Feeding patterns as proxies of production performance and feed efficiency in growing pigs from different genetic backgrounds

1 Introduction

Improving feed efficiency remains a research priority in the swine industry to optimize feed utilization (reviewed in Patience et al., 2015; Gilbert et al., 2017). Besides proper nutrient utilization and genetic predisposition for fast growth, daily feed intake supports livestock muscle accretion and body weight gain once metabolic requirements are covered. Yet, achieving an optimal feed intake uniformly across the entire group of pigs for profitable, healthy growth rates can be a struggle on-site, and evaluating their feeding behavior provides new insights to address this challenge (Brown-Brandl et al., 2013; Boumans et al., 2015; Ewaoluwagbemiga et al., 2021).

Feeding motivation impacts the likelihood of starting to eat and the intensity of the feeding behavior promoting feed intake until the animal is satisfied or feed-drive drifts (Kyriazakis and Tolkamp, 2011; Mason and Bateson, 2009; Tolkamp and D’Eath, 2016), and assessing how animals behave on-site provides new insights into their feeding motivation. Research on swine feeding behavior highlighted that pigs are driven to meet a desirable daily feed intake but there is variability among individuals as some are eager to consume larger feed rations than others (Nielsen, 1999; Boumans et al., 2015; Maselyne et al., 2015). In group-housed rearing systems, feed-eager pigs actively seek opportunities to fulfill their need for high feed intake, achieving higher body weights and faster growth rates than those in pigs with low appetite, particularly if animals compete for access to feeder space. In this case, dominant individuals displace and outcompete subordinates, especially during feeder rush hours. However, pigs can adjust their feeding behavior to achieve their desirable feed intake depending on factors such as their social context, housing conditions, and environmental factors (Nielsen, 1999; Bornett et al., 2000; Boumans et al., 2015, Boumans et al., 2017). Thus, analyzing how pigs tailor their feeding behavior (e.g., structure, pattern, and intensity) in commercial-like conditions can help elucidate how feeding patterns relate to performance outcomes (Boumans et al., 2015).

Besides feeding motivation, the way animals consume their daily feed intake (i.e., feeding patterns) and their circadian patterns can impact not only their total feed intake but also the efficiency of nutrient intake (Nielsen, 1999; Baumung et al., 2006; Ewaoluwagbemiga et al., 2021). Although pigs with different feeding patterns can attain similar daily feed intake, a feeding pattern with a high frequency of small and short feeding bouts is likely to yield low individual feed intake under conventional, group-housed rearing conditions (Fernández et al., 2011; Piles et al., 2025). Additionally, for the same daily feed intake, a lower feeding frequency with larger rations can improve feed efficiency of growing pigs (Le Naou et al., 2014). Implementation of electronic feeders in swine facilities enables calculation of the daily feed intake for each pig as well as their individual feeding pattern by integrating features from each feeding bout (e.g., feed intake, feeding rate, occupation time, time of occurrence) and their structure throughout the day (e.g., daily frequency, meal criterion, time between feeding bouts) (Brown-Brandl et al., 2013; Maselyne et al., 2015; Bus et al., 2024). Jointly, determining how feeding patterns prompt daily feed intake and other production outcomes provides new practical avenues to support the performance and efficiency of growing pigs on-site.

Most studies on swine feeding behavior used pure lines in breeding facilities, which poorly reflect the reality of commercial rearing barns raising three-way crossbreeds for meat production (Poullet et al., 2022; Bus et al., 2025). In addition, among the studies tackling the link between feeding metrics and production traits, very few looked at how the individual feeding pattern impacts pigs’ performance using multivariate approaches (Ewaoluwagbemiga et al., 2021; Núñez et al., 2023; Bus et al., 2024). Thus, the objective of this study was, first, to describe the feeding patterns of growing pigs from different commercial genetic backgrounds and, second, to assess to what extent these feeding metrics are useful for inferring performance outcomes in growing pigs. We hypothesized that the feeding pattern would differ between crossbreeds, and feeding metrics associated with pigs’ feeding drive (e.g., feed intake, feeding rate, feeding time) would act as behavioral indicators of rearing performance.

2 Materials and methods

This research was carried out at the Institute of Agrifood Research and Technology (IRTA) experimental pig facilities (Monells, Girona), and all animal procedures carried out in this study were approved by the IRTA Ethical Committee (Expedient FUE-2022-02967477, ID 01SQ7J04D) and were conducted in accordance with the guidelines outlined by the Spanish Royal Decree 53/2013 and European Union Directive 2010/63/EU on the protection of animals used for experimentation.

2.1 Study design

A total of 620 commercial crossbred pigs from eight genetic backgrounds were evaluated during rearing in four successive batches from April 2022 to September 2023. The eight genetic groups were commercial F2 pigs, offspring of a three-way cross of different commercial lines of Duroc, Pietrain, Landrace, and Large White breeds from five breeding companies (i.e., origin), which can differ in their breeding goals. In all cases, the finishing sires belonged to different Pietrain or Duroc lines. As for the dams, in most cases, they were F1 crosses of different Landrace and Large White lines, except for one of the groups where the maternal granddam was Duroc. Table 1 summarizes the experimental groups in this study.

