4.7. Statistical Analysis

Analyses were executed using SAS (version 9.4; SAS Institute, Cary, NC, USA). Prior to performing all analyses, the data were explored for normality and outliers using the PROC UNIVARIATE model in SAS and the Shapiro–Wilk test. NT-proBNP and cardiac troponin I data followed a non-parametric distribution and data was log-transformed to reach normality. One-way ANOVA was used to contrast differences among normal vs. high protein commercial diets and two-way ANOVA was used to contrast parameters for fava bean–based diets (fixed effects being fava bean variety and fermentation). In this current study, we used mixed-sex dogs to control the variation possibly connected with this factor. Preceding studies from our group have not discovered sex-related differences among (spayed/neutered) dogs [63] and because sex-related differences are not associated to the rationale of the present study, we did not incorporate this factor in the model. All post-hoc analyses were executed using the Fisher least significant difference (LSD) method. Differences were considered significant at p < 0.10.