![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
Previous Article | Next Article 
Antimicrobial Agents and Chemotherapy, November 2003, p. 3393-3399, Vol. 47, No. 11
0066-4804/03/$08.00+0 DOI: 10.1128/AAC.47.11.3393-3399.2003
Copyright © 2003, American Society for Microbiology. All Rights Reserved.
Division of Clinical Pharmacology, Clinical Research Initiative, Albany Medical College, Albany, New York,1 Clinical Pharmacology and Discovery Medicine, GlaxoSmithKline, Research Triangle Park, North Carolina2
Received 27 November 2002/ Returned for modification 17 May 2003/ Accepted 15 July 2003
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Another challenge is defining an optimal protease inhibitor concentration-time profile that will maximize viral suppression while allowing for a reasonable dosing schedule. An optimal level of drug exposure should be linked to a measure of viral susceptibility (e.g., peak concentration/the 95% effective concentration [EC95], area under the concentration-time curve [AUC]/EC95, time for which the concentrations are above the EC95 [time > EC95]).
One approach to treating patients who have experienced antiretroviral treatment is to provide a protease inhibitor drug exposure that is more likely to be effective against more resistant viral strains. Amprenavir concentrations in serum can be pharmacokinetically enhanced when the drug is dosed with the protease inhibitor ritonavir due to the ability of ritonavir to inhibit the cytochrome P-450-mediated metabolism of amprenavir (13, 14).
A measure of viral susceptibility relative to the EC50 for a wild-type virus is readily obtained by using phenotypic-resistance assays. These data are useful in combination with the known pharmacokinetics of amprenavir to maximize the likelihood of virological suppression. The purposes of the present study were to develop a target for antiviral concentration attainment for amprenavir in a hollow-fiber system (in vitro) and then to determine the likelihood of attainment of the target EC50 for a given virus when a dose of amprenavir-ritonavir is administered to patients (in vivo).
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Cell culture. The human T-lymphoblastoid cell line CEM and this line chronically infected with the IIIB strain of HIV type 1 (HIV-1) was originally obtained from the AIDS Research and Reference Reagent Program, AIDS Program, National Institute of Allergy and Infectious Diseases, Bethesda, Md. Cells were cultured in RPMI 1640 medium containing 10% (vol/vol) fetal bovine serum, 25 mM HEPES, and gentamicin (50 µg/ml; Paragon Biotech, Baltimore, Md.). The medium for hollow-fiber bioreactor culture was modified for growth in the absence of CO2 by the omission of Na2HCO3 during formulation. Hollow-fiber bioreactors with a 10,000-molecular-weight cutoff (Spectrum Laboratories, Inc., Rancho Dominguez, Calif.) were used for this study. Experimental hollow-fiber units (previously perfused for 16 to 24 h with test media) were initiated by introduction of 7 ml of a 1:100 mixture of HIV-infected and uninfected CEM cells at a density of 3 x 106 cells per ml into the extracapillary space. An initial sample was taken for determining the cell number, the percentage of infected cells, viability by trypan blue exclusion, and p24 level. In continuous-infusion experiments, the drug-containing medium was recirculated, and the replenishment of drug-containing medium was done daily to maintain adequate stability of the agent.
HIV antigen assay. Samples for the p24 antigen assays were taken at the time points indicated in Fig. 1 after the initiation of experiments. Aliquots of cells and tissue culture media were removed through the extracapillary space access ports on the bioreactors at the time points indicated in Fig. 1 and centrifuged for 5 min at 1,500 x g. The levels of HIV gag p24 protein in cell-free culture supernatants were measured by using the Coulter Immunology (Hialeah, Fla.) p24 enzyme-linked immunosorbent assay according to the manufacturer's guidelines. Some of the HIV p24 assays performed at SRA Life Sciences (Rockville, Md.) used p24 kits from Organon-Tecknika. Absorbance was measured and data were analyzed with a computer-supported microplate reader (Molecular Devices, Menlo Park, Calif.). Levels of the p24 protein were calculated by the SoftMax program provided by the manufacturer of the microplate reader. The levels of p24 at baseline (day 0) were subtracted from values obtained on later days.
