On farm precision feeding of gestating sows based on energy and amino acids on farrowing performances and feeding behavior over 3 consecutive gestations

Abstract

Gestating sows are often fed a single diet throughout their gestation cycle, leading to situations of nutrients deficiency or excess at the individual level. The purpose of this study was to characterize, over 3 consecutive cycles, the impact of a precision feeding (PF), i.e., dietary supplies adjusted at individual level in terms of quantity (energy intake) and quality (standardized ileal digestible (SID) lysine (Lys) intake), on gestating sows’ productive performance, feeding behavior, environmental outputs, and health status compared with sows fed a conventional feeding (CF, i.e., fixed SID Lys intake). At the start of the trial, 2 batches of 20 Landrace × Large White gestating sows were allocated to one of the 2 feeding strategies (PF vs. CF), based on their parity (1, 2, or 3 and more), body weight, and backfat thickness (BT) 3 days after insemination. The PF strategy consisted in mixing with automatic feeders 2 iso-energetic diets (9.8 MJ/kg of net energy with 3.3 and 8.5 g SID Lys/kg, respectively) in variable proportions at individual and daily levels, whereas for the CF strategy these proportions remained constant throughout gestation (73% and 27%, respectively, resulting in an SID Lys concentration of 4.7 g/kg). Sows were followed over 3 consecutive gestations and the sows remained allocated to the same strategy throughout the study. Some sows were culled before the end of the study and were replaced by other sows who therefore performed only 1 or 2 gestations during the study. Thus, 106 gestations and lactations from 51 sows were fully studied and their data analyzed. The PF strategy allowed the sows to reach more closely the expected BT values at farrowing across cycles than the CF strategy (P < 0.001). The PF sows consumed 16% less SID Lys per gestation than the CF sows (P < 0.001), resulting in a 4% improvement in N efficiency (P < 0.001), with no impact on performance at farrowing (P > 0.10). The sows consumed their daily ration in a single visit whatever the feeding strategy (P = 0.41), but CF sows spent more time in the feeder in cycles 2 and 3 (P < 0.001). Thus, compared to the CF strategy implemented by farmers, the individual and daily nutritional supplies implemented with the PF strategy were more efficient in enabling sows to achieve body condition objectives at farrowing over the long term, also with a reduced SID Lys intake and an improved N efficiency without negative impact on farrowing performances.

Precision feeding of gestating sows over 3 consecutive cycles improved the body condition at farrowing (by a better adequacy of energy supplies) and improved nitrogen efficiency (by reducing excess standardized ileal digestible lysine intake), without negative impacts on sows’ performance, feeding behavior or health but with a better potential longevity.

Introduction

In commercial farms, sows are frequently fed the same single diet during the whole gestation and more rarely with a transition diet or a mix of gestation and lactation diets at the entrance of the farrowing room (Quiniou et al., 2018). Nutritional requirements assessed through a factorial approach depend on parity, initial body condition, stage of gestation (Noblet et al., 1993; Dourmad et al., 2008), and expected target values at farrowing for body weight (BW), backfat thickness (BT), prolificacy, and piglets’ birth weight (Young et al., 2004; Quiniou, 2016). The minimum ratios between essential amino acids and energy requirements (Noblet et al., 1990; Gaillard et al., 2019) increase with the stage of gestation but they are not similar for all sows. It means that with a single diet, formulated for a given ratio between essential amino acids and energy, the adequacy between nutritional requirements and supplies is not achieved for each sow in the herd each day (Solà-Oriol and Gasa, 2017).

Accounting for differences in energy requirement is more and more frequent in commercial farms, but most often only at an average level based on parity and classes of BT measured at weaning or after insemination. Then only few different dietary energy allowances are implemented in practice for sows of the same batch, with decisions taken by farmers based on observations instead of measurements. In addition, as the increase in maternal BW and BT from the insemination to the farrowing depends on the amount of dietary energy supplied, unadjusted energy delivery is one of the factors that contribute to the inter-individual variability of sow’s characteristics body condition (BW and BT) at farrowing, which has important consequences on sows and piglets’ welfare and performances when sows are either too fat or too lean. In this case, for overfat sows the farrowing process is impaired and results in an increased rate of stillborn piglets. In addition, these sows have a poor appetite during the lactation and mobilize more their body reserves with consequences on reproduction after weaning and increased risk of being culled early (Niemi et al., 2017). On the contrary, too lean sows eat more but they farrow lighter piglets and produce less milk, which results in an increased risk of mortality and low body weight of piglets at weaning (Quiniou, 2016).

The aim of precision feeding (PF) is to improve the adequacy between the nutritional supplies and the requirements. For the present study, the adjustment occurs individually and daily. Through an individualized quantity and quality of the feed, the aim is to reach target values of sow characteristics at farrowing in terms of BW, BT, and litter weight at the individual level, to reduce the inter-individual variability of BT at farrowing at the batch level and to improve the efficiency of nutrient utilization towards reduced situations of deficiency and spillage at the sow’s level. In group-housed gestating sows, the assessment of nutritional requirements of each sow on a daily basis relies on the use of new technologies like automatic feeders equipped with multiple feed hoppers; they identify each sow and can deliver an individually and daily adapted mix of different diets. The quantity allocated per day per sow is mainly related to energy requirement (Gaillard et al., 2020). The daily adjustment of the dietary quality refers mainly to the amino acid content (based on lysine (Lys) requirement for the present study). Indeed, a recent trial carried out over 1 gestation, the PF strategy designed by Gaillard and Dourmad (2022) reduced the Lys intake by around 25% and nitrogen excretion by 19% compared to a conventional feeding (CF) strategy (1 diet for all the sows and all the gestation), without any negative impact on litter characteristics at birth. Furthermore, feeding behavior seemed slightly affected by the feeding strategy, as it resulted in the same number of feeding visits, but with + 0.5 non-feeding visit at the end of gestation for PF sows (Gaillard and Dourmad, 2022). The long-term effects (i.e., over several reproductive cycles) of such a PF strategy have not yet been studied. Therefore, the aim of this present paper was to estimate the impact of a PF strategy adjusted on energy and amino acids across 3 consecutive gestations on sow’s performances and behavior compared with a CF defined thereafter.

Material and Methods

Experimental farm

The protocol for the experiment was reviewed and approved by the Ethics Committee in Animal Experimentation (Rennes, France, reference APAFiS #24663). The experiment was conducted in accordance with the French legislation for commercial pig production and experimental animal care. It was performed between June 2021 and September 2022, at the Pig Physiology and Phenotyping Experimental Facility (UE3P), Saint-Gilles, France (doi: 10.15454/1.5573932732039927E12).

