Prepartum biochemical, ruminal, and behavioral indicators associated with postpartum fat-to-protein ratio in dairy cows

The milk fat-to-protein ratio (F/P) is widely used as an indicator of metabolic status in dairy cows during early lactation. Identifying prepartum factors associated with subsequent F/P ratio variation may improve the early detection of cows with differing metabolic adaptation patterns. Therefore, the aim of this study was to evaluate associations between prepartum biochemical, ruminal, and behavioral indicators and postpartum F/P ratio classification in dairy cows. Sixty-nine clinically healthy primiparous Holstein cows were monitored during the four weeks preceding calving and the first four weeks of lactation. Cows were classified according to their average postpartum F/P ratio as Group 1 (F/P ≤ 1.4; n = 35) or Group 2 (F/P > 1.4; n = 34). Blood biochemical analyses, intraruminal sensor measurements, and behavioral data were collected and compared between groups. Logistic regression and receiver operating characteristic (ROC) analyses were used to identify predictors of postpartum F/P ratio classification. Cows in Group 1 produced 25.4% more milk (p = 0.004),whereas Group 2 exhibited greater milk fat concentrations (p = 0.004). Before calving, ALT activity was 22% higher (p < 0.001), serum iron concentration was 24% higher (p = 0.007), and creatinine concentration was 9% lower (p = 0.042) in Group 1. Behavioral and ruminal measurements also differed between groups. Group 1 cows spent 11% more time ruminating (p = 0.006) and had 3% higher rumen pH values (p = 0.002), whereas Group 2 cows demonstrated 30% greater activity (p < 0.001) and 14% greater water intake (p = 0.035). Logistic regression identified creatinine, glucose, activity, water intake, and rumen pH as significant variables associated with postpartum F/P ratio classification. Among individual predictors, activity demonstrated the highest discriminatory ability (AUC = 0.689; p = 0.007). Significant differences in biochemical, ruminal, and behavioral characteristics were detectable before calving in cows subsequently classified according to postpartum F/P ratio. These findings indicate that sensor-derived and biochemical indicators may provide useful information for characterizing metabolic adaptation during the transition period and support their potential application in precision dairy herd management.

The transition period, defined as the interval extending from approximately three weeks before calving to three weeks after parturition, represents one of the most physiologically challenging stages in the productive life of a dairy cow (Mezzetti et al., 2021). During this period, animals undergo profound metabolic, endocrine, and behavioral adaptations to support the onset of lactation while maintaining homeostasis (Bruckmaier and Gross, 2017). The rapid increase in nutrient and energy requirements associated with milk production frequently exceeds dietary energy intake, resulting in negative energy balance and increased mobilization of body reserves (Martens, 2023). Consequently, transition cows are particularly susceptible to metabolic disorders that compromise health, welfare, and productivity (Kang et al., 2025).

Among the indicators commonly used to evaluate metabolic status in dairy cows, the milk fat-to-protein ratio (F/P) has received considerable attention (Arlauskaitė et al., 2025). Because milk fat concentration tends to increase during enhanced lipid mobilization, whereas milk protein concentration may decrease under conditions of inadequate energy supply, the F/P ratio provides a practical and non-invasive indicator of energy metabolism (Mao et al., 2026). Elevated F/P ratio values during early lactation have been associated with excessive body fat mobilization, subclinical ketosis, impaired production performance, and an increased risk of health disorders. As a result, F/P ratio is widely used in dairy herd management as an indirect marker of metabolic imbalance during the postpartum period (Arlauskaitė et al., 2025).

Although F/P ratio is routinely measured after calving, the physiological processes leading to alterations in milk composition begin before parturition (Schcolnik, 2016). Identifying prepartum indicators associated with subsequent F/P ratio status could improve the early recognition of cows at risk of metabolic disturbances and facilitate timely management interventions (Emam et al., 2024). Blood biochemical parameters, including markers of liver function, mineral status, energy metabolism, and protein turnover, have been widely investigated as potential indicators of metabolic adaptation during the transition period (Goselink et al., 2020). Likewise, behavioral and physiological traits such as rumination activity, water intake, physical activity, rumen temperature, and rumen pH provide valuable information regarding the health and metabolic status of dairy cows (Grigė et al., 2026).

