Main content area

Semen evaluation and in vivo fertility in a Northern Italian pig farm: Can advanced statistical approaches compensate for low sample size? An observational study

Elmi, Alberto, Banchelli, Federico, Barone, Francesca, Fantinati, Paolo, Ventrella, Domenico, Forni, Monica, Bacci, Maria Laura
Animal reproduction science 2018 v.192 pp. 61-68
acrosome reaction, data collection, farms, farrowing rate, litter size, livestock and meat industry, neonates, observational studies, piglets, regression analysis, semen, sows, spermatozoa, statistical models, viability
The evaluation of sperm functionality and morphology allows discerning between high and low quality ejaculates, but does not give detailed predictive information regarding in vivo fertility. The current developments in statistical modeling have helped in carrying out reproductive studies, but their biggest limitation is in the size of the dataset to be used. The aim of the present observational study was to evaluate whether advanced statistical approaches, such as mixed effects regression models and bootstrap resampling, can help in assessing the predictive ability of semen parameters in terms of in vivo fertility (farrowing rate and litter size), on a small/medium farm with a limited number of animals.Data regarding 33 ejaculates, including viability, subjective motility and acrosome reaction, were collected. Two hundred and thirty-five sows were inseminated with an outcome of 167 deliveries and 1734 newborn piglets. In order to evaluate the relationships among the parameters measured and fertility, mixed effects regression statistical models were used. Once the covariates to be included in the final models were identified, non-parametric bootstrapping was used. The results showed that the farrowing rate was highly associated with the total number of spermatozoa and subjective motility, while litter size was associated with percentage of acrosome reaction. In conclusion, the proposed statistical approach seemed to be suitable for studies regarding reproduction and fertility, even for relatively small sample sizes. Nonetheless, larger data sets are still preferable and required in order to achieve higher reliability.