Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Contents lists available at ScienceDirect
Preventive Veterinary Medicine
journal homepage: www.elsevier.com/locate/prevetmed
Quantification of Mycobacterium bovis transmission in a badger vaccine field
trial
I. Aznar
a,b,c,⁎
, K. Frankena
b
, S.J. More
a
,J.O’Keeffe
c
, G. McGrath
a
, M.C.M de Jong
b
a
UCD Centre for Veterinary Epidemiology and Risk Analysis, UCD School of Veterinary Medicine, University College Dublin, Belfield, Dublin 4, Ireland
b
Quantitative Veterinary Epidemiology group, Wageningen Institute of Animal Sciences, Wageningen University & Research, P.O. Box 338, 6700 AH Wageningen, The
Netherlands
c
Department of Agriculture, Food and the Marine, Kildare St., Dublin 2, Ireland
ARTICLE INFO
Keywords:
Mycobacterium bovis
Badgers
Vaccine efficacy for susceptibility
Vaccine efficacy for infectiousness
Bacille calmette-Guérin (BCG)
Transmission
Basic reproduction ratio
ABSTRACT
In the UK and Ireland, Bacille Calmette-Guérin (BCG) vaccination of badgers has been suggested as one of a
number of strategies to control or even eradicate Mycobacterium bovis infection in badgers. In this manuscript, we
present the results of a badger field trial conducted in Ireland and discuss how the novel trial design and ana-
lytical methods allowed the effects of vaccination on protection against infection and, more importantly, on
transmission to be estimated. The trial area was divided into three zones North to South (A, B and C) where
vaccination coverages of 0, 50 and 100%, respectively, were applied. Badgers were trapped over a 4 year period.
Badgers were assigned to either placebo or vaccine treatment, with treatment allocation occurring randomly in
zone B. Blood samples were collected at each capture, and serology was performed in these samples using a
chemiluminescent multiplex ELISA system (Enfer test). The analysis aimed to compare new infections occurring
in non-infected non-vaccinated badgers to those in non-infected vaccinated ones, while accounting for the zone
in which the badger was trapped and the infection pressure to which this individual badger was exposed. In total,
440 records on subsequent trappings of individual non-infected badgers were available for analysis. Over the
study period, 55 new infections occurred in non-vaccinated (out of 239 = 23.0%) and 40 in vaccinated (out of
201 = 19.9%) badgers. A Generalized Linear Model (GLM) with a cloglog link function was used for analysis.
Statistical analysis showed that susceptibility to natural exposure with M. bovis was reduced in vaccinated
compared to placebo treated badgers: vaccine efficacy for susceptibility, VE
S
, was 59% (95% CI = 6.5%–82%).
However, a complete lack of effect from BCG vaccination on the infectivity of vaccinated badgers was observed,
i.e. vaccine efficacy for infectiousness (VE
I
) was 0%. Further, the basic reproduction ratio as a function of
vaccination coverage (p) (i.e. R(p)) was estimated. Given that the prevalence of M. bovis infection in badgers in
endemic areas in Ireland is approximately 18%, we estimated the reproduction ratio in the unvaccinated po-
pulation as R(0) = 1.22. Because VE
S
was now known, the reproduction ratio for a fully vaccinated population
was estimated as R(1) = 0.50. These results imply that with vaccination coverage in badgers exceeding 30%,
eradication of M. bovis in badgers in Ireland is feasible, provided that the current control measures also remain in
place.
1. Introduction
Bovine tuberculosis (bTB, caused by infection with Mycobacterium
bovis) is a chronic inflammatory disease of bovidae (Bezos et al., 2014).
A control/eradication programme for bTB in cattle started in Ireland in
1959 not only to address the economic losses associated with the in-
fection (Caminiti et al., 2016), but also its zoonotic potential (Langer
and LoBue, 2014). In the first ten years of the control programme, with
a focus on measures to limit cattle to cattle transmission, the incidence
of M. bovis infection in cattle was reduced from 17% to 0.5% (More and
Good, 2006). Subsequently, progress has been slow, despite ongoing
application of intense control strategies, which raised concerns about a
role for one or more reservoirs of M. bovis maintaining transmission.
Over the years, this hypothesis has been confirmed, including work
highlighting high prevalence of infection in badgers (Meles meles)
(Corner et al., 2005). Since then substantial research has been con-
ducted to understand transmission of M. bovis between cattle and
badgers, and of potential strategies capable of reducing this transmis-
sion. One such strategy is the use of BCG (Bacille Calmette-Guérin)
badger vaccination (More and Good 2006).
https://doi.org/10.1016/j.prevetmed.2017.10.010
Received 27 April 2017; Received in revised form 18 October 2017; Accepted 20 October 2017
⁎
Corresponding author at: Centre for Veterinary Epidemiology and Risk Analysis, UCD School of Veterinary Medicine, University College Dublin, Belfield, Dublin 4, Ireland.
E-mail address: inma.aznar@ucd.ie (I. Aznar).
Preventive Veterinary Medicine 149 (2018) 29–37
0167-5877/ © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).
MARK
Fig. 1. Topographic map of the Irish badger vaccine field trial showing: number of farms and bovines, zone area and area of farmed grassland (sq km), and number of main and secondary
badger setts per zone. From north to south, zones A, B and C indicate vaccine (blue badger) and/or placebo (yellow badger) allocation. Estimated M. bovis prevalence in badgers at the end
of the first year is shown per zone (pie charts). Badger setts are represented as: all surveyed setts (grey dots), setts with at least one positive badger trapped in the first year (purple cross),
setts with at least one negative badger trapped in the first year (light green cross). (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)
I. Aznar et al. Preventive Veterinary Medicine 149 (2018) 29–37
30
Experimental challenge trials with M. bovis following BCG vacci-
nation by subcutaneous, mucosal, oral or intramuscular routes (Corner
et al., 2008; Lesellier et al., 2009, 2011; Murphy et al., 2014) have
demonstrated a reduction in disease progression in captive badgers. It
has been proposed that this observed reduction in the number of sites
with gross pathology and of general gross pathological severity scores
observed in these badgers, could translate to a reduction of badger
infectivity, and thus to a reduction in transmission in the field
(Chambers et al., 2011). Here, the expected reduction in transmission
due to a lower infectivity of badgers equates to what is known as
vaccine efficacy for infectivity (VE
I
). In the human field, it is not un-
common to find vaccines that, by helping to reduce pathology and
clinical symptoms in vaccinated and subsequently infected individuals,
achieve a reduction of the infectivity of these individuals and, as a
consequence, a reduction in transmission in the general population.