Table 1. Overview of the experimental design and groups by batch, eight genetic backgrounds, sex, and dietary composition.

Du, Duroc; Lw, Large White; Ld, Landrace; Pt, Pietrain.

Subscripts in genetic combinations refer to the population of origin.

¹All males were intact.

²Analyzed at the IRTA Animal Nutrition laboratory, Constanti, Spain.

Piglets were raised off-site during weaning and early growth at their respective breeding company barns. Pigs were then transported to control facilities at around 60 days of age and reared on-site under standardized conditions at IRTA Experimental Pig Farm in Monells (Girona, Spain). Pigs were reared separately by genetic background and sex (i.e., single-sex housing) in consecutive and identical rooms within the same barn. After a three-week habituation period, the study began when animals were 80 days of age (33 ± 6 kg) and continued until they reached market body weight (around 170 days of age and 119 ± 11 kg of body weight, depending on the genetic background), lasting approx. 90 days, as illustrated in Figure 1. All pigs were reared under the same husbandry and management procedures during the growing-fattening period.

2.2 Animal husbandry and management

Upon arrival (on day 1), pigs were housed in groups of 12 to 14 animals per pen (5 m × 2.8 m, 14 m²) and ear tagged (ISO FDX, Allflex) for individual identification, automatic feeder recognition, and traceability throughout the experiment. Each pen was equipped with one NEDAP electronic feeder (NEDAP N.V., Groenlo, the Netherlands) and one watering cup drinker for ad libitum access to feed and fresh water. Windows in the study rooms provided natural light, with daylight length varying among seasons. The barn’s automated system adjusted ventilation rate and heater activity to keep room temperature between 22 °C and 25 °C. Fan speed increased progressively from 50 to 100% max. rpm as room temperature exceeded 26 °C, and heaters turned on when room temperature fell below 21 °C. A three-phase feeding strategy was adopted to meet group-mean nutritional requirements during the growing period (phase I: 80–100 and phase II: 101–145 days of age) and the finishing/fattening period (phase III: 146–165 days of age). Pigs were fed the commercial diets because feed formulations were tailored for crossbreed-specific metabolic needs (analyzed feed composition is shown in Table 1).

2.3 Data collection

2.3.1 Production and feed efficiency traits

All pigs were individually weighed at the beginning of the study, every three weeks during rearing, and at the end of the study (see Figure 1 for more detail). Body weight (BW) records were used to calculate the individual average daily growth (ADG) using a linear model fitted to body weight measurements over the control period. Backfat thickness (BFT) was also individually measured every three weeks on each weighing day using PIGLOG 105 (Frontmatec A/S, Smørum, Denmark) ultrasound equipment. Automatic feeders provided individual feed intake (FI) records from each feeder visit, and this information was used to calculate the average daily feed intake (ADFI) for each animal. The individual coefficient of variation (CV) for the ADFI was also calculated for each period by dividing the standard deviation throughout the 20-day interval by the mean ADFI during this period. Then, the feed conversion ratio (FCR) was computed for each pig as the amount of FI required to gain one unit of body weight (ADFI divided by ADG). The calorie efficiency (CE) and protein efficiency (PE) were also calculated as additional feed efficiency metrics to account for numerical dietary differences by individually multiplying the previously computed FCR by the calorie and protein content of the corresponding diet for that experimental group and phase.

Feed efficiency and performance phenotypes were then estimated at four fixed ages for all pigs because age at weighing varied slightly across batches. Values for production traits (FI, BW, and BFT) were interpolated from random regression animal models described in Herrera-Cáceres et al. (2022) at 105, 125, 145, and 165 days of age (i.e., within the age range available for all pigs), and equal-length intervals were considered for computing average daily metrics (ADG and ADFI) and feed efficiency traits (FCR, CE, and PE) during the growing (period I: 105–125 days old) and fattening phases (period II: 145–165 days old).

2.3.2 Feeding metrics at the feeder visit level

Electronic feeders automatically registered the starting time, FI, and ending time of each feeder visit for each pig. Then, the occupation time (OT) and feeding rate per feeder visit (FR_visit; FI divided by OT) were calculated. From these metrics, average daily feed intake (ADFI), daily feeder occupation time (DOT), and daily number of feeder visits (DFV) were also computed. The hourly feeder occupation time at the pen level was also calculated by summing the OT of all penmates by date and time of day, in minutes, and dividing by 60. Night feed intake was estimated following Bus et al. (2025) as the proportion of daily feed intake that took place between 10:59 p.m. and 6:00 a.m. the following day. Only records from two 20-day periods (I: 105–125 days old and II: 145–165 days old), fixed across experimental groups, were employed in this study.

