Use of statistical methods in the rapid determination of antimicrobial susceptibilities of Escherichia coli

A rapid method for determination of MICs for Escherichia coli in broth cultures was developed. Each strain was tested against 11 concentrations of an anti-microbial agent in twofold dilution series as well as positive and negative controls. The agents tested were: ampicillin, nalidixic acid, nitrofurantoin, sulphamethoxazole, tetracycline, and trimethoprim. Extinction measurements by a continuous flow method, after 3–4 h incubation, were used to estimate growth of 203 bacterial strain/antimicrobial agent combinations and expressed as a percentage of that obtained in a control broth. The conventional MICs for these strains were determined by the twofold serial dilution tube method. The strains were also tested for disk susceptibility by the Stokes method. Percentages of growth obtained by the rapid method were used to predict conventional MIC in two ways: a cut-off point, and a linear regression equation using the per cent of growth in 1, 2, or all 11 concentrations. The cut-off method gave 79% agreement, within ±1 dilution, with conventional MIC. The regression models with 1, 2, or 11 concentrations gave 95, 95, and 98% agreement respectively. The regression model permitted determination of MIC from growth measurements in only three tubes. For sulphamethoxazole a larger number of tubes may be needed. Discriminant function analysis was used to categorize strains into susceptible and resistant on the basis of the growth measurements. This model gave 100% aggreement with conventional disk susceptibility testing. The model permitted prediction of susceptibility to sulphamethoxazole with 3 h incubation using four tubes.