|
|
Investigating temporal patterns |
|
The infection dynamics of BDD are poorly understood and
have not been subject to rigorous scientific investigation.
Our understanding of the factors governing the transmission at the
group and farm level is still mostly hypothetical; as some of these factors are
time-varying (e.g. environmental hygiene), they are likely to directly
influence the seasonal variability of the disease. For the stated objective of
developing effective intervention strategies, such knowledge is a prerequisite.
Few studies have been published specifically investigating BDD
incidence rates and temporal disease trends. Laven and Lawrence (2006) analysed
a database of veterinary treatments over an eight-year period recorded by a
network of 40 veterinary practitioners; they found that the morbidity of BDD
had a significant seasonal component, with fewer reports being made from June
to October. However, they inferred that the overall seasonality of BDD had
decreased since the early 1990s. They attributed this to the continued spread
of BDD, leading to an endemic disease pattern on affected farms, where the
disease persists into the grazing period. Holzhauer et al. (2006) compared the
results of their cross-sectional study with those of Frankena et al. (1991) and
concluded that the morbidity of BDD had increased in the intervening 15 years;
they suggested that this could possibly be associated with prolonged housing
periods.
A longitudinal study was carried out from February 2004 to December
2005 on four of the cross-sectional study farms. Three cohorts consisting of
calves, heifers and cows were selected (n=128); these were dynamic, i.e.
drop-outs were replaced. Sampling was performed fortnightly during the housing
season, and every four weeks during the grazing season (n=1548). At each visit,
foot inspection, blood sampling and animal hygiene scoring were carried out.
Similar statistical approaches were used as for the cross-sectional study (i.e.
exploratory data analysis followed by statistical modelling), but with specific
emphasis on detection of temporal trends. Models were formulated to account for
the covariance structure and autocorrelation in the data. For clinical BDD
(binary outcome variable), a generalized linear mixed model was fitted; the
seasonality of the incidence was detrended by inclusion of a sine function, and
a first-order transition model was fitted. For serology (continuous outcome
variable), no temporal detrending was required (see below); autocorrelation was
compensated for by fitting a first-order continuous autoregressive model
(CAR(1)).
We identified a strong
seasonal trend in the clinical incidence of BDD, with a maximum between
November and January; interestingly, the incidence rate started declining
before the end of the housing period, and continued into the grazing period.
Minimum incidence occurred between June and July. However, it remained
substantial throughout the grazing period, and newly-developing lesions were
observed all year round. There was some variability in clinical manifestation
and prevalence between farms, which was possibly related to farm management
practices such as footbathing, or climactic and environmental conditions, time
of turning out and re-housing, and farm-specific BDD infection dynamics. There
was also some variation in BDD incidence per year, with the incidence in 2005
being lower. Lesion presentation and development were variable and changed
dynamically, and recurrence was common. This complicated investigation of the
relationship between the humoral response and clinical disease. First calving
heifers did not have a higher incidence of lesions, but the lesions tended to
be more severe. Cows older than five years had a higher proportion of
regressing lesions. The incidence of lesions increased from the start of
lactation to a maximum around the peak of lactation, thereafter decreasing
slightly.
Combining the information from Laven and Lawrence (2006), Holzhauer
et al. (2006) and our study, it can be speculated that observed increases in
morbidity are due to the disease having become endemic on most farms (possibly
due to a prolongation of housing periods), as a result of which new lesions
continue to develop throughout the grazing period.
The incidence of
clinical BDD was positively associated with age. No lesions were seen in
animals younger than one year old on the date of sampling, and in few in-calf
heifers; in all but one case, those lesions that were observed were small,
acute, did not persist for more than two weeks, and did not lead to
seroconversion. This implies that clinical BDD occasionally occurred in the
in-calf heifers group, but did not lead to chronic lesions; it was not observed
to spread rapidly or widely. There was a sharp increase in the clinical
incidence after the first calving. Compared to older cows, the proportion of
acute lesions was much larger for these heifers. With increasing age, the
proportion of acute lesions decreased while that of chronic and regressing
lesions increased. BDD prevalence rose from the start of lactation, reaching a
maximum at the peak of milk production; it remained reasonably constant
thereafter, but decreased towards the end of the lactation. This suggests
development of some degree of resistance over the duration of the lactation.
