Intensive animal production often comes under review because of a range of practices that are considered to be adverse for the animals involved [1
]. The welfare issues encountered when pigs are intensively reared are related to overcrowding, restrictive space allocations, barren environments, and the isolation of individual animals [2
]. The farrowing crate has been criticized, as it imposes several of these welfare issues on the periparturient and lactating sow [3
]. The farrowing crate was introduced to intensive pig breeder farms with several aims: to reduce piglet mortality from sow crushing, provide a clean and hygienic environment for neonatal piglets to grow, and protect stock-people from sow aggression [1
]. Farrowing crates were initially devised with the aim of providing a safe working environment and maintaining pre-wean mortality as low as 10% [6
]. Despite these benefits, there is evidence that housing sows in farrowing crates leads to compromised sow welfare, as confinement results in an increased stress response at certain times during farrowing and lactation [3
]. Repetitive, bar-associated behaviours or stereotypies often develop prior to and during parturition as sows attempt to form nest areas in a restrictive environment [9
]. As a result, housing options that reduce the level of sow confinement during parturition and lactation have received attention in recent literature.
A significant concern for producers is that non-confinement of sows in farrowing pens leads to an increase in exactly what farrowing crates were designed to minimise: piglet mortality. Piglet mortality that was associated with crushing increased by 6–9% when the sows farrowed in open, confinement free pens versus farrowing crates [6
]. Whilst some investigations report exactly this [6
], the results from others suggest that the type of farrowing system results in little influence on piglet mortality [11
]. The reason as to why similar results have been achieved in pens and crates is unclear, but is most likely explained by a range of influences that would include design features, management procedures, sow factors, environmental factors, and experimental design flaws, such as insufficient statistical power [1
]. This inconsistency has led to limited large-scale commercial pen adoption, which is anticipated to continue until success factors related to farrowing pens are better understood.
The meta-analysis is a statistical tool that allows for the combination of results across multiple scientific studies and allows for the determination of important factors that affect key variables across experiments [14
]. A meta-analysis was applied to 45 experiments reported in 42 publications to determine the relative importance of pen design features in grower pig housing. From this, these authors determined the importance of factors, such as space allowance, enrichment, and group size on behaviour, growth, and efficiency which contribute to sustainable production [4
]. A similar methodology has been applied to gestation housing in sows [15
], which was able to evaluate physiology, behaviour, and production outcomes from multiple publications where group and stall housing were compared. To date, there is no published evidence of such an analysis on lactation sow housing. Thus, the aim of this investigation was to conduct a systematic review to identify the key variables that may impact on piglet output, and then perform a meta-analysis on these included publications to determine the factors that result in the comparable performance of farrowing pens. We hypothesized that overall, piglet mortality would be higher in pens than crates, but that there would be specific features of pens that result in similar performance.
2. Materials and Methods
Guidelines for conducting a systematic review were obtained from Wylie et al. [16
]. This publication complies with the publication guidelines that were provided by Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) [14
2.1. Search Strategy
The literature searches were conducted on the 21st of February 2018 in four databases; Scopus, BIOSIS Previews, Cab Abstracts, and Web of Science. A search protocol was designed to obtain any articles that provided data on sow farrowing performance and the design of farrowing accommodation. The search terms used were identified as being relevant by the authors and designed broadly to ensure that all publications conducted in farrowing pens were included. The specific terms varied based on the database in question, but all of the methods included the terms ‘farrow’ AND ‘sow’ AND ‘design’ OR ‘housing’ OR ‘system’. The actual search frames that were used for each database are provided below:
Search method for Scopus
TITLE-ABS-KEY (farrow* AND sow AND (design OR housing OR system))—856 results
Search method for BIOSIS previews
TS = (farrow* AND sow AND (design OR housing or system))—2546 results
Search method for Cab Abstracts
TS = (farrow* AND sow AND (design OR housing or system))—2330 results
Search method for Web of Science
TS = (farrow* AND sow AND (design OR housing or system))—961 results
2.2. Screening & Eligibility
Searching the four databases identified a total of 6695 articles. All of the articles collected from the online searches were downloaded to Endnote (X7.7.1). Hard copy conference proceedings from the Australasian Pig Science Association (APSA) and final reports from research supported by the Co-operative Research Centre for High Integrity Australian Pork in Australia were imported into the database and screened accordingly. The reference lists of included studies were scanned for potential new inclusions while using the study eligibility form. All of the publications were sorted by the first author, title, and abstract of the articles screened to remove duplicates, and a species scan was conducted to remove articles that did not focus on pigs. An inclusion/exclusion checklist was developed to identify papers conducted in a research area relevant to the hypothesis (Table 1
). The abstract and full paper were investigated to further examine studies with relevant titles. If there was uncertainty on whether an individual publication complied with the inclusion and exclusion criteria, a decision was made by discussion with all authors. If a publication was not accessible as a full text, the corresponding author was contacted with a follow-up email that was sent two weeks after the initial email if no response was received. The paper was excluded from the study if the authors did not respond within four weeks of the first contact.
