Abstract
Piling is the formation of abnormally dense aggregations of hens and may compromise their welfare and, in fatal cases, result in smothering. However, the housing and management contexts related to this behavior remain poorly understood in cage-free systems. This study aimed to characterize heterogeneity in management practices and assess whether flock size, nest type, and light-off routine were associated with reported recurrence and the reporting of piling in multiple location categories. A cross-sectional survey was conducted with 90 respondents, of whom 83 reported piling and were included in the analyses. Multiple correspondence analysis included eight active variables, bias-reduced logistic regression models evaluated the proposed associations, and open-ended responses were thematically categorized. Among respondents reporting piling, 57 (68.7%) described recurrent episodes. Corners (66.3%) and low walls (43.4%) were the most frequently reported locations. The first two dimensions of the multiple correspondence analysis explained 34.3% of the raw inertia and indicated that production scale, automation, feeding, and pullet-management practices varied jointly. There was no evidence that recurrence was associated with flock size (OR = 1.05; 95% CI: 0.61–1.83), automatic versus manual nests (OR = 1.92; 95% CI: 0.65–5.87), or non-abrupt versus abrupt light-off (OR = 1.04; 95% CI: 0.40–2.63). These characteristics were also not associated with reporting multiple location categories. In the open-ended responses, the predominant themes were pullet rearing, transfer, or acclimation (26.8%) and lighting conditions and management (25.0%). Farm characteristics therefore represented interdependent combinations rather than isolated factors capable of explaining the reported recurrence or spatial distribution of piling. In practical terms, the findings indicate that monitoring should prioritize the transfer and acclimation period, localized lighting contrasts, and areas where birds congregate. However, these conditions should be evaluated prospectively using objective measurements before being considered risk factors or used to support preventive recommendations.
Keywords: Cage-free systems, Laying hens, Management, Multiple correspondence analysis
Introduction
Concern about laying-hen welfare has increased demand for production systems that allow greater freedom of movement and opportunities to perform highly motivated behaviours. This transition from conventional cages towards cage-free housing changes both the behavioural opportunities available to hens and the management demands placed on producers (Xin et al. 2011; Hartcher and Jones 2017; Arno et al. 2023; Harju et al. 2024).
Despite these potential welfare benefits, cage-free housing presents distinct behavioural and management challenges. One of these is piling, defined as an unusually dense aggregation in which hens press against or climb over one another (Mazocco et al. 2024). Piling and smothering are related but distinct outcomes: smothering refers specifically to mortality caused by suffocation during a piling event (Gray et al. 2020; Chowdhury et al. 2024). Thus, piling may occur without mortality. Factors potentially associated with the initiation or development of piling include lighting conditions, temperature, ventilation, flock distribution, sudden noises, predator disturbance, and management practices (Barrett et al. 2014; Rayner et al. 2016; Winter et al. 2022b; Chowdhury et al. 2025).
Previous studies have described the occurrence, timing, location, and perceived or measured contexts of piling and smothering in commercial cage-free systems (Barrett et al. 2014; Rayner et al. 2016; Gray et al. 2020; Winter et al. 2022a, b; Chowdhury et al. 2024, 2025). Evidence from Brazilian farms remains limited, however. Farm size, nest configuration, lighting routines, feeding, pullet rearing, and other practices may covary, making it difficult to represent management heterogeneity with isolated variables or arbitrary farm categories.
This study characterized farm-management heterogeneity and reported manifestations of piling among Brazilian cage-free producers and evaluated whether flock size, nest type, and light-off routine were associated with reported recurrence and reports involving multiple location categories. We hypothesized that farm-management practices would covary along interpretable multivariable dimensions and that flock size, nest type, and light-off routine would be associated with both piling outcomes.
Materials and methods
Ethical considerations
The study used an anonymous questionnaire and was conducted in accordance with Brazilian National Health Council Resolution No. 510 of 7 April 2016 (Brasil 2016). According to the institutional assessment, formal research ethics committee review was not required because the survey addressed professional production practices and collected no identifiable personal or sensitive information.
Survey design and study population
This cross-sectional survey was distributed to commercial cage-free egg producers registered with the MIRA Laying Hen Initiative across Brazil. Cage-free was defined broadly as housing in which hens were not confined in cages; outdoor access was not required. Participation was voluntary, so the sample represents a self-selected producer network rather than a probability sample of Brazilian farms. Data were collected through Google Forms® from July to November 2023. The supplied anonymous file contained state but no respondent, farm, company, or timestamp identifier; consequently, unique farms and one response per farm could not be independently verified.
