Early lactation performance of primiparous dairy heifers reared with differing levels of early-life complexity

Animal management

This experiment was conducted at the Tasmanian Dairy Research Facility (TDRF) near Elliott in north-west Tasmania, Australia (41°080 S, 145°770 E; 155 m above mean sea level). All animal procedures were approved by the University of Tasmania Animal Ethics Committee (A0018141) under the Tasmanian Animal Welfare Act 1993. Sixty dairy heifer calves were enlisted in the initial treatment period of this experiment, which ran from August to November 2019, while heifers were aged 2–13 weeks. After the treatment period, heifers were mixed into a single replacement heifer group along with other unfamiliar non-experimental replacement heifers (~80), and all animals were managed identically by TDRF staff, following the farm’s commercial protocol. Data presented in the present paper were collected from 48 heifers who remained in the dairy herd between July and October 2021, when heifers were approximately 23–25 months old. This reflects a 20% dropout in experimental animals from treatment imposition to the period described in this study.

The following sections describe the management of heifers throughout the duration of this longitudinal study, with specific detail given to management over the transition period into the commercial milking herd. Table 1 offers a brief overview of the management of study heifers from birth until calving, and includes references to published studies undertaken at several points across this period.

Measures recorded

All data from the first month of lactation were collected without any interaction between researchers and animals, and experimental heifers were managed by the research farm identically to all other animals in the commercial milking herd in which they were housed. A timeline of data collection points is provided in Fig. 1.

Fig. 1.

Sampling timeline illustrating sampling periods for all variables measured in the present study. DIM, days in milk from day of calving; BCS, body condition score; LW, liveweight.

Data management

The initial treatment was imposed on 60 heifers; at their first calving at approximately 23 months of age, 50 heifers remained in the TDRF replacement herd. Dropout of the 10 lost heifers was explained by failure to get in calf, or genetics outside the farm’s longitudinal breeding goals.

One CC heifer did not calve and was pregnancy diagnosed as empty at the beginning of September 2021; this animal was removed from the experiment. Twenty-five experimental heifers calved during the EC peak, between 12 July 2021 and 23 July 2021 (+S = 6, −S = 9, CC = 10). Twenty-four experimental heifers calved during the LC peak, between 30 July 2021 and 8 August 2021 (+S = 10, −S = 8, CC = 6). As calving occurred according to success of conception and ensuing gestation length experimental protocol could not balance or otherwise influence the distribution of breed across calving peaks; these data within treatments is presented in Table 2.

Accounting also for the EC −S heifer which remained in the sick herd for her first month of lactation, the data analysed therefore pertain to 48 experimental heifers (+S = 16, −S = 16, CC = 16). For other heifers which spent any proportion of their first month of lactation in the sick herd, data from the days which they were not in the main or fresh herds were removed from analysis.

Data were also discarded on three study days for various reasons. All data from 11 August and 8 September were discarded, as a routine herd testing processes caused significant disruption to the daily routine and management of the herd. This affected data from 37 heifers and 14 heifers respectively. Electrical supply issues on 9 September resulted in significant disruptions and delays to milking; this affected data from 15 heifers. A total of 10 milk cortisol samples were also discarded from these dates, comprising three week 1 cortisol samples and four week 4 cortisol samples from 11 August, two week 4 cortisol samples collected on 8 September and one week 4 milk cortisol sample collected on 9 September.

Once these data were removed, mean DIM at heifer entry to the main herd from the fresh herd was 7.3 ± 2.7 (range 4–19 DIM). Mean day of entry to the main herd by treatment was 7.8 ± 4.0 DIM for +S heifers (range 5–19 DIM), 6.7 ± 1.8 DIM for −S heifers (range 4–10 DIM), and 7.5 ± 2.0 DIM (range 5–12 DIM) for CC heifers.

Statistical analysis

All data were analysed in RStudio ver. 4.3.1 (2023 R Core Team: The R Foundation for Statistical Computing, Vienna, Austria) with linear mixed models. A Restricted Maximum Likelihood (REML) approach was used to fit all models, and account for missing data across the datasets. REML can be reliably applied to unbalanced data and analyses incorporating both fixed and random effects (Dieters et al. 1995).

