ABSTRACT
Sixty days of ciprofloxacin administration at 500 mg every 12 h is currently recommended for the prophylaxis of inhalational exposure to Bacillus anthracis. We examined Bacillus anthracis (Δ-Sterne strain) in our hollow-fiber infection model. We measured the ciprofloxacin concentrations achieved and the number of organisms present before heat shock (total population) and after heat shock (spore population). We fit a mathematical model to these data. Monte Carlo simulation with differing initial spore burdens (3, 5, and 6.9 log10 CFU/ml) demonstrated that 35 days of this regimen would completely clear the spore burden in 95% of patients. Durations of 110 days did not achieve 99.9% eradication, irrespective of initial burden, because of between-patient variance in drug pharmacokinetics. Given the absence of person-to-person transmission for Bacillus anthracis, adverse drug effects with long-term ciprofloxacin administration, and the possibility of engendering resistance in bodily flora, shorter prophylaxis duration should be given consideration, along with careful monitoring of all exposed individuals.
In 2001, the United States experienced the intentional release of spores of the pathogen Bacillus anthracis, using the U.S. Postal Service as the delivery vehicle. These events gave rise to a considerable effort to identify new therapeutics, diagnostics, and vaccines for this pathogen (13). In 2002, a consensus statement was released regarding recommendations for management of anthrax, both for therapy and for prophylaxis (7). For prophylaxis, the primary recommendation was administration of ciprofloxacin at a dose and schedule of 500 mg every 12 h for 60 days. As mentioned in the article, the recommendations for prophylaxis were empirical.
In this study, we employed our hollow-fiber (HF) infection model (5) to evaluate two regimens of ciprofloxacin administration (500 mg every 12 h and 500 mg every 24 h) for 15 days and to fit a mathematical model to the data to identify the relationship between ciprofloxacin concentration-time profiles and the ability of the regimens to kill vegetative-phase organisms and to track the number of spores present over time (with the assumption that the ciprofloxacin would not have a microbiological effect on the spore form of the pathogen).
In many papers, drug doses are recommended as if the population of patients will develop the same concentration-time profile. This is manifestly untrue, as between-patient variance does exist. We therefore explored the impact of this true between-patient variance in pharmacokinetic parameters upon the clearance of B. anthracis spores by employing Monte Carlo simulation. The impact of differences in pharmacokinetic parameters across patients would allow evaluation of the current recommendation for a 60-day duration of ciprofloxacin prophylaxis.
MATERIALS AND METHODS
Bacteria, media, susceptibility testing, and mutation frequency to resistance.The Δ-Sterne strain of B. anthracis was evaluated. This strain lacks the pX01 and pX02 virulence plasmids containing the toxin and capsule genes, respectively. Ciprofloxacin powder was purchased from MP Biomedicals (Solon, OH).
MICs for ciprofloxacin were determined simultaneously by broth macrodilution and agar dilution methods in Mueller-Hinton II broth and Mueller-Hinton II agar (MHA) using the methods outlined by CLSI (3). MICs were read after 24 h of incubation at 35°C. Trailing endpoints were observed. After discussion with H. Heine, our coinvestigator at USAMRIID and a member of the CLSI advisory committee, the MIC was defined as the lowest ciprofloxacin concentration that resulted in ≥80% reduction in growth compared to the growth controls. Minimum bactericidal concentrations (MBCs) were determined using standard methods (11).
In vitro HF pharmacodynamic infection model.The HF infection model described previously (5) was used to study the response of B. anthracis to ciprofloxacin exposures, simulating human pharmacokinetics. HF cartridges (FiberCell Systems, Frederick, MD) consist of bundles of HF capillaries encased within a plastic housing. The fibers have numerous pores (50% cutoff of 25,000 Da) that permit the passage of nutrients and low-molecular-weight species, such as antibiotics, but exclude bacteria. Approximately 15 ml of extracapillary space lies between the fibers and the cartridge housing. Medium within the central reservoir was continuously pumped through the HFs, and low-molecular-weight compounds rapidly equilibrated across the fibers with the extracapillary space. Microorganisms inoculated into the extracapillary space were exposed to conditions approximating those prevailing in the central reservoir.
