Meta-analysis of poultry organ weights and their relationship with meat yield

Abstract

Late-stage mortality is a significant challenge for the poultry industry, leading to substantial economic losses, concerns about animal welfare, and operational sustainability. Heart-related conditions, including ascites syndrome, pulmonary hypertension syndrome, hypertrophic cardiomyopathy, and sudden death syndrome, contribute significantly to this issue. The increasing prevalence of these conditions is potentially linked to intense selection pressure aimed at maximizing meat yield, particularly breast meat. However, the precise relationship between meat yield, heart size and cardiovascular health remains unclear. To address this, a systematic literature review and meta-analysis were conducted to explore the relationship between breast meat yield and organ size (heart, lungs, liver), in which 91 publications meeting specific inclusion criteria were identified. Data extracted included variables such as live weight, portion yields (breast, leg, wing), organ weights (heart, lungs, liver), and the prevalence of heart-related conditions (pulmonary hypertension syndrome, ascites syndrome). A backward selection modeling approach was used to develop linear mixed models, treating the study as a random effect, to examine the relationship between organ weights as a percentage of body weight (% BW), meat yield and other animal attributes. The best heart weight model (% BW) included the effects of sex, species (chicken or turkey), bird purpose (meat or egg), breast meat yield (%), and live weight (g). The best liver weight model (% BW) included species, bird purpose, breast meat yield (%), and live weight (g). The best lung weight (% BW) model included heart weight (g). Model performance was evaluated using residuals analysis, root mean squared prediction error, and the concordance correlation coefficient. Findings suggest that laying hens have larger hearts relative to body weight compared to broiler chickens and turkeys. The liver and lung models revealed that broiler chickens had larger livers (% BW) compared to laying hens, and that lung weight (% BW) was negatively correlated to heart weight (g). These results highlight the potential need to consider organ health in breeding programs focused on meat yield.

INTRODUCTION

Intense genetic selection for meat yield in poultry has driven rapid growth and heavier body weights, leading to impressive gains in production performance. However, these advancements are offset by profitability challenges, potentially due to increased mortality rates, impacting both early and late life stages. Compared to early mortality, which is defined in chickens as deaths occurring before four weeks and in turkeys before 11 weeks, late mortality is more costly given the time and cumulative resources invested (Richter et al., 2024). Cardiovascular issues are among the most common causes of late mortality in poultry species (Cherian, 2007).

While factors influencing mortality are numerous and multifaceted, one of the hypotheses for why mortality rates have continued to rise in recent years is that intense selection for meat yield, in particular breast meat yield (BMY), might be impairing organ function due to reduced body cavity size. Body cavity size has rapidly diminished over the past 50 years in broilers (Havenstein et al., 2003a; 2003b) and turkeys (Havenstein, 2006; Havenstein et al., 2007). The reduced body cavity space restricts organ growth, resulting in organs that are too small to fully support the animal (Havenstein et al., 2003b). This could lead to alterations in the size, shape, and function of various organs, potentially making organ size a limiting factor for further gains in poultry production and efficiency (Gaya et al., 2006). Suboptimal organ function has obvious implications for animal health and welfare, as well as profitability, but poor organ function can also lead to economic losses via increased morbidity and mortality. Furthermore, it has been linked to the development of breast myopathies, such as white striping and woody breast syndrome, which has secondary economic implications (Xing et al., 2021; Ayansola et al., 2021).

Proper development and function of key organs, such as the liver, lungs, and heart, is critical for maintaining the overall health and general welfare of poultry. One of the predominant organ systems of concern is the cardiovascular system. The prevalence of cardiovascular issues in poultry is increasing (Druyan et al., 2007) and is potentially related to deficiencies in organ development and/or the inability of the developed system to properly supply oxygen and nutrients to the large mass of tissue seen in modern birds (Habibian et al., 2017). Increased demands for growth may overly strain the developing cardiovascular system, especially in early life, resulting in mortality (Closter et al., 2009, 2012; Dou et al., 2017; Li et al., 2022). Fast-growing broilers are more susceptible to ascites than slow-growing broilers due to oxygen shortage, which increases pulmonary arterial pressure, straining the heart and leading to high blood pressure, fluid accumulation in the abdominal cavity, and, potentially, death (Closter et al., 2009). Cardiovascular-related issues such as ascites syndrome, sudden death syndrome, and pulmonary hypertension, are a leading cause of mortality in poultry (Özkan et al., 2006). In 2009, mortality caused by ascites alone ranged from 5% to 8% in chicken populations worldwide and reached as high as 30% in flocks of heavier birds (Closter et al., 2009). The occurrence of ascites has since gone down in a large part due to genetic selection. However, this remains a prominent issue today, with mortality rates in North American broiler flocks reported as high as 6%, equating to a loss of over half a billion birds annually (National Chicken Council, 2024).

Another major organ of concern is the liver, which is the main organ involved in metabolism (Dou et al., 2017). Rapid growth and higher meat yields are associated with high metabolic demands, especially in chickens. Poor liver performance can result in reduced growth, poor feed conversion, impaired immune function, and even substantial mortality from these conditions. The liver's role as a regulator of metabolism and blood circulation is crucial for maintaining overall health, and any dysfunction can lead to systemic issues. For instance, birds affected by ascites often show signs of liver enlargement or congestion, with the organ appearing congested, shrunken, or even nodular, highlighting its vulnerability under stress conditions (Wang and Hacker, 1993). This damage is further compounded by factors like cold stress, which can increase lipid peroxidation and elevate serum enzymes, indicating severe liver cell damage (Fathi et al., 2023). Moreover, poultry with greater meat yield often have heavier and sometimes compromised livers, suggesting that high production demands can strain the organ (Xing et al., 2021). Given these risks, the liver is a critical organ to consider in poultry health management to prevent complications like impaired immune function and to ensure optimal growth and feed efficiency.

The development and growth of organs is complex; numerous factors can influence organ size, including genetic strain, sex, age, and nutrition (Tůmová and Chodová, 2018). This can make distilling the relationship between organ size and meat yield across studies challenging. Meta-analysis can serve as a robust statistical approach for quantitatively synthesizing research findings, assessing inter-study variability, identifying overarching trends, and deriving conclusions with greater statistical power than those based on individual studies alone (Sauvant et al., 2008). As breast meat yield remains one of the most economically significant traits in the poultry industry (Lorentz et al., 2011), quantifying the relationship between meat yield and organ weight may be informative for reducing cardiovascular susceptibility in breeding programs (Damaziak et al., 2015). Understanding a bird’s susceptibility to these issues is essential to reducing mortality, improving welfare, and increasing the profitability and sustainability of the industry. Therefore, the objective of this study was to employ a meta-analysis approach to quantify the relationship between the weights of key organs (heart, liver, and lungs) and breast meat yield in poultry, while also identifying additional factors that may influence organ weight.