Vaccines against smallpox, varicella, rubella, measles, hepatitis B and
whooping cough have been recognized as having an important VE
I
which contributes to the overall effect of these vaccines on the popu-
lation (vaccinated and non-vaccinated), this overall effect being re-
ferred to as herd immunity (Fine, 1993; Halloran et al., 1999; Stephens,
2008).
In addition, protection of badgers against M. bovis infection could
also be achieved as a consequence of reduced susceptibility. A reduc-
tion in susceptibility against infection would have both a direct and an
indirect effect in the general population, i.e. vaccinated individuals are
less likely to become infected (direct effect) and therefore, non-infected
badgers are less likely to become infected if surrounded by these less
susceptible individuals (indirect effect). Although this type of protec-
tion was not observed in laboratory trials, a reduction in susceptibility
could potentially be attained under natural conditions because the in-
fective dose that badgers are exposed to in the field is likely to be much
lower than that used in experimental trials (Corner et al., 2008;
Lesellier et al., 2011). This type of protection is referred to as vaccine
efficacy for susceptibility (VE
S
). VE
S
solely refers to the direct effect.
Knowledge of both vaccine efficacies is important as overall trans-
mission depends on both susceptibility and infectivity. However,
methods to quantify transmission after vaccination have only been used
in the last 20 years (Moerman et al., 1993; Stegeman et al., 1995; De
Jong and Kimman, 1994). In 1994, de Jong and Kimman designed an
experimental study that allowed quantification of the transmission
observed in pigs vaccinated against pseudorabies virus. In subsequent
experimental and field transmission studies, the effectiveness of vacci-
nation was evaluated based on estimation of R(p) or the basic re-
production ratio as a function of the proportion of the population that is
vaccinated (Moerman et al., 1993; Stegeman et al., 1995). R(p) is a
crucial parameter to understand the impact of vaccination on popula-
tion dynamics of M. bovis infection. If BCG vaccination is capable of
reducing transmission between badgers, then estimates of the minimum
vaccine coverage necessary to achieve eradication in badgers would be
essential when designing an eradication programme, based on
Diekmann et al. (1990). By examining R(p), the effects of combining
vaccination with other control methods in the same or different species
(e.g. the strategy of detection-and-removal of infected cattle from cattle
herds) can be calculated. This is extremely important in the case of
vaccination in badgers, as the ultimate goal is to help in the control or
eradication of M. bovis infection in cattle.
Aznar et al. (2011) presented a novel design of a badger vaccination
trial and developed a methodology to estimate both VE
S
and VE
I
as well
as R(p) based on incidence data (i.e. new M. bovis infections). The trial
design consisted of three badger populations vaccinated with different
vaccination coverages as suggested by Longini et al. (1998), but taking
into account that these vaccination coverages are achieved over time
rather than instantaneously. Here, we present the results of this badger
vaccine/placebo field trial. M. bovis transmission among badgers was
quantified as well as the effects of vaccination on the susceptibility and
infectivity of badgers. Based on these results, the impact of badger
vaccination on the M. bovis eradication programme in Ireland is re-
viewed.
2. Material and methods
2.1. Trial
The badger vaccine field trial ran from 2009 until 2013. The trial
area of approximately 750 square kilometres was divided into three
zones north to south (A, B and C respectively) (see Fig. 1). Using cages
and stopped wire restraints, a capture–tag–release regime was estab-
lished. Traps were fitted and left in the vicinity of every active sett for
10 days, with daily checks carried out by DAFM (Department of Agri-
culture, Food and the Marine) employees. After the 10 day period, traps
were moved to different setts, taking approximately 23 weeks to cover
the whole trial area (trappings occurring simultaneously in all three
zones). Each 23 week period constituted a “sweep”. A total of 8 sweeps
were carried out over the length of the trial (2 sweeps per year). The
last two sweeps involving some badger removal to allow for post-
mortem evaluations (in the study of Gormley et al. (2017), sweeps 7
and 8 were combined and presented as sweep 7).
At first capture, each badger was tattooed and microchipped, with
blood samples being collected at first capture and every subsequent
recapture (Gormley et al., 2017). Vaccination with an oral BCG vaccine
(Danish strain 1331, at a dose of 1 × 10
8
cfu of BCG administered in
the upper pharyngeal mucosa) suspended on a lipid formulation
(Ancelet et al., 2012; Gormley et al., 2017) was applied randomly to
50% of the badgers trapped in zone B and to all badgers trapped in zone
C. All badgers in zone A and the remaining 50% of the badgers trapped
in zone B received a placebo.
The Enfer chemiluminescent multiplex ELISA system (Whelan et al.,
2008, 2010; Aznar et al., 2014) had been previously optimized to be
used as the diagnostic test in this trial. The Enfer test optimization was
conducted using data obtained from a population of 215 badgers
trapped across 16 counties in Ireland (Murphy et al., 2010; Aznar et al.,
2014). These badgers had been thoroughly examined and a large
number of samples from tuberculous and non-tuberculous lesions were
taken for culture (culture was used as the gold standard). Details about
these badgers and culture methods are presented in Murphy et al.
(2010). A study of factors affecting the statistical power of this design
highlighted the importance of achieving close to 100% specificity in the
diagnostic test used (Aznar et al., 2013). Therefore, the Enfer test was
optimized to maximise sensitivity while retaining specificity at 99.99%
in order to avoid loss of power that would arise from a number of false
positive results randomly occurring in the mainly negative samples
from both vaccinated and unvaccinated animals (Aznar et al., 2014).
Test sensitivity did not play a major role in terms of study power,
however, there was a need for consistent test performance among
samples from all study animals throughout the trial period, including
those vaccinated and not vaccinated. Many steps were taken to achieve
this, including evaluating and comparing test results from vaccinated
and non-vaccinated animals from experimental studies. Differences in
terms of time to seroconversion were observed when the Enfer test was
applied to vaccinated and non-vaccinated captive badger groups. As a
result, a minimum time lag between two subsequent trappings of
215 days for all pairs of trappings was recommended (Aznar et al.,
2014).
