Reader Comments

Post a new comment on this article

Replication study in French Holstein and Montbeliarde cattle data

Posted by dboichard on 09 Mar 2014 at 23:43 GMT

In a recent article published by Plos One (Hinde K, Carpenter AJ, Clay JS, Bradford BJ (2014) Holsteins Favor Heifers, Not Bulls: Biased Milk Production Programmed during Pregnancy as a Function of Fetal Sex. PLoS ONE 9(2): e86169. doi:10.1371/journal.pone.0086169), the authors report a large effect of the sex of the calf on milk production, the female calves positively affecting milk production of their dams. According to these authors, this effect is observed for calves born at the onset of the lactation and also for calves the gestation of which overlaps with the lactation.

In this comment, we present a replication of this study in the two most numerous French dairy cattle breeds. We analyzed lactations in parity 1 to 3 from 2000 to 2008. This time period was chosen because since 1999 the sex of the calf has been a fully reliable information as part of the generalized animal identification system. And before 2009, use of sexed semen was negligible in France. Indeed it is known that sexed semen is used to inseminate the best cows and this can generate a bias in the estimation of the sex of calf effect on production. Similarly, dystocia is known to negatively affect subsequent production and is more frequent for male calves. To avoid such a bias due to dystocia, all data with birth conditions scored over 3 were discarded. The analyzed data sets included 1,434,000 and 7,467,000 lactations in Montbéliarde and Holstein breeds, respectively.

In a first analysis, we studied the effect of the sex of the calf the birth of which initiates the following lactation. The analyzed traits were milk, fat, and protein yields, and fat and protein contents. The statistical model was derived from the model used in the French genetic evaluation system for cattle (Robert-Granié et al, 1999) by adding the effect of the sex of the calf. This model also included the fixed effects of herd-year, of age at calving, of month of calving, and of preceding days dry in parity 2 and 3. The three latter effects were defined within year and region. The model also included the random effect of the breeding value of the cow and her permanent environment. The model accounted for heterogeneous variances between herds and year.

Results are presented in Table 1. Results presented correspond to the [(female calves) minus (male calves)] contrast. In Hilde et al’s paper, this contrast is strongly positive.

In first lactation, in Holstein as in Montbéliarde, the estimated effects were very low. A small favorable effect was observed for males in subsequent lactations of their dams. This effect reaches about 40 kg milk (eg about 0.5% of the mean), 0.6 kg fat, 0.6 kg protein. A small difference was also noticed for fat and protein contents (from 0.01 to 0.02%) in parity 2 and 3.

Table 1. Difference in milk production (female calf – male calf contrast) in the first 3 lactations
Montbéliarde breed Holstein breed
L1 L2 L3 L1-L3 mean L1 L2 L3 L1-L3 mean
Milk yield (kg) -1.9 -47.2 -40.8 -28.0 1.4 -42.7 -33.5 -23.3
Fat yield (kg) 0.66 -0.77 -0.53 -0.16 0.73 -0.63 -0.30 -0.06
Protein yield (kg) 0.01 -0.75 -0.53 -0.42 -0.12 -0.75 -0.50 -0.49
Fat content (%) 0.10 0.15 0.15 0.13 0.07 0.12 0.12 0.10
Protein content (%) 0.01 0.11 0.11 0.07 -0.01 0.07 0.07 0.03

These results were quite similar across years. For instance, in Holstein breed, the estimated effect on milk yield ranged from -62 to +86 kg for the 27 estimates (9 years x 3 parities) which represent less than 1% of the total lactation yield.

In a second study, we also analyzed the combined effect of the sex of the calf born and the sex of the calf carried during the first lactation of its dam. Editing rules were similar. However, to obtain the information about the sex of the second calf, cows were required to have a second calving. Table 2 presents the results for milk yield in first lactation, in both breeds.

Table 2. Difference in milk yield (kg) in 1st lactation according to the sex of calf born x sex of calf in gestation
Calf born x Carried calf Montbéliarde Holstein
Male x Male -2 -8
Male - Female 18 9
Female – Male -15 -16
Female - Female 0 0

As previously, in both breeds, differences in milk yield were very small.