2.3.3 Feeding metrics at the meal level

Feeder visits were grouped together into meals based on the lowest probability of starting a new feeder visit within the next minute after the last feeder visit (P_start) (Howie et al., 2009; Tolkamp et al., 2011; Gomes et al., 2021). Meal criteria were computed individually and by rearing period because the meal criterion may vary with age and among pigs (Fernández et al., 2011; Maselyne et al., 2015; Piles et al., 2025). A simple moving average was calculated over five center-aligned one-minute intervals to smooth out short-term ups and downs in the individual P_start curve as suggested by previous researchers (Howie et al., 2009; Piles et al., 2025). The individual meal criteria in this study averaged 33 min during the grower period and 30 min during the fattening period. Then, feeder visits close to each other (below the individual meal criterion) were merged to calculate the feed intake per meal (meal size), meal duration, number of feeder visits per meal (FVM), feeding rate per meal (FR_meal), and time to the next meal or interval between meals (IBM).

2.4 Statistical analyses

Data processing, statistical computing, and data illustration were performed using R software 4.3.1. Both feeding and performance data were analyzed separately by period (I: grower and II: fattening) to assess whether feeding patterns and their relationship with production efficiency varied between rearing periods with the pig as the experimental unit. First of all, the original dataset from the electronic feeders was filtered for outliers and abnormal observations using multivariate analyses (Piles et al., 2025) by examining the Q-Q plots and Hotelling’s T² from a principal component analysis of the behavioral traits. Observations with a distance greater than three times the Hotelling T² limit were removed (0.12% of observations). Feeder records were also evaluated for unrealistic observations with large feed consumption (2,000 g), long feeder visits (2,000 s), and fast feeding rate (200 g per min) (Herrera-Cáceres et al., 2020; Ewaoluwagbemiga et al., 2021; Piles et al., 2025). Pigs missing data for 5 days per period were also excluded (1% of total observations). The resulting dataset included 261,456 observations for further analyses at the feeder visit level. Across models, post-hoc analyses were employed for statistical pairwise comparisons between time of day and groups using the Bonferroni test to adjust p-values for multiple comparisons. Significance was set at p< 0.05.

2.4.1 Hourly pattern of feed intake

The following mixed model was used to estimate the 24-h pattern of feeding activity:

Model 1:FIijklmo=intercept+batchm+groupk+sexj+dayl+timeo+timeo*groupk+pigi+eijklmo

where FI_ijkmo is the feed intake at o^th time (24 levels) on the l^th day (throughout the 20-day interval) of the i^th pig nested within pen (random effect) with sex j^th (two levels: intact male or female), from the k^th genetic group (eight levels, Table 1) within the m^th batch. The residuals of the model are e_ijklmo.

2.4.2 Feeding patterns and performance traits across genetic backgrounds

The least-squares means of the feeding behavioral traits (at both the feeder visit and meal levels) for each genetic background were estimated from a linear mixed model with the individual as a random effect as follows:

Model 2a:Phenotypeijklm=intercept+batchm+dayl+groupk+sexj+pigi+ eijklm

where Phenotype_ijkm is the phenotype of a feeding trait throughout the 20-day interval for an individual_i of sex_j belonging to group_k in batch_l on day_l as defined above.

The conditional and marginal coefficients of determination (R²) were computed to assess the ability of the fixed and random factors to account for the variability of feeding traits for each period. The intraclass correlation coefficients (ICC) were also calculated to evaluate the repeatability of an individual phenotype for each feeding metric and period throughout the 20-day interval.

Production and efficiency metrics were analyzed with a fixed model since, within each period, there were no repeated measures for an individual:

Model 2b:Phenotypeijkl=intercept+batchl+ groupk+sexj+eijkl

where Phenotype_ijkl is the phenotype of ADG, BW, BFT, FCR, CE, PE, or meal criteria for an individual of sex_j belonging to group_k in batch_l as defined above.

2.4.3 Association between feeding traits and performance outcomes

The residuals of production metrics, efficiency metrics, and feeding traits were used for association analyses after scaling and standardizing the residual dataset. Pairwise Pearson correlation coefficients between the residuals of feeding traits and those of the production outcomes were estimated to evaluate the one-to-one association between them after correcting for systematic effects. Correlation matrices were subjected to cluster analyses to group together highly associated variables using heatmaps for visualization. Principal component analyses were also performed on the residual dataset to illustrate the relationship between feeding metrics and performance outcomes using the loading-weight plots of the first and second principal components for each period.

2.4.4 Explaining performance outcomes from feeding traits

Partial least-squares regression (PLSR) was employed to assess how feeding traits (including the ADFI) jointly explain performance outcomes (BW, ADG, BFT, FCR, CE, and PE) using the same residual dataset as in the preceding association analyses. The variable importance in projection (VIP) scores were calculated for all feeding variables, and those with VIP scores higher than one were considered plausible partial predictors of efficiency and production metrics. Based on these results, three PLSR models were developed according to the feeding metrics included: 1) all feeding metrics (PLSR_all: 14 variables), 2) only feeding metrics with high VIP scores at the meal and feeder visit level (PLSR_VIP: 6 variables), and 3) only feeding metrics with high VIP scores at the feeder visit level (PLSR_FV: 3 variables). The PLSR procedure performed a 10-fold cross-validation to estimate the optimal number of projections, and the R² adjusted for the number of optimal projections (adj-R²) was computed as a goodness-of-fit metric for the dependent variables in each PLSR model. The cross-validated R² (Q²) was also computed for model validation.