Recent advances in precision livestock farming have enabled continuous monitoring of animals through automated sensors and biosensor technologies (Neethirajan, 2023). Intraruminal boluses, automated milk analysis systems, and digital behavioral monitoring platforms provide high-frequency data that can reveal subtle physiological changes before the appearance of clinical signs (Brassel et al., 2019). These technologies offer new opportunities to identify early predictors of metabolic imbalance and improve dairy herd management (Sani et al., 2026). Despite growing interest in sensor-based monitoring, limited information is available regarding the relationship between prepartum biochemical, ruminal, and behavioral characteristics and postpartum F/P ratio classification. A better understanding of these associations could contribute to the development of predictive tools for identifying cows at increased risk of unfavorable metabolic adaptation after calving (Arlauskaitė et al., 2025). Therefore, the aim of this study was to evaluate prepartum blood biochemical, ruminal, and behavioral indicators in dairy cows subsequently classified according to their postpartum milk fat-to-protein ratio and to identify variables associated with an increased likelihood of developing elevated F/P ratio values during early lactation.

Animals, housing, and feeding management

The study was conducted in accordance with the ethical standards for animal experimentation approved by the State Food and Veterinary Service of the Republic of Lithuania (Approval No. 135834789). All procedures complied with national regulations governing animal welfare and institutional research requirements. The study took place at the Large Animal Clinic and Practical Training and Research Centre of the Lithuanian University of Health Sciences in central Lithuania from 29 April 2024 to 22 January 2026. The study was conducted in a commercial Holstein dairy herd comprising approximately 220 lactating cows. A total of 69 clinically healthy primiparous cows meeting the predefined inclusion criteria were prospectively enrolled for detailed physiological monitoring. Mean body weight of the monitored cows was 550 ± 45 kg. Each enrolled cow was followed from approximately four weeks before the expected calving date through the first four weeks after parturition. Animals were maintained in a ventilated free-stall housing system with unrestricted access to water. Feeding was based on a total mixed ration formulated according to production stage and nutritional requirements. The diet was formulated for a 550 kg Holstein cow producing approximately 35 kg of milk daily, providing 1.60 Mcal/kg net energy for lactation. Approximately 35 days before the expected calving date, dry cows were transitioned to the same total mixed ration as lactating cows in order to facilitate dietary adaptation during the transition period. Detailed feeding ration is provided in Table 1. Feed was distributed twice daily at 09:00 and 16:00 h. Milking was performed using automatic robotic milking units (DeLaval Inc., Tumba, Sweden). = The average milking frequency during the study period was 2.5-3.0 milkings per cow per day. The herd’s average annual energy-corrected milk yield in 2024 was 10,304 kg per cow, standardized to 4.1% milk fat and 3.4% protein. During the entire experimental period, animal disturbance and unnecessary handling were minimized to maintain optimal welfare conditions.

Measurements and sample collection

Groups creation

This prospective observational cohort study followed clinically healthy Holstein dairy cows during early lactation to evaluate the association between milk fat-to-protein ratio and physiological, biochemical, and behavioral parameters. Inclusion in the final dataset required the availability of complete records from all monitoring systems, including in-line milk analysis after calving, intraruminal sensor measurements, and serial blood sampling. Animals that developed any clinical disorders during the study, such as mastitis, lameness, metritis, displaced abomasum, or gastrointestinal disturbances, were excluded, as were cows with missing or incomplete records. Of the 73 animals initially enrolled, four were removed from the analysis because of incomplete data collection. The remaining 69 heifers remained free of clinical disease throughout the study.

All measurements were obtained once per week and performed at approximately the same time on each sampling day. Clinical examinations were conducted by the single licensed veterinarian with more than five years of clinical experience in dairy cattle medicine between 08:00 and 10:00 h to ensure consistency of assessment throughout the study period.

After completion of the first four weeks of lactation, the mean milk fat-to-protein (F/P) ratio was calculated for each cow using all milk composition measurements obtained during this period. Cows were then classified into two groups according to their mean F/P ratio: Group 1 (F/P ≤ 1.4; n = 34) and Group 2 (F/P > 1.4; n = 35). The value of 1.4 was selected based on previous studies identifying this threshold as a biologically relevant indicator of metabolic status and an increased risk of negative energy balance during early lactation (Arlauskaitė et al., 2025). This classification was subsequently used to compare prepartum physiological, behavioral, and biochemical characteristics between the two groups. The F/P ratio was selected because it is a routinely available, non-invasive indicator widely used in dairy herd management for monitoring metabolic adaptation during early lactation.