The trial was carried out under three licences issued by three dif-
ferent bodies: the Department of Health & Children (research licence,
B100/3187), the Department of Agriculture, Food & the Marine (clin-
ical trial licence, RL/08/06) and the Animal Research Ethics Committee
of University College Dublin (ethics approval, AREC-P-08-26).
2.2. Datasets
Two datasets were collected for analysis. The first consisted of data
I. Aznar et al. Preventive Veterinary Medicine 149 (2018) 29–37
31
collected in the field (using handheld computers) by operators in charge
of capturing and treating badgers in the trial area. This dataset con-
tained information on 2189 badger trappings (from the 1st of
September 2009 to the 12th of July 2013). Data recorded on the
handheld computers prior to the start of the trial and during the op-
erator’s training period were discarded (133 trapping records).
Information recorded at each of the trappings were: badger identifica-
tion (ID) data (badger ID, microchip and tattoo numbers), badger’s sett
ID, date of examination, presence of ectoparasites (ticks, fleas, lice) and
injuries, demographic data (age, sex, weight), type of diagnostic sam-
ples taken (faecal swabs, blood samples, pharyngeal swabs, DNA sam-
ples, others), vaccination data (date of vaccination, vaccine code), op-
erator name, comments, trial zone (A, B or C), and sweep number (1–8).
The second dataset, consisting of 1800 records, contained diagnostic
test information of blood samples taken each time a badger was
trapped. Blood samples were tested using the Enfer multiple antigen
ELISA system for detection of M. bovis antibodies (Enfer Scientific, Co.
Kildare, Ireland). Antibody responses were expressed as relative light
units (RLU) to a panel of 8 antigens: MPB83, MPB70, Rv3616c fragment
and full protein, ESAT-6 and CFP10, as well as purified protein deri-
vative from M. bovis (PPDb) and a peptide of MPB70. The optimization
process is described in detail by Aznar et al. (2014). Blood samples were
analysed twice with the Enfer test: first after the end of each sweep, and
a second time after the vaccine trial had ended. When both sets of re-
sults were compared, low repeatability for two antigens (MPB70 and
Rv3616c fragment) was observed. These two antigens were removed
prior to the final test optimization. The optimization was carried out
using the second set of test results and after removing the two men-
tioned antigens. For that, a stepwise logistic regression with analytical
weights (to optimize specificity versus sensitivity) on the converted
RLU obtained for the six remaining antigens was carried out (Aznar
et al., 2014). By assessing the ROC curve for the model results, a cut off
value equal to −1.95 was selected to achieve 99.99% (exact confidence
interval: 97.34–100%) specificity and 25.33% (exact confidence in-
terval: 20.80–42.24%) sensitivity (exact intervals instead of confidence
intervals were calculated as specificity was very close to 100%). Blood
samples were classified as positive or negative based on this cut off
value.
2.3. Data collation
The two datasets, containing capture and serology data were
merged (1759 trapping records). Data were collated to be analysed as a
Bernoulli experiment. For that purpose, the full dataset was organized
so that each entry contained information regarding two subsequent
trappings of a single badger (vaccinated or non-vaccinated) that tested
negative at the initial trapping. The first entry for an individual badger
was recorded the second time that a specific badger was trapped. A
badger that tested negative at its second trapping could then initiate a
new record in our dataset if trapped for a third time, and so forth. Each
entry line contained information on: infection status of the badger at
the initial and current trapping, current and previous examination date,
sweep number and zone where the badger was trapped each time,
whether the badger had been allocated to vaccine or placebo, and date
of treatment. Once a badger was allocated to either vaccine or placebo,
it remained as such for the rest of the study. Prior to the analysis, three
variables were calculated from the data recorded in the handheld
computers including: delta t (Δt) (i.e.time in days between two sub-
sequent trappings of an individual badger), and prevalence (Prev) and
fraction of infected vaccinated badgers (Fi; the fraction of the total
number of infected badgers that became infected after vaccination) at
the beginning of Δt in the zone where the badger was trapped. A badger
allocated to the vaccine treatment was considered vaccinated the day
after receiving the vaccine. Therefore, as we knew whether the badger
had been allocated to the vaccine or placebo treatment, a new variable
Vaccine status (Vs) was created that coded 0 for badgers allocated to the
placebo treatment (also for badgers allocated to the vaccine treatment
on the first date of treatment) and 1 for vaccinated badgers trapped at
least one day after they received the first vaccination.
2.4. Statistical analysis
The data collation, as well as the descriptive and statistical analyses,
were carried out using Stata
®
(version 14; Stata Corp., College Station,
TX, USA). As part of the descriptive analysis, crude transmission rate
parameters (beta transmission parameters) were calculated as the
number of new cases divided by number of susceptibles and prevalence
in each sweep, for the three zones. In order to help in visualizing pat-
terns, a non-parametric regression of the beta transmission parameters
(lowess smoothing) was conducted.
The purpose of the statistical analysis was to compare new infec-
tions occurring in vaccinated non-infected badgers to those occurring in
non-vaccinated non-infected ones while taking account of both the in-
fection pressure these badgers were exposed to and the trial zone (A, B
or C) badgers had been trapped in (Aznar et al., 2011). Data on 440
pairs of trappings (subsequent trappings of individual badgers) were
used in the statistical analysis. Only badgers that tested negative at the
initial trapping were included. Badgers were coded either 1 or 0, re-
spectively, depending on whether or not they tested positive at the
subsequent trapping.
Assuming “separable mixing”, whereby transmission depends only
on the infectivity of the donor and the susceptibility of the receptor
(Diekmann et al., 1990), the expected infection status of any uninfected
re-trapped badger (vaccinated or non-vaccinated) was modelled in the
total population using a generalized linear model (GLM). With this
model we aimed to explain new infections from three explanatory
variables: a) the vaccination status of the badger, b) the fraction of
infected vaccinated badgers, and c) the zone where the badger was
trapped. Details of the statistical model are elaborated below. If vac-
cination is effective, then we would expect infectivity to vary both
between the three zones and also over time due to differences in the
fraction of infected badgers that were vaccinated. It is important to note
that the percentage of vaccinated badgers increased over the duration
of the trial in zones B and C (Fig. 2).