The questionnaire was adapted from previous surveys (Barrett et al. 2014; Rayner et al. 2016; Winter et al. 2022a) and comprised 19 closed-ended fields and two optional open-ended fields (Supplementary Table S1). The screening field asked whether the respondent had ever experienced piling on the farm. Subsequent questions described usual or reported farm practices and experiences, but the instrument did not establish a common observation period or verify that every management response referred to the same flock and time as the reported piling episode. Age, location, perceived triggers, strain, season, and time of day allowed multiple responses.
Outcome terminology, ascertainment, and scope
Before questionnaire completion, piling was explained to respondents as hens pressing against or climbing over one another in an unusually dense aggregation. Smothering is conceptually distinct and denotes death by suffocation resulting from piling (Gray et al. 2020; Chowdhury et al. 2024). No quantitative threshold for group size, local density, or episode duration was specified, and events were not independently observed. Structured mortality data were not collected. Accordingly, the outcome is termed farmer-reported piling rather than confirmed piling or smothering.
The questionnaire did not distinguish production barns as single-tier litter, combined litter-and-slatted floor, or multi-tier aviary systems. It did not record litter substrate, amount, depth, age, moisture, or friability; usable floor area; illuminance or spatial light gradients at the reported site; or the number, type, position, and bird-access space of feeders and drinkers. Stocking density could not be calculated. These unmeasured characteristics were not included in the analyses.
Variables and statistical analysis
Ninety questionnaires were returned. Eighty-three respondents reported at least one piling episode and were included in all within-reporter analyses; seven reported no piling and contributed only to participant flow because skip logic left their farm-management fields blank. All 83 piling-reporting records were complete for the active multiple correspondence analysis variables and the regression models. Two pairs of records were identical across all supplied fields, including one pair among piling reporters. Because no identifiers were available, all records were retained in primary analyses and the later record in the positive pair was removed in a sensitivity analysis.
Closed-ended variables were summarized as counts and percentages. Multiple-response options were expanded separately using 83 respondents as the denominator, so percentages may sum to more than 100%. Flock size was harmonized into < 5,000, 5,000–10,000, 10,001–15,000, and > 15,000 hens. Transfer age was grouped as < 14, 14–16, and ≥ 17 weeks. State was treated as a geographic descriptor, while age, location, triggers, strain, season, and time of day described the reported experience rather than farm-management dimensions.
Multiple correspondence analysis (MCA) was performed among the 83 piling reporters using eight active categorical variables selected to represent available farm-level scale, housing, feeding, and pullet-management characteristics: flock-size category, manual or automatic nest, abrupt/staged/dimmer light-off, feeding frequency, electrified-wire use, pullet source, pullet-rearing system, and transfer-age category. Categories with sparse or overlapping source labels were combined as documented in Supplementary Table S1. Recurrence and the number of reported location categories were projected as supplementary variables and did not influence the axes. Raw eigenvalues and inertia were reported, with Benzécri-corrected inertia shown separately; category coordinates, contributions, and squared cosines (cos²) are provided in Supplementary Table S2 (Greenacre 2017). A sensitivity MCA used only five current production-management variables: flock size, nest type, light-off, feeding, and electrified wire.
Reported recurrence was designated the primary analytical outcome: repeated episodes (1) versus one reported episode (0). The secondary outcome was selection of more than one reported piling location category (1) versus one category (0); this does not establish that a single episode occurred simultaneously at several locations. Firth bias-reduced logistic regression was used because of the modest sample and sparse cross-classifications (Firth 1993; Heinze and Schemper 2002). Three conceptually selected farm predictors were entered simultaneously without stepwise selection: flock size as an ordinal score across the four categories, automatic versus manual nest, and non-abrupt light-off (staged or dimmer) versus abrupt light-off. The secondary model additionally included recurrence because repeated episodes provide more opportunities to observe several locations.
Adjusted odds ratios (ORs), 95% profile penalized-likelihood confidence intervals, and penalized likelihood-ratio p-values were calculated. The three focal exposure tests were adjusted by Holm’s method separately within each outcome; recurrence in the secondary model was an adjustment covariate and was not included in that multiplicity family (Holm 1979). Sensitivity analyses treated flock size as a nominal four-level factor or as > 10,000 versus ≤ 10,000 hens, retained three light-off categories, removed the later record in the exact positive duplicate pair, restricted the location model to respondents reporting repeated piling, and used conventional maximum-likelihood logistic regression. The revised models were not preregistered, and two-sided results were interpreted as hypothesis-generating with emphasis on effect estimates and precision.
