Introduction
Dairy cows in early lactation are susceptible to negative energy balance (NEB) because energy requirements for milk production exceed energy intake from feed (Bauman and Currie 1980). This NEB typically peaks around 2 weeks postpartum and persists for 4 to 6 weeks (Bisinotto et al. 2012; Van Knegsel et al.
2014). During NEB, body fat is mobilized, and non‐esterified fatty acids (NEFAs) are released into the bloodstream. Elevated NEFA promotes hepatic production of ketone bodies (e.g., β‐hydroxybutyrate [BHB]), and cows in NEB exhibit high blood concentrations of NEFA and BHB (Leroy et al.
2017). These alterations in energy metabolism have been reported to impair follicle, oocyte, and embryo quality; delay ovulation; and ultimately lead to reduced conception rates and reproductive disorders (Bisinotto et al. 2012; Leroy et al.
2005). Furthermore, Ospina et al. (2010) reported that elevated postpartum NEFA concentrations are strongly associated with prolonged days open.
Because prolonged NEB increases the risk of metabolic disorders (Ospina et al. 2010), simple and non‐invasive indicators are needed. Recently, milk components, particularly fatty acid (FA) composition, have been reported as potential biomarkers of energy metabolism and NEB in dairy cows (Churakov et al.
2021; Duffield et al. 1997). De novo FA (C4:0 to C14:0) are primarily synthesized in the mammary gland from ruminal acetate and butyrate and decrease during NEB, whereas preformed FA (≥ C18:0) originate from body fat mobilization and dietary fat and increase during NEB (Churakov et al.
2021; Gross et al. 2011).
Milk FA profiles serve as sensitive indicators of NEB and health status (Bach et al. 2019; Fukumori et al. 2021).
Furthermore, the milk fat‐to‐protein ratio (FPR) serves as a widely used simple indicator of NEB. During NEB, milk fat percentage increases while milk protein percentage decreases, making FPR a practical monitoring tool (Jenkins et al. 2015).
Elevated FPR has been associated with increased risk of subclinical ketosis (Jenkins et al. 2015; Jorjong et al. 2015).
Therefore, these indicators are expected to be applicable for assessing metabolic status and predicting reproductive performance in early lactation.
While numerous studies have reported indirect effects of milk FA composition on reproductive performance through NEB and metabolic diseases (Churakov et al. 2021), direct investigations of the relationship between FA composition and reproductive outcomes, such as days open and conception rate, using on‐farm dairy herd improvement (DHI) data from commercial herds remain limited. Understanding how NEB and changes in milk FA composition affect reproductive performance indicators such as days open and conception rate is crucial for on‐farm reproductive management and disease prevention.
DHI testing records on commercial farms represent an effective approach for this purpose. Regular sampling through DHI testing facilitates the collection of individual cow data on milk yield, milk composition, and FA profiles, serving as an important information source for on‐farm health and reproductive management (Parker Gaddis et al. 2016).
Although studies utilizing DHI data for early detection of metabolic status and disease risk are increasing (Mota et al. 2023), the direct relationship between milk FA composition and reproductive performance metrics such as days open and conception rate using on‐farm DHI data from commercial herds remains insufficiently elucidated. Understanding these relationships is important because it could potentially enable farmers to identify cows at higher risk of prolonged days open through routine DHI monitoring, which may provide a means for timely interventions to improve reproductive efficiency.
Furthermore, under heat stress conditions, conception rates are markedly reduced due to thermal damage to oocytes and embryos (Hansen 2019). Heat stress‐induced reductions in feed intake exacerbate NEB, leading to decreased de novo FA and increased preformed FA (Moore et al. 2024).
These changes are closely associated with reproductive function (Leroy et al. 2005), and because heat stress increases within‐farm individual variation in metabolic stress, variations in milk components and FA composition are expected to more clearly reflect differences in reproductive performance.
