Vet. Res. 38 (2007) 451–464 451c© INRA, EDP Sciences, 2007DOI: 10.1051/vetres:2007002

Original article

Risk factors associated with the prevalenceof tuberculosis-like lesions in fenced wild boar

and red deer in south central Spain

Joaquín Va,b,c*, Ursula Ha, Joseba M. Gd,Isabel G. F--a, Pelayo Aa, Ramón Jd,

Marta Bd, Christian Ga

a Instituts de Investigación en Recursos cinegéticos – IREC (CSIC-UCLM-JCCM),Ronda de Toledo s/n, 13080 Ciudad Real, Spain

b Ciudad Real Agricultural Engineering School (UCLM), Ronda de Calatrava 7,13004 Ciudad Real, Spain

c Present address: Wildlife Disease Ecology Team, Central Science Laboratory, Tinkley Lane,Nympsfield, GL103UJ Gloucestershire, United Kingdom

d NEIKER Instituto Vasco de I+D Agraria, C./Berreaga 1, 48300 Derio, Spain

(Received 27 January 2006; accepted 19 October 2006)

Abstract – In recent decades the management of large game mammals has become increasinglyintensive in south central Spain (SCS), resulting in complex epidemiological scenarios for diseasemaintenance, and has probably impeded schemes to eradicate tuberculosis (TB) in domestic live-stock. We conducted an analysis of risk factors which investigated associations between the patternof tuberculosis-like lesions (TBL) in wild boar (Sus scrofa) and red deer (Cervus elaphus) across 19hunting estates from SCS and an extensive set of variables related to game management, land useand habitat structure. The aggregation of wild boar at artificial watering sites was significantly asso-ciated with an increasing risk of detecting TBL in both species, which probably relates to enhancedopportunities for transmission. Aggregation of wild boar at feeding sites was also associated withincreased risks of TBL in red deer. Hardwood Quercus spp. forest availability was marginally asso-ciated with an increased risk of TB in both species, whereas scrubland cover was associated with areduced individual risk of TBL in the wild boar. It is concluded that management practices that en-courage the aggregation of hosts, and some characteristics of Mediterranean habitats could increasethe frequency and probability of both direct and indirect transmission of TB. These findings are ofconcern for both veterinary and public health authorities, and reveal tuberculosis itself as a potentiallimiting factor for the development and sustainability of such intensive game management systemsin Spanish Mediterranean habitats.

Mycobacterium tuberculosis complex / red deer / risk factors / tuberculosis / wild boar

1. INTRODUCTION

Tuberculosis (TB) is a low-induced im-munity [23] and chronic infectious dis-

* Corresponding author:joaquin.vicente@uclm.es

ease. Mycobacterium bovis, the causativeagent of bovine TB, infects a broad hostrange [6, 7]. Bovine TB in cattle is a ma-jor economic problem. In Spain, in previ-ous years, infection in cattle has been re-duced by testing-and-slaughter (the overall

Article available at http://www.edpsciences.org/vetres or http://dx.doi.org/10.1051/vetres:2007002

452 J. Vicente et al.

proportion of herds officially consideredTB free in the country in 2004 wasover 99%), but only a moderate decreasingtrend has been observed in southern areas,where the proportion of herds officiallyconsidered free of TB in 2004 was only95% [22]. The re-emergence of TB andcontinued failure to eradicate the disease inlivestock in many countries have been re-lated to reservoirs of infection in wildlifepopulations (e.g. [1, 36]). However, it issimplistic to assume that infection in cat-tle is due entirely to an inherent reservoirin wildlife [8]. Nevertheless, further under-standing of the main factors affecting thepersistence of TB in wildlife is likely tobe valuable in the development of sustain-able approaches to managing the diseasein livestock. During recent decades, ex-tensive changes in land-use in Spain havebeen followed by a marked increase inthe abundance and distribution of wild un-gulates [9, 30], and the development of acommercial recreational hunting industry,especially in south central Spain (SCS). Inorder to increase hunting harvests, high-wire fences have been employed on a largeproportion of estates to contain wild un-gulates. These populations have essentiallybecome captive, and artificial feeding andwatering is usually provided during all orpart of the year. TB in wildlife appearsto be endemic across a broad geographicregion, coincident with the traditional biggame hunting areas of SCS [11, 14, 26, 34].Data suggest that TB in wild boar (Susscrofa), red deer (Cervus elaphus) andcattle is an endemic and low incidenceinfectious disease [11, 26]. Although in-terspecific transmission rates have neverbeen described, recent work has shownthat wild boar and red deer share simi-lar strains from the Mycobacterium tuber-culosis Complex [11]. Across this area,tuberculosis-like lesions (TBL) in ungu-late populations have been employed asa criterion to evaluate disease distribu-tion using large numbers of animals [34].

This research revealed that TBL in bothspecies were spatially associated, and thatthe prevalence of TBL was consistentlyhigher in wild boars than in red deer [34].Within this complex multi-host scenario,the prevalence of TBL in wild boar didnot vary according to whether populationswere or were not cohabiting with red deer,suggesting involvement of wild boar as areservoir [34]. This situation is of concernfor veterinary and public health authorities(e.g. [7,11,21]) since game ungulates couldpotentially act as TB reservoirs for domes-tic livestock. Therefore, understanding therisk factors associated with TB infection inthese ungulates is fundamental to the de-velopment of effective TB control policies.

Epidemiologic evidence is often the bestfirst step in identifying disease reservoirsand revealing associated risk factors [13].To date no research has identified manage-ment and habitat correlates of TB in wildboar and red deer populations in SpanishMediterranean habitats. Such informationmay also provide insights into transmis-sion dynamics at the big game/livestockinterface. Our aim was to identify the ma-jor management and environmental factorsassociated with the occurrence of TBL inwild boar and red deer in these areas.