Antibiotic was infused over 1 h into the central reservoir at predetermined time points by syringe pumps. Antibiotic-containing medium was isovolumetrically replaced with drug-free medium, simulating a half-life of 4.5 h. The rate constant of elimination of antibiotic was the rate of fresh medium infusion divided by the volume of the medium in the total system. The system simulated a single-compartment model with exponential elimination.
For each experiment, 15 ml of B. anthracis suspension (107 CFU/ml) was inoculated into the extracapillary space of three HF cartridges, and the experiment (performed in duplicate) was initiated by infusing antibiotic. At predetermined time points an 800-μl sample of bacteria was collected from each HF system. Washed samples were quantitatively cultured onto drug-free MHA (total organisms). Half of each sample was subjected to “heat shock” (exposure to 65°C for 30 min in a water bath), which kills vegetative-phase bacteria but allows spore survival. Samples taken from the central reservoir over the first 48 h were assayed for ciprofloxacin concentrations to validate that the desired pharmacokinetic profiles were achieved. The measured drug concentrations were within 10% of the targeted values.
The experiment using ciprofloxacin at 500 mg every 12 h was repeated to document the reproducibility of the findings, as this is the recommended regimen (2) and was employed for the Monte Carlo simulations.
Ciprofloxacin concentration determinations.Mueller-Hinton II broth samples were diluted with high-pressure liquid chromatography water (0.050 ml of sample into 1.00 ml of water), and were analyzed by high-pressure liquid chromatography tandem mass spectrometry (LC-MS-MS). The LC-MS-MS system was comprised of a Shimadzu Prominence high-pressure liquid chromatography system and an Applied Biosystems/MDS Sciex API5000 LC-MS-MS.
Chromatographic separation was performed using a Phenomenex Luna phenyl-hexyl column, 5 μm; a 150- × 3.0-mm column; and a mobile phase consisting of 88% 0.1% formic acid in water and 12% 0.1% formic acid in acetonitrile, at a flow rate of 0.75 ml/min. Ciprofloxacin concentrations were obtained using LC-MS-MS monitoring of the MS-MS transition m/z 332 → m/z 288. Analysis run time was 3.0 min. The assay was linear over a range of 0.010 to 8.0 mg/liter (r2 > 0.996). The interday coefficients of variation ranged from 5.79 to 6.40%, with all coefficients of variation and variances in accuracy being <±5.0%.
HF system experimental design.The ciprofloxacin experiments consisted of two systems in which regimens of 500 mg once daily and 500 mg twice daily were simulated plus a no-treatment control. Samples were taken from each HF system over the 15-day experiment. The samples were washed twice to prevent drug carryover and were plated onto drug-free MHA. Heat shock was performed as described above. Spores represented greater than 99% of the total population at baseline. During an exposure, it is likely that this value or a greater percentage of spores would be experienced by the exposed patient.
Modeling of exposure-response relationship of the total bacterial population and the spore subpopulation to drug pressure.A mathematical model was constructed to describe exposure-response relationships of a mixed population of vegetative-phase organisms and spore-phase organisms in an HF infection model under differing amounts of antimicrobial pressure. The mass-balance equations (three parallel first-order inhomogeneous differential equations) that describe the time course of ciprofloxacin concentrations (equation 1), the vegetative phase-organisms (equation 2), and the spore-phase (equation 3) are shown below. $$mathtex$$\[\mathrm{d}X_{1}/\mathrm{dt}{=}R(1){-}(\mathrm{SCL}/V_{c}){\times}X_{1}\]$$mathtex$$(1) $$mathtex$$\[\mathrm{d}N_{v}/\mathrm{dt}{=}K_{g-v}{\times}E{\times}N_{V}{-}K_{\mathrm{kill}-V}{\times}M{\times}N_{V}{-}K_{V-S}{\times}N_{V}{+}K_{S-V}{\times}N_{S}\]$$mathtex$$(2) $$mathtex$$\[\mathrm{d}N_{S}/\mathrm{dt}{=}K_{V-S}{\times}N_{S}{-}K_{S-V}{\times}N_{V}\]$$mathtex$$(3) $$mathtex$$\[E{=}[1{-}(N_{V}{+}N_{S}){\times}\mathrm{POPMAX}]\]$$mathtex$$(4) $$mathtex$$\[M{=}(X_{1}/V_{c})^{H}/[(X_{1}/V_{c})^{H}{+}C_{50-K}^{H}]\]$$mathtex$$(5) Equation 1 describes drug pharmacokinetics in the HF system (a one-compartment open model with zero-order input and first-order elimination). X1 is the amount of drug in the central compartment. R(1) is a zero-order, time-delimited infusion into the central compartment (mg/h). SCL (liters/h) is the rate of clearance of drug from the system; Vc is the central compartment volume.
