Body condition score at calving, subclinical ketosis, postpartum body condition score losses, diseases, and fertility in Holstein cows: modelling confounding associations

Abstract

This study examined the association between energy metabolism-related variables and conception rate at first service (CRFS) in Holstein cows. It focused on identifying confounding relationships among key factors. Data came from a large dairy farm in Coahuila, Mexico. The dataset included 1056 lactations recorded in 2024. Of these, 419 were primiparous and 637 were multiparous cows. Only cows that completed a synchronization protocol and received timed artificial insemination (TAI) were included. A logistic regression model was used to predict CRFS. In the model, body condition score (BCS) at calving and postpartum diseases were not significant when subclinical ketosis (SCK) and BCS loss were included. This suggested confounding effects. Key associations were found: Cows with BCS ≥ 3.75 at calving were 5.55 times more likely to lose ≥ 0.75 BCS units by first breeding than cows with BCS ≤ 3.5. Cows with BCS ≤ 3.5 at calving were 0.45 times as likely to develop SCK compared to cows with BCS ≥ 3.75. Cows with postpartum diseases were 4.42 times more likely to develop SCK than healthy cows. Multicollinearity was observed between postpartum diseases and SCK and between BCS at calving and postpartum BCS loss. The best-fitting model for CRFS included: breeding season, milk yield at week 8 postpartum, parity, SCK, and postpartum BCS loss. SCK and BCS losses, as well as postpartum diseases, were key factors associated with CRFS and were also confounded by BCS at calving.

1. Introduction

The lactational performance of dairy cows is strongly influenced by effective nutritional and feeding strategies, in conjunction with adequate cow comfort (Lucy, 2007; Melendez & Risco, 2021). Maintaining cow health is also essential for achieving optimal milk production and reproductive performance. However, fertility in dairy cattle is a complex trait shaped by numerous interrelated factors, including body condition score (BCS) at calving, the extent of BCS loss during the postpartum period, and the presence of peripartum diseases (Mohtashamipour et al., 2020; Santos et al., 2009). Metabolic imbalances related to negative energy balance (NEB), such as ketosis, elevated non-esterified fatty acid (NEFA) levels, and fatty liver, also represent significant risk factors for reduced conception rates (Lucy, 2001; Melendez & Vizcaino, 2024).

These variables can impact fertility both independently and synergistically, complicating efforts to evaluate reproductive outcomes in dairy herds. For example, cows with high BCS at calving often display lower fertility at first postpartum insemination (Roche et al., 2009). This pattern may be the result of reduced dry matter intake (DMI), increased postpartum BCS loss, elevated milk yield, immune suppression, and heightened risk of early postpartum conditions such as hypocalcemia and metritis (Fricke et al., 2023; Melendez & Risco, 2021; Roche et al., 2009). Cows with higher BCS at calving are more prone to developing ketosis and experiencing greater BCS losses during early lactation (Fricke et al., 2023; Roche et al., 2009). These associations are further complicated by physiological responses such as increased lipolysis, which elevates NEFA and ketone bodies concentrations (Contreras et al., 2018; Melendez & Vizcaino-Serrano, 2024; Pinedo & Melendez, 2022). For example, subclinical ketosis (SCK) during the first month postpartum, defined by blood β-hydroxybutyrate (BHB) concentrations exceeding 1.2 mmol/L (McArt et al., 2012) or milk BHB levels above 0.128 mmol/L (Melendez et al., 2025), has been linked to reduced conception rates at first service (CRFS). Nevertheless, some studies suggest that elevated BCS at calving is not consistently associated with infertility, implying that additional confounding factors may influence this relationship (Pinedo et al., 2022).