2. MATERIALS AND METHODS

2.1. Study sites

Data were collected at the individuallevel, and at the hunting estate level (asa discrete management unit). A variety ofhunting estates may be found across SCS,which can be classified as containing cap-tive fenced cervid populations (large en-closures, generally at relatively high den-sities [12]) or having free-ranging nativecervid populations (not enclosed). In ad-dition, a small number of deer farms arepresent in Spain, although no data wereavailable to include them in the present

Risk factors and TB in big game in Spain 453

study. Both wild boar and red deer werepresent on all the study sites.

Study sites were located on the southof the central Spanish plateau. This is ahilly area (altitude of sampling sites rangedfrom 600 to 1100 m a.s.l.), which con-sists of the Montes de Toledo and SierraMorena mountain chains. The Guadianariver valley, a fragmented Mediterraneanhabitat, connects them both. The habitat isMediterranean and characterised by Quer-cus ilex forests and scrublands (dominatedby Cystus spp., Pistacia spp., Rosmarinusspp., Erica spp. and Phyllirea spp.) withscattered pastures and small areas of crops.Annual rainfall is extremely variable (rang-ing from 300 to 700 mm) and the climateis Mediterranean with a continental influ-ence. The wet season typically starts inSeptember–October and contributes mostof the annual rainfall. The dry season(from June to September) is when foodand water resources become limited forungulates. Seasonal streams cross manyof the estates but in most of them, wa-ter is retained in artificial waterholes allyear around. These waterholes consist ofsmall ditches across the riverbeds. In theseMediterranean woodland habitats, most ofthe hunting estates are devoted exclusivelyto recreational wild boar and red deer hunt-ing, and only a few still maintain tradi-tional livestock. Nevertheless, cattle hadbeen present on most estates until theywere fenced for hunting purposes, severaldecades ago. Other land use, such as arableand extensive sheep and goat farming aremore frequent in most non-fenced huntingareas.

2.2. Sampling and diagnosis

Data were collected from a cross-sectional national survey of TB infec-tion in wild boar and wild Iberian reddeer at 19 sites (Tab. I) in SCS, withina 50 000 km2 area concentrated on the

Table I. Mean prevalence of macroscopic TBcompatible lesions (%) in the wild boar and reddeer at sampling sites included in the risk factoranalysis.

Sampling Wild boar Red deer

site N Prevalence (%) N Prevalence (%)

1 33 33.3 18 22.2

2* 23 30.4 10 0.0

3* 20 50.0 36 5.6

4 18 88.9 25 44.0

5 20 55.0 29 20.7

6 20 40.0 14 7.1

7 13 61.5 39 2.5

8 13 38.5 10 0.0

9 34 58.8 13 23.1

10 11 18.2 24 4.2

11 15 46.7 27 0.0

12 41 80.5 150 18.7

13 62 51.6 38 13.2

14 14 100.0 30 30.0

15 13 61.5 11 18.2

16 3 100.0 31 19.4

17* 38 36.8 15 6.7

18 10 60.0 34 0.0

19 11 36.4 20 0.0* Indicate open estates.

provinces of Ciudad Real and Toledo (re-gion of Castilla-La Mancha, N = 14) andborder areas (regions of Andalucía, N = 4;and Extremadura, N = 1). Geographicalcoordinates are from 37◦ 13’ 48” Nto 39◦ 31’ 43” N in latitude and06◦ 34’ 06” W to 2◦ 25’ 54” W in lon-gitude (Fig. 1). Sampled hunting estatesranged in size from 408 to 19 328 hectares.We chose a sample of sites that exhib-ited a range of management factors, vari-ations in deer and wild boar abundanceand landscape diversity. Sampling tookplace during the regular hunting seasons(October–February), from October 1999to February 2004. We obtained data from574 red deer and 412 wild boar culled

454 J. Vicente et al.

Figure 1. Map of south central Spain (woodlands are shaded) showing the sampling sites. Numberinside the left semi-circle represents TBL prevalence in the wild boar, and left semi-circle darkportion represents TBL prevalence in the wild boar in relation to total semi-circle (100%). Rightsemi-circle similarly applies to the red deer.

by hunters in the different estates. In thefield, we arbitrarily sampled a random ageand sex stratified subset of individuals ofeach species. A necropsy was performed,including detailed determination of mor-phometry, weight, and sex. Based on tootheruption patterns, boars between 7 and12 months were classified as either juve-niles (N = 92), sub-adults of between 12and 24 months (N = 154), or adults ofover 2 years of age (N = 161) [29] (ap-proximate age could not be determined for7 individuals). In the case of the red deer,age was determined from sectioning in-cisor 1 [18] and animals were grouped intosubadults (3 to 4 years) N = 167 and(4) adults (≥ 5 years) N = 396 (approxi-mate age and sex could not be determinedfor 11 individuals).

The presence of TBL was diagnosed bynecropsy of the entire animal with macro-scopic inspection of lymph nodes and ab-dominal and thoracic organs. Lymph nodeswere dissected and sectioned serially andsystematically examined for gross lesions(see [10, 11, 34] for details). The presenceof TB infection at the level of the estatewas confirmed by culture and using datafrom a separate molecular study, whichwas based on a sub-sample of the sameanimals [11]. Some misclassification couldhave resulted from the omission of confir-matory cultures from all animals, but it wasimpossible to quantify how such bias mayhave influenced prevalence estimates forboth species. Nevertheless, TBL criterionprovided relative values that were com-parable between localities, and permitted

Risk factors and TB in big game in Spain 455

exploratory analyses of risk factors. Thediagnostic value of TBL for epidemiolog-ical analyses in the study area is furtherdiscussed in [34].

