|Home | About | Journals | Submit | Contact Us | Français|
Atazanavir is an HIV-1 protease inhibitor with high protein binding in human plasma. The objectives were first to determine the in vitro binding characteristics of atazanavir and second to evaluate whether plasma protein binding to albumin and alpha-1 glycoprotein acid (AAG) influences the pharmacokinetics of atazanavir in HIV-infected patients. For the in vitro study, atazanavir protein binding characteristics were determined in AAG- and albumin-containing purified solutions. Atazanavir was found to bind AAG on a high-affinity saturable site (association constant, 4.61 × 105 liters/mol) and albumin on a low-affinity nonsaturable site. For the in vivo study, blood samples from 51 patients included in trial ANRS 107—Puzzle 2 were drawn prior to drug intake at week 6. For 10 patients included in the pharmacokinetic substudy, five additional blood samples were collected during one dosing interval at week 6. Atazanavir concentrations were assayed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). Albumin concentrations, AAG concentrations, and phenotypes were also measured in these patients. Concentrations of atazanavir were modeled using a population approach. A one-compartment model with first-order absorption and elimination best described atazanavir pharmacokinetics. Atazanavir pharmacokinetic parameters and their interindividual variabilities were as follows: absorption rate constant (ka), 0.73 h−1 (139.3%); apparent clearance (CL/F), 13.3 liters/h (26.7%); and apparent volume of distribution (V/F), 79.7 liters (27.0%). Atazanavir CL/F decreased significantly when alanine aminotransferase and/or AAG levels increased (P < 0.01). The ORM1*S phenotype also significantly increased atazanavir V/F (P < 0.05). These in vivo results indicate that atazanavir pharmacokinetics is moderately influenced by its protein binding, especially to AAG, without expected clinical consequences.
Atazanavir is an azapeptide protease inhibitor (PI) with a distinct resistance and tolerance profile approved for use in combination treatment of HIV-1 in the United States and Europe (11, 12, 17). In the United States, one of the recommended antiretroviral regimens is the combination of atazanavir-ritonavir with tenofovir-emtricitabine. Recommended doses of atazanavir are 400 mg once daily taken with food without ritonavir in therapy-naive patients or 300 mg in combination with low-dose ritonavir (100 mg) once daily in antiretroviral-experienced patients or when combined with tenofovir (7). The major advantages of atazanavir are its simplicity of administration and a favorable adverse effects profile, especially on lipid parameters.
Major pharmacokinetic characteristics of atazanavir are variable absorption through the gut, 86% plasma protein binding on albumin and alpha-1 glycoprotein acid (orosomucoid [AAG]), and elimination through biotransformation which involved CYP3A. Although the consequences of CYP3A metabolism on first-pass effect and drug-drug interactions have been largely studied, there is a lack of information on consequences of protein binding on atazanavir pharmacokinetics.
It is now recognized that protein binding is an important modulator of protease inhibitor disposition, and unbound concentration inhibitor is considered as the active moiety which is available to cross cell membranes. Therefore, variations in the concentrations and structures of orosomucoid and albumin under physiological or pathological infections such as HIV infection are likely to influence protease inhibitor pharmacokinetics (3). Albumin concentrations could be significantly reduced in patients with liver disease such as coinfections with hepatitis B virus or C virus. Orosomucoid, also called alpha-1 glycoprotein acid, is a glycoprotein which is controlled by a cluster of three adjacent genes: AGP-A, which codes for the major protein ORM1; and AGP-B and AGP-B′, which code for the protein ORM2 (14). The proteins ORM1 and ORM2 differ by a sequence of 22 amino acids. The protein ORM1 is polymorphic, with three variants (ORM1*F1, ORM1*S, and ORM1*F2), whereas the protein ORM2 is generally monomorphic with one variant (ORM2*A) (8). This polymorphism could be responsible for interindividual variation in the plasma binding of protease inhibitors, which might influence their pharmacokinetic parameters.
The purpose of this work was to determine the in vitro binding characteristics of atazanavir. Then we evaluated whether albumin concentration and/or alpha-1 glycoprotein acid concentration and alpha-1 glycoprotein acid variants are pertinent covariates in the population pharmacokinetics analyses of atazanavir used in combination with ritonavir-tenofovir in HIV-infected patients included in a clinical trial (ANRS 107—Puzzle 2) (22).
