Introduction

Coccidiosis is a disease of major economic importance affecting domesticated avian species (Williams 2005). The disease affects birds of all ages. It poses serious public health and economic challenges to both commercial and smallholder poultry farmers globally on a continuous basis (Azeezah et al. 2012). The short direct life cycle and high reproductive potential of coccidians in poultry often leads to severe outbreaks of disease in small backyard flocks or modern poultry house (McDougald and Fitz-Coy 2008). In the tropics, the problem of diseases such as coccidiosis is mostly combated with the use of drugs which invariably add more to the cost of poultry production (Azeezah et al. 2012). Hence, alternatives approach is being sought worldwide as addition to the use of drugs and vaccine. Breeding for resistance to the disease will be one of the possible approaches as addition to the use of drugs and vaccine (Adeleke et al. 2015). Breeding involves identification of line, strain, or breed of animal that has potential to resist coccidiosis. Among potential strains of chickens for this genetic selection are Nigerian indigenous chickens. The unique adaptive features of the Nigerian indigenous chickens predisposing it to adapt to the local environment have been reported by several authors (Adebambo et al. 1999; Ikeobi et al. 2001; Peters et al. 2011; Ajayi et al. 2012). These features include disease resistance (Egena et al. 2014), hardiness, and ease of rearing (Ige 2013).

In evaluating coccidiosis resistance characteristics, several variables have been used as coccidiosis resistance variables. These variables include post-inoculation body weight gain, fecal oocyst shedding, and plasma levels of carotenoid, nitrite plus nitrate (Kim et al. 2006; Hong et al. 2009; Kim et al. 2010), and hematological and biochemical variables (Meskerem et al. 2013). Hematological variables have been reported to provide valuable information on the immune status of chickens (Ladokun et al. 2008). This information is useful for diagnostic management purposes as well as breeding programs for the genetic improvement of indigenous chicken. However, evaluation of coccidiosis resistance variables between healthy and infected chickens has been restricted to the use of univariate analysis only. Authors considered the variables individually using univariate analysis. Univariate analysis generally indicates testing for group differences on each of the coccidiosis variables without taking into account its relationships to the other coccidiosis variables. Therefore, in interpreting univariate analysis, caution should be exercised because univariate F tests do not account for correlations among the coccidiosis parameters or any potential increase in Type 1 error (probability of incorrectly rejecting null hypothesis) that results from several univariate analysis being carried out on all the variables (Tenko and George 2008).

Meanwhile, the use of multivariate (discriminant) analysis is considered to be more appropriate. This is due to the joint consideration of all measured coccidiosis parameters at once. Discriminant analysis is where two or more groups are known and one or more new observations are classified into one of the known groups based on the measured characteristics (Asamoah-Boaheng and Sam 2016).

The aims of this study were the following: (1) to use discriminant analysis to study the variability in Eimeria-infected and uninfected (normal) chickens, (2) to develop a model which could group the chickens in one of these two groups using only a few of the hematological variables. This is novel approach as to the best of our knowledge; no systematic effort has been reported that attempts to use body weight gain, hematology, and comprehensive statistical tools to find any diagnostic or predictive biomarkers for coccidiosis disease in chickens raised in hot humid tropical environment.

Materials and methods

Study site and experimental materials

The experiment was carried out at the Poultry Breeding Unit of the Directorate of Farm, Federal University of Agriculture, Abeokuta, Ogun State, Nigeria. A total of 230 chicks of Nigerian indigenous chickens were obtained through artificial insemination from parent stock kept at the Poultry Breeding Unit. Abeokuta is located within the rainforest zone of Southwestern Nigeria with latitude 7° 13′, 49° 46′ N, longitude 3° 26′, 11° 98′ E and altitude 76 mm above sea. The annual mean temperature and humidity were 34 °C and 82% respectively (Amujoyegbe et al. 2008). The indigenous chickens used in this study were generated from several years of selection of Nigerian local chickens in the Poultry Breeding Unit of the institution. The selection of the chickens started in 1995 when local chickens were sourced from local farmers in Southwest Nigeria. The chickens used in this study were made up of indigenous chickens with the feather distribution gene (Naked Neck), feather structure gene (Frizzle Feather), and normal feathered. More information about this population of chickens has been provided by Peters et al. (2011).

The chicks were brooded for 3 weeks on a commercial chick mash. Feed and water were supplied ad libitum from day-old on a ventilated deep litter system using wood shavings as bedding materials. At 3 weeks of age, the chicks were divided into two groups. In the first group, chicks were inoculated with Eimeria tenella (E. tenella) which was obtained from National Veterinary Research Institute, Vom, in Plateau State, Nigeria, through oral inoculation at the rate of 1 × 105 doses per chick. All inoculated chicks were raised in battery cages to avoid physical contact with their feces. In the second group, chicks were not inoculated with E. tenella and served as control group. Blood samples were collected from the wing web using a 2-ml sterile syringe and needle from each bird from the two groups. Blood samples were collected from both groups 2 weeks after inoculation of the first group into EDTA tube. The total red blood cells were assessed in a 1:200 dilution of blood in Hayem’s solution. The differential leukocyte counts were determined by staining blood films with Wright’s stain. Packed cell volume was measured using the microhematocrit method. Body weight was taken in the morning before the birds were fed and it was done using a weighing balance scale with sensitivity of 0.01 g. The post-inoculation body weight gain for days 3, 6, 9, 12, and 15 was estimated as \( \frac{\mathrm{final}\kern0.5em \mathrm{body}\kern0.5em \mathrm{weight}\hbox{-} \mathrm{initial}\kern0.5em \mathrm{body}\kern0.5em \mathrm{weight}}{\mathrm{number}\kern0.5em \mathrm{of}\kern0.5em \mathrm{days}} \).

