Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
00001.tif
(USC Thesis Other)
00001.tif
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer. Hie quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, prim bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize materials (e.g^ maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the bade of the book. Photographs inchided in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order. A Bell & Howell Information Company 300 North ZeeP Road. Ann Arbor. Ml 48106-1346 USA 313/ 761-4700 800/ 521-0600 PHARMACOKINETICS AND PHARMACODYNAMICS OF DIDANOSINE (DDI) IN PEDIATRIC PATIENTS WITH HIV-1 INFECTION BY ANTONIA PERICLE PERICLOU A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Pharmaceutical Sciences) May 1995 Copyright 1995 Antonia Pericle Periclou UM I Number: 9617135 UMI Microform 9(17135 Copyright 1996, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90007 This dissertation, written by ............. under the direction of hear:.... Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillment of re quirements for the degree of DOCTOR OF PHILOSOPHY Dean of Graduate Studies DISSERTATION COMMITTEE Chairperson z DEDICATION I dedicate my Ph.D. dissertation to my parents who taught me the importance of learning and educational excellence and who, throughout my studies, offered me all the support I needed to complete this work. ACKNOWLEDGMENTS I would like to acknowledge my academic advisor, Dr. Vassilios I. Avramis, for giving me the opportunity to work on this project and for his continuous guidance. I would also like to acknowledge my co-workers in the lab, Michelle M. Solorzano, Richard Kwock, Anis Khan, and Partha Nandy for providing an excellent environment to work in and offering me their help whenever I needed it. Finally, I would like to acknowledge the patients who were enrolled on the clinical trials ACTG #176 and ACTG #144, for providing me with the blood specimens that allowed me to complete my pharmacokinetic studies as well as the healthy volunteers Sunita Coutinho and Jonas Ekblom for their blood donation which allowed me to complete the pharmacodynamic potrion of my dissertation. CONTENTS LISTOFTABLES.................................................................................................... V III LIST OF FIGURES.....................................................................................................X ABSTRACT.............................................................................................................X IV CHAPTER PAGE I. INTRODUCTION........................................................................................ 1 1. Structure of HIV-1 1 2. HIV-1 reverse transcriptase (RT) 3 3. HIV-1 binding and infection 4 4. Pathogenesis of HIV-1 infection 8 5. Antiretroviral chemotherapy 10 6. HIV-1 treatment with AZT 12 7. Development of resistance of HIV-1 to AZT 15 8. Antiretroviral treatment of HIV-1 infection with didanosine 17 9. Structure and chemistry of ddl 18 10. Intracellular metabolism and mechanism of action of ddl 21 11. Clinical efficacy of ddl 23 12. Pharmacokinetics of ddl 26 13. jn vivo and £x vivo evaluations of the ddl and AZT combination treatment 28 14. Clinical efficacy of the ddl and AZT combination treatment in adult patients with HIV-1 infection 29 II. RATIONALE............................................................................................. 32 III........................................SPECIFIC AIMS.................................................. 37 IV. MATERIALS AND METHODS...................................... 37 1. Pharmacokinetic and population pharmacokinetic studies 39 2. Theory of population pharmacokinetic analysis 39 3. Methods of population pharmacokinetic analysis 41 a. Standard Two-Stage (STS) method 42 b. NONMEM (or First-Order method) 45 4. Phase I/ll combination study with ddl and AZT (ACTG #176) 50 a. Materials 50 b. Drug formulation and administration 50 c. Patients 51 d. Study design 52 e. Radioimmunoassay (RIA) of ddl 52 f. High performance liquid chromatography (HPLC) of ddl 53 g. Radioimmunoassay (RIA) of AZT 53 5. Pharmacokinetic analysis of ddl and AZT in the ACTG #176 study by the STS method 54 6. NONMEM population pharmacokinetic analysis of ddl and AZT in the ACTG #176 study 55 7. Effect of previous AZT antiretroviral therapy on the pharmacokinetic profiles of ddl 29 8. ACTG #144 phase ll/lll clinical trial 59 v a. Drug formulation and administration 59 b. Patients 60 c. Pharmacokinetic monitoring of patients 60 d. Analytic methods 61 9. NONMEM analysis of ddl in the ACTG #144 clinical trial 61 a. Estimation of the area under the curve (AUC) 63 b. Estimation of intra- and interindividual variabilities (68% confidence intervals) 64 10. Pharmacodynamic studies of ddl 66 a. Materials 66 d. Isolation of peripheral blood mononuclear cells (PBMC) 66 c. Separation of PBMC cells into T-lymphocytes and monocytes 67 d. ddl anabolism in PBMC cells 67 e. HPLC assay and quantitation of nucleotide mono-, di-, and triphosphates 68 f. Cellular anabolism of AZT in Jurkat T-cell lines 69 g. Cellular anabolism of ddl in vitro in T-cell lines 69 V. RESULTS............................................................................................... 72 1. Assays of ddl and AZT in pediatric patient plasma 72 2. Pharmacokinetic analysis of ddl alone and in combination with AZT by STS method 72 3. Population pharmacokinetic analysis of ddl by NONMEM in the ACTG #176 study 88 vi 4. Pharmacokinetic analysis of AZT by the STS method 91 5. Population pharmacokinetic analysis of AZT by NONMEM 104 6. Effect of prior AZT treatment on ddl pharmacokinetics 107 7. Population pharmacokinetics of ddl by NONMEM in the ACTG #144 study 107 8. Cellular anabolism of ddl in PBMC cells 122 9. Cellular anabolism of AZT in Jurkat T-cell lines 122 10. Anabolism of ddl in vitro in Jurkat T-cell lines in the absence and presence of AZT 131 11. Development of a mathematical model to explain the synergistic inhibition of HIV-1 by ddl and AZT 134 V.l. DISCUSSION 144 V.ll. CONCLUSIONS 165 V.lll. REFERENCES 168 vii LIST OF TABLES Tables Table 1. Mean pharmacokinetic parameters of ddl during a combination treatment of ddl and AZT (ACTG #176 clinical trial) Table 2. Statistical comparison between dose-independent PK parameters of ddl obtained during administration of ddl alone and in combination with AZT Table 3. Statistical comparison between AUC values of ddl obtained during administration of ddl alone and in combination with AZT Table 4. Population pharmacokinetic models of ddl in pediatric patients with H IV -1 infection receiving a combination treatment of ddl and AZT Table 5. Mean PK parameters of AZT during a combination treatment with ddl (ACTG # 1 7 6 clinical trial) Table 6. Statistical comparison between dose-independent PK parameters of ACT when AZT was administered alone and in combination with ddl Table 7. Statistical comparison of AUC values of ACT when ACT was administered alone and in combination with ddl Table 8. Population pharmacokinetic parameters of ACT in pediatric patients with H IV -1 infection receiving a combination treatment of ddl and AZT Table 9. Comparison of one- versus two-compartment model with first-order absorption, in the ACTG #144 clinical trial Table 10. Comparison of population pharmacokinetic models of ddl in the ACTG #144 clinical trial Table 11. Anabolism of ddl in human PBMC cells, lymphocytes and monocytes Table 12. Comparative anabolism of AZT in the Jurkat/0 and Jurkat/AZT-10 cell lines Table 13. Anabolism of ddl in Jurkat/0 and Jurkat/AZT-10 cells in the presence and absence of AZT Table 14. Anabolism of ddl in Jurkat/0 and Jurkat/AZT-10 cells following a 2-hour pre-incubation with 1 /iM AZT Table 15. Extent of inhibition of HIV-RT (vin h ib it* /v u n in hibirtd) by various concentration combinations of AZTTP and ddATP Table 16. Effect of dNTP pools on the velocity of the HIV-RT catalyzed reaction LIST OF FIGURES Figures Figure 1. Structure of HIV-1 virus Figure 2. HIV-1 infection cycle Figure 3. Structure of AZT Figure 4. A. Synthesis of viral DNA by reverse transcriptase. B. AZT inhibition of viral DNA synthesis by reverse transcriptase Figure 5. Structure of ddl Figure 6. Acid-catalyzed hydrolysis of ddl via the carboniun ion pathway Figure 7. Metabolism of ddl in human T-cells Figure 8. A typical HPLC chromatogram of ddl Figure 9. Comparison of linear calibration curves of ddl by the HPLC and RIA methods Figure 10. Plasma concentration-time profiles of ddl in pediatric patients after oral administration of four dose levels (ACTG #176) Figure 11. Log-normal frequency distribution of the clearance of ddl in pediatric patients with HIV-1 infection receiving a combination treatment of ddl and AZT Figure 12. Log-normal frequency distribution of the volume of distribution of ddl in pediatric patients with HIV-1 infection receiving a combination treatment of ddl and AZT Figure 13. Log-normal frequency distribution of the elimination rate constant of ddl in pediatric patients with HIV-1 infection receiving a combination treatment of ddl and AZT Figure 14. Log-normal and normal frequency distributions of the absorption rate constant of ddl in pediatric patients with HIV-1 infection receiving a combination treatment of ddl and AZT 81 Figure 15. Average AUC values of ddl per dose level and per PK evaluation, in the ACTG #176 clinical trial 85 Figure 16. Linear relationship between average AUC values of ddl and dose level of drug in pediatric patients with HIV-1 infection 87 Figure 17. Plasma concentration-time profiles of AZT in pediatric patients following oral administration of four dose levels (ACTG #176) 92 Figure 18. Log-normal and normal frequency distributions of the absorption rate constant of AZT in pediatric patients with HIV-1 infection receiving a combination treatment of ddl and AZT 96 Figure 19. Log-normal frequency distribution of the elimination rate constant of AZT in pediatric patients with HIV-1 infection receiving a combination treatment of ddl and AZT 97 Figure 20. Log-normal frequency distribution of the clearance of AZT in pediatric patients with HIV-1 infection receiving a combination treatment of ddl and AZT 98 Figure 21. Log-normal frequency distribution of the volume of distribution of AZT in pediatric patients with HIV-1 infection receiving a combination treatment of ddl and AZT 99 Figure 22. Average AUC values of AZT per dose level and per PK evaluation, in the ACTG #176 clinical trial 101 Figure 23. Linear relationship between average AUC values of AZT and dose level of the drug, in pediatric patients 103 xi Figure 24. Plasma concentration of ddl in patients receiving 50 m g/m 2 or 150 m g/m 2 ddl under fasting conditions 108 Figure 25. Plasma concentration-time profiles of ddl in patients receiving ddl under fasting conditions. A. 50 mg/m2 ddl. B. 150 m g/m 2 ddl. 109 Figure 26. Plasma concentration of ddl in patients receiving 50 m g/m 2 or 150 m g/m 2 ddl under fed conditions 110 Figure 27. Plasma concentration-time profiles of ddl in patients receiving ddl under fed conditions. A. 50 m g/m 2 ddl. B. 150 m g/m 2 ddl. 111 Figure 28. Predicted concentration-time profiles of ddl for a patient of average age (77.7 months) and weight (19.2 kg) in the ACTG #144 clinical trial receiving 50 m g/m 2 ddl under fasting conditions and with food 118 Figure 29. Predicted concentration-time profiles of ddl (50 m g/m 2 ) indicating intra- and interindividual variabilities associated with each predicted concentration, in a patient of average age and weight in the ACTG #144 study, receiving ddl on an empty stomach and with food 119 Figure 30. Predicted concentration-time profiles of ddl (150 m g/m 2 ) indicating intra- and interindividual variabilities associated with each predicted concentration, in a patient of average age and weight in the ACTG #144 study, receiving ddl on an empty stomach and with food 120 Figure 31. A typical HPLC chromatogram showing the separation of nucleotide triphosphates 123 Figure 32. Intracellular ddATP concentration in PBMC cells, lymphocytes, and monocytes following a 1 hour incubation with a radioactive mixture of 1 nM ddl 125 Figure 33. A typical HPLC chromatogram showing separation of AZT 126 xii Figure 34. Comparative anabolism of AZT in the Jurkat/0 and Jurkat/AZT-10 cells Figure 35. Comparative anabolism of ddl in Jurkat/0 and Jurkat/AZT-10 cell lines Figure 36. Mono-, di-, and triphosphate levels of ddl anabolites in the Jurkat/0 and Jurkat/AZT-10 cell lines after a 1-hour incubation with 1 nM ddl, in the absence and presence of 1 nM AZT Figure 37. ddl anabolism in Jurkat/0 and Jurkat/AZT-10 cells following a 2 hour pre-incubation with a 1 mM AZT ABSTRACT Development of resistance to 3’-a2ido-3’-deoxythymidine (AZT) after 3 to 6 months or more of continuous treatment and significant toxicity associated with AZT treatment in a considerable percentage of patients prompted the design of clinical trials to evaluate the use of ddl as an alternative antiretroviral to AZT monotherapy or in combination with AZT. Both drugs have wide bioavailability after oral administration. The pharmacokinetics (PK) of ddl alone and in combination with AZT were studied in pediatric patients with HIV-1 infection, in a phase I/ll (ACTG #176) and a phase ll/lll (ACTG #144) clinical trial to understand the potential drug interaction between these two drugs and the variability associated with the estimation of their PK parameters. The pharmaco-dynamics (PD) of ddl were studied in human peripheral mononuclear blood cells (PBMC) and in human T-cell lines (Jurkat) sensitive and partially resistant to AZT, to further elucidate the mechanism of drug interaction between ddl and AZT. Pharmacokinetic evaluations of ddl and AZT in the ACTG #176 clinical trial were performed using the Standard Two-Stage (STS) method of pharmaco kinetic analysis. Pharmacokinetic profiles of individual patients were analyzed by the computer software PCNONLIN version 4.0. A population pharmaco kinetic analysis was also performed using the computer program NONMEM xiv version III level 1. The NONMEM program provided population estimates of PK parameters by a simultaneous regression of the PK data obtained from all patients. NONMEM analyses of the ddl and AZT PK data