2.3. Management and environmentalvariables

We visited the sampling sites in Septem-ber 2002, immediately before the huntingseason, in order to obtain field estimatesof the relative abundance of wild boar andred deer and habitat variables. The habi-tat and management variables consideredin the study were chosen on the basis oftheir likely epidemiological relevance andpotential influence on the characteristicsof wild ungulates and livestock (Tab. II).Habitat use and structure in the 19 studyareas were recorded at points spaced ev-ery 200 m along linear transects (N = 20points per estate) and were used to cal-culate mean values for each estate. Inter-views were carried out with gamekeepersto characterise the management practicesemployed on each estate. Data was col-lected on general estate-related features,watering sites (availability and number ofartificial waterholes and natural water bod-ies on each estate), feeding practices, thenumber and location of livestock (and theirexposure to wild ungulates) and the pres-ence of other large game species (Tab. II).All variables were related to the year 2002.

2.4. Estimation of red deer and wildboar relative abundance

Wild boar relative abundance esti-mates were based on dropping frequencycounts [31]. Briefly, each count consistedof N = 40 transects of 100 m, dividedinto 10 sectors of 10 m in length. Drop-ping frequency was defined as the aver-age number of 10 m sectors with wildboar droppings present in each transect

(DF = ΣDi/n; where “D” is the num-ber of dropping-positive sectors and rangesfrom 0 to 10, and “n” is the number of100 m transects, usually 40). Red deer rel-ative abundance data was obtained by spot-lighting at night. Counts were transformedinto indices of abundance per km [35]and to deer per ha [17] (mean transectlength 14.84 ± 13.90 km). At least one ex-perienced observer (Joaquin Vicente andChristian Gortazar) participated in eachtransect.

2.5. Statistical analysis

Non-parametric Mann-Whitney testswere used to compare TBL prevalence be-tween fenced and open estates for wildboar and red deer, respectively. Quantita-tive exploratory analysis of risk factors forTBL was carried out using a two-stageanalysis. First, the associations betweenthe hypothesised risk factors and wild boarand red deer TBL prevalences (%) re-spectively, were analysed using SpearmanRank correlations (N = 19). All the fac-tors that captured the effect of any setof highly correlated variables for whichP < 0.1 were selected for inclusion inthe final models for each species (Tab. II).For categorical variables, nonparametricKruskall-Wallis analysis was used for ini-tial assessment of associations, and se-lected variables were then jointly evaluatedin a multiple logistic model. The individ-ual TBL status of wild boar (N = 407)and red deer (N = 565) were the re-spective response variables (as binomialvariables, i.e. present or absent) in two sep-arate models. Since sampling across differ-ent populations was not homogeneous inrelation to age and sex, and because pre-vious work suggested these factors wereinfluential [34], statistical analyses wereconducted at the individual level to controlfor them. Age was included as a continu-ous discrete explanatory variable in models

456 J. Vicente et al.

Table II. Variables included in the study, indicating which were significantly associated with TBL(excluding other highly correlated variables) in wild boara and red deerb models (rS , P < 0.1,N = 19).

Estate-related features and management practices

General: Geographic location, area (ha), boundaries limiting with big hunting estates (%), introductionsor purchasing (binomial), captures (binomial), predator control (binomial), agricultural lands for hunting(ha and % related to the estate), fencing status (binomial), years since fencinga, boundary fenced (%)

Supplemental feeding: Supplemental feeding for wild boar (binomial), Supplemental feeding for reddeer (binomial), feeding method (binomial: spreading or provided in feeders), wild boar and red deersharing feeding sites (binomial), wild boar feeding site density, red deer feeding site density, number ofwild boar per feeding site indexb (wild boar dropping frequency abundance index/wild boar feeding sitedensity), number of red deer per feeding siteb (deer per 100 ha/red deer feeding sites density), baitingprevious to hunting season (binomial)

Watering: Number of wild boar watering sites, number of red deer watering sites, wild boar wateringsite density, red deer watering site density, number of wild boar per watering site indexa,b (wild boardropping frequency abundance index/ wild boar watering site density), number of red deer per wateringsite (deer per 100 ha/red deer watering sites density)

Host demography: Wild boar dropping frequency abundance indexa,b, red deer kilometric abundanceindex (deer/km), deer density (deer/ha).

Other wildlife: Kilometric abundance indices: Mouflon, fallow deer, roe deer, Iberian hare, wild rabbit,fox, other carnivores.

Dropping or pellet abundance frequency indices: Iberian hare, wild rabbit, carnivore

Livestock: Livestock (cattle, goat and sheep, binomial), sheep density (per ha), historic livestock pres-ence (years), livestock sharing pastures, feeders or watering site with hunting (binomial)

Estate-related general environmental conditions (mean values for each estate)

Habitat availability (%): Scrublands, Dehesa (savannah-like habitat), Mediterranean hardwoodforesta,b (Quercus spp.), pine plantations, pastures, riparian habitat, agricultural areas

Land cover and habitat structure (each 200 m): Scruba /forest/grassa,b /soil covers (%), forest andscrub diversity/10 m, No. Quercus trees/10 m, soil compactness (cat.: 1-5)

for both wild boar and red deer and sex wasincluded as a categorical binomial explana-tory variable in the red deer model. Tostandardise comparisons, we included inthe analysis only wild boar over 7 monthsold and only red deer over 2 years old [34].The number of non-fenced facilities (2 outof 19 sampling sites, see Tab. II) was toosmall to support a separate statistical anal-ysis, and thus they were excluded from theanalysis. We used a stepwise strategy toobtain the final model.