The expected number of cases per unit of time C
E
(
)
can be for-
mulated as =⋅ −−⋅ ⋅
C
E
() S(1 e
)
βΔtPrev
t) where S is the number of suscep-
tible badgers and −−
(
)
e1**
βI t
N
Δis the probability that any of the sus-
ceptible badgers becomes infected (supplementary material, Section 1).
Then the complementary log–log (cloglog) link function results in an
estimate for log (β) taking ln(Prev*Δt) as offset (Aznar et al., 2011). This
model was run separately for vaccinated and non-vaccinated badgers,
allowing separation of the effects of vaccination in susceptibility and in
infectivity as explained in derivations presented in Section 1 of the
supplementary material. By separating these two effects, estimations of
VE
S
and VE
I
are possible. The model used was:
=+ + + + +cloglog E C β β Z β Z β Vs β Fi of fse
t
() BBCC
01, 1, 2 3
where Zcodes for zone (binary dummy variable 0/1 for each of the
zones, zone A being the reference), vs is the vaccination status of the
recipient badger, Fi is the fraction of vaccinated badgers among the
infected badgers at the beginning of the time interval in that same zone,
and β
0
,β
1,B
,β
1,C
,β
2
,β
3
, are the regression coefficients as estimated by
our model. For modelling purposes, once a badger tested positive to the
serological test, it was considered positive for the rest of the study and
therefore subsequent trappings of this badger were not included in the
analysis. As the number of predictors in the maximum model was small,
all possible combinations of predictors were examined (including in-
teraction terms). The final model was selected based on the lowest
value for the Akaike Information Criterion (AIC).
From this model, four transmission parameters: β
vv
,β
vu
,β
uv
and β
uu
were estimated. The first sub-index in these transmission parameters
I. Aznar et al. Preventive Veterinary Medicine 149 (2018) 29–37
32
indicates the vaccination status of the badger transmitting M. bovis
(whether it is from a vaccinated (v) or non-vaccinated badger (u)),
while the second sub-index refers to the vaccination status of the re-
cipient badger. The two vaccine efficacies and R(p) can then be cal-
culated from these four transmission parameters (see Section 1 of the
supplementary material). Using the regression coefficients from our
model, the transmission rate parameters, ignoring zone effects, can then
be estimated as:
β
uv
=e
β
0
+β
2
,β
vv
=e
β
0
+β
2
+β
3
,β
uu
=e
β
0
and β
vu
=e
β
0
+β
3
Vaccine efficacies were calculated as:
=− =− =− =− =−
=−
VE β
β
β
βeVE β
β
β
β
e
111and11
1,
suv
uu
vv
vu
βIvu
uu
vv
uv
β
2
3
noting that coefficient β
2
calculated for the variable (Vs) contributes to
the estimation of VE
s
, and the coefficient β
3
calculated for the variable
(Fi) contributes to the estimation of VE
I
, thus being able to estimate
both vaccine efficacies. The reproduction ratio as a function of the
proportion (p) of badgers vaccinated R(p) was determined as:
=− +
R
ppRpR( ) (1 )· (0) · (1)
Where =−
R
(0) 1
1prevalence
and R(1) = (1 −VE
S
)∙(1 −VE
I
)∙R(0)
3. Results
3.1. Vaccine field trial descriptive analysis
Overall, 1093 badgers were trapped over the 8 sweeps, with 435
badgers trapped in zone A, and 243 and 415 in zones B and C, re-
spectively. In total, 673 badgers were trapped once, 253 twice, 111
three times, 38 four times, 13 five times and 5 six times. An initial
concern over the vaccine trial design was the fact that no major physical
boundaries existed between the three zones. A large number of badger
movements across the three zones could have hampered the vaccination
gradient between the zones and therefore reduced the power of the
analysis. Such a large movement was not expected, nonetheless we can
confirm that it did not occur as only in 2% (22) of the subsequent
trapping events had badgers originally trapped in one zone been
trapped in a different zone at a later stage.
The prevalence of M. bovis infection, estimated as the overall per-
centage of positive trappings to the Enfer test at each sweep, ranged
between 12.5% and 37.8% (see Table S1 in Section 2 of the supple-
mentary material). At the beginning of the trial, the zone prevalence
(the percentage of positive trappings in each zone in sweep 1) was
higher, but not statistically different, in zone A (31.7%) compared to
zones B (19.0%) and C (23.5%) (p-value = 0.14). During the first year
of the trial (that is, considering sweeps 1 and 2 together to avoid the
effect of seasonality on trapping efforts), the prevalence was also
highest in zone A (26.9%) compared to zones B and C (20% and 25.2%,
respectively) but again, these differences were not statistically sig-
nificant (p-value = 0.49). Due to a procedural error, blood results for
70 samples taken from badgers during sweep 2 were not available (see
Section 4 of the supplementary material). The incidence of M. bovis
infection per sweep, defined as the number of newly infected badgers
(captured badgers that tested positive for the first time in sweep n)
divided by the number of susceptible badgers (badgers trapped in
sweep nthat had never tested positive or tested positive for the first
time in that sweep), varied over time and across zones, with the lowest
incidence being in sweep 5 in zone C (see Fig. S1 in Section 2 of the
supplementary material). In zones B and C, the proportion of BCG
vaccinated badgers increased from sweep 3–6, then decreased in
sweeps 7 and 8 (as the last sweeps involved badger removal) (Gormley
et al., 2017). At sweep 6, the proportion of vaccinated badgers in zones
B and C were 37.3% and 62.2% respectively (Fig. 2).
Crude transmission rate parameters in each sweep, for the three
zones, and a lowess smoothing of the transmission parameters are
presented in Fig. 3. During the trial, there was a non-significant de-
crease in these crude transmission rate parameters in zones B and C.
The possible change over time in crude transmission rate was less clear
for zone A. However, the overall initial transmission (at sweep 3) in this
zone (i.e. even before vaccination could have had an impact) was al-
ready lower compared to the other two zones (Fig. 3).