Blood biochemistry

Blood sampling was performed once weekly over eight consecutive weeks. To reduce circadian and management-related variation, samples were collected on the same weekday at 08:00 h before morning feeding. Approximately 10 mL of blood was withdrawn from the coccygeal vein into evacuated serum collection tubes (BD Vacutainer, Crawley, UK). Immediately after sampling, the material was transported to the Laboratory of Clinical Tests at the Large Animal Clinic, Veterinary Academy, Lithuanian University of Health Sciences. Serum biochemical analyses were conducted using a Hitachi 705 automated biochemical analyzer (Hitachi, Tokyo, Japan) together with DiaSys diagnostic reagents (Diagnostic Systems GmbH, Düsseldorf, Germany). Concentrations of non-esterified fatty acids (NEFA) were analyzed separately using an Rx Daytona automated wet chemistry analyzer (Randox Laboratories Ltd., London, UK). All analytical equipment underwent routine calibration and internal quality-control procedures according to manufacturer recommendations. Multipoint calibration standards supplied by DiaSys were used for calibration of the Hitachi analyzer with reagent-specific correction factors. The following biochemical variables were determined: albumin (ALB), alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyltransferase (GGT), calcium (Ca), creatinine (CREA), iron (Fe), glucose (GLUC), magnesium (Mg), non-esterified fatty acids (NEFA), phosphorus (PHOS), total protein (TP), triglycerides (TRIG), and urea (UREA).

Milk yield and composition recording

Milk yield and milk composition, including fat, protein, lactose, and the fat-to-protein ratio (F/P), were recorded using an in-line milk analysis system (Brolis Sensor Technology, Vilnius, Lithuania) integrated into an automatic robotic milking system. Cows were milked voluntarily using the automatic milking system, with an average milking frequency of 2.5–3.0 milkings per cow per day, depending on individual voluntary visits to the milking robot. Milk fat and protein concentrations were measured during every milking, with data acquired continuously every 5 s throughout the milking process. For each milking event, milk composition values were calculated as flow-weighted averages of fat and protein concentrations. Daily milk composition values were obtained by averaging all milkings performed within the same day. The primary study outcome was defined as the mean postpartum F/P ratio calculated from the daily values recorded during the first four weeks after calving. This four week mean F/P ratio was subsequently used to classify cows into the study groups. For comparisons with blood biochemical variables, milk composition values obtained on the same sampling day as blood collection were used to ensure temporal synchronization between physiological and milk production measurements.

Monitoring of rumen function and behavioral activity

Reticulorumen temperature, rumen pH, rumination time, physical activity, and water intake were continuously monitored using intraruminal SmaXtec sensor boluses (SmaXtec animal care technology ®, Graz, Austria). These devices enable long-term automated monitoring of physiological and behavioral indicators associated with dairy cow health and metabolic status. Each animal received one bolus approximately five weeks before the expected calving date to allow sufficient adaptation and stabilization of sensor recordings. Boluses were administered orally by the same trained veterinarian using a dedicated applicator according to the manufacturer’s protocol. The devices are gravity-balanced and designed for permanent retention within the reticulum. Prior to administration, each sensor was activated, linked to the corresponding ear-tag identification number, and integrated into the central data-management system. During placement, cows were restrained in self-locking head gates while the head was stabilized to ensure correct positioning of the bolus at the base of the tongue. Animals were subsequently monitored for two hours for possible adverse reactions. Recorded data were transmitted wirelessly through receiving antennas to the SmaXtec system, where signals were processed electronically, digitized, and stored in the central database. Data collection and visualization were performed using SmaXtec Messenger ® software (version 4). Continuous measurements of reticulorumen temperature, rumination activity, physical activity, and water intake were obtained throughout the monitoring period. Table 2 presents the physiological, behavioral, and ruminal parameters monitored throughout the experimental period.

