Abstract :
[en] Length of productive life (LPL) defined as the time from first calving to culling is an ambiguous trait resulting from many factors. This preliminary study aims to link LPL to several of these factors including lactation curve traits. Lactation curve traits for milk were provided by the Walloon Breeding Association for 20,766 primiparous Holstein cows calving from 2003 to 2014 in 395 larger herds (>50 cows). A linear model that included fixed effects of herd (H), calving year (CY), 4 classes of calving season (CS), 4 classes of milk production adjusted to 305 days (M305), and 9 classes of age at first calving (AFC) together with linear regressions on LPL expressed in days in milk (DIM), peak production (PK), persistency (PS) defined as a post-peak slope, was used to study variations of LPL. Effect of censored LPL was reduced by the CY effect. Observed means were 7,427 kg (M305), 393.7 days (DIM),45.9 months (LPL) and 27.7 months for AFC. As expected LPL was affected by herd (P<0.01) linked to management strategies among herds, which culling decisions, and by calving year (P<0.01) as animals born earlier are expected to be able to achieve a higher LPL. Season had a significant effect on LPL (P<0.01), as autumn seemed the least favorable season. Cows with intermediate yields tended to stay longer in the herd than low (M305<5,000) or high (M305>10,000) producing ones, a result that can be explained by voluntary culling of low producing cows and of involuntary culling for reasons such as diseases for very high producing cows. Results for AFC were less clear, later calvings (>28 month) being linked to lower LPL. For linear regressions of DIM on LPL, significance was close (P<0.06) for a positive relationship, but this could be an artifact due to the definition of LPL. Finally, the lack of significance for the regressions of PK and PS on LPL in this study should be considered a preliminary result for two reasons: (1) the simultaneous presence of other correlated effects as milk yield in the model; and (2) the use of a linear regression when non-linear relationships are more likely (intermediate optimum).
