Nyamwarob, P. Kitalad, J. Gathumad,. Gracea, G. Valee, J. Hargrovee, J. McDermotta

International Livestock Research Institute, P.O. Box 30709-00100, Nairobi, KenyaTrypanosomiasis Research Centre, Kenya Agricultural Research Institute, P.O. Box 362-00902, Kikuyu, KenyaDepartment of Agricultural Economics, University of Nairobi, P.O. Box 29053-00625, Nairobi, KenyaDepartment of Public Health Pharmacology and Toxicology, University of Nairobi, P.O. Box 29053-00625, Nairobi, KenyaSouth African Centre for Epidemiological Modelling and Analysis (SACEMA), Private Bag X1, Matieland, Stellenbosch University 7602, South Africa

r t i c l e i n f o

rticle history:eceived 2 December 2008eceived in revised form 31 August 2010ccepted 2 September 2010

eywords:ffectivenesssetseepellent-technologyrypanosomosisattleenya

a b s t r a c t

We conducted a field trial among Maasai cattle-keepers in Nkuruman and Nkineji areas ofKenya to evaluate the effectiveness of a synthetic tsetse-repellent technology developedfor the control of trypanosomosis in cattle. The technology was a repellent (2-methoxy4-methylphenol) emitted from dispensers attached to collars worn by cattle. Treatmentwas allocated at the herd level to ensure adequate protection of all the animals in a herd,with measurements of effectiveness conducted at the individual-animal level. The trialbegan in April 2005 and ran for 16 months including a baseline phase of 4 months. Werecruited 12 herds in each area using a restricted random-sampling technique and dis-tributed them equally into intervention (repellent) and control groups. Sample size wasdetermined using a formal power calculation. Effectiveness or minimal worthwhile dif-ference was defined as a 50% reduction in the incidence of trypanosome infection in thetreated versus control group (effectiveness below which the technology was consideredby experts as not viable compared to existing control techniques). All the animals in therecruited herds were screened monthly (buffy-coat technique) for trypanosome infections.The analysis followed the principle of intention-to-treat by which subjects are analysedaccording to their initial treatment assignment, regardless of the mechanical performanceof the device. Crude and adjusted effects of the technology were 23% (p < 0.001) and 18%

(p = 0.08) reduction in the infection incidence in the treatment compared to the controlgroups, respectively. The impact of the technology estimated in this study did not achievethe threshold of 50% reduction in the trypanosome infection incidence set a priori to indicateeffectiveness (p < 0.001). We therefore concluded that the prototype repellent technologypackage was not sufficiently effective in reducing trypanosome infection incidence undernatural tsetse challenge to merit commercial development.

∗ Corresponding author at: International Livestock Research Institute,.O. Box 30709-00100, Nairobi, Kenya. Tel.: +254 020 4223000.

E-mail address: [email protected] (B. Bett).

167-5877/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.prevetmed.2010.09.001

© 2010 Elsevier B.V. All rights reserved.

1. Introduction

Tsetse-transmitted animal trypanosomosis, the mostserious cattle disease of sub-Saharan Africa, constrains live-stock production in much of arid and semi-arid Kenya.It is estimated that 25% of the country (including 60% ofthe rangelands with pasture suitable for raising cattle) is

Page 2: Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya

rinary M

B. Bett et al. / Preventive Vete

infested with tsetse (Chemuliti et al., 2005). A variety of try-panosomosis control methods is available, including tsetsecontrol, drugs and trypanotolerant cattle. Tsetse-controlstrategies have been implemented over limited areas andhave not proven sustainable (Torr et al., 2005). Cattle keep-ers rely mainly on curative drug treatment, but continueto incur important losses from trypanosomosis and asso-ciated animal-health expense (Holmes, 1997). Researchershave therefore sought to develop control technologies thatcould improve the ability of cattle keepers to manage try-panosome infections more effectively and at lower cost,and which would thereby help address the threat of emerg-ing drug resistance (Mugunieri and Murilla, 2003; Rodericket al., 2000).

Saini and Hassanali (2007, 2002) recently developed anovel tsetse-repellent device for individual cattle and herdsconsisting of a synthetic tsetse repellent (2-methoxy 4-methylphenol; Patent No. Ke00185, 2004). The repellentwas synthesized by adding carbon to 2-methoxyphenoland in lab and field experiments was more potent than2-methoxyphenol and related analogues known to repeltsetse (Saini and Hassanali, 2007). The repellent was dis-pensed from two reservoirs attached to a collar at a rate of4.5 mg/h each. Preliminary studies (the equivalent of PhaseI and Phase II clinical research trials) to evaluate the safetyand efficacy of the technology showed that the repellentdoes not harm the animals and that the device can reducetsetse challenge and feeding efficiency by >80% (Munyua,2006; ILRI/ICIPE, 2001). Studies to develop the device andto identify a placement site on an animal recommendedsuspending the device on the neck for convenience despitethe device being more effective when placed towards thefeet (ILRI/ICIPE, 2001). However, these studies are yet to bepublished in peer-reviewed journals.

