This research received approval from the Kasetsart University Institutional Animal Care and Use Committee (Approval No. ACKU67-CSC-005) and the National Research Council of Thailand Ethics Review Committee (NRCT Permit No. U1-10350-2565). Furthermore, the study adhered to established animal welfare protocols. Three cohorts of 100 birds (each cohort consisting of 50 males and 50 females) were utilized, namely, KU-Phuphan (KP), Pradu Hang Dam (PHD), and commercial broilers (Ross 308).

The chickens were raised in an open-house system at the Kasetsart University Sakon Nakhon Campus farm, Thailand, in accordance with Good Agricultural Practices. Throughout the study, each chicken was tagged with a wing clip (National Band & Tag Company; KY, USA) to allow for identification.

Chickens were housed in open cages measuring 1.4 meters wide, 1.5 meters long, and 2.5 meters high. The building was constructed of concrete blocks 0.5 meters high and covered with floor-to-ceiling wire mesh to promote ventilation and prevent disease from pathogens. Rice husks served as a bedding material to protect against heat and to absorb moisture on the floor. Separate plastic containers (Mitr Kasetphan Co., Ltd., Thailand) were used to store water and food for the chickens. Each container held enough food for 10 chickens.

The density of the broiler chickens was maintained at 0.1 square meters per bird, in accordance with the(National Bureau of Agricultural Commodity and Food Standards, 2017) requirements. Rice husks were used as a litter material to provide warmth for the chicks. A 100 W light bulb supplied heat during the initial 6 weeks post-hatching, which helped to promote growth and to decrease mortality.

Nutritional management followed a two-phase feeding program: during 0–4 weeks, chickens received a commercially manufactured pelleted diet, containing minimum values of 21% crude protein (CP) and 3,200 kcal/kg metabolizable energy (ME); during 5–12 weeks, chickens received a diet with a minimum of 18% CP and 3,200 kcal/kg ME. All diets met the nutritional requirements for growing chickens(Council, 1994). Feed and water were replenished as needed. Vaccination protocols were implemented by farm veterinarians following standard Thai commercial broiler farm procedures. Management followed the National Bureau of Agricultural Commodity and Food Standards (2017)(National Bureau of Agricultural Commodity and Food Standards, 2017).

The sample size was calculated utilizing the pwr package(Champely, 2020) in the R statistical program(Team, 2024) based on the parameters: 1) three treatment groups: commercial broiler, PHD, and KP chickens; 2) effect size of 0.6; 3) significance level tested at 0.05; and (4) power level of 0.8. The minimum required sample size was 9.99 animals per group, rounded to 10 for practicality.

Body weight was obtained weekly during 0–12 weeks using a precision digital scale (Tanita KD-200; Tanita Corp.; Japan). The average daily gain was computed for the intervals 0–2, 0–4, 0–6, 0–8, 0–10, and 0–12 weeks. Outliers were identified via the Mahalanobis distance based on chi-square criteria.

Data on body weight during 0–12 weeks and mean daily increase values were examined for the intervals 0–2, 0–4, 0–6, 0–8, 0–10, and 0–12 weeks. Statistical analysis was conducted on growth parameters derived from the optimal model for each breed. Analysis was performed using the JASP software version 0.19.3(Team, 2024). Two-way analysis of variance was applied with the main effects being chicken breed (commercial broiler, KP, PHD) and sex (male, female), including breed × sex interaction effects. Mean comparisons were performed using Tukey’s honest significant difference test at the 95% confidence level, with significance tested at p<0.05.

The statistical model used for data analysis was:

y_ij= μ + B_i + S_j + (B×S)_ij + e_ij

where y_ij was the observation of the i^th breed and j^th sex, μ was the overall mean, B was the fixed effect of the i^th breed (i was broiler, KU-Phuphan, and Pradu Hang Dam), S_j was fixed effect of the j^th sex (j was male, female), (B×S)_ij was interaction effect between i^th breed and j^th sex, and e_ij was random error term.

Individual growth curve parameter estimates were derived for all chickens using five nonlinear models (Gompertz, Logistic, von Bertalanffy, Hossfeld, Weibull) with the equations shown in Table 1. Each model was fitted to body weight data from day 1 through to week 12 of age. Growth curve parameters for each model were estimated using the Levenberg-Marquardt algorithm via the nlsLM function of the minpack.lm package(Elzhov et al., 2023) in the R software environment(Team, 2024).

Each model’s performance and goodness-of-fit were evaluated using leave-one-out cross-validation (LOOCV), which is suitable for small sample sizes(James et al., 2013). Leave-one-out cross-validation (LOOCV) involves applying N-1 samples for the training set and a single sample for the testing set, iterating until each observation has been used once as the test set(James et al., 2013).