Equations 2 and 3 describe the rates of change of the vegetative- and spore-cell populations, respectively, over time. The model equation for describing the rate of change of the numbers of microorganisms in the vegetative population was developed based on the in vitro observation that Bacillus anthracis bacteria in the HF system are in logarithmic growth phase in the absence of drug and exhibit an exponential density-limited growth rate (equation 4). For the vegetative-phase organisms, first-order growth was assumed, up to a density limit. For the spore phase, we described the dynamics by first-order transfer constants between the vegetative phase and the spore phase (KV-S) and back again (KS-V). As the vegetative microorganisms approach maximal bacterial density, they approach stationary phase. This is accomplished by multiplying the first-order growth terms by E (equation 4; a logistic carrying function). The maximal bacterial density (“POPMAX”) is identified as part of the estimation process.
Equation 2 allows the antibacterial effect of the different drug doses administered to be modeled. For the vegetative population, there is an effect of the drug dose on this population. There is a maximal kill rate that the drug can induce for each population (Kk-V). The killing effect of the drug was modeled as a saturable event (M, equation 5) that relates the kill rate to drug concentration, where H is the slope constant and C50 (mg/liter) is the concentration at which the bacterial kill rate is half-maximal. We assumed that spore-phase organisms are not killed by ciprofloxacin (or any other antibiotic). The drug effect observed is a balance between growth and death induced by the drug concentrations achieved. NV and NS are the numbers of colonies per ml in the vegetative and spore populations, respectively.
Measured outputs were as follows: (i) change in ciprofloxacin concentration over time, (ii) changes in total population (total = NV + NS), and (iii) spore (NS) densities, where the total population (vegetative plus spore burden) was enumerated prior to heat shock and spore populations were enumerated on drug-free plates after heat shock (which kills the vegetative-phase organisms).
The different drug treatment regimens were simultaneously comodeled in a population sense using the population modeling program BigNPAG (10). Weighting was as the inverse of the estimated observation variance, so as to approximate the homoscedastic assumption. Point estimates of the weights were derived by fitting individual files first with the same model in the ADAPT II package of programs of D'Argenio and Schumitzky (4), using the maximum likelihood weighting option.
For the simulation portion of the evaluation, we added a term for intrinsic spore clearance, as identified in a previous publication (2). This first-order clearance rate constant was added to equation 3, and the rate constant KsporeCL was given a value of 0.0029 h−1 (0.07 day−1), as in that publication (2).
Monte Carlo simulation.We employed the ADAPT II package of programs of D'Argenio and Schumitzky (4) for simulations. Both normal and log-normal distributions were examined, and the choice between distributions was made on the basis of the fidelity with which the starting parameter values and their variances were recreated by the simulation. The simulation was 9,999 iterates.