The two open-ended fields were read together and coded descriptively at respondent level. Multiple themes could be assigned, but each respondent was counted at most once within a theme. Theme percentages used the 56 respondents who contributed any open text as the denominator and can sum to more than 100%. The coding summarized producer perceptions; it did not convert comments into measured exposures or causal evidence.
Results
Respondent flow and descriptive characteristics
Of 90 respondents, 83 (92.2%) reported having observed piling and seven (7.8%) reported none (Fig. 1). The proportion 83/90 describes responses in this self-selected network and is not an estimate of piling prevalence or incidence. The 83 reporting respondents were located in 10 states, most frequently São Paulo (41.0%), Rio Grande do Sul (25.3%), and Minas Gerais (16.9%). Farm, housing, and management characteristics are summarized in Table 1.
Fig. 1.

Participant flow for the cross-sectional survey. The seven respondents reporting no piling had no subsequent farm-management data and were retained for flow only; they were not treated as a comparison group
Table 1.
Geographic, farm, housing, and management characteristics among respondents reporting piling (n = 83)
| Variable | Category | n (%) |
|---|---|---|
| State | São Paulo | 34 (41.0) |
| Rio Grande do Sul | 21 (25.3) | |
| Minas Gerais | 14 (16.9) | |
| Bahia | 4 (4.8) | |
| Pernambuco | 3 (3.6) | |
| Ceará | 2 (2.4) | |
| Santa Catarina | 2 (2.4) | |
| Paraná | 1 (1.2) | |
| Sergipe | 1 (1.2) | |
| Espírito Santo | 1 (1.2) | |
| Flock size | < 5,000 hens | 23 (27.7) |
| 5,000–10,000 | 27 (32.5) | |
| 10,001–15,000 | 24 (28.9) | |
| > 15,000 | 9 (10.8) | |
| Nest type | Manual | 44 (53.0) |
| Automatic | 39 (47.0) | |
| Light-off routine | Abrupt | 36 (43.4) |
| Staged | 35 (42.2) | |
| Dimmer | 12 (14.5) | |
| Feeding | One or two meals/day | 24 (28.9) |
| Three meals/day | 24 (28.9) | |
| Ad libitum | 35 (42.2) | |
| Electrified wire | No | 71 (85.5) |
| Yes | 12 (14.5) | |
| Pullet source | Own farm | 56 (67.5) |
| Third party | 27 (32.5) | |
| Pullet-rearing system | Cage-free | 69 (83.1) |
| Conventional cages | 14 (16.9) | |
| Transfer age | < 14 weeks | 25 (30.1) |
| 14–16 weeks | 47 (56.6) | |
| ≥ 17 weeks | 11 (13.3) |
Values are respondent counts and percentages within each variable
Repeated piling was reported by 57 respondents (68.7%), whereas 26 (31.3%) reported a single occurrence. Corners were selected by 55 respondents (66.3%), low walls by 36 (43.4%), pen dividers by 18 (21.7%), nests by 14 (16.9%), the centre of the litter area by nine (10.8%), and slats by two (2.4%; Table 2).
Table 2.