Data analysis

Incomplete data due to blood clotting from chicks were excluded from the analysis. Preliminary analysis was carried out where homogeneity was tested. Because of deviation from normality of measurements taken, all data were log10 transformed before analysis. Effects of health status (uninfected or Eimeria-infected) on body weight gain and hematological variables in the chickens was determined. Means were separated using Tukey’s method. Canonical discriminant analysis was used to identify the combination of coccidiosis parameters that best separate the two groups (uninfected and infected). The combination of measurements that best discriminate between the two groups was selected and a discriminant function model was obtained from there. To identify infected and uninfected chickens, the unstandardized discriminant function procedure was employed (Tenko and George 2008). The ability of this function to identify infected chickens with Eimeria tenella from uninfected chickens was indicated as the percentage of individuals correctly classified from the samples that generated the function. Accuracy of the classification was evaluated using a priori method at p ≤ 0.05. All analyses were done using SAS (2010).

Results and discussion

The mean (± SE) values of the coccidiosis parameters presented in Table 1 showed significant differences (P < 0.01) in body weight gain at day 3 (BWG 3), monocyte (Mono), packed cell volume (PCV), and white blood cells (WBC) in male chickens. Only PCV and WBC as well as BWG 3, 6, and 12 were significantly different in female chickens. Reduction observed in PCV of infected male and female chickens was comparable to those observed by Fukata et al. (1997) and Meskerem et al. (2013) who reported lower counts of red blood cells and PCV in chickens infected with Eimeria tenella when they were compared to the uninfected chickens. Increase in WBC count obtained in this study was similar to results of Ricklefs and Sheldon (2007), who reported high counts of WBC in infected animals.

Table 1 Means and standard errors of body weight gain, fecal egg count, lesion scores, and hematological variables in the Eimeria-infected and non-infected Nigerian indigenous chickens
Full size table

Although the univariate statistics showed significant differences in some variables, the multivariate method provided better resolution. Wilk’s lambda was used for multivariate statistical test of group differences in male and female chickens (Table 2). The two groups do differ significantly when the coccidiosis resistance variables are considered simultaneously. This implies significant discriminant function (a linear combination of the parameters). This was evidenced by the selection of four variables (PCV, WBC, BWG 3, and lymphocytes) in male chickens and three variables (PCV, RBC, and BWG 3) out of the 12 variables assigned for stepwise discriminant analysis. A stepwise discriminant analysis was carried out to determine if any of the variables could be used to categorize chickens into one of the two groups (Eimeria-infected and uninfected chickens). The results implied that PCV, WBC, BWG 3, and lymphocytes were observed to be the most informative variables to effectively place infected and uninfected chickens in distinct group for male chickens, while in female chickens, PCV, RBC, and BWG 3 were the main variables in differentiating between infected and uninfected chickens. PCV and BWG 3 were found to be important variables in distinguishing infected and uninfected chickens in both sexes. This implied that factors leading to coccidiosis may also cause significant changes in other physiological characteristics such as hematology and body weight gain which may be used as substitute signs for coccidiosis disease. Data from the stepwise discriminant analysis was successively used to develop linear models representing the contribution of each of the important variables to be able to distinguish between the infected and uninfected groups. The discriminating variables extracted for each sex were included in a discriminant equation (Y):

$$ {\displaystyle \begin{array}{ll}\mathrm{Y}=-742.46+389.26\mathrm{PCV}-14.77\mathrm{WBC}-0.62\mathrm{BWG}3+527.17\mathrm{LYMP}& \mathrm{Male}\\ {}\mathrm{Y}=-140.35+162.609\mathrm{PCV}+55.62\mathrm{RBC}+20.50\mathrm{BWG}3& \mathrm{Female}\end{array}} $$
Table 2 Weight gain, FEC, lesion scores, and hematological variables pain the Eimeria-infected and non-infected Nigerian indigenous chickens
Full size table

With the discriminant equation, new measurements of PCV, RBC, and BWG 3 could be assigned into the equation to estimate discriminant scores using a pocket calculator. Positive discriminant score indicates the chicken is infected while negative discriminant score indicates uninfected chicken. A similar study was carried out by Alshamisi et al. (2013) in camels using only hematological parameters to distinguish fracture, lame, and normal camels.

The effectiveness of the model to be able to discriminate between the two chicken groups using coccidiosis variables in each of the model was also tested. The discriminant function was able to correctly classify 90.48% of the 63 infected chickens and 77.78% of the 45 uninfected chickens in male chickens, while in female, 98.33% of the 60 infected and 76.47% of the 51 uninfected chickens were studied (Table 3). Female infected chickens were more accurately differentiated than their male counterpart. Cross validation with the prior method indicated 85.19% of the original grouped correctly classified in male while 88.29% of the original group correctly classified in female. This study’s finding is relevant in order to establish measurable evaluation model for E. tenella resistance and to explore new coccidiosis resistance breeding method for chickens.

Table 3 Classification of Nigerian indigenous chickens based on the discriminant model for males and females
Full size table

Conclusion

Statistical models developed in this study as predictive tool could successfully differentiate infected from uninfected chickens using results from routine hematological tests. In both males and female chickens, over 90% of infected chickens could be successfully distinguished from the uninfected chicken group using four variables (PCV, WBC, LYMP, and BWG 3) in males and three variables (PCV, RBC, and BWG 3) in female chickens. We therefore conclude that in diagnosing and predicting chicken infected with Eimeria our two models could be used with minimum rate of misclassification without involving all clinical signs, biochemical, pathmorphological and histological analyses.