Mantel tests [19] were used to assessthe spatial association between TBL preva-lence in wild boar and red deer acrossdifferent estates, respectively. One matrix

was a measure of dissimilarity between ev-ery pair of estates in terms of their TBLprevalence (the difference), whilst the sec-ond was the Euclidean distance betweenestates. The Pearson correlation coefficientbetween the elements of the matrices wasused as the test statistic and its significancewas assessed by permuting the row labelsof one matrix relative to the other fivethousand times [19]. The distance betweenestates was assumed to be the distance be-tween their nearest sampling points usedto characterise environmental and manage-ment features, and was calculated with the“Distance Operator” tool of Idrisi 32 soft-ware version I32.21 (The Clark labs.�,

Risk factors and TB in big game in Spain 457

Table III. General statistics for the most relevant land uses, habitat data and management practicesat the estate level (N = 19).

Variable Mean ± SD Median (25 and 75% quartiles) RangeEstate area (ha) 4616 ± 4637 3000 (1700−6862) 408−19328Estate boundary fenced (%) 87.1 ± 32.0 100 (100−100) 0−100Time completely fenced (years since) 21 ± 12.6 19 (14−34) 0−43Cultivated dehesas or pastures for hunting (%) 8.5 ± 15.5 2.1 (0.1−13.3) 0−45. 5Wild boar feeding sites (No. per 100 ha) 0.3 ± 0.01 0 (0−0.53) 0−1.8Red deer feeding sites (No. per 100 ha) 0.3 ± 0.01 0 (0−0.65) 0−1.8Wild boar per feeding site index 89.2 ± 208.1 0 (0−95.1) 0−880Number of red deer per feeding site 32.1 ± 65.0 0 (0−49.0) 0−261.9Number of watering sites (No. per 100 ha) 0.6 ± 0.5 0.3 (0.1−0.6) 0.11−2.3Number of wild boar per watering site index 79.7 ± 61.6 77.9 (16.9−129.5) 0−188.9Number of red deer per watering site 59.5 ± 39.0 57.5 (29.7−92.9) 0−130.0Red deer abundance index (KAI) 9.9 ± 5.1 6.4 (5.6−13.7) 3.33−18.4Red deer abundance (deer per 100 ha) 24.4 ± 16.3 17.4 (11.0−40.9) 3.42−55.1Wild boar abundance (dropping frequency 0.6 ± 0.9 0.2 (0.1−0.7) 0−4index)Land uses (%)

Quercus spp. forest 30.7 ± 24.8 23.8 (4.8−35.3) 4.5−100Dehesa 28.5 ± 26.4 28.6 (0−55) 0−80.9Pine plantation 17.4 ± 27.2 0 (0−37.5) 0−71.4Pastures 0.6 ± 1.9 0.1 (0−0.3) 0−7.2Scrubland 20.6 ± 20.2 11.5 (4.8−35.0) 0−61.9

Habitat structure (% in 25 m radio)Scrub cover 32.7 ± 11.9 31.5 (22.7−41.7) 16.2−58.7Quercus spp. forest cover 22.1 ± 11.3 23.8 (4.8−35.2) 0.3−50.2Grass cover 30.9 ± 13.9 23.9 (4.8−35.2) 7.6−50.6Soil cover 29.02 ± 10.49 28.6 (21.3−38.9) 10−51.67

Clark University, Massachusetts, USA).Statistical significance was assumed wher-ever P < 0.05. We employed SPSS 10.0.6(SPSS Inc., 1999) and Genstat (Genstat 5Committee 1993) statistical packages.

3. RESULTS

3.1. Characteristics of the samplingestates and populations

Habitat use and structure, and manage-ment practices in the study estates aredescribed in Table III. The most commonhabitats in the study area were Quercusspp. forests (Evergreen oak Quercus ilex

as predominant species), open pasture withQuercus spp. trees (savannah-like habi-tats typically known as “Dehesas”) andMediterranean scrublands (Cistus spp. andQuercus spp. as predominant species).

Fifteen of the 19 study estates werecompletely fenced and two were incom-pletely fenced (60% and over 95% ofthe perimeter respectively, the latter con-sidered as fenced for statistical analyses)when the interviews were performed.Seven out of the 19 study sites fed wildboar artificially all year round and only12 provided food to wild boar in the formof bait immediately before and during thehunting season. Amongst the 19 sites, reddeer were fed regularly all year round at

458 J. Vicente et al.

eight, while 12 (including those feeding allyear round) provided food as bait both im-mediately before and during the huntingseason. Seven out of the 19 study estatesfed both wild boar and red deer all yearround. Domestic sheep were kept on onlythree of the study estates and neither cat-tle nor goats were present on any. Thirteenof the interviewed game managers reportedfinding emaciated and moribund wild boarand/or red deer, mainly in the summerand usually near watering sites, althoughthese were not specifically associated withTB infections.

3.2. Risk assessment

The distribution of TBL across wildboar and red deer populations from thestudy sites is shown in Table I and Figure 1.No statistical significant spatial associationwas found between the dissimilarity valuesfor the TBL prevalence per estate and theEuclidean distance between them for ei-ther species (wild boar Pearson r = −0.10,N = 171; P = 0.80, red deer Pearsonr = −0.13, N = 171, P = 0.87).

Those potential risk factors that wereselected for inclusion in subsequent mod-els are shown in Table II (the correlationcoefficients ranged from 0.39 to 0.64 inabsolute value). For both the wild boarand red deer these models included an ag-gregation index for wild boar at wateringsites (Fig. 2), the wild boar dropping fre-quency index (an abundance estimate) andthe availability of Quercus spp. forest habi-tat and grass cover. Aggregations of boarand red deer at feeding sites were includedonly in the red deer model. Scrub coverwas selected only in the final wild boarmodel (the only negative correlation, RS =−0.59, P < 0.01, N = 19).