3 Results

3.1 Hourly pattern of feed intake

Figure 2 illustrates the daily feeding activity of pigs from different genetic backgrounds during the grower and fattening periods. The overall feeding activity of growing pigs showed two feeding peaks at dawn and dusk, while feed intake decreased during midday and dropped during nighttime, with significant differences between crossbreed groups. In general, the four crossbreds from Pietrain boars and Large White × Landrace sows had more pronounced ups and downs than the three Duroc-sired crossbreds. Compared to these three, pigs from the four Pietrain-sired crossbreds ate significantly less during midday (grower period: 28.8%, P< 0.001) and at midnight (grower period: 43.7%, P< 0.01; and fattening period: 22.5%, P = 0.037). In addition, the FI difference during feeding peaks was higher in the fattening period as Pietrain-sired pigs ate 15.1% more during the morning peak (P = 0.044) and 22.9% more during the afternoon peak than Duroc-sired pigs (P< 0.001). Indeed, the Pietrain-sired crossbred from origin 5 showed the most pronounced feeding peaks during both periods (Figure 2) and had lower night feed intake than other Pietrain-sired crossbreds. Feeder occupation time at the pen level followed a similar pattern throughout the day and approached 90% occupancy, on average, during the grower afternoon peak (from 6:00 to 8:00 p.m.) in four of the Pietrain-sired crossbreeds (Supplementary Table S1).

3.2 Feeding patterns and performance traits across genetic backgrounds

Table 2 summarizes the ADFI and feeding pattern for each genetic background during both periods. More descriptive statistics of the feeding metrics by crossbreed and period are included in Supplementary Table S2. The feeding behavior of growing-fattening pigs varied significantly depending on their genetic background, and small changes in the feeding behavior between genetic backgrounds were associated with large differences in the ADFI, particularly during the fattening period. Overall, the three Duroc-sired crossbreds consistently had the highest ADFI during the growing and fattening periods, while the ADFI of Pietrain-sired crossbreds varied across genetic groups and over time. Certainly, the ADFI of Pietrain-sired pigs from origin 5 (those with the most pronounced feeding peaks in Figure 2) peaked during the fattening period, surpassing that of pigs crossbred with Duroc males. As also illustrated in Figure 2, Duroc-sired crossbreeds had the highest proportion of night FI, followed by the crossbred with Duroc in its maternal line, and then Pietrain-sired crossbreds (Table 2).

Table 2. Least-square means (± standard error) of the feeding metrics for each genetic background over both periods (n = 612 growing pigs).

ADFI, average daily feed intake; DOT, daily feeder occupation time; FV, feeder visit frequency; OT, feeder occupation time; FI_visit, feed intake per feeder visit; FR, feeding rate; FVM, feeder visits per meal; and IBM, interval between meals.

Du, Duroc; Lw, Large White; Ld, Landrace; Pt, Pietrain.

Subscripts in genetic combinations refer to the population of origin.

¹From 105 to 125 days old.

²From 145 to 165 days old.

^a–dDifferent letters indicate significant differences across genetic backgrounds within each period and metric (P< 0.05).

Looking at the feeding traits altogether, few feeding patterns appeared to differ among groups based on the frequency of feeder visits/meals, OT, and FI. These relative differences between genetic groups remained similar during both periods. For example, the group with the highest ADFI (Duroc2b-sired pigs from origin 2) had long and numerous feeder visits and meals, while the group with the lowest ADFI (Pietrain-sired pigs from origin 2) had the highest DFV and number of meals and the slowest FR (small feed intake and short occupation time per visit and meal). Compared with these Duroc2b-sired pigs from the same origin, Pietrain-sired pigs had higher ADFI CV during both periods (grower: + 2.9 ± 0.4%, P< 0.001; and fattening: + 3.1 ± 0.5%, P< 0.001). Two groups of Pietrain-sired pigs (those from origins 1 and 5) consistently had the fastest FR during both periods, yet their ADFI fell between the previous groups because those from origin 1 had the lowest DOT and those from origin 5 had low DFV and few meals.

Production and efficiency metrics across genetic types are outlined in Table 3. Overall, the growing pigs with the lowest ADFI (Pietrain-sired pigs from origin 2) were the lightest group, and those with the highest ADFI (Duroc2b-sired pigs from origin 2) were the heaviest group. Both genetic types also showed the lowest FCR during both periods. Backfat was the thickest in the three Duroc-sired genetic types with significant differences among them, with Duroc2b-sired pigs from origin 2 being leaner than the other two groups (Table 3). Among the Pietrain-sired crossbreeds, those with the most pronounced feeding peaks and fastest FR (origin 5) ate the most, grew the fastest, and were the heaviest and the most protein efficient.

