2010
Sow Reproductive Stayability and Lifetime Traits
Sow Reproductive Stayability and Lifetime Traits
By Linda Engblom, Ken Stalder, John Mabry, Justin Holl, William Herring and Matt Culbertson
Sow longevity is an important economic and animal well-being trait that can be improved by selection for more robust animals. Selection for sow longevity could be performed indirectly by selection for improved fertility, or directly by implementing a longevity trait. The longevity trait could be either longevity (e.g. days in herd), stayability (removal or not after a certain parity) or lifetime production (measured in number of piglets or piglet weight).
In addition, it is possible to construct additional traits, such as lifetime productivity per herd time. The objective of the study used for this article is to evaluate the data from F1 sows with known pedigrees to estimate variance components and heritabilities for several longevity traits and other reproductive traits.
The analyses were performed using different statistical tools to try to analyze the data including the binary stayability traits as accurately and efficiently as possible.
Details of the Study
Data included 12,725 pedigreed commercial F1 (Landrace x Large White) sows with first farrowings between 2004 and 2009 on a single farm. Records were treated as complete, ignoring the 30% censoring (incomplete records since the sows were still alive). The pedigree included information about five generations and the model included the fixed effect of year-month as well as the random effects of animal and common birth litter. Phenotypic mean, heritability (h2) and effect of common litter (l2) were calculated and are presented for analyzed traits.
Several reproductive and lifetime traits were analyzed fitting normal distribution in DMU: age at first farrowing (AFF), number born alive parity 1 (NBA1), number weaned parity 1 (NW1), accumulated born alive to parity 2- 4 (ABA2 to ABA4), removal parity (RPAR), length of productive life (LPL), accumulated lifetime born alive (LBA), lifetime number weaned (LNW) and lifetime born alive by parity (LBA/P). In addition, stayability (removal or not after a certain parity or event) traits were analyzed. This binary trait was assigned a value of 0 for a sow removed before an event and 1 for a sow staying until that specific event. The events were if the animals were mated as gilts (STAY0M), after farrowing parity 1 to 3 (STAY1M to STAY3M) and if they farrowed a parity 1 to 4 (STAY1F to STAY4F). In DMU the stayability traits were analyzed both with normal and binomial variance functions fitting Generalized Linear Mixed Model. The binomial variance function was fitted with both logit and probit link functions. In THRGIBBSF90 a threshold model was fitted for the stayability traits. The Gibbs sampler ran 300,000 rounds, 100,000 burn-ins, sampling every 100 giving results based on 2,000 samples.
Summary and Implications
Sow longevity can be improved by selection, but genetic evaluations are challenged for implementation due to the long time interval required to collect complete lifetime data and low reliability of early indicator traits. This study evaluated genetic parameters for several sow reproductive, stayability and longevity traits. Heritabilities for sow reproductive and longevity traits were estimated from 0.03 to 0.09. Heritabilities of the binary stayability traits were estimated from 0.00 to 0.14. The common litter effect was largest in early recorded traits and decreased with higher parity numbers. The results show that fitting the binary stayability traits to normal variance function give the lowest heritability estimates and the binomial variance function with logit function resulted in the highest estimates.
The results show that the non-normal variance functions all gave higher heritability estimates than the normal variance function for the analyzed stayability traits.
Editor’s Note: Linda Engblom, is a post doc research associate; Ken Stalder and John Mabry are professors in the Department of Animal Science at Iowa State University; Justin Holl, William Herring and Matt Culbertson are with Smithfield Premium Genetics Group in Rose Hill, NC. The authors gratefully acknowledge Dr. S. Tsurata at the University of Georgia for his help with the THRGIBBSF90 program. We also wish to acknowledge Smithfield Premium Genetics for making this study possible by providing the data. For more information on this study, go to www.benchmark.farms.com