Impact of Non-Confinement Accommodation on Farrowing Performance: A Systematic Review and Meta-Analysis of Farrowing Crates Versus Pens

Simple Summary The aim of this project was to review previously published research with a focus on the effects of farrowing accommodation on piglet performance. The specific design features were analysed to determine whether animals in loose housed farrowing pens or crates from loading to weaning contribute to differences in litter performance obtained from different farrowing house accommodation types. This was the first systematic review and meta-analysis conducted towards the farrowing performance of crates and pens. The relative risk of piglet mortality was 14% higher in farrowing pens than farrowing crates, which indicated that non-confinement of sows compromises post-natal piglet survival. Overall, the type of farrowing accommodation did not affect the number of stillborn piglets. However, the rate of stillborn piglets was lower in farrowing pens that were not enriched when compared with farrowing crates, also with no enrichment. There was no effect of housing type on the number of piglets born alive or the number of piglets weaned, although the sample size for the later was much smaller. Producers should anticipate an increase in mortality when piglets are reared by sows that are unconfined in the pen designs that are currently available, which supports the wider belief that crates are successful for reducing pre-weaning piglet mortality. Abstract There are conflicting reports regarding the effect of farrowing house accommodation on piglet performance. The aim of this investigation was to use a systematic review and meta-analyses to summarise the results of publications that focused on direct comparisons between full confinement conventional crates and various designs of loose-housed farrowing pens from loading until weaning. Literature searches in Scopus, BIOSIS Previews, Cab Abstracts, and Web of Science identified 6695 articles. Twenty-two publications were retained for the systematic review and individual meta-analyses after screening for inclusion criteria. The random effects meta-analyses were performed on crate versus pen for number of piglets born alive, number of stillborn piglets, pre-weaning mortality, and number of piglets weaned. Additionally, the modifiers of confinement length (no confinement from loading until weaning or partial confinement for shorter periods of time in the early stages post parturition), enrichment (no enrichment or enrichment provided), and pen size (small, medium, or large) were examined. There was a 14% increase in the relative risk of piglet mortality in farrowing pens when they were compared with crates (p = 0.0015). The number of stillborns per litter was not different between the pen and crate. However, when providing enrichment in the pens, there was an increase in stillborns within farrowing crates versus pens (p = 0.009). There was no overall effect on piglets that were born alive or number weaned. As there is no difference between piglets born alive and mortality is significantly higher in farrowing pens, a reduction in the number of piglets weaned was expected but not observed, which was possibly due to the lack of weaning details provided in the publications. This was the first systematic review and meta-analysis conducted on the performance of farrowing accommodation and identified that farrowing pens do compromise post-natal piglet survival. Future efforts should focus on improving sow comfort in the farrowing crate to maximize both piglet and sow welfare.


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

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. Table 1. Inclusion and exclusion criteria used to screen and identify publications relevant to the topic.

Inclusion Criteria Exclusion Criteria
Did the publication have farrowing pens included in this study? Did the publication include group housing accommodation during farrowing? Did the publication compare non-confinement or partial confinement farrowing accommodation with a traditional farrowing crate?
Did the publication include outdoor housing accommodation during farrowing? Did the publication include the required outcomes of born alive litter size, piglet mortality rate, number of stillborn piglets and number of piglets weaned? Did the publication include any descriptive or visual information on farrowing crate design?

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.

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.

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.

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.

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 m 2 (similar range to conventional farrowing crate), 5-7.5 m 2 (equivalent size of farrowing pens as defined in the Animal Welfare Code of Practice), and greater than 7.5 m 2 (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.

Database Search
The database search identified 6693 articles from four databases and two articles from external sources. After the duplicates were removed, there were 4483 articles remaining. Of these publications, 380 were deemed to be relevant when the inclusion and exclusion criteria were applied to the title. Twenty-two of these articles were examined in detail to enable the extraction of information for systematic review. The systematic review identified 32 individual comparisons between a farrowing crate and a farrowing pen from the twenty-two publications. Figure 1 presents a flow diagram of the publications that are involved in the screening and eligibility and the data used in each meta-analysis are shown in Appendix A. Data were extracted from 4385 litters, with data provided from 2182 and 2203 sows farrowing in crates and pens respectively. The articles were conducted in a range of countries, including Australia, Canada, China, Czech Republic, Denmark, Finland, Germany, New Zealand, United Kingdom, and USA, and were published from 1990 to 2016. Only 14% of articles were abstracts, conference proceedings or final reports while 86% of publications were peer-reviewed, full journal articles. The average sample size for the included studies was 68 (range = 6-394). Sample size was often restricted in these publications with 40% involving less than 20 sows per treatment while 22% of studies were conducted with more than 60 animals per treatment. The specific type of pen (e.g., Freedom) was not assessed, as the sample size for each type was too small for analysis. Studies using loose-housed farrowing pen systems from loading to weaning were compared to studies within farrowing pens that provided an option for confinement for a short period during parturition and early lactation.