Therefore, this study aimed to clarify the relationships between milk components and FA composition with days open in early lactation Holstein cows under heat stress and non‐heat stress conditions, and to evaluate their predictive ability for days open. We hypothesized that differences in days open would be reflected in milk components, particularly FA composition, with cows having shorter days open exhibiting higher de novo FA and lower preformed FA compared with cows with longer days open. Furthermore, we expected these differences to be more pronounced under heat stress conditions, as heat stress increases within‐farm individual variation in metabolic stress.
Materials and Methods
Ethics Statement
This study used only existing DHI testing data and did not involve direct handling of live animals; therefore, approval from an animal ethics committee was not required.
Meteorological Data
Meteorological data (mean temperature, maximum temperature, minimum temperature, and relative humidity) for this study were obtained from the Tsu Local Meteorological Observatory of the Japan Meteorological Agency (Tsu City, Mie Prefecture). These data were used to characterize environmental conditions during the study period and to distinguish between non‐heat stress and heat stress periods. Data were obtained from the official website of the Japan Meteorological Agency (https://www.jma.go.jp/jma/kishou/info/coment.html), covering the same period as the DHI data (March 2020 to September 2024).
Dairy Herd Improvement Data
DHI data were provided by dairy farmers enrolled in DHI testing programs (Table 1). This study collected DHI data from 461 multiparous Holstein cows that calved on three farms in Mie Prefecture, Japan, between March 2020 and September 2024. All three farms were located within approximately 15 to 25 km from the Tsu Local Meteorological Observatory, ensuring that the meteorological data adequately represented the environmental conditions experienced by the animals used in this study.
Because some cows contributed data from multiple lactations, the total number of observations was 1155. Data from the heat stress period comprised 258 observations (86 cows at 0 to 30 days in milk (DIM), 83 cows at 31 to 60 DIM, and 89 cows at 61 to 90 DIM), while data from the non‐heat stress period comprised 897 observations (224 cows at 0 to 30 DIM, 331 cows at 31 to 60 DIM, and 342 cows at 61 to 90 DIM).
Farm characteristics and management practices are shown in Table 2. All three farms fed total mixed rations with a forage‐to‐concentrate ratio of 33% to 37% forage and 63 to 67% concentrate (72% to 74% total digestible nutrients on a dry matter (DM) basis, 16% to 16.8% crude protein on a DM basis, and 33% to 35% neutral detergent fiber on a DM basis). None of the three farms supplemented FA throughout the year.
Housing systems consisted of free‐stall barns on two farms and a tie‐stall barn on one farm. Milking was performed two to three times daily.
DHI testing was conducted once monthly, recording milk yield, milk composition (milk fat, milk protein, lactose, milk FA composition, etc.), and reproductive performance (calving date, insemination date, pregnancy diagnosis results, days open). Milk yield was measured at each milking on the monthly DHI test day (two to three times on that day, depending on the farm's milking frequency), and the total daily milk yield was recorded. For milk composition analysis, an automatic sampling device was attached to the milking system, and milk samples were collected from each cow at a constant rate from the beginning to the end of milking.
Approximately 50 mL of the collected sample was used for analysis. The collected milk was immediately cooled in ice water and stored in a refrigerator at 4 °C. Samples without any preservatives were sent to the Milk Inspection Station of Tokai Dairying Federation of Cooperatives (Okazaki, Japan) on the morning of each sampling day.
Milk composition was measured using a Fourier‐transform infrared milk analyzer (CombiFoss 7 DC, FOSS, Hilleroed, Denmark) on the day following receipt. Days open were defined as the number of days from calving to the insemination date at which conception was confirmed by pregnancy diagnosis. Pregnancy diagnosis was performed by veterinarians at approximately 30 to 40 days after insemination using transrectal ultrasonography.