(This work was presented in part at the 14th Conference on Retroviruses and Opportunistic Infections, Los Angeles, CA, February 2007.)
Alpha-1 glycoprotein acid and serum albumin solutions were prepared in pH 7.4 phosphate-buffered saline. The concentrations of alpha-1 glycoprotein acid and serum albumin used were those found in normal patients and were 0.7 g/liter and 40 g/liter, respectively. These solutions were spiked with known amounts of atazanavir to yield the following final concentrations: 0, 500, 1,000, 1,500, 2,000, 5,000, 10,000, 15,000 and 30,000 ng/ml. Bound atazanavir and unbound atazanavir were separated by ultrafiltration of 500-μl samples using Centrifree devices (Amicon YM-300 filter system; Millipore Corp., Bedford, MA) at 3,000 × g for 2 h at 30°C. Atazanavir was then measured in the ultrafiltrate according to a method developed for amprenavir (1). The Amicon Centrifree YM-300 filter system (Millipore Corp., Bedford, MA) with a membrane molecular mass cutoff of 30,000 Da was used to ultrafiltrate plasma samples. The driving force for ultrafiltration was provided by centrifugation (Jouan, GR 4.11) at 2,000 × g at 30°C. The duration of centrifugation was 120 min to obtain an ultrafiltrate volume of at least 200 μl from a 500-μl plasma sample. Atazanavir concentrations in ultrafiltrate were measured by high-performance liquid chromatography (HPLC) with separation on a C18 column after liquid-liquid extraction and UV detection at 220 nm. The mobile phase consisted of pH 5.6 phosphate buffer-acetonitrile-methanol (480/110/110 [vol/vol]), and the flow rate was 1.1 ml/min. The retention times for atazanavir and 6,7-dimethyl-2,3-di-(2-pyridyl)-quinoxaline (internal standard from Sigma Aldrich Chemicals) were 15 and 30 min, respectively, in our chromatographic system.
The unbound fraction was the ratio of unbound atazanavir and total atazanavir concentrations. The graph of the bound drug concentration as a function of the unbound drug concentration depicts the protein binding system (24).
ANRS 107—Puzzle 2 was a randomized, open-label, multiple-dose trial evaluating the efficacy and safety of a combination of atazanavir-ritonavir and tenofovir in HIV-infected patients with multiple antiretroviral treatment failures. The study design is detailed elsewhere (22). From week 3 to week 26, all patients received atazanavir-ritonavir (300 mg atazanavir and 100 mg ritonavir once a day [QD]) plus tenofovir disoproxil fumarate (300 mg QD, equivalent to 245 mg of tenofovir disoproxil or 136 mg of tenofovir) and nucleoside reverse transcriptase inhibitors (NRTIs) selected according to the baseline reverse transcriptase genotype of the HIV isolate infecting each patient. Drugs were administered in the morning with a light continental breakfast.
Before inclusion, all patients gave their written informed consent. The protocol was approved by the Institutional Review Board of Saint Antoine Hospital, Paris VI University. HIV-infected patients were eligible for inclusion if they met the following criteria: documented treatment failure with at least two PIs and one nonnucleoside reverse transcriptase inhibitor, an HIV RNA level of >10,000 copies/ml, no change in antiretroviral treatment within the last month before inclusion in this study, and normal liver function.
Blood samples for biochemical and immunovirological measurements were drawn at screening and at regular time intervals after inclusion. Blood samples for the plasma atazanavir concentration assay were collected prior to drug intake in the morning at week 6, 1 month after starting the full antiretroviral treatment with atazanavir plus ritonavir plus tenofovir plus the optimized background treatment, according to the design of the ANRS 107 study (22). For 10 patients included in the pharmacokinetic substudy, additional blood samples were collected after dosing at times 1, 2, 3, 5, 8, and 24 h (23). Plasma samples were kept at −20°C until analysis. The actual times of drug administration and sampling were recorded.
Atazanavir concentrations were measured in patients’ plasma samples by validated liquid chromatography-tandem mass spectrometry (LC-MS/MS) assays (Bristol-Myers Squibb, Saint Nazaire, France). The lower limits of quantification (LOQ) were 1 ng/ml. Day-to-day variabilities for the quality control samples were 5.7%.