MATERIALS AND METHODS

Database Development

A systematic literature search was conducted in January 2024 using Web of Science, Google Scholar, PubMed, and ProQuest databases, as well as handsearching. There were no limitations imposed regarding publication date or language to the searches. The search terms used were: (chicken OR turkey) AND ("breast weight" OR "body weight" OR "live weight") AND ("heart weight" OR "lung weight" OR "liver weight" OR "organ weight") AND (ascites OR “aortic rupture” OR “pulmonary hypertension” OR “sudden death syndrome” OR “hypertrophic cardiomyopathy”) NOT (human OR “homo sapiens” OR people). These term combinations were applied to identify papers that included poultry organ weights and BMY, which formed the initial database for this study (N = 539 studies). Inclusion/exclusion criteria were then applied to evaluate papers within the database to identify the studies to be included in the meta-analysis (Figure 1). Papers that did not mention poultry, heart weight, liver weight, lung weight, or BMY in the titles, abstracts, or keywords were excluded from further examination, as well as if they were not primary research articles. Only studies on broiler chickens, laying hens, and turkeys were considered for this meta-analysis, other poultry species (i.e., quail, duck) were excluded due to a low number of studies (N_quail = 9, N_duck = 5). Full-length papers were reviewed to determine if both BMY and organ weight were included. Of the initial 539 studies identified, 91 met the criteria for inclusion in the meta-analysis and are summarized in Table 1. The literature funnel, describing the literature collection process, is illustrated in Figure 1.

Figure 1.

Literature funnel for analyzing the relationship between meat yield and organ weight.

Table 1.

Summary description of publications included in the meta-analysis, with information on species in the study, purpose of the birds, heart percent, liver weight, and lung weight. Organ weights represent averages across all study treatments if the study included multiple treatments.

Publication	Species	PP1	n2	Heart (% BW)3	Liver (% BW)3	Lung (% BW)3
Abdelfatah et al. (2023)	Chicken	Meat	4	0.83±0.026	3.60±0.084	-
Abuoghaba et al. (2021)	Chicken	Meat	6	0.58±0.007	2.64±0.061	-
Acar et al. (1995)	Chicken	Meat	5	0.56±0.016	-	-
Aghashahi et al. (2015)	Chicken	Meat	5	0.63±0.023	-	-
Ahmadipour et al. (2017)	Chicken	Meat	3	0.72±0.058	2.55±0.180	-
Ahmadipour et al. (2019a)	Chicken	Meat	3	0.72±0.058
Ahmadipour et al. (2019b)	Chicken	Meat	3	0.57±0.039	2.03±0.123	-
Azizian and Saki (2021)	Chicken	Meat	3	0.47±0.012	-	0.42 ± 0.031
Balog et al. (1994)	Chicken	Meat	6	-	2.64 ± 0.032	-
Balog et al. (2003)	Chicken	Meat	3	0.40±0.009	2.75 ± 0.085	-
Bölükbasi et al. (2005)	Chicken	Meat	4	0.58±0.013	-	-
Boostani et al. (2010)	Chicken	Meat	4	0.37±0.005	-	-
Bosco et al. (2014)	Chicken	Meat	6	0.37±0.022	1.45 ± 0.060	-
Buys et al. (1999)	Chicken	Meat	10	0.69±0.035	2.64 ± 0.078	0.73 ± 0.022
Camacho-Fernandez et al. (2002)	Chicken	Meat	8	-	2.36 ± 0.070	-
Damaziak et al (2012)	Turkey	Meat	4	0.43±0.048	1.45 ± 0.194	-
Damaziak et al (2015)	Turkey	Meat	8	0.67±0.057	1.50 ± 0.149	-
Daneshyar et al. (2007)	Chicken	Meat	2	0.70±0.030	2.39 ± 0.145	0.67 ± 0.005
Delfani et al (2023)	Chicken	Meat	3	0.59±0.026	2.54 ± 0.215	0.58 ± 0.012
Dou et al (2017)	Chicken	Meat	2	0.49±0.120	1.20 ± 0.300	0.38 ± 0.145
Fathi et al (2023)	Chicken	Meat	5	0.68±0.023	-	-
Galal et al (?)	Chicken	Meat	8	-	2.00 ± 0.082	0.22 ± 0.009
Gao et al (2017)	Chicken	Meat	5	-	2.58 ± 0.034	-
Gaya et al (2006)	Chicken	Meat	2	0.52±0.001	2.03 ± 0.004	-
Gholami et al (2020)	Chicken	Meat	16	0.62±0.001	-	-
Gopi et al (2014)	Chicken	Meat	9	1.07±0.034	2.72 ± 0.069	-
Govaerts et al (2000)	Chicken	Meat	4	0.39±0.006	1.92 ± 0.041	0.44 ± 0.014
Harash et al (2019)	Chicken	Meat	2	0.62±0.007	-	-
Hartinger et al (2021)	Chicken	Meat	3	0.54±0.008	2.25 ± 0.059	-
Haunshi et al (2021)	Chicken	Meat	2	0.43±0.045	1.63 ± 0.070	-
Huang et al (2011)	Chicken	Meat	3	0.49±0.035	-	-
Jahanpour et al (2020)	Chicken	Meat	15	0.47±0.014	-	-
Khajali et al (2011)	Chicken	Meat	4	0.59±0.015	2.30 ± 0.077	-
Kranen et al (1996)	Chicken	Meat	4	0.43±0.028	-	-
Li et al (2022)	Chicken	Meat	1	-	1.82 ± N/A	-
Lieboldt et al (2016)	Chicken	Egg	12	0.49±0.012	2.38 ± 0.045	-
Lorentz et al (2011)	Chicken	Meat	1	-	1.91 ± N/A	-
Malan et al (2003)	Chicken	Meat	7	-	-	0.46 ± 0.022
Malan et al (2007)	Chicken	Meat	9	0.71 ± 0.002	2.90 ± 0.024	0.09 ± 0.001
Martinez-Lemus et al (1998)	Chicken	Meat	2	0.58 ± 0.012	-	0.56 ± 0.032
Maxwell et al (1986)	Chicken	Meat	2	0.84 ± 0.157	-	-
McGovern et al (2000)	Chicken	Meat	4	0.48 ± 0.004	-	-
Milani et al. (2020)	Chicken	Meat	4	-	2.56 ± 0.070	-
Molenaar et al (2011)	Chicken	Meat	4	0.32 ± 0.012	-	-
Namakparvar et al (2014)	Chicken	Meat	5	-	-	0.55 ± 0.027
Oke et al (2021)	Chicken	Meat	7	0.42 ± 0.044	2.07 ± 0.281	-
Owen et al (1995)	Chicken	Meat	3	-	3.10 ± 0.291	0.66 ± 0.045
Özkan et al (2006)	Chicken	Meat	4	0.68 ± 0.018	2.21 ± 0.096	0.46 ± 0.018
Payvastegan et al (2017)	Chicken	Meat	4	0.44 ± 0.019	2.31 ± 0.079	-
Peterson et al (1973)	Turkey	Meat	4	0.12 ± 0.003	-	-
Poltowicz et al (2015)	Chicken	Meat	3	0.68 ± 0.028	3.92 ± 0.182	-
Portillo-Salgado et al. (2023)	Turkey	Meat	2	-	1.61 ± 0.182	-
Rahimi et al (2015)	Chicken	Meat	7	-	2.33 ± 0.082	-
Saedi et al (2010)	Chicken	Meat	2	0.60 ± 0.020	2.45 ± 0.100	-
Sandercock et al (2009)	Chicken	Egg	2	0.78 ± 0.020	-	-
Sandercock et al (2009)	Chicken	Meat	6	0.62 ± 0.036	-	-
Sharifi et al (2015)	Chicken	Meat	2	0.65 ± 0.070	2.63 ± 0.245	-
Tickle et al (2014)	Chicken	Meat	1	0.53 ± N/A	2.48 ± N/A	0.55 ± N/A
Toghyani et al (2011)	Chicken	Meat	3	0.55 ± 0.034	2.08 ± 0.019	-
Tůmová and Chodová (2018)	Chicken	Meat	6	0.56 ± 0.017	2.50 ± 0.120	-
Tůmová et al. (2020)	Turkey	Meat	2	0.36 ± 0.032	0.97 ± 0.067	-
Varmaghany et al (2021)	Chicken	Meat	5	0.50 ± 0.037	-	-
Venturini et al (2014)	Chicken	Meat	1	0.56 ± N/A	2.36 ± N/A	-
Wang Hacker (1993)	Chicken	Meat	4	0.80 ± 0.008	2.90 ± 0.082	-
Wideman Jr et al (1999)	Chicken	Meat	3	-	-	0.64 ± 0.030
Wideman Jr et al (1996)	Chicken	Meat	2	-	-	0.60 ± 0.024
Wilson et al (2016)	Chicken	Meat	2	0.53 ± 0.004	-	-
Yahav Plavnik (1999)	Chicken	Meat	8	0.42 ± 0.008	-	-
Zamani et al (2017)	Chicken	Meat	3	-	2.62 ± 0.037	-