Seasonal Classification
Using the meteorological data obtained, monthly mean temperature, mean humidity, and temperature–humidity index (THI) were calculated. THI was calculated using the following equation: THI = (temperature × 0.8) + [(humidity/100) × (temperature −14.4)] + 46.4 (Zimbelman et al. 2009).
The thermoneutral zone for dairy cattle has been reported to be 4 °C to 24 °C, with heat stress beginning to affect cows when temperatures exceed 25 °C (West 2003; Vermunt and Tranter 2011). Additionally, the THI threshold for the onset of heat stress has been reported as 72 or higher (Habeeb et al. 2018).
Based on these reports, months meeting both criteria of mean temperature ≥ 25 °C and THI ≥ 72 (July to September) were defined as the heat stress period, and all other months were defined as the non‐heat stress period (Table 3).
Statistical Analysis
Statistical analyses were performed using R software (version 4.3.0) within RStudio. Prior to statistical analysis, records judged to be biologically implausible were excluded. Additionally, observations with missing values for any variable were excluded from the analysis, and only complete data were used.
Days open were classified into two groups: ≤ 150 days (shorter) and ≥ 151 days (longer). This threshold was established to distinguish between cows with different reproductive performance relative to the average calving interval of approximately 420 to 430 days for Holstein cows in Japan (Livestock Improvement Association of Japan 2023). For both heat stress and non‐heat stress periods, subject cows were categorized into these two groups based on days open.
Data for milk yield and milk constituents up to 90 DIM were stratified into three periods: 0 to 30 DIM, 31 to 60 DIM, and 61 to 90 DIM. Linear mixed models were fitted using the lmer() function from the lme4 package. The model for milk yield and milk constituents included days open group (shorter vs. longer), DIM period (0 to 30, 31 to 60, 61 to 90 days), and their interactions as fixed effects, and animal nested within farm as random effects.
DIM period was included as a fixed effect to examine its interaction with days open because energy metabolism and milk composition change throughout early lactation. Normality of residuals and homogeneity of variance were examined for each continuous dependent variable after fitting the statistical models using the Shapiro–Wilk and Levene tests, respectively.
For variables where the interaction between days open and DIM was significant, a total of three multiple comparisons were performed to further investigate the interaction effect of days open across different DIM periods. Specifically, these comparisons were conducted between the “150 days or less” and “151 days or more” days open groups within each of the three DIM periods (0 to 30 DIM, 31 to 60 DIM, and 61 to 90 DIM). A Bonferroni correction was applied to adjust for multiple comparisons, with the adjusted significance level set at p < 0.0167 (0.05/3).
For variables where days open were significant, the discriminative accuracy for reproductive performance was evaluated using receiver operating characteristic (ROC) analysis. ROC analysis is a method employed to assess the diagnostic performance of a test by plotting sensitivity against the complement of specificity for various cut‐off points (Hanley and McNeil 1982). The performance of the ROC curve is quantitatively expressed by the area under the curve (AUC).
AUC values are typically interpreted as measures of diagnostic accuracy, with classifications such as excellent (0.9 to 1.0), good (0.8 to 0.9), fair (0.7 to 0.8), poor (0.6 to 0.7), or fail (0.5 to 0.6; Swets 1988). The AUC was calculated using the pROC package in R.
To further examine inter‐farm and within‐farm variability in milk components showing relatively higher discriminative ability in the ROC analysis (milk protein, SNF, and de novo FA [g/100 g milk]), additional linear models were fitted for both non‐heat stress conditions (61 to 90 DIM) and heat stress conditions (31 to 60 DIM). The models included days open group (shorter vs. longer), farm, and their interaction as fixed effects, and were fitted using the lm() function in R software.
Results
Associations of Days Open With Milk Components and Fatty Acid Composition Under Non‐Heat Stress Conditions
The results for milk yield and milk components under non‐heat stress conditions are shown in Figure 1. No significant differences between days open groups were observed for milk yield (Figure 1A) or milk fat percentage (Figure 1B). In contrast, the contents of milk protein (Figure 1C) and solids‐not‐fat (SNF; Figure 1D) were significantly higher in the group with shorter days open (p < 0.01).