Statistical analysis

All statistical analyses were conducted using IBM SPSS Statistics (version 31).

Descriptive statistics were generated for all blood biochemical, rumen function and behavioral variables. Prior to modeling, the dataset was filtered to include only pre-partum observations for all analyses aimed at identifying physiological predictors of post-calving F/P ratio classification.

To determine which pre-partum indicators were most strongly associated with post-calving F/P ratio category, a binary logistic regression model was fitted using a forward stepwise likelihood-ratio procedure (Forward: LR). All pre-partum variables—including blood metabolites and minerals, liver enzymes, rumen temperature, rumen pH, rumination time, physical activity and water intake—were entered as candidate predictors. Multicollinearity was evaluated using variance inflation factors (VIF), and all retained predictors exhibited acceptable independence (VIF < 3). Odds ratios (Exp(B)) and 95% confidence intervals were calculated for each retained variable. Model adequacy was evaluated using the Hosmer–Lemeshow goodness-of-fit test and Nagelkerke R². Statistical significance was accepted at p < 0.05.

To analyze physiological dynamics before calving, weekly changes in biochemical, mineral, metabolic and behavioral indicators during the four-week prepartum period (Weeks 1–4) were evaluated using linear mixed-effects models (LMM). Week was included as a fixed effect, and cow ID as the subject variable with repeated measurements across weeks. An autoregressive covariance structure of order 1 [AR(1)] was selected based on the lowest AIC and BIC values relative to alternative structures (CS, UN). Model parameters were estimated with restricted maximum likelihood (REML), and Type III tests of fixed effects were used to assess the influence of Week. When the Week effect was significant, Bonferroni-adjusted pairwise comparisons of estimated marginal means were conducted.

Postpartum repeated measurements (Weeks 5–8, corresponding to the first four weeks of lactation) were analyzed using linear mixed-effects models within the same framework applied to the prepartum data. For each dependent variable, the model included F/P ratio group (Group 1 vs. Group 2) as a between-subjects fixed effect, Week as a within-subjects fixed effect, and the Group × Week interaction; cow identity was specified as the subject (random effect) to account for the correlation among repeated measurements obtained from the same animal. Because F/P ratio group is a fixed attribute assigned to each cow, the postpartum Group × Week design was fully balanced (seven cows per group in every week, with no empty cells). The covariance structure was selected on the basis of the lowest AIC and BIC values among first-order autoregressive [AR(1)], compound symmetry (CS), and unstructured (UN) alternatives. Model parameters were estimated by restricted maximum likelihood (REML), and Type III tests of fixed effects were used to assess the influence of Group, Week, and their interaction. When a significant Week or Group × Week effect was detected, Bonferroni-adjusted pairwise comparisons of estimated marginal means were performed.

Receiver operating characteristic (ROC) analysis was performed to assess the discriminatory ability of individual prepartum biomarkers (Weeks 1–4) in relation to the dichotomous post-calving F/P ratio classification. ROC curves were generated for each variable (creatinine, glucose, activity, water intake, and rumen pH), and the area under the curve (AUC) with standard errors, asymptotic p-values, and 95% confidence intervals was calculated using non-parametric estimation. ROC analysis was used to quantify the stand-alone predictive performance of each biomarker and to complement the multivariable logistic regression model.

In the ROC analyses, each cow contributed a single prepartum value per biomarker, calculated as the mean of its four weekly prepartum measurements; the analyses were therefore performed at the animal level (one observation per cow), avoiding pseudoreplication of the repeated measurements. To make fuller use of the repeated prepartum measurements and to identify the most informative temporal representation of each biomarker, several complementary strategies were compared. First, ROC analyses were repeated separately for each prepartum week (Weeks 1–4) to determine the sampling time point with the greatest discriminative ability. Second, five summary measures of each cow’s weekly trajectory were derived per biomarker — the mean concentration, the temporal slope (linear regression of concentration on week), the minimum, the maximum, and the area under the concentration–time curve (trapezoidal rule) — and the discriminative ability of each was evaluated by ROC analysis. Areas under the curve (AUC) with 95% confidence intervals were estimated using the DeLong method, and the AUCs of the alternative summaries were compared with that of the mean concentration using the DeLong test for correlated ROC curves. Finally, a sensitivity analysis was conducted to establish the effect size detectable given the available sample size.