However, neither lactation stage, nor parity were significantly associated with
BDD by the transition model. Body hygiene score (BHS) was significantly
associated with incidence of clinical BDD. We were unable to incorporate group-
and farm-level variables, particularly housing hygiene, into our statistical
models. This was because inspection of the housing was unlikely to be
representative due to the variability in the housing hygiene, depending on when
it had previously been cleaned or scraped out. Also, creating compound indices
for these variables is complex and subjective.
Serology provided further
clues on the exposure to BDD-associated Treponema spp. As determined
both by the EDA and the CAR(1) model, the IgG2 titres were very low
at birth, showing a gradual rise up to the age of two to three years, which
coincides roughly with the age of first calving. Subsequently, the range of PP
values increased markedly, which was presumably due to some animals
seroconverting after infection, while other animals remained uninfected. The
mean serologic titre continued to rise up to an age of four to five years,
after which it remained more or less constant. There was no discernable
lactation stage effect, suggesting that exposure may be continuous.
These results are consistent with an endemic disease pattern and
provide evidence of development of partial immunity after long-term exposure.
Currently, we do not have the microbiological tools which enable us to identify
sources of the infection. Our serologic results suggest that young stock are
exposed to the infection at a low level, possibly as a consequence of spread of
infectious material by movement of farm materials, vehicles and personnel. The
force of infection appears to rise in the heifer groups; this may
be related to the relatively poorer hygiene in these groups. However, most
animals were immunologically naive upon first entering the lactating cows
group. The level of exposure was probably much higher in these groups.
In contrast to the clinical incidence, no seasonal effects were seen
for the serology. This was surprising, since we had established that BDD lesion
positive animals had significantly higher titres than lesion negative animals.
Also, our results indicated that the half-life of the IgG2
antibodies was short, decaying during or soon after regression of the lesions.
It was therefore logical to expect that the serology would show a similar
seasonal pattern as the incidence of clinical lesions. Two possible
explanations can be advanced for this.
Firstly, the relatively high
incidence of new lesions observed during the grazing period indicates that
there may be continuous exposure to the causative organisms. In accordance with
our causal web, the lesions could regress during the grazing period due to
drier and cleaner underfoot conditions; however, the continued exposure could
still be sufficient to maintain substantial antibody titres, and in some cases
bring about development of new lesions. From our CAR(1) model, autocorrelation
was found to persist for a duration of about 230 days.
|
|
Fig. W6. Sample
variogram of residuals from the CAR(1) model for BDD serology. Each point
represents the half-squared difference between a pair of residuals, plotted for
all time intervals. The red line is kernel smoothed with a bandwidth of 150
days; the dotted line is the 'process variance'. Autocorrelation decays to zero
when the kernel smoothed line crosses the process variance (refer to Diggle et
al. (2000) for details) |
|
The second explanation relates to the association
between the humoral response and clinical disease. From the serological
distributions of clinical negatives and positives, it was observed that the
number of clinical negatives with high IgG2 titres was more
substantial than the number of clinical positives with low titres. Further
investigation showed a clear age effect, with older cows having a much higher
proportion of such serologic false positives. It is therefore likely that older
animals, who have presumably had a prolonged exposure to the organisms, are
more likely to seroconvert without developing clinical BDD.
Inspection of
plots of individual animals relating the serological profile to clinical
disease were difficult to interpret due to the variability of the lesion
presentation between sampling observations. Progression of the disease was not
necessarily linear, where this is defined as transition from acute lesions to
chronic lesions, which then regress. In many instances, acute lesions regressed
spontaneously, while in others new foci of infection developed out of chronic
or regressing lesions. The duration of persistence of the lesions was also
highly variable. Recurrence or relapses were common.
In initial
studies (Walker et al., 1997; Demirkan et al., 1999), the serology was strongly
associated with clinical BDD, and very few inconsistent results were reported.
It is possible that in this period, animals had only been relatively recently
exposed to the infection, and hence showed a pattern consistent with primary
infection; no degree of immunity (the mechanisms of which are unknown) had yet
developed. As the disease became endemic in successive years and animals were
repeatedly challenged, the development of a degree of immunity has led to older
animals seroconverting without developing clinical BDD.
Another
explanation of the findings of these and more recent studies, such as that of
Trott et al. (2003), is that they were experimental case-control studies rather
than population-based in design; and smaller numbers of animals were studied.
It is therefore possible that the extent of these inconsistent serologic
results could simply not be determined. |
|
|
|
|
©
Willem Daniel Vink 2006 |