Table 3. Least-squares means (± standard error) of the performance outcomes for each genetic background across both periods (n = 612 growing pigs).

BW, body weight; BFT, backfat thickness; ADG, average daily growth; FCR, feed conversion ratio; CE, calorie efficiency; PE, protein efficiency.

Du, Duroc; Lw, Large White; Ld, Landrace; Pt, Pietrain.

Subscripts in genetic combinations refer to the population of origin.

¹From 105 to 125 days old.

²From 145 to 165 days old.

^a–dDifferent letters indicate significant differences across genetic backgrounds within each period and metric (P< 0.05).

3.2.1 Intra-individual repeatability of feeding metrics

Table 4 captures the mixed model goodness-of-fit statistics and the intra-individual correlation (i.e., repeatability) for the feeding metrics. In general, higher conditional R² values (i.e., overall goodness of fit) were achieved for metrics calculated at the feeder visit level than for those at the meal level. Looking at differences between periods, the model’s explanatory capacity was higher when using observations from the growing period than when using those from the fattening period. Based on the marginal R², the variability explained by the fixed factors was above 10% for the ADFI, DOT, and FR_visit in both periods, driven mainly by differences across genetic backgrounds.

Table 4. Goodness-of-fit statistics (conditional coefficient of determination: Cond. R² and marginal coefficient of determination: Marg. R²) and repeatability (intraclass correlation coefficient: ICC) of each feeding trait for the mixed model including genetic background, sex, batch, and day as fixed effects, and animal as a random effect.

¹From 105 to 125 days old.

²From 145 to 165 days old.

The individual effect explained substantial variability in feeding metrics during both periods, although repeatability estimates varied widely across traits (Table 4). Within-subject observations were strongly correlated for most metrics at the feeder visit level (r > 0.60) but were poorly correlated for those at the meal level (r< 0.25). Among metrics at the feeder visit level, FR showed the highest repeatability, particularly during the grower period. As for the daily feeding metrics, DFV and DOT had the highest within-animal correlation, while the repeatability of ADFI and night feed intake was moderate to low, and repeatability was lowest for meal frequency (Table 4).

3.3 Relationship between feeding behavior and performance metrics

Figure 3 illustrates the overall relationships between feeding metrics and performance outcomes based on principal components 1 and 2, which jointly explained 44% and 48% of the total variability across growing and fattening periods, respectively. The FR, FI, and OT per visit grouped together in the upper-right quadrant beside meal size and the IBM, while the frequency of meals and FV, meal duration, and FV per meal were located in the opposite quadrant. Production and feed efficiency metrics, together with the ADFI, clustered in the bottom-right quadrant in the same direction as the preceding groups of feeding traits in either PC1 or PC2. The ADFI was the foremost driving variable within this cluster, and the FR_visit appeared to be particularly associated with performance outcomes during the growing period. The proportion of night feed intake and DOT seemed to have little influence on the overall variability of feeding metrics and performance outcomes during both periods (Figure 3).

Full correlation coefficients between residuals of feeding behavior and performance metrics are provided in Supplementary Table S3 and give a more detailed picture than the associations observed in principal component analyses. Overall, the DFV was strongly and negatively associated with both FI and OT per visit (r = -0.80 and -0.70, respectively). In addition, the FR and FI, both per visit and per meal, were negatively correlated with DOT (r< -0.60) and positively with ADFI during the two periods (r between 0.31 and 0.53). Figure 4 illustrates the heatmap of the correlation coefficients between feeding metrics and performance outcomes. Besides ADFI, the FR and FI per visit were the feeding behavioral traits most strongly associated with production and feed efficiency traits during both periods, correlating positively with BW (r = 0.33 and 0.50 in the grower and fattening periods, respectively), BFT (r > 0.24), and ADG (r = 0.30 and 0.18).

3.4 Explaining performance outcomes from feeding traits

In the PLSR analysis considering all feeding metrics (PLSR_all model), Figure 5 shows the VIP scores for each trait to assess their potential to explain production and feed efficiency traits (BW, ADG, BFT, FCR, CE, and PE). Overall, these results demonstrate that the individuals’ ADFI and their FR and FI at both the feeder visit and meal levels were the most important descriptors of the performance profile of growing pigs during both rearing periods. Based on this VIP analysis, two additional PLSR models were assessed: one including only the most relevant feeding metrics (PLSR_VIP model) and another considering only ADFI plus FR and FI at the feeder visit level (PLSR_FV model).