|
Sensitivities of viral isolates to amprenavir. The sensitivities of the viral isolates to amprenavir were determined by dose-response determinations in an MTT-based cytotoxicity assay. The viral isolates were aliquoted and stored until titrations of the isolates were performed. MT-2 cells were infected at 106 cells per ml with a dilution of each isolate to be tested, resulting in a concentration which produces an adequate cytopathic effect (usually 2,000 times the 50% tissue culture infective dose) within the assay period of 5 to 7 days. After 1 h, cells were centrifuged free of the virus and the cells were suspended at 105 per ml. A total of 100 µl of cells was added to the central 60 wells of a microtiter plate that had previously received 100 µl of drug in growth medium. Twofold serial dilutions of the drug were set horizontally, resulting in a 2- to 3-log concentration range. Controls included uninfected cells, untreated wells, and cells infected with a virus(es) of known susceptibility to the drug in question.
Pharmacokinetic data.
Written statements of informed consent were obtained from patients before their enrollment in this clinical trial, and human experimentation guidelines of the U.S. Department of Health and Human Services were followed in the conduct of clinical research. Thirteen patients failing current therapy, with plasma HIV-1 RNA concentrations of 
1,000 and 
50,000 copies/ml at screening, enrolled in a clinical trial (ESS4006) of amprenavir and ritonavir administered at doses of 600 and 100 mg, respectively, twice a day (BID). The patients also received tenofovir, abacavir, and a third nucleoside reverse transcriptase inhibitor based on genotypic-resistance testing. Patients were 18 years of age or older and had a CD4+-lymphocyte count of 50 cells/mm3; 7 days of exposure to amprenavir, abacavir, and tenofovir; and HIV-1 isolates susceptible to amprenavir, abacavir, tenofovir, and the third nucleoside reverse transcriptase inhibitor. For all drugs, susceptibility was defined as an HIV strain with a 4.0-fold change in the 50% inhibitory concentration from that of the control virus. The pharmacokinetics of amprenavir was studied at steady state at day 14 of therapy. The dose was administered after an 8-h fast with 240 ml of water, and patients continued to fast for 3 h after being dosed. Sampling times were predose and at 0.25, 0.5, 0.75, 1.0, 1.5, 2.0, 2.50, 3.0, 4.0, 5.0, 6.0, 8.0, 10.0, and 12.0 h. Adherence to therapy was assessed by utilizing patient diary cards. Samples were analyzed by a high-performance liquid chromatography technique (13).
Pharmacokinetics-pharmacodynamics model.
A population pharmacokinetic analysis for amprenavir was performed by using a nonparametric adaptive grid (NPAG) (9) with an optimized 
. The NPAG program is a population modeling program in which there are no assumptions made about the underlying parameter distributions. The underlying distribution is one of the things discovered by the analysis. The optimized is a new addition to the program relating to the problem of the choice of weights. This is a scalar that multiplies the original model for the relationship between concentration and the variance of the observation. The scalar is reestimated and optimized with each cycle.
A two-compartment model with first-order absorption and elimination was used. A lag time for absorption was also built into the model. We assumed that total observation variance was proportional to assay variance. The initial weight was the inverse of the square of the value for the standard deviation of the observation estimated from the assay regression line. This value was multiplied by the scaling factor (), which was optimized with each cycle.
A maximum a posteriori (MAP)-Bayes analysis using the NPAG population-of-one utility and the population mean parameter vector was performed. From this, a predicted-observed plot was generated, and this plot was one measure of goodness of fit. A Monte Carlo simulation for amprenavir concentrations was performed for 2,500 subjects. The mean parameter vector and full covariance matrix were embedded in subroutine PRIOR of the ADAPT II package of programs of D. Z. D'Argenio and A. Schumitzky (user manual, Biomedical Simulations Resource, University of Southern California, Los Angeles, 1992). Both normal and log-normal parameter distributions were evaluated. A log-normal distribution was chosen, as this best recaptured the central tendency of the parameters and their dispersions.
Determination of fractional target attainment.