In the following descriptions of the analyses, ‘Heifer’ denotes the individual animal from whom data was collected. In all analysis, ‘Heifer’ was nested within ‘Rearing group,’ and this was included in all statistical models as a random effect to account for the individual group a heifer was reared with (i.e. group 1–6 during rearing, wherein +S Replicate one heifers belonged to Group 1, +S Replicate two heifers belonged to group 4, and so on). ‘Treatment’ denotes +S, −S or CC early-life management protocol.

‘Calving Peak’ was included in most statistical models to account for potential differences in experiences between the EC and LC heifers during their first month of lactation (such as proportion of mature cows in the herd, and differences in pasture growth and quality), except for cortisol analysis, for which number of records precluded analysis to this level. Where appropriate, ‘Days in Milk’ (DIM) or ‘Week in Milk’ was included in the model to account for the day or week of lactation each data point corresponded with. For BCS, LW and ADG analysis, week differences were analysed using the previously described 3-day means, labelled in analysis as ‘Week_1st3DayDIMS.’ ‘Herd Number’ accounted for whether heifers were in the main or fresh herd. Within specific analyses, the interactions between two variables were also explored.

Each dataset was fitted to a linear mixed model. Variable-specific factors were initially chosen for inclusion in each full model, before ANOVA comparison of models were conducted to ascertain whether the inclusion or exclusion of each fixed factor had a significant effect on the outcome of the model. Final models were derived from the outcomes of these ANOVA comparisons.

In analysis of milk yield, BCS, LW and ADG, initial analysis tested the effects of treatment, week in milk (1–4; calculated for BCS, LW and ADG using mean values from the first 3 days of each week), and their interaction, treatment × week in milk; calving peak (1/EC or 2/LC), and the interaction of treatment × calving peak and treatment × calving peak × week on outcome variables.

Daily milk yield

In data collected in this study, there was a significant interaction between treatment × calving peak × week on milk yield (P < 0.001). Across most treatments and calving peaks, milk yield generally followed an upward trend over the first 3 weeks of lactation (Fig. 2). Least-squares means (LS-means) for milk yields in week 4 were, however, slightly lower than in week 3 for all treatments in the LC peak, but only for the +S heifers in the EC peak (Table 3, Fig. 2).

Fig. 2.

Milk yield over weeks, by treatment and calving peak (graphed data represent LS-means ± s.e. from the linear mixed model MilkYield ~ Treatment × CalvingPeak × weekInMilk + (1|RearingGroup/Heifer)).

A significant interaction between treatment × calving peak across weeks (P = 0.011) is reflected in the values presented in Table 3, with −S heifers producing the least milk in the EC peak but the most in the LC peak.

Body condition score

There was a significant interaction effect of treatment × calving peak × week on BCS (P = 0.001; Table 4). Heifer BCS generally followed a downward trend in most treatments and calving peaks across the 4-week period, with a more linear and greater magnitude of decline observed in LC than EC heifers (Fig. 3). The −S heifers generally had the lowest BCS across the EC peak, but the highest across the LC peak.

Fig. 3.

Body condition scores over weeks, by treatment and calving peak (graphed data represent LS-means ± s.e. from the linear mixed model BCS ~ Treatment × CalvingPeak × Week_1st3dayDIMS + (1|RearingGroup/Heifer)).

Liveweights and average daily gain

Only week in milk significantly affected ADG (P < 0.001). While LW generally followed a downward trend across the first 4 weeks of lactation (effect of week on LW P < 0.001), weekly changes in ADG generally trended upwards but without following any linear pattern; LW was lost over the 28-day study period, but the magnitude of weight loss appeared to lessen with time (Table 5 and Supplementary S1). While the least-squares means for LW of EC +S and CC heifers were within 10 kg of their LC counterparts, the EC −S heifers entered lactation 57.5 kg lighter than LC −S heifers (LS-means 440.32 ± 10.58 vs 497.81 ± 11.35, Supplementary S2). This difference is reflected in a significant interaction effect of treatment × calving peak on LW (P = 0.004).