Experimental design

Initially, 2 batches of 20 Landrace × Large White gestating sows were involved in the experiment. At the beginning of the study (cycle 1), the sows were allocated to 1 of the 2 feeding strategies (PF vs. CF), based on their parity (1, 2, or 3 and more), BW, and BT 3 days after insemination (i.e., initial BW and BT). They were followed during 3 consecutive gestations and remained allocated to the same strategy throughout the study. Some sows were culled after weaning after the first or second gestation(s) of the trial under the performance-based replacement policy of the experimental farm, i.e., before the end of the study. They were replaced by primiparous sows who could perform only 1 or 2 gestations during the study. All sows that completed at least 1 gestation cycle were used for the analyses.

Housing conditions, equipment, and management

Sows were moved to the gestation room and group-housed 3.1 ± 0.5 d after insemination. Each batch of 20 gestating sows was housed in a room of 7.5 × 8.0 m, on concrete floor with fresh straw distributed every morning. The feed was delivered using 2 automatic feeders (Gestal, JYGA Technologies Inc., QC, Canada), each equipped with 2 hoopers allowing the delivery of a mixture of 2 diets in individualized proportions. The 2 automatic feeders were connected to the same computer, so that each sow could access its daily ration using either one or both automatic feeders. Sows had free access to water using 2 drinking troughs with an integrated sow scale (Asserva, Lamballe, France). Ventilation set point temperature was 21 °C in the gestating rooms. The sows were moved to the farrowing room about 1 week (106.9 ± 0.7 d) before the planned farrowing date and farrowed at 115.4 ± 1.2 d of gestation after the first insemination.

Diets and feeding strategies

Two diets were formulated on an iso-net energy basis (9.8 MJ/kg) and different amino acid concentrations. As Lys is in most cases the first limiting essential amino acid in pigs, the diets were formulated in such a way that the minimum levels between standardized ileal digestible (SID) essential amino acids and SID Lys follow the fixed ideal protein profile for gestating sows (van Milgen and Dourmad, 2015). The SID Lys concentration was 3.3 and 8.5 g/kg in the GL and GH diets, respectively (Table 1). For the PF strategy, proportions of GL and GH diets were changed over the gestation on an individual and daily basis according to the assessed Lys requirement of each sow. Both diets were mixed in constant proportions (73% GL and 27% GH, respectively) for the CF strategy, resulting in a SID Lys concentration of 4.7 g/kg. From the entrance into the farrowing room to farrowing, all sows received a standard gestation diet (so-called GS diet), which formulation corresponded to the mix of GL and GH delivered before to CF sows. From the day after farrowing up to weaning, all sows received a blend made of 34% GL diet and 66% of a lactation diet available on the farm which presented a high concentration in amino acids (so-called LH diet with a 10.6 g SID Lys/kg), so that the SID Lys concentration in the resulting diet was close to the value observed in a standard lactation diet. Lactation feed allowance followed a progressive rationing plan which is different for primiparous and multiparous sows. Details on ingredients and chemical composition of diets GL, GH, GS, and LH are shown in Table 1.

Table 1.

Ingredients and chemical composition of sows’ diets (on an as-fed basis)

Diet type1	GL	GH	GS	LH
Ingredients, g/kg
Barley	400	256	353	230
Wheat	254.1	272.6	230.0	230.0
Maize	100	120	100	120
Wheat bran	150	100	150	100
Soybean meal 48%	—	162.6	70.8	219.9
Vegetable oil	20	19.9	20	20
Sugar beet pulp	50	—	50	—
Cane molasses	—	30	—	30
Limestone	11.5	15.3	12.5	18.0
Monocalcium phosphate	2.78	6.6	2.5	9.0
Salt	4.5	4.5	4.5	4.5
Premix2	5	5	5	5
l-Lysine 50	1.07	4.59	0.65	6.28
dl-Methionine	—	0.54	—	1.40
l-Threonine	—	0.77	—	1.45
l-Tryptophan	—	—	—	0.66
l-Valine	—	0.55	—	2.76
Phytase	0.05	0.05	0.05	0.05
Acidifying agent	1	1	1	1
Calculated chemical composition, g/kg
Dry matter	879.9	874.9	880.5	875.4
Crude protein	105.7	160.4	130.0	181.5
Ash	46.6	59.4	50.6	66.8
Crude fat	41.0	39.9	41.1	39.9
Crude fiber	49.3	38.4	50.3	39.3
NDF	198.3	151.8	193.9	147.9
ADF	61.4	48.2	62.6	49.2
ADL	13.1	10.3	12.9	10.0
Lys	4.4	9.6	5.9	11.8
Met	1.8	2.9	2.1	3.8
Cys	2.4	3.0	2.7	3.2
Met+Cys	4.2	5.9	4.8	7.0
Thr	3.5	6.3	4.5	7.8
Trp	1.3	1.9	1.6	2.4
Ile	3.7	6.3	4.9	7.2
Val	5.1	7.7	6.3	9.4
SID3 Lys	3.3	8.5	4.7	10.6
SID Met	1.5	2.6	1.8	3.5
SID Cys	2.0	2.6	2.2	2.8
SID Met + Cys	3.5	5.2	4.0	6.3
SID Thr	2.6	5.3	3.5	6.7
SID Trp	1.0	1.6	1.3	2.1
SID Ile	3.0	5.5	4.1	6.3
SID Val	4.0	6.5	5.1	8.2
Total calcium	7.6	9.8	8.2	11.5
Total phosphorus	4.6	5.5	4.7	6.2
Digestible phosphorus	2.6	3.4	2.6	4.0
Calculated energy, MJ/kg
Digestible energy	13.3	13.7	13.5	13.8
Metabolizable energy	12.8	13.1	12.9	13.1
Net energy	9.8	9.8	9.7	9.7

¹Corresponding to a gestation diet with a low Lys content (GL) or a high Lys content (GH), a standard gestation diet (GS), and a lactation diet with a high Lys content (LH).

²Provided the following amounts of vitamins and trace elements in units per kilogram of feed: vitamins: A = 10,000 IU, D3 = 1,500 IU, E = 45 IU, biotin = 0.2 mg; trace minerals: Cu (E4 sulfate) = 10 mg, Zn (E6 oxide) = 100 mg, Mn (E5 oxide) = 40 mg, Fe (E1 carbonate) = 24 mg, Fe (E1 hepta sulfate) = 28.2 mg, Fe (E1 mono sulfate) = 27.5 mg, I (E2 calcium iodate) = 0.60 mg, Se (E3 sodium selenite) = 0.25 mg.

³SID, standardized ileal digestible.

Lys, lysine; Met, Methionine; Cys, Cysteine; Thr, Threonine; Trp, Tryptophan; Ile, Isoleucine; Val, Valine.