Most previous studies on tsetse repellents have eitherused traps, electric-screen and wind-tunnel experiments orcontrolled experiments involving a few herds to gauge theefficacy of the repellent odours. This study was the first toevaluate a tsetse-repellent device in the field under naturaltsetse challenge.

2. Materials and methods

2.1. Study areas

The study was conducted in Kenya in Nkuruman Sub-Location, Kajiado District and in Nkineji, an area thatstraddles Megwara and Maji Moto Sub-Locations, NarokDistrict. The areas were described in Bett et al. (2008).

2.2. The tsetse repellent technology

The prototype device we used had two reservoirs madeof aluminium pipes (10 mm in diameter and 10 cm long),each with corked tygon-silicon tubing as the diffusion point(Fig. 1). The reservoirs were attached to a collar by wire

and the collar was tied around the animal’s neck so thatthe tygon-silicon tubes were suspended ventrally. A singledevice was applied on every animal in a herd. Each reservoirwas filled with 7 ml of the repellent fluid to be replenishedmonthly.

edicine 97 (2010) 220–227 221

2.3. Phases of the study

The study was implemented over two successive phasesstarting with a baseline phase conducted between April andAugust 2004 to assess the similarity of treatment groups,followed by an intervention (longitudinal phase) carriedout between August 2004 and August 2005. Just before thestart of the longitudinal phase (month 0), all animals weretreated with diminazene aceturate administered intramus-cularly at 7 mg/kg to eliminate all tryapanosomes acquiredbefore the start of the trial.

2.4. Sample-size determination

Parameters used for estimating sample size weredeveloped during a workshop with leading experts intsetse and trypanosomosis research. We considered theminimal worthwhile difference to be 50% reduction in try-panosomosis incidence in repellent-treated animals versuscontrols. This was a generously small reduction belowwhich the technology would clearly not offer a viable alter-native to existing control techniques either in terms ofefficacy or cost. The confidence and power of the studywere set at 95% and 80%, respectively, and an intra-herdcorrelation coefficient of 0.4 was used based on Otte andGumm (1997). This was conservatively high because theherd-level treatment was likely to enhance intra-herd clus-tering of measurements estimated at animal level.

Trypanosome infection incidence per animal per yearwas the primary outcome. Assuming a monthly incidenceof 10%, the expected annual incidence was 72% in the con-trol group and 46% (an expected 50% reduction) in thetreatment group. Estimated monthly incidence was basedon a previous study by Mwangi et al. (1998) in Nkuru-man which reported monthly incidence ranging between4.6 and 30.4%, depending on the level of tsetse challenge.The sample size was estimated using the method describedby Dohoo et al. (2003, p. 41) for comparing two indepen-dent proportions in a one-sided test, and then was adjustedbased on the level of clustering within herds using:

n′ = n(1 + �(m − 1)), (1)

where n′ is the adjusted sample size; n, unadjusted samplesize; �, intra-herd correlation coefficient, in this case 0.4;m, average herd size.

Table 1 shows the various estimates of sample sizesobtained under different input parameter scenarios. Usingyearly instead of monthly measures of incidence, we wereable to detect the required worthwhile difference in a man-ageable number of herds (24 herds each having between 50and 200 animals) with 12 assigned to each group.

2.5. Selection of the study herds

Within each area, six villages were recruited that metcriteria of sufficient tsetse challenge and non-overlapping

grazing zones. A sampling frame of households with cat-tle herds ≤200 animals was developed. Households wereclassified by number and size (<50, 50–99, and ≥100 ani-mals) of cattle herds owned and grazing patterns. Herdsizes of 200 animals or more would have required more
Page 3: Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya

222 B. Bett et al. / Preventive Veterinary Medicine 97 (2010) 220–227

Screw cap

One of the two wire strands used to fasten the dispenser onto a neck collar

Aluminium pipe used as a reservoir

Tygon-silicon tube (4 cm long) inserted to the distal end of the pipe (diffusion area)

Protective housing made of a garden hose used to guard aluminium pipe-tygon-silicon tube connection

-repelle

dw1atc(ptSphbtwrabcorfit

2

b

TTp

t

w

Cork

Fig. 1. The structure of a tsetse

evices and sampling materials for monitoring and soere not included. Households were then grouped into

2 spatial clusters (evenly distributed between the tworeas) based on grazing patterns, each corresponding to areated herd. Grazing was communally managed within theluster and the clusters generally coincided with villagesparticularly in Nkineji). In Nkuruman, transhumance wasractised in which herds moved between villages duringhe year according to wet and dry season grazing areas.erial numbers were assigned to all the herds in the sam-ling frame, by village, and both treatment and controlerds selected using a restricted randomization processased on random numbers generated in MS Excel (usinghe worksheet function = randbetween). Treatment herdsere selected first; the randomization process was then

epeated to select a control herd in each cluster frommongst those herds having approximately the same num-er of animals as the selected treatment herd for thatluster. All animals in herds were ear-tagged. Developmentf the sampling frame, generation of random numbers andecruitment of herders and herds was performed by therst author. All herders selected agreed to participate inhe trial.

