If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Identification of the associations of cow feeding behavior with productivity is important for supporting recommendations of strategies that optimize milk yield and composition. The objective of this study was to identify associations between measures of feeding behavior and milk production using data collated from studies of the feeding behavior of lactating dairy cows. A database containing behavior and production data for 132 dairy cow-week observations (mean of 7 d of consecutive data per cow) was assembled from 5 studies. Cows averaged (mean ± standard deviation) 1.8 ± 0.9 lactations, 108.4 ± 42.7 d in milk, and 654.6 ± 71.4 kg of body weight during each observation week. Production data included dry matter intake (27.0 ± 3.1 kg/d), milk yield (43.0 ± 7.0 kg/d), milk fat content (3.60 ± 0.49%), and milk protein content (3.05 ± 0.25%). Behavioral data included feeding time (230.4 ± 35.5 min/d), feeding rate (0.13 ± 0.03 kg/min), meal frequency (9.0 ± 2.0 meals/d), meal size (3.2 ± 0.9 kg/meal), daily mealtime (279.6 ± 51.7 min/d), and rumination time (516.0 ± 90.7 min/d). Data were analyzed in multivariable mixed-effect regression models to identify which behavioral variables, when accounting for other cow-level factors (days in milk, parity, and body weight) and dietary characteristics (forage level, nutrient content, and particle distribution), were associated with measures of production. Dry matter intake was associated with feeding time (+0.02 kg/min) and tended to be associated with rumination time (+0.003 kg/min) and meal frequency (+0.2 kg/meal). Similarly, milk yield was associated with feeding time (+0.03 kg/min) and rumination time (+0.02 kg/min), and tended to be associated with meal frequency (+0.3 kg/meal). Milk fat yield was associated with meal frequency (+0.02 kg/meal). Overall, our results suggest that milk yield and component production may be improved in situations where cows are able to increase their time spent feeding, in more frequent meals, and time spent ruminating.
). Dry matter intake is largely a function of feeding behavior, affected by changes in meal size, duration, and frequency, as well as feeding time and rate (
); alternatively, consuming more frequent, smaller meals, in a more consistent pattern across the day may stabilize rumen conditions, reducing risk of SARA and improving milk fat production (
The body of literature on the effect of housing, nutrition, and management strategies on dairy cow feeding behavior is continually increasing. Identification of the associations of cow feeding behavior with productivity is important for supporting recommendations of strategies that optimize milk composition and yield. Thus, the objective of this study was to identify associations between measures of feeding behavior and milk production using data collated from studies of the feeding behavior of lactating dairy cows.
A database containing behavior and production data for 132 dairy cow-week observations was assembled from 5 studies conducted at the University of Guelph, Kemptville Campus Dairy Education and Innovation Center (Kemptville, Ontario, Canada). In all studies, cows were kept in the same experimental pen, where cows had access to free stalls with waterbeds (DCC Waterbeds, Advanced Comfort Technology Inc., Reedsburg, WI), which were bedded with wood shavings, TMR was provided (1 to 3×/d) in roughage intake feed bins (Insentec B.V., Marknesse, the Netherlands), and milked on a set schedule (2 or 3×/d) using an automatic milking system (Lely A3 Next, Lely Industries N.V., Maassluis, the Netherlands) with no additional feed provided at the milking unit. Table 1 describes details of the studies used to create the database, including number of animals, parity distribution, and details of dietary composition. All diets fed were similar in ingredient composition, but varied in nutrient content and physical particle structure. Table 2 describes the average parity, DIM, and BW of cows during periods when data were collected in each study. Use of cows in each study was approved by the University of Guelph's Animal Care Committee; cows were managed according to the guidelines set forth by the Canadian Council on Animal Care (
Studies are as follows: (1) Hart et al., 2014; (2) Hart et al., 2013; (3) DeVries and Chevaux, 2014; (4) King et al., 2016a; and (5) King et al., 2016b.
Particle size of TMR as determined by Penn State Particle Separator, which has a 19-mm screen (long), 8-mm screen (medium), 1.18-mm screen (short), and a pan (fine).
