The Use of Discrete Characters in Discriminant Analysis for Diagnosis of Pulmonary Tuberculosis and for Classification of Patients Differing in Treatment Efficiency Based on Polymorphisms at Nine Codominant Loci: HP, GC, TF, PI, PGM1, GLO1, C3, ACP1, and ESD

Abstract

Discriminant analysis was used to differentiate patients with pulmonary tuberculosis (N = 106) from healthy individuals (N = 328) and patients whose treatment was efficient (N = 71) from those whose treatment was inefficient (N = 35). The analysis involved the data on nine polymorphic codominant loci: HP, GC, TF, PI, PGM1, GLO1, C3, ACP1, and ESD. The loci were selected by significance of differences in genotype frequencies between tuberculosis patients and healthy controls (GC, TF, PI, C3, ACP1) or between the two groups of patients differing in treatment efficiency (HP, GC, PI, PGM1, C3, ESD). Discrimination was based on a graphic method of Bayes classification procedure with a single-variate nomograph allowing easy estimation of the a posteriori probabilities for an individual to be classified. The two groups of patients proved to be discriminated sufficiently well (probability of misclassification P _err = 0.24), whereas discrimination between tuberculosis patients and healthy individuals was less efficient (P _err = 0.33). The method was proposed as a means of predicting the efficiency of treatment in pulmonary tuberculosis. Along with clinical, roentgenological, and laboratory examination, discriminant analysis may be employed as an accessory test in diagnostics of pulmonary tuberculosis, especially when the diagnosis is questionable.