2.3. Inclusion and Exclusion Criteria
A publication was selected if the experimental design included a comparison of a non-confinement in a farrowing environment at any point from parturition to weaning with traditional farrowing crates. In addition, the methodology was required to include details on the design of the farrowing area and to measure piglet performance (mortality rate, number of stillborn piglets, and number of weaned piglets). Any study that involved outdoor, free range, or group housing during parturition and lactation was excluded. Publications that were written in a language other than English were excluded if a translatable version was not accessible. Similarly, articles that were published prior to 1990 were excluded, if there were no accessible copies.
2.4. Data Collection
If the inclusion criteria were satisfied, data from each publication collected in the format that is outlined in Table 2
. All data was present in the included publications and as a result, no authors were contacted for obtaining additional data that was no published in the original article.
2.5. Quality Assessment
A quality assessment form was adapted (Table 3
) to enable the analysis of the quality of each publication through a weighted comparison of separate studies [16
]. The rating system provided two individual scores; the first score measured the quality of generic experimental design and the second score quantified the specific details of the farrowing environment. These two scores were combined, and each article was assigned a rating out of 36, with a higher score indicating that the experiment was robust and relevant to the topic of this review. One reviewer conducted the quality scoring on all of the publications that were included in the systematic review and meta-analysis.
The extracted data were analysed in individual meta-analysis models while using four data subgroups; number of piglets born alive (n = 28), number of stillborn piglets (n = 27), total piglet mortality from parturition to weaning (n = 30), and number of piglets weaned (n = 15). The sample size varied for each analysis, based on the data that were published in each article. The data were represented as the total number present in each litter, rather than as a percentage of total born or born alive piglets, with sow being an experimental unit. Any articles that reported these values as a percentage of total born or born alive were transformed while using the reported litter sizes to a value that represented a total number per litter. Each data-point included in the meta-analysis demonstrated a comparison between a farrowing crate and one type of pen. If there was more than one pen-type included in an article, they were included as separate data-points.
2.7. Statistical Analyses
Random effects meta-analyses were performed on the database, while using the metaphor package R statistical software Version 3.2.5 (R Core Team, Vienna, Austria) to examine whether pooled effect sizes for crate versus pen housing altered number of piglets born alive, number piglets stillborn, pre-weaning mortality, and number of piglets weaned [17
]. A pooled estimate of the mean relative risk of these traits and the corresponding 95% confidence intervals were calculated by random effects logistic regression model (binomial-normal model) to allow for heterogeneity in the analysis. Each measure of piglet viability was examined in separate meta-analyses to compare the effect of farrowing environment (pen or crate) by considering individual inter-publication variation. The measure of heterogeneity (I2) indicated the variation between studies. The measures of piglet mortality and number of stillborns were analysed while using relative risk. Risk, as opposed to odds, is calculated as the number of piglets in the group who died divided by the total number of piglets in the group. A relative risk greater than one indicated increased likelihood of the stated outcome being achieved in the treatment group, less than one indicated there was a decreased likelihood in the treatment group, and a ratio of one indicated no difference, that is the outcome is just as likely to occur in the treatment group as it is in the control group. The born alive and weaned number of piglets were continuous variables and, hence, were analysed while using standardized mean difference (SMD). Estimates of the traits and their associated confidence intervals were calculated by transforming the mean log-risk and its confidence interval back into the probability scale. The Q-test was used to assess statistical heterogeneity between studies and the I2 was calculated to describe the amount of inconsistency of findings across studies. Post hoc exploratory meta-regression analyses were performed, which included enrichment (provided or not provided), confinement (no confinement from loading until weaning or partial confinement for shorter periods of time in the early stages post parturition), and pen area (small, medium and large) to evaluate potential moderators to explain heterogeneity. The presence or absence of straw as a source of enrichment for sows within farrowing accommodation was recorded. Each pen environment was classified as small, medium, or large with ranges of 2.8–4.9 m2
(similar range to conventional farrowing crate), 5–7.5 m2
(equivalent size of farrowing pens as defined in the Animal Welfare Code of Practice), and greater than 7.5 m2
(greater space allocation than farrowing pen defined in the Animal Welfare Code of Practice), respectively. Funnel plots to test the asymmetry and publication bias of individual studies were conducted [18
]. A p
-value of less than 0.05 was considered to be significant.