As we studied only one strain of B. anthracis, we fixed the model parameters having to do with the interaction of the drug and the organism. The study of a single strain is a mandatory limitation, as this is a select pathogen. We used the Δ-Sterne strain, as this can be employed in a biosafety level 2 setting. The only other strain available for this setting would be the Sterne strain, which is the parent of the Δ-Sterne strain and would add little biological variability. The standard deviations for the volume and clearance terms were derived from previously published studies (9) of ciprofloxacin pharmacokinetics in healthy subjects, as we felt this would best represent the population that would be receiving ciprofloxacin prophylaxis in the event of an intentional anthrax release. As part of the simulation, we examined three different starting challenges of anthrax spores, the challenge number in the HF system (8.7 × 106) and smaller challenges of 1 × 105 and 1 × 103. This allowed the impact of differing spore challenge size to be evaluated with regard to clearance. The actual HF study was performed only with an initial inoculum of 8.7 × 106 organisms/ml.
RESULTS
MIC/MBC of the B. anthracis Δ-Sterne strain for ciprofloxacin.The modal MIC/MBC of ciprofloxacin for the Δ-Sterne strain was 0.06 mg/liter for the MIC and 1.0 mg/liter for the MBC.
Control system growth.The no-treatment control grew well over time, starting at 7.13 log10 CFU/ml. Initially, the number of spores was very high, differing from the total by only 0.1 log10 CFU/ml, indicating that vegetative-phase cells accounted for only 20% of the total population (vegetative-phase cells plus spores). Over time, the total population grew to approximately 8.5 log10 CFU/ml. As there was no drug pressure brought to bear and as the environment was nutritionally replete, the number of spores slowly declined, being 1.5 logs lower than the total on day 1 and 2.5 log10 CFU/ml lower on day 4 and reaching 3.7 log10 CFU/ml lower on day 14.
Model fit to the data.The model described in Materials and Methods was fit to the data. The mean point estimates of the parameters and their standard deviations are shown in Table 1.
Parameter estimates for the model describing the changes in ciprofloxacin concentrations and the consequent changes in the total population (vegetative plus spore phase) and in the spore burden
The overall fit of the model to the data after the Bayesian step was quite acceptable for all three outputs (drug concentrations, total population, and spore population). For ciprofloxacin concentrations, the observed-predicted plot had an equation as follows: observed = 1.021·predicted + 0.013; r2 = 0.975; P ≪ 0.001; mean weighted error (MWE) = −0.154; bias-adjusted mean weighted squared error (B-AMWSQE) = 0.354.
For the total (vegetative- plus spore-phase) population, the equation was as follows: observed = 0.999·predicted + 0.098; r2 = 0.977; P ≪ 0.001; MWE = −0.441; B-AMWSQE = 0.953.
For the spore (heat-shocked) population, the equation was as follows: observed = 1.029·predicted − 0.159; r2 = 0.926; P ≪ 0.001; MWE = −0.226; B-AMWSQE = 4.56.
Figure 1A to F shows the fit of the model to the data for the two regimens for the three outputs. Figure 1A and B show that the fit of the model to the data was good. The desired concentrations were achieved to within 10% at all time points. In the same way, Fig. 1C and D show that the total population was well fit by the model and that the heat shock estimate of the spore burden was likewise well fit (Fig. 1E and F).
(A) Fit of the model (•) to the ciprofloxacin concentration-time data (▵) for a regimen of 500 mg every 12 h. (B) Fit of the model (•) to the ciprofloxacin concentration-time data (▵) for a regimen of 500 mg every 24 h. (C) Fit of the model (•) to the total counts (vegetative plus spore phase) of Bacillus anthracis (▾) for a regimen of 500 mg every 12 h. (D) Fit of the model (•) to the total counts (vegetative plus spore phase) of Bacillus anthracis (▾) for a regimen of 500 mg every 24 h. (E) Fit of the model (•) to the spore counts as derived after heat shock of Bacillus anthracis (▾) for a regimen of 500 mg every 12 h. (F) Fit of the model (•) to the spore counts as derived after heat shock of Bacillus anthracis (▾) for a regimen of 500 mg every 24 h.
The original experiment and the repeat experiment were examined for repeatability by linear regression. Over 15 days, the correlation coefficient (r) was 0.9427 and the coefficient of determination (r2) was 0.889 (P = 0.000044).