In conclusion, our study does not confirm the results reported by Hilde et al. The effect of the sex of calf hardly affects the lactation of its dam. A small effect is observed in parity 2 and 3 and is similar across breeds. It is in favour of males, and not of females as reported by Hilde et al, and markedly smaller than their estimate. Similarly, the estimated effect of the sex of the calf in gestation on the simultaneous lactation is very small.
A major difference between the analysis of Hilde et al and ours is the statistical model used. In our approach, the model included a contemporary group (ie a herd-year effect) which limits the risk of bias.

A Barbat, R Lefebvre, D Boichard
INRA, UMR1313 GABI, Jouy en Josas, France

No competing interests declared.

RE: Replication study in French Holstein and Montbeliarde cattle data

bbradfor replied to dboichard on 02 Apr 2014 at 02:59 GMT

In a comment on our recent publication (Hinde et al., 2014), Drs. Barbat, Lefebvre, and Boichard presented results that did not appear to agree with our findings that dairy cows produced more milk following, or even during, gestation of a heifer. We appreciate the interest of this group and their efforts to replicate our findings in a large, independent data set.

Statistical models employed in genetic evaluations are complex; they are designed to account for as much non-genetic variation as possible so that individual genetic merit can be predicted with precision (VanRaden and Wiggans, 1991). However, such models are optimized for making inferences on random effects (individual animal deviation from the mean) rather than the fixed effects. As detailed in their comment, Barbat and colleagues included many factors in their model that we did not include in our analysis. Although these factors might lead to smaller standard errors around the predicted means, we suspect that over-parameterization of this model led to an erroneous conclusion that calf sex did not influence milk yield in the French data.

Among the parameters estimated in the genetic model of Barbat et al. is a “permanent environment” effect. Kruuk and Hadfield (2007) defined the permanent environment effect as “environmental effects on individual’s phenotype that are constant across (or common to) repeated measures on that individual”. Critically, this effect is mathematically-derived and the biological basis for it is undefined, though some have suggested in recent years that this parameter may capture epigenetic differences that are only now being investigated (Singh et al., 2012).

Our primary concern with the model used by Barbet et al. is that the sustained, multi-lactation effect of bearing a heifer in the first gestation fits perfectly with the definition of a permanent environment effect. We suspect that with this parameter in the model, the effect of calf sex is already accounted for, rendering the direct effect of calf sex negligible when added as a separate variable. Our confidence in our own findings is bolstered by very similar results reported in a large, independent Canadian analysis (Beavers and Van Doormaal, 2014). Importantly, the Canadian data were evaluated using a relatively simple model that did not carry the risk of accounting for calf sex in a permanent environment variable.

Fortunately, this is an easily-testable hypothesis. We request that Barbet and colleagues re-evaluate their data using the same model, with the exception that the permanent environment effect be removed. Our hypothesis is that the effect of calf sex in this revised model will more closely match our own findings.

Again, we appreciate the time and interest of our colleagues in the genetics field, and we hope that continued evaluation of this question will help to clarify the biological impact of fetal sex on the dam.

B. J. Bradford, K. Hinde, A. J. Carpenter, and J. S. Clay


Beavers, L. and B. Van Doormaal. 2014. Is sex-biased milk production a real thing? Canadian Dairy Network. Available at

Hinde, K., A. J. Carpenter, J. S. Clay, and B. J. Bradford. 2014. Holsteins favor heifers, not bulls: biased milk production programmed during pregnancy as a function of fetal sex. PLoS ONE 9(2):e86169.

Kruuk, L. E. B. and J. D. Hadfield. 2007. How to separate genetic and environmental causes of similarity between relatives. Journal of Evolutionary Biology 20(5):1890-1903.

Singh, K., A. J. Molenaar, K. M. Swanson, B. Gudex, J. A. Arias, R. A. Erdman, and K. Stelwagen. 2012. Epigenetics: a possible role in acute and transgenerational regulation of dairy cow milk production. Animal 6(3):375-381.

VanRaden, P. M. and G. R. Wiggans. 1991. Derivation, calculation, and use of national animal model information. J. Dairy Sci. 74(8):2737-2746.

No competing interests declared.