Application of Principal Component Analysis as a Prediction Model for Feline Sporotrichosis

Abstract

Sporotrichosis is a worldwide zoonotic disease that is spreading and causing epidemics in large urban centers. Cats are the most susceptible species to develop the disease, which could cause significant systemic lesions. The aim was to investigate and to identify predictive indicators of disease progression by correlations between the blood profile (hematological and biochemical analytes) and cutaneous lesion patterns of 70 cats diagnosed with Sporothrix brasiliensis. The higher occurrence in male cats in this study could be related to being non-neutered and having access to open spaces. Principal component analysis (PCA) with two principal components, followed by binary logistic regression, and binary logistic regression analysis, with independent variables and backward elimination modeling, were performed to evaluate hematological (n = 56) and biochemical (n = 34) analytes, including red blood cells, hemoglobin, hematocrit, leukocytes, segmented neutrophils, band neutrophils, eosinophils, lymphocytes, monocytes, total plasma protein, albumin, urea, creatinine, and alanine aminotransferase. Two logistic regression models (PCA and independent variables) were employed to search for a predicted model to correlate fixed (isolated) and disseminated cutaneous lesion patterns. Total plasma protein concentration may be assessed during screening diagnosis as it has been recognized as an independent predictor for the dissemination of cutaneous lesion patterns, with the capability of serving as a predictive biomarker to identify the progression of cutaneous lesions induced by S. brasiliensis infections in cats.