[From prevalence to predictive values: considerations on the antimicrobial susceptibility testing dealing with change of susceptibility rates].
Résumé
The aim of the study was to quantify the influence of the variation of the susceptibility prevalence (frequencies of susceptible, intermediate and resistant isolates) on the predictive value of antimicrobial susceptibility testing reports performed with disk diffusion method, under several scenarios of distinct susceptibility prevalence. Prevalence variation effect was assessed through a modeling approach that enabled to take into account the technical variability and to control prevalence through scenarios. Results show how the prevalence impacts on the disk diffusion performance with level of discrepancies varying with prevalence. The phenomenon may be quantitatively noteworthy for some antimicrobial agents considering the prevalence recorded with time or location, leading to lowered performances. For instance, with amoxicillin+clavulanic acid, predictive values of susceptible and resistant reports varied respectively from 70 to 96 and from 33 to 97 %. For a 18-mm diameter, the probability the isolate is truly susceptible varied from 16 to 73 % according to the prevalence tested. Characteristic of the prevalence effect are described and consequences on zone diameter breakpoint policy raised. They include to (i) reevaluate disk diffusion breakpoint consistency when the prevalence impact is noteworthy and (ii) estimate the consequences of an international harmonized breakpoint policy on prediction quality and appropriate patient management.