3.2. Statistical analysis
The dataset consisted of 440 records (239 originated from non-
0 20 40 60 %0 20 40 60%
0 20 40 60 %
050 100 150
050 100 150
050 100 150
1 2 3 4 5 6 7 8
A
B
C
Fig. 2. Total number of badgers caught at each sweep (left vertical axis) and percentage
of captured badgers that were vaccinated (right vertical axis), including polynomial
(n = 4) smoothing of this percentage per sweep and zone (right vertical axis).
I. Aznar et al. Preventive Veterinary Medicine 149 (2018) 29–37
33
vaccinated badgers and 201 from vaccinated badgers). A total of 55
(23.0%) and 40 (19.9%) new infections occurred in non-vaccinated and
vaccinated badgers, respectively. Vaccination status of the badger re-
ceiving the vaccine was the only statistically significant explanatory
variable in the model. Nonetheless, all variables (except the interaction
terms) were kept in the final model as that was the model with the
lowest AIC (490.2). Using the coefficient obtained for recipient vacci-
nation status, we calculated vaccine efficacy for susceptibility, VE
S
,as
59% (95% CI = 6.5–82%); that is, a 59% reduction in susceptibility of
vaccinated compared to unvaccinated badgers was achieved (see
Table 1), but no significant effect of vaccination on infectivity was
observed in this trial (hence VE
I
= 0%).
In addition to the main model, two more statistical analyses were
conducted. As Zone B resembles the classic 50:50 vaccine-placebo trial
design (but with a change in vaccination coverage over time and the
availability of longitudinal data on infection), estimation of the direct
effect of vaccination on susceptibility was possible in this zone only.
The model showed a similar outcome for VE
S
(54%, 95%
CI = 0.0–79.9%) (see Table S2 in Section 3 of the supplementary ma-
terial). A lower initial crude beta transmission parameter was observed
in zone A compared to zones B and C (Fig. 3), for reasons that are not
clear. Due to this lower initial transmission parameter observed in zone
A, and as our design only required a minimum of two populations
vaccinated at different vaccination coverages, the model was run again
using data from zones B and C only. Similar results were obtained in
terms of both the effect of vaccination on susceptibility and infectivity,
with VE
S
= 60% (95% CI = 8.8-83.0%) and no significant effect of
vaccination on infectivity (VE
I
= 0%) (see Table S3 in section 3 of the
supplementary material).
We finalized our analyses by estimating R(p) for a range of vaccine
coverages, as it is the impact of the combination of both vaccine effi-
cacies that determines the feasibility of using vaccination as a strategy
to achieve M. bovis eradication in badgers. The average M. bovis pre-
valence in badgers in Ireland declined between 2007–2013, with an
average prevalence from May 2007 to May 2011 equal to 17.7%, and
from June 2011 to April 2013 equal to 9.9% (Byrne et al., 2015). For
any population where an infection is at the endemic steady state, the
fraction of susceptible individuals equals 1/R. Thus, for a badger pre-
valence equal to 18%, we can calculate
==
−
R
(0) 1.22
,
1
1prevalence
and
for a prevalence equal to 10%, =
R
(0) 1.1
1
.AsVE
S
= 59% and using
the higher prevalence estimate (18%), the reproduction ratio for a fully
vaccinated population can be calculated as R(1)=(1−VE
S
)*R(0)
=0.50. These results indicate that adding vaccination to the current
control strategies in Ireland, eradication of M. bovis infection in badgers
can be achieved with any vaccination coverage above 30% (Fig. 4).
0 0.002 0.004 0.006
3 4 5 6 7 8 3 4 5 6 7 8 3 4 5 6 7 8
A B C Fig. 3. Crude transmission rate parameters (beta, in
blue) and lowess smoothing (in red) per sweep esti-
mated separately for the three trial zones. (For inter-
pretation of the references to colour in this figure le-
gend, the reader is referred to the web version of this
article.)
Table 1
Results of the final generalized linear model including the estimated regression coeffi-
cient, p-value and 95% confidence interval for all explanatory variables (fraction of in-
fected vaccinated badgers, vaccination status and zone) and constant in a model fitted in
data from all three zones of the Irish badger vaccine field trial. Only Vs, the vaccination
status of the recipient, is significant, with the other variables retained to control for
confounding.
Variable Coef p-Value (95% CI)
Constant −6.07 < 0.001 −6.38 to −5.77
Zone
A Reference
B 0.55 0.083 −0.07 to 1.17
C 0.63 0.193 -0.32 to 1.58
Vs −0.90 0.034 −1.73 to −0.07
Fi 1.37 0.119 −0.35 to 3.10
0
0.2
0.4
0.6
0.8
1
1.2
0 102030405060708090
Fig. 4. Basic reproduction ratio for badger to badger transmission as function of vacci-
nation coverage, given R(0) = 1.22 and a VE
S
= 59%.
I. Aznar et al. Preventive Veterinary Medicine 149 (2018) 29–37
34
4. Discussion
In this manuscript, the effect of BCG vaccination on M. bovis
transmission between badgers in the field has been quantified for the
first time. Here, separate estimates on the effects of vaccination on both
protection against infection and on the infectivity of badgers that be-
come infected subsequent to vaccination are presented. The vaccine
efficacy estimates presented in this paper contribute to a better un-
derstanding of the biological processes underpinning the protection
against transmission achieved by BCG vaccination in the field. While no
direct protection against infection following vaccination was reported
in experimental trials (where vaccinated badgers were challenged with
different doses and different strains of M. bovis)(Corner et al., 2008,
2010), we observed a 59% protection against infection of vaccinated
badgers in the field. The difference between our findings and those
obtained in laboratory trials is not surprising, as the route of infection,
infection dose, number of infection events to achieve this dose, etc
occurring in the wild are unknown. Nonetheless, we cannot confirm
whether or not the observed protection against infection is due to a
lower infection dose in the wild compared to experimental trials
(Corner et al., 2008; Lesellier et al., 2011). A reduction in the total
infectivity of vaccinated and subsequently infected badgers in the field
had been anticipated based on the reduction in disease progression
observed in vaccinated compared to non-vaccinated badgers in ex-
perimental studies (Chambers et al., 2011). However, no reduction of
infectivity was found in our study. The lack of effect of BCG vaccination
on infectivity in the general badger population is thus at odds with the
hypothesis that vaccination, by reducing disease progression, reduces
the infectivity of vaccinated and subsequently infected badgers. From
this study, we cannot determine whether a similar reduction in disease
progression to that observed in experimental studies was found in the
field as no post-mortem data were available. Nevertheless, if that re-
duction in disease progression does exist, we did not find a concurrent
reduction in infectivity. The lack of effect of vaccination on infectivity
has implications in terms of the effectiveness of BCG badger vaccination
in Ireland (or how much reduction of transmission is achieved by
vaccination). The effectiveness of a vaccination programme is the result
of both the effect of vaccination on susceptibility and infectivity. Here,
as there is no added reduction in transmission due to a reduction in
infectivity (one type of indirect effect of vaccination), the total reduc-
tion in transmission or effectiveness achieved by vaccination is equal to
VE
S
.
