1. Introduction
In dairy farm systems cows are exposed to high social and physiological demands, including different potentially stressful situations in the daily routine, such as changing group structures, separation from conspecifics, frequent handling by humans, e.g., during milking, or potentially aversive routine procedures involving restraint, novelty, or noise, e.g., during veterinary procedures [
1] or claw trimming [
2,
3].
When an animal experiences a stressful situation, a number of behavioral and physiological responses can be activated to help the organism to cope with temporal stressors [
4]. Sustained stress, however, has been found to be related to reduced productivity and increased disease susceptibility in cows: investigations by Holtenius et al. [
5], for instance, indicated associations between metabolic stress indicators and incidences of mastitis treatments. More recently, Ivemeyer et al. [
6] found lower physiological stress responses associated with increased mastitis curing rates during lactation.
One central component in the animal’s physiological stress response is the secretion of glucocorticoids (particularly cortisol in cattle) from the hypothalamic–pituitary–adrenal (HPA) axis with a major role in energy regulation. Thus, concentrations of glucocorticoids (GCs) or their metabolites are used as stress indicators and have accordingly been extensively used in animal welfare research (reviewed in [
7,
8]). Besides the traditional method of measuring GC concentrations in blood serum/plasma, also noninvasive matrices are used, e.g., saliva, milk, or urine [
9]. A well-established noninvasive method to assess stress levels over longer periods is to analyze fecal cortisol metabolites (FCMs) by an enzyme immunoassay (EIA), developed by Palme and Möstl [
10]. In contrast to cortisol measures in blood or saliva, FCMs reflect the cortisol secretion over a longer period, with a delay of the gastrointestinal passage rate of 9–15 h in cattle, and thus, are less prone to short-term variations during the day (reviewed in [
7]).
Applied properly, FCMs are considered a powerful tool. Although a number of methodical challenges need to be observed [
7], previous investigations showed that different short-term and prolonged or recurrent stressors can be related to an increase of FCM concentrations: Palme et al. [
11], for instance, demonstrated that a 2 h road transport of cows was followed by a significant increase in FCM concentrations. Möstl et al. [
12] described that after transport into an unfamiliar stable the cortisol excretion via feces of test cows was elevated for about 1 week and declined thereafter to baseline values. In further investigations, FCM measures have been used to detect mid-term stress after claw trimming [
3] and during a 14-day-period of overstocking [
13], as well as chronic or recurrent stress when providing cows inappropriate lying surfaces [
14]. Moreover, FCMs have been found to be associated with the prevalence of hock-lesions [
15], which again could be attributed to chronic or recurrent stress due to suboptimal housing design.
In addition to the aforementioned potential stressors, also human–animal contacts during routine work can be a major source of stress in farm animal species. Previous investigations on the human–animal relationship showed associations between stockpersons’ attitudes and behaviors towards the animals and the animals’ levels of fear of humans, measured e.g., by means of cows’ avoidance distances towards an experimenter [
16,
17]. Indications for associations of the increased fear levels with physiological stress responses are their effects on milk ejection and milk yield [
18,
19,
20], success at first insemination [
21], and aspects of udder health [
6,
22]. Cows’ acute physiological stress responses in the presence of humans have been investigated by Hemsworth et al. [
23], who found lower milk cortisol concentrations in primiparous cows that had been handled positively during calving compared to those without additional human handling. Additionally, Breuer et al. [
24] reported effects of positive and negative handling on dairy heifers’ blood cortisol concentrations in the presence of humans. More recently, Lürzel et al. [
25] found significantly lower salivary cortisol values before and after an isolation test in calves, which had experienced 40 min of gentle interactions (stroking and gentle talking) during the first 4 weeks of life, compared to control calves. A number of studies in pigs indicated chronic stress responses in animals that are fearful towards humans (e.g., [
26,
27]). In cattle, however, mid- or long-term effects of the quality or quantity of human–animal contact have not yet been investigated.
Furthermore, to date no cross-sectional investigations have been conducted to explore the interplay of different potentially influencing factors (
Figure 1) on the cows’ physiological stress levels, including characteristics of herd, housing, management, and human–animal contact.