Farmer-reported manifestations and perceived contexts of piling among respondents reporting piling (n = 83)
| Characteristic | Response category | n (%) |
|---|---|---|
| Recurrence | One reported episode | 26 (31.3) |
| Repeated episodes | 57 (68.7) | |
| Number of reported location categories | One | 48 (57.8) |
| Multiple | 35 (42.2) | |
| Age at occurrence* | 1–4 weeks | 11 (13.3) |
| 5–12 weeks | 26 (31.3) | |
| 13–20 weeks | 32 (38.6) | |
| 21–30 weeks | 31 (37.3) | |
| 31–40 weeks | 18 (21.7) | |
| 41–50 weeks | 7 (8.4) | |
| ≥ 51 weeks | 3 (3.6) | |
| Reported location* | Corners | 55 (66.3) |
| Low walls | 36 (43.4) | |
| Pen dividers | 18 (21.7) | |
| Inside nests | 14 (16.9) | |
| Centre of litter area | 9 (10.8) | |
| On slats | 2 (2.4) | |
| Perceived trigger* | Other cause | 40 (48.2) |
| Predator near barn | 38 (45.8) | |
| Storm, wind gusts, or thunder | 28 (33.7) | |
| Localized sunlight | 21 (25.3) | |
| Tarp noise | 20 (24.1) | |
| Full moon | 19 (22.9) | |
| Vaccination management | 10 (12.0) | |
| Nest position | 4 (4.8) | |
| Feeder/drinker position | 2 (2.4) | |
| Season* | Summer | 44 (53.0) |
| Winter | 39 (47.0) | |
| Spring | 31 (37.3) | |
| Autumn | 24 (28.9) | |
| Time of day* | Night | 53 (63.9) |
| Afternoon | 25 (30.1) | |
| Morning | 21 (25.3) |
* Multiple responses were allowed; each option uses 83 as the denominator and percentages within these blocks can sum to more than 100%. ‘Multiple location categories’ means that more than one checkbox was selected across the respondent’s experience and does not establish simultaneous sites in one episode. Perceived triggers were not measured exposures and should not be interpreted causally. Structured mortality was not collected
Multiple correspondence analysis
The MCA included 21 modalities across eight active variables. Dimension 1 had an eigenvalue of 0.324 and accounted for 19.9% of raw inertia; Dimension 2 had an eigenvalue of 0.234 and accounted for 14.4%, for 34.3% cumulatively (Table 3; Fig. 2). Benzécri-corrected shares were 74.8% and 22.5%, respectively; these percentages refer only to corrected inertia and should not be interpreted as 97.3% of the original data variation. Dimension 1 contrasted smaller flocks, manual nests, earlier transfer, conventional-cage rearing, one or two daily meals, and third-party pullets with automatic nests, larger flocks, dimmer-controlled light-off, electrified wire, and transfer at 14–16 weeks. It was therefore interpreted cautiously as a scale/automation and pullet-management gradient.
Table 3.
Summary of the multiple correspondence analysis of eight active farm-level categorical variables (n = 83)
| Dimension | Raw eigenvalue | Raw inertia (%) | Cumulative raw inertia (%) | Benzécri-corrected share (%) | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 0.324 | 19.9 | 19.9 | 74.8 | |||||||
| 2 | 0.234 | 14.4 | 34.3 | 22.5 | |||||||
| 3 | 0.161 | 9.9 | 44.3 | 2.5 | |||||||
| Modalities contributing at least 5% to Dimension 1 or Dimension 2 | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Active modality | n | D1 coordinate | D1 contribution (%) | D1 cos² | D2 coordinate | D2 contribution (%) | D2 cos² | ||||
| Flock size = 10,001–15,000 | 24 | 0.78 | 6.8 | 0.25 | 0.07 | 0.1 | 0.00 | ||||
| Flock size = 5,000–10,000 | 27 | 0.06 | 0.0 | 0.00 | -0.87 | 13.2 | 0.37 | ||||
| Flock size = < 5,000 | 23 | -1.18 | 14.8 | 0.53 | 0.48 | 3.4 | 0.09 | ||||
| Flock size = > 15,000 | 9 | 0.75 | 2.4 | 0.07 | 1.20 | 8.4 | 0.18 | ||||
| Nest type = Automatic | 39 | 0.82 | 12.2 | 0.60 | 0.10 | 0.2 | 0.01 | ||||
| Nest type = Manual | 44 | -0.73 | 10.8 | 0.60 | -0.09 | 0.2 | 0.01 | ||||
| Light-off = Dimmer | 12 | 0.97 | 5.2 | 0.16 | 1.23 | 11.7 | 0.26 | ||||
| Light-off = Staged | 35 | -0.03 | 0.0 | 0.00 | -0.66 | 9.9 | 0.32 | ||||
| Feeding = Ad libitum | 35 | 0.39 | 2.5 | 0.11 | 0.63 | 9.1 | 0.29 | ||||
| Feeding = One or two meals/day | 24 | -0.68 | 5.2 | 0.19 | 0.22 | 0.7 | 0.02 | ||||
| Feeding = Three meals/day | 24 | 0.11 | 0.1 | 0.01 | -1.14 | 20.2 | 0.53 | ||||
| Electrified wire = Yes | 12 | 1.05 | 6.1 | 0.19 | 0.79 | 4.8 | 0.11 | ||||
| Pullet source = Third party | 27 | -0.64 | 5.1 | 0.20 | -0.06 | 0.1 | 0.00 | ||||
| Rearing system = Conventional cages | 14 | -1.06 | 7.3 | 0.23 | 0.11 | 0.1 | 0.00 | ||||
| Transfer age = 14–16 weeks | 47 | 0.54 | 6.3 | 0.38 | -0.43 | 5.5 | 0.24 | ||||
| Transfer age = < 14 weeks | 25 | -0.80 | 7.4 | 0.27 | 0.79 | 10.1 | 0.27 | ||||
The Benzécri percentages are shares of corrected inertia and are not percentages of the original data variation. Raw inertia is therefore reported prominently
All active modalities, including coordinates, contributions, and cos² for the first two dimensions, are shown in Supplementary Table S2. Recurrence and number of location categories were supplementary and did not determine the axes
Fig. 2.