Quality Assessment Scoring
A quality assessment score was provided for each publication. The scores associated with each publication for the twenty-two quality assessment criteria are presented below ( Table 4). The overall quality of all the publications was high with half of the publications obtaining a score greater than 70% (range = 25-31) and 95% of publications obtaining a score greater than 50% (range = 19-31). Only one publication was assigned a score of less than 50%. The primary aim of 72.7% of publications was to compare the traditional farrowing crate with a non-confinement farrowing accommodation from loading until weaning. The remaining articles focused on behaviour analysis as the primary focus while still providing farrowing performance data. Over 81% of publications involved a comparison between two housing types, while 18.2% involved an experimental design with more than three treatments. The dimensions of each housing type, with distinctions for the separate sow and creep areas, were provided for 86.4% of publications, while the remaining publications provided minimal detail on either total area or total size. The design of the farrowing environment was described in varied detail. The lighting conditions were only described in 4.5% of publications, while creep area heating was defined in all but one of the articles. The flooring material in the creep and sow areas were described in all the publications whereas ventilation and environmental design of sheds was clearly described in only 27.3% of the articles. Only 36.4% of publications described the components in detail of the farrowing space that would enable the protection of piglets or restraint of the sow. However, there were still over 90% of publications that mentioned the presence of these protective designs. Enrichment was provided for sows in 68.2% of the studies that were included in the review. Over 59% of the experiments were designed for both farrowing crates and pens to be housed in the same rooms. Clear protocols and definitions were provided for fostering in 50% of publications. However, within the top 12 studies, 75% provided clear definitions of the methods for fostering. Similarly, while only 27.3% of publications provided detailed piglet mortality definitions and values, 68.2% of studies separated piglet mortality into smaller sub-groups that were based on the cause of death. Overall, the quality of data collection for farrowing and weaning performance was moderate with only 50% and 27.3% of publications, respectively, providing comprehensive figures.

Qualitative Systematic Review
The systematic review identified 28 articles that reported the number of piglets born alive in each litter. The average number of piglets born alive was 12.54 in both farrowing crates (range = 8.4-17.1) and farrowing pens (range = 8.8-17.1). The results in 50% of articles indicated a decrease in piglets that were born alive in farrowing pens, 32% indicated an increase in born alive and the remaining articles found no change between farrowing housing types in piglets born alive.
Twenty-seven articles recorded the number of stillborn piglets for farrowing pens and crates. Stillborn piglets were recorded as a total number of stillborns in all litters for each farrowing accommodation type. Crates resulted in 77.81 stillborns (range = 5-416) across all litters, while 81.37 (range = 8-440) stillborns were found in farrowing pens. Over 59% of articles reported a decrease in the number of stillborns that were found within farrowing pens, while 37% of published work indicated the contradictory result, an increase in stillborns in pens.
Pre-weaning mortality was recorded as a total number of piglet deaths in each farrowing accommodation across all litters. An increase in pre-weaning mortality in farrowing pens was observed in 56% of publications, while 40% of published data indicated that the farrowing crates had more piglet mortality. The average pre-weaning mortality for farrowing crates was 101.66 piglets/trial (range = 7-506), which was lower than 123 piglets/trial (range = 9-590) in farrowing pens.
The number of piglets weaned per litter was published in 15 articles. The number weaned was 9.73 piglets/litter in farrowing crates (range = 7.1-12) and 9.81 piglets/litter in farrowing pens (range = 7.54-12.3). One-third of articles indicated a decrease in the number of piglets weaned/litter in farrowing pens, 46% measured an increase, while 20% recorded no difference between housing types.

