• No se han encontrado resultados

VISIONES DE MUNDO, SABERES OTROS, A PROPÓSITO DE LA MEMORIA, EL TERRITORIO Y LA RESISTENCIA

3.5 Memoria, entramada de sentidos

Most animals used in the study were lambs or hoggets (only 30 animals were greater than one year of age), hence the effect of age at inoculation was difficult to evaluate. Only one study used adult sheep with a wide age range up to until 11 years (Delgado et al., 2012; Delgado et al., 2013).

We tested the effect of age at inoculation as a continuous variable (linear or not), or using categories corresponding to production groups relevant to the mathematical model of ovine paratuberculosis (lambs/hoggets/two-tooth and ewes) or as a dichotomous variable (less/more than one year old) in all models. The only model where a significant age effect

156

could be detected was the model of probability of progression. The odds of presenting markers of progression at any point in time were 12 times higher in animals inoculated before

one year old (versus older than one year old). This resulted into estimating a different χ

parameter in these two categories of age, which would have important repercussions for the simulated infection dynamics.

For all other parameters estimated, there was no difference between age groups, due to no significant age effect detected in the GAMMs. This, however, means that all estimates were driven by dynamics observed in young animals which were the majority in the dataset.

6.5.5.1

Effect of strain type

The strain type used in experimental studies (“ovine” or type I versus “bovine” or type II) is not always explicitly given, especially in older studies. In recent studies, the strains used for inoculation in experimental challenge were usually typed by PCR, which can be considered virtually perfect to distinguish the two main types of MAP. In numerous older studies, however, this was not the case. In such cases we used robust bacteriological and

epidemiological criteria (educated guess) with the help of expert advice (Karen Stevenson10

and Marian Price-Carter11) to determine the likely strain type. We also checked the robustness

of the results obtained by running all models with different classification systems, i.e. if all the

strains that were not typed had been ovine or alternatively bovine. The inferences remained the same in these models compared with models based on the educated guess.

The only model in which the strain type had a significant effect was modelling the prevalence

of active infection (model 1): the bovine strain appeared as a risk factor for the occurrence of

histology lesions (any grade) compared to ovine strain. The effect in the model of the presence

of mild-moderate lesions in the guts (model 3) was similar but not significant. This is in line

with the literature where an increased number of lesions, and more extensive lesions, were described in sheep inoculated with bovine MAP compared to ovine MAP, despite a tendency to later recover from the lesions caused by the bovine strain (Chapter 4). The main conclusion is that if the bovine strain seems more infectious than the ovine strain in sheep (more likely to cause active infection), the fate of sheep in terms of progression to clinical paratuberculosis is not significantly different between both strains.

10

Moredun Research Institute, Pentlands Science Park, Bush Loan, Penicuik, EH26 0PZ, Scotland, UK

11

Hopkirk Research Institute, AgResearch Grasslands, Tennent Drive, Private Bag 11008 Palmerston North 4442, New Zealand

157

6.5.5.2

Effect of breed

Breed was partially collinear with strain type of MAP as all Romney and crosses and the vast majority of Merino and crosses were inoculated with an ovine strain while most other British breeds were inoculated with a bovine strain. This made it difficult to evaluate the effect of one versus the other. Moreover, breed was a proxy for the geographical region where the study was performed. In final models, the covariate strain type (only two levels) was preferred since breed did not improve the models over the effect of the strain type when the latter was significant.

In experiments undertaken in Australasia, both Romney and Merino crosses were infected with similar MAP strains, thus location and strain type did not confound the effect of breed. Results for these (not shown) suggest no obvious difference between those breeds in their susceptibility to MAP.

6.5.5.3

Effect of experimental conditions

One of the primary goals was to measure the effect of inoculum dose, thus the ‘infectiousness’ of MAP. We discarded one paper (Dukes et al., 1992) and another 10 sheep from an experiment that was included in the meta-analysis (Stewart et al., 2004) because these animals were dosed with tissue homogenate and the dose was reported as the weight of tissue, with no further attempt to enumerate the number of MAP. In our model, the “dose” corresponded to the total oral dose supposedly received by each animal (at the logarithmic scale). Most of the time, MAP was given in several inocula (up to 10) at various time intervals, with typically a few days between each dose. The total dose was the sum of all doses administered. The time

from inoculation to post-mortem was calculated from the date of the first inoculum. To

evaluate a potential effect of repeated doses, we tested the total number of doses

administered as a continuous covariate in the two main models model 1 (active infection) and

model 2 (progression). This was also a proxy for duration of the dosing regimen, although time intervals varied across studies. This effect was not significant and not kept in the models. Experimental data are needed to measure the effect of MAP on the pathology of paratuberculosis. However, the effect of chronic exposure to MAP in a natural environment, although impossible to estimate, might trigger a different host response.

We also evaluated the possible bias due to MAP enumeration methods used to quantify the inoculum in experimental studies, on the estimated effect of dose. Each enumeration method comes with inherent biases and we classified them into two categories (Chapter 5). Culture-

158

based methods, counting viable/cultivable MAP) might underestimate MAP numbers. Physical methods counting total number of MAP might overestimate numbers by counting dead MAP cells. We used an interaction term between these two categories and the inoculum dose in the

two main models (model 1 and model 2). The effect of dose was very similar within each

stratum and the interaction was not significant. Hence, the bias induced by enumeration methods was small compared to the amplitude of the effect of varying the inoculum doses itself. In this respect, this meta-analysis failed to generate robust conclusions about which enumeration method would seem more appropriate.