To assess whether pre-partum metabolic profiles differed between cows later categorized into divergent post-calving F/P ratio groups, we compared a comprehensive panel of biochemical markers between Group 1 and Group 2.

Milk production differed between cows classified according to their average postpartum F/P ratio (Figure 1). Cows in Group 1 (F/P ≤ 1.4) produced 25.4% more milk than cows in Group 2 (F/P > 1.4) (p = 0.004.

Several prepartum blood biochemical parameters differed between cows subsequently classified according to their postpartum FPR (Figure 2). ALT activity was approximately 22% higher in Group 1 than in Group 2 (p < 0.001). AST activity was also higher in Group 1 by approximately 10%; however, this difference was not statistically significant (p = 0.076).

Iron concentration was 24% higher in Group 1 compared with Group 2 (p = 0.007). In contrast, CREA concentration was approximately 9% lower in Group 1 (p = 0.042). TP concentration was numerically lower in Group 1 by approximately 4%, although the difference was not statistically significant (p = 0.095).

No significant differences were observed between groups for GLUC (p = 0.217), NEFA; (p = 0.338), P (p = 0.701), TRIG (p = 0.221), UREA (p > 0.45), or GGT (p = 0.558). Likewise, ALB, Ca, and Mg concentrations did not differ significantly between groups (p > 0.30).

Several prepartum rumen and behavioral parameters differed between cows subsequently classified according to their postpartum F/P ratio (Figure 3). Rumen temperature was lower in Group 1 than in Group 2, with a difference of approximately 0.5% (p = 0.005). Rumination time was approximately 11% greater in Group 1 compared with Group 2 (p = 0.006).

Activity level differed markedly between groups, with cows in Group 1 exhibiting approximately 30% lower activity values than cows in Group 2 (p < 0.001). Water intake was also lower in Group 1 by approximately 14% (p = 0.035). In contrast, rumen pH was approximately 3% higher in Group 1 than in Group 2 (p = 0.002).

Analysis of the postpartum period using linear mixed-effects models confirmed the differences identified by group comparison. A significant main effect of group was observed for milk fat-to-protein ratio, which was higher in Group 2 than in Group 1 (p = 0.001), for milk yield, which was higher in Group 1 (p = 0.043), for milk fat concentration, which was higher in Group 2 (p = 0.026), and for physical activity, which was higher in Group 2 (p = 0.007). A significant main effect of week reflected the expected temporal changes in milk yield and composition across early lactation. Group × Week interactions were non-significant for the majority of variables, indicating that between-group differences were broadly consistent across the first four weeks of lactation.

Analysis of the postpartum period using linear mixed-effects models confirmed the differences identified by group comparison. A significant main effect of group was observed for milk F/P ratio ratio, which was higher in Group 2 than in Group 1 (p = 0.001), for milk yield, which was higher in Group 1 (p = 0.043), for milk fat concentration, which was higher in Group 2 (p = 0.026), and for physical activity, which was higher in Group 2 (p = 0.007). A significant main effect of week reflected the expected temporal changes in milk yield and composition across early lactation. Group × Week interactions were non-significant for the majority of variables, indicating that between-group differences were broadly consistent across the first four weeks of lactation.

To identify prepartum variables associated with postpartum F/P ratio classification, a forward stepwise logistic regression analysis was performed (Table 2). Of all evaluated blood biochemical, rumen, and behavioral variables, five predictors were retained in the final model: CREA, GLUC, activity level, water intake, and rumen pH.

In contrast, rumination time was markedly higher in the low F/P group, with an increase of approximately 11% compared with cows that later maintained higher F/P values (p = 0.006). This suggests that animals predisposed to a lower postpartum F/P ratio spent more time ruminating before calving.

Physical activity, expressed as activity units (0–100 scale), was significantly lower by around 30% in cows that later developed a high F/P ratio (p < 0.001). This reduction indicates markedly decreased pre-partum behavioral activity in this group.

Daily water intake was also significantly lower in the group 2, showing a reduction of approximately 14% (p = 0.035). Reduced water consumption prior to calving may reflect early physiological shifts associated with energy or hydration status.