Concentrations of amprenavir in simulated patients were corrected for the effect on viral susceptibility (EC50) by the addition of 1.5 mg of human 
-1 acid glycoprotein/ml and 4 mg of human serum albumin/dl. In this experimental series, protein binding changed the EC50 upward in vitro by 7.4-fold. The experimentally determined EC50 was corrected for the difference between the EC50 and EC95 by multiplying the EC50 by a factor of 4. Previous experience has demonstrated that this is a reasonable approximation of the EC90-95 (4). It is important to recognize that it would be inappropriate to employ the actual EC50 instead of a multiple of this number, as the goal of therapy should be to shut down rounds of viral replication as completely as possible. Obviously, the EC50s turn down viral replication only by half.
An EC50 range of 0.002 to 0.160 µM was evaluated. Concentrations in plasma from the Monte Carlo simulation were generated at 0.1-h intervals throughout a steady-state dosing interval. For each of the simulated plasma amprenavir concentrations, the values were divided by a factor of 29.6 (4 x 7.4), thereby correcting for protein binding and the difference between the EC50 and the EC90-95. For an EC50 ranging from 0.002 to 0.160 µM, we then determined the fraction of the 2,500 simulated subjects that had free-drug concentrations in excess of four times the EC50 (approximately the EC95) for 80% of a 12-h interval. The entire dosing interval was used in the calculations of time (in hours) for which concentrations were above the EC95 (time > EC95), including time before the peak concentration.
A second target was determined from the minimum or trough concentration (Cmin)/EC50 ratio based on a previously published relationship among exposure, viral load decrease, and the probability of maintaining a viral log decrease through 8 weeks in a population highly experienced with retrovirals (D. S. Stein, Y. Lou, J. Johnson, and S. Randall, Abstr. 2nd Int. Workshop Clin. Pharmacol. HIV Ther., poster 5.6, 2001). The previously described relationship determined the link between the Cmin/EC50 and the probability of maintaining at least a 1.0-log10 time-weighted average change in viral load from baseline (in copies per milliliter) for combination therapy that included amprenavir. In the previous study, the protein binding correction was 10-fold rather than the 7.4-fold used in the in vitro model based on 90% protein binding. A protein binding-adjusted Cmin/EC50 ratio of 5 was associated with approximately a 90% probability of maintaining at least a 1-log time-weighted change in viral load from baseline to week 8.
|
RESULTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
80% of the dosing interval) is 97.4%. For an isolate with reduced susceptibility to amprenavir, i.e., for which the amprenavir EC50 is 0.050 or 0.080 µM, the likelihood of target attainment falls to 91.0 or 75.8%, respectively.
|
0.03 µM. |
DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In this evaluation, we chose to investigate how to optimize the protease inhibitor component of an antiretroviral regimen. It is important to note that in clinical care, one would also have the antiviral activities of other agents used in the combination therapy to drive antiviral activity. The first thing that needed to be defined was the pharmacodynamically linked variable. To do this, we studied amprenavir in our hollow-fiber pharmacodynamic model. As was demonstrated previously for the protease inhibitor atazanavir (4), the time that free-drug concentrations remain above four times the EC50 (time > 4 x EC50) is the pharmacodynamically linked variable. This can be clearly seen in Fig. 1, where there is an untreated control hollow-fiber unit and three treated hollow-fiber units. All three treated units have identical 24-h AUC exposures to amprenavir. The difference is that one unit has a continuous infusion at a concentration of 4 times the EC50, which generates a 24-h exposure of 24 times 4 times the EC50. For the other two units, the 24-h AUCs are identical (96 times the EC50), but in one unit, the AUC is broken up into two equal exposures of 48 times the EC50, administered 12 h apart, while in the other, it is broken into three equal parts (32 times the EC50), administered every 8 h. If the ratio of the AUC to four times the EC50 were the pharmacodynamically linked variable, all treated units would give the same degree of viral suppression. If the ratio of the peak concentration to four times the EC50 ratio were linked to viral suppression, then dosing every 12 h (q12h) would be best. Neither of these outcomes occurred; rather, continuous infusion was best and administration q8h had an equivalent outcome, with q12h dosing giving the least viral suppression, clearly demonstrating that time > 4 x EC50 is the appropriate dynamic variable for evaluation.