Milk cortisol concentration

There was no significant interaction effect of week number (1, 2 and 4, corresponding to DIM 7, 14 and 28) with treatment on milk cortisol concentrations (P = 0.315), nor was there an effect of herd on cortisol concentration (i.e. fresh or main herd, P = 0.333).

Sample number had a significant effect on milk cortisol concentration for all heifers (P = 0.009; raw cortisol data presented in Supplementary S3). Cortisol concentrations were highest at 14 DIM (LS-means 1.491 ± 0.199) and lowest at 28 DIM (LS-means 0.716 ± 0.267), corresponding to a decrease of 70.2% between the two sampling periods. The initial samples, collected at seven DIM sat between these two concentrations (LS-means 1.008 ± 0.274).

Early-life treatment and physiological mobilisation

A calving peak × treatment × week interaction for both milk yield and BCS appeared to largely be due to differences between −S heifer groups between calving peaks. The greatest observable difference was discrepancies in LW at calving between −S EC and −S LC heifers. In means taken over the first three DIM, −S EC heifers were an average of 50.7 kg lighter (per LS-means) than their LC counterparts, compared to an approximately 10 kg difference between +S and CC calving peaks. Despite this, −S heifers across both calving peaks calved at a similar BCS. The −S heifers in the EC peak had the lowest milk yield and BCS (by LS-means) over the first 4 weeks of lactation, compared to heifers of the +S and CC treatments; however, the opposite was true for −S LC heifers. Given the mobilisation of body reserves necessary for dairy cattle to reach peak lactation, it is possible that −S LC heifers were more physically mature, or had greater body reserves to support establishing milk production and the physiological stress associated with the process. This may have supported these heifers to maintain a higher level of production across the first 4 weeks of lactation. Regardless, there is no apparent indication as to why this difference in −S groups across calving peaks occurred. Distribution of pure-breed animals was identical in −S groups across calving peaks. The most likely explanation is that these disparities are simply the result of naturally occurring variation between two small groups of animals.

Physiological mobilisation in early-lactation heifers

When exploring heifer data outside of treatment effects, we found that the LS-means for BCS in the first 3 d after calving in the present study ranged from 5.10–5.33 in EC heifers to 4.94–5.25 in LC heifers on a 10-point scale. This is slightly higher than the mean BCS of 4.72 at calving across all pasture-based Jersey and Friesian study animals reported by Roche et al. (2007). Industry body Dairy Australia (2015) recommend a target BCS between 4.5 and 5.5 at calving on an 8-point scale (approximately equivalent to 4.4–6.1 on a 10-point scale, per Roche et al. 2004). Research suggests that the ideal BCS on a 10-point scale at calving for ongoing milk yield is between 5.0 and 6.5 (Roche et al. 2007). The results of the present study thus indicate that study heifers across were in relatively good commercial condition, meeting both industry recommendations and BCS required to meet their yield potential in their first lactation.

Collectively, the physiological changes observed in heifers in this study are typical of those observed in commercial herds and other research (e.g. Morton and McBride 2004; reviewed by Bauman and Currie 1980; McNamara et al. 2008). The yield of our primiparous heifers generally followed an upward trend in the first 4 weeks following calving, while BCS declined along with their liveweight. Roche et al. (2007) observed a nadir (lowest point) in body condition score (BCS) at an average of 48 days post-calving in pasture-based Jersey and Friesian dairy cattle; these results are in keeping with the trend observed in our own. The rate of weight loss, as determined from ADG calculated from weekly changes in automatically collected daily liveweights, decreased in magnitude as the first month of lactation progressed.

Patterns of cortisol concentration during early lactation

Higher apparent fearfulness or nervousness in dairy cattle has separately been loosely correlated with reduced milk yield (Hedlund and Løvlie 2015). In previous tests with the heifers sampled in the present study, we found that increasing early-life complexity appeared to improve heifer responses to novelty and mixing with mature animals (Field et al. 2023a, 2024b). While we had hypothesised that, given these differences, heifers from more complex early-life treatments (+S and −S) would have lower concentrations of cortisol in their milk samples, the final statistical model for milk cortisol analysis ultimately did not include treatment in its terms.