No significant differences between days open groups were observed for lactose (Figure 1E) or milk urea nitrogen (MUN; Figure 1F). FPR (Figure 1G) was significantly higher in the group with longer days open (p < 0.05).
The results for milk FA composition under non‐heat stress conditions are shown in Figures 2 and 3. For individual FA composition, C14:0 (g/100 g milk; Figure 2A) was significantly higher in the shorter days open group (p < 0.01). For C14:0 (% of total FA; Figure 2B), the interaction between DIM and days open was significant (p < 0.05), with the shorter days open group showing significantly higher values at 0 to 30 DIM and 31 to 60 DIM.
C16:0 (g/100 g milk; Figure 2C) tended to be higher in the shorter days open group (p = 0.08). No significant differences were observed for C16:0 (% of total FA;
Figure 2D). C18:0 (g/100 g milk; Figure 2E) was higher in the longer days open group (p = 0.05). For C18:0 (% of total FA; Figure 2F), the interaction between DIM and days open was significant (p < 0.05), with the longer days open group showing significantly higher values at 0 to 30 DIM and 31 to 60 DIM.
No significant differences were observed for C18:1 (g/100 g milk; Figure 2G). For C18:1 (% of total FA; Figure 2H), the interaction between DIM and days open was significant (p < 0.05), with the longer days open group showing significantly higher values at 0 to 30 DIM and 31 to 60 DIM. As a group of FA profile, de novo FA were significantly higher in the shorter days open group both as g/100 g milk (Figure 3A) and % of total FA (Figure 3B; p < 0.01).
For mixed FA, the interaction between DIM and days open was significant for both g/100 g milk (Figure 3C) and % of total FA (Figure 3D; p < 0.05), with the shorter days open group showing significantly higher values at 0 to 30 DIM. Preformed FA (g/100 g milk; Figure 3E) were higher in the longer days open group (p < 0.05). For preformed FA (% of total FA; Figure 3F), the interaction between DIM and days open was significant (p < 0.05), with the longer days open group showing significantly higher values at 0 to 30 DIM and 31 to 60 DIM.
Associations of Days Open With Milk Components and Fatty Acid Composition Under Heat Stress Conditions
The results for milk yield and milk components under heat stress conditions are shown in Figure 4. No significant differences between days open groups were observed for milk yield (Figure 4A) or milk fat percentage (Figure 4B). In contrast, the contents of milk protein (Figure 4C; p < 0.05) and SNF (Figure 4D; p < 0.01) were significantly higher in the group with shorter days open, and lactose (Figure 4E; p = 0.05) tended to be higher.
No significant differences between days open groups were observed for MUN (Figure 4F). FPR (Figure 4G) was significantly higher in the group with longer days open (p < 0.05).
The results for milk FA composition under heat stress conditions are shown in Figures 5 and 6. For individual FA composition, C14:0 (g/100 g milk; Figure 5A) showed a significant interaction between DIM and days open (p < 0.01), with the shorter days open group showing significantly higher values at 0 to 30 DIM and 31 to 60 DIM. C14:0 (% of total FA; Figure 5B) was significantly higher in the shorter days open group (p < 0.01).
C16:0 (g/100 g milk; Figure 5C) also showed a significant interaction (p < 0.05), with the shorter days open group tending to show higher values at 0 to 30 DIM. No significant differences were observed for C16:0 (% of total FA; Figure 5D) or C18:0 (g/100 g milk or % of total FA; Figure 5E,F). C18:1 was higher in the longer days open group both as g/100 g milk (Figure 5G; p = 0.05) and % of total FA (Figure 5H; p < 0.01).