investigated potential interactions between ddl and AZT in plasma when these two drugs were administered in combination. NONMEM analyses also provided information as to whether AZT treatment, prior to the initiation of a combination treatment with ddl and AZT, influences the estimation of the PK parameters of ddl. According to the study design, AZT and ddl were administered orally every 6 and 12 hours, respectively, at doses which ranged from 60 to 180 mg/m2 for both drugs. The pharmacokinetics of both ddl and AZT were adequately described by the one-compartment model with first-order absorption. Pharmacokinetic evaluations indicated that both ddl and AZT were characterized by linear kinetics. The PK profiles of ddl alone and in combination with AZT demonstrated a linear increase in the average plasma concentration and median area under the curve (AUC) from 3.04 jxM-hr to 10.68 /xM-hr with increasing ddl doses. The elimination half-life of ddl was independent of the dose level. When ddl and AZT were co-administered on a chronic regimen, STS analysis indicated no pharmacokinetic interaction between them as determined by the values of the elimination half-life, volume of distribution, and clearance. NONMEM analysis of the ddl and AZT PK data provided similar PK parameter estimates as the STS method. Investigation of a possible PK interaction between ddl and AZT in plasma by NONMEM was performed using two different types of analysis. In the first type of population analysis, the regression model used an indicator variable to distinguish between the PK data of ddl obtained after co-administration of ddl with AZT from the data obtained after administration of ddl alone. In the second type of analysis, the regression model provided linear relationships between the AZT plasma concentrations measured at various times and PK parameters. NONMEM analysis of the ddl data showed that the concurrent presence of AZT in plasma or the simultaneous plasma levels of AZT did not alter the pharmacokinetic behavior of ddl in pediatric patients. The NONMEM program was used to determine the population pharmaco kinetics of orally administered ddl and to evaluate the effect of food on the bioavailability (extent of absorption) and rate of absorption (ka) of ddl, in pediatric patients with HIV-1 infection, as part of the ACTG #144 phase ll/lll clinical trial. The effect of variables such as age, weight, height, and serum creatinine levels, on the pharmacokinetic parameters of ddl was also quantified during population parameter estimation. NONMEM analysis indicated that the rate of absorption of ddl decreased in the presence of food, but the extent of absorption remained unchanged. The absorption rate constant was estimated to be 4.92 h'1 (absorption half-life of 8.5 min) and 1.4 h'1 (half-life of 29.7 min) under fasting conditions and with food, respectively. The clearance (Cl) of ddl decreased linearly with increasing age of the patients while the apparent volume of distribution (Vd) decreased linearly with increasing body weight. For a patient of average age (77.7 months) and average weight (19.2 kg) in this study, the clearance was 23.27 L/h/m 2 and the volume of distribution was 35.95 L/m2. The coefficients of variation for ka, Cl, and Vd were 79.7%, 59.6%, and 60.3 % of their respective population means. The intraindividual coefficient of variation was 40.0%. Both AZT and ddl are precursors of active drugs which must be activated intracellularly to their respective mono-, di-, and triphosphate anabolites in a sequential manner. The triphosphate anabolites of AZT and ddl, AZTTP and ddATP, are the active species which inhibit replication of HIV-1. Pharmacodynamic studies with ddl in PBMC cells and in separated lymphocytes and monocytes isolated from the blood of healthy volunteers indicated that the activation of ddl to its mono-, di-, and triphosphates was higher in lymphocytes than in either PBMC cells or monocytes. The intracellular concentrations of ddl, ddAMP, and ddADP were not different in the PBMC cells and lymphocytes. However, the concentration of the triphosphate anabolite ddATP was significantly higher in the lymphocytes than in the PBMC cells. The concentrations of ddl and its mono-, di-, and triphosphates were statistically significantly higher in PBMC and T-cells than in monocytes. The ddATP intracellular concentrations in PBMC, T-cells, and monocytes following a 1 hour incubation with 1 nM ddl were 23.03 ± 5.20 nM (3.16x10'3 ± 0.54x10'3 pmole/106 cells), 32.31 ± 8.76 nM (4.23x1 O'3 ± 1.12x10* pmole/106 ), and 10.85 ± 1.72 nM (3.18x1 O '3 ± 0.52x1 O '3 ), respectively. The anabolism of ddl was also investigated in the human leukemia cell lines Jurkat E6-1 (Jurkat/O) and Jurkat E6-1/A2T-10 (Jurkat/AZT-10). No statistically significant differences were observed in the activation of ddl to its phosphorylated anabolites in these two cell lines, suggesting that cellular resistance to AZT did not confer cross-resistance to ddl. Further examination of the ability of AZT to modulate ddl anabolism in the Jurkat/0 and Jurkat/AZT-10 cells indicated that the concurrent exposure to 1 juM AZT or a 2-hour pre-incubation of the cells with 1 nM AZT did not affect the anabolism of ddl. The pharmacokinetic and pharmacodynamic studies of ddl indicated that the improvement observed in the clinical status of pediatric patients receiving a combination treatment with ddl and AZT could not be attributed to a drug-drug interaction between ddl and AZT in plasma, or due to increased ddATP intracellular levels in the presence of AZT. Therefore, a mathematical model was developed to explain the synergistic inhibition of HIV-RT by ddl and AZT based on Jackson’s equation which describes the rate of DNA synthesis by DNA polymerase using the four natural triphosphate substrates dCTP, dATP, dTTP, and dGTP. This model describes the rate of viral DNA synthesis by xviii HIV-RT, a DNA polymerase-like enzyme with the same substrates, in the presence and absence of one or more Michaelis-Menten competitive inhibitors, such as ddATP and AZTTP. According to this model, the rate of DNA synthesis by RT, in the presence of both AZTTP and ddATP at concentrations half of their IC^, values, w as 44.4% of the uninhibited reaction, suggesting a synergistic inhibition on HIV-RT by AZTTP and ddATP. Hence, these two drugs can be synergistic against HIV-RT, as it has been documented by independent biology and clinical investigative studies. xix I. INTRODUCTION The human immunodeficiency virus (HIV-1) (Figure 1) w as isolated by two groups of investigation, one lead by Montagnier in France and the other by Gallo in U.S.A. HIV-1 is a retrovirus which infects cells of the immune system, primarily CD4+ T-lymphocytes, leading to the development of AIDS, a disease which is characterized by an irreversible breakdown of the cellular immune system in the host (Gallo, 1987). Retroviruses contain their genomic information in the form of RNA and utilize the host cell enzymatic mechanisms to replicate, a process which eventually destroys the host cells (Gallo, 1987, Gallo, 1988). 1.1. Structure of HIV-1 The HIV-1 virus is roughly spherical in shape. The outer surface, or envelope of HIV-1 consists of a lipid bilayer which is studded by a number of proteins, som e of which are of human origin (Greene, 1993). These proteins belong to the class I and class II major histocompatibility complex (MHC) proteins which play important roles in immune response. In addition, the HIV-1 envelope is spanned by protein structures, which play a crucial role in HIV-1 binding and entry into target host cells. Each of these protein structures consists of four protein molecules of the glycoprotein gp120 located on the outside surface of the virus and the same number of the glycoprotein gp41 which is embedded 1 Spike of envelop* glycoprotein (gpi 20) glycoprotein (gp4l) Lipid layer R N A with protein surround (p7/p9) Nucieocapsid (pi7) Reverse transcriptase Ribonucleic protein (P24) Figure 1. Structure of HIV-1 virus (Adapted from Boucher et al., 1994). 2 in the membrane. The layer below the envelope consists of the matrix protein p17, which surrounds the core or inner capsid of the virus (Haseltine, 1988, Greene, 1993). The inner capsid of the virus is made of the protein p24 and it contains the genetic material of the virus. The genetic material of HIV-1 is contained in two strands of RNA, about 9200 nucleotide bases long. Attached to the RNA are molecules of the enzyme reverse transcriptase (RT) which are responsible for the transcription of viral RNA into DNA. The inner capsid of the virus also contains the enzymes integrase and protease. The enzyme protease is a dimer with a symmetrical structure (Mitsuya et al., 1991). This enzyme cleaves the precursor forms of itself, RT, and other core proteins into active forms during the final stage of viral assembly into mature particles (Hirsch, 1990, Greene, 1993). The enzyme integrase is an endonuclease that catalyzes integration of proviral DNA into the cellular genome (Mitsuya et al., 1991). 1.2. HIV-1 reverse transcriptase (RT) The RT of HIV-1 exists as a heterodimer consisting of subunits of molecular weights 51,000 and 66,000 dalton with common N-terminal sequences (Boucher et al., 1994). The 66,000 dalton form of RT contains a 15,000 dalton protein domain with ribonuclease H (RNase H) activity. The RNase H activity of RT is responsible for the degradation of the viral RNA template from the RNA-DNA hybrid to allow formation of double-stranded viral DNA (Mitsuya et al., 1991). In vivo, the HIV-1 RT is error-prone and it produces approximately 5-10 nucleotide errors per genome in every replication cycle (Mitsuya et al., 1991). Unlike human DNA polymerases, the HIV-RT can not correct transcription errors by exonucleolytic activity (Roberts et al., 1988). The high replication error rate of the HIV-1 RT is believed to be responsible for the emergence of drug-resistant HIV-1 variants during the selective pressure induced by antiretroviral chemotherapy. 1.3. HIV-1 binding and infection HIV-1 infects cells primarily through the CD4+ surface antigen. Cells which bear CD4+ are primarily the helper subset of T-lymphocytes. These cells participate in the activation of other components of the immune system such as B-cells and killer T-cells (NK cells), which attack cells infected by the virus (Greene, 1993). Other immune cells may also express CD4+, including 10- 20% of circulating monocytes, 5-10% of B lymphocytes (Weber et al., 1988), dendritic macrophages, Langherhans cells, and som e brain glial cells (Gallo, 1988, Haseltine, 1988, Weber et al., 1988). In vitro studies have shown that HIV-1 can also infect CD8+ T-cells, muscle cells, fibroblastoid cells, and neuronal cells which may not express CD4+ (Weber et al., 1988, Tateno et al., 1989). CD4+ and CD8+ are membrane glycoproteins that normally bind to 4 class II and class I major histocompatibility complex (MHC) antigens, respectively. HIV-1 infection begins with the binding of gp120 of the virus to the CDA* receptor of a host cell (Figure 2) (Boucher et al., 1994). Recent studies have shown that the antigen CD26 may be needed for infection of T-cells by serving as a cofactor for CD4, through its interaction with the V3 loop of the gp120 of the HIV-1 virus (Cohen, 1993, Callebaut et al., 1993). The V3 loop is one of the five hypervariable regions in gp120 portion of the envelope gene. Substitutions in this domain are very important in determining cellular tropism and antigenicity of gp120 (Boucher et al., 1994). CD26 is a 110-KD glycoprotein (Kameoka et al., 1993) with dipeptidyl Peptidase IV (DPP IV) enzymatic activity (Callebaut et al., 1993). Experiments performed by Callebaut et al. indicated that the infectivity of murine NIH 3T3 cells was greatly reduced when these cells were treated with monoclonal antibodies directed against CD26. In addition, the levels of infectious virus were approximately 30 times higher in NIH 3T3 cells transfected both with plasmid vectors expressing human CD4 and CD26 than in murine cells transfected with either plasmid alone (Callebaut et al., 1993). Binding of gp120 to the CD4+ receptor is followed by the fusion of the viral membrane with the cell membrane and the release of the viral capsid (inner 5 The HIV-1 replication cycle and the role o f reverse transcriptase Replication c y c le of A A M tf .W V .tf DNA V 'M r j’ ir.'./l'V M n 0 NA M A A A M A r t.V » /A t integrase host cell DNA rev erse / tra n sc rip ta se I I AAAAAAWVWVW MAAAAAMMAAA i Tw o RNA c o p ies transcription viral R N A transcript viral DNA integrated into host DNA p ro c e s s in g viral mRNA cell nucleus g e n o m ic v iral RNA translation ’ vtral proteins assem w y a protease F igure 2. HIV-1 infection cycle (Adapted from Boucher et al., 1994). 6 core unit made up of proteins and viral RNA) into the cytoplasm. In the cytoplasm, the viral RNA is transcribed into DNA by the enzyme reverse transcriptase (RT), which is available in each infectious virus particle. The ribonuclease H (RNase H) activity of the reverse transcriptase degrades the RNA template after one strand of viral DNA has been made; this allows the synthesis of double-stranded viral DNA (Gallo, 1988, McKinney, 1991). The newly synthesized proviral DNA migrates into the cell nucleus where it integates into the host DNA, in a reaction catalyzed by the viral enzyme integrase. It is believed that the integrase cleaves and removes the terminal b ases from each 3’ end of the proviral DNA. The viral 3’ ends are then joined to the target DNA, while the 5’ ends of the LTR (Long Terminal Repeat) sequences remain unjoined (Fujiwara et al., 1988). The LTR sequences contain many signals that allow the retrovirus to function, such as the promoter and enhancer sequences and polyadenylation signals. The viral genome directs the host cell to copy the DNA of the integrated virus into RNA which, in turn, moves into the cytoplasm. There, it follows two possible pathways: the viral mRNA is translated into proteins, which undergo post-translational processing to become functional and viral origin RNA is attached to a copy of RT (Haseltine, 1988). The viral enzyme protease plays a crucial role in the secondary processing of certain viral proteins (Mitsuya et al., 1991). The viral proteins and two copies of RT-containing RNA are then assembled into mature virus particles, which bud out of the host cell using part of the host membrane. When many viral particles are released, the integrity of the cellular membrane is lost and the host cell is killed. 