Given the intricacy of these interconnected factors, employing multivariable modeling approaches is crucial. These methods enable the identification of confounding variables, multicollinearity, and interaction effects, offering a better understanding of their collective impact on fertility. Moreover, extrinsic factors like heat stress, housing type, management systems, calving and transition management, dietary composition, and use of additives introduce further complexity and variability into fertility assessments (Cavallini et al., 2022; Fricke et al., 2023; Magro et al., 2024; Mammi et al., 2021). To reduce environmental variability and enable more robust multifactorial analysis, it is advantageous to evaluate cow-level traits such as BCS, disease status, and productivity within the same environment, particularly when animals originate from the same dairy operation. This information may also help AI software better predict which cows are at risk for low fertility in dairy herds. Therefore, the aim of this study was to analyze historical records from a large commercial dairy herd to determine whether BCS at calving, BCS loss from calving to first breeding, and the presence of SCK and other diseases during early postpartum act independently, interrelate as confounders, or influence the CRFS through interactions with other factors in Holstein cows.

2. Materials and methods

2.1. Dairy farm

The study was conducted on a commercial dairy farm located in Torreón, Coahuila, Mexico. The dairy comprised 2400 Holstein cows, housed in a dry-lot system, milked, and fed a total mixed ration three times a day. The herd consisted of 43 % first-lactation cows, 27 % second-lactation cows, and 29 % third-lactation or higher. Cows produced, on average, 35 kg of milk per cow daily, with 3.55 % fat and 3.20 % protein content. The average somatic cell count (SCC) for the herd was 130,000 cells/ml, and the milk urea nitrogen level was 11 mg/dl. The herd's average pregnancy rate was 24 % annually, varying from 15 % in summer to 34 % in cooler months (below 25 °C).

Following calving, cows underwent a postpartum health monitoring program during the first ten days of lactation. Each cow was evaluated for BHB in the blood between days 5 and 8 in milk using a portable field meter (FreeStyle Optium Neo Abbott). Cows with a BHB concentration ≥ 1.2 mmol/L (SCK) were treated with daily propylene glycol (300 ml orally) for three days, followed by a re-evaluation of their BHB levels. Subsequently, the cows entered an ovulation synchronization program utilizing a double ovsynch protocol coupled with timed artificial insemination (TAI). This double ovsynch program commenced at 50 ± 3 days in milk, with the first insemination occurring around 77 ± 3 days in milk on average. Pregnancy outcomes were assessed via ultrasound between 28- and 34-days post-insemination, resynchronizing cows diagnosed as not pregnant for a second TAI. Pregnant cows were re-evaluated between 42- and 48-days post-insemination to confirm pregnancy status. Additionally, cows were assessed for BCS at calving and during the first insemination, utilizing a scale from 1 to 5 with increments of 0.25 points, based on the method outlined by Ferguson et al. (1994). Any early postpartum diseases such as milk fever, retained fetal membranes, metritis, and subclinical ketosis were recorded in the dairy herd management software (DairyComp 305, Valley Ag Software, CA, USA).

2.2. Study design

For this investigation, historical records of lactations between 2022 and 2023 from the farm's database were obtained. As the objective of the study was to evaluate variables related to energy nutrition and energy balance of cows and their association with CRFS after an ovulation synchronization protocol and TAI. The independent variables of the study were: BCS at calving, BCS at the first AI, change in BCS between calving and first insemination (BCS loss), milk yield at week 8 postpartum, lactation number, breeding season, and the incidence of early postpartum diseases (retained fetal membranes, metritis, and/or milk fever) and cows with SCK after all cows were assessed for blood BHB concentration between 5 and 8 days postpartum (mmol/L). These diseases were included based on consistent diagnoses and records established by the farm veterinarian. Other diseases, such as clinical mastitis or others identified by dairy personnel (e.g., milkers, maternity pen managers, etc.), were not included in the analysis because these records were less reliable. Milk yield was averaged over 7 days in week 8 postpartum (kg/day) to calculate the daily mean milk production for that week.

BCS at calving, BCS loss, and BHB concentrations were dichotomized to enable the assessment of interaction or confounding and interpretation of the data. Confounding is a distortion of the true relationship between an exposure and an outcome variable due to a third variable (the confounder). The confounder is associated with both the exposure and the outcome but is not on the causal pathway (Dohoo et al., 2009). Meanwhile, interaction is a condition where the effect of an exposure on an outcome differs depending on the level of a third variable. The third variable modifies the strength or direction of the association (Littell et al., 2006).