(%)
DM (%)
CP (%)
NDF (%)
NEL (Mcal/kg)
Long
Medium
Short
Fine
1
6 PP and 6 MP
104 ± 32
52.4
50.4
17.1
31.7
1.60
12.6
46.5
31.9
9.0
2
7 PP and 5 MP
171 ± 31
64.0
48.9
16.7
33.0
1.70
16.7
44.6
31.5
7.1
3
2 PP and 10 MP
95 ± 17
57.7
54.7
17.9
34.4
1.66
7.2
45.7
31.9
15.3
4
4 PP and 8 MP
77 ± 23
60.3
55.8
16.7
30.8
1.65
1.8
45.0
37.0
16.2
5
4 PP and 8 MP
98 ± 23
60.3
55.4
17.2
33.4
1.64
3.6
43.9
37.0
15.5
1 All diets were TMR composed of corn silage, legume/grass haylage, high-moisture corn, grain supplement, and protein concentrate.
4 Mean DIM at the start of each data collection period for each cow (cow-week) on each treatment within each respective study.
5 Particle size of TMR as determined by Penn State Particle Separator, which has a 19-mm screen (long), 8-mm screen (medium), 1.18-mm screen (short), and a pan (fine).
Behavior data, as summarized in Table 2, were collected similarly in all source studies. Dry matter intake and feeding behavior were recorded automatically by the roughage intake feed bins (Insentec B.V.), as validated by
. Data from the feed bins were used to calculate DMI (kg/d), feeding time (min/d), and feeding rate (kg/min). Meal criteria were individually calculated for each cow, as described by
, and applied to the feeding data to calculate meal frequency (no./d), meal length (min/meal), daily mealtime (min/d; daily mealtime includes feeding time as well as nonfeeding time within meals while cows had their head outside the feed bin), and meal size (kg of DM/meal). Rumination behavior data were collected by automatic rumination detection devices (Lely Qwes-HR collars, Lely Industries N.V.), as validated by
Production data, as summarized in Table 2, were collected similarly in all source studies. Milk yield data were automatically recorded daily, at each milking, by an automated milking system (Lely A3 Next, Lely Industries N.V.). Milk samples were collected from each milking for either 3 d (
), during each experimental period using the Lely Shuttle Sampling Device (Lely Industries N.V.) and sent to a DHI testing laboratory (CanWest DHI, Guelph, Ontario, Canada) for analysis of milk fat and protein percentage using a near-infrared analyzer (FOSS System 4000 Infrared Transmission Analyzer, Foss, Hillerød, Denmark). The yield of 4% FCM (kg/d) was calculated (
) as 0.4 × milk yield (kg/d) + 15.0 × fat yield (kg/d). Energy-corrected milk was calculated using the following equation: ECM = (0.327 × kg of milk) + (12.95 × kg of fat) + (7.2 × kg of protein) (
Cows were individually exposed to either 2 or 3 treatments within the studies their data were sourced from. For the current analyses, the experimental unit was the cow-week observation, each of which was the average of daily data collected for a cow during 7-d data collection periods, per treatment, in each respective study. Data were averaged on a per-week basis to improve the accuracy of the estimate of the true mean for each predictor and outcome variable. Prior to analyses, all data were screened for normality using the UNIVARIATE procedure of SAS version 9.4 (SAS Institute Inc., Cary, NC).
Mixed multivariable linear regression models were built to assess whether there were associations between behavioral (Table 2), cow level (parity, DIM, and BW; Table 2), and dietary predictors (forage level; nutrient content: DM%, CP%, NDF%, and NEL; particle size distribution; Table 1) and production outcomes (Table 2). Models were constructed using the MIXED procedure of SAS (SAS Institute Inc.). Multiple observations per cow within each source study (i.e., different treatments) were accounted for in each model by including treatment in the repeated statement. Thus, the random effects were study and cow within treatment and study (subject of the repeated statement). Compound symmetry was selected as the covariance structure on the basis of best fit according to Schwarz's Bayesian information criterion. Degrees of freedom for fixed effects were estimated using the Kenward-Roger option in the MODEL statement.
All predictor variables hypothesized to be related to the outcome of interest were initially screened for unconditional associations in univariable analyses. Any predictor variables that were liberally associated with the outcome (P < 0.25) were considered for inclusion in the final model. Continuous variables were assessed for linearity with the outcome variables. If a variable was nonlinear, a quadratic term was constructed and tested in the model. If the quadratic terms were not significant, then logarithmic and square root transformations were tested. If none of the transformations appeared to improve linearity, the variable was categorized.
