Susceptibility Tests of Anaerobic Bacteria: Statistical and Clinical Considerations

Abstract

In considerations of susceptibilty testing of bacteria, linear regression has been used to relate minimal inhibitory concentration and zone size. Although lacking in statistical validity, this technique has been used for aerobic bacteria. Reports of anaerobic testing show much greater variation about the regression line; the errors of misclassification have not been analyzed. A bivariate “error rate-bounded” classification scheme is proposed for relating minimal inhibitory concentration and zone size of bacteria. This method requires the following clinical input: (1) specification of breakpoints of susceptibility as determined by minimal inhibitory concentrations and the pharmacology of the antibiotic, (2) determination of the relative importance of two types of errors, false resistant classification and false susceptible classification, and (3) determination of the acceptable rate of error of false classification. Based on this input the classification scheme determines the zone size, which gives the minimal error of misclassification with the known data on zone diameter and minimal inhibitory concentration.