Prediction errors were calculated in each iteration to determine the metrics of performance, consisting of: coefficient of determination (R²); root mean square error (RMSE); mean absolute error (MAE); Akaike Information Criterion (AIC); and Bayesian Information Criterion (BIC), as presented in Table 2. The model with the highest R² and the lowest RMSE, MAE, AIC, and BIC values was considered the most accurate for predicting chicken body weight during 0–12 weeks.

The three chicken breeds demonstrated different growth trajectories (Table 3). The commercial broiler chickens had a rapid growth rate of 6 weeks, significantly faster than for the domestic breeds. Growth increased rapidly until week 4 and thereafter decreased as it reached maturity. Conversely, the KP and PHD chickens showed slower but consistent growth throughout he 12 weeks, eventually achieving similar final body weights, despite varying growth patterns. Throughout the study, the broilers had consistently higher body weights, especially from hatching to 6 weeks, averaging over 1,600 g, in contrast to the 588 g for the KP and 456 g for the PHD samples (Figure 1), highlighting the substantial growth advantage of commercial strains against indigenous breeds.

There were no significant sex differences in body weight in commercial broiler chickens during the 6 weeks. Conversely, the KP chicken demonstrated notable male-female disparities from 6 weeks of age (p<0.05), with the males reaching 1,465 g at 12 weeks, while the females reached 1,153 g, reflecting a 27% advantage for males. The PHD chickens displayed a consistent pattern, with males significantly heavier than females from age 2 weeks (p<0.05); at 12 weeks, the males averaged 1,475 g, while the females averaged 1,126 g, showing a 31% advantage. Across all breeds, sexual dimorphism was negligible during early development (0–4 weeks) but increased significantly between 6 and 12 weeks (Table 3).

There were significant breed effects in growth performance. The commercial broilers reached market weight in 5 weeks, whereas the KP and PHD chickens required the complete 12-week growth period to reach similar body weights. In addition, there were significant differences in the body weight among breeds (p<0.05) from hatching to 6 weeks, with the broilers consistently being the heaviest, followed by KP and PHD. No significant changes were observed between the KP and PHD chickens beyond 8 weeks (Table 3). The interaction effects of breed and sex were significant at 2 and 4 weeks, but not at subsequent ages.

This study found that breed and sex significantly affected the ADG ( Table 3). The commercial broilers showed rapid growth during the first 6 weeks, consistent with the intensive selection program that had aimed at achieving early weight gain and increasing efficiency in commercial production. Conversely, the KP and PHD chickens exhibited a continuous and consistent increase in ADG, suggesting different metabolic mechanisms suited to resource-limited environments.

Sexual dimorphism in ADG became increasingly evident with age in the indigenous breeds, as males consistently outperformed females from mid-development forward (p<0.05), suggesting hormone-driven growth responses approaching sexual maturity. Conversely, the commercial broilers displayed no significant sexual dimorphism during their growth period, indicating that the emphasis on rapid early growth may have diminished sex-specific differentiation. The eventual convergence of ADG for the KP and PHD samples in later phases indicated similar genetic potential, despite divergent initial trajectories, potentially signifying compensatory expansion in the slower-starting PHD line.

A comparative analysis of the five growth models revealed significant correlations of specific functions between breeds (Table 4). The Gompertz model was the most appropriate for broiler chickens, providing the highest predictive performance (R² = 0.999, AIC= 68.13, BIC = 67.92), followed by the Weibull model. In contrast for the two indigenous chicken breeds, the Hossfeld model provided the most appropriate

predictive performance (KP chickens: R² = 0.999, AIC= 115.15, BIC = 117.41; and PHD chickens: R² = 0.999, AIC = 115.73, BIC = 117.99), while the logistic model provided the poorest performance modeling across all breeds, indicating that this model still had limitations in its application to these chicken populations.

Investigation of breed, sex, and their interaction effects revealed significant differences across several growth curve variables (Table 5). The age at the inflection point (TI) demonstrated highly significant breed effects (p<0.001), with the PHD chickens having the highest values; however, sex effects were mostly not significant. On the other hand, weight at the inflection point (WI) and maximum increment (MI) were significantly different at high levels for both breed (p<0.001) and sex (p<0.01), with the males being generally heavier and having a greater maximum growth rate than the females.

The main growth curve characteristics indicated that there were significant variations in asymptotic weight (A) depending on breed and sex, with the broilers having the highest values (Table 5). In addition, the maturation rate (k) differed among breeds, with the broilers maturing the fastest. Sexual dimorphism was consistent across breeds, with the males attained higher asymptotic weights and maximum increments, whereas females reached their inflection points earlier.

The current study has provided important information to guide breeding strategies and production management in Thai native chickens in tropical environments. The growth pattern analysis supports body weight prediction at different ages and guides the development of breed- and sex-specific feeding programs for native chickens. The observation that model fit varies among breeds indicates the need for breed-specific approaches. The Hossfeld model can be used as a tool for predicting growth patterns and planning production in Thai native chickens. Selective breeding of native chickens is essential for increasing production efficiency while maintaining their unique characteristics, which helps preserve biodiversity and ensure nutritious foods.