Once the effectiveness of BCG vaccination was calculated, in order
to assess its impact, it was necessary to estimate the ongoing badger to
badger transmission. The reproduction ratio for badger to badger
transmission under the current control options in Ireland was calculated
as 1.22 assuming a badger prevalence of 18%. Based on surveillance
data collected from badgers culled as part of an interim badger culling
regime in Ireland during 2007–2013, an average national prevalence of
14.1% was estimated (Byrne et al., 2015). However, this includes two
partial prevalence estimates (17.7% for May 2007 to May 2011, and
9.9% for June 2011 to April 2013), noting that differing methods were
used during these periods to differentiate M. bovis from non-tuberculous
mycobacteria (biochemical tests to May 2011, and PCR techniques
subsequently). In this paper, we used 18% as a conservative prevalence
estimate. The formulae used for calculating
R
(0)
is a basic formulae
used to assess transmission in badgers assuming that there is no
transmission between cattle and badgers. Although this is likely not the
case, we can use this number as an approximation, and conclude that if
we were to vaccinate all badgers in Ireland, we would be able to reduce
transmission by 59%, with the resulting
=
R
(1) 0.5
, which is sub-
stantially below 1 indicating that eradication in badgers would be
feasible. Further and by estimating
R
(p)
or the reproduction ratio for a
range of vaccine coverages (p), we were able to assess what was the
minimum vaccination coverage necessary to eradicate. The most re-
levant finding in this manuscript was that in Ireland, vaccination of
badgers with a vaccination coverage equal to or higher than 30% is
sufficient to eradicate M. bovis infection in badgers, as long as current
control strategies also remain in place in both cattle and badgers. The
outcomes of this study will have major implications for the control of
M. bovis infection in Ireland, not only in badgers but also in cattle. It is
important to note that if any or some of the control strategies currently
in place have an effect on badger to badger transmission, then mod-
ifications to any strategy would have repercussions on the effectiveness
of the badger vaccination programme (as the reproduction ratio for
badger to badger transmission would change also). For similar reasons,
it is not possible to predict the effectiveness of BCG vaccination in
badgers in a different country, with different transmission character-
istics between badgers.
In this study the infection status of individual badgers was de-
termined by whether or not these badgers tested positive to the Enfer
test. Table S1 shows the prevalence of infection as measured by this
test. Prevalence values varied between sweeps and zones with pre-
valence in sweep 2 in zone 1 being much lower (12.5%) than that ob-
served in the same zone in sweep 1 (31.7%). The second lowest pre-
valence observed in the whole study was in zone C in sweep 5 (16.3%)
with the rest being between 20.0–37.8%. We are not aware of any
specific reasons why these prevalence values changed and we assume
that these differences are due to randomness. Tuberculosis is a chronic
disease with latent and reactivation periods and with serology varying
through the different disease stages. If the badger population in the trial
differed in terms of disease profile from the 215 badgers in which the
test was optimized (i.e. a larger proportion of badgers in a chronic
phase in the field trial), the sensitivity of the Enfer test in the trial could
be higher or lower than the 25.3% achieved during test optimization.
Indeed, prevalence estimates very much depend on the representa-
tiveness of the gold standard panel for the population tested as there is
not yet a gold standard test for M. bovis infection in badgers. In a pre-
vious study where factors affecting study power were explored, it was
shown how high test specificity was paramount (Aznar et al., 2013).
Test sensitivity did not play an important role in our ability to detect an
effect if BCG vaccination really worked. The fact that we found an effect
(VE
S
= 59%) suggests that both sensitivity and specificity were suffi-
ciently large and did not affect the study power. The low sensitivity of
the test used will also have an effect on incidence and prevalence values
and therefore on the beta transmission parameter. We note, however,
that the aim of this paper was not to provide true values for these
parameters, but rather to use them to estimate VE(s) by comparing
them in the vaccinated and non-vaccinated groups.
Badger capture data from this vaccine trial has been previously
analysed (Gormley et al., 2017). Two vaccine efficacies were reported
from this earlier analysis, one for badgers enrolled during sweeps 1 and
2 (VE = 36%) and other for badgers enrolled during sweeps 3–6
(VE = 84%). In that study, the direct effect of vaccination was esti-
mated by comparing hazard rates of badgers trapped in zone A (0%
vaccination coverage) to that of badgers trapped in zone C (100%
vaccination coverage). In addition to the different serological tests used
in both studies (incidence in badgers was measured with the BrockTB
Stat-Pak lateral flow serology test, (Chambers et al., 2008)), the
methodology in which badgers were enrolled for the analysis and the
statistical methods used to compute VE estimates were also different.
Data from zone B were not used in the prior analysis, despite badgers
from this zone being the ideal population for measuring the direct effect
(as vaccinated and non-vaccinated badgers would have been exposed to
the same infection pressure).
In a badger vaccine field trial carried out in the UK (Carter et al.,
2012), badger setts (rather than individual badgers) were allocated to
either vaccine or placebo. From that field study, estimates of the direct
effect of BCG vaccination on susceptibility in badgers have been re-
ported with two estimates depending on the diagnostic tests used:
VE
S
= 54% (95% CI = 12-74%) for the more sensitive test (described
as “triple test
V
”) and VE
S
= 76% (95% CI = 48-89% for the less
I. Aznar et al. Preventive Veterinary Medicine 149 (2018) 29–37
35
sensitive test (“dual test”). Nonetheless due to the study design of
choice, separation of the effects of vaccination in susceptibility and
infectivity was not possible in either Gormley et al. (2017) or Carter
et al. (2012), leading to two biases in the estimate of VE
S
. Firstly, the
indirect effect of BCG vaccination is included in the estimate of VE
S
(although this estimate should only reflect the direct effect of vacci-
nation), and secondly, if the infectivity of vaccinated and non-vacci-
nated badgers differs, then this difference in infectivity has to be taken
into account also when estimating VE
S
. In hindsight, and based on our
results, we now know that such a difference does not exist (VE
I
=0as
the coefficient for Fi was not statistically significant). Nonetheless, it is
important that this issue is highlighted so it can be considered in the
design of future vaccine field trials.