Multiple correspondence analysis of farm-level characteristics among respondents reporting piling (n = 83). (A) Raw-inertia scree plot; the dashed line is the 1/Q eigenvalue threshold expressed as a percentage of total raw inertia (Q = 8 active variables). (B) Active-category map for Dimensions 1 and 2; labels are shown for modalities contributing at least 5% to either displayed dimension. Proximity represents co-occurrence patterns, not discrete farm groups. Full coordinates, contributions, and cos² are provided in Supplementary Table S2
Dimension 2 primarily contrasted three daily meals, flocks of 5,000–10,000 hens, staged light-off, and transfer at 14–16 weeks with dimmer use, earlier transfer, ad libitum feeding, and the largest flock category. It was not interpreted as a technology tier. Supplementary centroids for recurrence and number of location categories lay near the origin, visually indicating substantial overlap. In the five-variable sensitivity MCA, the first two dimensions accounted for 26.9% and 17.7% of raw inertia and retained a similar scale/automation-management interpretation for Dimension 1.
Bias-reduced logistic regression
In the primary recurrence model, there was no evidence that flock-size category (OR per category increase 1.05; 95% CI 0.61–1.83; raw p = 0.867; Holm p = 1.000), automatic versus manual nest (OR 1.92; 95% CI 0.65–5.87; raw p = 0.237; Holm p = 0.712), or non-abrupt versus abrupt light-off (OR 1.04; 95% CI 0.40–2.63; raw p = 0.940; Holm p = 1.000) was associated with reporting repeated rather than one-time piling (Table 4). In the secondary model, none of the three farm characteristics was associated with reporting multiple location categories after adjustment for recurrence (all Holm p = 1.000). Repeated versus one-time piling had OR 1.75 (95% CI 0.68–4.76; p = 0.252) in that model. Confidence intervals were wide and remained compatible with lower or higher odds.
Table 4.
Firth bias-reduced logistic regression estimates for reported recurrence and reporting multiple piling location categories among respondents reporting piling (n = 83)
| Outcome | Predictor | Adjusted OR | 95% profile penalized-likelihood CI | Raw PLR p | Holm-adjusted p |
|---|---|---|---|---|---|
| Repeated vs. one-time piling | Flock-size category, per one-category increase | 1.05 | 0.61–1.83 | 0.867 | 1.000 |
| Repeated vs. one-time piling | Automatic vs. manual nest | 1.92 | 0.65–5.87 | 0.237 | 0.712 |
| Repeated vs. one-time piling | Non-abrupt vs. abrupt light-off | 1.04 | 0.40–2.63 | 0.940 | 1.000 |
| Multiple vs. one reported location category | Flock-size category, per one-category increase | 1.17 | 0.69–1.97 | 0.559 | 1.000 |
| Multiple vs. one reported location category | Automatic vs. manual nest | 1.48 | 0.53–4.13 | 0.452 | 1.000 |
| Multiple vs. one reported location category | Non-abrupt vs. abrupt light-off | 1.34 | 0.55–3.30 | 0.524 | 1.000 |
| Multiple vs. one reported location category | Repeated vs. one-time piling (adjustment) | 1.75 | 0.68–4.76 | 0.252 | — |
The recurrence outcome compared repeated episodes (n = 57) with one reported episode (n = 26). The multiple-location outcome compared > 1 selected category (n = 35) with one category (n = 48) and was adjusted for recurrence. Flock size was coded ordinally from 0 (< 5,000) to 3 (> 15,000). Reference categories were manual nest and abrupt light-off; non-abrupt combined staged and dimmer routines. Holm adjustment was applied to the three focal farm-exposure tests separately within each outcome. Recurrence was an adjustment covariate and was not part of the secondary multiplicity family. OR, odds ratio; CI, confidence interval; PLR, penalized likelihood ratio
Open-ended responses and sensitivity analyses
Fifty-six of the 83 piling reporters contributed text to at least one open-ended field. The most frequent themes were pullet rearing, transfer, or acclimation (15/56; 26.8%), lighting conditions and management (14/56; 25.0%), housing design or resource distribution (9/56; 16.1%), and thermal or ventilation conditions (7/56; 12.5%; Fig. 3). These themes represent perceptions, not measured exposures. Alternative flock-size and light-off codings, removal of one positive record from the exact duplicate pair, restriction of the secondary model to recurrent reports, and conventional logistic regression did not materially alter the conclusions (Supplementary Table S3). Predictor variance-inflation factors were 1.04–1.40.