.6. Animal sampling

Methods used in sampling the recruited animals haveeen described previously (Bett et al., 2008). Briefly, all

able 1he number of herds and corresponding number of animals that would be requireower at the various levels of a priori incidence and expected average herd-size,

A priori tryps incidence (outcomemeasure)

Average herd-size, m

50 100

Monthly incidence (10 vs. 5%)a 140 herds 7010 cattle 138 herds, 1Yearly incidence (72 vs. 46%)b 12 herds 594 cattle 12 herds 117

a A priori trypanosome infection incidences in the control group, based on prehe targeted 50% reduction.

b A priori trypanosome infection incidences in the control and treatment grouphere t = 12 months.

nt dispenser used in the study.

the ear-tagged animals were sampled once every month.They were restrained in a crush, their ear veins prickedwith a lancet and blood collected using a pair of hep-arinised capillary tubes. Presence of trypanosomes in theblood was determined using the buffy-coat technique (BCT)(Murray et al., 1977). Trypanosomes were identified byspecies based on their motility characteristics observedon wet preparations of the buffy coat and later confirmedthrough thin blood smears stained with Giemsa. All animalsfound positive or having a packed-cell volume <22% weretreated with diminazene aceturate at 7 mg/kg body weight(administered intramuscularly). Similarly, other cases suchas wounds, tick-borne infections and worm infestation thatwere observed or presented by the cattle owners weremanaged appropriately. Calves, weaners and lean animalswere also given broad-spectrum anthelmintics every 4months.

2.7. Examination and refilling of the repellent dispensers

The dispensers were designed to hold sufficient repel-lent for at least a month after being filled. We soon realizedthat the device was sensitive to abrasions and tension

(especially at the point of diffusion), causing leakage. At thetime of sampling, the condition of the devices and whetheror not they contained the repellent were recorded. Thedevices were described as: (i) good—if both of the reser-voirs still had repellent; (ii) average—if at least one of the

d to detect a 50% reduction of trypanosome infection incidence with 80%m.

200 400

3,816 cattle 137 herds 27,429 cattle 137 herds 54,653 cattle0 cattle 12 herds 2322 cattle 12 herds 4627 cattle

viously published data (see text), and in the treatment group, assuming

estimated from monthly incidence rates using the formula: 1 − (1 − p)t

Page 4: Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya

rinary M

B. Bett et al. / Preventive Vete

reservoirs had repellent; or (iii) poor—if neither reservoirhad repellent. Missing devices were also recorded. At eachmonthly sampling, dispensers were replenished and dam-aged ones replaced.

2.8. Data analysis

Data were stored in a relational database designed usingMicrosoft® Access and statistical analyses conducted inSTATA 8.2 (StataCorp., 2003). The level of confidence forall the statistical tests was 95%. Six herds (three at eachsite) recruited at the beginning as single herds were laterrecognized to be composed each of two sub-herds man-aged independently. The sub-herds were therefore treatedas separate herds for purposes of analysis, increasing thesample size to 30 herds.

High rates of the repellent dispenser malfunction wereobserved, so the analysis maintained the principle ofintention-to-treat whereby subjects are analysed accord-ing to original treatment assignment (Dohoo et al., 2003;Peduzzi et al., 2002). Trypanosome infection incidence wasestimated at animal level using both the first and mul-tiple diagnoses determined by BCT. The use of the firstinfection was considered most appropriate for this anal-ysis as a type of survival model with a single failure andconsistent with the a priori trypanosomosis incidence usedto estimate the sample size. Trypanosome infection, how-ever, is a recurrent event and including repeated infectionscaptures additional relevant information to understandfactors predisposing an animal to repeated infections andthe effectiveness of an intervention such as the repellentdevice in such a population. The denominator for estimat-ing incidence was the number of animal-years betweenthe dates the longitudinal phase of the study began anddetection of infection, withdrawal from the study, or ter-mination of the study. Animals introduced into study herdsduring the course of the longitudinal study were recruitedat the following monthly sampling. Possible reduction inrisk of infection due to treatments we administered orthose given by stock owners was ignored when derivingincidence based on multiple infections per subject. This isbecause diminazene aceturate has a variable prophylacticeffect in cattle ranging between a few days to a few weeks(Peregrine and Mamman, 1993).

2.8.1. Multivariable analysesWe did not anticipate doing multivariable analysis

when designing the study given the fact that we selectedthe study herds using a random sampling technique thatwas expected to provide a balanced design. Changes inthe grazing patterns observed during the driest timesof the year, however, disrupted this design. For ran-domised controlled trials, Peduzzi et al. (2002) suggestedthat when a variable is imbalanced between the treat-ment groups, a multivariable analysis should be donetogether with a crude analysis because the former would

give greater precision in estimating the treatment effect.Multivariable analyses were therefore conducted usingPoisson-regression models incorporating general estima-tion equations to correct for repeated measures in time. Themodels adopted a continuous-time first-order autoregres-

edicine 97 (2010) 220–227 223

sive correlation pattern given its suitability for controllingwithin-subject correlation decreasing in time in a longitu-dinal study design. The Huber/White/Sandwich estimatorof variance was used to estimate standard errors asdescribed by Caroll et al. (1998).