TB prevalence in both the wild boarand red deer did not differ when com-paring open (N = 3) and fenced estates(N = 16) (Mann-Whitney test, P always

> 0.05). The range of management prac-tices, host abundances and habitat descrip-tors for open estates overlapped with thoseof fenced estates. Nevertheless, open es-tates were excluded from the models sincethey were numerically under-represented.

The definitive logistic models fitted ade-quately (Hosmer and Lemeshow goodness-of-fit test: χ2 = 4.48, d.f. 8, P = 0.81,χ2 = 11.33, d.f. = 8, P = 0.18 for wildboar and red deer, respectively). However,neither model explained the majority of thevariation in the data (R2

Nagelkerke = 0.15,

R2Nagelkerke = 0.13 for the wild boar and

red deer respectively). In both cases, thenull hypothesis (H0) of β = 0 was rejected(Wald χ2, P < 0.05 in both models). Theresults of the wild boar and red deer mod-els are shown in Table IV. For significantterms in the models, confidence intervalsfor the Wald statistic did not span the nullOdds ratio (exponentiated β) value in anycase (Tab. IV). The aggregation of the wildboar at artificial watering sites was signif-icantly associated with an increased riskof TBL presence in both species (Figs. 2aand 2b). Aggregation of the wild boarat feeding sites significantly increased therisk of TBL presence in red deer (the in-dex value was over twice as high in TBLpositive animals compared to negative an-imals). Scrubland cover was significantlyassociated with a decreased individual riskof TBL presence in the wild boar, andthere was a marginal positive association(P = 0.06, Tab. IV) between the availabil-ity of Quercus spp. forest habitat and therisk of testing positive for TBL.

4. DISCUSSION

The results of this study confirm pre-vious findings by the authors [34], insuggesting that differences exist betweenestates in terms of not only the pres-ence, but in particular in the prevalenceof TBL. To date it has been unusual to

Risk factors and TB in big game in Spain 459

Figure 2. Mean TBL prevalence (%) in relation to the estimated potential aggregation of wildboar (a) and red deer (b) at watering sites at estate level (N = 19).

find a TBL-free ungulate game popula-tion in SCS. The main finding of thepresent study is the association betweenthe aggregation of hosts (mainly a conse-quence of hunting management in Mediter-ranean habitats) and the presence of TBLin potential TB reservoirs. Previous re-

ports suggested similar relationships forother infectious pathogens in the studyarea [31]. Our spatial analyses of the TBLpattern across estates revealed that man-agement units close together in space didnot tend to be similar in terms of TBLprevalence, which suggests that there was

460 J. Vicente et al.

Table IV. Final logistic regression models for the association between hunting estate characteristicsand the individual TBL status of wild boar and red deer respectively.

Variable β (S.E.) Wald95%CI d.f. P Sig. Exp(β)95%CI

Red deer model

Sex 0.55 (0.27) 3.96 (2.02−5.90) 1 0.04 1.73 (1.01−2.96)

Wild boar/watering site 0.01 (0.002) 15.99 (12.79−19.19) 1 < 0.001 1.01 (1.01−1.02)

Wild boar/feeding site 0.001(0.0005) 3.87 (1.94−5.80) 1 0.05 1.00 (1.00−1.01)

Wild boar model

Age class 0.54 (0.15) 13.23 (9.55−16.91) 1 < 0.001 1.71 (1.28−2.28)

Wild boar/watering site 0.01 (0.002) 10.08 (8.66−12.09) 1 0.001 1.01 (1.00−1.01)

Quercus spp. forest 1.23 (0.65) 3.52 (1.66−5.38) 1 0.06 3.43 (0.95−12.47)

Scrubland cover –0.02 (0.01) 6.35 (3.175−9.52) 1 0.01 0.98 (0.96−0.99)

no spatial aggregation (which would leadto spatial autocorrelation in our data). Inother words, after accounting for indi-vidual factors (which are species-specific)management scale factors seemed to bemore influential in explaining differencesin TBL prevalence between estates thantheir spatial location. This suggests thatthe management characteristics of each es-tate are the major determinants of diseasetransmission, and should therefore be con-sidered as the appropriate target for controlpolicies. For example, in the present pa-per estates that were geographically closesometimes had very different levels ofTBL (e.g. study sites in the Toledo Moun-tains exhibited TBL prevalence from 18%to 90% in wild boars, Fig. 1). The samplingdesign we employed included a broad va-riety of hunting management practices andMediterranean landscapes from throughoutSCS (Tab. III). For instance, the numberof red deer per feeding site, the index ofwild boar per feeding site and the pro-portion of open grassland (cultivated dehe-sas, pastures for big game) varied widelyamongst estates, although unfenced popu-lations were under-represented. Red deerand wild boar on fenced estates can beconsidered as semi-domestic livestock andare now present in SCS at far higher den-sities than several decades ago [9, 30],

where TB has become established as a self-maintained, endemic disease [14, 34]. Thepotential influence of game managementintensification on TB dynamics may be il-lustrated by the positive association at theestate level between the time at which anestate was fenced and the prevalence ofTBL in wild boar populations.