Traditionally, farrowing crates have been the preferred housing type to enhance piglet survivability. The restrictive nature of a farrowing crate prevents sow movement, leading to a reduction in the number of piglets that are crushed by the sow; the primary cause of piglet death [19
]. The main benefit of a farrowing pen is an increase in the freedom of movement, which ensures that sows can conduct a normal range of behaviour, particularly during farrowing [21
]. As the sow undergoes hormonal changes that lead to restlessness and erratic posture changes, the increased range of movement within a pen is expected to be associated with an increase in piglet death due to sow overlay [22
]. The current finding determined that total piglet mortality was 14% more likely in a pen than in a crate supports this notion. Farrowing crates were designed to reduce movement of the sow that could cause overlaying or squashing of piglets [25
]. By removing the restrictive structures sows can perform more posture changes and conduct these changes at a greater speed, which heightens the risk of overlays [10
]. However, there are other factors outside of this that may lead in increases in piglet deaths. Farrowing pens are often larger in size than crates, and so specifically designed areas (creeps), which meet thermal needs of newborn piglets become harder to locate [3
]. A crate with a smaller area has fewer spaces that can cause piglet deaths that are associated with exposure to cold temperatures. Most farrowing crates have a separate heated creep area to accommodate for the different temperature requirements for piglets and sows. Ultimately, the farrowing crate was designed to ensure piglet comfort and survival [1
]. Death from exposure is more likely if a piglet fails to locate the creep. Novel projects that increase the likelihood of piglets remaining in the creep area within a pen may act to limit chilling and the associated deaths that are caused by overlays for non-viable piglets [27
]. Additionally, whilst pen size was shown to exert little influence on piglet deaths in our analyses, the numbers of animals used in most investigations were likely too few to examine such impacts on the exact causes of piglet mortality. The last way in which piglet mortality may be increased under pen conditions is the willingness of the stockperson to interact with the sow and her litter when housed in farrowing pens [9
]. Farrowing crates provide a safer environment for stock-people to work in, especially during the period immediately following farrowing when hormonal changes in the sow often result in high levels of aggression [7
]. Under pen conditions, the reduced confinement and high levels of sow aggression can make stockperson interventions that aimed at improving piglet survival more difficult. One study cited significant farm differences in mortality comparisons between crates and pens [28
]. Whilst there would have been animal, environmental, and nutritional differences between farms, personal communication with the authors indicated that the farm with pen performance comparable to crates employed stock-people with exceptional neonatal piglet care skills. Obviously, no detail on the level of stockperson skill was included in any of the publications that were included in this review and meta-analysis, and so we could not objectively examine this variable.