Independent variables included age, sex, colour of theanimal, prevailing weather, village, area and herd size(Table 3). Cattle were categorised as calves or adults.Calves have lower risk of contracting trypanosomosis com-pared to adults (Torr et al., 2006). In both areas, adultanimals were taken out for grazing, while calves wereconfined around homesteads resulting in different expo-sures. Male cattle have higher susceptibility to the infectionthan females (Dolan, 1998; Rowlands et al., 2001; Torret al., 2007). Since tsetse are attracted more to dark thanlight surfaces (Sterverding and Troscianko, 2004), the coatcolour of cattle was categorised as brown, black, white orother.

Trypanosomosis incidence is often higher during thewet season because of increased tsetse challenge (Bett etal., 2004; Cherenet et al., 2004). But seasonal variation ingrazing patterns of pastoral herds also influences exposureto tsetse. Herders often avoid densely infested areas in thewet season when pasture is available elsewhere but mightuse them in the dry season when pasture is scarce. Thismight contribute to high rates of infection also during thedry season as reported by Ngaira et al. (2002) in easternparts of Kenya. Prevailing weather was classified as wet ordry based on the amount of monthly rainfall recorded ineach area. For Nkineji, March to August were wet months,while for Nkuruman, December and March to May werewet.

Herders sometimes changed grazing patterns in the dryseason, and as a result, treated herds were not necessarilyfound in the same villages as their corresponding controlherds. Village (where herds were found at time of sam-pling) was therefore included as a fixed effect to control forvillage-level effects including tsetse challenge. Tsetse chal-lenge could not be used as an independent factor becauseit was not unconditionally associated with trypanosomosisincidence (Bett et al., 2008). Parameter estimates gener-ated for the dummy variables representing village weretreated as nuisance factors (i.e., parameter estimates ofthese dummy variables are not reported although thesefactors are included in the model to obtain adjusted effectof treatment). Area was coded as Nkuruman or Nkineji,and captures variation in ecological or management factorsbetween sites that might not have been explicitly capturedby other variables. Some of these factors include variationin tsetse populations and trypanosome species as well aspresence of a research station in Nkuruman that increasedherder awareness there. Herd size was fitted as a con-tinuous linear variable in the models. It can be a strongpredictor for trypanosomosis; increasing the size of a herdfrom one to 12 has been estimated to increase the meannumber of tsetse flies visiting 4-fold and the mean proba-

bility of feeding from 54 to 71% (Torr et al., 2007). However,the probability of an individual animal being bitten by afly declines with an increase in herd-size. The linearityassumption was tested by fitting quadratic terms for thevariable. The quadratic terms were not significant, so the
Page 5: Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya

224 B. Bett et al. / Preventive Veterinary Medicine 97 (2010) 220–227

Table 2Crude incidence rates ratio for treatment per animal-year (Nkineji, Narok and Nkuruman, Kajiado; Kenya; 2004–2005).

Tryps infection Treatment Incidence Incidence rates ratio

Estimate 95% CIa Estimate 95% CIa

First cases only Repellent 0.80 0.73–0.86 0.77 0.67–0.88

ls

3

3

iiiappeo1tacwr

fitl(cvd

Fv

Control 1.03All cases Repellent 0.98

Control 1.26

a One-sided confidence interval.

inearity assumption was retained. The models are pre-ented in Table 3.

. Results

.1. Trypanosomosis incidence

Both Trypanosoma congolense and T. vivax were presentn the study areas. A positive outcome comprised annfection caused by either of these parasites singly orn combination. A total of 876 primary cases in 964nimal-years was recorded. This represented an overall try-anosome infection incidence of 0.91 (95% CI: 0.86–0.96)er animal-year. When multiple infections were consid-red, a total of 1193 cases in 1069 animal-years werebserved giving an overall incidence of 1.12 (95% CI:.05–1.18) per animal-year. The observed incidences in thereated and control groups and crude incidence rate ratiore given in Table 2. The ratio estimates indicate a statisti-ally significant 22–23% reduction in incidence associatedith the treatment (p < 0.001), but does not achieve the 50%

eduction targeted.Fig. 2 shows the trends in the monthly animal-level

rst case incidence in treatment and control groups forhe two study areas. Also shown by the broken line is the

og-transformed average number of tsetse per trap per dayFTD). Note, though, that the fly densities are not directlyomparable between areas because fly-trapping efficacyaries according to tsetse species and the mix of speciesiffered across the two areas; G. pallidipes and G. longipen-

1

10

100

1000

0

2

4

6

8

10

12

14

16

121086420-2-4

Try

pano

som

e in

fect

ion

inci

denc

e (%

)

Month

Nkineji

Aug 05Apr 04

ig. 2. Monthly trypanosome infection incidence determined at the animal-levelillages in Nkineji, Narok and Nkuruman, Kajiado Districts, Kenya (2004–2005). M

0.96–1.110.91–1.05 0.78 0.69–0.871.18–1.35

nis were common in Nkuruman whereas tsetse found inNkineji were G. swynnertoni (which is known to be difficultto trap [Williams et al., 1992]) and very low densities of G.pallidipes.