¹Poultry Purpose

²Treatment means per study

³Organ weight, as a percent of body weight, mean ± standard error across study treatments. NA indicates that standard errors could not be reported due to the study containing only one treatment mean

Data extracted from each paper was collated in an Excel database and included study characteristics (location, year, etc.), species (chicken and turkey), purpose of the birds (meat and egg), selection intention (commercial and local), experimental treatments (diet, strain, breed, sex, etc.), slaughter age, age at measurement, sex, carcass component weights (whole breast, leg, wing, dressing percent, etc.), organ weights (heart, lung, gizzard, liver, etc.), diet factors, and cardiovascular issue occurrence. The ‘selection intention’ variable was determined based on whether the article stated if the birds were a commercial or a local/regional breed. Not all papers reported all variables, thus the number of studies and observations included in each data column varied.

Within the process of preparing the dataset for analysis, steps were taken to refine and optimize the variable representation. The variable representing the purpose of the bird initially consisted of four levels (broiler, breeder, layer, multi-purpose), but was collapsed to two categories: egg-type and meat-type birds. Additionally, categorical variables exhibiting minimal variation (e.g., selection intention, ventricular hypertrophy index, pulmonary hypertension syndrome incidence), and those variables with over 85% of observations falling into a single category were omitted from the analysis to avoid redundancy and limit potential bias, unless the remaining variation was deemed biologically relevant (e.g., cardiovascular issue incidence). Furthermore, new variables were created to capture important interactions. For example, a composite variable called ‘poultry purpose’ was created to represent the size differences, growth rate differences and selection intention between species of birds by combining the species and purpose of the birds to three categorical variables: chicken egg-type (laying hens), chicken meat-type (broiler chickens), and turkey meat-type birds (turkeys). The observations that for sex fell under the ‘unsexed population’ categorization were excluded from model development.

There was variation between studies in how consistently carcass portions and organ weights were reported. Therefore, organ weights were converted to measurement units of weight as a % BW. This unit was selected instead of e.g., weight in g or kg, to allow easier comparison between turkeys and chickens, by accounting for the substantial size differences between these two species.

Before developing the model, descriptive statistics were computed using SAS Studio (version 3.81, SAS/STAT, SAS Institute Inc., Cary, NC) including the mean ± standard error, median, minimum, maximum, and standard deviation for both dependent and independent variables. The results of these calculations are detailed in Table 2, which presents the descriptive statistics for heart weight, liver weight, and lung weight, all expressed as a percentage of body weight (% BW).

Table 2.

Descriptive statistics of the data included in the meta-analysis, separated by poultry purpose.