Conversely, rumen pH was significantly higher in the low F/P group, with an increase of approximately 3% (p = 0.002).

A forward stepwise logistic regression analysis was performed to identify the most informative pre-partum predictors associated with an increased likelihood of belonging to the high post-calving F/P ratio group. Out of the full panel of blood biochemistry, rumen function, and behavioral indicators, only five variables remained in the final model: CREA, GLUC, activity units, water intake, and rumen pH (Table 3). These parameters collectively provided the best discriminatory power for classifying cows into their respective post-calving F/P ratio groups.

Higher pre-partum CREA concentrations were associated with an increased probability of belonging to the group 2 (OR = 1.06, p = 0.008), suggesting that altered muscle metabolism or renal filtration dynamics before calving may contribute to later metabolic imbalance. GLUC emerged as a strong positive predictor (OR = 79.74, p = 0.014), indicating that cows with higher circulating glucose prior to calving exhibited substantially greater odds of developing elevated postpartum F/P ratios. Although the confidence interval was wide, the effect direction remained consistent and statistically significant.

Behavioral variables also contributed meaningfully to prediction. Increased activity was associated with higher odds of high F/P ratio classification (OR = 1.87, p = 0.005), suggesting that cows displaying heightened pre-partum behavioral arousal may experience underlying metabolic stress. Similarly, increased water intake before calving was positively associated with high F/P ratio outcomes (OR = 1.06, p = 0.008), possibly reflecting shifts in rumen osmolality or feed intake patterns.

In contrast, rumen pH demonstrated a strong protective effect. Lower pre-partum pH had the likelihood of belonging to the high-F/P group (OR = 0.000097, p = 0.011).

To evaluate the predictive performance of individual prepartum biomarkers for identifying cows at risk of developing a high post-calving F/P ratio, receiver operating characteristic (ROC) curve analysis was performed for variables measured during the four weeks preceding calving (Figure 4). Activity showed the highest discrimination (AUC = 0.689), whereas CREA, GLUC, water intake, and rumen pH demonstrated weak or non-significant predictive value.

ROC analysis was performed to evaluate the discriminatory capacity of individual prepartum biomarkers (Weeks 1–4) for predicting classification into the high post-calving F/P ratio group (Table 4).

Among all tested biomarkers, prepartum activity exhibited the strongest predictive performance, achieving an AUC of 0.689 (95% CI: 0.551–0.828; p = 0.007). This reflects moderate and statistically significant discrimination, suggesting that lower prepartum behavioral activity is associated with increased risk of developing a high F/P ratio after calving.

Creatinine showed weak but non-significant discriminatory ability (AUC = 0.623; 95% CI: 0.475–0.771; p = 0.103), indicating limited standalone predictive value. Glucose did not demonstrate meaningful discrimination between F/P groups (AUC = 0.563; 95% CI: 0.410–0.715; p = 0.421).

Water intake showed weak and non-significant discriminatory ability (AUC = 0.627; p = 0.096), while rumen pH demonstrated no reliable discriminative capacity (AUC = 0.411; p = 0.255). Higher pre-partum rumen pH was associated with a reduced likelihood of belonging to Group 2 (OR = 0.000097, p = 0.011). Temporal changes in prepartum physiological variables are presented in Supplementary Tables S3 and S4. These analyses were intended to characterize normal physiological adaptation during late gestation and, because they were descriptive rather than predictive, are not discussed further in relation to postpartum F/P classification.

To exploit the longitudinal structure of the data, the discriminative ability of the prepartum biomarkers was examined across sampling weeks and across alternative trajectory summaries (Supplementary Tables S1, S2). ROC analyses performed separately for each prepartum week showed that discrimination was broadly stable across Weeks 1–4, with no single week providing a consistent, marked advantage. When each biomarker’s weekly trajectory was summarized per cow, the mean concentration performed at least as well as the temporal slope, minimum, maximum, and area under the concentration–time curve; the DeLong test indicated no significant improvement of any alternative summary over the mean for the principal predictors (all p > 0.05). These results indicate that the mean prepartum concentration, used in the primary analysis, captures the discriminative information available in the repeated measurements as effectively as more complex trajectory-derived measures. With approximately 34 and 35 cows per group, the study had 80% power at a two-sided α of 0.05 to detect a standardized mean difference of d ≈ 0.68, corresponding to an AUC of approximately 0.68; individual biomarkers with true discriminative ability below this threshold would therefore be underpowered, and single-marker AUCs close to this value should be interpreted with corresponding caution.