In the present study, there was a significant effect of sample number (1–3, reflecting d 7, 14 and 28 of lactation) on milk cortisol concentration (ng/mL). The highest concentrations were recorded from milk collected at 14 DIM, approximately a week after the day of transition from the fresh to main milking herd, and the lowest concentration 2 weeks later, at 28 DIM. Previous research has found that milk cortisol concentration is significantly higher in early lactation than mid-late lactation (Shutt and Fell 1985; Fukasawa et al. 2008; Gellrich et al. 2015). Shutt and Fell (1985) recorded milk cortisol concentrations of 4.4 ± 1.3 ng/mL in the 1–2 d following parturition, which dropped to 0.35 ± 0.06 ng/mL as lactation established. Fukasawa et al. (2008), meanwhile, defined early lactation as any time between 7 and 90 DIM and found significant differences between cortisol concentrations in samples taken over this period compared to those collected months later, with no effect of parity. In the present study, the three samples were collected within a sampling timeline more comparable to that described in Gellrich et al. (2015). These authors sampled twice weekly in weeks 2 and 3 post-partum and weekly thereafter, but used a modified ELISA rather than an RIA kit, so concentration of cortisol in samples cannot be directly compared (Gellrich et al. 2015). Regardless, a similar trend was found in the results of these authors, wherein samples from the week encompassing 8–14 DIM were significantly higher than the relatively stable concentration of cortisol in samples collected in the following weeks.

There are two stressors most likely to have influenced the pattern observed in milk cortisol concentrations across 7, 14 and 28 DIM in this study. Firstly, the first sample was taken just before the mean date of entry to the main herd, while the second was taken after most heifers had been in the main herd for 1 week. The slight increase in cortisol concentration between these two sampling points, both higher than the sample collected at 28 DIM, may indicate higher social stress experienced by heifers as they transitioned from the relatively small fresh herd into the main milking herd. This transition would be characterised by ongoing destabilisations to the established social structure of the herd, each followed by increased agonistic interactions between animals allowing the group to effectively determine hierarchy and dominance relationships (von Keyserlingk et al. 2008). Aggression is usually highest between the first 1–2 h after mixing, although regrouping effects can last up to 15 d post-mixing, and aggression is usually directed towards subordinate animals such as smaller, less experienced heifers (Brakel and Leis 1976; Kondo and Hurnik 1990; Bøe and Færevik 2003; reviewed by Verdon 2023). Integration, and the associated increase in agonistic interactions, has previously been correlated with peaks in faecal cortisol concentrations (Wagner et al. 2012). The third sample in this study, taken approximately 21 d after entry to the main herd, was significantly lower than that taken after approximately 7 d in the main herd. This may reflect a stabilisation of individual heifer position within the herd’s social structure, resulting in fewer received agonistic interactions and thus reduced social stress.

Secondly, the pattern of cortisol concentration peaking and then subsiding observed in the present study may reflect the physiological stress associated with early lactation NEB, which mobilises the hypothalamic-pituitary-adrenal (HPA) axis response similarly to other stressors, triggering the release of cortisol into the body (Bremel and Gangwer 1978; Crofton et al. 2015). The NEB lasts approximately 3–4 weeks until physiological processes stabilise, and condition is regained as the animal metabolises more energy than is required for continued milk production and body condition maintenance once again (Roche et al. 2007; Mellor 2017; reviewed by McNamara et al. 2008). Previous research resulting in similar cortisol patterns indicates that the physiological stress of the NEB is a likely cause of heightened early-lactation concentrations compared to concentrations of samples taken even by the third week of lactation (Shutt and Fell 1985; Fukasawa et al. 2008; Gellrich et al. 2015; Sgorlon et al. 2015). Despite these conclusions, it would be expected that heifers in the present study would remain in a NEB for at least 12–20 d after sampling concluded, associated with continued heightened cortisol concentrations (Roche et al. 2007; Harder et al. 2019). Given the reducing magnitude of negative ADG over weeks in the present study, however, the magnitude of physiological stress may have been less significant at 28 DIM post-parturition compared to 14 DIM; such an effect would support the present results. It is likely that both social and physiological stressors influenced patterns of early lactation milk cortisol concentrations, but it is not possible to disentangle the effects of the two in the present study.

