The Epidemiology of Mycobacterium bovis Infection in Cattle 45
4.1.1 A conceptual model of bovine
tuberculosis progressionAfter infection, susceptible animals (S) enter an
occult (O), or unreactive, period where they are
infected, but do not react in diagnostic tests.
Early experimental and field studies established
an occult period between inoculation and cattle
reacting to tuberculin. Francis (1947) con-
cluded that this occult period had a duration of
20 to 30 days but could range between 8 and 50
days. These early insights are consistent with
more modern data suggesting that artificially
challenged animals are detectable within 3
weeks of infection (Thom et al., 2006).
The duration of latency in infectious dis-
eases, including murine M. tuberculosis infec-
tion, can, at least theoretically, vary with
infectious dose (Meynell and Meynell, 1958).
This raises the possibility that experimental esti-
mates may underestimate the occult period due
to the (comparatively) large doses used in chal-
lenge studies. However, animals given as low a
dose as one colony forming unit have also been
demonstrated to become test positive within the
same 3-week timescale (Dean et al., 2005). Epi-
demiological data from Great Britain provides
further evidence that the occult period must be
relatively short. Although latency periods in
general are poorly identified by transmission
models (Conlan et al., 2012, 2015; Bekara et al.,
2014; Brooks-Pollock et al., 2014; O’Hare et al.,
2014), occult periods greatly in excess of the
experimentally derived range of 8 to 50 days
would generate unrealistically high rates of
recurrence in low incidence areas based on esti-
mates of cattle-to-cattle transmission rates
(Conlan et al., 2012).
There is marked confusion in the literature
in the use of the term ‘latency’, which is assumed
to mean clinical latency for much of the human
literature, but usually refers to epidemiological
latency for most studies of the cattle disease.
Despite over a century of study, the rela-
tionship between infection, diagnostic status,
shedding of live bacteria and infectiousness of
animals are still poorly characterized. Shedding
of bacteria by naturally infected animals is inter-
mittent and unpredictable, which along with
culture times of up to 3 months devalues efforts
to isolate M. bovis from live animals in the field.
Experimentally challenged animals have been
shown to shed bacteria within the first 10 to
60 days after intranasal challenge (Neill et al.,
1988, 1989; Kao et al., 2007). Subsequently,
shedding is intermittent and unpredictable, with
the frequency of shedding held to increase with
the progression of visible signs of disease and
pathology.
This intermittency in shedding makes
quantifying a period of epidemiological latency
for bovine tuberculosis particularly difficult.
This interval between infection and infectious-
ness is of critical importance for transmission as
it sets the natural timescale over which changes
in transmission, and therefore the impacts of
control, are expected to be manifest. Some math-
ematical models of bovine TB assume that there
is an additional ‘reactive’ latent period (R) where
animals are reactive to the skin test but are not
yet infectious (I). Barlow’s original model of
bovine TB in New Zealand herds assumed a
range of epidemiological latency (the period
spent in O and R states combined) of 180 to 600
days, while acknowledging a likely shorter range
of 87 to 226 days in experimentally infected
calves (Neill et al., 1992). Subsequent attempts
to estimate this timescale rigorously from test-
and-slaughter surveillance data sets have only
served to increase this uncertainty (Conlan
et al., 2012, 2015; Bekara et al., 2014; Brooks-
Pollock et al., 2014; O’Hare et al., 2014).
Brooks-Pollock et al. (2014) estimated a
much longer average (epidemiological) latent
period of 11.1 years (95% credible interval of
3.29–25.7 years) from their national level mod-
els of bovine TB spread within Great Britain.
Estimates from herd-level models fitted to the
same British data aggregated at the herd level
vary considerably (Conlan et al., 2012, 2015;
O’Hare et al., 2014). However, this variability
can be attributed to the very different prior
assumptions made by authors in different stud-
ies, with vague assumptions on latency (Conlan
et al., 2015) leading to longer estimated periods
of latency consistent with Brooks-Pollock et al.
(2014). A further apparent inconsistency in
these recent modelling studies is that more par-
simonious models, where animals are assumed
to be infectious immediately on meeting infec-
tion, but at a lower overall rate of infectiousness,
fit the British data equally well as models that
include a latent period (Conlan et al., 2012,
2015).