For instance, BCS at calving was categorized as ≤ 3.5 or ≥ 3.75 to distinguish between cows with low and high BCS. Similarly, BCS losses from calving to first breeding were classified as ≤ 0.5 or ≥ 0.75 units to identify significant changes in BCS losses. Lastly, SCK was determined based on BHB concentrations, using the cut-off value of ≥ 1.2 mmol/l based on McArt et al. (2012). Categorizing variables facilitates the testing and interpretation of potential confounding or interaction.

2.3. Statistical analysis

Statistical analyses consisted of fitting logistic regression models to predict the CRFS.

Models were defined as:

Logistic regression model:

$$\text{Logit}\left(\mathrm{p}\right)=\mathrm{a}+{\mathrm{b}}_1{\mathrm{X}}_1+{\mathrm{b}}_2{\mathrm{X}}_2+\dots .{\mathrm{b}}_k{\mathrm{X}}_k$$

Where:

• π = log of the odds of the event (conception: yes, no)

• α = intercept

• β₁ = parameter of X₁

• X₁ = parity number (primiparous, multiparous)

• β₂ = parameter of X₂

• X₂ = BCS at calving (≤ 3.5, ≥ 3.75)

• β₃ = parameter of X₃

• X₃ = BCS change calving to first breeding (≤ 0.5, ≥ 0.75)

• β₄ = parameter of X₄

• X₄ = subclinical ketosis (yes, no)

• β₅ = parameter of X₅

• X₅ = milk yield week 8 (kg/day)

• β₆ = parameter of X₆

• X₆ = breeding season (summer [April-September], no summer [October-March])

• β₇ = parameter of X₇

• X₇ = presence of diseases (yes: the presence of at least one of the diseases, no: none of

• the defined diseases were present).

• β_k = parameter of X_k

• X_k = potential biological interactions

A backward elimination procedure was conducted, incorporating all variables and double interactions of biological significance. The multivariate modeling aimed to determine whether there were confounding or interactions among BCS at calving, BCS loss, the presence of other diseases, and SCK while controlling for lactation number, milk yield at week 8, and breeding season. Explanatory variables were retained in the model if the P-value was less than or equal to 0.05. The best-fitting model was chosen based on the Akaike Information Criterion (AIC). Adjusted odds ratios and 95 % confidence intervals were reported. The statistical analysis was conducted using the PROC GLIMMIX of the statistical package SAS 9.4. The GLIMMIX procedure enables researchers to specify a generalized linear mixed model and to perform confirmatory inference in such models for binary outcomes (e.g., conception: yes, no) (Littell et al., 2006).

In the absence of interaction, if two or more explanatory variables were associated with the CRFS and were also related to each other, confounding may have been present. Confounding was considered when two explanatory variables were independently associated with the conception outcome in univariate analysis, and one explanatory variable was also related to the other, following a logical sequence of causality over time. When both related explanatory variables were included in a logistic regression model, if one variable remained significant while the other was not, it suggested a confounding effect. If the significant variable was removed from the model and the previously non-significant variable became significant, with its adjusted odds ratio increasing by 30 % compared to the crude odds ratio, a confounding association between the two variables was established (Dohoo et al., 2009). To better understand potential confounding associations, a path analysis diagram was constructed.

A multicollinearity diagnostic evaluation between explanatory variables was assessed by estimating the TOL VIF and COLLIN statistics in the PROC REG procedure of SAS. The larger the eigenvalue and condition index between two variables, the greater the multicollinearity.

3. Results

3.1. Descriptive statistics

A comprehensive dataset containing 1056 lactation records was obtained from the herd management system software. After applying inclusion criteria (only cows bred by TAI with consistent health and production records) and correcting data entry errors, a final dataset of 688 lactation records was used for analysis.

