Growth curve of broiler chicken submitted the cassava meal diet

Curva de crecimiento de pollos de ceba alimentados con dieta de harina de yuca

L. R. R. Lucena leandroricardo_est@yahoo.com.br

Universidade Federal Rural de Pernambuco, Brazil

M. A. C. Holanda

Universidade Federal Rural de Pernambuco, Brazil

M. C. R. Holanda

Universidade Federal Rural de Pernambuco, Brazil

The research was in the aviary of Fazenda São João, located in the district of Santa Rita, municipality of Serra Talhada-PE, in the micro region of the Sertão do Pajeú, mesoregion of the Sertão de Pernambuco, under license number 127/2019 of the ethics committee on the use of animals of the Federal Rural University of Pernambuco.

Were used 450 male broiler chickens of the Cobb lineage, with one day life, starting weight of 42 grams, vaccinated on the first day still in the hatchery, against Mareck, Newcastle, Gumboro and revaccinated at 14 days against Newcastle and Gumboro.

The birds were housed in an aviary built in masonry, with ceramic tiles and concrete floors, lined with bed of inert material (rice husk) at a height of 15 cm, keypad with galvanized wire screen and curtain to prevent drafts and control the environment temperature.

During the first 14 days of life, a 150 watt incandescent lamp was used with heat source for broiler chickens. Aviary was divided in 25 experimental plots, each measuring 2 m², with a density of 9 birds/m².

Experimental design was completely randomized with five treatments and five replications, where each experimental unit was composed of 18 birds.The treatments consisted of a control diet based on corn and soybean meal, and four test diets containing 25, 50, 75 and 100 % inclusion of integral meal of cassava roots supplemented with endogenous enzymes, in the quantity of 500 grams per ton of feed.

Cassava roots were acquired in the municipality of Araripina-PE, posteriorly the roots were processed and dehydrated in the sun for five days until they lost maximum moisture to obtain dry meal. A sample was collected and taken to the laboratory for chemical analysis that presented the following results:88.56% dry matter, 2.54% crude protein, 0.62% lipids, 5.32% crude fiber, 10.84% neutral detergent fiber (NDF), 3.96% acid detergent fiber (ADF), 84.92 % organic matter, 3.52 % ash, 0.18 % calcium and 0.09% phosphorus. The gross energy of 4,123 kcal/kg was determined in the IKA 200 calorimeter.

The result of the chemical composition was used to formulate the experimental diets together with the metabolizable energy of 12,502 MJ/kg (determined in a metabolism experiment carried out previously with chicks, this experiment was carried out before formulating the diets).The multi-enzyme complex was composed of galactosidase 35 U/g, galactomannanase 110 U/g, xylanase 1,500 U/g, β-glucanase 1,100 U/g, and was mixed to the premix in a Y-type mixer for mixing low level ingredients in the diets and used in the proportion of 500 grams per ton of feed for the test diets, for greater availability of nutrients contained in whole cassava meal.

From the first day of life the birds received experimental diets according to the treatments, following the nutritional recommendations of Rostagno et al. (2017) (table1, 2, 3 and 4).

Table 1
Chemical composition and calculated of the experimental diets for broiler chickens from 1 to 7 days of age as a function of the levels cassava meal

¹Premix vitamin/kg: Folic Acid 106.00 mg; Pantothenic 2,490 mg; Antifungal 5,000 mg; Antioxidant 200 mg; Biotin 21mg; Coccidiostatic 15,000 mg; Choline 118,750 mg; Vitamin K3 525.20 mg; niacin 7,840 mg; Pyridoxine 210 mg; Riboflavine 1,660 mg; Thiamine 360 mg; Vitamin A 2,090,000 UI; Vitamin B12 123,750 mcg; Vitamin D3 525,000 UI; Vitamin E 4,175 mg. Cu 2,000 mg; I 190 mg; Mn 18,750 mg; Se 75 mg; Zn 12,500 mg.

Table 2
Chemical composition and calculated of the experimental diets for broiler chickens from 8 to 21 days of age as a function of the levels cassava meal

¹Premix vitamin/kg: Folic Acid 106.00 mg; Pantothenic 2,490 mg; Antifungal 5,000 mg; Antioxidant 200 mg; Biotin 21mg; Coccidiostatic 15,000 mg; Choline 118,750 mg; Vitamin K3 525.20 mg; niacin 7,840 mg; Pyridoxine 210 mg; Riboflavine 1,660 mg; Thiamine 360 mg; Vitamin A 2,090,000 UI; Vitamin B12 123,750 mcg; Vitamin D3 525,000 UI; Vitamin E 4,175 mg. Cu 2,000 mg; I 190 mg; Mn 18,750 mg; Se 75 mg; Zn 12,500 mg.