The present analyses of risk factorsconstitute a correlational study, so causeand effect cannot be demonstrated, andwe therefore consider our conclusions asexploratory. These analyses included vari-ables related to both management practicesand habitat structure; nevertheless, it is dif-ficult to isolate the effect of a particularcomponent. For example, the aggregationof animals at waterholes created on naturalriverbeds is a factor related to both man-agement and natural features of Mediter-ranean habitats. Management of ungulatesin these habitats tends to promote ag-gregation of ungulates, and subsequentlyenhances opportunities for transmission.This may arise because ungulates comeinto contact with a higher proportion ofindividuals either at the inter- and intra-specific levels, or with a more heavilycontaminated habitat (i.e. direct and/or in-direct mycobacteria transmission). In par-ticular, we found that the potential ag-gregation of wild boar at watering sites

Risk factors and TB in big game in Spain 461

significantly increased the risk that an in-dividual was TBL positive, in both hostspecies. These findings were in agreementwith the known health risks to small gamein southern Spain associated with artificialfeeding and water provision [15, 20]. Thefrequency of contact between wild mam-mals at watering sites could increase dur-ing the drought season (summer), since thefew water holes are more intensely used.TB spread may occur by ingestion or in-halation of nasal and oral excretions fromcoughing or sneezing of infected individ-uals or from infected dust particles [5].Spread may also occur indirectly fromcontaminated vegetation, water, mud orfomites [24]. It is known that oral excretionfrom affected mandibular lymphnodes oc-curs through fistulae [10]. Wild boar activ-ity around these places (such as wallowing,brushing, drinking, defecating, urinating,and mating) may enhance the possibili-ties of environmental contamination andTB transmission. Watering sites may alsoconcentrate mating behaviour and brushingat the end of the summer in red deer [4].In addition, the practice of providing ar-tificial water at these sites may cause theformation of a multiple “piosphere” effect,which consists in the existence of a zoneof high ungulate utilisation in woody veg-etation extending far beyond the area [2],and subsequently, an increased risk of indi-rect transmission. Models for each speciessuggest that interspecific indirect transmis-sion at these sites is probably more likelyto occur from the wild boar to red deer.Both the wild boar and red deer haveshared free access to watering sites, al-though more research is needed concerningthe use and inter and intraspecific interac-tions at these points. For some potentialhosts the main route of TB infection maybe ingestion of infected carrion [28], and inthe study sites described here TB-infectedcarcasses and subsequent scavenging bywild boars both occur (the authors, un-published observations). In addition, se-

vere droughts in Mediterranean habitatsand seasonal scarcity of water resourcesmay exacerbate the effect of TB infectionon hosts [3] and, watering (and also feed-ing) sites may attract tuberculous animalsin the advanced stages of infection (theauthors, unpublished observations). Water-ing sites used by domestic livestock couldalso be frequented by wild ungulates.

The presence of TBL in red deer waspositively associated with the abundanceof wild boar per feeding site index. Thismay be explained in part by indirect inter-specific transmission opportunities, whichseem most probable in the direction ofwild boar to red deer than vice versa.Deer have free access to wild boar sup-plementary food since this is provided byspreading on the ground. The prevalenceof TBL in the wild boar is higher than inred deer, and they may contaminate theenvironment with mycobacteria in placeswhere the two species co-exist [10]. Incontrast, red deer are fed with cereals orconcentrate pellets, frequently given in el-evated feeders, which are not accessibleto wild boar. M. bovis can survive for ex-tended periods of time on different typesof animal feeds (unpublished data citedby [27]1). The importance of transmissionby direct contact (aerosol transmission)or by indirect contact via contaminationof feed sites has been demonstrated inwhite-tailed deer (Odocoilus virginianus)in Michigan [10, 25]. However, we failedto detect any significant relationships con-sistent with intraspecific transmission ofTB amongst deer at feeding sites. Higherprevalences of TBL in red deer in thepresent study area compared to previouslypublished estimates have been suggestedto be due in part to interspecific transmis-sion from wild boar [34], but confirmationrequires more observational and experi-mental research.

1 Palmer M.V., personnal communication.

462 J. Vicente et al.

A decreased TBL prevalence in wildboar (and in red deer in the initial stage ofanalysis) was seen in estates with higherpercentages of scrub cover. Scrublandsprovide food for deer to browse but sincethey are less attractive to wild boar, inter-actions and opportunities for transmissionbetween the two species are less likely.Conversely, the presence of TBL in indi-vidual wild boar and in red deer at the pop-ulation level, were marginally (P = 0.06)and positively associated with the avail-ability of Quercus spp. forest (Mediter-ranean hardwood). Miller et al. [21] sug-gested that woodland areas provide shady,moist conditions under which M. bovismight survive longer in the environment.Quercus spp. woodlands could provide thebest available environment for mycobacte-ria persistence in Mediterranean habitats,where rainfall is highly variable amongstseasons, but retention of moisture in wood-lands may favour the environmental per-sistence of mycobacteria [32, 33]. Suchhabitats become more important in an epi-demiological sense if they are positivelyselected by hosts. Quercus spp. acornsfrom Mediterranean woodlands are inten-sively foraged by wild ungulates duringthe autumn. If environmental contamina-tion exists, wild boar and red deer feedingin the area (by rooting and muzzling whilesearching for acorns) could either ingestor inhale mycobacteria. Acorn grazing isalso a common practice in free-roamingIberian pigs in large areas of southernSpain [26] and hence infectious interac-tions could occur.

In both species, levels of TBL in un-fenced estates overlapped with the rangefound for fenced populations (Tab. II), sug-gesting that factors other than fencing andprovisioning may be also important de-terminants of TBL prevalence. This find-ing also raises the possibility that cap-tive wildlife may act as disease reservoirsand risks for native free-ranging wildlifein addition to domestic livestock. Since

unfenced populations were numericallyunder-represented in the present study, fur-ther research is required for a robust com-parison of fenced and unfenced estates.