3.2. Multivariable analysis

Results of the multivariable (GEE) Poisson modelsare presented in Table 3. The models fitted the datawell—the magnitudes of their respective scale parametersare given below each model. The adjusted effect of thetechnology was not more than 18% reduction in diseaseincidence and not statistically different from zero (p = 0.08)at alpha = 0.05. Again, the null hypothesis of the effect beingless than the 50% reduction targeted could not be rejected.The models further show that controlling for the effect ofother covariates, males had a higher incidence of the dis-ease than females and there was a lower infection incidencein the wet compared to dry months.

3.3. The defects of the device

We put collars on 891 animals at the start of the longi-tudinal study. Because the study population was open, thetotal number of animals/dispensers inspected varied with

time. Over the longitudinal phase, we did 9713 inspections(7478 of which were in Nkineji). Of the inspections car-ried out, 1187 (16% [95% CI: 15–17%]) and 175 (8% [7–9%])classified the dispensers as being in good condition (havingthe repellent remaining in both dispensers) at Nkineji and

1

10

100

1000

0

2

4

6

8

10

12

14

16

121086420-2-4

Flie

s pe

r tr

ap p

er d

ay (

Log

10)

Month

Control

Treatment

Log (FTD)

Nkuruman

Apr 04 Aug 05

and log10 transformed numbers of flies per trap per day (FTD) for studyonth 0 indicates the beginning of the longitudinal study.

Page 6: Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya

B. Bett et al. / Preventive Veterinary Medicine 97 (2010) 220–227 225

Table 3Poisson-regression multivariable models of trypanosome infection incidence at the animal-level and the effect of tsetse-repellent technology (Nkineji,Narok and Nkuruman, Kajiado; Kenya; 2004–2005).

Variable Level First-case-only model All-cases-model

Cases Animal-years IRRa p > |Z| Cases Animal-years IRR p > |Z|

Estimate 95% CIb Estimate 95% CI

Treatment Repellent 403 506 0.82 0.70–0.99 0.08 546 557 0.83 0.67–1.03 0.15Control 473 458 1.00 647 512 1.00

Age Calf 258 279 1.17 1.00–1.37 0.10 345 305 1.04 0.85–1.28 0.73Adult 618 685 1.00 848 764 1.00

Sex Male 248 243 1.42 1.21–1.66 <0.001 343 272 1.40 1.14–1.71 0.01Female 628 721 1.00 850 798 1.00

Colour White 114 105 1.22 0.99–1.51 0.12 155 119 1.08 0.87–1.35 0.56Black 202 214 1.04 0.87–1.25 0.69 286 240 1.06 0.76–1.47 0.78Other 72 70 1.00 0.76–1.30 0.98 103 78 1.25 0.97–1.61 0.14Brown 488 575 1.00 649 634 1.00

Herd-size 1.00 0.99–1.00 0.31 1.00 0.99–1.00 0.51Season Wet 387 433 0.61 0.53–0.70 <0.001 616 509 0.77 0.64–0.91 0.01

Dry 489 531 1.00 577 561 1.00Area Nkuruman 246 322 1.20 0.50–2.90 0.73 294 338 1.69 0.68–4.19 0.34

Nkineji 630 642 1.00 899 731 1.00

Scale parameter = 1; Wald �2 (20)=173, p ≤ 0.001 Scale parameter = 1; Wald �2 (20)=80, p ≤ 0.001

a Incidence rates ratio.b One-sided confidence interval.

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1211109876543210

Prop

ortio

n of

ani

mal

s pe

r pr

otec

tion

leve

l

Good

Average

Poor

Lost/stolen

Nkineji

Aug 04 Aug 05

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1211109876543210

Nkuruman

Aug 04 Aug 05

imal-levan, Kaji

Month

Fig. 3. The distribution of the states of the repellent dispenser at the antsetse repellent evaluation trial carried out in Nkineji, Narok and Nkurum

Nkuruman, respectively. We found repellent remaining inat least one dispenser in 41% (3085) of the inspections inNkineji and in 27% (595) in Nkuruman. The distributions ofthe defects of the devices observed over time are shown inFig. 3.

