02.02.2013 Views

Proceedings of the 10th International Colloquium on Paratuberculosis

Proceedings of the 10th International Colloquium on Paratuberculosis

Proceedings of the 10th International Colloquium on Paratuberculosis

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Subsequently, (II) under <str<strong>on</strong>g>the</str<strong>on</strong>g> proposed model, we assessed <str<strong>on</strong>g>the</str<strong>on</strong>g> associati<strong>on</strong> between time to<br />

MAP infecti<strong>on</strong> and fecal shedding status. The risk change over time for seroc<strong>on</strong>versi<strong>on</strong> or<br />

acquiring MAP infecti<strong>on</strong> was also estimated under I and II, respectively.<br />

MATERIALS AND METHODS<br />

The model<br />

We assumed that <str<strong>on</strong>g>the</str<strong>on</strong>g> unobserved time to infecti<strong>on</strong> (tri) is Weibull distributed:<br />

f ( tr ) = � � � � tr<br />

i<br />

i<br />

��1<br />

i �<br />

e<br />

��<br />

�<br />

i* tri<br />

b zi<br />

,where i e *<br />

� = , zi <str<strong>on</strong>g>the</str<strong>on</strong>g> covariate vector for <str<strong>on</strong>g>the</str<strong>on</strong>g> i th individual, b a vector <str<strong>on</strong>g>of</str<strong>on</strong>g> unknown<br />

regressi<strong>on</strong> coefficients and ñ <str<strong>on</strong>g>the</str<strong>on</strong>g> shape parameter <str<strong>on</strong>g>of</str<strong>on</strong>g> <str<strong>on</strong>g>the</str<strong>on</strong>g> Weibull distributi<strong>on</strong>. The<br />

unobserved tri was modeled as a functi<strong>on</strong> <str<strong>on</strong>g>of</str<strong>on</strong>g> <str<strong>on</strong>g>the</str<strong>on</strong>g> observed time to detecti<strong>on</strong> <str<strong>on</strong>g>of</str<strong>on</strong>g> infecti<strong>on</strong> (ti)<br />

and <str<strong>on</strong>g>the</str<strong>on</strong>g> latent infecti<strong>on</strong> period (ui):<br />

tr = t � u<br />

i<br />

i<br />

i<br />

With appropriate prior informati<strong>on</strong> <strong>on</strong> ui <str<strong>on</strong>g>the</str<strong>on</strong>g> model is fully specified. Extended model<br />

descripti<strong>on</strong> with step by step explanati<strong>on</strong> <str<strong>on</strong>g>of</str<strong>on</strong>g> <str<strong>on</strong>g>the</str<strong>on</strong>g> corresp<strong>on</strong>ding WinBUGS code is available<br />

up<strong>on</strong> request from <str<strong>on</strong>g>the</str<strong>on</strong>g> corresp<strong>on</strong>ding author.<br />

Data<br />

Data were collected from February 2000 to March 2003 from 1912 cows, in 8 Danish dairy<br />

herds. Briefly, milk samples were obtained 11 times/year from all lactating cows and tested<br />

by an indirect milk ELISA. Fecal samples were collected from all lactating and n<strong>on</strong>-lactating<br />

cows 4 times/year and were cultured <strong>on</strong> LJ or HEYM. Fecal culture (FC) positive cows were<br />

divided in two shedding groups: FCHigh if positive with high bacterial counts (>10 cfu/g) or<br />

many c<strong>on</strong>secutive FC-positive tests and FCLow o<str<strong>on</strong>g>the</str<strong>on</strong>g>rwise. Animals not shedding or shedding<br />

undetectable levels <str<strong>on</strong>g>of</str<strong>on</strong>g> MAP were classified as FCNeg. An animal c<strong>on</strong>tributed a censored<br />

observati<strong>on</strong> if it was still milk-ELISA negative when it exited <str<strong>on</strong>g>the</str<strong>on</strong>g> study for a reas<strong>on</strong> not<br />

associated with MAP infecti<strong>on</strong>.<br />

Priors<br />

For <str<strong>on</strong>g>the</str<strong>on</strong>g> proposed model we had to specify prior informati<strong>on</strong> about ui from <str<strong>on</strong>g>the</str<strong>on</strong>g> inducti<strong>on</strong> <str<strong>on</strong>g>of</str<strong>on</strong>g><br />

MAP infecti<strong>on</strong> to milk seroc<strong>on</strong>versi<strong>on</strong>. For naturally infected dairy cattle, we chose a prior<br />

value <str<strong>on</strong>g>of</str<strong>on</strong>g> ui equal to 1300 days (3.5 years) with a 95% c<strong>on</strong>fidence interval extending from<br />

1000 (2.7 years) to 1600 (4.4 years) days: That is a normal N~(1300, 4.4 x 10 -5 ).<br />

RESULTS<br />

Estimated model parameters under (I) <str<strong>on</strong>g>the</str<strong>on</strong>g> standard and (II) <str<strong>on</strong>g>the</str<strong>on</strong>g> proposed Weibull model are<br />

given in Table 1. The shape <str<strong>on</strong>g>of</str<strong>on</strong>g> <str<strong>on</strong>g>the</str<strong>on</strong>g> Weibull hazard – as expressed by ñ – and <str<strong>on</strong>g>the</str<strong>on</strong>g> baseline<br />

hazard – expressed by b0 – differed between models I and II. Median time to milk<br />

seroc<strong>on</strong>versi<strong>on</strong> was 2638 [95%Credible intervals (CrIs): 2542; 2743], 1880 (1758; 2016) and<br />

1722 (1594; 1869) for FCNeg, FCLow and FCHigh cows, respectively. The corresp<strong>on</strong>ding median<br />

time to MAP infecti<strong>on</strong> was 1726 (1420; 2160), 356 (255; 502) and 227 (153; 339) days. The<br />

risk <str<strong>on</strong>g>of</str<strong>on</strong>g> MAP infecti<strong>on</strong> decreases with time (ñ=0.56 (0.51; 0.61) 1 under I). This decrease and increase,<br />

respectively, is significantly steeper for shedders (heavy or low) than cows not shedding or<br />

shedding undetectable levels <str<strong>on</strong>g>of</str<strong>on</strong>g> MAP (see Figure 1).<br />

152<br />

2

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!