Category1	Effect	n2	Mean3		SE	Median4	Min5	Max6
Heart weight
Chicken Egg	Heart (% BW)7	14	0.53	0.030		0.51	0.43	0.80
	BMY (% BW)	2	18.00	0.500		18.00	17.50	18.50
	Live weight (g)	14	1012.36	48.238		992.00	554.00	1262.00
Chicken Meat (Female)	Heart (% BW)	6	0.44	0.049		0.44	0.29	0.59
	BMY (% BW)	10	26.11	2.447		26.70	16.60	35.02
	Live weight (g)	16	2132.78	165.489		2035.00	1220.0	3803.00
Chicken Meat (Male)	Heart (% BW)	154	0.60	0.014		0.58	0.27	1.26
	BMY (% BW)	81	25.67	0.703		24.78	14.43	36.57
	Live weight (g)	269	2113.94	42.799		2116.40	437.00	4578.00
Turkey Meat (Female)	Heart (% BW)	9	0.47	0.085		0.45	0.12	0.84
	BMY (% BW)	8	21.95	1.685		22.73	13.54	28.83
	Live weight (g)	24	4137.59	756.173		2955.57	349.00	12697.00
Turkey Meat (Male)	Heart (% BW)	9	0.45	0.086		0.43	0.12	0.86
	BMY (% BW)	8	23.54	2.100		23.12	12.86	32.42
	Live weight (g)	25	6433.96	1346.250		4106.00	376.00	22396.30
Liver weight
Chicken Egg	Liver (% BW)	12	2.38	0.045		2.36	2.15	2.65
	BMY (% BW)	2	18.00	0.500		18.00	17.50	18.50
	Live weight (g)	14	1012.36	48.238		992.00	554.00	1262.00
Chicken Meat	Liver (% BW)7	161	2.44	0.041		2.47	0.9	4.24
	BMY (% BW)	128	27.16	0.593		29.20	13.22	36.57
	Live weight (g)	36	2116.76	36.063		2150.80	63.12	4578.00
Turkey Meat	Liver (% BW)	16	1.44	0.098		1.47	0.90	2.07
	BMY (% BW)	16	22.74	1.317		22.80	12.86	32.42
	Live weight (g)	49	5309.21	789.933		3900.00	349.00	22396.30
Lung weight
Chicken Meat	Lung (% BW)	68	0.46	0.026		0.50	0.09	0.87
	Heart weight (g)	247	13.66	0.656		12.18	0.30	79.67

¹How animals are categorized by model to show estimates.

²Number of treatment means.

³Average value of each trait per category.

⁴Median value of each trait per category.

⁵Minimum value of each trait per category.

⁶Maximum value of each trait per category.

⁷(% BW), percent body weight.

Model Development

Models were developed and tested using PROC CORR, PROC MIXED and PROC GLIMMIX in SAS (SAS version 3.81, SAS/STAT, SAS Institute Inc, Cary NC.). For all models developed in this study, a beta distribution was specified in GLIMMIX due to the dependent variable (heart, liver, lung) being expressed as a proportion of BW, and a LOGIT link function was used. All model parameter estimate results presented are therefore on the model (beta) scale.

Before multivariable modelling, univariable analysis was conducted using PROC GLIMMIX to individually examine each independent variable and its relationship with heart weight (% BW), liver weight (% BW), and lung weight (% BW) as the outcome variables of interest. If a variable at the univariable model level had observations from 10 or more studies and a P-value of < 0.1 then the variable moved on to consideration in multivariable modelling (Table S1).

A multivariable generalized mixed model approach was applied, with a random effect of study on the model intercept term according to St-Pierre (2001), as follows:

$$\begin{array}{c}\hfill Y_{ijk}=B_0+B_{1_{ijk}}*X_{1_{ijk}{n}_{ijk}}+B_{2_{ijk}}*T_{1_{ijk}{n}_{ijk}}+B_{3_{ijk}}*M_{1_{ijk}{n}_{ijk}}+S_i+e_{ijk},\end{array}$$ (1)

where i =1, … , n studies, j = 1, ... , n_i observations, B₀ = overall intercept, S_i = random intercept by study, X_1ij = fixed effects of the model for the ith study and jth observation for continuous variables, B_1ijk = slope coefficient of the X_1ij fixed effect for continuous variables, T_1ij = fixed effects of the model for the ith study and jth observation for categorical variable levels k = 1, … , n, B_2ijk = coefficient for categorical variable T_1ij levels k = 1, … , n, M_1ij = fixed effects of the model for the ith study and jth observation for the interaction of continuous and categorical variables, B_3ijk = slope coefficient of the fixed effect for the interaction of continuous and categorical variables M_1ij, and e_ij = residual error of the model. The Y variables considered for analysis were heart weight, liver weight, and lung weight (% BW).

A correlation analysis (Table S2) was performed using PROC CORR on the dependent and independent variables, prior to multivariable model development. This analysis was performed to identify potential variables related to the independent variables heart weight, liver weight, and lung weight (% BW) and to identify collinear and redundant variables within the database. A cut-off value of r > 0.40 (Grewal et al., 2004) was used to identify substantial collinearity between variables. If two variables were collinear, the variable with the lower univariable P-value was chosen. If the variables had identical P-values, then the more biologically relevant variable was chosen.

Multivariable models were developed for the heart and liver (% BW) utilizing the structure described in Eq. 1. To develop multivariable models, a backward stepping modeling approach was implemented by first creating a ‘full’ model for each independent variable (while avoiding collinearity between terms), and then sequentially removing the least significant term (highest P-value) at each step and evaluating the significance of the remaining variables. Two-way interactions were also assessed for significance and consideration in the models.

Throughout model building, the residuals and distribution of the random study effect estimates were visually assessed for normality, while the Akaike Information Criterion corrected (AICc) value was evaluated for model improvement or deterioration. To account for variability in sample size and SEM across experiments, a weight statement was implemented in PROC GLIMMIX to weight the regression by the corresponding treatment mean sample size. This was done by dividing the population for each treatment mean by the average population across the dataset to center the weighting factor around 1 (St-Pierre, 2001).

The Cook’s distance test (Cook, 1977) was used to identify observations with a significant influence on the estimates produced by the model, using a normal approximation. The Cook’s distance quantifies the change in fitted response values of the regression after observation removal (Cook, 1977) and this statistic was computed for each observation for each of the models. Observations with a Cook’s distance >4/n, where n represents the total number of observations, were considered influential and were removed from the data set in a stepwise fashion until no further improvement to the model was observed. After identifying and removing any influential points, the PROC GLIMMIX procedure in SAS was used to develop models based on the beta distribution, incorporating a ‘variance components’ (VC) matrix structure of the random effect.

To aid interpretation of the model’s parameters which exist on the Beta scale, a behavior analysis was conducted to facilitate evaluation of the driving variable impacts on the back transformed outcome variables. First, the models developed were implemented to predict the outcome variable of interest on the beta scale (Y_B), and one driving (X) variable was changed at a time, while the others were held constant. The Y_B predictions were back-transformed to the data scale (Y_BT) using the formula:

$$Y_{BT}=\left(\frac{EXP\left(Y_B\right)}{1+EXP\left(Y_B\right)}\right)\times 100$$ (2)

then, the magnitude of a ±10% change in the Y variable was divided by the magnitude change it caused in the back-transformed X variable (Y_BT), and multiplied by 100, indicating the magnitude of impact in % change on the data scale.