Table 3
Chemical composition and calculated of the experimental diets for broiler chickens from 22 to 35 days of age as a function of the levels cassava meal

¹Premix vitamin/kg: Folic Acid 106.00 mg; Pantothenic 2,490 mg; Antifungal 5,000 mg; Antioxidant 200 mg; Biotin 21mg; Coccidiostatic 15,000 mg; Choline 118,750 mg; Vitamin K3 525.20 mg; niacin 7,840 mg; Pyridoxine 210 mg; Riboflavine 1,660 mg; Thiamine 360 mg; Vitamin A 2,090,000 UI; Vitamin B12 123,750 mcg; Vitamin D3 525,000 UI; Vitamin E 4,175 mg. Cu 2,000 mg; I 190 mg; Mn 18,750 mg; Se 75 mg; Zn 12,500 mg.

Table 4
Chemical composition and calculated of the experimental diets for broiler chickens from 36 to 42 days of age as a function of the levels cassava meal

¹Premix vitamin/kg: Folic Acid 106.00 mg; Pantothenic 2,490 mg; Antifungal 5,000 mg; Antioxidant 200 mg; Biotin 21mg; Coccidiostatic 15,000 mg; Choline 118,750 mg; Vitamin K3 525.20 mg; niacin 7,840 mg; Pyridoxine 210 mg; Riboflavine 1,660 mg; Thiamine 360 mg; Vitamin A 2,090,000 UI; Vitamin B12 123,750 mcg; Vitamin D3 525,000 UI; Vitamin E 4,175 mg. Cu 2,000 mg; I 190 mg; Mn 18,750 mg; Se 75 mg; Zn 12,500 mg.

To evaluate the performance of broiler chickens weight according to age and inclusion of cassava meal, regression model adjustments were proposed: exponential, Weibull, logistic, Gompertz, power, hyperbolic tangent, and gamma (table 5).

Table 5
Regression models evaluated

where, Yiis the observed weight of the i-th broiler chickens after birth; Tiis the i-th evaluation day; Mandiis the percentage of cassava added to the diet of the i-th broiler chickens after birth and εiis the i-th error associated with weight, where presents exponential parameter distribution α to exponential model, Weibull distribution of parameters α and γ, normal distribution of mean 0 and constant variance σ² to logistic, Gompertz, power and hyperbolic tangente and Gamma distribution of parameters α and β. The metrics ω, β0, β1 and β2are the parameters associated with the model.

The following criteria evaluated the models: Coefficient of Determination of the Model (R²), Akaike's Information Criterion (AIC) and Sum of Square of Residuals (SSR).

Let ${\widehat{Y}}_i$ the values of the i-th broiler chickens weight after model adjustment and $\overline{Y}$ mean broiler chickens weight, define SSR for this study by the following expression:

The coefficient of model determination is expressed by:

The Akaike information criteria (AIC), as defined by Akaike (1974), are given by:

where, L(x\$\widehat{\theta })$ is the maximum likelihood function, defined as the production of density function and p is the number of model parameters.

Cluster analysis using the Ward method was used to verify which models are most similar to their adequacy criteria. Posteriorly, residue analysis was performed to validate the quality of the model that best adjusted to the weight growth of the broiler chickens according to age and the different levels of inclusion of cassava meal in their diet.Validation of the model was performed through studentized residues, analysis of leverage and influential points and quantile-quantile plot of distribution normal.

Let hat matrix (H),

and,

where, $h_{ii}$ are the diagonal elements of matrix H. Assume that any observation that exceeds twice the average ($h_{\mathrm{ii}}\mathrm{>2p/n}$) is remote enough from the rest of the data to be considered a leverage point.

Studentized resisuals defined by:

where, $e_i$ is the residue of the i-th observation of the model (difference between the observed and adjusted weight).

To detect a point of influence we use Cook’s distance, defined by:

if $D_i\mathrm{>2}\sqrt{\frac{p}{n}}$ , denoted influential point.

The R-project version 2.13.1 for windows software was used to perform the analyzes.

Mean weight of the birds in relation to the lifetime and the different diets with cassava meal are shown in table 6. For all evaluation periods, verified that there was not difference (p-value> 0.05) in the broiler chickens weight in relation to the different levels of cassava meal in diet (table 6).

Table 6
Broiler chickens weight according to lifetime and inclusion of cassava meal in diet

The results of this study corroborate with findings of Sousa et al. (2012) that verified a difference in the weight gain of broiler chickens fed up to 20% of cassava meal in the initial phase (1-21 days), while in the final phase (22-40 days) there was not difference in the weight gain. Carrijo et al. (2010), Souza et al. (2011) and Holanda et al. (2015), found no difference in the weight gain of free-range broiler chickens fed different levels of cassava meal.

Table 7 shows that the models exponential, Weibull, logistic and Gompertz presented explanatory power of less than 0.90, in addition to presenting the largest sums of squares of the residues, indicating a poor adequacy of these models to explain the broiler chickens weight as a function of age and percentage of cassava meal introduced in their diet.

Table 7
Adjusted regression models and model adequacy criteria to growth broiler chickens weight fed with levels of cassava meal in the diet

R²- model determination coefficient; SSR-sum of squares of residues; AIC- Akaike information criterion; ${\widehat{Y}}_i$ is the adjusted weight of model of the i-th broiler chickens after birth; T is the lifetime; Mand is the percentage of cassava

Table 8 shows the estimates of the parameters of the models with their respective standard errors, test statistics and p-value, showing the significance of each parameter.