Absolute numbers of CD4 lymphocytes, plasma HIV RNA levels, blood chemistry parameters (albumin, total bilirubin, total cholesterol, triglycerides, and alanine and aspartate aminotransferases [ALAT and ASAT, respectively]) were determined according to standard assays (22).
Plasma alpha-1 glycoprotein acid concentrations were measured by nephelometry (BN Prospec, Dade Behring). Alpha-1 glycoprotein acid phenotypes (ORM1*S-ORM2*A, ORM1*F1S-ORM2*A, and ORM1*F1-ORM2*A) were determined on plasma samples by isoelectric focusing according to the method developed by Eap and Baumann, with some modifications (10).
Data at week 6 were analyzed using a population approach with the first-order method (WinNonMix version 2.0.1, Pharsight Corporation, Mountain View, CA) and parameterized with the apparent volume of distribution (V/F), the first order absorption rate constant (ka), and the apparent clearance (CL/F). The statistical model for the observed plasma concentrations of the drug Cij in patient i at time tij was given by the equation Cij = f(tij,θi) + ij, where θi is the pharmacokinetic parameter vector of patient i, ij is the residual error, and f is the pharmacokinetic model.
An exponential random-effect model was chosen to describe interindividual variability, θi = θ exp (ηi), where θ is the population mean vector of the pharmacokinetic parameters and ηi represents the random effect vector. Random effects were assumed to follow a normal distribution with zero mean and variance matrix Ω, which was supposed to be diagonal. Residual variability was modeled using a proportional error model.
Goodness-of-fit plots (observed versus predicted population and individual concentrations and weighted residuals versus predicted concentrations and versus time) were examined for each model. The tested covariates were age, weight, body mass index, creatinine clearance at week 6, alanine aminotransferase (ALAT), and aspartate aminotransferase (ASAT) at baseline, plasma alpha-1 glycoprotein acid and albumin concentrations at week 6, alpha-1 glycoprotein acid phenotypes, coinfections (hepatitis B infection and/or hepatitis C infection), ritonavir trough and average concentrations, and combined NRTIs. Continuous variables were centered on their median. The sex of the patients was not tested as a covariate because there were only two females in the population. The effects of covariates were tested on each parameter (correlation test for continuous variables and analysis of variance [ANOVA] for categorical variables) using Statgraphics version 5.1 (Manugistics, Inc., Rockville, MD). The covariates that were found to have a significant effect (P < 0.05) were then evaluated in the population analysis. The effect of a covariate was assessed by the likelihood ratio test. In the forward inclusion process, a covariate was retained in the model if there was a decrease greater than 3.84 in the objective function (P < 0.05, 1 degree of freedom) and if there was a decrease in the interindividual variability of the associated pharmacokinetic parameter. From the best model including covariates, a backward elimination procedure was then used to test whether all covariates selected should remain in the final model. When deletion of a covariate (P < 0.05) significantly increased the log likelihood (>3.84), that covariate was kept in the model.
In a solution of 0.7 g/liter of alpha-1 glycoprotein acid, atazanavir was found to bind to one high-affinity saturable site at about 31.1%. The association constant was 4.61 × 105 liters/mol, and the number of sites was 0.61. In a solution of 40 g/liter of serum albumin, the mean unbound fraction of atazanavir was found to be 21.3% and was constant up to 15,000 ng/ml of total atazanavir. This suggested that atazanavir binds to one low-affinity nonsaturable site. The number of sites and the affinity constant were not differentiated and the product was estimated as 6,352 liters/mol. When atazanavir total concentrations were above 15,000 ng/ml, the unbound atazanavir fraction increased to 30.9%, the protein binding sites becoming saturated (Fig. (Fig.11).
(i) Patients. Fifty-one patients completed the study at week 6, and 10 patients were included in the pharmacokinetic substudy. The characteristics of these 51 patients are summarized in Table Table1.1. The NRTIs combined with ritonavir-boosted atazanavir plus tenofovir were lamivudine (n = 40), abacavir (n = 30), didanosine (n = 18), zidovudine (n = 15), zalcitabine (n = 5), and stavudine (n = 2). Twenty-three patients were hepatitis virus B and/or C coinfected.
The alpha-1 glycoprotein acid concentrations were 1.0 g/liter (0.6 to 1.5) at week 6 and were elevated in 15 patients (28.8%), accounting for their age and sex.