Based on the parity (1, 2, or 3 and more), BW and body conformation assessed visually by the experimental farm technician 3 d after insemination, each CF sow was allocated to a small, medium, or high feeding level. A 3-phase strategy, following an high-low-high feeding pattern, was used, with phase 1 between insemination and day 35 of gestation (recovery of maternal reserves mobilized during the previous lactation), phase 2 between days 36 and 85 of gestation (energy supply close to maintenance requirement: 0.440 MJ/kg BW^0.75; Noblet et al., 1990), and phase 3 between day 86 of gestation and entrance into the farrowing room. Details for each group are available in Supplementary Table S1. For PF sows, individual total feed allowance and dietary proportions of diets GL and GH were calculated based on energy and Lys requirements estimated by an enhanced version of the InraPorc model (Dourmad et al., 2008) updated for gestating sows (Gaillard et al., 2020). The energy requirement was assessed for the whole gestation period based on individual age, BW, and BT at insemination, and target values for litter birth weight, BT, and BW at farrowing after calibration of a sow’s profile. For this purpose, historical performances of the herd were used to fit a BW curve according to age at farrowing, which was used in our study to estimate the expected maternal BW at farrowing. Expected litter size and litter weight at birth were considered according to the parity. Thereafter, the total feed allowance for the whole gestation was calculated from the total energy requirement divided by the dietary energy content. Then, it was partitioned per day in a 2-phase program with daily feed allowance increased by 500 g after day 86 of gestation (corresponding to phase 3 for CF) compared to previous allowance (from insemination to day 85 of gestation), to match the increase in energy requirement for fetal growth.

Measurements, data collection, and calculations

Each sow was equipped with 2 radio-frequency identification (RFID) ear-tags, 1 for the identification by the automatic feeders and 1 for the identification by the drinking troughs. Both equipment recorded each individual visit (duration and quantity eaten or drunk) each day. Overlapping visits at the same automaton and visits without duration were removed from the database. These data were used to characterize feeding and drinking behaviors of sows on a daily basis. Sow’s BW and BT were measured manually at the beginning of the gestation (around the day 3 of gestation) during the transfer from the service room to the gestation room, at the end of the gestation (around day 107 of gestation) when sows were moved to the farrowing room, and at weaning (on Thursday, after around 4 wk of lactation). An individual electronic scale (Schippers, the Netherlands, precision ± 0.5 kg) was used. An ultrasound portable device (ECM Imago, France) was used to measure the BT at the level of the last rib 6.5 cm from the midline on the right and left sides of the sow, and the average of both values was considered in the study.

In the gestation room, health status was monitored once per week by trained observers. The dirtiness of different body parts (on both sides instead of only one), bursitis, lameness (by observing the gait), or any other health problems such as coughs or injuries were noted using the Welfare Quality protocol (Welfare Quality®, 2009) as the reference method.

At farrowing, individual birth weight, and status (alive, dead, or mummified) of piglets were recorded. Depending on the number of functional teats, some piglets were cross-fostered in between the sows independently of the feeding strategy. From week 3 onward, piglets had access to creep feed. Piglets were weighed at weaning, i.e., at around 4 weeks of age. Piglets’ mortality between birth and weaning was also recorded, and piglets weighed at that time. Born alive and stillborn piglets were considered to calculate the litter size at birth, average birth weight per piglet and per litter. The average daily gain (ADG) of the litter was calculated based on BW gain of piglets alive at weaning and the BW gain of piglets that died before, divided by the duration of the lactation. The ADG per piglet was obtained from the ADG of the litter, divided by the suckled litter size, i.e., number of weaned piglets and dead piglets taken into account on a prorate temporis.

Feed costs were estimated on the basis of the average price of the diets observed over the last semester 2023: 322.23 ($348.00), 393.05 ($424.49), 339.72 ($366.90) €/ton, and 371.09 ($400.78) €/ton for diets GL, GH, GS, and LH, respectively.

Feed intake, performance at farrowing, and sow’s initial BW and BT were averaged by parity and feeding strategy and used to calculate N and P retention and excretion with the InraPorc model.

Statistical analyses

Statistical analyses were carried out using the R studio software (R Core Team, 2022). A linear mixed model with the lme function of the nlme package (Pinheiro and Bates, 2000) was used to assess the effect of the feeding strategy on the variables of interest followed during each gestation: sow characteristics (BW and BT), litter characteristics (number of alive piglets, birth weight, ADG) intakes (feed, nitrogen, SID Lys), and nutrient (N and P) retention and excretion. The sow was considered as the experimental unit. The feeding strategy (CF vs. PF), the cycle (1, 2, or 3), and their interaction were used as fixed effects. Parity of the sow at the start of the trial was divided into 3 classes (1: parity 1, 2: parity 2, 3+: parity 3 and more) and used as a covariate. The sow was considered as a random effect. An autocorrelation structure of order 1 (corAR1 function) was used to take into account the correlations between results for the consecutive gestations of a given sow. To analyze daily data, the stage (day) of gestation was taken into account as a fixed effect as well as its interactions with other fixed effects. A general linear mixed model with glmer function of the lme4 package (Bates et al., 2015) was used to test the effect of the feeding strategy on binary, counting and percentage variables such as those related to health (occurrence of lameness and bursitis and percentage of dirtiness) and feeding behavior (number of visits, time spent in the feeder). The same fixed effects were used as for the linear mixed model. Binary and percentage data were normalized with Binomial transformation. Counting data were normalized with Poisson transformation. For pairwise comparisons, the Tukey’s test for multiple comparisons of means was used. Assumptions of regression models (linearity, normal distribution of error, homoscedasticity of errors and independence of the observations thanks to the autocorrelation structure) were evaluated for each model used. To compare the measured values to the expected values (BW, BT, litter weight at farrowing) and to compare cumulative data over 3 cycles, a Student’s test or a Welch test was used depending whether the variances were equal or not for both feeding strategies, respectively.

Residual standard deviation was used and calculated as follows:

$${\displaystyle \begin{array}{l}{\displaystyle \sqrt{\frac{{\displaystyle \sum _{i=1}^n}{(Y_i-{{\displaystyle \widehat{Y}}}_i)}^2}{(n-p)}}}\end{array}}$$

with n is the sample size, p is the number of parameters to estimate, Y_i is the observed value and ${\widehat{Y}}_i$ predicted value (Gu et al., 1992). The effect was considered significant when P value was <0.05 and as a tendency when P value ranged between 0.05 and 0.10.

Results

Over the 3 reproductive cycles of the 2 batches, 118 gestations were initiated from 58 sows, of which 11 failed because the sows were not confirmed 25 d after insemination. In addition, one PF sow died because of rectal prolapsus at 35 d of gestation in cycle 1. At the end, 106 gestations and lactations from 51 sows were fully studied. The number of sows in each parity class, in CF and PF, is the following: 13 and 10 (parity 1), 4 and 6 (parity 2), and 11 and 7 (parity 3 and more) respectively. During the trial, besides the sows removed from the experiment due to death or not confirmed, 8 CF sows and 2 PF sows were culled after weaning (parity 5.8 on average) mainly due to locomotive disorders and poor performances in maternity. In total, 20% of each batch was therefore replaced during the experiment. A total of 11 PF sows and 9 CF sows completed 3 cycles of gestation/lactation, 6 PF and 9 CF completed only 2 cycles and 4 PF and 12 CF only 1 cycle.

