When a drug dose is administered to patients it becomes part of the non-deterministic process of pharmacokinetic variability. In other words, a particular dose does not lead to a specific concentration-time profile in all patients, but rather a distribution determined in part by alleles of genes encoding enzymes involved in xenobiotic metabolism, the particular physique of each patient as is the case of pyrazinamide [6
], or even dietary considerations. This means that in some patients, despite patients taking all their drug doses low drug concentrations could still be encountered, which could lead to emergence of drug resistance. Thus, resistance emergence could occur in part due to non-deterministic causes that have nothing to do with DOTS or default.
The response of the pathogen to a particular drug concentration-time profile is itself related to several PK/PD factors. For M. tuberculosis
, PK/PD factors have been derived in monotherapy studies in the hollow fiber system (HFS) [7
]. First, the shape of the concentration-time curve has been related to resistance emergence for each of the first line anti-TB drugs. Studies with isoniazid and pyrazinamide revealed that the relationship between drug exposure and population of drug-resistant M. tuberculosis
was a series of curves that changed with time, starting with a “U” shaped curve, which then evolved over time to an inverted “U” curve (). In other words, the relationship is defined by a quadratic function, with time as part of the defining characteristics of the leading coefficient (see reference 11
). Therefore, in interpreting indices at which resistance can be suppressed, the duration of therapy should be taken into consideration. Rifampin resistance emergence and suppression are linked to the peak concentration (Cmax
) to MIC, with optimal suppression of resistance at a free drug Cmax
/MIC of 175 [10
]. Isoniazid resistance emergence was demonstrated to be closely linked to both Cmax
/MIC and AUC/MIC [13
]. On the other hand, both pyrazinamide and ethambutol resistance emergence were associated with the % time concentration persisted above MIC (%TMIC
]. The lessons are obvious, resistance emergence to a drug depends on the drug exposure achieved, and in many situations the actual shape of the concentration-time curve, which often differ from the PK/PD parameter linked to microbial kill.
Change in size of drug-resistant M. tuberculosis population with exposure and time
These PK/PD results, as well as the exposures associated with optimal kill, can be used for several purposes. The first is to refine susceptibility breakpoints. Setting susceptibility breakpoints using the PK/PD approach does not just rely on the MIC distribution in wild type isolates, but also on the doses and the drug exposures achieved by the doses in patients, given pharmacokinetic variability. An isolate is defined as being drug-resistant if it has an MIC that precludes it from being effectively killed by antibiotic concentrations achieved in at least 90% of patients given a particular dose. Put simply, if an isolate cannot be effectively killed at site of infection in most patients by a drug after taking the maximum tolerated dose, then it is resistant to that drug. Using this approach, new critical concentrations for each of the first line anti-TB drugs, as well as moxifloxacin, were recently proposed, as shown in . The most dramatic changes are proposed for isoniazid and rifampin, and thus the definition of MDR-TB itself. These are the two drugs in which PK/PD studies have been performed using at least two independent models (mice, hollow fibers and guinea pigs) [10
] and population pharmacokinetic studies are available from at least 3 independent groups [18
]; utilizing any of these studies leads to the same conclusion on breakpoints, so that it is unlikely that bias from any one PK/PD model can be invoked. Nevertheless, further work is needed to confirm these proposed susceptibility breakpoints. The breakpoints we have proposed, as well as the currently accepted breakpoints, need to be examined in large datasets of combination therapy clinical studies, and each breakpoint examined for whether it can predict microbiologic failure.
The second use of PK/PD exposures that could suppress resistance is to design doses and dosing schedules that could suppress drug resistance emergence. This is because it has been demonstrated for virtually each anti-TB agent examined; that the dose associated with maximal kill is not necessarily the one that prevents resistance emergence. Indeed, for moxifloxacin and ciprofloxacin such exposures associated with maximal bactericidal effect were also the ones associated with maximal amplification of the resistant sub-populations [7
]. Thus, computer aided clinical trial simulations have been utilized to determine doses that best suppress resistance suppression for moxifloxacin, pyrazinamide, and rifampin. In HFS studies, a 24h AUC/MIC greater than 53 was associated with suppression of drug resistance. Monte Carlo simulations revealed that this target could be achieved by 59% of patients treated with 400 mg of drug and by 93% in patients treated with 800mg of moxifloxacin daily [7
]. Similarly, Goutelle et al examined if rifampin doses of 600 mg and 1200 mg could adequately achieve the Cmax
/MIC of 175 needed to suppress drug resistance [18
]. The standard dose of 600 mg a day fared badly, while 1,200 mg performed better. We performed similar studies for pyrazinamide’s ability to achieve % time above MIC of ≥67% needed to suppress resistance emergence [11
]. Doses ≥3G a day administered daily achieved the target in >90% of patients. All such doses advocated by this approach still need to be shown to be safe for patients. In the case of pyrazinamide, however, toxicodynamic analysis and meta-analysis recently suggested low rates of hepatotoxicity even at these high doses, provided they are administered no more than 2 months [12