Five patients had the phenotype ORM1*S-ORM2*A, 19 had the phenotype ORM1*F1-ORM2*A, and 28 were heterozygous and presented the phenotype ORM1*F1S-ORM2*A. The allele ORM1*F2 was not observed in this rather small sample. The allele ORM2*A was monomorphic and observed in all sera of patients. The ORM1 allele frequencies were as follows: 0.635 for ORM1*F1 and 0.365 for OMR1*S. Good agreement was found between observed and expected values, assuming a Hardy-Weinberg equilibrium (P > 0.5). Total plasma alpha-1 glycoprotein acid concentrations were not found to be significantly dependent on alpha-1 glycoprotein acid phenotypes.
A one-compartment model with first-order absorption and elimination was used to describe atazanavir pharmacokinetics. The population parameter estimates, their percent relative standard error of estimation (RSE), and their interindividual variability for this basic model are shown in Table Table22.
From the basic model, we tested the effects of the covariates on the individual estimates of the random effects. Using the estimated individual parameters, significant effects of (i) hepatitis B and/or C infection on ka (P = 0.030), (ii) plasma alpha-1 glycoprotein acid concentration, didanosine, ALAT level, ASAT level, ritonavir trough, and average concentrations on CL/F (P = 0.001, 0.028, 0.037, 0.017, 0.009, and 0.006, respectively), and (iii) plasma alpha-1 glycoprotein acid concentration and the ORM1*S alpha-1 glycoprotein acid phenotype on V/F (P = 0.009; 0.008, respectively) were found.
According to the likelihood ratio test, the final population model had plasma alpha-1 glycoprotein acid (AAG) concentration and ALAT level effects on CL/F (P < 0.01 and P < 0.01, respectively) and ORM1*S alpha-1 glycoprotein acid phenotype (ORM1*S) effects on V/F (P < 0.05). The equations are (i) CL/Fi = [CL/F − βALATCL/F (ALATi/29) − βAAGCL/F (AAG/1.0)] × exp(ηCL/Fi), where 29 IU/liter and 1.0 g/liter are the median ALAT and AAG levels, respectively, and (ii) V/Fi = (V/F × βORM1*SV/F) × exp(ηV/Fi).
The population parameters of this final model and their relative standard errors of estimation are given in Table Table2.2. The goodness-of-fit plots (not shown) were all very satisfactory for the basic and final models.
The apparent volume of distribution was estimated to be 79.7 liters and increased about 2-fold in patients with the ORM1*S phenotype. The absorption rate constant was estimated to be 0.73/h. The apparent clearance was estimated to be 8.1 liters/h when ALAT was 29 IU/liter and the alpha-1 glycoprotein acid level was 1.0 g/liter. The median half-life was 6.9 h (range, 4.0 to 18.2 h). We found an increase of 34% in mean atazanavir area under the curve (AUC) (47,410 ng·h/ml versus 35,432 ng·h/ml; P = 0.0004) and an increase of 12% in mean half-life (7.8 h versus 6.9 h; P = 0.03) in patients who had an elevated ALAT level (≥40 IU/liter), compared with patients with a normal ALAT level. A 36% increase in mean atazanavir AUC (49,644 ng·h/ml versus 36529 ng·h/ml) and a 28% increase in mean half-life (9 h versus 7 h) were observed in patients with elevated alpha-1 glycoprotein acid (according to their age and sex), compared with those with a normal alpha-1 glycoprotein acid level.
The interindividual variabilities of absorption rate constant, apparent volume of distribution, and apparent clearance were 139.3%, 27.0%, and 26.7%, respectively. They were slightly decreased from the basic model by the incorporation of the covariates. Residual variability was 19.9%.
Figure Figure22 shows the model-predicted concentrations and the observed concentrations versus time when ALAT and plasma alpha-1 glycoprotein acid concentrations were 29 IU/liter and 1.0 g/liter, respectively, in patients with or without the alpha-1 glycoprotein acid ORM1*S phenotype.
