TABLE 1.

Types of experiments that can be performed with widely used nonclinical PD infection models^a

Study objective	Static time-kill model	One-compartment system (“chemostat”)	Two-compartment hollow fiber system	Mouse infection model
1. Dose-range study: killing of predominant population	Yesb	Yesb	Yesb	Yesb
2. Dose-range study: suppression of resistance	±c	±c	Yesc	±c
3. Dose-fractionation study: killing of predominant population	No	Yes	Yes	Yes
4. Dose-fractionation study: suppression of resistance	No	±	Yes	±
5. Combination therapy: killing of predominant population	Yes	Yes (short term)	Yes	Yes
6. Combination therapy: suppression of resistance	±	±	Yes	±
7. Toxin suppression by drugs	Yes	±	Yes	Yes
8. Dissecting the interaction of the parent drug and metabolites on antimicrobial effect	±d	±d	Yesd	No
9. Effect of physiological state of bacteria on drug activity	±	±	Yes	±
10. PD index for drug toxicity	No	No (unless toxicity is acute)	Yes	±e

^aPD, pharmacodynamic; ±, study objective can potentially be addressed in this system.

^bBacterial strains which display the lowest mutation frequency of resistance should be avoided in dose-range studies; instead, strains which best represent the most commonly observed mutation frequencies are preferred.

^cStrains with a relevant resistance mechanism(s) should be chosen for in vitro studies. The MIC₅₀ and MIC₉₀ for the pathogen of interest may be used to guide strain selection.

^dA biologically active metabolite(s) needs to be available, since it is most likely not formed in the in vitro system.

^eSome dosage regimens (e.g., those used to assess time over a toxicity threshold) may also lead to high peak concentrations, especially for short-half-life drugs, which complicates the interpretation of these studies.