4. Discussion

Baseline comparison of the treatment groups confirmedthat they were adequately matched (data not shown).Treatments were not blinded from the livestock own-ers and the research team and this could be regardedas introducing a potential source of bias. No change was

observed, however, in herding practices among treatedherds—indicating that lack of blinding did not influenceherders’ behaviour. The need for blinding had been consid-ered during the study-design stage but a suitable strategycould not be devised because, given resources available, to

Month

el at each monthly monitoring event over the longitudinal phase of theado Districts; Kenya (2004–2005).

equip the control group with the limited collar and dis-penser equipment supplies available would have requireda substantial reduction in sample size. Potential for biaswas considered low because the trial was implemented byan independent research team.

Although the BCT test is the standard for field sur-veys, it has low sensitivity because it cannot detecttrypanosomes when parasitaemia is <200–1000 try-panosomes/ml (Murray et al., 1977). There is very scantyinformation on sensitivity of the BCT test—the few esti-mates that have been published vary by trypanosomespecies examined, host type and area where the study wasconducted. Delafosse et al. (2006) assumed that this esti-

mate ranged between 20 and 80% while determining trueprevalence of trypanosomosis in cattle in a tsetse infestedarea of Lake Chad. Studies conducted by Manuel Monzónet al. (1990) in horses in Argentina demonstrated that thetest had a sensitivity of 63.4% when it was used to diag-
Page 7: Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya

2 rinary M

ndsspi(imisctgtoofenm

wmcmtaoeppsv

ectn5rmwiWcre

mtroivmttpti

26 B. Bett et al. / Preventive Vete

ose T. evansi. Gall et al. (2004) also reported that the testetects a third to a quarter of cases positive on PCR. In ourtudy, resource limitations did not permit using more sen-itive, expensive tests. However, trypanosomosis screeningractices that are known to improve the diagnostic abil-

ty and reliability of the test (suggested by Maudlin et al.2004)) such as sampling animals very early in the morn-ng, collecting venous blood from peripheral ear veins and

inimizing the time between blood collection and screen-ng were always followed. The use of a test with a lowensitivity might cause misclassification bias, which in thisase could be classified as being non-differential becausehe errors were expected to be constant across treatmentroups. Such a bias would dampen the magnitude of thereatment effect (Dohoo et al., 2003). However, the impactf this bias on the incidence rate ratio becomes apparentnly when the specificity of a test is not perfect. Given theact that BCT test has a high (perfect) specificity (Delafosset al., 2006), we expect that misclassification bias mightot have influenced the observed treatment impact esti-ates.Use of the device did not attain the minimal worth-

hile reduction on trypanosome infection incidence foreasures using first or multiple cases per subject. The

rude effect (∼23%) was different from the adjusted esti-ate (∼18%), justifying the use of multivariable analysis

o correct for confounding, even though blocking by villagend restricted randomization had been applied. The pairingf treatment and control herds by village was intended toxpose the pairs to identical grazing conditions and try-anosomosis risk, but unanticipated changes in grazingatterns for individual herds (particularly in the dry sea-on when herders sought pastures and water outside theirillages) might have introduced confounding.

Our Phase III trial (the definitive assessment of theffectiveness of the repellent technology under typi-al field conditions) found the novel tsetse-repellentechnology developed for trypanosomosis control wasot effective in reducing disease incidence by at least0%. The estimated 18% reduction in trypanosomosisisk associated with the collars in the field trial wasuch lower than the >80% reduction in feeding observedhen the repellent was tested against a stationary ox

n behavioural experiments (Saini and Hassanali, 2002).hile design effects of the dispenser are likely to have

ontributed to this incompatibility, auxillary analysis (noteported here) suggested that its contribution was mod-st.

Other factors related to the nature of the repellentay explain the discrepancy. First, Rogers (1988) showed

hat trypanosomosis risk is more sensitive to mortalityate and feeding intervals of tsetse and incubation periodf trypanosomes in the vector than to other parametersncluding the probability that the vector feeds on a givenertebrate host. Second, he also showed that trypanoso-osis prevalence is more responsive to changes in the

ransmission parameters in hosts that are less preferred bysetse (for blood meal) compared to those that are highlyreferred. Tsetse blood-meal analyses conducted duringhis study showed that cattle were the most preferred hostn Nkineji and only the 5th most preferred in Nkuruman

edicine 97 (2010) 220–227

after warthog, elephant, zebra and buffalo, in that order(Bett et al., 2008). Therefore, we expect less than antici-pated effectiveness of the repellent in these areas becauseit will be more difficult to reduce chances of a fly feeding ona highly preferred host compared to a less preferred one.Third, the repellence of human odour to savannah tsetse(Vale, 1974) is similar to that of 2-methoxy 4-methyl phe-nol yet human odour can become ineffective when flieshave been starved (Vale and Cumming, 1976). Fourth, sincethe 2-methoxy 4-methyl-phenol operates by blocking theperception of attractants (Saini and Hassanali, 2007), it willbe most effective in situations where odour attraction isparticularly important. This would be consistent with theobservation of significant reductions in feeding on station-ary cattle (Saini and Hassanali, 2002) since odour is theprimary attractant to these baits (Vale, 1974). In the fieldtrial, the cattle were mobile much of the time, and in thissituation, visual attraction plays a more important role(Vale, 1974).