A reduction in M. bovis incidence in cubs from vaccinated setts
compared to those from non-vaccinated setts was also observed in
Carter et al. (2012). In that study, the observed reduction in incidence
in cubs is reported as an indirect effect of vaccination. However, it is
not possible to distinguish whether this reduction in incidence is due to
the indirect effect achieved by a reduction in susceptibility or to a re-
duction in the infectivity of vaccinated infected badgers compared to
non-vaccinated infected ones. Based on our results, the reduction in
incidence among cubs was likely due to a reduction in susceptibility of
the vaccinated adult badgers in the sett.
5. Conclusion
In summary, we have presented a new methodology to estimate
both VE
S
and VE
I
providing further knowledge on the biological ways in
which BCG vaccination works in badgers. We have also presented sci-
entific arguments that support the crucial role of appropriate trial de-
sign in order to obtain accurate estimates. Further, we have estimated
the impact of vaccination in the current badger transmission Ireland
and concluded that a minimum vaccination coverage of 30% is neces-
sary to achieve eradication of M. bovis infection in badgers. As a result
of this work, policy makers can now make informed decisions con-
cerning the best strategy or combination of strategies to achieve era-
dication. These results could also be used to guide the best vaccination
route to achieve the minimum vaccine coverage needed.
Conflict of interest
No conflict of interest to declare
Acknowledgments
This work was funded by the Department of Agriculture, Food and
the Marine (DAFM). We thank Enfer Scientific for undertaking the as-
says used in this study and Jamie A. Tratalos for his work on the re-
conciliation of the various project databases. We also thank all field
staffinvolved in the trial, Wayne Martin for reviewing this manuscript
prior to submission, and Eamonn Gormley and his team for preparing
blood samples for analysis by Enfer.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at http://dx.doi.org/10.1016/j.prevetmed.2017.10.010.
References
Ancelet, L.R., Aldwell, F.E., Rich, F.J., Kirman, J.R., 2012. Oral vaccination with lipid-
formulated BCG induces a long-lived, multifunctional CD4(+) T cell memory im-
mune response. PLoS One 7, e45888.
Aznar, I., McGrath, G., Murphy, D., Corner, L.A.L., Gormley, E., Frankena, K., More, S.J.,
Martin, W., O’Keeffe, J., de Jong, M.C.M., 2011. Trial design to estimate the effect of
vaccination on tuberculosis incidence in badgers. Vet. Microbiol. 151, 104–111.
Aznar, I., More, S.J., Frankena, K., de Jong, M.C.M., 2013. Estimating the power of a
Mycobacterium bovis vaccine trial in Irish badgers. Prev. Vet. Med. 111, 297–303.
Aznar, I., Frankena, K., More, S.J., Whelan, C., Martin, W., Gormley, E., Corner, L.A.L.,
Murphy, D., de Jong, M.C.M., 2014. Optimising and evaluating the characteristics of
a multiple antigen ELISA for detection of Mycobacterium bovis infection in a badger
vaccine field trial. PLoS One 9, e100139.
Bezos, J., Álvarez, J., Romero, B., de Juan, L., Domínguez, L., 2014. Bovine tuberculosis:
historical perspective. Res. Vet. Sci. 97, S3–S4. http://dx.doi.org/10.1016/j.rvsc.
2014.09.003. ISSN 0034-5288.
Byrne, A.W., Kenny, K., Fogarty, U., O’Keeffe, J.J., More, S.J., McGrath, G., Teeling, M.,
Martin, S.W., Dohoo, I.R., 2015. Spatial and temporal analysis of metrics of tu-
berculosis infection in badgers (Meles meles) from the Republic of Ireland: trends in
apparent prevalence. Prev. Vet. Med. 122, 345–354.
Caminiti, A., Pelone, F., LaTorre, G., De Giusti, M., Saulle, R., Mannocci, A., Sala, M.,
Della Marta, U., Scaramozzino, P., 2016. Control and eradication of tuberculosis in
cattle: a systematic review of economic evidence. Vet. Rec. 179, 70–75.
Carter, S.P., Chambers, M.A., Rushton, S.P., Shirley, M.D.F., Schuchert, P., Pietravalle, S.,
Murray, A., Rogers, F., Gettinby, G., Smith, G.C., Delahay, R.J., Hewinson, R.G.,
McDonald, R.A., 2012. BCG vaccination reduces risk of tuberculosis infection in
vaccinated badgers and unvaccinated badger cubs. PLoS One 7, e49833.
Chambers, M.A., Crawshaw, T., Waterhouse, S., Delahay, R., Hewinson, R.G.,
Lyashchenko, K.P., 2008. Validation of the BrockTB Stat-Pak assay for detection of
tuberculosis in Eurasian badgers (Meles meles) and influence of disease severity on
diagnostic accuracy. J. Clin. Microbiol. 46, 1498–1500.
Chambers, M.A., Rogers, F., Delahay, R.J., Lesellier, S., Ashford, R., Dalley, D., Gowtage,
S., Davé, D., Palmer, S., Brewer, J., Crawshaw, T., Clifton-Hadley, R., Carter, S.,
Cheeseman, C., Hanks, C., Murray, A., Palphramand, K., Pietravalle, S., Smith, G.C.,
Tomlinson, A., Walker, N.J., Wilson, G.J., Corner, L.A., Rushton, S.P., Shirley, M.D.,
Gettinby, G., McDonald, R.A., Hewinson, R.G., 2011. Bacillus Calmette-Guerin vac-
cination reduces the severity and progression of tuberculosis in badgers. Proc. R. Soc.
B 278, 1913–1920.