Table 1 presents the results of the univariate analysis for CRFS ( %) in Holstein cows (n = 688). Independent factors such as BCS at calving, BCS loss from calving to first service, breeding season, postpartum diseases, lactation number, and SCK were significantly associated with CRFS (P ≤ 0.05). Cows bred in summer, those with postpartum diseases, those with SCK, multiparous cows, cows with BCS at calving ≥ 3.75, or cows with a BCS loss ≥ 0.75 units from calving to breeding had lower CRFS compared to cows bred in winter, cows without diseases, primiparous cows, cows with BCS at calving ≤ 3.75, or cows with a BCS loss ≤ 0.50 units, respectively.

Table 1.

Univariate analysis for conception rate at first service (CRFS, %) in Holstein cows (n = 688).

Item	Level	CRFS	Chi-Square P-value
BCS at calving	-≤ 3.5 -≥ 3.75	63.1 % 53.8 %	0.02
Breeding season	-Summer (hot) -Winter (cool)	50.9 % 62.8 %	< 0.01
Postpartum diseases	-Yes -No	40.95 % 61.08 %	0.02
Lactation	-Primiparous -Multiparous	56.1 % 46.8 %	< 0.01
Subclinical ketosis	-Yes -No	46.0 % 58.6 %	0.02
BCS loss calving-TAI	-≤ 0.5 -≥ 0.75	62.2 % 52.3 %	0.02

Table 2 shows the univariate analysis for SCK ( %) in Holstein cows (n = 688). Factors such as BCS at calving, BCS loss from calving to first breeding, and breeding season were not significantly associated with SCK (P > 0.05). However, other postpartum diseases and lactation number were significantly associated with SCK occurrence (P ≤ 0.05). Cows with postpartum diseases other than SCK and primiparous cows had a higher incidence of SCK compared to cows without diseases and multiparous cows, respectively.

Table 2.

Univariate analysis for subclinical ketosis (SCK, %) at day 5–8 postpartum by BCS in Holstein cows (n = 688).

Item	Level	SCK1	Chi-Square P-value
BCS2 at calving	-≤ 3.5 -≥ 3.75	7.60 % 9.60 %	0.24
Breeding season	-Summer (hot) -Winter (cool)	8.99 % 7.67 %	0.55
Postpartum diseases	-Yes -No	20.9 % 5.59 %	< 0.01
Lactation	-Primiparous -Multiparous	13.1 % 5.5 %	<0.01
BCS loss calving-TAI	-≤ 0.5 -≥ 0.75	7.10 % 9.50 %	0.28

¹subclinical ketosis= blood BHB ≥ 1.2 mmol/L.

²BCS= body condition score.

Table 3 displays the least square means for blood BHB concentrations (mmol/L) in relation to various study variables. Blood BHB concentrations did not differ significantly by breeding season (P > 0.05). However, significant differences were observed based on lactation number, disease presentation, and BCS at calving (P ≤ 0.05). Primiparous cows, cows with disease presentation, and cows with BCS at calving ≥ 3.75 had higher blood BHB concentrations.

Table 3.

Least square means for blood BHB concentrations (mmol/L) for study variables In Holstein cows (n = 688).

Variable	Blood BHB concentrations 1SEM= 0.07 mmol/L	P-value
Lactation
-Primiparous	1.03	<0.01
-Multiparous	0.94
Diseases
-No	0.74	<0.01
-Yes	1.23
BCS at calving
-≤ 3.5	0.96	<0.01
-≥ 3.75	1.01
Breeding season
-Cool	0.98	0.81
-Hot	0.99

¹SEM= standard error of the mean.

3.2. Multivariable modeling

Table 4 presents the adjusted odds ratios (OR) and 95 % confidence intervals (CI) for the best-fitting logistic regression model predicting CRFS in Holstein cows. Significant predictors of CRFS included breeding season, milk yield at eight weeks postpartum, lactation number, SCK, and BCS loss from calving to first breeding. In contrast, BCS at calving and postpartum diseases were not significant in the final model.

Table 4.