The present cross-sectional investigation aimed at exploring the mid-term physiological stress level by means of FCMs in lactating cows on different German organic dairy farms. In detail, we aimed to explore possible (1) important influences of management, housing, and human–animal contact on FCM levels, while taking potentially confounding effects of cow lactation status (DIM) and the day time of sampling into account, (2) associations between cows’ fear behaviors towards humans and FCM levels, and (3) associations between FCM levels and milk yield as well as udder health (somatic cell scores, SCS).
4. Discussion
The present investigation aimed at examining associations between fecal cortisol metabolites (FCMs) reflecting the cows’ mid-term physiological stress levels and farm factors, the cows’ behaviors and milk yield as well as udder health. Particular attention was paid to factors of human–animal contact.
4.1. Level and Range of FCM Levels in the Investigated Sample
Since measurements of GC metabolites in fecal samples strongly depend on the methods used, but analytical methods vary between different laboratories, opportunities to compare between investigations are limited. Higher concentration levels found do not necessarily indicate higher adrenocortical activities (i.e., a stronger stress response), but may have their basis in methodical issues [
7]. For example, time periods and temperatures at sample collection and storage have to be considered [
33]. Thus, meaningful comparisons, also within the same species, are only possible, if exactly the same methods were used [
8].
The medians of repeated FCM measures found in the present investigation ranged on a relatively low level between 2.2 and 47.6 ng/g at animal level (median = 11.0 ng/g, mean = 12.2 ng/g) compared to previous investigations using the same method (e.g., [
15,
34,
35]): Rouha-Mülleder et al. [
15] investigated singularly measured FCM levels on 80 dairy farms and found concentrations at farm level ranging from 30 to 157 nmol/kg (median = 77 nmol/kg, corresponding to approx. 23.4 ng/g). In an investigation exploring dairy cow coping capacity during a change from conventional to automatic milking, Weiss et al. [
35] measured average FCM concentrations ± sd of 134 ± 12 ng/g at animal level during the control period. Before dry-off, Bertulat et al. [
34] measured baseline FCM concentrations in 80 late lactating dairy cows ranging from 30.0 to 184.9 ng/g at animal level. The maximum values correspond to some higher values that we found at sample level (max = 159.5 ng/g). Lower FCM values in nonlactating dairy cows in New Zealand, ranging from 6.0 to 8.2 ng/g [
14], resulted from a commercial corticosterone assay, thus values cannot be directly compared.
Due to considerable differences between extraction and assay methods, no threshold or target has been defined for FCM measures in cattle, so far [
7]. However, it appears that the results of the present investigation range on a relatively low level.
4.2. Identified Factors Influencing FCM Measures
The multivariable modelling resulted in a final model including altogether eight factors. As potential confounders, day time of sampling and cow’s DIM were integrated as fixed factors in the minimal model and remained significant (
p < 0.001) in the final model. However, effect sizes of day times were very low (ranging between 0.03 and 0.09) and low for DIM (0.14), suggesting that these factors should be considered, but did not strongly affect the present analyses. This may be explained by only a weak increasing effect of advanced pregnancy [
7] in the sample, since late lactating cows (>200 DIM at the first sampling date) were not included, and part of the cows being in early lactation and not yet pregnant again. Physiological changes and challenges in early lactation may have even counteracted: Fukasawa et al. [
36] found higher cortisol concentrations measured in milk samples of cows in early lactation (7–90 DIM: 0.39 ng/mL) compared to cows in later lactation stages (e.g., 91–180 DIM: 0.22 ng/mL).
The factors included by means of forward selection, referred to different aspects of management, housing, and human–animal contact. Three factors relating to human–animal contact were included in the final model: contact time per cow, active habituation of heifers to milking, and manual provision of concentrates. The contact time per cow describes the quantity of human–animal contact during routine work in min/d. Higher contact times were associated with lower FCM levels, suggesting that prolonged contact to humans can decrease cows’ overall stress levels. One reason for this could be reduced fear towards humans due to an increased quantity of human–animal contact of positive quality as found in earlier investigations [
16,
37,
38]. However, in the present study no substantial association between FCM levels and fear-indicating responses in the avoidance distance (AD) test and regarding the cows’ expressive behaviors during a standardized human–animal interaction (QBA) could be detected. Therefore, reduced cows’ fear responses towards humans appeared to be less crucial here. Further mechanisms, such as earlier detection and solving of e.g., technical or social problems in the herd due to more human presence in the stable, need be considered, too.