Nutritional supplies

On average for all the sows and cycles, observed feed intake and SID Lys intake over the whole gestation were in agreement with supplies implemented (Table 2). Observed daily feed intake and SID Lys intake were 0.04 kg/d and 0.2 g/d lower than expected, respectively (P < 0.001). As a result of the application of feeding strategies, PF sows ate daily 0.20 kg more than CF sows in cycle 1 (P = 0.002) and this difference doubled for cycles 2 and 3 (P < 0.001). A PF sow ate + 22 kg, + 43 kg and + 46 kg more than a CF sow, over the consecutive 3 cycles (P = 0.002). On average for all cycles, PF sows consumed less SID Lys, both on a daily basis (−1.7 g/d) and during gestation (−168 g SID Lys) (P < 0.001). Over the first 12 weeks of gestation, PF sows consumed 2.2 g SID Lys/d less than CF sows (Figure 1; P < 0.001). Over the last 2 wk in the gestation room, SID Lys intake of PF and CF sows was similar (on average 14.5 g/d). Across all weeks in the gestation room, the proportion of GL diet was significantly different between primiparous and multiparous sows and was significantly different between parity 2 and parity 3 and more, from week 3 onward (Supplementary Figure S1; P < 0.001).

Table 2.

Cumulated and daily feed intake and SID Lys intake regarding feeding strategy (CF vs. PF) through the different cycles

Strategy	Cycle						RSD1	P-value
	1		2		3			Strategy2	Cycle2	Strategy × Cycle2	Observed vs. Expected3
	CF	PF	CF	PF	CF	PF
Number of sows	30	21	18	17	9	11
Expected in the gestation room, per sow
Feed, kg/d	2.59	2.77	2.61	2.99	2.67	3.09	0.22	—	—	—	<0.001
Feed, kg	263	282	266	305	272	314	4.6	—	—	—	0.365
SID Lys, g/d	12.2	10.3	12.3	10.6	12.6	10.5	0.90	—	—	—	<0.001
SID Lys, kg	1.24	1.05	1.25	1.08	1.28	1.06	0.02	—	—	—	0.350
Observed in the gestation room, per sow
Duration, d	101	102	102	102	102	102	0.1	0.966	0.913	0.876	—
Feed, kg/d	2.55a	2.75b	2.57a	2.99c	2.59a	3.05c	0.26	0.002	0.741	<0.001	—
Feed, kg	258a	280b	262a	305c	264a	310c	27	0.197	0.452	0.020	—
Diet GL, kg	188a	258b	191a	292c	193a	305c	30	<0.001	0.477	<0.001	—
Diet GH, kg	70a	22b	71a	13c	71a	5c	12	<0.001	0.907	<0.001	—
Nitrogen, kg	4.98	4.93	5.05	5.27	5.10	5.29	2.9	0.421	0.471	0.137	—
SID Lys, g/d	12.0a	10.2b	12.1a	10.6c	12.2a	10.3bc	1.3	<0.001	0.739	<0.001	—
SID Lys, kg	1.22a	1.04b	1.23a	1.08b	1.24a	1.05b	0.10	<0.001	0.529	0.941	—
Feed intake in the farrowing room 4
Duration before farrowing, d	8.4ab	9.1b	8.4ab	8.5ab	8.3ab	8.2a	1.2	0.037	0.744	0.073	—
Gestation diet, kg/d	2.51	2.50	2.71	2.74	2.62	2.56	0.27	0.707	0.010	0.719	—
Gestation diet, kg	21.4ab	22.6ab	22.6ab	23.1b	21.8ab	20.9a	2.9	0.039	0.541	0.075	—
Duration of lactation, d	28.1ab	27.4a	28.3b	27.9ab	28.7b	28.6b	1.8	0.037	0.150	0.168	—
Lactation diet, kg/d	6.61a	6.98b	7.06b	7.07b	6.74ab	6.48a	1.30	0.046	0.663	0.057	—
Lactation diet, kg	185	191	200	197	193	186	38	0.204	0.362	0.163	—

¹RSD, residual standard deviation.

²Fixed effects of the variance analysis model, with initial parity of the sow as a covariate.

³Welch test for daily data and Student test for gestation data.

⁴The gestation diet was delivered before farrowing, and the lactation diet thereafter.

a, b, c, and d values within a row that do not share a common superscript are significantly different (P ≤ 0.05).

CF, conventional feeding; PF, precision feeding; SID Lys, Standardized ileal digestible lysine; GL, gestation diet with a low Lys content GH, gestation diet with a high Lys content.

Figure 1.

Weekly evolution of the standardized ileal digestible lysine (SID Lys) intake during gestation regarding cycle (1, 2, or 3) and strategy (CF vs. PF). The ‘cloud’ corresponds to the average standard deviation of the 3 cycles for each strategy. P value of the Student’s test of the effect of the feeding strategy for each week of gestation was performed. (***P < 0.001; NS = P > 0.10). CF: conventional feeding; PF: precision feeding.

Over the 3 gestation cycles, PF sows consumed 109 kg more feed on average than CF sows and 550 g less SID Lys (Supplementary Table S2). From the arrival in the farrowing room to parturition, there was no difference in daily feed intake between the 2 feeding strategies (Table 2; P = 0.71). The feeding strategy affected daily feed intake during lactation (P = 0.05). On average, PF sows consumed 0.37 kg/d more than CF sows in cycle 1, but 0.26 kg/d less in cycle 3, the same average daily feed intake was observed in cycle 2 (interaction Strategy × Cycle: P = 0.06).

Performance of the sow and the litter

As presented in Table 3, there was no significant effect of the feeding strategy on the initial BW and initial BT (P > 0.10). The initial BW increased with average parity from cycle 1 to cycle 3, by +21 kg on average across cycles (P < 0.001). The interaction between the feeding strategy and the cycle was significant on final BW (P = 0.05) and final BT (P < 0.001), with PF sows being heavier and fatter in cycle 2 (+ 15 kg and + 1 mm compared to CF sows, respectively), and cycle 3 (+ 20 kg and + 3 mm compared to CF sows, respectively). Unlike final BW which was well achieved (P = 0.60), observed final BT was significantly different from expected BT (P = 0.002), and BT of CF sows was 4 mm lower than the target value in cycle 3 (interaction Strategy × Cycle: P < 0.001).

Table 3.

Observed and expected body weight (BW) and backfat thickness (BT) at gestation group-housed entrance and at maternity entrance regarding feeding strategy (CF vs. PF) through the different cycles

Strategy	Cycle						RSD1	P-value
	1		2		3			Strategy2	Cycle2	Strategy × Cycle2	Observed vs. Expected3
	CF	PF	CF	PF	CF	PF
Number of sows	30	21	18	17	9	11
Parity	2.8	2.9	3.4	3.5	3.7	4.7	0.5	0.899	<0.001	0.666	—
Age, days	566	566	656	652	667	828	4	0.602	<0.001	0.537	—
Expected 4
Final BW5, kg	271	267	289	288	300	305	11	—	—	—	0.601
Final BT, mm	19.0	19.0	19.0	19.0	19.0	19.0	3.7	—	—	—	0.002
Observed
Initial6 BW, kg	198	200	218	219	220	242	10	0.517	<0.001	0.924	—
Initial6 BT, mm	15.3	15.4	14.3	14.5	12.8	14.5	4.4	0.756	0.139	0.444	—
Final7 BW 5, kg	268a	274ab	283b	299c	288bc	320d	17	0.879	<0.001	0.045
Final7 BT, mm	17.9b	18.4b	17.7b	19.7b	14.6a	17.9b	1.4	0.450	0.002	<0.001

¹RSD, residual standard deviation.