Atazanavir is a potent and safe HIV PI with a pharmacokinetic profile that allows once-daily oral administration and which can be optimized by adding low-dose ritonavir. All patients included in this study received ritonavir-boosted atazanavir, and a relationship was found between atazanavir clearance and ritonavir concentrations. However, variation in ritonavir concentrations was not found to impact atazanavir pharmacokinetics to a significant extent and therefore was not retained in the final model. It is known that protein binding affects PI disposition (3) and could reduce the antiviral effect of these drugs (13). It was reported that atazanavir binds to both alpha-1 glycoprotein acid and albumin to a similar extent, but binding parameters were unknown (12). Our in vitro data demonstrate that atazanavir has a higher affinity to alpha-1 glycoprotein acid than to albumin. However, alpha-1 glycoprotein acid is present at about 1/40 the concentration of albumin in plasma. Consequently, the binding of atazanavir to alpha-1 glycoprotein acid was saturable above 6000 ng/ml, whereas the saturation of albumin appears above 15,000 ng/ml of atazanavir. A similar result was already found for saquinavir (15). As for many basic drugs, alpha-1 glycoprotein acid has a high affinity but albumin has a higher capacity for binding (16). In patients treated with atazanavir-ritonavir at the recommended daily dose of 300 mg atazanavir and 100 mg ritonavir, trough plasma atazanavir concentrations are around 1,000 ng/ml. Consequently, alpha-1 glycoprotein acid is the major protein involved in atazanavir binding. These findings are essential because the alpha-1 glycoprotein acid concentration can vary considerably as a result of disturbances of homeostasis. It increases during acute or chronic inflammation and infectious disease. HIV-infected patients are therefore likely to exhibit increased concentrations of alpha-1 glycoprotein acid (3, 16, 21). Moreover the alpha-1 glycoprotein acid polymorphism could be responsible for interindividual variation in the plasma binding of protease inhibitors, which might influence their disposition.
The impact of proteins that bind atazanavir on atazanavir pharmacokinetics was evaluated in 51 HIV-infected patients having failed several lines of previous antiretroviral treatment and included in the ANRS 107—Puzzle 2 trial. Unfortunately, the unbound atazanavir fraction could not be measured in these patients as remaining sample volumes were too low.
This is the first population model analysis to explore the possible influence of alpha-1 glycoprotein acid concentrations, alpha-1 glycoprotein acid phenotypes, albumin concentrations, ALAT and ASAT levels, and/or hepatitis B and C coinfections on atazanavir pharmacokinetic parameters.
In this very advanced population, alpha-1 glycoprotein acid concentrations remained elevated in 29% of the patients even though they had received a new treatment for 6 weeks. This could be explained by the weak virological response to the new treatment and the lack of effect of the antiretroviral treatment on the inflammation due to HIV infection (22). The relative frequencies of alpha-1 glycoprotein acid phenotypes found here are close to those previously described in healthy subjects or in HIV-infected patients (5, 8). We did not find ORM1*F2 on this population, but this variant is present only at a low allelic frequency (9, 19). As already described in healthy subjects, alpha-1 glycoprotein acid plasma concentrations were not found to differ between the 3 phenotypes (18).
To determine atazanavir pharmacokinetic parameters and the influence of protein binding, we used a population approach because only 10 patients had a complete pharmacokinetic profile. The pharmacokinetics of atazanavir was described by a one-compartment model with first-order absorption and first-order elimination with random effects on ka, V/F, and CL/F. This structural model was similar to that of Dailly et al. and Colombo et al., although Colombo et al. described a first-order absorption with a lag time (4, 6). Our estimation of atazanavir pharmacokinetic parameters, apparent volume of distribution, absorption rate constant, apparent clearance, and half-life are in the same range as those of previous studies (4, 6, 17, 23). The atazanavir apparent clearance of 9.8 liters/h demonstrates that atazanavir has low extraction ratio with clearance depending on the unbound fraction and intrinsic clearance. The atazanavir apparent volume of distribution of 78 liters (1.2 liters/kg) indicates that plasma protein binding is not a restricting factor to body distribution. Assuming that pharmacologic effect is related to exposure to unbound drug concentrations (AUCu), which after oral administration, depends on the fraction absorbed through the gut wall, the dose and the intrinsic clearance as demonstrated by Benet and Hoener (2), any change in total drug concentrations should not have clinical consequences. Large interindividual variability was found for ka (139.3%) as variable absorption of PIs has already been described.