Spearman correlation coefficients were calculated for all predictor variables that were considered for inclusion in the multivariable model to detect issues of collinearity. Consequently, if the correlation coefficient was greater than |0.8|, then either the variable that made the most biological sense was used or that with the lower P-value. A confounding variable was defined as a nonintervening variable whose removal resulted in a >25% change in the coefficients of significant variables in the final model. Predictor variables in each model were checked for interaction by including all biologically appropriate 2-way interaction terms into the model. Variables were retained in the final multivariable model if they were significant at P ≤ 0.05 or tendencies at 0.05 < P ≤ 0.10, if they acted as a confounder, or were part of a significant interaction term.
Dry matter intake varied across cows in the study data sets (Table 2). When controlling for parity and BW, DMI was positively associated with feeding time, and tended to be associated with rumination time and meal frequency (Table 3). Specifically, for every hour increase in feeding time per day, DMI is predicted to increase by 0.96 kg/d (Table 3); this equates to a predicted 3.4 kg/d difference in DMI between those cows in our data set with the least and greatest feeding time (Table 2). For every extra meal per day, DMI is predicted to tend to increase by 0.19 kg/d (Table 3); this equates to a predicted 2.3 kg/d difference in DMI between those cows with the least and greatest meal frequency (Table 2). The associations with feeding time and meal frequency are not unexpected, as any alteration in DMI must be accompanied by a concomitant change in feeding behavior (
). Thus, cows in our data set were achieving higher DMI through spending more time feeding, distributed among a greater frequency of meals, rather than only consuming their feed at a faster rate, in larger meals. The association of DMI with feeding time is consistent with research conducted by
, who also identified positive association between feeding time and DMI while studying the risk factors associated with metritis in dairy cattle. The tendency for an association of DMI and meal frequency is less consistent with the literature.
reported a positive correlation between DMI of TMR and meal frequency; however, the correlation was even stronger for meal size in that study. Similarly,
found highly positive correlations between meal duration, daily mealtime, and DMI in the higher-producing cows observed in their study. It should be noted that production levels varied in these studies (
: 21.3 kg/d of DMI and 39.4 kg/d of milk yield), and were all lower than the current study (27 kg/d of DMI and 43.0 kg/d of milk yield). This indicates that at higher production levels, cows may not necessarily be able to increase DMI by simply changing the size of meals and rate at which feed is consumed within a meal, but rather need to consume more meals over a longer period of time within the day. It is also possible that these differences between studies may be related to differences in diet composition. It should be noted, however, that the range of dietary forage content was similar across these studies to those used for the current analysis, with the exception of
who fed a lower forage (40% of DM) diet in addition to more moderate levels of forage (50–60% of DM). In that study, meal frequency and size did not vary with dietary forage content. In the current study, both larger meal size and faster feeding rate were also associated with DMI at a univariable level; however, given the high correlation between these variables and meal frequency and feeding time, respectively, they could not be included in the same predictive model of DMI. Meal frequency and feeding time were chosen for multivariable modeling given that they accounted for more variability in DMI at a univariable level. Further, previous associations of measures of feeding behavior with DMI have not accounted for differences in cow-level factors (such as parity and BW), as done in the current study (Table 3).
Table 3Final multivariable linear regression models for factors associated with milk yield (n = 132 cow-weeks)
Feeding time was also associated with milk yield, and meal frequency tended to be associated with milk yield (Table 3). Specifically, for every hour increase in feeding time per day, milk yield is predicted to increase by 1.74 kg/d (Table 3); this equates to a predicted 6.2 kg/d difference in milk yield between those cows in our data set with the least and greatest feeding time (Table 2). For every extra meal per day, milk yield is predicted to tend to increase by 0.3 kg/d (Table 3); this equates to a predicted 3.6 kg/d difference in milk yield between those cows with the least and greatest meal frequency (Table 2). Similar associations with 4% FCM and ECM were also observed (data not shown). These associations between feeding behavior and milk yield are likely explained by the association of these variables with DMI.