Model Evaluation

Models were moved forward into model evaluation if all the fixed effects included in the model were significant (P<0.05) and the residuals and random effect distribution were normal. Models were evaluated for predictability using mean square prediction error (MSPE) (Bibby and Toutenburg, 1978), calculated as:

$$\begin{array}{c}\hfill \mathit{MSPE}=\sum _{i=1}^n\frac{{(Y_{\mathit{pred}}-Y_{\mathit{obs}})}^2}{n},\end{array}$$ (3)

where Y_obs is the observed value, Y_pred is the predicted value and n is the total number of observations. The square root of mean square prediction error (RMSPE) provides an estimate of the overall prediction error. RMSPE is commonly expressed as a portion of the observed mean (RMSPE, %).

The MSPE was decomposed into three components, error due to random disturbance (ED), error due to deviation of the regression slope from unity (ER), and the error in central tendency due to overall bias (ECT). These fractions of MSPE were calculated as:

$$\begin{array}{c}\hfill \mathit{ECT}={(\bar{O}-\bar{P})}^2,\end{array}$$ (4)

$$\begin{array}{c}\hfill \mathit{ER}={(S_p-R\times S_o)}^2,\end{array}$$ (5)

$$\begin{array}{c}\hfill \mathit{ED}=(1-R^2)\times S_o^2,\end{array}$$ (6)

where $\overline{O}$ and $\overline{P}$represent the observed and predicted means, respectively, S_p is the standard deviation of predicted values, S_o is the standard deviation of observed values, and R is the Pearson correlation coefficient. The ECT, ER and ED are displayed as a % of the total MSPE.

Additionally, the agreement between predicted and observed values was assessed using the concordance correlation coefficient (CCC) (Lin, 1989), which simultaneously assesses the precision and accuracy of a model. The CCC ranges from -1 to 1, with -1 indicating perfect unrelatedness, 0 indicating no relationship, and 1 indicating perfect relatedness. The CCC was calculated as:

$$\begin{array}{c}\hfill \mathit{CCC}=R\times C_b,\end{array}$$ (7)

where R is the Pearson correlation coefficient, an indication of precision, and C_b is a bias correction factor, an indication of accuracy. C_b ranges from 0 to 1, with 1 indicating no deviation of the regression line from the line of unity. The C_b variable is calculated as:

$$\begin{array}{c}\hfill C_b=\frac{2}{(v+\frac{1}{v}+{\mu }^2)},\end{array}$$ (8)

Where v and m are explained by the following equations:

$$\begin{array}{c}\hfill v=\frac{S_o}{S_p},\end{array}$$ (9)

$$\begin{array}{c}\hfill \mu =\frac{\bar{O}-\bar{P}}{\sqrt{(S_o-S_p)}},\end{array}$$ (10)

in which v quantifies the scale shift in standard deviation between predicted and observed values and $\mu$ provides a scaled metric of bias.

Evaluation metrics were calculated using both conditioned predictions and raw predictions. Conditional predictions include the effect of both fixed and random effects while raw predictions include only fixed effects. A visual assessment of the raw and conditioned results of each model was also performed by plotting the predicted vs. observed and predicted vs. residual values using PROC SGPLOT in SAS.

RESULTS

Throughout, model parameters presented are on the beta scale (Table 3), while impact sizes were calculated via application of behavior analysis as described in the Materials and Methods section above and are thus reported as impacts (percent change / percent change *100) on the original data scale (Table 4).

Table 3.

Parameter estimates (± standard error) for models developed to quantify the effect of various factors on organ size as a percentage of body weight.

Model	Variable2	Category3	Parameter Estimates			P-value	LSM4
Heart (% BW)1	BMY% × poultry purpose2					0.0014
		Chicken Egg-type	0.07	±	0.024
		Chicken Meat-type	0.01	±	0.013
		Turkey Meat-type	0.06	±	0.026
	Sex					0.0005
		Female					0.61 ± 0.043
		Male					0.73 ± 0.046
	Live weight (g)		-1.40e−4	±	2.900e−5	<0.0001
Liver (% BW)	BMY% × poultry purpose2					0.0058
		Chicken Meat-type	0.01	±	0.005
		Turkey Meat-type	1.85e−3	±	0.009
	Live weight (g)		-5.00e−5	±	1.20e−5	0.0001
Lung (% BW)	Heart weight (g)		0.03	±	0.009	0.0063

¹(% BW), percent body weight.

²Interactions or isolated variables in each model.

³How animals are categorized by model to show estimates.

⁴LSM, Least Squared Means

Table 4.

Behaviour analysis¹ of univariable and multivariable models for heart weight (% BW), liver weight (%BW), and lung weight (% BW). Multivariable models are shown by comparing the results between sex and poultry purpose for heart weight (% BW) and poultry purpose for liver weight (% BW).

Univariable analysis
Variables	Heart weight (%BW)		Liver weight (% BW)				Lung weight (%BW)
Change in X Variable	+10%	-10%	+10%		-10%		+10%		-10%
Change in Y Variables:
Live weight	-1.0%	+1.0%	-1.2%		+1.2%		-		-
BMY (%)	-5.3%	+5.6%	-		-		-17.05%		+20.45%
Liver weight (% BW)	+6.3%	-6.0%	-		-		-6.2%		+6.5%
Heart weight (% BW)	-	-	+6.5%		-6.1%		+5.8%		-5.5%
BMY (g)	-	-	-1.2%		+1.2%		-		-
Heart weight (g)	-	-	-		-		+3.4%		-3.2%
Gizzard weight (% BW)	+4.1%	-4.0%	-		-		-		-
Bursa (% BW)	-	-	+2.4%		-2.4%		-		-
Wing weight (% BW)	-	-	-		-		+10.1%		-9.2%
Multivariable analysis
Categories2	Heart weight (%BW)			Liver weight (% BW)				Lung weight (%BW)
Change in X Variables	+10%	-10%		+10%		-10%		+10%		-10%
Change in Y Variables:
Chicken egg (female)	+3.3%	-3.6%		-		-		-		-
Chicken meat (female)	+2.9%	-3.3%		-		-		-		-
Turkey meat (female)	+2.3%	-2.8%		-		-		-		-
Chicken meat (male)	+0.7%	-0.7%		-		-		-		-
Turkey meat (male)	+2.5%	-2.7%		-		-		-		-
Chicken meat	-	-		+2.1		-2.1		-		-
Turkey meat	-	-		-2.2		+2.2		-		-

¹Behavior analysis, expected change (in %) in the predicted Y-variable (either univariable or multivariable), with a ±10% change in the driving X variables.