1.4. Pathogenesis of HIV-1 infection HIV-1 infection causes a slow but progressive degenaration of the immune and central nervous system s (Haseltine, 1988). During the early stages of the infection (acute phase) which last for 3-6 weeks viral replication is mainly constrained within the lymphoid tissue such as the lymph nodes, there is an increased production of antibodies, which neutralize the virus, and activation of NK cells, which destroy infected host cells. However, a large number of infected cells remain in the lymph nodes and in the blood circulation. Over time (latent phase), the virus spreads to organs other than the lymphoid tissue and greatly damages the immune system rendering the infected individuals very susceptible to opportunistic infections (Greene, 1993, Boucher et al., 1994). CD4+ lymphopenia, the profound deficiency in the numbers of circulating T-lymphocytes, is one of the first clinical diagnostic abnormalities noted in patients with HIV-1 infection. Cell death of CD4+ bearing molecules, however, could only partially be attributed to the direct attack of the cells by the virus. A number of possible theories exist. CD4+ cells often die due to the formation of syncytia or multinucleate cells, which are unable to perform the normal cellular functions and eventually undergo cell lysis. The formation of syncytia is a result of the binding of gpl20, present on the surface of an infected cell, to the CD4+ receptor of a healthy cell. It is also very likely that cytotoxic T-cells recognize infected CD4+ lymphocytes or viral envelope pro teins adhered to the CD4+ receptor (MHC) of uninfected cells, and kill them (Eales, 1988, Haseltine, 1988). Another theory suggests that the phenomenon of apoptosis or programmed cell death may play a role in the reduction of CD4+ cells (Greene, 1993, Boucher et al., 1994). HIV-1 is believed to interfere with the transfer of chemical m essages between and within CD4+ cells initiating cell death (apoptosis) when a specific antigen binds to the cell (Boucher et al., 1994). Apoptosis normally occurs in the thymus gland causing depletion of CD4+ thymus cells. The phenomenon of apoptosis has been used by some scientists to explain the rapid progression of AIDS in infants, whose immune systems are still developing (Greene ,1993). The thymus gland is primarily responsible for the development of the immune system in infants. Finally, a group of researchers (Greene, 1993) suggests that the decline in the number of CD4+ cells circulating in the blood could be due to the damage of the lymph nodes by the HIV-1 virus. This theory is based on scientific evidence indicating significant replication of HIV-1 in the scores of the lymph nodes (Greene, 1993). 9 In addition to CD4* lymphopenia, HIV-1 infection can cause abnormalities in the function of CD4+ bearing cells (Eales, 1988). Disruption of the normal function of T-helper cells can have an impact on the production of lymphokines, such as interleukin-2 and y-interferon, both of which are essential in the maturation and differentiation of B-cells, the phagocytic and antigen presenting function of monocytes and macrophages, the activation of NK cells, and the control of cytotoxic T-cells. The HIV-1 virus does not only deteriorate the immune system but it also causes serious neurological diseases such as peripheral neuropathy and dementia (Mitsuya et al., 1987). Neuropsychological deterioration, a s expressed by the loss of developmental milestones and intellectual function, is one of the most destructive manifestations of HIV-1 infection in children (Pizzo et al., 1988). It is possible the HIV-1 virus is carried across the blood- brain barrier by infected macrophages (Gartner et al., 1986, Mitsuya et al.. 1987, Ho et al., 1988). 1.5. Antiretroviral chemotherapy Since its identification as the etiological agent of AIDS, HIV-1 has been the subject of a large amount of research into antiretroviral therapy. Zidovudine or azidothymidine (AZT) was the first licensed drug to be used against HIV-1. In 1987, Fischl et al. demonstrated the ability of AZT to increase the CD4+ cell 10 counts, slow clinical progression, and improve the overall survival of patients with AIDS or AIDS-related complex (Fischl et al., 1987). Volberding et al. showed that low doses of AZT (500 mg per day) were more effective in slowing down the progression of the disease in asymptomatic HIV-infected patients with CD4+ cell count less than 500 cells/mm3 than in patients with AIDS (Volberding et al., 1990). Recent approaches of anti-HIV therapy are based on the intervention of virtually every step in the life cycle of HIV-1 (McKinney, 1991, Mitsuya et al., 1991). Viral entry has been attempted to be inhibited by administration of recombinant soluble CD4+ (Clapham et al., 1989, Mitsuya et al., 1991). Benzodiazepine and dipyridodiazepinone derivatives have been tested for inhibition of the RT of HIV-1, while interferon-a has been used in an attempt to inhibit retroviral release from the cell membrane (Mitsuya et al., 1991). However, most of the studies concerning the development of antiretroviral therapy against HIV-1 have been focused on the targeting of the viral RT. Antiviral drugs are required to be effective not only against T-lymphocytes, the main target of HIV-1, but also against monocytes/m acrophages since studies suggest that these cells may serve as reservoirs for the persistence of HIV-1 in infected individuals, allowing viral replication for long periods of time with minor cytopathic effects (Gartner et al., 1986, Ho et al., 1986, Richmann, 11 1988). The drugs used as inhibitors of viral DNA transcription are usually either pyrimidine or purine nucleoside analogs (antimetabolites). 1.6. HIV-1 treatment with AZT The thymidine analog 3’-azido-3’-deoxythymidine or AZT (zidovudine) was the first drug to enter the AIDS clinical trials (Fischl et al., 1987) (Figure 3). AZT has been administered to both adult and pediatric patients and has been shown to be quite effective in decreasing the mortality and frequency of opportunistic infections. AZT is converted to the monophosphate AZTMP by the enzyme thymidine kinase, with a K„, value of 3.0 nM as compared to a K „, value of 2.9 nM for the phosphorylation of thymidine by the sam e enzyme (Furman et al., 1986). Enzymatic conversion of AZT to AZTMP proceeds with a maximal rate equal to 60% of the rate with thymidine. AZTMP is subsequently converted to the diphosphate AZTDP by thymidylate kinase (dTMPk). The K„, values of thymidylate kinase for AZTMP and dTMP were estimated to be 8.6 nM and 4.1 (iM, respectively, but the phosphorylation rate to AZTDP was only 0.3 % of the phosphorylation rate of dTMP to dTDP. The active anabolite of AZT against HIV-1 is the triphosphate AZTTP. The mechanism of action of AZT against HIV-1 (Figure 4) is believed to involve competition between the AZTTP anabolite and the cellular nucleoside triphosphates for RT sites, and premature termination of viral DNA chain synthesis because the AZT triphosphate lacks the 3’-hydroxyl group required HOCHi Figure 3. Structure of AZT. 13 fttt i - KMo r 3' d M w um i Figure 4.A. Synthesis of viral DNA by reverse transcriptase. B. AZT inhibition of viral DNA synthesis by reverse transcriptase (Adapted from Yarchoan et at., 1989). for chain elongation of DNA (Furman et al., 1986, Yarchoan et al., 1989, Merigan, 1991). Inhibition of HIV-1 RT by AZTTP is favored about 100 times over inhibition of the cellular polymerase a (Hirsch, 1990, Yarchoan et al., 1986). AZT is usually administered orally. The plasma levels of AZT achieved, easily exceed 1 nM which is the concentration of AZT that can efficiently block the cytopathic effect of HIV-1 jn vitro, in T-cell lines, and completely inhibit production of reverse transcriptase without inhibition of the functions of T- and B-cells (Mitsuya et al., 1985). The average bioavailability of AZT ranged from 0.6 to 0.68 (Yarchoan et al., 1986, Balis et al., 1989, Gitterman et al., 1990) in pharmacokinetic studies of this drug. Absorption of AZT into the systemic circulation was rapid. Peak plasma levels were obtained 30 to 90 minutes following oral administration of the drug (Yarchoan et al., 1989). Elimination of AZT was also rapid with an estimated half-life of approximately 1 hour (Yarchoan et al., 1986). A major elimination pathway of AZT is its glucuronidation to the metabolite GAZT (Yarchoan et al.,1986, Singlas et al., 1989, Macleod et al. 1992). Urinary excretion of GAZT represented 60% of the administered AZT dose (Singlas et al., 1989). 1.7. Development of resistance of HIV-1 to AZT Long-term administration of AZT is associated with a number of toxic events 15 primarily myelosuppression (Richmanetal., 1987, Hobbsetal., 1991,Merigan, 1991, Mitsuya et al., 1991) and myopathy (Dalakas, 1993) as well a s the development of resistance (Johnson et al., 1991, McKinney, 1991, Merigan, 1991, Mitsuya et al., 1991) after 6 months or more of therapy (Larder et al., 1989). Larder et al. demonstrated the development of HIV-1 strains with reduced sensitivity to AZT in patients treated with AZT for periods longer than 6 months (Larder et al., 1989). AZT-resistant HIV-1 isolates were found to have common mutations occuring in the RT gene. Three mutations were common in all resistant HIV-1 strains studied and they involved the aminoacid substitutions Asp6 7 -» Asn, Lys70 - * • Arg , and Thr215 - * • Phe or Tyr. The fourth aminoacid substitution, Lys2 1 9 -»Gin, was found in some of the resistant HIV-1 strains. A ll four of these aminoacid substitutions occured in the NH2 -terminal domain of RT. The NH2 -terminal of RT is believed to b e important for nucleotide recognition and polymerase function (Larder et al., 1989). Avramis et al. showed that activation of AZT to the mono-, di-, and triphosphate anabolites was reduced in Jurkat (human leukemia) cell lines with various degrees of resistance to AZT (Avramis et al., 1989). Activation of AZT in the Jurkat E6-1/AZT-100 cell line, developed by continuous exposure to 100 mM AZT, was diminished 28-fold as compared to the less AZT-resistant cell line, Jurkat E6-1/AZT-10. The difference in the activation of AZT to AZTTP was found to be due to the lack of thymidine kinase (TK) activity in the Jurkat E6-1/AZT-100 cell line. It was, therefore, hypothesized that HIV-1 with resistant phenotypes to AZT develop following viral replication in the presence of subinhibitory concentrations of AZTTP (Avramis et al. 1993). Similar observations were documented in PBMC cells from pediatric patients, where it was shown that AZTTP levels were reduced after 8-24 months on AZT monotherapy (Avramis et al., 1993). Hence, AZT is not indicated for use in treating AIDS as a single agent. Another nucleoside analog drug, didanosine (2’,3’-dideoxyinosine), entered clinical trials as an antiretroviral drug against HIV-1 infection for use alone and in combination with AZT. 1.8. Antiretroviral treatment of HIV-1 infection with didanosine Dideoxynucleosides similar in structure to AZT were tested in early phase clinical trials soon after AZT was approved (Yarchoan et al., 1988, Yarchoan et al., 1989, Cooley et al., 1990, Lambert et al., 1990). Both ddl and zalcitabine (ddC) were made widely available to patients in randomized clinical trials. Clinical studies with AZT indicated that patients who could not tolerate AZT or whose clinical condition deteriorated during AZT treatment were offered these alternative therapeutic options. Didanosine has demonstrated activity against HIV-1 in various lymphocyte cultures. Jn vitro experiments indicated that ddl concentrations >10 mM could completely protect ATH8, a tetanus toxoid-specific T-cell line, against the cytopathic effect of HIV-1 virus and inhibit the expression of HIV-1 p24 gag protein (an index of viral infectivity and replication in vitro) in H9 cells (Mitsuya et al., 1986). Phase I clinical trials in adult and pediatric patients, designed to determine the maximally tolerated dose (MTD), showed that ddl was effective in increasing the CD4+ cell count and decreasing the p24 antigen levels (Kahn et al., 1992, Shelton et al., 1992) and was at least as effective as AZT in slowing the progression of HIV-1 disease (Kahn et al., 1992). The available information on ddl, thus far, indicates that ddl could be used as an alternative antiretroviral drug to AZT therapy. Yet, the long-term efficacy of ddl and the criteria for its optimal use still remain to be determined. Didanosine was the second antiretroviral drug approved by FDA for the treatment of HIV-1 infection. 1.9. Structure and chemistry of ddl Didanosine (figure 5) is an analog of the naturally occuring nucleoside inosine, with molecular weight 236.2, and it lacks the 2-/9-hydroxyl group at the 3’- position of the deoxyribose ring (Shelton et al., 1992). Didanosine is very susceptible to acid hydrolysis in the gastric fluid where it degrades to hypoxanthine. The half-life of degradation of ddl in 0.1 N HCI is less than one minute (Hartman et al., 1990, Shelton, 1992). Experiments indicated that acid hydrolysis of ddl proceeds via the carbonium ion pathway (figure 6) (Rawn, 1983, Anderson et al., 1988). The great instability of ddl in acidic conditions is due to the absence of the 2’- and 3’-hydroxyl groups, which function to 18 HN HOCH Figure 5. Structure of ddl. 19 HOHiC HOH,C J (slow) Hypoxan thine HOHX : . r \ HOH,C + H20 1 Resonance-stabilized carbonium ion HOH,C OH dideoxyribose Figure 6. Acid-catalyzed hydrolysis of ddl via the carboniun ion pathway. 20 destabilize the developing carbonium ion through an inductive effect (Anderson et al., 1988). 