The present prospective cohort study demonstrated that clinically healthy primiparous dairy cows subsequently classified according to their postpartum milk F/P already exhibited measurable differences in biochemical, ruminal, and behavioral characteristics before calving. These findings suggest that variation in postpartum F/P ratio is preceded by differences in physiological adaptation during late gestation, supporting the concept that metabolic trajectories associated with early lactation are established before parturition rather than arising exclusively after calving.

Postpartum milk production and composition differed between the two F/P groups. Cows with a lower mean F/P ratio (≤ 1.4) produced more milk and had higher lactose concentrations, whereas cows with a higher F/P ratio (> 1.4) exhibited greater milk fat concentrations. This pattern is consistent with the physiological role of lactose as the primary osmotic regulator of milk secretion, where increased lactose synthesis promotes greater milk volume (Girdauskaitė et al., 2025). Conversely, elevated milk fat concentrations contribute directly to an increased F/P ratio and are generally considered to reflect greater mobilization of body fat reserves during early lactation. Consequently, the observed differences in milk composition support the biological relevance of postpartum F/P ratio as an indicator of metabolic adaptation during the transition period (Schcolnik, 2016; Mutsvangwa and Nwakalor, 2023).

Importantly, cows that subsequently developed different postpartum F/P ratios already displayed distinct biochemical profiles before calving. Higher prepartum ALT activity and serum iron concentrations, together with lower creatinine concentrations, suggest that metabolic adaptation before parturition differed between the two groups. Although all measured values remained within physiological reference ranges, these differences may reflect subtle variation in hepatic metabolism and nutrient utilization during the transition period (Wood et al., 2013). The liver plays a central role in coordinating glucose production, lipid metabolism, and nutrient partitioning before and after calving; therefore, even modest differences in hepatic function before calving may influence the cow’s ability to adapt to the increased metabolic demands of early lactation (Goselink et al., 2020; Grigė et al., 2026). Rather than representing evidence of disease, these findings indicate that normal physiological variation before calving may contribute to the subsequent differences observed in postpartum F/P ratio.

In contrast, conventional metabolic biomarkers such as GLUC, NEFA, TRIG, Ca, Mg, P, ALB, and UREA did not differ significantly between groups. This observation suggests that classical blood biomarkers alone may not be sufficiently sensitive to identify cows that subsequently develop different postpartum F/P ratios. The absence of differences in NEFA is particularly noteworthy, as elevated postpartum F/P ratios are commonly associated with increased lipid mobilization and negative energy balance (Zhou et al., 2025). Because all cows were clinically healthy, physiological adaptations associated with later F/P classification may have remained below the detection threshold of traditional biochemical indicators during the prepartum period. These findings support the need for complementary approaches capable of detecting more subtle physiological changes before calving (F1000Research, 2025).

Behavioral and ruminal indicators obtained through continuous sensor monitoring appeared to discriminate future postpartum F/P groups more clearly than most blood biochemical variables. Cows that later exhibited lower postpartum F/P ratios spent more time ruminating and maintained higher ruminal pH before calving, suggesting more stable rumen function during the transition period (Baek et al., 2022). Adequate rumination stimulates saliva secretion and enhances ruminal buffering capacity, thereby promoting a more favourable fermentation environment and supporting nutrient utilization. These physiological characteristics may facilitate a smoother metabolic adaptation after calving and contribute to maintaining a lower postpartum F/P ratio (Reddy et al., 2025).

Conversely, cows that subsequently developed higher postpartum F/P ratios demonstrated greater physical activity and increased water intake before calving. Although the underlying biological mechanisms require further investigation, these findings indicate that behavioral alterations may precede the metabolic changes reflected by postpartum F/P ratio (Reddy et al., 2025). Together with the rumination and rumen pH results, these observations suggest that continuous sensor-derived measurements provide valuable information on physiological adaptation during the transition period that may not be captured by conventional blood biomarkers alone (Girdauskaitė et al., 2025).