Finally, the electric-screen experiments used to deter-mine the efficacy of the repellent (considered to be >80%[Saini and Hassanali, 2007]) might have overestimated thetrue effect of the device because such experiments do notdistinguish flies that fail to feed (successfully repelled)from those that probe on the target hosts without suc-cessfully feeding to repletion (not repelled). Given thatprobing (by infected tsetse) is sufficient for trypanosomo-sis transmission (Roberts, 1981), the reduction in feedingrecorded in the experiment may not translate into anequivalent reduction in infection. There are several rea-sons, therefore, to suspect that the repellent might notreduce infection rates to the degree anticipated from theinitial experiments measuring tsetse challenge and feedingefficiency.

5. Conclusion

Applying the repellent device to all cattle in a herd undernatural field conditions did not significantly reduce try-panosome infection incidence at the animal level at theminimal worthwhile difference. Given that the trial designprovided for maximum protection of treated herds (i.e., allanimals in the herd were provided the device), the resultsclearly indicate that the current prototype does not offeradequate levels of protection to merit further commercialdevelopment.

Acknowledgements

We thank all the technical staff from Epidemiology andEntomology Divisions, Trypanosomiasis Research Centre,Kenya Agricultural Research Institute for helping in datacollection. We also thank Hippolyte Affognon, Stephen Torrand Oumar Diall for helping in the design of the study.Logistical support was provided by International Livestock

Research Institute, Trypanosomiasis Research Centre ofKARI, University of Nairobi and International Centre forInsect Physiology and Ecology. This work was funded byInternational Fund for Agricultural Development, TechnicalGrant Assistance Number 554.
Page 8: Field trial of a synthetic tsetse-repellent technology developed for the control of bovine trypanosomosis in Kenya

rinary M

B. Bett et al. / Preventive Vete

References

Bett, B., Irungu, P., Nyamwaro, S.O., Murilla, G., Kitala, P., Gathuma, J., Ran-dolph, T.F., McDermott, J., 2008. Estimation of tsetse challenge and itsrelationship with trypanosomosis incidence in cattle kept under pas-toral production systems in Kenya. Vet. Parasitol. 155 (3–4), 287–298.

Bett, B., Orenge, C., Irungu, P., Munga, L.K., 2004. Epidemiological factorsthat influence time-to-treatment of trypanosomosis in Orma Borancattle raised at Galana ranch, Kenya. Vet. Parasitol. 120, 43–53.

Caroll, R.J., Wang, S., Simpson, D.G., Stromberg, A.J., Ruppert, D.,1998. The Sandwich (robust covariance matrix) Estimator,Technical Report. Preprint available at http://stat.tamu.edu/ftp/pub/rjcarroll/sandwich.pdf.

Chemuliti, J.K., Mugunieri, G.M., Okoth, S., Munene, S., Ndungu, J.M.,2005. Area-wide pan African tsetse and trypanosomosis eradicationcampaign for the central South of Mount Kenya belt—prospects andchallenges. In: Proceedings of the 28th Meeting of International Sci-entific Council for Trypanosomiasis Research and Control (ISCTRC).STRC/AU, Addis Ababa, Ethiopia.

Cherenet, T., Saini, R.A., Panandam, J.M., Nadzr, S., Speybroeck, N., van denBossche, P., 2004. Seasonal prevalence of bovine trypanosomosis ina tsetse-infested zone and a tsetse-free zone of the Amhara Region,North-West Ethiopia. Onderstepoort J. Vet. Res. 71 (4), 307–312.

Delafosse, A., Thébaud, E., Desquesnes, M., Michaux, Y., 2006. Epidemiol-ogy of Trypanosoma vivax infection in cattle in the tsetse area of LakeChad. Prev. Vet. Med. 74, 108–119.

Dohoo, I., Martin, W., Stryhn, H., 2003. Controlled Trials. Veterinary Epi-demiologic Research, AVC Inc., PEI, Canada, pp. 185–205.

Dolan, R.B., 1998. The Orma Boran: A Trypanotolerant East African Breed.Fifteen years of Research on Galana Ranch in Kenya. Kenya Trypanoso-miasis Research Institute, Kikuyu, Kenya, p. 96.

Gall, Y., Woitag, T., Bauer, B., Sidibe, I., McDermott, J., Mehlitz, D., Clausen,P.H., 2004. Trypanocidal failure suggested by PCR results in cattle fieldsamples. Acta Trop. 92 (1), 7–16.

Holmes, P.H., 1997. New approaches to the integrated control of try-panosomosis. Vet. Parasitol. 72 (2–3), 121–135.

ILRI/ICIPE, 2001. Getting Tsetse Repellent Out of the Laboratory and tothe Farmer: Enhancing the Transfer, Delivery and Adoption of NewControl Technology for Improved Livestock Health and Productivity,Full design document, Nairobi.