Limitations and experimental design in pastoral dairy research

Designing experiments conducted in real-world dairy systems is an inherently complex process, and the results of the present study must be interpreted with prudence (Tempelman 2009). Low replication and sample sizes in the present study likely account for the few treatment effects found, particularly in terms of production indices. Engstrom et al. (2010) contend that it is difficult to find treatment differences in productivity, particularly milk yield, using a single dairy herd. The authors suggest that 350 cows per treatment may be needed to detect 1 kg differences in milk production (Fleiss 1981; Engstrom et al. 2010; Yandell 2017). This being said, significant effects of feeding treatment interactions on milk production and post-partum ADG have been found using 38 dairy cows (n = 13, 14 and 11 cows per treatment; Stein et al. 2006). Hedlund and Løvlie (2015), meanwhile, found personality effects on behaviour during milking, including indicators of fear, stress and discomfort, using data from 56 milking cows of two breeds. These effects were correlated with lower milk yields during first lactation (Hedlund and Løvlie 2015). Alongside low-level replication and small sample sizes, the effects of early-life treatments in the present study were likely too weak to influence heifer personality and response to the stressors present in early lactation to the point of differences in productivity parameters as identified in these previous studies. Greater replication would better-placed to detect robust longitudinal treatment effects.

Achieving group-level replication in pasture-based dairy systems is significantly more complex than when cows are housed indoors, hence the dearth of welfare research arising from pastoral systems compared to outputs from housed systems. Weather patterns, forage quantity and quality, and individual grazing behaviour are difficult to replicate across experimental units, while the greater space requirements and logistical complications associated with rotationally grazing cattle, and a tendency for block calving in herds managed at pasture are barriers to achieving high rates of group replication (Bransby 1989; Tempelman 2009; Bello et al. 2016). These conditions present challenges to producing replicable and easily interpreted results from pasture-based dairy research (Lean and Lean 2010). Pastoral dairy researchers must evaluate trade-offs between achieving high group-level replication and sufficient statistical power, and what is realistic and feasible in practice (Bello et al. 2016). Compromises in replication, while sometimes necessary, result in claims of pseudoreplication or insufficient statistical power (Oksanen 2001). Tempelman (2009) suggests that statistical robustness requires that at least two pens are used per treatment group in a single-dairy study; while lower than reported in research conducted in housed dairy systems (n = 2 groups/treatment) rearing two groups per treatment in the present study thus provided low-level replication. Further replication would be preferable, and would provide more biologically valid results. It is important to note here that in the present study, variability as a result of group effects within treatment was close to zero for most variables, suggesting that despite this statistical requirement, the potential effect of rearing group on the results obtained was negligible (Tempelman 2009).

Given the aforementioned complexities, it is important to acknowledge that before definitive conclusions can be drawn from novel research such as that described in the present study, similar replicated studies must be repeated with other animal cohorts and in other environmental settings. In this way, future research can elucidate the tangible welfare benefits afforded by potential treatment differences which are biologically valid and can be repeated in varying contexts and across a wider sample of herds. The data collected in this novel exploration into the effects of early-life experiences on first-lactation performance should be carefully interpreted until such a time as similar studies explore this topic with greater replication.

Lastly, in the present study, treatment effects may not have existed, or may have been diminished by the stress associated with early lactation, and particularly the physiological demand on the body. The relationships between stressors faced by lactating cattle, milk cortisol concentrations both chronic and acute, and behavioural indicators of stress and stress-prone individuals require further investigation. Despite little effect of treatment found with low sample size and replication, understanding the experiences of first-lactation heifers during their transition to the milking herd and productive life is important. The present study effectively describes the experiences of a group of pasture-housed dairy heifers during their first month of lactation. The early-life treatments and data collection methodology described in this study may inform future studies able to achieve greater replication, as research expands into this novel area.