Table 5 summarizes the goodness-of-fit statistics for each production and feed efficiency trait obtained from the three PLSR models across both periods. Adjusted R² values calculated for the three PLSR analyses support the conclusion that, overall, feeding metrics can partially explain production and feed efficiency traits during rearing. However, their explanatory capacity varied based on the period, performance trait of interest, and feeding metrics considered in the analyses. Higher goodness of fit was obtained for outcomes calculated during the grower period than for those during the fattening period. Regarding differences across performance traits, feeding traits may largely explain the variability of BW, moderately that of ADG, and weakly that of BFT and feed efficiency traits (FCR, PE, and CE). Comparing results of the three PLSR models, the fully parametrized model (PLSR_all) achieved a goodness of fit slightly higher than that achieved with “reduced” models including only key feeding metrics from previous VIP analyses (PLS_RVIP and PLS_RFV models). Compared to the PLSR_FV, the PLSR_VIP model additionally including the FR and FI at the meal level yielded similar adjusted R² values (Table 5). Cross-validated R² values obtained for each production and efficiency trait were similar to the values presented in Table 5 across models and periods (Supplementary Table S4).

Table 5. Goodness-of-fit statistic (adjusted coefficient of determination: adj-R²) for the three PLSR models to assess the explanatory capacity of feeding metrics for jointly describing the performance outcomes of growing pigs across both rearing periods.

PLS_Rall: model including all feeding metrics at both visit and meal levels (14 variables); PLS_RVIP: model including the most important feeding metrics based on the VIP analysis (6 variables); PLS_RFV: model including the most relevant feeding metrics only at the feeder visit level (3 variables).

¹From 105 to 125 days old.

²From 145 to 165 days old.

4 Discussion

This study started by describing the feeding behavior of commercial growing pigs under commercial-like rearing conditions as a first step to establish baseline reference values from which to draw conclusions about crossbreed-specific feeding patterns. The variability of feeding metrics between and within the different genetic backgrounds and its relationship with production performance were analyzed during both growing and fattening periods. Lastly, this study evaluated to what extent feeding metrics can explain production outcomes and feed efficiency traits of growing pigs from different commercial crossbreeds.

The daily feeding activity of growing pigs followed a bimodal distribution with two feeding peaks at dawn and dusk and overall little feed intake at night (Nielsen et al., 1995, Nielsen et al., 1996; Baumung et al., 2006; Poullet et al., 2022). The timing of feeding peaks was consistent across crossbreeds, while peak amplitude changed remarkably depending on the genetic background. Duroc-sired crossbreeds generally showed smooth feeding peaks and consistently had higher night feed intake than their counterparts throughout rearing. In contrast, the feeding peaks of Pietrain-sired crossbreds were sharp and became more pronounced during the fattening period. Previous studies showed similar hourly feeding patterns in pure-line pigs with no clear differences between Duroc and Pietrain breeds reported by Fernández et al. (2011), whereas Baumung et al. (2006) illustrated that Pietrain pigs ate less during the day and at night than Large White and Landrace pigs. Overall, Duroc-sired crossbreeds in our study ate more during the daytime and at night than the other crossbreeds, resulting in high daily feeder occupation time, particularly at young ages (spending more than one hour daily). This was consistent with results reported for Duroc and Pietrain pigs from pure lines (Fernández et al., 2011). From a management standpoint, acknowledging that some pig populations consume the majority of their daily feed intake in a narrow time window raises concerns about feeder space and feeding competition (Brown-Brandl et al., 2013). Pigs must indeed freely access the feeder trough during feeding peaks, particularly in subpopulations with poor feeding flexibility and strong feeding synchronization, to ensure feed intake and performance are not constrained by limited feeder access (Nielsen, 1999; Bornett et al., 2000). Previous research concluded that early signs of social constraint on feeder access appear at both the behavioral and performance levels in groups of 15 pigs per pen with a single-space feeder (Nielsen et al., 1995; Boumans et al., 2018). The pig-to-feeder ratio in our study was always below 14 pigs per feeder, and feeders did not reach full occupancy during feeding peaks.

Feeding metrics are interconnected at different levels, creating a feeding pattern that reflects how an individual consumes its daily feed (Nielsen, 1999; Baumung et al., 2006; Bus et al., 2025). Overall, a high frequency of feeding bouts (either meal or feeder visit) strongly correlated with small feed intake, short occupation time per feeding bout, and short intervals between feeding bouts. Within feeding bouts, feeder occupation time correlated negatively with feed intake and feeding rate. These relationships between feeding traits in our commercial crossbreeds were similar to previous results in pure lines (Labroue et al., 1997; Poullet et al., 2022; Piles et al., 2025). Interestingly, and regardless of feeding frequency, high daily feed intake was positively associated with fast feeding rate and large intake per feeding bout (in line with Labroue et al., 1997). Both can act as indicators of feeding motivation since high feeding rate, on its own, may not be enough to result in high daily feed intake (Nielsen, 1999; Bornett et al., 2000). In our study, the two crossbreeds with the fastest feeding rate had intermediate daily feed intake due to short feeding bouts or a low frequency of feeding bouts. Previous studies also reported that a feeding pattern with large feed intake per visit/meal, low feeding frequency, and fast feeding rate is associated with high daily feed intake, high body weight, and fast growth rate (Baumung et al., 2006; Fernández et al., 2011; Piles et al., 2025). Research examining feeding behavior over time showed that the feeding pattern of growing pigs was overall consistent for Large White, Landrace, and Duroc pure breeds while that of Pietrain pigs varied periodically depending on the feeding traits (daily feed intake and feeding rate remained stable) (Fernández et al., 2011). Our results also support the conclusion that the feeding pattern of growing pigs remained generally stable across genetic backgrounds during the rearing periods assessed in this study, despite numerical differences in feeding metrics as pigs aged (Nielsen, 1999; Baumung et al., 2006; Bus et al., 2024; Piles et al., 2025). Nonetheless, the within-individual variability of daily feed intake was high (15% on average) regardless of the crossbreed and rearing period.