²Categories, how the multivariable model mentioned in the Materials and Methods section results in the levels of the poultry purpose variable.

Univariable analysis

Heart. The database variables associated with heart weight (% BW) in univariable models with the lowest error probability were poultry purpose (P = 0.001), live weight (g) (P < 0.0001) (unfavorable correlation), BMY (%) (P < 0.0001) (favorable correlation), and liver weight (% BW) (P < 0.0001) (favorable correlation) on the beta scale (Table S1). The average back transformed LSMean values for the poultry purpose factor were 0.0067, 0.0056, and 0.0035 for laying hens, broiler chickens, and turkeys, respectively. The behavior analysis of models developed indicates that a 10% increase in live weight (min: 349.0 g, max: 22,396.30 g) resulted in a 1.0% predicted decrease in heart weight (% BW), while a 10% increase in BMY (%) (min: 12.86%, max: 36.57%) resulted in a 5.3% predicted decrease in heart weight (% BW), and a 10% increase in liver weight (% BW) (min: 0.90%, max: 4.24%) resulted in a 6.3% predicted increase in heart weight (% BW).

Additionally, other variables significantly associated with heart weight at the univariable level included gizzard weight (% BW) (P<0.0001), where a 10% increase in gizzard weight (min: 0.70%, max: 4.03%) resulted in a 4.1% predicted increase in heart weight (% BW), and ventricular hypertrophy presence (Yes or No) (P=0.004) (Table 3).

Liver. The factors associated with liver weight (% BW) in univariable models with the smallest error probability were poultry purpose (P=0.001), live weight (P<0.0001) (unfavorable correlation), heart weight (% BW) (P<0.0001) (favorable correlation), and breast weight (BrW, g) (P<0.0001) (unfavorable correlation) (Table S1) on the beta scale. The average back transformed LSMean values for the poultry purpose factor were 0.02382, 0.02376, and 0.01378 for laying hens, broiler chickens, and turkeys, respectively. The behavior analysis of models developed indicate that a 10% increase in live weight (% BW) (min: 349.0 g, max: 22,396.30 g) resulted in a 1.2% predicted decrease in liver weight (% BW), while a 10% increase in heart weight (% BW) (min: 12.86%, max: 36.57%) resulted in a 6.5% predicted increase in liver weight (% BW), and a 10% increase in BrW (min: 95.2 g, max: 6,580.50 g) resulted in a 1.2% predicted decrease in liver weight (% BW).

Additionally, other variables significantly associated with liver weight (% BW) at the univariable level included bursa weight (% BW) (P=0.02) (Table 3), where a 10% increase in bursa weight (% BW) (min: 0.70%, max: 4.03%) resulted in a 2.4% predicted increase in liver weight (% BW).

Lung. The factors associated with the lungs weight (% BW) in univariable models with the lowest error probability were BrW (g) (P<0.0001) (unfavorable correlation), heart weight (g) (P =0.006) (favorable correlation), and liver weight (% BW) (P<0.0001) (favorable correlation) (Table S1) on the beta scale. The behavior analysis of models developed indicates that a 10% increase in BMY (min: 12.86%, max: 36.57%) resulted in a 17.0% predicted decrease in lung weight (% BW), while a 10% increase in liver weight (% BW) (min: 0.90%, max: 4.24%) resulted in a 6.1% predicted decrease in lung weight (% BW).

Other variables significantly associated with lung weight at the univariable level included wing weight (% BW) (P<0.0001) (Table 3), where a 10% increase in wing weight (% BW) (min: 0.70%, max: 4.03%) resulted in a 10.1% predicted increase in lung weight (% BW).

The models developed for lung weight (% BW) were limited to univariable models due to a lack of data for developing more complex multivariable models. The best univariable model for lung weight (% BW) included the fixed effect of heart weight (g) (P<0.006) (favorable correlation) (Table 3 and Table S2, respectively). The behavior analysis of models developed indicates that a 10% increase in heart weight (min: 0.30 g, max: 79.67 g) resulted in a 3.4% increase in lung weight (% BW).

Multivariable analysis

Heart. The best performing multivariable model for heart weight (% BW) included the interaction between poultry purpose and BMY (%) (P = 0.001), plus sex (P = 0.0005), and live weight (g) (P < 0.001). The parameters for this model are presented in Table 3.

The results of the behavior analysis conducted on this heart weight (% BW) model is presented in Table 4. For females, laying hens had the largest average heart weight (% BW) across species and their purposes. Within species, the heart weight (% BW) for broiler chickens and turkeys were more impacted by increases in BMY and live weight than laying hens (significantly higher regression coefficients).

The behavior analysis of this model indicates that a simultaneous 10% increase in live weight and BMY (%) for laying hens resulted in a 3.3% predicted increase in heart weight (% BW). Within female birds, turkeys had hearts that were more impacted (2.3% increase) by simultaneous 10% increases in BMY (%) and live weight than broiler chickens (2.9% increase). For males, turkey hearts (2.5% increase) were less impacted by simultaneous increases in BMY (%) and live weight than broiler chickens (0.7% increase).

Liver. The best performing model for liver weight (% BW) included the interaction between poultry purpose and BMY (%) (P = 0.011), plus live weight (P < 0.01). The parameters for this model are presented Table 3.

The behaviour analysis of the liver weight (% BW) model are presented in Table 4. Within meat-type birds, broiler chickens had livers that were less impacted (2.1% increase) by a simultaneous 10% increase in BMY (%) and live weight than turkeys (2.2% decrease). There were no laying hens included in the liver weight (% BW) model.

Evaluation

Model evaluation was based on pseudo-AICc, RMSPE, and CCC (Table 5), and residual plots for selected models are shown in Figure 2.

Table 5.

Evaluation of model equations for heart (% BW) model, liver (% BW) model, and lung (% BW) model.