Adjusted Odds Ratios and 95 % confidence intervals (CI) for the best-fitting logistic regression model for predicting CRFS ( %) in Holstein cows (n = 688)¹.

Effect	Point Estimate	95 % CI	P-value
Breeding season: Cool vs Hot	1.25	1.09–2.24	0.01
Milk yield at wk 8 pp: Per 1 kg of milk	1.03	1.005–1.050	0.01
Lactation number: Prim. vs Mult.	1.60	1.21–2.12	0.01
Subclinical ketosis: No vs Yes	1.66	1.23–3.42	< 0.01
BCS loss: ≤ 0.5 vs ≥ 0.75	1.51	2.09–3.08	0.01

¹No interactions were detected. Goodness of fit parameters for the model: AIC= 1284.9 and SC= 1314.7.

Notably, the adjusted OR for SCK predicting CRFS increased by 31.7 % when BCS at calving was included in the model. This suggests that BCS at calving acted as a confounding variable in the relationship between SCK and CRFS. This is because BCS at calving was also associated with SCK in the multivariate analysis. As shown in Table 5, cows with a BCS at calving ≤ 3.5 were 0.45 times less likely to develop SCK than cows with a BCS ≥ 3.75.

Table 5.

Adjusted Odds Ratios and 95 % confidence intervals (CI) for the best-fitting logistic regression model for predicting subclinical ketosis in Holstein cows (n = 688)¹.

Effect	Point Estimate	95 % CI	P-value
Breeding season: -Cool vs. Hot	0.410	0.198–0.852	0.01
Lactation number: -Primiparous vs. Multiparous	3.386	1.801–6.366	< 0.01
BCS at calving: -≤ 3.5 vs. ≥ 3.75	0.452	0.246–0.831	0.01
Diseases: -Yes vs. No	4.42	2.38–8.32	< 0.01

¹No interactions between predictive variables were detected (P > 0.05).

Additionally, postpartum diseases were significant predictors of CRFS in the multivariable model, while SCK was not. When postpartum diseases were removed from the model, SCK became significant. Since postpartum diseases were also associated with the occurrence of SCK, they contributed to confounding the effect of SCK on CRFS.

BCS loss from calving to first breeding was also confounded by BCS at calving. The crude OR for BCS loss predicting CRFS changed by 31.3 % when BCS at calving was included in the model. Furthermore, BCS at calving was significantly associated with postpartum BCS loss. As shown in Table 6, cows with a BCS at calving ≥ 3.75 were 5.55 times more likely to lose ≥ 0.75 BCS units from calving to first breeding than cows with a BCS ≤ 3.5.

Table 6.

Adjusted Odds Ratios and 95 % confidence intervals (CI) for the best-fitting logistic regression model for predicting BCS losses from calving to first breeding in Holstein cows (n = 688)¹.

Effect	Point Estimate	95 % CI	P-value
Lactation number: -Primiparous vs. Multiparous	2.35	1.39–3.98	< 0.01
BCS at calving: -≥ 3.75 vs. ≤ 3.5	5.55	4.16–11.5	< 0.01

¹No interactions between predictive variables were detected (P > 0.05).

Fig. 1 presents a path analysis diagram showing both crude and adjusted odds ratios (OR) for CRFS ( %). In the univariate model, cows with a BCS at calving ≤ 3.5 had 1.18 times higher odds of conceiving at first service compared to cows with a BCS ≥ 3.75. This association was statistically significant. However, when subclinical ketosis (SCK) was included in the multivariable model, the adjusted OR for BCS at calving predicting CRFS decreased to 0.95 and lost significance.

Fig. 1.

Path Analysis Diagram with crude and adjusted OR for CRFS ( %).