Manuel Monzón, C., Antonio Mancebo, O., Pablo Roux, J., 1990. Compari-son between six parasitological methods for diagnosis of Trypanosomaevansi in the subtropical area of Argentina. Vet. Parasitol. 36, 141–146.

Maudlin, I., Holmes, P.H., Miles, M.A., 2004. The Trypanosomiases. CABIPublishing, Oxfordshire, UK.

Mugunieri, G.L., Murilla, G., 2003. Resistance to trypanocidaldrugs—suggestions from field survey on drug use in Kwale district,Kenya. Onderstepoort J. Vet. Res. 70 (1), 29–36.

Munyua, P.M., 2006. Toxicological assessments of a tsetse fly repellent(Patent No. Ke00185, 2004) on the health of exposed animals. MScThesis, University of Nairobi.

Murray, M., Murray, P.K., McIntyre, W.I.M., 1977. An improved parasito-logical technique for the diagnosis of African trypanosomiasis. Trans.R. Soc. Trop. Med. Hyg. 71, 325–326.

edicine 97 (2010) 220–227 227

Mwangi, E.K., Stevenson, P., Gettinby, G., Reid, S.W., Murray, M., 1998.Susceptibility to trypanosomosis of three Bos indicus cattle breeds inareas of differing tsetse fly challenge. Vet. Parasitol. 79, 1–17.

Ngaira, J.M., Bett, B., Karanja, S.M., 2002. Animal-level risk factors for Try-panosoma evansi infection in camels in eastern and central parts ofKenya. Onderstepoort J. Vet Res. 69, 263–271.

Otte, M.J., Gumm, I.D., 1997. Intra-cluster correlation coefficients of 20infections calculated from the results of cluster-sample surveys. Prev.Vet. Med. 31, 147–150.

Peduzzi, P., Henderson, W., Hartigan, P., Lavori, P., 2002. Analysis of ran-domized control trials. Epidemiol. Rev. 24 (1), 26–38.

Peregrine, A.S., Mamman, M., 1993. Pharmacology of diminazene: areview. Acta Trop. 54, 185–203.

Roberts, L.W., 1981. Probing by Glossina morsitans and transmission ofTrypanosoma (Nannomonas) congolense. Am. J. Trop. Med. Hyg. 30,948–951.

Roderick, S., Stevenson, P., Mwendia, C., Okech, G., 2000. The use of try-panocides and antibiotics by Maasai pastoralists. Trop. Animal HealthProd. 32 (6), 361–374.

Rogers, D.J., 1988. A general model for the African trypanosomiases. Par-asitology 97, 193–212.

Rowlands, G.J., Leak, S.G.A., Peregrine, A.S., Nagda, S.M., WoudyalewMulatu, d’Ieteren, G.D.M., 2001. The incidence of new and the preva-lence and persistence of recurrent trypanosome infections in cattle insouthwest Ethiopia exposed to a high challenge with drug-resistantparasites. Acta Trop. 79, 149–163.

Saini, R., Hassanali, A., 2002. Attractants and repellents for tsetse—wheredo we go from here? Integrated Vector Control including synergisticuse of drugs and bait technologies for the control of trypanosomiasisand tick borne diseases. In: Proceedings of the Joint ICTTD-2/ICPTVWorkshop, Institute of Tropical Medicine, 10–12 April 2002. FAO,Antwerp, Belgium, pp. pp31–32.

Saini, R.K., Hassanali, A., 2007. A 4-alkyl-substituted analogue of guiaiacolshows greater repellency to savannah tsetse (Glossina spp.). J. ChemEcol. 33 (5), 985–995.

Sterverding, D., Troscianko, T., 2004. On the role of blue shadows in thevisual behaviour of tsetse flies. Proc. R. Soc. Lond. B (Suppl.) 271,S16–S17.

Torr, S.J., Hargrove, J.W., Vale, G.A., 2005. Towards a rational policy fordealing with tsetse. Trends Parasitol. 21, 537–541.

Torr, S.J., Mangwiro, T.N.C., Hall, D.R., 2006. The effects of host physiol-ogy on the attraction of tsetse (Diptera: Glossinidae) and Stomoxys(Diptera: Muscidae) to cattle. Bull. Entomol. Res. 96, 71–84.

Torr, S.J., Prior, A., Wilson, P.J., Schofield, S., 2007. Is there safety in num-bers? The effect of cattle herding on biting risk from tsetse flies. Med.Vet. Entomol. 21 (4), 301–311.

Vale, G.A., 1974. The responses of tsetse flies (Diptera. Glossinidae) tomobile and stationary baits. Bull. Entomol. Res. 64, 545–588.

Vale, G.A., Cumming, D.H.M., 1976. The effect of selective eliminationof hosts on a population of tsetse flies Glossina morsitans morsitansWestwood (Diptera, Glossinidae). Bull. Entomol. Res. 66, 713–729.

Williams, B., Dransfield, R., Brightwell, R., 1992. The control of tsetse fliesin relation to fly movement and trapping efficiency. J. Applied Ecol.29, 163–179.