1.10. Intracellular metabolism and mechanism of action of ddl Didanosine enters target cells (e.g., T-lymphocytes) by passive diffusion where it undergoes sequential intracellular activation by host kinases to form 2’,3’- dideoxyadenosine-triphosphate (ddATP). ddATP is the active anabolite of ddl against the reverse transcriptase of HIV-1 (Figure 7) (Ahluwalia et al., 1987, Back et al., 1992, Nair et al., 1992, Shelton et al., 1992,). Activation of ddl to ddATP proceeds via three phosphorylation steps. Initially, ddl is converted to the monophosphate ddIMP by the enzyme 5’-nucleotidase. The mono phosphate ddIMP can then follow two metabolic pathways. In one pathway, the enzyme adenylosuccinate synthetase catalyzes the conversion of ddIMP to the intermediate 2’,3’-dideoxyadenylosuccinate, which is subsequently converted to ddAMP by the enzyme adenylosuccinate lyase (Ahluwalia et al., 1987, Larder et al. ,1989, Nair et al., 1992). The relative convertion rate of ddIMP to 2’,3’-dideoxyadenylosuccinate was 2% of the rate measured with the physiological substrate IMP (Ahluwalia et al., 1987). The relative velocity (Vm ax ) of the enzymatic conversion of 2’,3’-dideoxyadenylosuccinate was 1.44% that of the natural substrate while the relative efficiency of this reaction was 1.85% that of adenylosuccinate (Nair et al., 1992). The convertions of ddAMP to ddADP and ddADP to ddATP are catalyzed by the enzymes adenylate kinase o II ddl S'-nucleoddase O H ? ddIMP 1. adenviosuccinate syniAetue 2. adenviosuccinate lywe ddAMP jPNM kinase ddADP | PN D kinase ddATP (active moiety against HIV-i) Figure 7. Anabolism of ddl in human T-cells. NHj 2 2 or dAMP kinase and nucleoside diphosphate kinase, respectively. In the second pathway, ddIMP is converted to 2’ ,3’-dideoxyxanthosine-5’ - monophosphate by the enzyme IMP dehydrogenase which, in turn, converted to 2,,3,dideoxyguanosine~5’-monophosphate (ddGMP) by th e enzyme GMP synthetase (Rawn, 1983). In addition to the intracellular formation of ddATP, ddl can be catabolized to hypoxanthine. Hypoxanthine can then be used in the generation of endogenous nucleotides or degraded further to uric acid by xanthine oxidase. The precise mechanism by which ddl suppresses HIV-1 replication is believed to involve competition of ddATP with endogenous dATP for binding sites on the reverse transcriptase and/or the incorporation of ddAMP at the 3’ end of the growing viral DNA strand, thereby terminating DNA replication and interrupting the production of virus (Yarchoan et al., 1989). The K ,,, value of HIV-RT for the naturally occuring substrate dATP was estimated to be 11.2 ± 2.9 juM , whereas the K , value of HIV-RT for the inhibitor ddATP was 0.22 ± 0.12 mM (Hao et al., 1988). 1.11. Clinical efficacy of ddl The effectiveness of a drug as an anti-HIV agent is generally evaluated by quantifying changes in a number of clinical parameters induced as a result of drug administration to HIV-infected patients (McKinney, 1991). HIV-1 infection 23 causes a number of clinical complications which determine the survival and quality of life of the patients. The clinical efficacy of an antiretroviral drug is generally determined by measurements of surrogate markers, such as the number of CD4+ lymphocytes, viral p24 antigen levels, the number of incurrent opportunistic infections (e.g., PCP, candidiasis, CMV retinitis, menigitis), growth parameters (head circumference, weight, height), neuropsychological and neurological function (developmental parameters), and safety and tolerance (e.g., hematologic or neurologic toxicity). Surrogate markers of disease influenced by drug activity, in adults, is usually associated with a positive change in CD4+ count and decreased levels of HIV- 1 p24 antigen. In children, however, demonstration of drug activity may rely more heavily on measurements of neurologic developmental function and change of growth as related to antiretroviral therapy (Pizzo, 1990). Measurements of p24 antigen levels and CD4+ counts are not as reliable indicators of drug activity against HIV-1 in children as in adults. Infants, for example, become symptomatic before the CD4+ count has decreased to less than 500 cells/mm3 while the p24 antigen level is not consistanly elevated (Butler et al., 1991). The clinical efficacy of ddl has been evaluated in adult and pediatric patients. In a phase I/ll study conducted by Butler et al. (Butler et al., 1991), in 43 24 children with symptomatic HIV-1 infection, ddl was administered orally in three divided doses ranging from 60-540 m g/m 2/day for 24 weeks. In this study, an increase in the CD4+ count that met the criteria for a response occured in 50% of the patients who had an initial CD4+ count above 100 cells/mm3 as compared with 10% of the patients with an entry CD4+ count of less than 100 cells/mm3. A significant decline in the median p24 antigen levels from 272 pg/ml at baseline to 77 pg/ml, at 20-24 weeks of ddl administration, was observed in 27 of 32 patients who had detectable p24 levels at entry. In terms of toxicity, pancreatitis was observed in one patient receiving 360 mg/m2/day ddl after 12 weeks of therapy and in another patient receiving 540 mg/m2/day ddl after 14 weeks of therapy. Pancreatitis was also one of the major toxicities observed in patients with HIV-1 infection receiving treatment with ddl (Shelton et al, 1992, Valentine et al., 1993). No peripheral neuropathy was observed in any patient during this study, though this was the most frequent serious toxicity observed in adult phase I clinical trials with ddl (Valentine et al., 1990). A simultaneous evaluation of ddl pharmacokinetics in these patients indicated that the clinical efficacy of ddl depended on the plasma concentrations achieved after oral administration of ddl, and more specifically, on the area under the curve (AUC) of the drug. The increase in the CD4+ count, the decline in the median p24 antigen level, and the degree of improvement in IQ score correlated significantly with higher AUC values. 1.12. Pharmacokinetics of ddl The pharmacokinetics of ddl have been studied during phase I clinical trials in HIV-infected adult and pediatric patients after intravenous and oral administrations of the drug. The disposition of ddl in adult patients could be adequately described either by the one-compartment or the two-compartment open models with first-order absorption (Hartman et al., 1991). The latter one, however, was primarily used to describe data from patients who received the highest evaluated dose of the drug which was 6.4 m g/kg or 320 mg for an adult patient of 50 kg and 448 kg for a patient of 70 kg. Didanosine demonstrated linear pharmacokinetic behavior in adults (Knupp et al., 1991) over the intravenous dose range of 0.4-16.5 mg/kg and the oral dose range of 0.8-10.2 mg/kg, as indicated by measurements of the clearance (Cl) and the steady-state volume of distribution (V,,). In addition, ddl did not accumulate with repeated intravenous administration after two weeks of intravenous treatment. Didanosine is administered orally as a buffered solution or immediately after the ingestion of an antacid because of its susceptibility to degradation by acid hydrolysis in the gastric fluid. Absorption of orally administered ddl occurs rapidly with an estimated absorption half-life of 30 minutes in near neutral gastric pH conditions (Shelton et al., 1992). 2 6 The bioavailability of ddl was characterized by a high degree of interindividual variability (Hartman et al., 1991, Knupp et al., 1991). Variations in gastric environment and mobility and/or in transit time, which are physiological parameters that can be influenced heavily by disease conditions, could be responsible for the wide interindividual variabilty in drug bioavailability (Hartman et al., 1991). The bioavailability of ddl could also be influenced by the presence or absence of food in the Gl tract since the presence of food influences gastric pH and mobility. A preclinical study by Shyu et al. (Shyu et al., 1991), indicated that the relative bioavailability of ddl was reduced by about 50% when ddl was administered with food as compared to the fasting state. To account for this difference, it was speculated that food intake increased gastric acid secretion in the postprandial state and prolonged the residence time of the drug in the stomach, thereby allowing for more extensive drug degradation (Shyu et al., 1991). In a separate study in adult patients with HIV-1 infection, using escalating doses of ddl, the highest administered oral dose (15.2 mg/kg) was associated with a relative bioavailability value of only 24% as opposed to 43% for doses in the range of 0.8-10.2 mg/kg (Knupp et al., 1991). Oral administration of ddl in HIV-infected children under fasting conditions (Balis et al., 1992) resulted in a mean relative bioavailability value of 19% ± 17% which was considerably lower than that determined in adult patients under fasting conditions. It is 27 possible that the lower bioavailability of ddl in children was due to inadequate neutralization of gastric acid since the antacid dose w as only 50% of that used in adult trials. The pharmacokinetic behavior of ddl after oral administration w as characterized by rapid elimination of th e drug from the central compartment (plasma). The plasm a half-life of ddl w as variable with an estimated range of 0.8-1.9 hours (Shelton et al., 1992). In on e study, the presence of food did not appear to affect th e elimination half-life or the time to reach Cm ax, which had a mean value of 0.5 hours (Shyu et al., 1991). Despite the fact th at ddl has a very rapid plasma half-life, its active anabolite against HIV-1, ddATP, has a prolonged intracellular half-life of 24 hours, thereby allowing ddl to be administered twice a day (Hartman et al., 1991, Knupp et al., 1991). This is an advantage over AZT whose active anabolite, AZTTP, has an intracellular half-life of about 3 hours, necessitating administration of AZT four to six times a day (Furman et al., 1986). 1.13. In Vitrp and ex vivo evaluations of ddl and AZT drug combinations The efficacy of th e ddl and AZT combination was investigated in T-cell lines infected with HIV-1 (Avramis et al., 1991) and in the PBMC cells of patients infected with the HIV-1 virus (Cox et al., 1993). The drug combination of ddl 28 and AZT at 1:1 molar ratio, produced a 2-log (100-fold) synergism in the concentration range of 10"8 M as determined by the decrease in DNA synthesis and inhibition of viral p24 antigen production (Avramis et al., 1991). In addition, the ddl and AZT drug combination was synergistic in the PBMC cells isolated from three patients with HIV-1 infection, prior to the beginning of AZT treatment, but this drug combination became antagonistic in the PBMC cells obtained from one of the patients and additive in the cells of another patient after 18-28 months treatment with AZT (Cox et al., 1993). 1.14. Clinical efficacy of combination treatments with ddl and AZT in adult patients with HIV-1 infection The efficacy of a combination treatment of ddl and AZT was evaluated in 69 patients with HIV-1 infection and compared to zidovudine treatment alone (Collier et al., 1993). The combination treatment of ddl and AZT resulted in larger and increased CD4+ counts, which were sustained for longer periods of time, as well a s more frequent decreases in plasma HIV-1 RNA titers. HIV-1 RNA titers provide a measure of viral load using the PCR technique. HIV-RNA titers are becoming increasingly more popular as a measure for assessing a drug’s clinical efficacy because they measure viral load in all individuals, regardless of whether the individuals are p24-positive or p24-negative or whether they have high or low CD4+ counts (Boucher et al., 1994). 29 In a separate study, the efficacy of the ddl and AZT combination treatment was evaluated in a randomized pilot study conducted in 41 adult patients with AIDS or with symptomatic HIV-1 infection and was compared to an alternating regimen of ddl and AZT (Yarchoan et al., 1994). The CD4+ counts in patients on the simultaneous regimen were significantly higher than in the patients on the alternating regimen, during weeks 6-45. In addition, patients on the simultaneous regimen had a significantly greater weight gain and overall greater decrease in HIV-1 p24 antigenemia and frequency of opportunistic infections than patients on the alternating regimen, though the difference was not statistically significant. Hence, the results of this study suggest that ddl and AZT are more effective against HIV-1 when administered simultaneously than when given alone or sequentially. The results of these two clinical studies verified the earlier pre-clinical observations which showed that ddl and AZT were highly synergistic in T-cell lines and PBMC cells infected with HIV-1 (Avramis et al., 1991, Cox et al., 1993). The scientific investigations with ddl prior to the beginning of this project provided som e information on the pharmacokinetics, clinical efficacy, and pharmacology of ddl. However, more studies were needed to thoroughly examine the pharmacokinetics and pharmacodynamics of this drug. The PK 3 0 studies would have to address how differences in the conditions of oral administration of ddl and in the physiological parameters of the patients influence the PK behavior of this drug. The PD studies would have to investigate the mechanism(s) of drug synergism between ddl and AZT and determine whether the concurrent and sequential presence of the two drugs influences their plasma kinetics and/or their activation to their respective anabolites ddATP and AZTTP. 31 II. RATIONALE The inability of AZT to provide an effective treatment for HIV-1 infection and the limitations associated with its long-term use prompted the design of clinical trials to evaluate other drugs as potential antiretroviral therapeutic agents. Earlier studies with ddl, jn vitro and in adult patients, demonstrated that ddl was effective in the treatment of HIV-1 infection when used alone and in combination with AZT. For years, most clinical trials had been conducted in adult patients. However, it becam e apparent that the results obtained in clinical trials with adult patients could not necessarily be extrapolated to pediatric patients. Several facts lead to this consideration. HIV-1 infection in children is usually acquired perinatally and is more accelerated than in adults. In addition, pharmacokinetic studies with a number of drugs indicated differences in many PK parameters between children and adult patients due to developmental changes in body water/fat compartments, plasma protein binding, the kidney’s ability for eliminating processes, and an age-dependency in the activity of hepatic enzymes (Gibaldi et al., 1983). These differences, therefore, suggested the possibility of a different response to the virus and the need for a different or modified anti-HIV treatment in pediatric patients. 