The individual effect indeed largely explained the within-population variability of the feeding metrics, supporting the conclusion that traits intrinsic to the individual are associated with differences in their feeding behavior and their daily feed intake, which consequently impacts performance outcomes (Reyer et al., 2017; Herrera-Cáceres et al., 2020; Núñez et al., 2023). Among these traits, stress resilience and sensitivity, immunocompetence, social status and interactions, sensory capability, behavioral flexibility, motivational state, and wellbeing impact the between-individual variability of daily feed intake in growing pigs (Nielsen, 1999; da Silva et al., 2014; Maselyne et al., 2015; Boumans et al., 2015, Boumans et al., 2017, Boumans et al., 2018; Óvilo et al., 2025). Low daily feed intake is seen in pigs with poor physical health, those undergoing chronic stress, and those outcompeted for access to the feeder (Hoy et al., 2012; Boumans et al., 2018; Bus et al., 2021). In group-housing systems with reduced feeder space per pig, social effects can further jeopardize the feed intake and performance of those penmates left behind (Piles et al., 2025), and on-site strategies that help improve pen-level performance outcomes must also be considered. Therefore, a wide range of factors regulate the feeding behavior of pigs in the same environment (see reviews: Maselyne et al., 2015; Boumans et al., 2017; Bus et al., 2021), ultimately determining their feed intake in the short term (e.g., feeding bout) and long term (e.g., lifetime feed intake), which affects their rearing performance.

Within individuals, the repeatability of feeding phenotypes across days in both periods was high for feeder visit metrics, intermediate for the average daily feed intake, and low for meal metrics and night feed intake (in line with Bus et al., 2024). Feeder visit metrics may indeed be the most suitable for characterizing pigs’ overall feeding pattern (Maselyne et al., 2015) and evaluating its link with individual production traits as previously suggested (Nielsen et al., 1996; Baumung et al., 2006; Núñez et al., 2023). Meanwhile, meal metrics, night feed intake, and even the daily feed intake may be sensitive to daily fluctuations such as environmental stress or early signs of disease onset and poor welfare (Poullet et al., 2022; Tusingwiire et al., 2024; Bus et al., 2025). As mentioned above, previous studies also reported that high intra-individual variation in daily feed intake was common under conventional rearing conditions (Tusingwiire et al., 2024; Bus et al., 2025). When analyzing the variability of feeding metrics longitudinally, daily changes in meal metrics were sensitive enough to generate tail damage and lameness alerts (Bus et al., 2025), flag stressful conditions, and spot resilient individuals (Tusingwiire et al., 2024). Lastly, the proportion of night feed intake may be an indicator of feeding plasticity and compensatory feeding in growing pigs depending on genetic background (Bornett et al., 2000; Poullet et al., 2022; Bus et al., 2025). Nonetheless, investigating the implications of night feed intake requires further research to elucidate potential welfare concerns. Acknowledging the peculiarities of feeding metrics (meal versus feeder visit) and how they relate to pigs’ feeding behavior can help generate robust indicators of desirable individual traits (reviewed in Maselyne et al., 2015; Bus et al., 2025).