Fig. 3.

Respondent-level themes in open-ended responses among piling reporters (n = 56 contributors). Both open-text fields were read together; multiple coding was allowed, and each respondent was counted at most once within a theme. Percentages can sum to more than 100%. Themes describe producer perceptions and do not represent measured exposure prevalence, effect size, or causality
Discussion
This survey represented heterogeneity in farm management regarding pillingthrough continuous multivariable dimensions. The main inferential finding was the absence of evidence that flock-size category, nest type, or light-off routine was associated with reported recurrence or with the reporting of piling in multiple location categories among respondents who had already reported the behavior. These findings should be interpreted alongside the wide confidence intervals: the data neither demonstrate equivalence nor exclude potentially meaningful associations.
The first MCA dimension summarized the joint occurrence of characteristics related to scale, automation, feeding, and pullet management. Smaller flocks, manual nests, conventional-cage rearing, earlier transfer, and fewer scheduled meals were located on one side of the dimension, whereas larger flocks, automatic nests, dimmer use, electrified wires, and transfer at 14–16 weeks were located on the opposite side. This pattern indicates that technology and production scale cannot be clearly separated in this sample. The second dimension was more closely related to contrasts involving feeding, light-off routine, transfer age, and flock size and should not be interpreted as a validated technology or welfare score. The five-variable sensitivity analysis supported the primary scale/automation interpretation without demonstrating the existence of discrete farm types.
The secondary outcome requires additional caution. Selecting more than one location category may combine observations from different episodes and does not demonstrate that piling occurred simultaneously at several locations. Respondents who reported repeated episodes also had more opportunities to observe the behavior at different locations, which justified adjusting for recurrence and conducting a sensitivity analysis restricted to recurrent reports. Neither analysis provided evidence that the selected farm characteristics were associated with reporting multiple locations. Walls, corners, partitions, and nests remain biologically plausible congregation sites and have been described in previous studies (Barrett et al. 2014; Rayner et al. 2016; Gray et al. 2020; Winter et al. 2022a). However, the underlying spatial mechanisms could not be evaluated without information on barn layout, local bird density, and bird movement.
The open-ended responses most frequently mentioned aspects related to pullet rearing, transfer or acclimation to layer housing, and lighting conditions. Transfer to the production barn may temporarily alter bird distribution and resource use, whereas localized light spots and pronounced contrasts between light and dark areas may attract hens and produce spatial concentrations. These mechanisms are plausible and have previously been discussed or evaluated (Gray et al. 2020; Winter et al. 2022a, b), but they were not directly measured in the present survey. The frequency of a theme represents what respondents chose to mention, rather than the magnitude of its effect, the strength of its association, or its causal importance.
Concerning lighting inside the barns should also be distinguished from the prospective Australian finding that the risk of smothering was greater in shaded outdoor range areas (Chowdhury et al. 2026). The settings and outcomes differed: the present study considered localized indoor light and farmer-reported piling, whereas the cohort study evaluated outdoor shade and confirmed smothering. These observations may reflect different behavioral processes rather than contradictory evidence. Similarly, the higher risk of smothering reported in aviaries compared with flat-deck systems in another Australian cohort (Chowdhury et al. 2025) cannot be directly extrapolated to the present study because production-barn configuration was not recorded.