Corner, L.A.L., O’Meara, D., Costello, E., Gormley, E., 2005. Tuberculosis in badgers: true
prevalence, diagnostic methods and epidemiology (poster). Society for Veterinary
Epidemiology and Preventive Medicine. . https://fr.slideserve.com/theodore-estes/
tuberculosis-in-badgers-true-prevalence-diagnostic-methods-and-epidemiology
Nairn, Scotland.
Corner, L.A.L., Costello, E., Lesellier, S., O’Meara, D., Gormley, E., 2008. Vaccination of
European badgers (Meles meles) with BCG by the subcutaneous and mucosal routes
induces protective immunity against endobronchial challenge with Mycobacterium
bovis. Tuberculosis 88, 601–609.
Corner, L.A.L., Costello, E., O’Meara, D., Lesellier, S., Aldwell, F.E., 2010. Oral vaccina-
tion of badgers (Meles meles) with BCG and protective immunity against en-
dobronchial challenge with Mycobacterium bovis. Vaccine 28, 6265–6272.
De Jong, M.C.M., Kimman, T.G., 1994. Experimental quantification of vaccine-induced
reduction in virus transmission. Vaccine 12, 761–766.
Diekmann, O., Heesterbeek, J.A.P., Metz, J.A.J., 1990. On the definition and the com-
putation of the basic reproduction ratio R
0
in models for infectious diseases in het-
erogeneous populations. J. Math. Biol. 28, 365–382.
Fine, P.E.M., 1993. Herd immunity: history, theory, practice. Epidemiol. Rev. 15,
265–302.
Gormley, E., Nı´ Bhuachalla, D., O’Keeffe, J., Murphy, D., Aldwell, F.E., Fitzsimons, T.,
Stanley, P., Tratalos, J.A., McGrath, G., Fogarty, N., Kenny, K., More, S.J., Messam,
L.L., Corner, L.A.L., 2017. Oral vaccination of free-living badgers (Meles meles) with
Bacille Calmette Guérin (BCG) vaccine confers protection against tuberculosis. PLoS
One 12, e0168851.
Halloran Jr., M.E., Longini, I.M., Struchiner, C.J., 1999. Design and interpretation of
vaccine field studies. Epidemiol. Rev. 21, 73–88.
Langer, A.J., LoBue, P.A., 2014. Public health significance of zoonotic tuberculosis caused
by the Mycobacterium tuberculosis complex. In: Thoen, C.O., Steele, J.H., Kaneene,
J.B. (Eds.), Zoonotic Tuberculosis: Mycobacterium Bovis and Other Pathogenic
Mycobacteria, third edition. John Wiley & Sons, Inc., Chichester, UK.
Lesellier, S., Corner, L.A.L., Costello, E., Lyashchenko, K., Greenwald, R., Esfandiari, J.,
Singhd, M., Hewinsone, R.G., Chambers, M., Gormley, E., 2009. Immunological re-
sponses and protective immunity in BCG vaccinated badgers following endobronchial
infection with Mycobacterium bovis. Vaccine 27, 402–409.
Lesellier, S., Palmer, S., Gowtage-Sequiera, S., Ashford, R., Dalley, D., Davé, D., Weyer,
U., Salguero, F.J., Nunez, A., Crawshaw, T., Corner, L.A.L., Hewinson, R.G.,
Chambers, M.A., 2011. Protection of Eurasian badger (Meles meles) from tuberculosis
after intra-muscular vaccination with different doses of BCG. Vaccine 29, 3782–3790.
Longini Jr., I.M., Sagatelian, K., Rida, W.N., Halloran, M.E., 1998. Optimal vaccine trial
design when estimating vaccine efficacy for susceptibility and infectiousness from
multiple populations. Stat. Med. 17, 1121–1136.
Moerman, A., Straver, P.J., de Jong, M.C., Quak, J., Baanvinger, T., van Oirschot, J.T.,
1993. A long-term epidemiologic study of bovine viral diarrhea infections in a large
herd of dairy cattle. Vet. Rec. 132, 622–626.
More, S.J., Good, M., 2006. The tuberculosis eradication programme in Ireland: a review
of scientific and policy advances since 1988. Vet. Microbiol. 112, 239–251.
Murphy, D., Gormley, E., Costello, E., O’Meara, D., Corner, L.A.L., 2010. The prevalence
and distribution of Mycobacterium bovis infection in European badgers (Meles meles)as
determined by enhanced post mortem examination and bacteriological culture. Res.
Vet. Sci. 88, 1–5.
Murphy, D., Costello, E., Aldwell, F.E., Lesellier, S., Chambers, M.A., Fitzsimons, T.,
Corner, L.A.L., Gormley, E., 2014. Oral vaccination of badgers (Meles meles) against
tuberculosis: comparison of the protection generated by BCG vaccine strains Pasteur
and Danish. Vet. J. 200, 362–367.
Stegeman, A., van Nes, A., de Jong, M.C.M., Bolder, F., 1995. Assessment of the effec-
tiveness of vaccination against pseudorabies in finishing pigs. Am. J. Vet. Res. 56,
I. Aznar et al. Preventive Veterinary Medicine 149 (2018) 29–37
36
573–578.
Stephens, D.S., 2008. Vaccines for the unvaccinated: protecting the herd [Editorial
commentary]. J. Infect. Dis. 197, 643–645.
Whelan, C., Shuralev, E., O’Keeffe, G., Hyland, P., Kwok, H.F., Snoddy, P., O’Brien, A.,
Connolly, M., Quinn, P., Groll, M., Watterson, T., Call, S., Kenny, K., Duignan, A.,
Hamilton, M.J., Buddle, B.M., Johnston, J.A., Davis, W.C., Olwill, S.A., Clarke, J.,
2008. Multiplex immunoassay for serological diagnosis of Mycobacterium bovis in-
fection in cattle. Clin. Vaccine Immunol. 15, 1834–1838.
Whelan, C., Whelan, A.O., Shuralev, E., Kwok, H.F., Hewinson, G., Clarke, J.,
Vordermeier, H.M., 2010. Performance of the Enferplex TB assay with cattle in Great
Britain and assessment of its suitability as a test to distinguish infected and vacci-
nated animals. Clin. Vaccine Immunol. 17, 813–817.
I. Aznar et al. Preventive Veterinary Medicine 149 (2018) 29–37
37
