The statistical analysis was performed using a completely randomized design and the general linear model (GLM) pro-cedure of SAS 9.2 (SAS Institute, 2009). The model only included the level of supplementation. Pens were the experimental units for all analysis. Orthogonal polynomial contrasts were used to determine the linear and quadratic effects of the increas-ing levels of inclusion and Duncan multiple range test was used to compare means. Significance was declared at P