Evaluation Parameters	Heart (% BW)1 Model	Liver (% BW) Model	Lung (% BW) Model
Conditional
n	91	98	43
AICc	-12.28	-725.17	5.17
Mean ± SE	0.54 ± 0.129	2.21 ± 0.566	0.48 ± 0.228
SD	0.12	0.50	0.27
RMSPE2	0.0005	0.0026	0.0004
RMSPE (%)3	9.57	11.69	9.05
ECT (%)4	2.86	7.21E−28	8.67E−17
ER (%)5	1.42	1.79	0.20
ED (%)6	98.58	98.21	99.80
CCC7	0.93	0.91	0.98
R8	0.94	0.92	0.98
Cb9	0.99	0.99	1.00
Raw
n	91	98	43
AICc	-12.28	-725.17	5.17
Mean ± SE	0.53 ± 0.066	2.20 ± 0.423	0.46 ± 0.055
SD	0.07	0.39	0.26
RMSPE2	0.0012	0.0049	0.0022
RMSPE (%)3	21.98	22.12	46.36
ECT (%)4	0.22	0.04	0.53
ER (%)5	2.85	0.04	1.52
ED (%)6	96.92	99.92	97.96
CCC7	0.45	0.60	0.16
R8	0.60	0.66	0.35
Cb9	0.76	0.91	0.45

¹(% BW), percent body weight.

²Root mean square prediction error.

³Root mean square prediction error expressed as a percentage.

⁴Error due to bias expressed as a percentage of MSPE.

⁵Error due to regression slope deviation expressed as a percentage of MSPE.

⁶Error due to disturbance expressed as a percentage of MSPE.

⁷Concordance correlation coefficient.

⁸Pearson correlation coefficient.

⁹Bias correction factor.

Figure 2.

Comparison of conditional predictions vs observed values and comparison of residuals vs conditioned predictions for the developed multivariable heart (% BW) model (2A, 2B), lung (% BW) model (2C, 2D), and liver (% BW) model (2E, 2F).

Heart. The best multivariable model for heart weight (% BW) had the smallest pseudo-AICc value, a RMSPE value of 9.57%, and a high CCC value (0.93) when the conditioned predictions were evaluated. This model also had reasonably good performance when raw predictions were evaluated, with a CCC value of 0.45 and a RMSPE value of 21.98%. The comparison of raw to conditioned predictions revealed that the random effect, study, accounted for a moderate proportion of model performance.

The conditioned predicted versus observed plot for heart weight (% BW) is provided in Figure 2A. Overall, the predictions and observations are closely aligned and the slope of the line of best fit produced matches the line of equality. The residual versus predicted plot for heart weight (Figure 2B) illustrates no patterns and a slope and intercept that is not significantly different from zero.

Liver. The best model for liver weight (% BW) had the smallest pseudo-AICc value, a low RMSPE value (11.69%), and a high CCC value (0.91), when the conditioned predictions were evaluated. This model also demonstrated good performance when raw predictions were evaluated, with a CCC value of 0.60 and a RMSPE value of 22.12%. Based on a comparison of the raw versus conditioned model evaluation metrics, the random effect of the study appears to have contributed moderately to the performance of the liver weight (% BW) model.

The conditioned predicted versus observed plot for liver weight (% BW) is provided in Figure 2C. Overall, the predictions and observations are closely aligned and the slope of the line of best fit produced matches the line of equality. The residual versus predicted plot for heart weight (Figure 2D) illustrates the desired outcome, no patterns and a slope not significantly different from zero.

Lung. The best model for lung weight (% BW) had the smallest pseudo-AICc value, a low RMSPE value (9.05%), and a very high CCC value (0.98) when the conditioned predictions were evaluated. However, when raw predicted values were evaluated, this resulted in CCC value of 0.16 and a RMSPE value of 46.36%. This result indicates a significant amount of variation was explained by the random effect of the study, as indicated by the substantial difference between the conditional and raw model evaluation results.

The conditioned predicted versus observed plot for lung weight (% BW) is provided in Figure 2E. Overall, the predictions and observations are closely aligned and the slope of the line of best fit produced matches the line of equality. The residual versus predicted plot for heart weight (Figure 2F) illustrates the desired outcome—a nearly flat line with a slope not significantly different from zero, effectively capturing the variation.

DISCUSSION

The objective of this study was to identify and quantify the relationship between heart, liver, lung weight, and metrics of meat yield in multiple poultry species. Understanding these relationships is crucial for improving the livability, the ability of a bird to survive and thrive throughout the production cycle, of multiple poultry species by utilizing the findings to optimize organ weight in relation to meat production. In general, this study found significant negative associations between BMY and organ size across literature data, however the strength and magnitude of these associations depends on several factors like the poultry purpose.

Heart

Poultry purpose. The quantified differences in heart size (% BW) between broiler chickens and laying hens are possibly reflecting differences in their cardiovascular and reproductive system demands. This difference could be attributed, in part, to the strong directional selection of broiler chickens and turkeys over decades for increased growth rates, feed efficiency, and meat yield (Closter et al., 2012), which may increasingly restrict the body cavity as birds grow. In contrast, turkeys seem to have a more balanced response to increases in BMY and live weight, suggesting that turkeys are less adversely affected by the intensive selection pressures associated with meat production, particularly in maintaining heart development relative to body weight gain. This resilience could be attributed to physiological differences between the two species, such as how turkeys distribute weight gain over a longer growth period. These findings highlight the importance of species-specific consideration in breeding programs, especially in relation to cardiovascular health.

Carcass Traits. The strong relationship between BMY (%) and heart weight (% BW) aligns with previous work which has found that heart mass and body mass are highly correlated, and that body weight is correlated with heart mass and function (Grubb, 1983). This may indicate that larger, heavier birds experience some restriction on heart growth. This has the potential to detrimentally impact heart function and pulmonary pressure in these larger, heavier birds leading to adverse health effects (Habibian et al., 2017). However, our findings suggest that poultry purpose and sex-based differences play a role in how this relationship manifests. In this study, female broiler chickens showed stronger unfavourable correlations between heart traits, BMY, and live weight when compared to males. These stronger associations in females suggest that their cardiac development may be more sensitive to selection for increased muscle mass and body size. This pattern could reflect underlying sex-based differences in physiological capacity, metabolic rate, or hormonal regulation of growth and tissue allocation (Malan et al., 2003; Havenstein et al., 2007; Namakparvar et al., 2014). However, it is important to note that these conclusions are based on observed correlations rather than causal relationships, and the sex imbalance in the dataset may have influenced the apparent magnitude of these associations. In contrast, female turkeys and male turkeys were similarly impacted by BMY and live weight. These findings could be attributed to selection pressure for carcass traits not being as high in turkeys as it is in chickens. Alternatively, these findings could suggest that turkeys are less adversely affected by the intensive selection pressures associated with meat production, particularly in maintaining heart development relative to body weight gain. There could be a need for targeted breeding strategies, based on sex, species, and purpose of the birds, that balance production efficiency with cardiovascular health to reduce potential welfare concerns.