The association of rumination time with DMI (Table 3) is also not consistent in the literature. In research with dry cows,
recently found that rumination time was a significant, but smaller contributor, in a DMI prediction model; their model predicted increases of 0.031 kg of DMI/h of rumination time. This is in contrast to the present model that predicts an increase in 0.2 kg of DMI/h of rumination time (Table 3). The difference between studies is likely due to
also using milk yield (FCM) in their predictive model; given that greater milk yield is inherently resultant of increased DMI, the FCM yield would be expected to account for much of the variance in their model. It should be noted, however, that the magnitude of the observed association in the present study is still very limited, as the current model (Table 3) only predicts a 1.4 kg/d difference in DMI between those cows with the least and greatest rumination time (Table 2). It is long established that rumination activity of cattle is correlated with total dietary NDF intake (
A new Nordic evaluation system for diets fed to dairy cows: A meta analysis.
in: Sauvant D. Van Miligen J. Faverdin P. Friggens N. Modelling nutrient digestion and utilisation in farm animals. Wageningen Academic Publishers,
Wageningen, the Netherlands2010: 112-120
). In the present analysis, dietary NDF concentration was similar across study data sets (Table 1), and thus it is not surprising that rumination time would increase with DMI level, and thus total NDF intake. One could also hypothesize that greater rumination may also contribute to greater DMI, as rumination may contribute to more effective particle size reduction, improved digestibility, and faster passage rate (
Greater rumination time was also associated with greater milk yield in the present study (Table 3). Specifically, for every hour increase in rumination time per day, milk yield is predicted to increase by 1.26 kg/d (Table 3); this equates to a predicted 8.7 kg/d difference in milk yield between those cows in our data set with the least and greatest rumination time (Table 2).
similarly found that higher-producing cows tended to have fewer ruminating bouts per day, but each bout was longer, resulting in a tendency for these cows to ruminate longer each day. Interestingly, those researchers also demonstrated that the increased rumination was proportional to increases in DMI, as cows spent less time ruminating per unit of DMI as intake and production increased (
). In the present analysis, total rumination time, and not rumination time per unit of DMI, was associated with milk yield. This suggests that, when accounting for cow-level factors, those cows with greater DMI, were then ruminating more, and also producing more milk as result of that greater nutrient intake.
Milk fat content and yield data of study cows are presented in Table 2. When controlling for parity, DIM, and dietary NDF concentration, milk fat content was associated with rumination time quadratically (Figure 1; Table 3). Two explanations are possible for this nonlinear association of milk fat content with rumination time. Evidence indicates that greater milk yield has a dilution effect on milk components (
). Thus, for those cows that were ruminating the most, milk fat may have been lower because they were also the highest production cows. This would then explain why milk protein content was negatively associated with rumination time (Table 3). It could also be that those high-production cows were increasing their rumination time to compensate for low ruminal pH. Low ruminal pH is known to be associated with milk fat depression (
When controlling for parity and BW, milk fat yield was positively associated with meal frequency (0.018 ± 0.0092 kg of fat/meal; P = 0.05), whereas protein yield was positively associated with rumination time (0.00033 ± 0.00013 kg of protein/min rumination; P = 0.01) and feeding time (0.00084 ± 0.00029 kg of protein/min of feeding time; P = 0.005). These associations are explained by the previously described associations of these behavioral predictor variables with total milk yield.
Overall, the results of this multi-study analysis identify associations between feeding behavior and production outcomes in dairy cows. Factors found to be associated with greater DMI and production include rumination time, meal frequency, and feeding time. These results suggest that nutritional, management, and housing factors that improve time spent feeding, in more frequent meals, and time spent ruminating may have a positive effect on milk yield and component production.
Acknowledgments
Thank you to the research and barn staff at the University of Guelph, Kemptville Campus, Dairy Education and Innovation Center (Kemptville, ON, Canada) for their contributions to the data collected in the summarized studies. This project was financially supported by a Natural Sciences and Engineering Research Council of Canada (NSERC; Ottawa, ON, Canada) Discovery Grant (T. J. DeVries).
REFERENCES
Aikman P.
Reynolds C.
Beever D.
Diet digestibility, rate of passage, and eating and rumination behavior of Jersey and Holstein cows.
A new Nordic evaluation system for diets fed to dairy cows: A meta analysis.
in: Sauvant D. Van Miligen J. Faverdin P. Friggens N. Modelling nutrient digestion and utilisation in farm animals. Wageningen Academic Publishers,
Wageningen, the Netherlands2010: 112-120
30073-0&id=gr1.jpg){kind=link}