In the univariate analysis, BCS at calving was significantly associated with CRFS, with non-overconditioned cows exhibiting a 1.18-fold increase in the odds of conception. These cows also experienced less BCS loss postpartum (crude OR = 0.18) and had a lower incidence of SCK (crude OR = 0.45). However, in the multivariable model, the association between BCS at calving and CRFS became nonsignificant (adjusted OR = 0.95), indicating that its initial effect was likely confounded by postpartum variables. Cows without SCK had a crude OR of 1.26 for CRFS, and those with lower postpartum BCS loss showed a crude OR of 1.15. When these factors were included in the multivariable model, their adjusted odds ratios increased to 1.66 and 1.51, respectively, further emphasizing their independent predictive strength. These shifts highlight that BCS at calving functioned as a confounding variable, influencing the associations between postpartum BCS loss, SCK, and fertility outcomes. Additionally, cows without postpartum diseases had a 1.52-fold increased odds of conception and were less likely to develop SCK (OR = 0.22). Multicollinearity was detected between SCK and postpartum diseases, and between BCS at calving and BCS loss. When both SCK and postpartum diseases were included in the same model, SCK lost statistical significance, while postpartum diseases retained a strong association with CRFS. Conversely, removing postpartum diseases restored SCK’s significance, suggesting that both variables are associated with CRFS, and postpartum diseases also mediate the occurrence of SCK. Notably, the removal of BCS loss did not render BCS at calving significant in the model, reaffirming that BCS at calving serves as a confounder rather than a direct predictor of fertility. This distinction underscores the importance of evaluating postpartum variables when interpreting fertility outcomes.

In the univariate model, the crude OR for SCK predicting CRFS was 1.26. After including BCS at calving in the multivariable model, the adjusted OR for SCK increased to 1.66 (95 % CI = 1.23–3.42). This suggests that BCS at calving acted as a confounding variable in the relationship between SCK and CRFS. It indicates that high BCS at calving may negatively affect CRFS primarily through its association with SCK.

Similarly, the relationship between BCS at calving and postpartum BCS loss showed that cows with a BCS ≤ 3.5 were 0.18 times less likely to lose ≥ 0.75 BCS units postpartum than cows with a BCS ≥ 3.75. In univariate analysis, cows losing ≤ 0.5 BCS units had 1.15 times higher odds of conceiving at first service than those losing ≥ 0.75 units. However, after including BCS at calving in the multivariable model, the adjusted OR for postpartum BCS loss predicting CRFS increased to 1.51 (95 % CI = 2.09–3.08). This further supports the role of BCS at calving as a confounding factor in evaluating the impact of postpartum BCS loss on CRFS.

Additionally, no association was found between SCK and postpartum BCS loss, indicating that both variables were independently associated with CRFS.

Regarding postpartum diseases, healthy cows had 1.52 times higher odds of conceiving at first service compared to cows with postpartum conditions such as hypocalcemia, retained fetal membranes (RFM), and/or metritis. Healthy cows were also 0.22 times as likely to develop SCK compared to cows with postpartum diseases.

While SCK was not a significant predictor of CRFS when postpartum diseases were included in the model, it became significant again when postpartum diseases were excluded. In that case, the adjusted OR for SCK predicting CRFS was 1.66 (95 % CI = 1.23–3.42).

Multicollinearity diagnostics revealed a substantial degree of collinearity between postpartum diseases and SCK, as indicated by elevated eigenvalues and condition index values. A similar pattern was observed between BCS at calving and postpartum BCS loss. When SCK was excluded from the model, postpartum diseases became a statistically significant predictor of CRFS, suggesting that its independent effect had previously been masked by collinearity with SCK. In contrast, removing BCS loss did not make BCS at calving significant, indicating that BCS at calving functions primarily as a confounding variable rather than an independent predictor.

Given the study’s emphasis on variables more directly related to energy metabolism, SCK was retained in the final model, while postpartum diseases were excluded. However, it is acknowledged that postpartum diseases negatively affect CRFS and contribute to the confounding structure surrounding SCK. Accordingly, the final model identified five significant predictors of CRFS: breeding season, milk yield at eight weeks postpartum, lactation number, SCK, and BCS loss from calving to first breeding.