3 2 A drug’s pharmacokinetic behavior is influenced both by metabolism and organ development and function, such as renal function. I, therefore, decided to perform an extensive population pharmacokinetic evaluation to estimate the population PK parameter estimates of ddl in the pediatric patient population. These studies would allow me to investigate whether differences in the developmental stages, gender, and organ function of these patients influence PK parameter estimation. Such information could be very useful in improving the design of ddl pediatric dosage regimens because they would be accomodating a child’s individual characteristics. A study conducted by Butler et al. in 43 pediatric patients with HIV-1 infection showed that higher AUC values of ddl significantly correlated with improved clinical status. Based on this study, it became apparent that a higher clinical improvement could be achieved if ddl was administered under conditions which maximized its relative bioavailability. Preliminary PK studies with ddl implied the possibility of reduced bioavailabilty of orally administered ddl in the presence of food (Shyu et al., 1991). Therefore, I proceeded to further investigate whether food lowered the bioavailability of ddl. Earlier successes in the treatment of leukemia in pediatric patients with combinations of antileukemic drugs, which were also purine and pyrimidine analogs, suggested that a similar approach of combination treatments could 3 3 be m ore effective in the treatment of HIV-1 than monotherapies. Several factors lead to the consideration of a concurrent combination treatment with AZT and ddl. Although these two nucleoside analog drugs share a common mechanism of action, the inhibition of HIV-RT, AZT is a pyrimidine analog while ddl is a purine analog, thus they are activated by different kinase enzymes (Ahluwalia et al., 1987). Their respective active triphosphates, AZTTP and ddATP, could enhance the potency of each other when used together, resulting in a synergisitic antiviral effect and a more clinically efficacious treatment. This potentially enhanced efficacy could result in the possibility of using lower d oses of each agent and reduced host toxicities. in vitro data demonstrated a significant (100-fold) synergistic antiviral effect of these two drugs in combination without decreased T-cell viability (Avramis et al., 1992). In addition, Cox et al. indicated that the ddl + AZT drug combination was synergistic in viral isolates from three patients prior to AZT treatment. However, after 18-28 months treatment with AZT, the ddl + AZT drug combination became antagonistic in one isolate, additive in another isolate, and remained synergistic in the third isolate (Cox et al., 1993). Therefore, it appeared that a combination treatment with ddl and AZT could be beneficial to patients with HIV-1 infection, especially if started early, prior to the development of resistance to AZT. 3 4 The concurrent administration of ddl and AZT could have implications on the pharmacokinetic characteristics of either drug and thus, affect the efficacy of the combination treatment. Pharmacokinetic drug-drug interactions between ddl and AZT could affect the absorption, transport, metabolism, and renal elimination of either drug. The transport of AZT into PBMC cells would not be expected to be inhibited by ddl since it has been shown that AZT enters cells by diffusion and does not use the nucleoside transport system (Wientjes et al., 1992). In addition, the renal clearance of ddl and AZT could be affected because both ddl and AZT are excreted in the urine by active tubular secretion (Knupp et al., 1991, Wientjes et al., 1992). The catabolism of ddl to inactive anabolites, such as hypoxanthine, a major metabolic pathway for ddl, could be affected if AZT is a substrate for or inhibits the purine nucleoside phosphorylases, the enzymes which convert ddl into hypoxanthine (Back et al., 1992, Wientjes et al., 1992). Although in vitro studies on the ability of ddl to inhibit the glucuronidation of AZT to GAZT indicated that ddl exerted no inhibitory effect on the metabolism of AZT, jn vivo drug interactions could be more pronounced (Macleod et al., 1992). Thus, taking into consideration the number of possible PK interactions between ddl and AZT, I investigated whether co-administration of these two drugs would affect the PK profile of either drug (Periclou et al, 1994a). The results of these investigations could help determine the impact that a PK interaction between ddl and AZT could have on the clinical efficacy of these drugs. Recent clinical trials in adult patients with HIV-1 infection indicated that the ddl and AZT combination treatment was m ore clinically efficacious than either AZT monotherapy (Collier et al., 1993) or an alternating regimen of ddl and AZT (Yarchoan et al., 1994). In addition, G ao et al. showed that AZT had a better anti-HIV activity in activated cells while ddl preferentially protected resting cells (Gao et al., 1993). It has been hypothesized by Collier et al., that the improved clinical efficacy of the ddl an d AZT simultaneous combination could be attributed to the fact that this combination regimen protects both the resting and dividing cells against HIV-1. T he results from these studies and the observed synergism between ddl and AZT, in vitro (Avramis et al., 1992) and vivo (Cox et al., 1993), prompted m e to design pharmacodymamic (PD) studies to further investigate the activation of ddl in human PBMC cells and T- cell lines sensitive and partially resistant to AZT (Periclou et al, 1994b, Periclou et al., 1994c). Investigation of ddl anabolism in resistant cell lines would determine whether cellular resistance to AZT also confers resistance to ddl. This would be an important finding since many patients who receive ddl have had prior AZT treatment. In these PD studies, ddl would be used alone, simultaneously with AZT, and after pre-incubation of the cells with AZT. The PD studies could provide a biochemical rationale for the synergistic interaction between ddl and AZT (Periclou et al., 1995) and hence, for the improved clinical status of the patients on the combination treatment. 3 6 III. SPECIFIC AIMS SPECIFIC AIM #1 A. To evaluate assays for the quantitation of ddl in the plasma of pediatric patients with HIV-1 infection. B. To investigate the pharmacokinetics of ddl using the Standard Two-Stage (STS) method and by means of the curve-fitting program PCNONLIN. SPECIFIC AIM #2 A. To investigate whether the population pharmacokinetic parameters of ddl obtained during ddl monotherapy differ from the estimates obtained during a combination treatment with AZT, in pediatric patients with HIV-1 infection. B. To investigate whether the population pharmacokinetic parameters of AZT obtained during AZT monotherapy differ from the estimates obtained during a combination treatment with ddl, in pediatric patients with HIV-1 infection. C. To investigate whether there are differences in the bioavailabilty and other PK parameters of ddl, under fasting and fed conditions, in pediatric patients with HIV-1 infection. 3 7 SPECIFIC AIM # 3 To investigate the effect of AZT monotherapy prior to the beginning of a combination treatment with ddl and AZT, in pediatric patients with HIV-1 infection. SPECIFIC AIM # 4 To determine whether ddl is metabolized to its active anabolite, ddATP, in peripheral blood mononuclear cells (PBMC) and whether there is preferential anabolism of ddl between monocytes and lymphocytes. B. To determine whether AZT increases activation of ddl to ddATP in vitro in T-cell lines, sensitive and partially resistant to AZT. C. To develop a theoretical biochemical rationale that could explain the synergistic interaction between AZT and ddl. 3 8 IV. MATERIALS AND METHODS In order to fulfill these specific aims, I participated in the clinical investigation of two national studies sponsored by the AIDS clinical trials group (ACTG). These studies were the ACTG #176 and ACTG #144 and were conducted in pediatric patients with HIV-1 infection. IV.1. Individual and population pharmacokinetic studies The pharmacokinetics (individual and population) of ddl were evaluated during a combination antiretroviral therapy of ddl and AZT, as a nested study, during the phase I clinical trial (ACTG #176) in pediatric patients with HIV-1 infection. In addition, the population pharmacokinetics of ddl were evaluated as part of the ACTG #144 phase ll/lll clinical trial which was designed to evaluate the efficacy, safety, and tolerance of two oral doses of ddl, 50 m g/m 2 and 150 mg/m 2, administered under fasting conditions and with food. IV.2. Theory of population pharmacokinetic analysis Population pharmacokinetic modeling describes the pharmacokinetic behavior of a drug in a large group of subjects in terms of the average values of the pharmacokinetic parameters (fixed effects) and their variances (random effects) (Whiting et al., 1986). A primary concern of population pharmacokinetic analysis is the estimation of interindividual variability and the identification of the 3 9 variability sources (covariates) by detecting relationships between kinetic parameters and subject characteristics such as age, gender, weight, and renal function (Whiting et al., 1986, Peck et al., 1986, Ludden, 1988). Such an analysis can help make individual treatment more effective, by m eans of Bayes Theorem for conditional probabilities. Bayes Theorem combines population parameters with measurements of drug concentrations from an initial trial regimen in a given patient, to predict the m ost probable parameter values for this patient (Ludden, 1988). Analysis of pharmacokinetic data for a drug requires the development of a mathematical model, which is described both by a structural (pharmacokinetic) model and a variance model (Steimer et al., 1985). The mathematical model relates the actual data, y,j, to the model parameters for an individual i by the expression: y« = f(ei.*ij) + [Eq-i] where yy is the jth observation for individual i (e.g., the j* * concentration value). 6, is the vector of parameter values for the i th individual. f(0jlxij) is the set of values (e.g., concentrations) predicted by the pharmacokinetic model with parameter values 6,, as a function of the independent variable x^ (e.g., time). 4 0 €jj is the measurement error which accounts for the deviations observed between the model predictions and the actual observations. Each individual parameter vector 6, varies from person to person according to a random distribution (Racine-Poon et al., 1990) which is characterized by 6, the vector of mean values of the parameters in the population. The equation relating 6, to 6 is given by: 0( = 0 + rj, [Eq.2] where is the dispersion about the mean population parameters of the model for the ith patient. In the cases when a pharmacokinetic parameter is also related to individual characteristics, such as age and weight, 0, is determined by the expression: ei = 9(Z|,0) + 0, [Eq.3] where z, is the covariate vector. IV.3. Methods of population pharmacokinetic analysis A number of methods are available for population pharmacokinetic analysis. 41 The conventional a n d simple m ethod for population pharmacokinetics is the Standard Two-Stage (STS) method. The Nonlinear Mixed Effects Model (NONMEM) and the Nonparametric Maximum Likelihood Approach (NPML) are more complicated m ethods than ST S and offer a greater flexibility in handling clinical data, by accomodating sp a rse data collected during routine patient care. IV.3.a. Standard Two-Stage (STS) pharmacokinetic method The STS method is a relatively sim ple method which determines population values, first by estimating the individual parameters, and th en by combining them to yield the population parameter estimates (Ludden, 1988, Steimer e t al., 1985, Grasela et al., 1986). In the first stage, the individual param eter estimates 0j are determined by fitting a pharmacokinetic m °d e l to the d a ta (Steimer et al., 1985, Grasela et al., 1986). This can be d o n e by minimizing the ordinary least squares objective function with respect to 6, as shown below: m(i) 0(0) = % [yi j -f(0l,xij )]2 [Eq.4] j = 1 where m(i) is the num b ©r of observations for th e i* * subject. 42 In addition to the ordinary least squares method, individual parameter estimation can be performed by weighted least squares or by maximum likelihood (ML) estimation. Individual parameter estimation may require the definition of a variance model (for ML estimation). A general variance model is given by the equation: Var(e) = a2 + B2[f(0j)xij)]2 [Eq.5] where a represents a constant background variance and the second term of the equation expresses a functional dependence of the variance on the magnitude of the observation. In the second stage, the population values are determined and expressed as the mean population vector estimate, 0, and the variance-covariance matrix estimate, n. For normally distributed parameters, 0 and n are estimated by equations 6 and 7, respectively (Steimer et al., 1985): N 0 = (1/N) z 0, [Eq.6] i=1 N n = (1/N) z(0r 0)(0,-0)T [Eq.7] i=1 4 3 where 6, and 8 are as defined previously, the superscript T represents the transpose of a matrix, and N is the number of individuals. The estimates of 6 and n, for log-normally distributed parameters, are obtained by the following equations: N 0 = (1/N) E ln(6j) [Eq.8] i = 1 N n = (1/N) E (In0, - In6)(ln8j - ln0)T [Eq.9] i = 1 The vector 0 in equation 8 is equal to the geometric mean. The STS method may yield population parameter estimates which are biased because the estimated parameter value for each individual, after a single drug dose or achievement of steady state, is taken to represent the true parameter value for that individual (Ludden, 1988). In addition, since each 0, is assumed to be estimated with equal precision, each individual parameter contributes equally to the calculated mean covariance. However, the individual parameter estimates may not precisely represent the true individual parameter values due to intra-subject variation. The STS method is more accurate if there is a large number of observations per subject. This method should be used for obtaining initial estimates for the more complex methods NONMEM and NPML. 44 IV.3.b. NONMEM population pharmacokinetic modeling (or First-Order method) NONMEM population PK modeling was developed by S.L. Beal and L.B. Sheiner in 1979. A major advantage of this method over the STS analysis is the fact that it can yield population pharmacokinetic parameter estimates from data obtained during routine care of the patients, even when only one data point per individual is available (Ludden, 1988, Steimer et al., 1985, Beal, 1984). With NONMEM, data from all patients are considered as one set and population estimates are obtained using a mixed (fixed and random) effects regression model. The general form of the mixed effects regression model (Beal, 1984) is given by the equation: Y ij = F(e,x,,i?