²Fixed effects of the variance analysis model, with initial parity of the sow as a covariate.

³Welch test for BT and Student test for BW.

⁴Expected value simulated with InraPorc model.

⁵BW: maternal BW + conceptus BW.

⁶Initial = at the transfer between the service room and the gestation room (3.1 ± 0.5 d of gestation).

⁷Final = at the transfer between the gestation room and the farrowing room (106.9 ± 0.7 d of gestation).

a, b, c, and d values within a row that do not share a common superscript are significantly different (P ≤ 0.05).

CF, conventional feeding; PF, precision feeding; BW, body weight; BT, backfat thickness.

Additional Supplementary Table S3 shows the results of sows followed over the 3 gestations. The PF sows were older than CF sows from the beginning of the study, with an initial parity of 2.7 vs. 1.7. Subsequently, PF sows were 20 kg heavier at the start of each gestation. Throughout the study, final BT of PF sows remained stable and close to expected BT, while BT of CF sows decreased and became 4 mm below the expectation at the end of cycle 3 (interaction Strategy × Cycle: P = 0.004).

At farrowing (Table 4), no significant effect of the cycle and the feeding strategy was observed on litter size (17.5 ± 4.5 total born piglets on average) and individual birth weight (1.43 ± 0.27 kg on average). Litter weight increased by 2.0 kg per cycle (P = 0.02). Sows farrowed 2.5 piglets more than expectations based on past results collected in the herd (P < 0.001) resulting in a heavier total litter weight than expected (+ 1.5 kg, P = 0.01). Over the 3 gestation cycles (Supplementary Table S4), PF sows delivered 2.6 more live piglets and 1.1 more stillborn piglets, with a cumulated additional litter weight of 5.5 kg compared to CF sows, but this difference was not significant (P > 0.10).

Table 4.

Expected and observed litter characteristics at birth and performance observed until weaning regarding feeding strategy (CF vs. PF) through the different cycles

Strategy	Cycle						RSD1	P-value
	1		2		3			Strategy2	Cycle2	Strategy × Cycle2	Observed vs. Expected3
	CF	PF	CF	PF	CF	PF
Number of sows	30	21	18	17	9	11
Expectations at birth, per litter
Total born piglets	14.7	14.6	15.1	15.0	16.0	15.9	4.8	-	-	-	<0.001
Birth weight, kg	21.8	21.5	23.5	23.4	24.7	24.5	1.4	-	-	-	0.014
Observations
At birth
Alive piglets	15.4	16.0	16.5	16.1	17.2	19.6	6.0	0.712	0.324	0.411	-
Stillborn piglets	1.1	1.0	1.5	1.2	0.8	0.8	1.0	0.685	0.442	0.406	-
Total born piglets	16.5	17.0	18.0	17.3	18.0	20.5	6.2	0.742	0.238	0.538	-
Birth weight, kg/piglet	1.35	1.39	1.44	1.55	1.52	1.40	0.25	0.337	0.167	0.436	-
Birth weight, kg/litter	21.5	23.0	25.2	26.2	26.2	28.6	7.5	0.741	0.022	0.714	-
After cross-fostering
Initial litter size	12.8	13.3	13.6	13.5	12.6	13.6	2.8	0.817	0.457	0.980	-
Litter size at weaning	11.4	11.9	12.3	12.2	11.2	11.3	2.2	0.688	0.534	0.983	-
Final4 weight, kg/piglet	7.76	7.84	8.37	8.78	8.31	7.97	0.97	0.235	0.054	0.327	-
Final4 weight, kg/litter	87.7	93.6	103.0	107.0	91.8	90.1	25.8	0.268	0.377	0.559	-
Average daily gain from birth to weaning
Piglet, g/d	233	241	252	264	252	246	51.4	0.212	0.078	0.299	-
Litter, kg/d	2.81	2.99	3.14	3.31	2.86	2.80	0.84	0.301	0.864	0.620	-

¹RSD, residual standard deviation.

²Fixed effects of the variance analysis model, with initial parity of the sow as a covariate.

³Welch test.

⁴Final = at weaning.

CF, conventional feeding, PF, precision feeding.

After cross-fostering (Table 4), litter size averaged 13.2 ± 3.2 piglets of which 11.8 ± 2.1 were weaned for both feeding strategies (P > 0.10). Litter size and BW of piglets and litters at weaning and ADG per piglet and per litter from birth to weaning were similar for both strategies (P > 0.10). The cycle tended to influence the average weight of piglets at weaning (P = 0.05) and their ADG (P = 0.08), but not the corresponding values at litter scale. For the sows followed over 3 cycles (Supplementary Table S4), no significant effect of the strategy was observed on their lactation performances (P > 0.10).

Feeding and drinking behaviors

Sows ate their daily ration in a single visit, regardless of the cycle or feeding strategy (Table 5; P > 0.10). Feeding visits represented 18% of total visits to the feeder and 55% of the time spent in the feeder. With the succession of gestation cycles, CF sows spent more and more time in the feeder for feeding visits whereas it remained stable for PF sows (interaction Strategy × Cycle: P < 0.001). Regarding the gestation cycle, CF sows spent more time in the feeder (i.e., + 1.7, + 12.7, + 8.6 min/d) than PF sows in cycles 1, 2, and 3, respectively (interaction Strategy × Cycle: P < 0.001). The feeding strategy affected the number of non-feeding visits, with + 0.3, −0.2 and + 0.9 additional non-feeding visit from PF than CF sows in cycles 1, 2, and 3, respectively (interaction Strategy × Cycle: P < 0.001).

Table 5.