As a group of FA profile, de novo FA were significantly higher in the shorter days open group both as g/100 g milk (Figure 6A; p < 0.05) and % of total FA (Figure 6B; p < 0.01). No significant differences were observed for mixed FA (Figure 6C,D). Preformed FA tended to be higher (g/100 g milk; Figure 6E; p < 0.1) and were significantly higher (% of total FA; Figure 6F; p < 0.05) in the longer days open group.
Discriminative Ability of Milk Components and Fatty Acids for Predicting Days Open Using ROC Analysis
ROC analysis was performed to evaluate the discriminative ability of milk components and FA composition for days open groups. Only variables that showed significant effects of days open in the linear mixed models were included in the ROC analysis. Under non‐heat stress conditions, ROC analysis was conducted for milk protein, SNF, FPR, C14:0, C18:0, C18:1, de novo FA, and preformed FA; however, AUC values from these variables were less than 0.65 (Table 4).
In contrast, under heat stress conditions, ROC analysis was conducted for milk protein, SNF, lactose, FPR, C14:0, C18:1, de novo FA, and preformed FA, showing higher discriminative ability than those in non‐heat stress conditions. Particularly at 31 to 60 DIM, SNF (AUC = 0.699), milk protein (AUC = 0.676), de novo FA (AUC = 0.670), and C14:0 (AUC = 0.655) showed AUC values in the “poor” range according to Swets (1988) criteria (Table 5).
Although these indicators showed relatively higher discriminative ability compared with non‐heat stress conditions, their AUC values did not reach the “fair” range (AUC ≥ 0.7). To understand the factors contributing to this limited discriminative ability, variability patterns across farms and within farms were examined for the indicators with relatively higher AUC values (milk protein, SNF, and de novo FA) in the following section.
Inter‐Farm and Within‐Farm Variability in Key Milk Components
The results of inter‐farm and within‐farm variability analysis for milk components, showing relatively higher discriminative ability in the ROC analysis (milk protein, SNF, and de novo FA [g/100 g milk]), are shown in Figures 7 and 8.
Under non‐heat stress conditions (61 to 90 DIM, Figure 7), milk protein was affected only by days open (p < 0.01). SNF was significantly affected by both days open (p < 0.01) and farm (p < 0.01), with no significant interaction (p = 0.56). De novo FA showed significant effects of days open (p < 0.05) and farm (p < 0.01), with a marginal interaction (p = 0.06).
Under heat stress conditions (31 to 60 DIM, Figure 8), all three indicators showed significant effects of days open (milk protein, SNF, and de novo FA: all p < 0.01). Farm effects were significant for milk protein (p < 0.01) and SNF (p < 0.01), but not for de novo FA (p = 0.25). Marginal interactions were observed for milk protein (p = 0.07) and SNF (p = 0.07).
Discussion
Milk fatty acid (FA) profiles measured by Fourier‐transform infrared spectroscopy have become widely available through DHI testing, enabling their use as biomarkers for metabolic status and reproductive performance (Churakov et al. 2021; Woolpert et al. 2016).
This study investigated the relationships between milk components, FA composition, and days open in early lactation Holstein cows under different environmental conditions, and evaluated their predictive ability for reproductive performance using on‐farm DHI data. Linear mixed model analysis revealed that cows with shorter days open consistently exhibited higher milk protein, SNF, de novo FA, and C14:0, while cows with longer days open showed higher FPR and preformed FA, with these differences being more pronounced under heat stress conditions. Furthermore, ROC analysis demonstrated that SNF (AUC = 0.699) and de novo FA (AUC = 0.670) at 31 to 60 DIM showed moderate but limited discriminative ability for days open groups under heat stress conditions, though predictive ability was insufficient under non‐heat stress conditions (all AUC < 0.65).
Under non‐heat stress conditions, the group with shorter days open exhibited higher milk protein and SNF, while the group with longer days open showed higher FPR. Additionally, cows with shorter days open had higher de novo FA and C14:0, whereas cows with longer days open showed higher preformed FA, including C18:0 and C18:1. These results suggest that cows with shorter days open maintained better energy status throughout early lactation.