The current study represents an initialidentification of the potential epidemiolog-ical risk factors that may be important inthe development of strategies aimed at con-trolling TB in Spain. Wild game ungulatesare not currently considered in disease con-trol programmes, but recent studies, in-cluding the present work, suggest that todo so could have significant consequencesfor disease control efforts. Host popula-tions enclosed by a fence (more or lesspermeable) are similar in many ways to ex-tensively managed livestock, and thereforecould be referred to as captive or semi-domesticated animals. However, a precisedefinition for large game mammals infenced estates is problematic since man-agement systems vary widely [12,16]. Thisis even more complex for the ubiquitouswild boar, which is easily able to “under-cross” fences.

Although bovine TB in wildlife is likelyto have originated from domestic cat-tle [24], wildlife reservoirs may act assources of infection for livestock in someareas, and therefore could hamper diseasecontrol programmes. A possible link be-tween the presence of livestock in thepast and subsequent detection of Mycobac-terium tuberculosis complex strains in boarand deer current big game was establishedin our study area by [11]. Because TBis extremely difficult to eradicate fromwildlife once established, the priority is toclearly identify the route of transmissionto cattle, and then control it. Although ac-tivities are now clearly segregated acrossthe majority of estates, extensive livestockproduction (cattle, goats and swine) inmarginal areas of SCS Spain still over-laps with those devoted to hunting. Ourresults suggest that indirect transmission ofTB in grazing areas across the livestock-game interface is likely to occur. Game

Risk factors and TB in big game in Spain 463

ungulates that can easily cross fences mayplay an important role in transmission atthe wildlife/livestock interface. Foragingexcursions of livestock into woodlands de-voted to hunting (a practice still commonin some marginal lands in SCS) may alsobe a route of transmission between biggame and domestic livestock. As men-tioned above, watering sites for domesticlivestock may also attract wild ungulates,especially during the dry season, with sub-sequent enhanced opportunities for infec-tious interactions. Investigations into thepatterns of dissemination of bacilli in theenvironment and the mode of transmissionof infection between wild ungulates andlivestock are research priorities. The de-velopment of effective and sustainable TBmanagement policies will need to take ac-count of the relationships between estatehusbandry practices and the transmissionof infection at the large game-domesticlivestock interface.

ACKNOWLEDGEMENTS

This study was supported by projectAGL2001-3947, MCYT and FEDER andproject AGL2005-07401 MEC. This is alsoa contribution to the agreement between FG-UCLM and Grupo Santander, and to the agree-ment between Yolanda Fierro and the UCLM.We wish to thank A. Peña, and all our col-leagues at IREC for their kind help. We aregrateful to Neil Walker for statistical advice andD. Richard Delahay for its comments on themanuscript and edition of the English.

REFERENCES

[1] Artois M., Delahay R., Guberti V.,Cheeseman C., Control of infectiousdiseases of wildlife in Europe, Vet. J. (2001)162:141−152.

[2] Brits J., Van Rooyen M.W., Van Rooyen N.,Ecological impact of large herbivores on thewoody vegetation at selected watering pointson the eastern basaltic soils in the KrugerNational Park, Afr. J. Ecol. (2002) 40:53−60.

[3] Caron A., Cross P.C., Du Toit J.T.,Ecological implications of bovine tu-berculosis in African buffalo herds, Ecol.Appl. (2003) 13:1338−1345.

[4] Carranza J., Fernandez-Llario P., GomendioM., Correlates of territoriality in rutting reddeer, Ethology (1996) 102:793−805.

[5] Cousins D.V., Mycobacterium bovis infec-tion and control in domestic livestock,Rev. Sci. Tech. Off. Int. Epizoot. (2001)20:71−85.

[6] De Lisle G.W., Joyce M.A., Yates G.F.,Wards B.J., Hoyle F.P., Mycobacteriunavium infection in a farmed deer herd, N. Z.Vet. J. (1995) 43:1−3.

[7] Delahay R.J., De Leeuw A.S., Barlow A.M.,Clifton-Hadley R.S., Cheeseman, C.L., Thestatus of Mycobacterium bovis infection inUK wild mammals: a review, Vet. J. (2003)164:90−105.

[8] Gilbert M., Mitchell A., Bourn D.,Mawdsley J., Clifton-Hadley R., WinW., Cattle movements and bovine tuber-culosis in Great Britain, Nature (2005)435:491−496.

[9] Gortazar C., Herrero J., Villafuerte R.,Marco J., Historical examination of the sta-tus of large mammals in Aragon, Spain,Mammalia (2000) 64:411−422.

[10] Gortazar C., Vicente J., Gavier-Widen D.,Pathology of bovine tuberculosis in theEuropean wild boar (Sus Scrofa), Vet. Rec.(2003) 152:779−780.

[11] Gortazar C., Vicente J., Samper S., GarridoJ., Fernández-De-Mera I.G., Gavín P., JusteR.A., Martín C., Acevedo P., De La PuenteM., Höfle U., Molecular characterization ofMycobacterium tuberculosis complex iso-lates from wild ungulates in South-CentralSpain, Vet. Res. (2005) 36:43−52.

[12] Gortazar C., Acevedo P., Ruiz-Fons F.,Vicente J., Disease risks and overabundanceof game species, Eur. J. Wild. Res. (2006)52:81−87.

[13] Haydon D.T., Cleaveland S., Taylor L.H.,Laurenson M.K., Identifying reservoirsof infection: a conceptual and practicalchallenge, Emerg. Infect. Dis. (2002)8:1468−1473.