Effects of the feeding strategy (CF vs. PF) through the different cycles on feeding and drinking behaviors defined through the number of visits and time spent in the automaton per day

Strategy	Cycle						RSD1	P-value
	1		2		3			Strategy2	Day2,3	Cycle2	Strategy x Cycle2
	CF	PF	CF	PF	CF	PF
Number of sows	30	21	18	17	9	11
Feeding behavior
Number of visits to the feeder per day
Total	5.5a	5.7ab	5.6ab	5.4a	5.9b	6.7c	1.1	0.147	<0.001	0.228	<0.001
Feeding	1.0	1.0	1.0	1.0	1.0	1.0	0.2	0.411	0.428	0.603	0.519
Non-feeding	4.4a	4.7ab	4.6ab	4.4a	4.8b	5.7c	1.3	0.096	<0.001	0.129	<0.001
Time spent in the feeder per day, min
Total	54.9a	53.2a	63.6b	50.9a	66.5c	57.9ab	3.6	0.949	<0.001	<0.001	<0.001
Feeding	29.9a	29.9a	33.4b	31.5b	35.9c	29.7a	2.0	0.897	<0.001	<0.001	<0.001
Non-feeding	24.9ab	23.3a	30.2b	19.5a	30.6b	28.2b	4.5	0.334	<0.001	<0.001	<0.001
Drinking behavior
Number of visits to the trough per day
Total	10.0b	11.1b	8.9a	7.8a	14.9c	10.7b	1.6	0.840	<0.001	<0.001	<0.001
Drinking	9.3b	10.0b	8.2a	7.5a	13.0c	10.0b	1.4	0.769	<0.001	<0.001	<0.001
Non-drinking	0.7b	1.1b	0.7b	0.3a	1.9c	0.7b	1.2	0.900	<0.001	0.455	<0.001
Time spent in the trough per day, min
Total	9.7b	10.0b	7.4a	8.4a	9.7b	9.5b	1.3	0.981	<0.001	<0.001	<0.001
Drinking	9.5b	9.8b	7.3a	8.3a	9.6b	9.3b	1.3	0.992	<0.001	<0.001	0.010
Non-drinking	0.2b	0.2b	0.1a	0.1a	0.1a	0.2b	0.7	0.800	<0.001	<0.001	<0.001
Water intake, L/d	13.9b	12.9b	11.4a	10.7a	17.6c	13.6b	1.6	0.810	<0.001	0.622	<0.001

¹RSD, residual standard deviation.

²Fixed effects of the variance analysis model, with initial parity of the sow as a covariate.

³From days 3 to 106 of gestation, the following are all followed in the gestating room.

a, b, c, and d values within a row that do not share a common superscript are significantly different (P ≤ 0.05).

CF, conventional feeding, PF, precision feeding.

The drinking behavior consisted mainly in drinking visits to the drinking troughs, with 92% of drinking visits taking 98% of the time spent in the drinking trough. The feeding strategy affected the number of drinking and non-drinking visits regarding the gestation cycle (P < 0.001). In cycle 1, PF sows visited more often the trough than CF sows, for both drinking and non-drinking visits. The opposite was observed in cycles 2 and 3 (P < 0.001). Time spent drinking depended on strategy and gestation cycle (interaction Strategy × Cycle: P = 0.01), with PF sows spending more time to drink in cycles 1 and 2 than CF sows and the opposite occurring in cycle 3. The water intake remained more or less constant for PF sows with 12.3 L/d over the 3 gestation cycles which was not the case for CF sows who consumed 5 L/d more in cycle 3 compared to previous cycles (17.6 vs. 12.6 L/d on average, respectively).

Health

Figure 2 is a 3 factors, i.e., cycle, feeding strategy and gestation week, figure. For instance, if the focus is on the dirtiness, there is an orange circle at week 15, cycle 3 and strategy CF meaning that on average around 50% of the body of sows is dirty for the CF sows, on the given week and cycle. Feeding strategy had no effect on the occurrence of lameness and bursitis throughout the different cycles (Figure 2; P > 0.10). A total of 3.6% of observations reported lameness events in cycle 1, and 1.6% in cycles 2 and 3 (Figure 2). There was twice less bursitis observed in cycle 3 than in cycles 1 and 2 (0.30 vs. 0.60 on average; P < 0.001) mainly due to a significant drop in the occurrence of bursitis in the second part of gestation of cycle 3 (Figure 2). The feeding strategy affected the dirtiness of sows regarding the gestation cycle (interaction Strategy × Cycle: P = 0.04). CF sows were dirtier than PF sows in cycle 1, less dirty in cycle 3, and not different from PF sows in cycle 2 (interaction Strategy × Cycle: P = 0.04). The gestation week had an impact on sow dirtiness with sows being 56% dirtier over the last 3 weeks of gestation than at the beginning of gestation (P < 0.001).

Figure 2.

Weekly evolution of (A) lameness (0 = absence vs. 1 = all sows are lamed), (B), bursitis (0 = absence vs. 1 = all sows have bursitis), and (C) dirtiness (0 = 100% sow’s body clean vs. 1 = 100% sow’s body dirty) observed over gestation and cycles for CF and PF sows. CF: conventional feeding; PF: precision feeding.

Feed cost and nutrient excretion

On average, PF sows consumed 43 and 46 kg more feed per gestation in cycles 2 and 3, respectively, which resulted in additional feed costs of + 11 € ($12) per gestation for PF sows compared to CF sows (Table 6; interaction Strategy × Cycle: P = 0.04). However, as PF sows mainly consumed GL diet, which is low in Lys and less expensive than all diets, the PF strategy has saved 11 € ($12) per ton of feed (P < 0.001). The 2 feeding strategies resulted in equivalent nitrogen intakes per gestation, but PF sows excreted 207 g (5%) less N per gestation, in particular for cycle 1, than CF sows (P = 0.009), with an improved N efficiency by 4.2% compared to CF sows (P < 0.001). Nitrogen excretion increased by +123 g on average by gestation cycle mainly due to an increase in sow parity (P < 0.001).

Table 6.

Effect of the feeding strategy (CF vs. PF) through the different cycles on feed costs and nutrient excretions

Strategy	Cycle						RSD1	P-value
	1		2		3			Strategy2	Cycle2	Strategy × Cycle2
	CF	PF	CF	PF	CF	PF
Number of sows	30	21	18	17	9	11
Feed cost
From total feed intake, €	86.6a	91.3a	87.8a	98.9b	88.7a	100.0b	8.6	0.504	0.456	0.037
Assuming GS diet is used, €	87.8a	95.1b	88.9a	104.0c	89.8a	105.0c	9.1	0.197	0.452	0.020
Feed, €/t	335c	326b	335c	324a	335c	323a	1.6	<0.001	0.729	<0.001
Nitrogen balance during gestation
Intake, kg	5.68	5.70	5.77	6.13	5.82	5.91	0.32	0.346	0.127	0.342
Retained, kg3	1.24a	1.49bc	1.19a	1.54c	1.22a	1.32b	0.16	<0.001	<0.001	0.755
Output, kg3	4.44b	4.21a	4.58c	4.59c	4.60c	4.59c	0.25	0.009	<0.001	0.147
Efficiency4, %	21.9b	26.1d	20.6a	25.0c	20.9a	22.2b	1.5	<0.001	<0.001	0.859
Phosphorus balance during gestation
Intake, kg	1.42a	1.50b	1.45a	1.62c	1.45a	1.58c	0.09	0.006	0.126	0.052
Retained, kg3	0.27a	0.35bc	0.26a	0.37c	0.26a	0.32b	0.04	<0.001	0.006	0.303
Output, kg3	1.15a	1.15a	1.19ab	1.25b	1.19ab	1.26b	0.06	0.632	<0.001	0.024
Efficiency4, %	18.9a	23.2c	18.1a	23.0c	18.0a	20.0b	2.01	<0.001	<0.001	0.580

¹RSD, residual standard deviation.