Sex. A significant difference was found between the sexes of birds in relation to heart weight, where males generally had smaller hearts relative to their body weight compared to females (across species and purpose). Male lines of meat birds are often selected for higher meat yield and faster growth rates, although this distinction is not universal. In broiler chickens, where both sexes are utilized for meat production, the selection dynamics may need to become even more complex. In this dataset, males outnumbered females among broiler chickens and turkeys, as the females in the studies were divided between broiler chickens, turkeys, and laying hens. Tůmová et al. (2020), showed that organ weights (g) were similar during early growth stages but began to diverge due to sex as the birds matured. For example, the mature weight of male turkeys was 1.85 times greater than that of females but also have a mortality rate 2.5 times higher. Male turkeys also had larger organ weights overall, with average heart weights of 58.3 g compared to 50 g in females (Tůmová et al., 2020). Additional evidence from Tumová and Chodová (2018) indicated that, across dietary groups, males exhibited higher slaughter weights, heart weights, and liver weights than females at the same age, likely due to genetic selection for increased BMY in males. While statistically significant, it should be acknowledged that there were relatively few females in the dataset (8.16%), so the effect of sex in this analysis should be interpreted with caution.

Liver

The factors included in the final model for liver weight were the interaction of poultry purpose and BMY as well as live weight.

Poultry purpose. Based on the models developed, the physiological adaptation of liver weight in birds is influenced by both the species and intended purpose of the bird. Liver size plays a critical role in supporting metabolic demands, particularly in fast-growing commercial poultry strains (Dou et al., 2017). One study found that at hatching, a low performing commercial strain had the largest liver to body weight ratio, but found from weeks 6 to 18, the high performing commercial strain showed the highest liver proportion (Lieboldt et al., 2016). Chickens bred for meat production appear to have livers better suited to support their rapid growth and larger body size, with an increase in liver weight observed as body weight increased. In contrast, turkeys exhibit much smaller livers relative to their body size, with a decrease in liver weight relative to body weight increases. This could indicate turkeys and chickens differ in their ability to meet the metabolic demands of their body, which could be due to variations in breeding objectives and physiological differences such as size and growth rates. This variation could result in a disproportionate development of internal systems, including the liver, potentially limiting their ability to meet the metabolic demands of their larger body mass. These findings emphasize the importance of understanding how species and breeding objectives influence not only muscle growth but also the internal organ systems required to support these traits.

Carcass traits.BMY and live weight were significantly related to liver weight, emphasizing the physiological link between bodily growth and organ development. Our analysis revealed distinct rates of change differences in liver weight among bird groups, underscoring the undesirable allometric relationship between muscle development and organ size. This link was observed in previous work conducted by Tickle et al. (2014) showing a link between body mass development and a reduction in organ development in Ross broiler chickens, which could indicate a trade-off between carcass growth, like BMY, and internal organs, such as the liver. The liver weight (% BW) model showed that turkey livers were more impacted by increases in BMY and live weight in comparison to broiler chicken livers. These findings highlight how breeding practices focused on meat production can have broader physiological effects, impacting not just muscle growth but also the internal organ sizes critical to supporting these traits. The smaller liver size observed in turkeys relative to their body mass could impact metabolic health, which highlights the need for organ traits to be considered in breeding practices.

Lungs

The scarcity of studies having multiple shared factors surrounding lung weight and function lead to a lack of lung data available in this study. A more accurate model could be developed as more data surrounding lung weight or function becomes available. The large difference between conditional and raw CCC values indicate that the studies with lung data vary considerably between studies. This variation could be due to different genetic strains, breeds of birds, diets, or a number of environmental factors such as location, temperature, and type of housing. Lung capacity being impacted by increases in heart weight has the potential to reduce oxygen delivery and carbon dioxide removal from tissues, as well as cardiovascular performance by increasing pulmonary pressure (Buys et al., 1999; Closter et al., 2012). However, it should be noted that it cannot be determined how lung capacity truly affects cardiovascular performance in this population. These findings indicate the importance of gaining a better understanding for what impacts the lungs of poultry.

Data limitations and opportunities

In order to compare chickens and turkeys, all organ weights were expressed as a percent of body weight. Analyzing weight as a percent of BW does cause some challenges in interpretation, as you cannot identify if it is the numerator and/or the denominator changing. For example, heart size could remain consistent despite increased body mass, leading to broiler chickens and laying hens having the same heart weight (g) but different heart weight (% BW). However, expressing organ weights in this way allowed examination of organ weight relative to bird size, which helped to avoid limiting data by modeling the species separately.

Cardiovascular incidence was included as a search term for this systematic review since a biological relationship was expected between disease incidence and organ size; the dataset therefore included both diseased and healthy animals. Although cardiovascular issue incidence was nonsignificant in every model, the use of diseased and healthy animals gave a dataset with potentially more extreme values. This may be more informative than when examined on all healthy birds, aiding in the development of a more robust model. However, it is important to note that disease may also impact organ weights or liveweight of the birds observed.

There were some additional issues with data balance between categories. For instance, the data was heavily skewed toward males, making the female group potentially unrepresentative in comparison. Parameter estimates derived from the female data may be less reliable due to the limited sample size. Addressing this issue through weighted analyses, data augmentation techniques (synthetic or simulated data), or targeted data collection strategies could help improve the imbalance.

Several significant univariate models contained variables that were excluded from consideration in multivariable models due to limited observations or high multicollinearity with other variables, such as strain, Pectoralis major percent, dietary components like crude protein and fat content, and age at slaughter and body weight. Although these variables are individually important, their exclusion likely results in some additional variation between and within studies left unexplained.

Interestingly, but perhaps as expected, on a raw basis all the models had a lower CCC value and a higher RMSPE value. The magnitude of this discrepancy, however, indicates that there remains considerable between study variation that is not captured by the fixed effects included in the model. This indicates that more data may be required to develop more accurate predictive models for use in the industry.

Poultry have been genetically selected for rapid muscle growth, but this emphasis on size may have contributed to more stress on their cardiovascular and respiratory systems, particularly due to the intense weight gain associated with rapid growth and limits placed on internal cavity size. The extent of impact varies depending on the bird's breeding purpose (egg or meat production), with broiler chickens and turkeys experiencing more intense impacts of body size changes than laying hens. Understanding these physiological relationships and what contributes to them could not only enhance the overall health and longevity of poultry but also improve meat production sustainability.

Disclosures

The authors declare no conflicts of interest.