Extension of the observation period.As we were predicting the future with the model, we chose to perform a longer observation experiment of >60 days with several drugs, one of which was ciprofloxacin (data not shown). We documented that spore counts declined below 50 colonies per ml at approximately the point predicted by the model. Addition of an intrinsic spore clearance rate demonstrated that, at mean, the spores were completely cleared prior to 35 days, quite consistent with the Monte Carlo simulation findings below.
Monte Carlo simulation.Figure 2 shows the simulated ability of a ciprofloxacin regimen of 500 mg every 12 h to clear the spore burden, as a function of changing initial spore burdens. Figure 1E demonstrates from data that the exposures employed here (free drug area under the concentration-time curve = 31.1 mg·h/liter) for this regimen cause the spore burden to decline from slightly less than 7 log10 CFU/ml to 2.9 log10 CFU/ml over a period of 15 days. It needs to be recognized that in an outbreak situation, where many people are provided prophylaxis, not everyone will have the same drug exposure. Here, we allowed a modest between-patient variability of 20% (9), suggested as appropriate for a large, otherwise healthy population receiving prophylaxis. The impact of this between-patient variability is substantial. The 50% of patients who had ciprofloxacin exposures greater than that achieved in vitro (free drug area under the concentration-time curve = 31.1 mg·h/liter) were, in general, those who cleared their spore burden completely in the first 5 to 10 days (depending on the initial burden—3.0, 5.0, or 6.9 log10 CFU/ml of spores). It should be recognized that these times are somewhat foreshortened, as we have added an intrinsic rate of spore clearance to the model (spore clearance rate constant = 0.0027 h−1), dependent on immunity, as suggested previously (2). The HF model system has no immune function. Other simulations not including the intrinsic spore clearance did not reveal a major impact of the time to complete clearance for different proportions of the population up to the 95% level (data not shown).
Percentages of simulated subjects who have completely cleared their spore burdens over time for starting burdens of 3 log10 CFU/ml (•), 5 log10 CFU/ml (▴), and 6.9 log10 CFU/ml (▪).
It should also be noted that 95% of the population would be expected to clear their burden by 5 weeks of therapy for any of the initial burdens. By the same token, to capture the last 5%, therapy duration requires a large increase, with 99% of the populace completely cleared by 36 to 82 days of therapy, depending on the initial spore burden. We extended the simulations to 110 days. None of the initial spore burdens were cleared by day 110 at the level of 99.9% of the simulated population.
DISCUSSION
Bacillus anthracis remains a major concern as an agent of bioterror. Ciprofloxacin has been recommended as the agent of first choice for both treatment and prophylaxis for Bacillus anthracis. In the setting of prophylaxis, large populations may seek antibiotic therapy. Currently, the recommendation is for 60 days of continuous therapy. Such a duration raises a number of concerns: side effects, emergence of resistance for bystander organisms, and adherence to the regimen (1, 8).
In this set of experiments, we were able to demonstrate that the regimen of 500 mg every 12 h was able to engender a multilog decline in both the vegetative-phase organisms and the number of B. anthracis spores. At first consideration, this may seem unusual, as spores are nearly invulnerable to the effects of antimicrobials. When one looks at Table 1, it is clear that the rate constant for transition from vegetative state to the spore phase is more rapid (but not significantly so) than the rate constant from spore to vegetative phase. Upon selective pressure (in this case ciprofloxacin exposure, but in humans or animals, immune pressure would also come to bear), the major movement would be from vegetative to spore phase. However, there is also a slower transition back to vegetative phase. This suggests that drug must be present continuously and for a long period to have a high probability of killing the organism during the transition back to vegetative phase. It should be noted that the MBC is considerably higher than the MIC (16-fold). Also, the trough ciprofloxacin concentrations exceed the MIC for the vast majority of subjects for this regimen. This helps to explain the success of the regimen (12). As long as the regimen is adhered to, there will be a high probability that there will be enough drug around to suppress vegetative-phase cells that have just transitioned back from spores. There may not be enough to kill them immediately, but they will not amplify. The next dose will have a high likelihood of achieving a sufficient concentration for a sufficient period of time to then kill the organism before it can transition back to spore phase. Ultimately, this allows clearance of the total spore burden in most patients.