²Fixed effects of the variance analysis model, with initial parity of the sow as a covariate.

³Values simulated with InraPorc model.

⁴Efficiency calculated as retention/intake.

a, b, c, and d values within a row that do not share a common superscript are significantly different (P ≤ 0.05).

CF, conventional feeding; PF, precision feeding; GS, standard gestation diet.

Feed formulation for SID Lys has concomitantly influenced phosphorus excretion through feeding strategies and cycles (P = 0.02). Phosphorus excretion stayed stable over the 3 cycles for CF sows but increased by 50 g/cycle on average for PF sows. Phosphorus efficiency was 4.0% greater for PF than CF sows (P < 0.001).

Discussion

Adequacy of the feeding plan (energy supply)

Whatever the cycle, PF feeding strategy was efficient to make sows achieve the expected BT, in contrast with CF feeding strategy. This result can be partly explained by the fact that PF sows were delivered more feed than CF sows, representing a difference of 109 kg of feed after 3 gestation cycles. Working with primiparous sows only, Amdi et al. (2013) reported that the impact of feed allowance during gestation was greater than initial BT at insemination on BT at farrowing. Although the PF sows consumed more feed over the 3 gestation cycles, they consumed 550 g less SID Lys mainly due to a 2.2 g/d average reduction over the first 12 wk of each gestation. Without any difference on performances at farrowing, these results would be related to a decreased amino acid supply in excess as performed with the CF strategy. They are in agreement with those obtained by Stewart et al. (2021) with CF sows receiving the same amount of SID Lys per day on average, and PF sows requirements assessed using the NRC model (NRC, 2012) without accounting for any difference among parities on litter size and weight at farrowing. Solà-Oriol and Gasa (2017) showed that, with the same average piglet birth weight but with 2 additional piglets per litter, SID Lys requirements were increased by 1 and 2 g/d at the end of gestation for gilts and multiparous sows, respectively. Dietary blending proportions had different dynamics according to parity in our experiment. Primiparous sows were supplied less low Lys diet than multiparous sows, whose diet consisted entirely of low Lys diet up to day 80 of gestation. These results are in line with SID Lys requirement dynamics mentioned by Gaillard et al. (2020) according to parity with a slight increase in primiparous sows’ requirements associated with maternal growth. The strong increase from day 80 of gestation up to end of gestation, i.e., a decrease of low Lys diet in the dietary blending, for all sows, was due to a switch of requirements from maternal tissue growth or maintenance to fetal and mammary growth but still with a gap between primiparous and multiparous sows due to maternal growth.

In our study, past performances of the herd were used to estimate litter size and litter weight per parity. But the sows performed better than expectation due to the continuous genetic progress for prolificacy, which finally led to a slight underestimation of sows’ needs for fetal development. Thomas et al. (2021) suggested an intake of 11 g of SID Lys per day, after observing no impact on performance except with a reduction in the stillbirth rate when the increase in SID Lys reached 18 g/d during gestation. Contrary to this recommendation, and in view of our results and those of Stewart et al (2021), it is therefore possible with precision feeding to considerably reduce Lys intake (an intake between 9 and 10 g SID Lys/d, representing a 17% reduction compared with a supply of 12 g SID Lys/d currently used on commercial farms) over the first 3 mo of gestation without impacting sow’ performances. This can be a real advantage when the protein-rich ingredients take up a large proportion of the formulation cost.

Litter characteristics and performances (litter size, piglet weight at weaning, ADG) were not affected by the feeding strategy and it was difficult to show significant differences in maternity performance with such small groups, especially for the last gestation cycle 3, with 9 and 11 individuals for the CF and PF strategy respectively. However, these results are in agreement with Hansen et al. (2021) who carried out an experiment similar to ours with only gilts at the start of the trial. These authors also focused on long-term impacts on piglets, i.e., performances up to slaughter and carcass quality. They did not find any significant differences between piglets born from PF or CF sows, expect for the ADG after weaning which was greater due to a greater average daily feed intake in pigs born from PF sows. In contrast, Cloutier et al. (2024) reported a positive effect of precision feeding over 3 cycles on stillbirth rate and total piglet mortality before weaning. According to these authors, PF strategy allowed to satisfy the increasing needs for fetal growth at the end of gestation and thus reduced mortality during farrowing. This was not observed in our trial, performed with mixed parity sows. In our trial, CF sows always had greater SID Lys intakes than PF sows during the first 13 weeks of gestation, and equivalent SID Lys intakes to PF sows for the last 2 wk in the gestation room. PF sows were fed according to their needs, so CF sows were never deficient in SID Lys, so we couldn’t expect lower performance, in particular for stillbirth rate, as reported by Thomas et al. (2021). Amdi et al. (2014) reported that fat primiparous sows (19.0 mm on average) had litters with heavier piglets than thin primiparous sows (14.4 mm on average) whatever the daily quantity allowed (3 groups of feed allowance: 1.8, 2.5, or 3.5 kg/d). Due to difference in BT at farrowing between our 2 feeding strategies, we expected to see similar differences but our CF sows were leaner at farrowing of cycle 3 and consumed more during lactation than PF sows. This kind of compensation lead to similar weaning weights of the litters between the strategies. With a total litter weight difference of 5.5 kg at weaning between our 2 feeding strategies over the 3 gestation cycles and a standard deviation of 9.8, more sows per strategy (i.e., 84 instead of 9 or 11) would have been required to prove a significant difference (power t-test) and the effect on mixed parity sows may have been less important than the one obtained on primiparous sows by Amdi et al. (2014).

Precision feeding, effects on feeding and drinking behavior and on sow health

Sows are fed restrictively during gestation, which may affect their behavior and induces stereotypies (D’Eath et al., 2018). However, feed is delivered below ad libitum to avoid sows becoming over-weight or overfat and exhibit health and welfare problems such as lameness, difficult farrowing with a high stillbirth rate (Zhou et al., 2018; Lavery et al., 2019) or reduced lactation feed intake, associated with more important mobilization of maternal reserves during lactation and reproductive failure after weaning (Koketsu et al., 1996). The aim was to preserve health and to achieve optimal conditions for farrowing and the upcoming lactation, as well as sow’s longevity (D’Eath et al., 2018). A change in feeding or drinking behavior could reflect a decreased health status or welfare (Cornou and Kristensen, 2013; Durand et al., 2021). Reduced feed intake below regular feed allowance and time spent feeding would indicate health issues (Gaillard et al., 2021) and may reduce the average birth weight and increase the number of stillbirths (Vargovic et al., 2021a).