„e,) [Eq-10] In equation 10: yy represents the value of the j* observation for individual i. F is a nydimensional vector-valued function. 8 is the vector of population model parameters. X jj is the vector of all values of fixed effects associated with a particular observation y .y . rjj is the vector of all values of random interindividual effects associated with the I th individual or the deviation of an individuals PK parameter estimate from the population estimate for that parameter. 4 5 e,j is the vector of all values of random intraindividual effects associated with y ,| or th e deviation of an observation from its model-predicted value. In other words, equation 10 is a relationship between the experimental observations a n d the model-predicted values (drug concentrations) which takes into account inter- and intra-subject variability. In this model, ry , and e( J are assum ed to be normally distributed with expected values E(r},)=0 and E(e,j)=0, respectively. Parameter estimation by NONMEM modeling proceeds with the linearization of the general model (Beal, 1984) given in equation 10 in terms of its random effects. Linearization of equation 10 gives rise to the linear random-effects model described by equation 11: F(0.xij,f7i.€ij) = f||(e,Xjj) + Gij(0,x|j)r7i + Hij(e,xij)cij [Eq.11] where fy is the function (structural PK model) which predicts plasma drug concentrations a t each time point 0 of every patient (i), and Gy and H ,j are both matrix-valued functions with row dimension equal to n,j. A special case of the linear random effects model is the first-order (FO) model which is the form of the mixed effects regression model used in population 4 6 parameter estimation by the NONMEM program. The FO model is the result of the linearization of equation 10 using a first-order Taylor series expansion in the random effects r > , and en , evaluated at their expected values (i.e., zero). y,j = Ffe.Xy.O.O) + aF/arjite.Xjj.O.O)*?, + 3F/acl] (6,xijI0 > 0)€l j [Eq.12] In equation 12, aF/arjiO.Xy.O.O) and aF/a«y(©,x,j,0,0) are matrices of the partial first derivatives of F with respect to and ex ] , respectively, evaluated at their expected values (zero). The vector y^ as defined in equation 12 is used in the computation of the first two statistical moments of the data, namely the expectation of y(j, Efyy), and its variance, Varfyy). The expectation of yy, E(y,j), is the term F(0,Xij,O,O) in equation 12. If we define the matrices of partial derivatives of F with respect to rj, and £|j in equation 12 by G(1 ) and H(1), respectively, then the variance of y,j is given by: Varfy) = G(1 ,T nG< 1 ) +diag(Hijs i j T ) [Eq.13] where G< 1 > is as defined above, G< 1 ) T is the transpose matrix of G< 1 ). and n is the variance-covariance matrix of the random interindividual effects r? j (Beal, 47 1984, and Sheiner, 1984). The population parameter estimates are obtained by minimizing an extended least squares objective function defined in terms of y,j, E(y,j), and Var(y,j) with respect to the population parameters. NONMEM modeling provides not only estimates of PK parameters in the population (population parameters) but also estimates of inter- and intraindividual variabilities. Interindividual variability refers to the deviation of an individual’s PK parameter estimate from the population estimate of that parameter. Intraindividual variability represents the deviation between an individual’s observations and the corresponding model predictions. Interindividual and intraindividual variabilities can be modeled as additive, multiplicative or exponential. An additive error model for clearance, for example, would be written by the expression: Cl; = Cl + r > , [Eq.14] where Cl, is the estimated clearance for the i * * 1 individual, Cl is the mean or population clearance, and rj, is the deviation of the individual’s clearance from the population clearance. A multiplicative error model or constant coefficient of variation model would be written as: 48 Cl, = Cl(1 + r> ,) [Eq.15] whereas an exponential error model would be written as Cl, = Cl x EXP(f),) [Eq.16] Although the NONMEM program appears to perform very well in estimating population pharmacokinetic parameters, it is limited by the fact that it relies on approximations (linearization of the model), it assum es a normal or lognormal population parameter distribution, and it does not provide an estimate of the whole probability distribution of a parameter. A major concern of population pharmacokinetic analysis is the reduction of interindividual variability observed in the estimation of population parameters, in order to design individual dosage regimens which produce drug levels within the therapeutic range. The phrase "reduction of variability" refers to the identification of individual patient characteristics (called covariates) that are either demographic variables (gender, age, weight, e.t.c.) or biological variables (body weight, creatinine clearance, e.t.c.) which contribute to the interindividual variability associated with the estimation of population PK parameters (Beal, 1984, Steimer et al., 1985). Part of interindividual variability can be explained by introducing relationships between PK parameters and these covariates. The NONMEM program can accomodate covariates in the 4 9 estimation of population parameters. Covariates are entered as explanatory variables in mathematical relationships which explicitly relate the covariates to pharmacokinetic parameters. The drug clearance, for example, may be related to serum creatinine (S ^ ,,) and body weight (Steimer et al., 1985) by the expression: Cldrua = 6, + e2(Sawt) + 03(weight) [Eq.17] Then, the two first moments are determined by E(0/Zj) and Var(0/Zj). IV.4. Phase I/ll combination study with ddl and AZT (ACTG #176) IV.4.a Materials The AZT radioimmunoassay (RIA) kit was purchased from INCSTAR Corporation, Stillwater, Minnesota. Didanosine and the RIA kit for ddl were purchased from SIGMA Co., St. Louis, Misouri. 0 6 mG was used as an internal standard for the HPLC assay of ddl and was donated by NCI. A ll other chemicals and solvents were of analytical HPLC grade. IV.4.b Drug formulation and administration AZT was manufactured and supplied by Burroughs Wellcome and distributed by NCI. AZT was administered orally either as a strawberry flavored oral solution with a concentration 10 mg/ml or as 100 mg capsules, every 6 hours, on an empty stomach. Didanosine was manufactured by Ben Venue 5 0 Laboratories and distributed by NCI. It was supplied as a sterile freeze-dried powder reconstituted with 16.5 ml of 0.9% Sodium Chloride, USP, to yield a solution of concentration 15 mg/ml. Didanosine was administered orally every 12 hours on an empty stomach, two minutes after taking a dose of antacid, immediately preceeding the morning and evening doses of AZT. IV.4.C. Patients Twenty-two pediatric patients (21 evaluable) with HIV-1 infection were entered by the Childrens Hospital Los Angeles AIDS Program, as a sub-set of 68 patients, into a phase I/ll clinical trial (ACTG #176), which evaluated the efficacy of the combination of AZT and ddl in pediatric HIV-1 infection. The patients were assigned into two arms, A and B. Patients in group A had not received any prior antiretroviral therapy with AZT or ddl but they could have received antiretroviral therapy with agents other than ddl and AZT, which was discontinued at least 14 days prior to their entry on this study. Patients in group B had demonstrated hematologic toxicity while receiving AZT prior to their entry on this study. Patients in both groups were between the ages 3 months and 21 years of age with class P-2 symptomatic HIV-1 infection or asymptomatic patients who had a total CD4+ cell count less than 500 per mm3 (class P-1b). A ll patients entered on the study were free of opportunistic infections at the time of entry, had a total WBC count > 1500/mm3, platelet count > 75,000/mm3, hemoglobin > 9.5 g/dL, creatinine < 2 times the upper 51 limit of normal value, total bilirubin or liver transaminases < 1 0 times the upper limit of normal, and no history of acute or chronic pancreatitis. lV.4.d. Study design This w as an escalating dose study for AZT and ddl. The patients in group A were assigned into one of eight dose levels in which the AZT dose ranged from 90 mg/m2 q6h to 180 mg/m 2 q6h and the ddl doses ranged from 60 m g/m 2 q12h to 135 mg/m2 q12h. The patients in group B received a fixed AZT d o se of 60 m g/m 2 q6h and a ddl dose which ranged from 90 m g/m 2 ql2h to 180 m g/m 2 q12h. The AZT/ddl combination study began with a single dose of AZT on day 1, a single dose of ddl on day 2, and single doses of AZT and ddl on day 3. The regular schedule of AZT (q6h) and ddl (q12h) began on day 4 and continued for 24 weeks. Pharmacokinetic blood samples were obtained immediately prior to drug administration and at 30,90, and 180 minutes after administration of the drug(s), on days 1, 2, and 3 and weeks 4 and 12 or as indicated. IV.4.e. Radioimmunoassay of ddl The concentration of ddl was measured in plasma by a ddl-3H radioimmunoassay (RIA) developed by Sigma Co., St. Louis, Misouri. The ddl- 3 H RIA is a competitive binding immunoassay based on the competition of labeled and unlabeled ddl for a limited number of binding sites in the rabbit 5 2 antiserum to ddl. In these assays, most of the plasma samples had to be diluted 100-fold in order to obtain drug concentrations within the range of the standard curve (0.3-10 ng/ml) recommended by the manufacturer. A standard curve was generated every time plasma specimens were analyzed for ddl. The plasma levels of ddl were quantitated based on the standard curve obtained on the day of analysis. The standard curve of ddl from many assays w as linear with an average R2 value of 0.976. IV.4.f. High Performance Liquid Chromatorgaphy (HPLC) assay of ddl An HPLC assay for the analysis of ddl w as developed in order to assess the validity of the RIA. The analysis of ddl was accomplished using a Bondapak micro C18 reverse phase column (Waters Associates, Millford Mass.) using a W aters Associates HPLC system. The sample was eluted with 84 parts of 95.24 mM ammonium acetate solution, pH 6.5, and 16 parts of 50% methanol, under isocratic conditions, at a flow rate of 0.8 ml/min. The compound O6- methylguanine (06mG), a substituted purine, was used as an internal standard. A known concentration of 0 6mG was added to plasma containing ddl and an aliquot of the intact sample w as injected directly into the HPLC without prior extraction. IV.4.g. Radioimmunoassay of AZT The plasma concentration of AZT was measured by a radioimmunoassay 5 3 developed by INCSTAR Corporation, Stillwater, Minnesota. This RIA uses iodine-125 labeled AZT. The standard curve of AZT was linear with a mean R2 value of 0.984. IV.5. Pharmacokinetic analysis of ddl and AZT by STS method in the ACTG #176 study Pharmacokinetic analysis and estimation of PK parameters were performed by nonlinear regression using the computer software PCNONLIN version 4.0 on an IBM PS/2 Model 90-486 computer. This version of PCNONLIN includes a curve-fiting initial estimation program. The one-compartment open model with first-order absorption provided best-fits to estimate the PK parameters of ddl and AZT. The data sets were assigned into four groups according to the dose of ddl or AZT received. Two kinetic evaluations were performed. In the first evaluation, the plasma concentrations of each time point were averaged within a given dose level, and the mean concentration values were fitted by the one-compartment model with first-order absorption using the PCNONLIN program. This method provided PK profiles (concentration vs time plots) for these drugs at each dose level as reported elsewhere (Balis et al., 1992, Knupp et al., 1991). 