The time that this requires will be dependent on the initial spore burden. The impact of spore burden can be seen in Fig. 2. From a public health perspective, one would like to provide prophylaxis for a period of time so that the vast majority of patients would completely clear their burden, so that clinical pulmonary anthrax does not result.
The initial burdens examined here run a range from a very small initial exposure to a massive one. Heine and colleagues examined the prophylactic effect of ciprofloxacin in BALB/c mice exposed to 50 50% lethal doses of Bacillus anthracis in an inhalation chamber (6). The spore burden at the initiation of therapy approximated 107 CFU/g of lung tissue, a value concordant with that used in this evaluation. Brief exposures to aerosols in an open environment would be unlikely to exceed such a challenge.
When one examines Fig. 2, it is clear that the time to total clearance of the initial spore burden does not differ greatly by virtue of the size of the initial burden, at least out to clearance of 90 to 95% of the initial population. For 90% population clearance, the range is from 11 to 23 days. For 95% clearance, the range is from 16 to 35 days. These numbers are trivially larger in the absence of the rate of intrinsic spore clearance.
Providing prophylaxis for longer periods is certainly reasonable. However, to get to the level of 99% population clearance, the duration ranges from 36 to 82 days, as a function of initial burden. Even at 110 days, we do not attain 99.9% clearance. This leads to an explicit evaluation of the balance of cost and benefit.
Even modest (20%) between-patient variability in pharmacokinetic parameter values ensures that complete spore burden clearance will be unlikely for at least 0.1% of the exposed population, even with prolonged (110-day) prophylaxis. However, 95% clearance can be achieved even for a massive initial spore burden at 35 days. There has never been a documented case of person-to-person transmission for Bacillus anthracis. Therefore, the therapeutic “downside” for shortening the duration of prophylaxis to 5 weeks should be minimal. No more than 5% of patients (and very likely far fewer) should have spores to germinate and cause clinical disease. The shortened prophylactic course needs to be married to continued observation for the patients exposed. However, this is no less the case for longer prophylactic courses, because of the inability to clear 99.9% of the population, even with 110 days of prophylaxis.
The “upside” of the foreshortened course is that many fewer patients will suffer continuing adverse events, there will be less likelihood of engendering resistance in “innocent bystander” organisms, and the likelihood of engendering ciprofloxacin-resistant isolates of B. anthracis will also be reduced. That such resistance does occur under antibiotic pressure has been documented in vitro (5) and in vivo (6).
In summary, while ciprofloxacin at 500 mg every 12 h is highly likely to be an effective regimen for anthrax prophylaxis, the current recommendation of 60 days of duration for such prophylaxis may be excessive, as durations longer than 5 weeks are unlikely to cause complete clearance of large initial spore burdens in more than 95% of the exposed population. Shorter courses have the benefit of limiting bystander damage and are also likely to be conservative, as nonhuman primate challenge data have demonstrated that even relatively short prophylactic courses of fluoroquinolone drive a high probability of survivorship, even with exposures exceeding 200 50% lethal doses, likely because of generation of active immunity by the exposure because of the delay in therapy initiation after exposure (J. Estep, Battelle Memorial Institute, personal communication). Consideration should be given to shortening the duration of ciprofloxacin prophylaxis for Bacillus anthracis exposure to approximately 5 weeks.
ACKNOWLEDGMENTS
This study was funded by 1 PO1 AI060908-01A1, “Choosing Drug Doses for Biodefense Pathogens.”
All the authors state that they have no conflicts of interest.
FOOTNOTES
- Received 4 April 2008.
- Returned for modification 14 May 2008.
- Accepted 18 August 2008.
↵▿ Published ahead of print on 25 August 2008.
- American Society for Microbiology