Although provision of attractive feed by hand is considered a pleasant human–animal interaction and was shown to be associated with less fearful responses towards humans [
39], manual concentrate provision was associated with increased FCM levels in the present investigation. This might be explained by increased arousal, physical activity [
40], and possibly also social stress during feeding, when the attractive feed is provided.
Active habituation of heifers to the milking parlor or automatic milking system (AMS) reported by the farm managers was associated with increased FCM levels. Similarly, in a previous investigation of the present data on farm level, Ivemeyer et al. [
6] found active habituation of heifers to milking associated with impaired udder health. Possibly, extra habituation efforts reflect a necessity due to problems with nervous heifers rather than additional positive human–animal contact.
From the field of investigated management factors, increased FCM levels were found in herds where diseased or lame cows were separated from the herd. Although for the diseased or lame cows the separation might provide some protection and promote recovery, apparently a disadvantage can be the associated changes of group structure which may increase social stress: Von Keyserlingk et al. [
41] observed that regrouping of dairy cows can disrupt behavior and impair production over days following the event, suggesting an impact of stress due to increased agonistic interactions.
With regard to housing, two factors were included in the final model: housing type and cow:lying space ratio. FCM levels were significantly lower on farms with straw yards compared to farms with raised cubicles. Additionally, Palme et al. [
42] and Fisher et al. [
14] found differences in FCM levels of cows housed in different systems. Fisher et al. [
14] analyzed FCMs from nonlactating cows that had been moved from pasture to four different floor types for four days, respectively. The lowest concentrations were measured when the animals were kept on a deformable floor; the highest concentrations were found when they were kept on concrete floor or on a section of the farm laneway. Accordingly, Palme et al. [
42] showed that cows on straw yards had significantly lower FCM values than those housed in systems with mainly raised cubicles.
In the present sample, associations found regarding cows’ access to resources raise several questions: the significantly lower FCM levels on farms offering a generous cow:lying place ratio compared to farms offering a suboptimal ratio conformed to expectations. However, FCM levels were higher when the animals had access to currently recommended lying place provision compared to suboptimal space. The same pattern was visible in the data regarding the cow:feeding place ratio. This result may call the current recommendations into question where partly even less than one lying or feeding place per cow are foreseen. However, the lower FCM levels under suboptimal conditions might be explained by nonlinear associations: typically, GC levels increase in response to a challenge or stressor, and thus, as a rule high levels can indicate stress. However, low levels can also be attributed to a downregulating response after long-term exposure to stressors. For instance, a recent investigation found lower plasma cortisol and FCM concentrations in domestic horses showing poor welfare in terms of depressive-like behaviors compared to the control animals [
43]. Inconsistent patterns and nonlinear associations were also found in wild animal species with regard to associations with reproduction rate and fitness [
44].
4.3. Associations between FCMs and Cow Behavior
Although the cows’ fear behaviors towards humans ranged markedly on the investigated farms, both measured by AD (0.0–170.0 cm) and QBA (−1.713–2.355), no association between the cows’ FCM levels and AD was found. Additionally, the correlation to QBA was almost negligible, though significant (rs = −0.149, p = 0.008). The direction of this correlation is contrary to expectations, but should not be over-rated because of the low effect size.
Previous investigations that found associations between positive handling treatments and the animals’ physiological responses in the presence of humans used GC measures reflecting the acute stress level, i.e., cortisol concentrations in milk [
23], blood [
24], and in saliva [
25]. Apparently, in animals that are more fearful towards humans, human–animal contacts can result in an immediate release of stress hormones. However, the present results suggest that this is not reflected in the general physiological stress level, with FCM levels reflecting the cortisol secretion over a longer period [
4].