Table 8
Estimative, standard error, t value and p-value of parameters models

Lucena et al. (2017) verified that the exponential, Weibull and Gompertz models presented explanatory power of 0.993, 0.916 and 0.948, respectively. Rizzi et al. (2013) observed that the Gompertz model was the most adequate to explain the growth of broiler chickens with explanatory power greater than 99%, these divergent results of this research, what can be explained by the introduction of increasing levels of cassava in the diet of the broiler chickens causing a loss of yield of these models, as these authors only evaluated weight growth as a function of the birds lifetime.

The hyperbolic tangent model presented explanatory power of 0.975 and sums of residual squares of 0.90.These criteria classify these models with good precision in estimating of the broiler chickens weight, however, these results are inferior to those presented by the power and gamma models, (table 7). Michalczuk et al. (2016), Liu et al. (2015), Zhao et al. (2015), Selvaggi et al. (2015) and Mohammed (2015) presented similar results for the logistic model, while the results for the hyperbolic tangent model corroborate with the describes by Lucena et al. (2017), that is, for all the researches reported, the weight behavior of the animals is similar when using these models.

Power and gamma models showed the highest model determination coefficients, lowest sums of squares of the residues and lowest Akaike information criteria, (table 7).These criteria indicate that these models are the most efficient to estimate the broiler chickens weight as a function of lifetime and introduction of cassava meal. Similar results were reported by Lucena et al. (2017) where they verified that the power model was the most adequate to explain the broiler chickens weight with precision of 0.997 followed by the gamma model with an explanatory power of 0.989.

Due to the different selection goals applied by geneticists in the last decades, growth parameters of broiler genotypes can differ in several characteristics, including those that affect the potential growth curves, with weight and maturation rates (Sakomura et al. 2011).

The differences between the functions in the growth rate directly reflect on the behavior in the data adjustment. Nonlinear functions have been used extensively to represent changes in broiler chickens weight as a function of age, so that the genetic potential of animals can be valued (Kuhi et al. 2019).

Early estimation of weight at maturity and growth rate in relation to body size can be important for selection purposes, given its association with other characteristics and the economy of production (Kuhi et al. 2019). The exploration of these parameters in growth models by adjusting curves using age with live weight can positively improve economic returns (Salako 2014).

Success in studying the growth characteristics of broiler chickens will help to define more adequate diets to cover high nutritional requirements during the growth phase, from hatching to age at the point of slaughter.In addition, selecting the best function based on your ability to describe the relationship between live weight and age is the first step in developing a genetical improvement program (Selvaggi et al. 2015). Growth curve parameters provide an opportunity to plan selection strategies, modifying dietary practices or genetic makeup of the shape of the growth curve (Selvaggi et al. 2015).

Figure 1 shows that the power model presented better estimates of broiler chickens weights than the Gamma model, because the power model showed only a discrepant value from the observed weight of the chickens that occurred on the 28th day, while the Gamma model presented two weight discrepant occurrences (28th and 42nd day).

Figure 1
Estimates of the broiler chickens weight in the power (a) and gamma (b) models

Through Ward cluster method using the metrics of model adequacy criteria, verified the formation of two groups of modelswhen using a cutting height greater 60, a group formed by the power and gamma models (models that presented higher R² and lower SSR and AIC), and the second formed by the others models (models that did not present criteria similar to the gamma and power models) (figure 2).

Figure 2
Cluster of adjusted regression models to growth broiler chickens weight fed with levels of cassava meal in the diet

Evaluating the three criteria of adequacy of the model, the cluster analysis and the estimates of the broiler chickens weights, the power model was proposed with most adequate to explain the growth of broiler chickens as a function of the lifetime and the different percentages of cassava in their diet.

After defining the power model with most appropriate, the analysis of the residues was performed (figure 3).No discrepant residues were diagnosed (figure 3a),because none is outside the limits of [-2; 2],also no residual leverage or influence was detected (figure 3b and 3c) because no point exceeded the criteria defined by the dotted lines, the assumption of normality of the residues was diagnosed in the quantile-quantile graph of the normal distribution, where the residues are within the confidence bands (figure 3d).

Figure 3
Analysis of residues of the power model in broilers that consume cassava meal

Cassava meal in the dietary supplementation of broiler chickens, in addition to promoting better zootechnical performance, decreases production costs, because for diets without inclusion of the cassava meal the production cost was higher because more corn was used ($0.27 per kg of feed for 0%; $0.26 per kg of feed for 25%; $0.24 per kg of feed for 50%; $0.23 per kg of feed for 75%; $0.21 per kg of feed for 100%),while the cost using 100% inclusion of cassava meal was lower because it used half quantity of corn for the control diet.

In many practical problems, such as parameter estimation, function values are uncertain or subject to variation. Therefore, a highly accurate solution is not necessary. In these situations, all you want is an improvement in the adjustement of the function, what can be observed in the use of the power model.

Weight growth of birds fed cassava meal can be estimated using the power regression model. The use of the power model provides information on the best level of inclusion of cassava meal (100%) and the best time for slaughtering birds (42 days) maximizing the weight in 3,295 g.