4. Discussion

This study aimed to determine the independent effects of energy metabolism–related variables on the probability of CRFS in dairy cows. Achieving this objective required unravelling complex interactions and confounding relationships among variables within a multivariate logistic regression framework. Although backward elimination was used to refine the model, biological plausibility and physiological relevance were prioritized to ensure meaningful interpretation.

All six dichotomous variables (BCS at calving, breeding season, postpartum diseases, lactation number, SCK, and BCS loss between calving and TAI) were significantly associated with CRFS in univariate analyses. However, statistical significance in univariate models does not imply causality. Instead, it may reflect inter-variable confounding or collinearity, which complicates the identification of truly independent predictors.

4.1. Impact of BCS at calving on CRFS

BCS at calving may influence CRFS indirectly through a cascade of metabolic and physiological disruptions. Elevated BCS is linked to reduced feed intake, increased risk of dystocia and metabolic disorders, and greater postpartum BCS loss, all contributing to NEB (Santos et al., 2009, 2010). Given the typical 75–80-day interval between calving and first AI, it is unlikely that BCS at calving exerts a direct effect on conception. However, NEB and reduced feed intake during early lactation impair endocrine function and reproductive performance by altering the profiles of metabolic hormones and key metabolites (Bertoni & Trevisi, 2013; Chapinal et al., 2012; Lucy, 2001, 2007).

Cows with excessive BCS tend to consume less feed and experience greater BCS loss postpartum (Chebel et al., 2018; Roche et al., 2009). This predisposes them to metabolic disorders such as hypocalcemia, ketosis, RFM, and metritis, a condition consistently associated with reduced CRFS (Melendez, 2024; Mohtashamipour et al., 2020; Ribeiro & Carvalho, 2018). These disorders are also linked to dystocia and stillbirth (Gröhn & Rajala-Schultz, 2000), and their detrimental effects on fertility may stem from impaired uterine health and inflammation-mediated reproductive dysfunction (Fricke et al., 2023; Hildebrand et al., 2023).

4.2. BCS loss and fertility outcomes

Evidence from a retrospective study conducted in Colorado, USA, involving 12,042 lactations, demonstrates a negative association between pronounced postpartum BCS loss and CRFS (Hernandez-Gotelli et al., 2023). Similarly, Holstein cows with lower BCS at the time of first breeding and greater BCS reduction from calving to AI showed reduced odds of conception (Pinedo et al., 2022). In contrast, cows that maintained or gained BCS during this period experienced fewer peripartum diseases (Barletta et al., 2017), reinforcing the association between metabolic steadiness and improved reproductive outcomes (Fricke et al., 2023).

Importantly, the directionality of the relationship between BCS, disease, and fertility requires clarification. In this study, disease diagnoses were limited to the first week postpartum. Therefore, it is biologically implausible that BCS loss occurring over the first 70 days of lactation could cause early-onset disease. This suggests that the causal pathway more likely reflects disease-induced BCS loss rather than the reverse.

Prepartum BCS dynamics also influence periparturient physiology. Even when cows calve with a normal BCS, BCS loss during the dry period has been associated with increased risk of uterine disease and reduced conception rates at both first and second AI (Chebel et al., 2018). A large-scale study involving 43,396 lactations further reported higher incidences of RFM, metritis, and mastitis in cows that lost BCS prepartum (Melendez et al., 2020). These findings highlight the importance of longitudinal BCS monitoring throughout the transition period. Unfortunately, such monitoring was not feasible in the present study due to limitations in data availability.

4.3. Subclinical ketosis and CRFS

Subclinical ketosis presents a multifaceted challenge due to its strong association with elevated BCS at calving and increased incidence of RFM, metritis, and hypocalcemia. Its confounding role within the fertility framework is evident. Although SCK has been identified as the most economically troublesome postpartum disorder (Rasmussen et al., 2024), its direct impact on fertility, once adjusted for comorbidities, is less pronounced than that of metritis or RFM.