[14] Hermoso de Mendoza J., Parra A., Tato A.,Alonso J.M., Rey J.M., Peña A., Garcia-Sanchez A., Larrasa J., Teixido J., ManzanoG., Cerrato R., Pereira G., Fernandez-LlarioP., Hermoso de Mendoza M., Bovine tuber-culosis in wild boar (Sus scrofa), red deer(Cervus elaphus) and cattle (Bos taurus)

464 J. Vicente et al.

in a Mediterranean ecosystem (1992–2004),Prev. Vet. Med. (2006) 74:239−247.

[15] Höfle U., Gortazar C., Ortíz J.A., Knispel B.,Kaleta E.F., Outbreak of Trichomonosis in awoodpigeon wintering roost, Eur. J. Wildl.Res. (2004) 50:73−77.

[16] Hunter D.L., Tuberculosis in free-ranging,semi-free-ranging and captive cervids,Rev. Sci. Tech. Off. Int. Epizoot. (1996)15:171−181.

[17] Jarvinen O., Vaisanen R.A., Estimating rel-ative densities of breeding birds in the linetranscect method, Oikos (1975) 26:316−322.

[18] Klevezal G.A., Kleinenberg S.E., Age de-termination of mammals from annual layersin teeth and bones, Academy of SciencesUSSR Translation Dept. of the Interior andNational Science Foundation, U.S. Dept.of Commerce, Clearinghouse for FederalScientific and Technical Information,Springfield, 1967.

[19] Manly B.F.G., Randomization, Bootstrapand Monte Carlo methods in Biology,Chapman and Hall, London, 1997.

[20] Millan J., Gortazar C., Villafuerte R.,Mycobacterium avium disease in wild red-legged partridges (Alectoris rufa), Eur. J.Wildl. Res. (2004) 50:97−99.

[21] Miller R., Kaneene J.B., Fitzgerald S.D.,Schmitt S.M., Evaluation of the influenceof supplemental feeding of white-tailed deer(Odocoileus virginianus) on the prevalenceof bovine tuberculosis in the Michiganwild deer population, J. Wildl. Dis. (2003)39:84−95.

[22] Ministry of Agriculture, Programa nacionalde erradicacion de tuberculosis bovina pre-sentado por España para el 2006, SecretariaGeneral de Agricultura, Madrid, 2005.

[23] Mollenkopf H.J., Kursar M., KaufmannS.H., Immune response to postprimary tuber-culosis in mice: Mycobacterium tuberculosisand Mycobacterium bovis bacille Calmette-Guerin induce equal protection, J. Infect.Dis. (2004) 190:588−597.

[24] Morris R.S., Pfeifer D.U., Jakson R., Theepidemiology of Mycobacterium bovis infec-tions, Vet. Microbiol. (1994) 40:153−177.

[25] O’brien D.J., Schmitt S.M., Fierke J.S.,Hogle S.A., Winterstein S.R., Cooley T.M.,Moritz W.E., Diegel K.L., Fitzgerald S.D.,Berry D.E., Kaneene J.B., Epidemiology ofMycobacterium bovis in free-ranging white-tailed deer, Michigan, USA, 1995-2000,Prev. Vet. Med. (2000) 54:47−63.

[26] Parra A., Fernandez-Llario P., Tato A.,Larrasa J., García A., Alonso J.M.,De Mendoza M.H., De Mendoza J.H.,Epidemiology of Mycobacterium bovisinfections of pigs and wild boars using amolecular approach, Vet. Microbiol. (2003)97:123−133.

[27] Palmer M.V., Waters W.R., Whipple D.L.,Shared feed as a means of deer-to-deer trans-mission of Mycobacterium bovis, J. Wildl.Dis. (2004) 40:87−91.

[28] Ragg J.R., Mackintosh C.G., Moller H., Thescavenging behaviour of ferrets (Mustelafuro), feral cats (Felis domesticus), pos-sums (Trichosurus vulpecula), hedgehogs(Erinaceus europaeus) and harrier hawks(Circus approximans) on pastoral farmlandin New Zealand: Implications for bovine tu-berculosis transmission, N. Z. Vet. J. (2000)48:166−175.

[29] Sáenz De Buruaga M., Lucio A.J., PurroyJ., Reconocimiento de sexo y edad enespecies cinegéticas, Diputación Foral deAlava, Vitoria, 1991.

[30] Sáez-Royuela C., Tellería J.L., The increasedpopulation of the wild boar (Sus scrofa L.) inEurope, Mammal Rev. (1986) 16:97−101.

[31] Vicente J., Segalés J., Höfle U., Balasch M.,Plana-Durán J., Domingo M., Gortazar C.,Epidemiological study on porcine circovirustype 2 (PCV2) infection in the Europeanwild boar (Sus scrofa), Vet. Res. (2004)35:243−253.

[32] Vicente J., Fierro Y., Gortazar C.,Seasonal dynamics of the fecal excre-tion of Elaphostrongylus cervi (Nematoda,Metastrongyloidea) first stage larvae inIberian red deer (Cervus elaphus hispanicus)from South Spain, Parasitol. Res. (2005)95:60−64.

[33] Vicente J., Fernández de Mera I.G., GortazarC., Epidemiology and risk factors analysisof elaphostrongylosis in red deer (Cervuselaphus) from Spain, Parasitol. Res. (2006)98:77−85.

[34] Vicente J., Höfle U., Garrido J.M., Fernándezde Mera I.G., Juste R., Barral M., GortazarC., Wild boar and red deer display highprevalences of tuberculosis-like lesions inSpain, Vet. Res. (2006) 37:107−119.

[35] Whipple J.D., Rollins D., Schacht W.H.,A field simulation for assessing accuracyof spotlight deer surveys, Wildl. Soc. Bull.(1994) 22:667−673.

[36] Wobeser G., New and emerging diseases,The wildlife interface, Can. Vet. J. (2002)43:793.