Confinement during the peri-parturient period has been linked to an increased incidence of stillbirth rates [29
]. The phenomenon is now commonly referred to as the confinement-stillbirth hypothesis [30
]. Investigations into the physiological underpinnings have identified that confined sows exhibit an increased level of cortisol prior to farrowing [31
], a decrease in post-expulsion oxytocin pulse [21
], as well an extended farrowing duration and inter-piglet birth intervals [32
]. However, since its inception, some have refuted the link between sow confinement and the incidence of intra-partum piglet mortality [33
]. Additionally, higher salivary cortisol has been observed in loose-housed farrowing accommodation as compared to confined sows [34
], with no effect on stillborn rates, farrowing duration, or birth interval [35
]. The results from the current meta-analysis would suggest that there is no overall improvement in the incidence of stillbirths in pens when compared with crates. However, when no enrichment was included for farrowing pens, the relative risk of stillborn piglets was 22% lower in farrowing pens versus farrowing crates without enrichment. This cements the idea that simply allowing the sow a greater freedom of movement in combination with a nesting substrate leading up to and during farrowing reduces the risk of intra-partum piglet death. This identifies that the provision of enrichment was beneficial for penned sows [36
]. The presence of straw in farrowing pens presents a substrate that could be a challenge for piglet movement towards the teats. Small low-viability piglets would have difficulty moving through the straw-based nest. Mortality that is associated with exposure to cold temperatures would increase if the piglets are incapable of moving through the dense bedding [38
], while alternate studies suggest that straw provisions can improve temperature regulation for piglets [3
]. However, chilling does present an issue with the classification of dead piglets, as difficulty in differentiating between stillborn and exposure-based deaths could lead to false positive classifications. This is due to physical similarities in stillborn, low-viability, and exposure piglets, as well as the positioning of the piglets near the rear of the sow. The refinement of current methods for classification of piglet status when booking-in litter information would be improved through assessment of more vigorous autopsy procedures.
The observed pre-weaning mortality was significantly higher in farrowing pens, and so this would then logically impact on the number of pigs weaned. However, lactation housing bore no impact on how many piglets were weaned. Most piglet deaths occur within the first 36 h post-farrowing [19
]. A common husbandry technique that was adopted within farrowing houses is cross-fostering, which involves the movement of piglets from one sow to another [39
]. Given that this process generally occurs at 24h after farrowing, any piglets that die prior to this fostering process can be replaced. With this reasoning, pre-weaning mortality can be higher in pens (when it occurs prior to fostering) with the number of pigs weaned remaining constant. The systematic review found that 50% of published literature failed to describe the fostering protocol used and only one article indicated that fostering occurred within treatment. While some publications did describe the protocol implemented, there was a lack of consistency in the reporting of several performance outcomes, which introduced variation in the data presented, making a meta-analysis unfeasible. However, individual piglet viability is variable, with differences in piglet vitality between pen and crate litters [40
]. Additionally, fostered piglets that are moved from one litter to another may experience greater stress in the crucial peri-natal period. Future studies should identify whether there are differences in growth potential between fostered and non-fostered piglets in different birth and rearing environments.
Meta-analyses are an innovative tool that can be used to assess overall trends in publications and lead to the instigation of production changes. They have been applied to allow the refinement of housing design and carcass features in pig production [4
]. There are several areas that could be enhanced by applying this technique to allow for the refinement of animal management, including the assessment of environmental conditions, reproductive management, or tracking disease transmission through herds. However, the limitation in data availability that was identified in the current study must be addressed to maximise the significance of any outcomes. The quality of previous literature is a major limitation for both literature and systematic reviews. The published literature included in the current study often lacked enough detail in the presentation of farrowing accommodation performance data, while also providing limited definitions for calculations used to obtain this data. Future publications should ensure that environmental and farrowing performance data should be more detailed and transparent when assessing any farrowing trials.
This study identified that current farrowing pen designs that are available in commercial production are flawed in one major area—pre-weaning piglet mortality. While other performance outcomes such as piglets born alive and number of piglets weaned remain consistent across pen and crate housing types, there is an increase in piglet mortality. This study determined that the number of piglets born alive does not differ, which indicates that reproductive development of piglets in utero is not affected by housing. Additionally, the management of un-confined sows is sufficient to ensure that stillborn rates do not increase in loose-housing. This suggests that the main concern for future research in farrowing pens should be focused on reducing piglet mortality. Adjusting environmental conditions and the refinement of farrowing pen design can allow for housing that is conducive for piglet survival. A systematic review should be conducted on the intervention strategies that have been used to reduce piglet mortality to determine the viable mechanisms that can be implemented in commercial production.