Cows with serum BHB concentrations ≥1000 μmol/L postpartum exhibit lower conception rates and prolonged intervals to pregnancy (Walsh et al., 2007). Meta-analytic evidence confirms reduced odds of conception in ketotic cows (Raboisson et al., 2014), and recent studies have quantified decreased CRFS in cows with milk BHB levels ≥ 0.128 mmol/L (Melendez et al., 2025). Mechanistically, SCK compromises oocyte and embryo viability through elevated concentrations of NEFA and BHB, which trigger endocrine disruption and inflammatory responses (Chirivi et al., 2023; Hildebrand et al., 2023; Leroy et al., 2006). These disruptions include altered IGF-1 signaling and impaired GnRH/LH axis dynamics, potentially leading to early embryonic loss (Piechota et al., 2015; Sammad et al., 2022).

Despite these effects, the fertility consequences of uterine disorders such as metritis and RFM remain substantial and independent of their association with SCK. Multicollinearity among postpartum disorders complicates causal attribution. Ribeiro et al. (2016) demonstrated that both uterine and systemic diseases not only reduce CRFS but also increase the risk of pregnancy loss. Emerging evidence further suggests that uterine inflammation may exert long-term effects on folliculogenesis and oocyte quality, persisting months beyond clinical resolution (Leblanc, 2023).

One limitation of this study is the inability to account for several additional fertility-related factors, particularly energy balance dynamics and BCS fluctuations during the dry period. For instance, cows with optimal BCS at calving may have experienced significant BCS loss during dry-off, a metabolic scenario distinct from cows that maintained their BCS throughout. Prepartum BCS loss has been linked to increased susceptibility to peripartum disorders such as hypocalcemia, RFM, and metritis (Chebel et al., 2018; Melendez, 2024), which may subsequently impair reproductive performance.

Another limitation involves the analytical treatment of variable interactions. In this study, interactions involving more than three fertility-related variables could not be statistically evaluated. For example, the relationships among BCS at calving, SCK, and postpartum diseases, each associated with CRFS, may be influenced by additional modifiers such as parity (primiparous vs. multiparous) or seasonality (summer vs. non-summer breeding). These moderating effects could alter the magnitude or direction of observed associations.

Nevertheless, this study provides valuable insights into key relationships, including the link between BCS at calving and peripartum disorders (including SCK), the association between BCS loss and CRFS, and the complex interplay of confounding, interaction, and multicollinearity among these variables.

5. Conclusion

In this study, the optimal multivariable logistic regression model for predicting CRFS included five significant predictors: breeding season, milk yield at eight weeks postpartum, lactation number, SCK, and BCS loss between calving and first breeding. Multiparous cows inseminated during the summer, likely exposed to heat stress, who exhibited SCK within the first postpartum week, elevated milk production at week eight, substantial BCS loss during the postpartum period, and early postpartum disease incidence demonstrated significantly reduced CRFS.

These findings indicate that BCS at calving functions primarily as a confounding variable rather than an independent predictor of fertility. Its influence appears to be mediated through its associations with other postpartum conditions, including SCK, disease occurrence, and BCS loss.

Notably, variables not captured in the present analysis, such as BCS dynamics during the dry period, may also exert considerable influence on CRFS. These unmeasured factors could confound observed relationships, particularly those involving BCS at calving and subsequent postpartum outcomes. This underscores the importance of incorporating longitudinal data across the transition period to improve causal inference and enhance the physiological relevance of predictive models for reproductive performance.

Ethical statement

This study did not use live animals. The investigation was a data analysis study design

CRediT authorship contribution statement

Pedro Melendez: Writing – original draft, Validation, Supervision, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Julian Bartolome: Writing – review & editing, Validation, Methodology, Conceptualization. Gerardo Gonzalez: Writing – review & editing, Validation, Resources, Project administration, Investigation, Funding acquisition, Conceptualization. Gustavo Lastra-Duran: Writing – review & editing, Validation, Supervision, Project administration, Methodology, Formal analysis. Pablo Pinedo: Writing – review & editing, Validation, Software, Methodology, Investigation, Formal analysis, Conceptualization.

Declaration of competing interest

The authors declared no conflict of interest.