4.4. Associations between FCMs and Milk Yield and Udder Health
Associations between FCM level and milk recording data were investigated to quantify potential effects on cows’ health and milk production. Only a slight, but significant positive correlation between FCM concentrations and somatic cell scores at test days as close as possible to the feces sampling was found in the present investigation (r
s = 0.109,
p = 0.005). A previous multivariable analysis of the present data on farm level regarding effects on udder health showed that lower median herd FCM levels were related to enhanced mastitis cure rates [
6]. The current results reflect only a low general influence of stress on udder infections that relate to immunosuppressive effects of longer lasting increased GC levels.
The daily milk yield (energy corrected milk, ECM), which was on average on a very moderate level in the investigated sample, was not related to the cows’ FCM levels. Previous investigations showed heterogeneous results: Rouha-Mülleder et al. [
15] found FCM levels negatively correlated with the daily milk yield. According to their interpretation, distress might have led to increased amounts of residual milk. Additionally, in a study of Bertulat et al. [
34] regarding stress after sudden dry-off, high-yielding cows had the lowest baseline value of FCM concentrations before drying-off and low yielding cows had the highest concentrations. In contrast, Pesenhofer et al. [
3] and Fukasawa et al. [
36] found no significant correlation between milk yield and FCM basal values. Major influences on milk yield are the animal’s genetic potential [
45,
46], the feeding level [
47,
48,
49,
50], lactation stage [
51], and health [
52,
53] whose joint effects on GC levels are not clear-cut. Therefore, within herd or well controlled investigations of possible associations might produce more meaningful results regarding possible stress effects on performance than cross-sectional designs.
4.5. Limitations of the Study
Using FCM measures in order to evaluate the physiological stress level is a noninvasive alternative that has been validated and repeatedly applied in cattle (reviewed in [
8]). However, concentrations of FCMs measured by means of immunoassays always have to be interpreted as relative values. Furthermore, not every elevated value of FCMs or other measures of glucocorticoid concentrations can be interpreted as distress [
54], since also events such as mating (in cattle: [
55,
56]; in horses: [
57]) or environmental enrichment (in pigs: [
58]) can increase cortisol levels. Consequently, it is not possible to define absolute threshold values. Furthermore, uncertainties still exist regarding the interpretation of low FCM values, as also found in the present investigation. U-shaped associations due to downregulating processes as a consequence of prolonged distress [
44] cannot be ruled out.
Beside the measure used, also the cross-sectional approach implies limitations: on-farm investigations always involve a great number of varying factors and factor combinations so that the associations found, especially in relatively small samples like in the current study, should not be generalized without further external evidence. Moreover, the broad number of potential influencing factors leads to methodical uncertainties regarding their selection within the multivariable modelling. Including or excluding a single factor into or from statistical modelling can result in a different combination of factors included in the final model. In addition, associations with FCM values may be interlinked with a combination of further potential influencing factors that have not been considered, but affect results in an opposite direction.
Thus, cross-sectional studies such as the present investigation can only identify patterns of associations. To prove causal relationships requires specific experimental investigations under reasonably controlled conditions.
5. Conclusions
The present cross-sectional investigation showed that FCM measures reflecting the physiological stress level in dairy cows are not straightforward to interpret. With, in general, relatively low FCM levels, no substantial associations on individual level with fear-indicating behavior towards humans, milk yield, or udder health could be detected. The multivariable analysis of external influences on the cows’ FCM levels on farm level yielded some unexpected results which show that the multifaceted nature of stress physiology is difficult to grasp under on-farm conditions. For instance, farms that separated diseased or lame cows had higher FCM levels, possibly because the associated group changes are disadvantageous. Additionally, farms that fed concentrates by hand and that actively habituated heifers to the milking procedure had higher FCM levels, in the latter case possibly because they had more nervous heifers that needed habituation. In addition, a suboptimal animal to lying place ratio was associated with lower FCM levels than lying place provision according to current recommendations. Here, a downregulating response after long-term exposure to stressors might have played a role. However, conforming to expectations, results indicated that generous resource provision and higher comfort around resting as well as increased human contact time per cow contribute to lower physiological stress levels.