5 4 In the second PK evaluation, the concentration-time profiles of ddl and AZT of each individual patient were analyzed by the PCNONLIN program in order to obtain estimates of the PK parameters of ddl and AZT for each patient. Averages of these estimates were then derived per dose level. This method was used to produce average PK parameter estimates. In order to estimate total body clearance (TBCI) and volume of distribution, previously reported values of fraction absorbed (F) for each drug were used i.e., 20% for ddl and 60% for AZT (Yarchoan et al., 1986, Balis et al., 1992). IV.6. NONMEM population pharmacokinetic analysis of ddl and AZT in the ACTG #176 study A population kinetic analysis was performed utilizing the computer program NONMEM version III level 1, developed by the NONMEM Project Group of the University of California, San Fransisco. NONMEM analysis of ddl and AZT was accomplished using the PREDPP and NM-TRAN packages of the NONMEM program. All pharmacokinetic profiles were regarded to be obtained from separate individuals, although more than one PK evaluation were performed for most individuals. This was a necessary step in order to stabilize the NONMEM program run, since NONMEM analyses tend to perform better with increasing number of individuals. The one- compartment open model with first-order absorption (ADVAN2.TRANS2 or 5 5 ADVAN2.TRANS1 of PREDPP) was used to fit the data as indicated by the STS analysis of the data. The structural PK model f,j of ADVAN2, TRANS2 subroutines was, therefore, described by the equation: FXoka j (e -iaw n . e'(ka'i)t) [Eq.18] (V d.ik.,1 - Cl,) where kai is the absorption rate constant for individual i, F is drug bio availability, Cl, is systemic clearance, and Vd, is the apparent volume of distribution. The average values of the PK parameters Cl and Vd (for TRANS2) or Vd and ke, (TRANS'!) estimated by the STS method, were used as initial estimates in the NONMEM analysis of the data. Both intra- and interindividual variabilities were described by the "constant coefficient of variation" (CCV) model because the frequency distributions of the individual parameters were skewed. Therefore, interindividual variabilities were described by the equations: k.,i = ka (1 + rjka,,) [Eq.19] Cl, = Cl (1 + rja ,) [Eq.20] Vd ii = Vd (1 + r,vdi) [Eq.21] Intraindividual variability was also described by the CCV model as shown in equation 22: 5 6 Observed C,j = predicted C( j (1 + e,j) [Eq.22] Since ddl and AZT were administered in combination, the structural PK model was modified to account for any possible influences of plasma AZT on the estimates of ddl PK parameters. Two different types of analysis were attempted (Williams et al., 1992). In the first analysis, the regression model expressed plasma AZT as an indicator variable which was linearly related to clearance: Clj = e, + e 2 x AZTD | [Eq.23] where Cl, is the predicted clearance of ddl for the I th individual, e, is the clearance of ddl in the absence of AZT. e 2 is the increase or decrease in ddl clearance in the presence of AZT in plasma. AZTD , is an indicator variable that took the value of 1 when AZT was present in plasma and the value of 0 when there was no AZT in plasma. In the second type of analysis, the clearance of ddl was expressed as a linear function of the AZT plasma levels, which were measured by the RIA assay: Cl = e, + e2 x AZTC [Eq.24] where e , was the clearance of ddl when the plasma concentration of AZT was 0, AZTC was the simultaneous plasma concentration of AZT, and e 2 was the 57 proportionality constant relating changes in ddl clearance to AZT plasma concentration. Improvement of the fit of the data, as a result of incorporation of covariates in the regression model, was determined based on the value of the minimum objective function (Likelihood Ratio Test), the magnitude of the standard errors of the parameter estimates, the plots of weighted residuals against the predicted concentrations, and decreases in the estimates of interindividual variabilities. According to statistical theory (Likelihood Ratio Test), the difference in the values of the objective function (Log Likelihood difference) of a reduced model (n-1 parameters), lr, and a full model (all n parameters present), lf, can be approximated by the chi-square (x2 ) distribution with q degrees of freedom, where q is the number of parameters whose values are fixed in the reduced model (NONMEM Project Group, Guide V, Antal et al., 1989). A covariate was considered to improve the fit of the data if the difference C2 = l r - 1 , was statistically significant at the P<0.01 level. The more conservative values P < 0.01 or P < 0.005 are often chosen by many investigators over the value of P < 0.05, as the statistical criteria for evaluating models, because of the asymptotic nature of the x2 -test (Grevel et al., 1988, Maitre et al., 1987). 5 8 IV.7. Effect of prior antiretroviral treatment with AZT on the pharmacokinetic profiles of ddl. The effect of previous antiretroviral therapy with AZT on the population pharmacokinetic parameters of ddl was investigated using the NONMEM program. Previous AZT treatment was modeled as an indicator variable that was assigned the value 1 for patients treated with AZT prior to their entry onto the ACTG #176 study and the value of 0 for patients not treated with AZT prior to their entry onto this study. The NONMEM model which was used to evaluate the effect of previous treatment with AZT on the pharmacokinetics of ddl was: Cl = e, + e 2 x AZTp [Eq.25] where e , was the clearance of ddl in patients without prior treatement with AZT, AZTp was the indicator variable, and e2 was the increase or decrease in ddl clearance in patients who had been on AZT monotherapy prior to their entry on the ddl + AZT combination study. IV.8. The ACTG #144 phase ll/ill clinical trial. IV.8.a. Drug formulation and administration Didanosine was supplied by Bristol-Myers Squibb in a lyophilized form. It was reconstituted to oral dosing solutions mixed with the appropriate amount of antacid (Maalox TC™, Extra Strength Maalox Plus™ liquid or Mylanta Double 5 9 Strength™) to achieve ddl-antacid concentrations that corresponded to the 50 m g/m 2 and 150 mg/m2 dose levels, ddl was administered every 12 hours on an empty stomach (fasting state) and in the presence of food (fed state). IV.8.b. Patients A minimum of 300 pediatric patients were enrolled on the ACTG #144 phase ll/lll multicenter, randomized, double-blinded outpatient study. The objective of this study was the evaluation of the efficacy, safety, and tolerance of two doses of ddl, in children with symptomatic HIV-1 infection, who were intolerant to zidovudine (ZDV), who had demonstrated disease progression after 6 months of ZDV treatment, or both. The patients were between the ag es of 3 months and 18 years. At the time of entry, the patients had a serum creatinine concentration < 2.0 mg/dl, hemoglobin > 7.0 g/dl, a platelet count > 50,000 mm3, and pancreatic amylase and lipase < 2 times the upper limit of normal. Patients had to be free of opportunistic infections (Ol) or successfully treated for Ol and had remained stable and free of acute bacterial infections. In addition, patients were permitted to receive immunoglobulin therapy and prophylaxis treatment for Pneumocystis carinii pneumonia (PCP). IV.8.C. Pharmacokinetic monitoring of patients Intensive and limited PK evaluations were performed on 92 patients (90 evaluable) receiving ddl on an empty stomach, at least 2 hours after a meal, 6 0 and with food, ten minutes after the child started receiving a balanced meal appropriate for his/her age. The intensive PK evaluation involved blood sampling at the pre-treatment period and at 15,30,60,90,120,180, and 240 minutes after the administration of ddl. The limited PK evaluation required blood samples to be drawn at the pretreatment period and at random but precisely known times after drug administration. IV.8.d. Analytic methods The plasma samples were initially analyzed for ddl using an HPLC method developed by Bristol-Myers Squibb. The majority of the samples, however, were assayed using a ddl3 -H radioimmunoassay (RIA) developed by Sigma Co., St. Louis, Missouri, in a number of institutions participating in the ACTG #144 study. IV.9. NONMEM analysis of ddl in the ACTG #144 clinical trial The pharmacokinetic data of this study were fitted by two structural PK models: the one- and the two-compartment models with first-order absorption. Two different parametrizations of each structural PK model were examined. The one-compartment model was expressed either in terms of the PK parameters ka, Cl, and Vd or in terms of the parameters ka, Cl, and Vd. The two-compartment model was expressed in terms of the parameters ka, the distribution rate constant a, the terminal phase elimination rate constant p, and the ratio A/B, where A and B are the coefficients defined in equation 26 shown below: C = A e^ + Be'* + Ce‘ kat [Eq.26] The other parametrization of the two-compartment model provided estimates for ka, Cl, the intercompartmental clearance Q, and the apparent volumes of distribution of the central (VJ and peripheral (Vp) compartments. Comparison between the one- and two-compartment models was performed using the Likelihood ratio test and the standard errors of the parameter estimates. These analyses indicated that the one-compartment model with first-order absorption could adequately describe the PK data. A ll further population PK analyses were performed using the ADVAN2, TRANS2 subroutines of the NONMEM-PREDPP load module, which represented the one-compartment model with first-order absorption. Inter- and intraindividual variabilities were described by the Constant Coefficient of Variation (CCV) model. The influence of food on the population PK behavior of ddl was investigated by modeling the covariate FOOD as an indicator variable which was assigned the value of 0 when ddl was administered under fasting conditions and the value of 1 when the drug was administered with food. In order to investigate whether the relative bioavailability of ddl decreased in the presence of food, the biovailability of ddl 6 2 was assigned the fixed value 0.2 under fasting conditions, an average value reported in the literature for pediatric patients (Balis et al., 1992), while the bioavalability of ddl in the presence of food was estimated by direct comparison of the ddl plasma data of the fasting and fed states, as shown below: F = 0.2 if FOOD=0 and F= 0(1) if FOOD= 1 A comparison of the relative bioavailability of ddl in the presence and absence of food was also performed using a model which simultaneously accounted for a possible difference in the rate of absorption of ddl under fasting and fed conditions. Demographic variables such as age, weight, gender, height, and serum creatinine concentration were linearly related to the PK parameters ka, Cl, and Vd in order to investigate whether these variables could partially explain interindividual variability. IV.9.a. Estimation of area under the curve (AUC) The area under the curve (AUC) values of the predicted concentration-time profiles were estimated by the trapezoidal rule using Sigma Plot 5.1 (Jandel Scientific). AUC values were obtained for the time interval 0-4 hours (A U C ^) 6 3 and they were also extrapolated to infinity (AUC^) using the following equation: AUCq *. = A U C ^ + C ^/k,, [Eq.27] where C4 h is the predicted plasm a concentration at 4 hours and ke, is the elimination rate constant. IV.9.b. Estimation of Intra- and interindividual variabilities (68% confidence intervals) The 68% (± SD) confidence intervals of the predicted concentrations were estimated from the inter- and intraindividual variances using the first-order model (Beal, 1984). The first-order model is a first-term Taylor series approximation of the nonlinear regression model with respect to the random effect parameters » ? , and C y (Beal, 1984). The variance of a measured concentration, y,., w as estimated from the equation: Var(yy ) = G(1 )nG (1 )T + (fya)2 [Eq.28] where y( j is the j,h concentration measurement of the i* individual, G( 1 ) is the vector of the partial derivatives of the model F, in equation 29 shown below, with respect to the interindividual effect parameters rj,, i.e. dF/dr?ka i, dF/dria i, and dF/di7vd,i> evaluated at the population parameters, n is the covariance matrix of the random effect param eters % and a2 is the variance of the 6 4 intraindividual error e,, (Maitre et al., 1987, Vohez et al., 1990). The term G(1 ,nG (1 ,T in equation 28 can be more specifically written as shown below: q(»)qG(,)T 3F/3nci “ u . c t a F / a n , , . w e i.M ® *ei ® c t.v d X 3 F / a n e , ® V d . U «Vd.C1 « * v d 3 F / d n V d The complete model F which included intra- and interindividual variabilities was given by the following equation (Eq.29): FX .M 1 +»w..<)exp(((-a(i+nei.i)/Vd (i +nvd.i)) x t) -exp(-k.(i +nUil) x t))) x (i +etJ ) F - -------------------------------------------------------------------- ^ • 0 ■ * , , i i i » , i ) V d ( i * a o + t j i e i) In the above equation, inter- and intraindividual variabilities were defined by the constant coefficient of variation model. The partial derivatives of F with respect to rjk . 4 , rja i, and nV d i, evaluated at the expected values of the random parameters (i.e., zero) were determined by appropriate differentiation of the equation for F and they were defined as snown in equations 30-32: •F X # k .C I (e*k “ - e‘(e,/v < > t) FX.k/ (k,vd - a)te-k “ m------------------------------------- + (Eq.301 (k,vd - a>2 (k.vd - a> * FX.k,a (e-< c ,/v < )t - e-k * * ) F X 0 k.ate(-c l/V d )t SF/enci.* * --------------------------------------------------------(E q.31) (k.vd - a )2 (k.vd • a>vd 65 (e k * ‘ - e(' cl/V d)t) FX ck.a-[k. -(a/Vd )]te'(cl/Vd)t 3F/9»lvd.i = (k.V„ - O)2 + (k.vd-a)* [Eq.32] IV. 10. Pharmacodynamic studies of ddl IV.10.a. Materials The PBMC cells were obtained from the blood of healthy volunteers. Didanosine and the Ficoll Histopaque (1.077g/ml) were obtained from Sigma Chemical Co., St. Louis, Mo. The radioactive 3H-ddl had a specific activity of 31 Ci/mmol and was obtained from Moravek Biochemicals Inc., Brea, CA. AZT w as obtained from Sigma Co., St. Louis, MO. The Jurkat E6-1 cell line (Jurkat/0) was provided by Dr. K. Weinberg, Division of Research Immunology and Bone Marrow Transplantation, Childrens Hospital Los Angeles. The Jurkat E6-1/AZT-10 T-cell line (Jurkat/AZT-10) was developed in our laboratory by exposure of the Jurkat/0 cells to an initial concentration of 0.1 m M that was gradually increased to 10 nM. IV.IO.b. Isolation of PBMC cells A volume of 20-30 ml whole blood from a healthy donor was centrifuged at 2000 rpm for 5 minutes, and the plasma was removed and stored at 0 °C. The cellular fraction was diluted 1:4 with cold PBS, layered onto Ficoll- Histopaque, and centrifuged at 1200 rpm for 25 minutes. The PBMC cells 6 6 were extracted from the interphase, washed two times with Hank’s Balanced Salt Solution (HBSS), and counted using a Coulter counter. Then, the cells were resuspended in RPM11640 enriched buffer which contained L-glutamine and penicillin-streptomycin so that there were approximately 20x106 PBMC/5 ml or 4x106 cells/ml. IV.10.C. Separation of PBMC cells into T-lymphocytes and monocytes An aliquot of 8.3 ml fetal calf serum (Gemini Bioproducts, Calavasas, CA) was added to each 100 mm polystyrene tissue culture dish (Corning 25020). The fetal bovine serum (FBS) was aspirated from the surface of the tissue culture dishes after 5 minutes. The PBMC cells were layered onto each dish so that there were approximately 60x106 PBMC cells per dish. After a 2 hour incubation at 37 °C, the lymphocytes were removed from the supernatant and the monocytes were removed from the bottom of the tissue culture dishes by vigorous pipeting with 15 ml of HBSS placed on the dishes for 15 minutes (Weinberg, personal communications). IV.IO.d. ddl anabolism in PBMC cells, lymphocytes, and monocytes An aliquot of 1x107 PBMC celss, lymphocytes or monocytes was incubated with a 1 mM mixture of 3H-ddl and cold ddl for 1 hour, at 37 °C. At the end of the incubation period, the cells were centrifuged at 2000 rpm for 5 minutes and the pellet was washed with 1 ml cold PBS. The ddl anabolites were extracted 6 7 from the cells with perchloric acid (PCA). The PCA extraction involved the suspension of the PBMC cells in 0.4N PCA and placing on ice for 5 minutes. This step was followed by centrifugation at 2000 rpm for 5 minutes. The supernatant, which contained the nucleotides, was neutralized with 10N KOH. The resulting mixture was centrifuged again in order to remove the pellet formed after the addition of KOH. The supernanant was saved for subsequent nucleotide analysis by HPLC. IV.lO.e. HPLC assay and quantitation of nucleotide mono-, di-, and triphosphates. The nucleotides of ddl were separated by HPLC on an anion exchange column (SAX-10, Custom LC, Inc., Houston, Texas) by gradient elution which utilized two solvent systems. Solvent A was 0.005 M ammonium phosphate solution, pH 2.8, and solvent B was 0.75 M ammonium phosphate solution, pH 3.6. Elution proceeded with a gradient increasing from 0 to 100% solvent B at 42 minutes. Detection of the nucleotides occured at 260 nm which is the absorption wavelength of nucleotides (Avramis et al., 1982). The HPLC eluates were collected every minute by fraction collector (ISCO Retriever III). Each eluate fraction was mixed with 10 ml of liquid scintillation coctail (Biosafe II, RPI). Determination of the nucleotide levels was achieved by tritium counting. 