In order to set a specific goal for regimen evaluation, we fit an inhibitory sigmoid maximal-effect model to the data (Fig. 2). Clearly, the degrees of viral suppression were virtually identical with the continuous-infusion and q8h hollow-fiber units. The latter attained drug concentrations at least four times the EC50 for approximately 80% of the dosing interval. We chose this as the amprenavir exposure target that would indicate adequate therapy for that portion of an individual's regimen. The necessity to keep drug concentrations in excess of four times the EC50 for 80% of the dosing interval was also seen in previous work with atazanavir in our laboratory (4). It should be recognized that in the case of atazanavir, the dosing interval was 24 h. Therefore, the dosing interval coverage number of 80% is independent of the duration of the dosing interval, at least out to 24 h.
We also wished to examine the attainment of a target that was based on clinical (in vivo) data. To do this we employed a previously developed relationship for amprenavir relating a protein binding-adjusted Cmin/EC50 ratio to the probability of maintaining at least a 1.0-log10 time-weighted decline in HIV RNA concentrations from baseline to week 8 in a highly antiretroviral-experienced population. A Cmin/EC50 ratio of 5 was associated with approximately a 90% probability of attaining this goal.
The enhancement of protease inhibitor drug levels by adding a small amount of ritonavir is an important new approach to antiretroviral therapy (10). With this approach, the concentration of amprenavir is increased and the drug can be administered less frequently and with fewer capsules so as to optimize patient adherence. In the clinical research study performed, 600 mg of amprenavir was boosted with 100 mg of ritonavir, with both drugs being given on a q12h dosing schedule. For all of the subjects, the concentration-time profiles obtained during serial pharmacokinetic sampling (178 plasma samples) were population modeled. The model fit the data well, as demonstrated in Fig. 3, where, after MAP-Bayes estimation, the predicted-value-observed-value plot had a regression line with a slope of 0.993 (1.0 is a perfect regression slope) and an intercept of 0.176 (0.0 is a perfect intercept). The overall r2 was 0.925, with quite acceptable measures of bias and precision.
In order to obtain an estimate of how well a specific dose of drug will be able to attain its therapeutic target in a population, the full variability of the pharmacokinetic parameter values needs to be taken into account. That is, the relatively small pharmacokinetic trials that are performed do not allow an estimate of the number of patients that will fall in the tails of the distribution to be obtained. Monte Carlo simulation allows the investigator to obtain an idea of the true impact of between-patient variability in the pharmacokinetics of the drug on the ability of that dose of drug to attain a therapeutic goal. It has recently been demonstrated through Monte Carlo simulation (4, 6) that such an estimate can be made, and the predictions have been borne out in a clinical setting.
We performed a 2,500-subject Monte Carlo simulation for the combination of amprenavir and ritonavir, and steady-state amprenavir concentrations were simulated. We then examined the fraction of the simulated subjects that had their free-drug concentrations exceed four times the EC50 (approximately the EC95) for 80% of the 12-h dosing interval. These results (Fig. 4A) are quite impressive. Target attainment exceeded 90% when subjects received amprenavir up to an EC50 of 0.050 µM. Target attainment exceeded 50% for the population receiving amprenavir up to an EC50 of 0.120 µM. The results imply that this combination can be used without monitoring the concentrations of amprenavir when the EC50 is less than or equal to 0.050 µM (target attainment approximately 90%).
For patients with viral isolates for which the amprenavir EC50s were greater than 0.050 µM, there is still a reasonable expectation of attaining the exposure target, indicating that this can be part of a regimen for salvage patients and that when other active antiretroviral agents are used, a high probability of success can be anticipated; but for these patients, drug concentration monitoring might be utilized as part of the patients' management programs. It is important to emphasize that the protease inhibitor component of the regimen is only one part of a multidrug therapeutic combination. Even if the protease inhibitor part is not completely optimized, it is still possible to achieve maximal viral suppression due to the additional nonprotease antiretroviral drugs.