In agreement with other studies (Vargovic et al., 2021b; Gaillard and Dourmad, 2022), sows of both feeding strategies in our experiment consumed their ration in a single feeding visit. Compared to other studies (Olsson et al., 2011; Iida et al., 2017; Vargovic et al., 2021b), our sows spent twice much time on their feeding visits regardless of the feeding strategy. It can be attributed to the feeding system used as our values were in the same range as those obtained by Gaillard and Dourmad (2022) and Durand et al. (2023) with the same equipment, i.e., a single-way automatic feeder. This clearly shows that sows modify their feeding behavior according to automatic feeder settings by reducing consumption time when there is more competition for access to the feeder, i.e., more sows per feeder (Olsson et al., 2011). The influence of this parameter on the characteristics of the feeding behavior is much more important than the type of feeding strategy.

Feed intake and duration of feeding visits are not a reliable indicator for anticipating health disorders, but rather for helping farmers to identify animals in need of care (Vargovic et al., 2021b). Non-feeding behaviors are better indicators to detect a change in health or welfare status, and a large variation from the mean value may indicate an event or problem (Gaillard et al., 2021). Indeed, in the trial performed by Durand et al. (2023) with sows facing a feeding challenge (with only one feeder available instead of two), the number and duration of feeding visits remained stable, while the number and duration of non-feeding visits decreased. In our trial, PF sows made more non-feeding visits but spent less time on non-feeding visits than CF sows, with no apparent relation to any deterioration in health status.

Drinking behavior is often characterized by high inter-individual variability and daily intra-individual variability (Massabie, 2001; Rousselière et al., 2017). The amount of water recorded depends on the type of trough and the type of sow management, and also influences water spillage (Junge et al., 2012). In our trial, PF sows consumed less water than CF sows with values twice greater than those reported by Rousseliere et al. (2017) but within the same range as reported by Massabie (2001). We may have overestimated water consumption because of spillage which is known for bowl-type troughs (Junge et al., 2012) and our CF sows may have increased their water intake to appease hunger (Massabie, 2001), which was probably more important than in PF sows who consumed more feed during the trial. As cycles 1 and 3 took place in summer and spring, it might contribute to the greater amount of water used by the sows (for consumption or to wet themselves and increased their evaporative heat loss), compared to cycle 2 which took place in winter.

We used the welfare quality (2009) variables to characterize sows’ state of health or welfare, and its evolution during the study. Depending on the week or gestation, the observers were different, which may have had influenced the results obtained, particularly for the occurrence of bursitis, which is not necessarily reliable between observers (Czycholl et al., 2016), unlike the other parameters monitored (Friedrich et al., 2020). We did not see any effect of the nutritional strategy on health status over the long term, but the sows’ culling policy in the farm might have hidden a possible effect of the nutritional strategy. Contrary to the results obtained by Díaz et al. (2013), with sows housed in group of 8 during the gestation whether on slatted floor or rubber slat mats, we noted an increase in sow dirtiness during the last 3 wk of gestation with our sows who were kept in the same stable group of 20 over the entire gestation period on concrete floor. Sow dirtiness is greater at mixing when sows are in a dynamic group, and it may be an indicator of poor well-being due to decreased available space per sow (Zurbrigg and Blackwell, 2006) and competition for resting areas. Being heavier at the end of gestation could influence the mobility of sows, which would lie down in available areas without taking into account of the dirtiness of the floor.

The economic and environmental aspects of precision feeding over several cycles

Based on the individual feed intake of the sows in our trial, PF strategy represented an extra cost of+11 € ($12) per gestation and per sow compared to the CF strategy. Using an iso-energetic blend of a gestation diet and a lactation diet, i.e., CF strategy used in our experiment, instead of single conventional gestation diet with the same energy and SID Lys intake saved an average of 2.6 € ($2.8) per gestation and per sow and are in line with results shown by Moehn et al. (2011) when adequate feeding equipment is available. In our trial, PF sows received 37 kg more feed per gestation on average than the CF sows thus explaining the extra cost for PF sows. This additional intake is mainly an energy intake, because even with this significant difference in feed quantity, PF sows consumed 16% less SID Lys per gestation to meet the individual needs of each PF sow. Assuming a similar intake during gestation with both feeding strategies, the PF strategy would have saved 11 € ($12) per ton of feed, i.e., around 9 € ($10) per sow for 3 gestation cycles by increasing the proportion of low Lys diet over the first 12 wk of each gestation. These values are in line with previous studies (Stewart et al., 2021; Gaillard and Dourmad, 2022) showing a very slight reduction in feed costs by applying a PF strategy when quantity intakes are equivalent.

A PF adjusted for energy and SID Lys reduced the SID Lys intake but resulted in equivalent cumulative N intake and output per gestation in our trial. The dynamics of intake differed according to sow needs, resulting in a 4% improvement in N efficiency per gestation compared to CF strategy. Several studies (Gaillard et al., 2019, 2020; Stewart et al., 2021; Gaillard and Dourmad, 2022) highlighted that PF based on energy and protein helps to reduce N intake and output, which is not the case here. This contrast can be related to the greater feed intake by PF sows in our study, which was not the case in previous studied performed in vivo or in silico with iso-energetic intakes for PF and CF sows. A greater feed intake for PF sows led to an increase in fecal N for PF than CF sows. We hypothesized that CF sows with a negative energy balance from feed and over supplies of amino acids led to an increase in urinary N output (Pedersen et al., 2019; Johannsen et al., 2024). This greater urinary N output compensated lower fecal output for CF compared to PF sows which explained the better retention of PF sows despite equivalent of N intakes and outputs for the 2 feeding strategies.

In our trial, phosphorus output of PF sows was greater than CF sows, in line with the results of Stewart et al. (2021) but in contrast with results of Gaillard and Dourmad (2022). To reduce phosphorus excretion into the environment, the adequacy of phosphorus supplies should be also considered in PF, in addition to energy and amino acids, as simulated by Gaillard et al. (2019).

Conclusion

The PF adjusted for energy and Lys over 3 gestation cycles improved nitrogen efficiency by 4% and reduced Lys intake by 16%, without impacting either performance of sows and litters, feeding and drinking behaviors, or health over the trial follow-up period. Precision feeding enabled to better reach body condition targets at farrowing than conventional feeding, which should improve sow longevity. The reduction in feed costs and N outputs by 3% and 5% respectively with the PF strategy would be even more important that the amino acid concentration of the diet used in the CF strategy is high and associated with an important spillage of amino acids. In our study, the effects of precision feeding on other nutrients but amino acid and N intake, like phosphorus, were nuanced, indeed phosphorus supply was not adjusted to requirements but resulted arithmetically from blending diets based on amino acid requirement, with their specific phosphorus contents. Therefore, accounting for phosphorus requirement in PF, in addition to energy and amino acids, is expected to reinforce the benefit of this strategy at the farm level.

Glossary

Abbreviations

ADG

average daily gain

BT

backfat thickness

BW

body weight

CF

conventional feeding

Lys

lysine

PF

precision feeding

RFID

radio-frequency identification

SID

standardized ileal digestible

Conflict of Interest Statement

The authors declare no conflicts of interest.