Meta Analysis
The data used for the meta-analyses is reported in the Appendix A (Tables A1 and A2). Total piglet mortality had a moderate amount of heterogeneity (I2 = 69.81%, p = 0.002). The relative risk of piglet mortality in a farrowing crate was 14% lower than a farrowing pen (Figure 2). Several external factors were examined as moderators. There was no effect of confinement type (whether sows were unconfined for the entire lactation period or had partial confinement; p = 0.853), sow enrichment with straw (p = 0.801), or relative pen size (p = 0.206) on pre-weaning mortality in penned sows. A funnel plot regression test that is shown in Figure 3 indicates that there was no publication bias (z = 0.538, p = 0.591).
The number of stillborn piglets recorded in each litter had a moderate amount of heterogeneity (I2 = 57.93%, p = 0.001). The relative risk of stillborn piglets was comparable in farrowing pens and farrowing crates (Figure 4). There was no effect when comparing the confinement type (p = 0.706), or between crates and pens when the pen size was considered to be standard (p = 0.089). However, when including enrichment as a moderator there was a significant difference between crates and pens. There was no difference in the relative risk of stillborn piglets in crates and pens when enrichment was provided, but when there was no enrichment of the farrowing pen, stillborn piglets in crates were 22% higher than in pens (p = 0.009). When enrichment was provided, the relative risk of stillborn piglets was decreased by 10% in crates relative to pens (p = 0.099). A funnel plot regression test, as shown in Figure 5, indicates that there is no publication bias (z = −0.223, p = 0.823).
The number of piglets born alive in each litter had a high amount of heterogeneity (I2 = 0%, p = 0.303). The farrowing crates and pens had no difference in the number of piglets born alive ( Figure 6) and, within farrowing pens there was no effect of confinement type (p = 0.786), sow enrichment (p = 0.597) or pen size (p = 0.659).
The number of piglets weaned from each litter had a low amount of heterogeneity (I2 = 43.96%, p = 0.021). There was no difference in the number of pigs weaned between crates and pens (Figure 7), and there was no effect when comparing the moderators of confinement type (p = 0.567), sow enrichment (p = 0.765), or pen size (p = 0.333) within pens.      Grey areas indicate publications of significance, the dotted line represents the relative risk of 1 shown by a line of no-effect and white areas show non-significance. If no publication bias is present, the data-points will be organized symmetrically. between crates and pens when the pen size was considered to be standard (p = 0.089). However, when including enrichment as a moderator there was a significant difference between crates and pens. There was no difference in the relative risk of stillborn piglets in crates and pens when enrichment was provided, but when there was no enrichment of the farrowing pen, stillborn piglets in crates were 22% higher than in pens (p = 0.009). When enrichment was provided, the relative risk of stillborn piglets was decreased by 10% in crates relative to pens (p = 0.099). A funnel plot regression test, as shown in Figure 5, indicates that there is no publication bias (z = −0.223, p = 0.823).   The number of piglets born alive in each litter had a high amount of heterogeneity (I2 = 0%, p = 0.303). The farrowing crates and pens had no difference in the number of piglets born alive ( Figure  6) and, within farrowing pens there was no effect of confinement type (p = 0.786), sow enrichment (p = 0.597) or pen size (p = 0.659).  The number of piglets weaned from each litter had a low amount of heterogeneity (I2 = 43.96%, p = 0.021). There was no difference in the number of pigs weaned between crates and pens (Figure 7), and there was no effect when comparing the moderators of confinement type (p = 0.567), sow enrichment (p = 0.765), or pen size (p = 0.333) within pens.

Discussion
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 [19,20]. The main benefit of a farrowing pen is an increase in the freedom of movement, which ensures

Discussion
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 [19,20]. 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][23][24]. 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 [10,26]. 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 [1,3]. 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 [36,37]. 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 [4,41]. 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.

Conclusions
This was the first systematic review and meta-analyses conducted into the influence of farrowing pen housing on the piglet traits that are important for farrowing house performance. The relative risk of pre-weaning mortality was 14% higher in farrowing pens when compared with crates. The number of piglets born dead was comparable between the crates and pens with enrichment, but the relative risk of stillbirth was increased by 22% in crates versus pen without enrichment. The number of pigs weaned was unaffected by lactation housing design, but this result could reflect an flawed experimental design and data inclusion in publications.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data and in the writing of the manuscript.