Correct terminology is essential for interpreting the findings. Participants received a qualitative explanation of piling, but no quantitative threshold was established for group size, density, or episode duration. Smothering requires death from suffocation resulting from piling, and the questionnaire included neither a structured mortality measure nor independent confirmation of the cause of death . Therefore, the outcome assessed was farmer-reported piling, not confirmed smothering. Differences in recognition, recall, and the period of experience considered by each respondent may have introduced outcome misclassification and variation in the opportunity to report recurrence.
Other limitations include voluntary participation by producers affiliated with the MIRA network and the absence of a standardized observation period. Current or usual farm characteristics may not have been identical to the conditions present when previous piling episodes occurred, and some management practices may have been modified after an episode. Restricting the analyses to respondents who reported piling prevents inferences regarding the determinants of occurrence and may produce selection-related associations. Bird uniformity at transfer was recorded inconsistently, and several environmental, structural, and resource-distribution characteristics were not collected. Although multiplicity was controlled within each family of tests and the sensitivity analyses yielded similar interpretations, the limited sample size reduced the precision of the estimates.
Future prospective studies should include flocks with and without piling, apply a standardized operational definition, distinguish nonfatal episodes from confirmed cases of smothering, and record the number of birds and time at risk. Repeated objective measurements should include production-barn configuration, usable floor area and stocking density, litter conditions, illuminance and spatial light gradients, temperature, ventilation, feeder and drinker distribution, environmental disturbances, and bird movement before and during episodes. Such study designs will be required to determine whether the contexts perceived by producers correspond to reproducible risk or protective factors.
Conclusion
Among Brazilian cage-free farmers who reported piling, practices related to farm scale, housing, feeding, and pullet management varied along continuous multivariable dimensions rather than forming clearly separated farm types. The selected farm characteristics were not independently associated with reported recurrence or with reporting multiple location categories, although the confidence intervals were wide. Open-ended responses most frequently referred to pullet transfer or acclimation and lighting conditions as risk factors for piling. These findings are hypothesis-generating and do not estimate the incidence of piling, identify determinants of its occurrence, establish causality, or confirm smothering.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
The authors thank the Brazilian cage-free egg producers who voluntarily participated in the survey and shared their experiences, making this study possible.
Author contributions
Caroline Citta Mazocco: Conceptualization; Methodology; Validation; Investigation; Data curation; Writing – original draft. Robson Mateus Freitas Silveira: Conceptualization; Validation; Formal analysis; Investigation; Visualization; Writing – review and editing. Rosangela Poletto: Conceptualization; Validation; Investigation; Visualization; Writing – review and editing. Iran José Oliveira da Silva: Conceptualization; Resources; Supervision; Project administration; Writing – review and editing; Conceptualization; Methodology; Validation; Investigation.
Funding
The Article Processing Charge (APC) for the publication of this research was funded by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) (ROR identifier: 00x0ma614).
Data availability
The de-identified questionnaire data and reproducible analysis code will be made available on reasonable request, subject to the authors’ and institution’s data-sharing requirements.
Declarations
Ethical approval
According to the institutional assessment, the study was exempt from formal ethics committee review under Brazilian National Health Council Resolution No. 510/2016 because it involved an anonymous survey of professional production practices and did not collect identifiable personal or sensitive information.
Generative AI and AI-assisted technologies
During the preparation of this manuscript, the authors used ChatGPT (OpenAI) to translate the text from Portuguese into English and to improve its readability, cohesion and coherence. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.
Consent to participate
Participation was voluntary. Respondents were informed of the study purpose before answering, and submission of the questionnaire was taken as consent to participate.
Conflict of interest
Robson Mateus Freitas Silveira is an Associate Editor of Tropical Animal Health and Production. He was not involved in the editorial handling or peer review of this manuscript.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1007/s11250-026-05455-2.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Robson Mateus Freitas Silveira, Email: robsonsilveira@usp.br.
Iran José Oliveira da Silva, Email: iranoliveira@usp.br.
