46 A.J.K. Conlan and J.L.N. Wood
Taken together, these inconsistencies sug-
gest that the period of epidemiological latency is
poorly identifiable from testing data. Models can
fit the patterns of transmission within, and likely
between, herds equally well with very short or
long average periods of latency by trading off
against other unknown parameters such as the
average rate of transmission and test character-
istics. Extreme estimates of latency, longer than
the average lifetime of an animal within this
population, may seem unrealistic at first sight,
but in fact point to inadequacies in our concep-
tual model of infection (Fig. 4.1). The standard
compartmental model we have outlined assumes
all animals progress through the compartments
of the model at fixed (average) rates. With no
heterogeneity between individuals, estimating a
low rate of progression to infectiousness is the
only mechanism through which the model can
capture the variability observed between out-
breaks. O’Hare et al. (2014) found evidence that
the inclusion of such heterogeneity in the form
of a few ‘super-spreading’ cattle can improve the
fit of herd level models, but this could also be
generated by seasonal differences in contact
rates (Bekara et al., 2014) or by a difference in
individuals’ rates of progression as previously
discussed for models of human tuberculosis
infection (Blower et al., 1995).
The final compartment of our conceptual
model, accounts for severely infected animals
with extensive lesions which nonetheless exhibit
no reaction to tuberculin (Francis, 1947;
Monaghan et al., 1994). Such ‘anergic’ animals
(A) were thought to be relatively rare even in
endemically infected herds, and have thus been
considered to be unlikely to play a significant
role in the transmission of bovine TB in regu-
larly tested herds (Barlow et al., 1997). This
assumption is supported by the relatively small
numbers of animals presenting at slaughter-
houses with lesions from regularly tested herds
(Frankena et al., 2007; Olea-Popelka et al., 2012;
Shittu et al., 2013) – and are more the inevitable
consequence of gaps in surveillance, not least
with an imperfect test (Mitchell et al., 2006),
rather than necessarily a failure of testing. How-
ever, in uncontrolled populations where bovine
tuberculosis is endemic or emerging, anergic
animals may well be masking an even higher
burden of infection within herds (Thakur et al.,
2010; Firdessa et al., 2012).
4.1.2 The hidden burden of infectionThe structure of epidemic models for bovine TB
reflects the importance of diagnostic test perfor-
mance in both controlling and estimating the
rate of transmission. Tuberculin testing remains
the imperfect gold standard for quantifying the
prevalence of infection and demonstrating free-
dom from infection for herds and animals
(Monaghan et al., 1994; de la Rua-Domenech
et al., 2006; Nuñez-Garcia et al., 2017). The sen-
sitivity and specificity of tuberculin testing are
well known to vary with respect to the format of
the test (e.g. single intradermal, comparative),
potency of tuberculin (Downs et al., 2013) and
even the compliance of veterinarians to protocol
(Humblet et al., 2011). Variability in compliance
is in some sense understandable given the con-
siderable health and safety challenges, for both
veterinarians and farmers, of administering the
test. Conscious and unconscious decisions con-
cerning the time spent per animal and the han-
dling of borderline reactions will also be
influenced by the epidemiological context of the
herd being tested. This veterinary discretion,
although a complicating factor for the interpre-
tation of epidemiological data, has been argued
to be a strength of the statutory system allowing
the impacts of disease, if not the disease itself,
to be more appropriately managed (Enticott,
2012).
Less appreciated, however, is the dynamic
relationship between transmission, test sensitiv-
ity and the hidden burden of infection missed by
testing. Transmission models for bovine tuber-
culosis are structured to account for the system-
atic reduction in test sensitivity for early and late
infections. Within our conceptual framework,
‘true’ reactor animals that are actually infected
are detected from the R and I compartments
(Fig. 4.1) with an efficiency determined by the
diagnostic test sensitivity. Exposure to other
environmental sources of Mycobacterium is
likely to be an important contributory factor to
false positive reactions to the skin test. To
account for this, models allow false positive ani-
mals to be detected from any compartment with
a risk determined by the test specificity parame-
ter. However, post-mortem inspection and
culture, which can itself be considered as an
insensitive but perfectly specific diagnostic test,
can in principle identify animals at any stage of