We found that the apparent clearance of atazanavir decreases with increasing plasma alpha-1 glycoprotein acid concentration. A similar effect was previously reported for lopinavir and indinavir clearance (5). This result suggests that elevated alpha-1 glycoprotein acid concentrations could decrease the protein-unbound fraction of atazanavir, leading to increased total atazanavir concentrations, but without change in unbound atazanavir concentrations linked to drug efficacy (2).
Interestingly, we found that atazanavir apparent volume of distribution decreases with increased with alpha-1 glycoprotein acid concentrations and increases with the ORM1*S phenotype. This suggests that atazanavir may preferentially bind to the ORM1*F1 variant, compared with the ORM1*S variant. Unfortunately, this hypothesis could not be tested in vitro as human plasma samples tested for alpha-1 glycoprotein acid variants were not available. In contrast, it was suggested that lopinavir and indinavir may preferentially bind to the ORM1*S variant (5). There are conflicting data on preferential binding of neutral and basic drugs such as quinidine to ORM1*S and ORM1*F1 (18, 20). In vitro drug-binding studies are warranted to determine the possible different capacities of the ORM1 variants.
Overall, increases in alpha-1 glycoprotein acid lead to a modest 28% increase in atazanavir half-life which should not have any clinical consequence, but could contribute to the interindividual variability of atazanavir pharmacokinetics, although to a lesser extent than absorption or CYP3A activity.
We found that the apparent clearance of atazanavir decreases with increasing ALAT. This was expected as atazanavir, like other PIs, is extensively metabolized by hepatic CYP3A isoenzymes, which are decreased in patients with liver failure and directly affect the intrinsic clearance. However, such a relationship is remarkable as liver dysfunction was mild to moderate in our patients. A 42% increase in AUC and longer terminal half-life (12 versus 6 h) have been reported in volunteers with hepatic impairment (Child-Pugh grade B or C), compared with healthy volunteers (12). In the present study, we found a similar increase in atazanavir AUC and half-life in patients who had an elevated ALAT level. Our data, although limited, could suggest that high atazanavir concentrations are expected with severe liver dysfunction. As previously described, age, body weight, and body mass index did not influence atazanavir pharmacokinetics (4). Most of our patients were males, so no influence of sex on atazanavir pharmacokinetics could be detected. In keeping with our in vitro data, atazanavir pharmacokinetics was not influenced by albumin concentrations; however, in these patients, there was a small variability in albumin concentrations.
In conclusion, this study demonstrates that atazanavir pharmacokinetics is modestly influenced by its protein binding, especially to alpha-1 glycoprotein acid. The effect of alpha-1 glycoprotein acid concentrations or polymorphisms or liver enzyme elevations on the unbound moiety able to cross biologic membranes and exert a pharmacologic effect is unknown as is, therefore, the potential clinical significance.
This study was supported by ANRS (National Agency for AIDS Research), Paris, France.
We thank Gilead Sciences and Bristol-Myers Squibb for helpful discussions and support for carrying out this trial.
We thank the following investigators who participated in the ANRS 107—Puzzle 2 trial: Hôpital Saint Louis, Paris, E. Oksenhendler, O. Carel, J. M. Molina, M. Tourneur, P. Palmer, I. Madelaine-Chambrin, and N. Mezreb; Hôpital de Bicêtre, Kremlin Bicêtre, J. F. Delfraissy, C. Goujard, M. Mole, Y. Quertainmont, N. Idri, and C. Le Tiec; Hôpital Européen Georges Pompidou, Paris, M. Kazatchkine, A. Aouba, N. Bengrait, A. Si-Mohamed, and B. Sabatier; Hôpital Saint-Antoine, Paris, D. Bollens, F. Besse, B. Gaujour, L. Morand-Joubert, A. Charrois, and A. Daguenel-Nguyen; Hôpital Sainte-Marguerite, Marseille, J. A. Gastaut, I. Poizot-Martin, V. Marin-Frixon, C. Tamalet, P. Colson, and C. Penot-Ragon; Hôpital de l'Archet, Nice, J. P. Cassuto, M. Poirée, C. Ceppi, S. Benhamou, and M. C. Rigault; and Hôpital Purpan, Toulouse, P. Massip, L. Cuzin, M. Obadia, J. Izopet, A. Jaafar, and I. Peyranne.
Published ahead of print on 7 December 2009.