References
- Arno A, Silveira RMF, da Silva IJO (2023) Characterization, typification and holistic consumer perception of welfare in laying poultry in Brazil: a machine learning approach. J Agric Sci 161(5):743–753. 10.1017/S0021859623000552 [DOI]
- Barrett J, Rayner AC, Gill R, Willings TH, Bright A (2014) Smothering in UK free-range flocks. Part 1: incidence, location, timing and management. Vet Rec 175:19. 10.1136/vr.102327 [DOI] [PubMed] [Google Scholar]
- Brasil, Ministério da Saúde, Conselho Nacional de Saúde (2016) Resolução nº 510, de 7 de abril de 2016. Diário Oficial da União, Brasília, DF, no. 98, seção 1, pp 44–46. https://www.gov.br/conselho-nacional-de-saude/pt-br/atos-normativos/resolucoes/2016/resolucao-no-510.pdf/view. Accessed 16 Sept 2026
- Chowdhury P, Hemsworth PH, Fisher AD, Rice M, Galea RY, Taylor PS, Stevenson M (2024) Descriptive epidemiology of smothering in Australian commercial free-range layer hen farms. Prev Vet Med 223:106098. 10.1016/j.prevetmed.2023.106098 [DOI] [PubMed] [Google Scholar]
- Chowdhury P, Hemsworth PH, Fisher AD, Rice M, Galea RY, Taylor PS, Stevenson M (2025) Risk factors for smothering in three commercial free-range layer poultry farms, Australia 2019–2022. Prev Vet Med 242:106568. 10.1016/j.prevetmed.2025.106568 [DOI] [PubMed] [Google Scholar]
- Chowdhury P, Fisher AD, Rice M, Taylor PS, Stevenson M, Hemsworth PH (2026) The effect of shadowed areas on smothering risk in three commercial free-range layer poultry farms, Australia 2019–2022. Prev Vet Med 254:106944. 10.1016/j.prevetmed.2026.106944 [DOI] [PubMed] [Google Scholar]
- Firth D (1993) Bias reduction of maximum likelihood estimates. Biometrika 80:27–38. 10.1093/biomet/80.1.27 [DOI] [Google Scholar]
- Gray H, Davies R, Bright A, Rayner A, Asher L (2020) Why do hens pile? Hypothesizing the causes and consequences. Front Vet Sci 7:616836. 10.3389/fvets.2020.616836 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Greenacre M (2017) Correspondence analysis in practice, 3rd edn. Chapman & Hall/CRC, Boca Raton. 10.1201/9781315369983 [DOI] [Google Scholar]
- Harju C, Lähtinen K, Heinola K, Väre M, Bonnefous C, Collin A, Cozma V, Kliphuis S, Parrott PA, Rodenburg TB, Spinu M, Niemi J (2024) Consumers’ views on egg quality and preferences for responsible production – results from nine European countries. Br Food J 126:623–644. 10.1108/BFJ-03-2023-0183 [DOI] [Google Scholar]
- Hartcher KM, Jones B (2017) The welfare of layer hens in cage and cage-free housing systems. Worlds Poult Sci J 73:767–782. 10.1017/S0043933917000812 [DOI] [Google Scholar]
- Heinze G, Schemper M (2002) A solution to the problem of separation in logistic regression. Stat Med 21:2409–2419. 10.1002/sim.1047 [DOI] [PubMed] [Google Scholar]
- Holm S (1979) A simple sequentially rejective multiple test procedure. Scand J Stat 6:65–70. 10.2307/4615733 [DOI] [Google Scholar]
- Mazocco CC, de Castro Júnior SL, Silveira RMF, Poletto R, da Silva IJO (2024) Laying hens: Why smothering and not surviving?—A literature review. Animals 14(11):1518. 10.3390/ani14111518 [DOI] [PMC free article] [PubMed]
- Rayner AC, Gill R, Brass D, Willings TH, Bright A (2016) Smothering in UK free-range flocks. Part 2: investigating correlations between disease, housing and management practices. Vet Rec 179:252. 10.1136/vr.103701 [DOI] [PubMed] [Google Scholar]
- Winter J, Stratmann A, Toscano MJ, Cardwell JM, Nicol CJ (2022a) Piling behaviour in British layer flocks: observations and farmers’ experiences. Appl Anim Behav Sci 253:105686. 10.1016/j.applanim.2022.105686 [DOI] [Google Scholar]
- Winter J, Toscano MJ, Stratmann A (2022b) The potential of a light spot, heat area, and novel object to attract laying hens and induce piling behaviour. Animal 16:100567. 10.1016/j.animal.2022.100567 [DOI] [PubMed] [Google Scholar]
- Xin H, Gates RS, Green AR, Mitloehner FM, Moore PA Jr, Wathes CM (2011) Environmental impacts and sustainability of egg production systems. Poult Sci 90:263–277. 10.3382/ps.2010-00877 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The de-identified questionnaire data and reproducible analysis code will be made available on reasonable request, subject to the authors’ and institution’s data-sharing requirements.