Milk protein is primarily influenced by dietary protein utilization and is also related to feed intake and energy status, with protein synthesis being promoted in cows with alleviated NEB (Woolpert et al. 2016). Similarly, SNF content is a sensitive indicator of energy status.
Because SNF primarily comprises protein and lactose, and lactose showed no differences between days open groups in this study, the elevated SNF in the group with shorter days open likely reflects the higher milk protein content, indicating better energy balance (Churakov et al. 2021). In contrast, FPR was significantly higher in the group with longer days open.
FPR serves as a practical indicator of energy balance status because it reflects the balance between fat mobilization (increasing milk fat percentage during NEB) and protein synthesis (decreasing milk protein percentage during NEB), and has been associated with increased risk of subclinical ketosis (Jenkins et al. 2015; Jorjong et al. 2015).
In this study, the elevated FPR in the group with longer days open indicates that these cows experienced more severe NEB, which is closely associated with impaired reproductive function.
De novo FA, including C14:0, was higher in the group with shorter days open. Churakov et al. (2021) reported that de novo FA decreases during NEB, while preformed FA increases with body fat mobilization.
In cows with alleviated NEB, elevations in blood NEFA and BHB are suppressed, and negative effects on follicular development and embryogenesis are mitigated (Butler 2003; Leroy et al. 2017). Interestingly, mixed FA (C16:0 and C16:1, predominantly C16:0) showed significantly higher values in the group with shorter days open 0 to 30 DIM, but these differences disappeared after 31 DIM.
Mixed FA have intermediate characteristics because they originate from both de novo synthesis and body fat mobilization (Gross et al. 2011). During the early postpartum period, body fat mobilization is most active, representing a transitional phase in which the origin of C16:0 shifts from de novo synthesis to mobilized adipose tissue (Palmquist et al.
1993). After 31 to 60 DIM, NEB is alleviated, and fat mobilization decreases, which likely explains the disappearance of differences between groups. These physiological changes are reflected in mixed FA patterns in milk during the early stage of lactation.
Conversely, higher preformed FA (C18:0, C18:1) in the group with longer days open suggests prolonged NEB and sustained mobilization of body reserves, which are closely associated with reproductive disorders (Butler 2003; Leroy et al. 2017).
In this study, milk yield and milk fat percentage were not significantly associated with days open, whereas milk FA composition showed clear differences between groups. This suggests that qualitative changes in milk components, particularly FA profiles, are more sensitive indicators of reproductive performance than quantitative measures such as total milk fat percentage. SNF, FPR, and FA composition may therefore serve as valuable biomarkers for assessing energy status and predicting reproductive outcomes (Fukumori et al.
2021).
Under heat stress conditions, milk components and FA composition showed enhanced discriminative ability for days open groups compared with non‐heat stress conditions. The group with shorter days open exhibited higher milk protein, SNF, and lactose, while the group with longer days open showed higher FPR. Additionally, cows with shorter days open had markedly higher de novo FA and C14:0, whereas cows with longer days open showed higher preformed FA, particularly C18:1.
Heat stress exacerbates NEB due to decreased feed intake and metabolic changes (Wheelock et al. 2010; Bagath et al. 2019), and worsening NEB induces changes in milk components that are reflected in reproductive performance (Baumgard and Rhoads 2013; Fukumori et al.
2021). Particularly, the group with shorter days open showed significantly higher C14:0 values at 0 to 30 DIM and 31 to 60 DIM. Because C14:0 is synthesized in the mammary gland from ruminal fermentation products (acetate and butyrate), higher C14:0 in cows with shorter days open suggests that maintaining adequate energy intake and ruminal fermentation even under heat stress is associated with better reproductive performance.
Conversely, preformed FA, particularly C18:1, was significantly higher in the group with longer days open under heat stress conditions (both as g/100 g milk and % of total FA), indicating prolonged body fat mobilization and sustained NEB, which negatively affect reproductive function (Butler 2003; Leroy et al. 2017).
