Genetics of twinning rate in Italian Holsteins

Twinning is undesirable in dairy cattle because it is associated with higher risks of abortion, calving difficulty, and metabolic and reproductive issues. Moreover, twin calves are lighter at birth and generally have a lower survival rate. As a result, twinning leads to substantial economic losses for the farmer. In the current study, genetic variance and h² of twinning rate (TR) were estimated as the first step to investigate the feasibility of reducing or at least curbing its increase in the Italian Holstein population through genetic strategies. Calving records (n = 1,625,859) were registered between 1992 and 2022 in 1,830 Holstein herds. A binomial logistic regression model was used to investigate the odds of TR across parities and calving seasons. The h2 and repeatability of TR were estimated using single-trait linear animal, linear sire, linear direct-maternal, threshold animal, threshold sire, and threshold direct-maternal models, which accounted for the fixed effect of parity, the random effects of herd-year-season of calving, permanent environment, and the residual and, depending on the model, the random animal, sire, or direct-maternal genetic effects. Moreover, a multiple-trait approach was adopted considering TR in different parities as different traits to estimate h2 within parity, as well as genetic and phenotypic correlations between parities. Similarly to the single-trait approach, linear and threshold animal, sire, and direct-maternal models were used. All models included the fixed effect of calving season, the random effects of herd-year of calving and the residual and, depending on the model, the random animal, sire, or direct-maternal genetic effects. The overall TR was 2.71% and 90% of the herds had TR from 0.00% to 4.49%. The greatest TR was observed after parity 2 (odds ratio ∼5.20) compared with parity 1, and in summer (odds ratio = 1.32) compared with winter. The h² increased with parity, ranging from 0.005 (parity 1) to 0.029 (parity ≥4) with linear models, and 0.061 (parity 1) to 0.142 (parity 3) with threshold models. Regardless of the model used, the genetic correlations between parities ranged from moderate to strong (0.66-0.99). Also, genetic correlations were stronger between multiparous than between primiparous and multiparous cows. Pearson correlations between sires EBV for TR obtained from single-trait linear and threshold models were close to unity, hinting at a limited re-ranking of bulls. This result suggests that there is room to carry out genetic evaluation for TR with the linear animal model (which is easier to be implemented in routine genetic evaluation than the threshold model) and manipulate the occurrence of twins through genetic strategies in the Italian Holstein population in order to stabilize or reduce TR.