The multivariable logistic regression analysis demonstrated that postpartum F/P ratio was associated with a combination of prepartum biochemical and sensor-derived variables rather than with a single physiological marker. CREA, GLUC, ACT, water intake, and rumen pH remained independently associated with postpartum F/P classification, highlighting the multifactorial nature of metabolic adaptation during the transition period (Hofstra and Roelofs, 2026). These findings support the concept that physiological changes preceding calving involve the interaction of energy metabolism, rumen function, and behavioral responses rather than isolated alterations in individual blood metabolites (Cabezas-Garcia et al., 2021).

Among the retained predictors, behavioral variables appeared particularly informative. Activity demonstrated the highest discriminatory ability in ROC analysis, whereas CREA, GLUC, water intake, and rumen pH showed limited predictive performance when evaluated individually. This finding suggests that no single prepartum biomarker is sufficient to reliably identify cows that subsequently develop different postpartum F/P ratios. Instead, combining continuously monitored behavioral data with conventional biochemical measurements may improve the early recognition of cows undergoing different metabolic adaptation trajectories before calving (F1000Research, 2025). The moderate predictive performance observed for activity also indicates that behavioral changes may precede measurable alterations in routine blood biomarkers (Grigė et al., 2025).

The present findings have practical implications for precision dairy herd management. The identification of measurable prepartum differences between cows subsequently classified according to postpartum F/P ratio suggests that continuous monitoring technologies may facilitate earlier recognition of animals with different metabolic adaptation patterns (Yılmaz and Orman, 2026). In particular, behavioral and ruminal indicators derived from automated sensor systems may complement routine biochemical analyses by providing continuous, non-invasive information that is difficult to obtain through conventional clinical monitoring alone (Arlauskaitė et al., 2025). Integrating these technologies into transition cow management may improve the identification of animals requiring closer observation or targeted nutritional management before the onset of clinical metabolic disturbances (Paudyal, 2021).

Several limitations should be considered when interpreting the results. The study included only clinically healthy primiparous Holstein cows from a single commercial herd, which may limit the generalizability of the findings to multiparous cows, other breeds, or different management systems. In addition, the postpartum milk F/P ratio was used as a practical herd-level indicator of metabolic adaptation rather than a definitive diagnostic marker of negative energy balance. Although the F/P ratio is widely applied in dairy herd management because it is routinely available and non-invasive, it is influenced by multiple biological and management factors, including diet, stage of lactation, and individual cow variation. Furthermore, calculating the mean F/P ratio over the first four weeks postpartum may have reduced the ability to capture short-term metabolic fluctuations occurring during the transition period, and the postpartum F/P classification was not validated against direct measurements of β-hydroxybutyrate, hepatic lipid accumulation, or other established indicators of negative energy balance. Although both the prepartum and postpartum repeated measurements were analyzed using linear mixed-effects models, these analyses were primarily designed to characterize weekly dynamics and overall group differences rather than to model individual longitudinal trajectories across the full transition period. Because the prepartum predictors of postpartum F/P classification were identified from aggregated prepartum data using logistic regression and ROC analysis, the study could not pinpoint the specific prepartum week during which the measured variables were most strongly associated with the postpartum outcome. Accordingly, the reported associations should be interpreted as observational rather than causal relationships. Future prospective studies involving larger and more diverse dairy populations, together with direct metabolic measurements, repeated biomarker assessments, and longitudinal statistical approaches, are warranted to validate these findings and further clarify the biological mechanisms linking prepartum physiological characteristics with postpartum metabolic adaptation.

Prepartum biochemical, ruminal, and behavioral characteristics differed between dairy cows subsequently classified according to their postpartum milk F/P ratio, indicating that variation in metabolic adaptation is already evident before calving. Continuous sensor measurements, particularly behavioral and ruminal indicators, complemented conventional biochemical analyses and provided valuable information for distinguishing cows with different postpartum metabolic adaptation patterns. These findings support the potential application of integrating automated sensor technologies with routine biochemical monitoring to improve the early identification of cows at risk of unfavorable metabolic adaptation during the transition period. Further prospective studies involving larger and more diverse dairy populations are warranted to validate these associations and assess their practical application in precision dairy herd management.