Based on our model, we have used 80% of the dosing interval for free-drug concentrations to exceed four times the EC50 as a target. In the clinic, for a 12-h dosing interval, this would translate to h 9.6 were there no lag time to absorption. As demonstrated in Table 1, there is a modest lag time with a central tendency of 0.616 (mean) to 0.470 (median) h and a 95% confidence interval from 0 to 1.82 h. To take this lag time into account, and to be conservative with regard to the clinical circumstance, we recommend that if the concentration in serum is monitored, the h 12 trough concentration would be most appropriate for determining whether or not a target of free drug more than four times the EC50 is attained, as monitoring levels at the time point of 80% of the dosing interval (h 9.6) is not practical. To be explicit, were one to monitor trough concentrations of amprenavir, we would recommend dividing the total drug concentration by a factor of 7.4 to adjust for the impact of drug binding to serum proteins. The EC50 should be multiplied by 4. The target, then, would be to have the protein binding-adjusted drug concentration at trough exceed four times the EC50.
The second, clinical-data-based target gives a similar target attainment rate distribution (Fig. 4B). It is important to note the relationship between the time > 4 x EC50 variable and the Cmin/EC50 ratio. Clearly, the goal is to have the number of rounds of viral replication be as low as possible, as this will provide the highest probability of driving the viral load to a level below assay quantitation as well as lowering the probability of the emergence of resistance. A measure of the time above threshold is bounded from above at 100%, while the Cmin/EC50 is not bounded. For the in vitro-derived target, adequate suppression of viral turnover is provided by having the protein binding-corrected drug concentrations exceed a target value that provides near-maximal effect (here, four times the EC50) for a large fraction (80%) of the dosing interval. For the Cmin/EC50 ratio, one must be careful in interpreting these values. A ratio of 1.0 guarantees only that the drug concentrations will provide half the maximal effect. When drugs are compared with regard to their Cmin/EC50 ratios, the meaning of the targets, the methodologies used for the protein binding corrections, the EC50 determinations, and the viral strains compared should be clear, and the ratios should, of necessity, have values of >1.0. It is also important to recognize that the clinical analysis used an average decline in viral load by taking the AUC of the viral load to the point of evaluation and dividing by the time. This will, of necessity, give a smaller viral load decline because of the effect of the viral loads determined early on in therapy. Further, the protein binding adjustment was done differently than in the in vitro evaluation, where the correction factor used was 10.0, not 7.4 (D. S. Stein et al., Abstr. 2nd Int. Workshop Clin. Pharmacol. HIV Ther.). The difference in the correction factor along with the need for a free target Cmin/EC50 ratio of 5 rather than 4 accounts for the differences between Fig. 4A and B.
In summary, we have demonstrated that, as for another protease inhibitor (atazanavir), time > 4 x EC50 is the pharmacodynamically linked variable for amprenavir. It should be recognized that the target set here for attainment is very conservative and that attainment of this target results in shutdown of virtually all rounds of viral replication in an in vitro model. We have also examined the variability of a ritonavir-enhanced protease inhibitor in the population and showed that the regimen studied performs very well up to an EC50 of 0.040 to 0.050 µM (95 to 90% target attainment rates). Even with more resistant isolates with EC50s around 0.120 to 0.140 µM, target attainment rates still exceed or approximate 50%. Amprenavir can be used with confidence in patients infected with viruses sensitive to protease inhibitors. For antiretroviral-experienced patients infected with less sensitive isolates, ritonavir-enhanced amprenavir still provides a reasonable probability of attaining an optimal exposure target, especially when combined with other active antiretroviral agents. The monitoring of serum drug concentrations may have a role in the treatment of patients infected with the more resistant, drug-experienced isolates. If the therapeutic monitoring of drug concentrations is considered, it should probably be performed relatively early in the therapeutic course to maximally improve the probability of good, long-term suppression.
|
ACKNOWLEDGMENTS |
|---|
This work was supported in part by a grant from GlaxoSmith Kline.
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Clin. Vaccine Immunol. | Clin. Microbiol. Rev. |
|---|---|
| J. Clin. Microbiol. | ALL ASM JOURNALS |
), BID (•), TID (+), and no-drug (
) regimens are followed.