In contrast to de novo and preformed FA, the interaction effect for mixed FA observed under non‐heat stress conditions was not evident under heat stress conditions. This likely reflects that body fat mobilization occurs more persistently under heat stress (Roche et al. 2009), obscuring the characteristic transition in C16:0 origin from de novo synthesis to mobilized adipose tissue during the early postpartum period.
Because heat stress increases within‐farm individual variation in metabolic stress, variations in de novo FA and C14:0 more sensitively reflect differences in energy metabolism status under these conditions (Wheelock et al. 2010; Moore et al. 2024).
This explains why milk components and FA composition showed higher discriminative ability for days open groups under heat stress compared with non‐heat stress conditions.
ROC analysis revealed enhanced discriminative ability for days open groups under heat stress conditions compared with non‐heat stress conditions. Under non‐heat stress conditions, AUC values for all tested variables were less than 0.65 across all DIM periods (Table 3), indicating limited discriminative ability. In contrast, under heat stress conditions, particularly at 31 to 60 DIM, SNF, milk protein, de novo FA, and C14:0 showed relatively higher discriminative ability, with AUC values in the “poor” range according to Swets (1988) criteria (Table 4).
Although these values do not reach the “fair” range (AUC ≥ 0.7), they suggest that these indicators may provide supplementary information for identifying cows at higher risk of prolonged days open, particularly when used in combination.
As shown in Figures 7 and 8, the variability in milk components showing relatively higher discriminative ability in the ROC analysis (milk protein, SNF, and de novo FA [g/100 g milk]) was attributed to both inter‐farm differences in management practices and within‐farm individual cow variation. Under non‐heat stress conditions, there was considerable overlap in distributions between days open groups, with many individuals showing similar values. In contrast, under heat stress conditions, the separation between days open groups was more distinct, particularly for SNF and de novo FA.
This difference suggests that heat stress amplifies inter‐farm and within‐farm individual variation in metabolic stress responses (Wheelock et al. 2010), causing variations in milk components and FA composition to more clearly reflect differences in reproductive performance. As observed in the violin plots, substantial inter‐farm differences in median values were evident, particularly under heat stress conditions, where farm‐level patterns contributed to the separation between days open groups.
However, within‐farm individual variation also remained considerable, with some cows in the longer days open group showing values comparable to those in the shorter days open group. These findings indicate that both inter‐farm differences in management practices and within‐farm individual variation in metabolic status and heat stress adaptation contribute to the limited predictive ability. Considering this inter‐farm variability, accurate prediction of reproductive performance using a single indicator alone remains challenging, and incorporating farm‐level management information could potentially improve discriminative performance.
Future improvements in predictive accuracy could be achieved through two complementary approaches: first, by integrating farm‐level management information to account for inter‐farm variability; and second, by combining multiple biomarkers to more comprehensively capture individual metabolic status (Gross et al. 2011). These findings demonstrate the practical utility of milk component monitoring using DHI data and have the potential to contribute to on‐farm individual management, reproductive strategy planning, and achievement of sustainable dairy farming.
This study has several limitations. First, because DHI testing was conducted once monthly, day‐to‐day variations in early lactation may not have been adequately captured (Yamazaki et al. 2024).
Second, the heat stress period examined in this study was limited to 3 months (July to September). The impact of heat stress may vary by region and year, and longer‐term data collection and validation under different climatic conditions are necessary to confirm the generalizability of these findings. Third, this study was based on data from three farms and could not completely control for differences in management practices and genetic backgrounds.
In the future, the collection of detailed monitoring data, including more frequent measurements and the accumulation of large‐scale data considering diverse management environments and genetic backgrounds, will be necessary.
Funding
The authors have nothing to report.
Conflicts of Interest
The authors declare no conflicts of interest.