Feeding metrics can help elucidate how feeding patterns relate to rearing performance in growing pigs. Results from multivariate analyses highlighted that, among all considered feeding metrics, the leading traits explaining the greatest variability in performance outcomes were those related to feeding motivation: daily feed intake, feeding rate, and feed intake per feeding bout. To date, very few studies have directly investigated the link between feeding motivation and desirable production outcomes in growing pigs (Nielsen, 1999; Boumans et al., 2015, Boumans et al., 2017, Boumans et al., 2018), although previous research supported the key role of these feeding metrics as predictors of pig performance (Labroue et al., 1997; Herrera-Cáceres et al., 2020; Núñez et al., 2023). Among the performance outcomes evaluated in this study, body weight and growth rate were best explained by feeding rate, feed intake per feeder visit, and daily feed intake. Growing pigs with faster feeding rate and larger feed intake per feeding bout reached a higher body weight more quickly regardless of their feeding frequency or feeding time (Herrera-Cáceres et al., 2020; Núñez et al., 2023; Bus et al., 2024). Although this result is intuitive, it also indicates that jointly assessing these three feeding metrics from electronic feeders during rearing may help identify pigs that may fall behind growth rate expectations and require customized management and care to support attainment of target body weight, either for market requirements or gilt/boar development (Brown-Brandl et al., 2013). Interestingly, explanatory accuracy improved only minimally when other feeding and meal metrics were added to the model to explain pigs’ performance profile. Previous research already noted that feeding rate and feed intake were the most relevant feeding traits at the individual level (Nielsen, 1999; Maselyne et al., 2015), and that feeder visit metrics are more accurate for estimating feeding time and rate than meal metrics (discussed in Maselyne et al., 2015). Therefore, assessing these key feeding metrics at the feeder visit level can provide practical insights to support specific management decisions in swine facilities with electronic feeders (Hoy et al., 2012; Brown-Brandl et al., 2013; Reyer et al., 2017). In addition, Núñez et al. (2023) demonstrated that the selection response for growth rate and feed efficiency improved when these feeding metrics were added to multivariate models to support breeding decisions.

Feed-driven pigs overall ate more and accumulated more fat than their counterparts with low appetite, but subcutaneous fat depot was only moderately explained by the feeding pattern in this study. Previous research noted that night feeding may lead to higher fat accumulation and lower feed efficiency in disturbed pigs for a week (Van Erp et al., 2020), while our results do not indicate a clear association between night feeding and production and efficiency metrics. Nonetheless, our goal was to assess the overall link between feeding patterns and performance outcomes across different commercial crossbreeds of growing pigs, and some of the relationships may be breed-specific. Additionally, feed efficiency traits were poorly explained by feeding metrics, particularly during fattening, despite the positive associations observed between feeding patterns and feed efficiency metrics previously described (Poullet et al., 2022; Núñez et al., 2023; Piles et al., 2025). Improving feed efficiency in growing pigs requires a multifactorial approach that integrates feeding patterns, digestibility, nutrient utilization, and genetic potential for optimal growth and low-fat deposition (Patience et al., 2015; Gilbert et al., 2017). From a feeding behavior standpoint, eager eaters can be less feed efficient than their counterparts due to high feed intake, although they grow faster and are heavier (Núñez et al., 2023; Piles et al., 2025). Indeed, swine selection programs solely focused on feed efficiency metrics often result in individuals with low daily feed intake that take longer to reach market body weight (Gilbert et al., 2017; Reyer et al., 2017). So, within this interplay, feeding patterns that promote daily feed intake in addition to other factors must be considered jointly to improve feed efficiency without compromising productive outcomes (Ewaoluwagbemiga et al., 2021). Our results support the role of feeding patterns, particularly feeding rate and feed intake per feeder visit, as proxies of production outcomes in commercial growing pigs, with straightforward implications for advancing swine management and breeding programs through precision livestock farming practices.

Results from this study support the conclusion that key feeding behavioral traits can help explain differences in the body weight and growth rate of growing pigs from multiple commercial crossbreeds. The extent to which feeding patterns can explain the performance profile of growing pigs may be breed- or crossbred-specific, and within-crossbreed analyses may indeed yield greater explanatory ability than analyzing different groups of animals together. Yet, a large population of pigs from the same genetic background is needed for this type of analysis, while our approach can help describe common feeding patterns that may explain performance outcomes across commercial crossbreeds. Additionally, our analysis can help elucidate how feeding traits are associated with the current performance profile to flag individuals falling behind performance objectives. Still, gauging to what extent feeding patterns during early development can predict performance outcomes at the end of fattening is a follow-up research question arising from this study. Lastly, this research was conducted under commercial-like conditions with small, single-sex groups with one feeder per pen. Validating these results in commercial pig farms with larger group sizes and more feeders per pen is advised due to greater environmental complexity and more social interactions that impact feeder use.

5 Conclusions

The feeding activity, feeding pattern, and eating intensity are impacted by the genetic background of the crossbreed, with a strong individual effect and high intra-individual repeatability, particularly in feeding metrics at the feeder visit level. Feed intake per feeder visit, feeding rate per feeder visit, and daily feed intake jointly served as core descriptors of production performance and, to a much lesser extent, of feed efficiency traits. These results support the conclusion that real-time records from electronic feeders can help characterize the feeding pattern of growing pigs and gauge their performance profile during rearing. Meanwhile, meal metrics and other feeding traits such as night feed intake and daily feed intake showed high intra-individual variability across days. Results from this study suggest that analyzing the feeding behavior of growing pigs can support and refine the decision-making process behind smart livestock practices in swine farming and help tailor management and breeding decisions to ongoing challenges.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fanim.2026.1938801/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Juliana Petrini, Embrapa Gado de Leite, Brazil

Reviewed by: Wim Gorssen, ETH Zürich, Switzerland

Moyosore Joseph Adegbeye, University of Africa, Nigeria

Ivan Bianchi, Instituto Federal Catarinense, Brazil