68 IV.lO.f. Cellular anabolism of AZT in T-cell lines. The anabolism of AZT was investigated in the Jurkat/0 and Jurkat/AZT-10 cell lines in order to verify differences in the ability of these two ceil lines to metabolize AZT to its active anabolite AZTTP. A volume of Jurkat/0 and Jurkat/AZT-10 cells equivalent to 1x107 cells was centrifuged at 2000 rpm for 5 minutes in order to separate the cellular pellet from the suspension. The pellet was, then, incubated for 2 hours at 37 °C with a radioactive mixture of AZT of a known specific activity. At the end of the incubation period, the cellular pellet was used to extract the AZT anabolites from the cells with PCA as desrcribed above, in section IV.IO.d. The specific activity of the radioactive AZT mixture was determined by an HPLC method and scintillation counting of the HPLC eluates. The HPLC analysis of AZT was performed on a Bondapak micro C18 reverse phase column (Waters Associates, Milford, MA) using a Waters Associates HPLC system. The sample was eluted with 50 parts of 0.2 M ammonium acetate, pH 6.5, and 50 parts of 50% methanol, using isocratic conditions and a flow rate of 0.5 ml/min. iV.lO.g. Cellular anabolism of ddl in T-cell lines The ability of AZT to modulate the cellular anabolism of ddl was investigated in Jurkat E6-1 (Jurkat/0) and Jurkat E6-1/AZT-10 (Jurkat/AZT-10) T-cell lines. 6 9 Jurkat/0 and Jurkat/AZT-10 are human T-cell lines which are sensitive and partially resistant to AZT, respectively. Two types of experiments were performed. In one experiment, a volume of Jurkat/0 and Jurkat/AZT-10 cells, each equivalent to 1x107 cells, was incubated at 37 °C with 1 nM radioactive mixture of ddl or with a radioactive mixture of 1 /jM ddl and 1 nM AZT, for 1 hour. In the second experiment, 1x107 cells were pre-incubated with 1 m M AZT for 2 hours. At the end of the pre-incubation period, the cells were washed with PBS and then incubated with a radioactive mixture of 1 nM ddl for 1 hour. Following incubation with ddl, in both experiments, the cellular pellet was obtained by centrifugation at 2000 rpm for 5 minutes. The anabolites of ddl were extracted from the cells with PCA and separated by HPLC as described above. Their intracellular levels were determined by scintillation counting. 7 0 ddl Anabolism schem e 1x107 cells 1 hour incubation with 1 mM 3H-ddl at 37 °C 1 hour incubation with a a mixture of 1 nM 3H-ddl and 1 nM AZT or 1 hour incubation with 1 nM 3H-ddl after a 2 hour pre incubation of the cells with 1 /iM AZT / PCA extraction of ddl anabolites from cellular pellet i HPLC separation of ddl anabolites | of i Quantitation of ddl anabolites by tritium counting 71 V. RESULTS. V.1. Assays of ddl and AZT in plasma from pediatric patients Plasma specimens were analyzed for ddl by both HPLC (Figure 8) and RIA methodologies as described in Materials and Methods. The equations describing the calibration lines were Y=0.746X + 0.219, R2= 0.985 and Y= 1.012X - 0.022, R2=0.997, with excellent linearity a s indicated by the squared residual values for the HPLC and RIA methods, respectively (Figure 9). The minimum concentration of ddl detected was 400 nM by HPLC and 1.06 nM by RIA. A comparison of the two methods showed that the RIA was approximately 400 times more sensitive than the HPLC method. Thus, the quantitative results of ddl in plasma were determined by the RIA method. AZT was analyzed by RIA only, as described in Materials and Methods. The linear equation for the quantitation of AZT was Y = -68.12X + 68.11, R2=0.990. V.2. Pharmacokinetic analysis of ddl alone and in combination with AZT by STS method. The pharmacokinetic analysis of ddl was performed using the computer software PCNONLIN, version 4.0. The PK profiles of ddl administered alone and in combination with AZT in pediatric patients with HIV-1 infection were adequately described by the one-compartment model with first-order absorption (Figure 10). Estimates were obtained for the absorption rate 72 2.50 0) * * 2.00 O > C M I 1.50. o X 0.50 0.00 0.50 1.50 2.00 1.00 x 1 0 mi nutes Figure 8. A typical HPLC chromatogram of ddl in human plasma, ddl was eluted at 11.48 minutes. The internal standard, 0 6mG, was eluted at 13.70 minutes. 73 a. 10.00 o o C 1.00 o 9 ^ o . i o r co C O C O ^ 0.01 V 0.01 O HPLC • RIA 0.10 1.00 10.00 Cone, of ddl, pM Figure 9. Comparison of linear calibration curves of ddl with HPLC and RIA methods. The m easured concentrations of ddl are depicted in the Y-axis and the theoretical concentrations in the X- axis. The lower ddl concentrations could not be detected by HPLC (open symbols). The line for the RIA assay was Y - 1.012X • 0.022, R2=0.997 and for the HPLC:. Y = 0.746X + 0.219, R2=0.985. 74 # dal b lo n a ■ d d lf AZT *a 0 .1 * o ■ d d lf AZT "O 0 .1 T im e, hours R gure 10. Plasma concentration-time profiles of ddl in pediatric patients after oral administration of four dose levels (ACTG #176). Dose level # 1 :60mg/m2, (A); dose level # 2 :9 0 m g/m 2, (B); dose level #3:135 m g/m 2, (C); dose level #4: 180 m g/m 2 , (D). Solid circles depict ddl concentrations when the drug was administered alone and solid squares depict ddl concentrations when ddl and AZT were administered in combination. 75 constant (ka), the elimination rate constant (k6 l), total body clearance (Cl), the the apparent volume of distribution (Vd), and the area under the curve (AUC) for almost every individual PK evaluation performed on this study. Reliable PK parameter estimates, except for the AUC, could not be provided for two PK evaluations due to the quality of the data and the limited number of PK sam ples provided (4 samples per PK evaluation including a pretreatment sample). In addition, estimates were obtained for the peak and trough ddl concentrations, the maximum (Cm ax ) ddl plasma concentration, and the time, W . required to reach Cm ax (Table 1). The frequency distributions of clearance and volume of distribution were found to be log-normal a s shown in figures 11 and 12, respectively. The frequency distribution of the elimination rate constant was log-normal for the most part, although there w as a small group of elimination rate constants much greater than the rest of the estimates (Figure 13). Estimation of the absorption rate constant (ka) was the least reliable parameter, probably due to the limited PK sampling and the very fast absorption of this drug. The log-normal and normal frequency distributions of ka are shown in figure 14. The arithmetic m eans of the absorption half-life of ddl were 14.94 ± 20.75 minutes (mean n=14 ± SD) when ddl was administered alone and 13.33 ± 17.33 minutes (mean n=39 ± SD) when ddl was administered in 7 6 Table 1: Mean PK parameter estimates of ddl when ddl was administered in combination with AZT. Dose (mg/m2 ) Peak [ddl] (mM) Trough [ddl] (mM) ^max (mM) ^m ax (min) 60 2.74±1.59 0.17±0.08 4.81 ±3.49 14.19±16.59 90 3.51 ±3.27 0.28±0.28 7.57±8.68 19.50±23.70 135 3.87±2.46 0.48±0.33 6.37±3.73 14.46±12.18 180 6.02±2.97 0.55±0.17 8.66±4.80 23.52±25.72 77 8 >- z 6 U 1 z> o W 4 q; * Li_ 2 0 1 10 100 ddl C le a r a n c e ( L / h ) Figure 11. Log-normal frequency distribution of the clearance (Cl) of ddl in pediatric patients with HIV-1 infection receiving a combination treatment with ddl and AZT. The histogram represents the distribution of clearance in this pediatric population. The curve represents the theoretical normal distribution that could be fitted to this set of data. 78 FREQUENCY 10 8 6 4 2 0 10 100 Vd of ddl (L) Figure 12. Log-normal frequency distribution of the volume of distribution (Vd) of ddl in pediatric patients with HIV-1 infection receiving a combination treatment with ddl and AZT. The histogram represents the distribution of Vd in this pediatric population. The curve represents the theoretical normal distribution that could be fitted to this set of data. 12 > — O z UJ Z D O UJ OH 10 8 Pf ■ 1 1 i i < ii iii ■ 0.1 1 10 k el ° f d d I ^ 100 Figure 13. Log-normal frequency distribution of the elimination rate constant (kel) of ddl in pediatric patients with HIV-1 infection receiving a combination treatment with ddl and AZT. The histogram represents the distribution of k# , in this pediatric population. The curve represents the theoretical normal distribution that could be fitted to this set of data. 8 0 20 6 2 8 4 n n 0 0.1 1 10 100 15 12 >- O Z 9 U J 3 a U J 6 Q £ 3 0 - 2 0 - 1 0 0 10 20 _ 30 40 50 k a o f d d l (h 1) Figure 14. A. Log-normal frequency distribution of the absorption rate constant (k j of ddl in pediatric patients with HIV-1 infection receiving a combination treatment with ddl and AZT. The histogram represents the distribution of k, in this pediatric population. The curve represents the theoretical normal distribution that could be fitted to this set of data. B. Normal frequency distribution of kr 81 combination with AZT. The geometric means were 5.56 minutes (Confidence interval or Cl=2.54 -12.13) and 5.36 minutes (Cl=3.50 - 8.20), respectively. The average values of the elimination half-life, clearance and volume of distribution were obtained using the geometric mean (Equation 8) since the frequency distributions of these parameters were clearly log-normal. On day 2, when ddl was administered alone, the average values of the elimination half- life, Cl, and Vd were 20.82 minutes (n = 14, Cl = 12.38 - 34.96), 24.36 L/h (n=14, Cl = 15.24 - 38.94), and 12.19 L/m2 (n = 14, Cl=8.14 - 18.26). On the days when ddl and AZT were co-administered, the average values of these parameters were 20.15 mimutes (n= 39, C l=14.71 - 27.61), 18.80 L/h (n=39, Cl = 14.85 - 23.80) and 9.21 L (n=39, C l=7.23 -11.73), respectively. Table 2 summarizes the pharmacokinetc results. The Student’s t-test for independent samples was used to compare the estimated PK parameters on day 2, when ddl was administered alone and on the days when ddl was administered in combination with AZT. The t-test was performed on the log-transformed values of the PK parameters since the frequency distributions appeared log-normal. Statistical analysis indicated that there were no appreciable alterations in the PK parameters of ddl when it was administered with AZT, in pediatric patients with HIV-1 infection, as shown in Tables 2 and figure 10. 82 Table 2: Statistical comparison of dose-independent PK parameters1 of ddl on day 2 when ddl was administered alone and on the days (other days) when ddl and AZT were administered in combination. Day ka (h‘1 ) k„ (h-1 ) Cl2 (L/h) Vd3 (L) day 24 7.49 (3.43-16.36) 1.99 (1.19-3.33) 24.36 (15.24-38.94) 12.19 (8.14-18.26) Other 7.76 days5 (5.07-11.87) 2.06 (1.51-2.82) 18.80 (14.85-23.80) 9.21 (723-11.73) P-value6 0.934 0.908 0.299 0.249 1 The average values of the PK parameters represent the geometric means. The numbers in parentheses represent the 95% confidence interval of the PK parameters. ^ h e clearance was calculated by the equation CI=FX0/AUC based on the assumption that the mean oral bioavailability of ddl was F=0.2, as indicated in the literature (Balis et. al., 1992). *The apparent Vd was calculated by the equation Vd=CI/k„. *The number of samples in the comparison was n=14. *The number of samples in the comparison was n=39. ®The P-values were obtained using the t-test on the log-transformed data. 83 Figure 15 depicts the average ddl plasma AUC values on day 2 (ddl alone), day 3 (first time in combination with AZT), and on weeks 4 and 12 after co administration with AZT in children with HIV-1 infection. Statistical comparisons were performed using the Wilcoxon Rank Sum test for independent samples (Table 3). The median AUC value of ddl on day 2 was not significantly different from the median AUC value of ddl during coadministration with AZT, at any dose level. The average AUC of ddl in the presence of AZT increased linearly with increasing dose level of ddl (Figure 16). The 60 m g/m 2 ddl dose level had a statistically significantly lower AUC value than the 90 mg/m2 (P=0.008), the 135 mg/m2 (P = 0.002), and the 180 m g/m 2 (P=0.0006) dose levels. No other significant differences were found between the AUC values of other dose levels (Table 3). The equation which described the average PK profile of ddl, in pediatric patients was: Cjj = 0.146X,, (exp'2 0 5 1 - exp'769* ) [Eq.33] The values of the PK parameters in the above equation represent the geometric means of the parameters. Estimation of these means was based on all individual PK parameter estimates because statistical analysis (t-test) indicated no difference between the dose-independent parameters of ddl when administered alone and in combination with AZT. 8 4 2 a . 1 2 d2 43 w4w12 d2d3w4w 12 d2 43 w4wf2 dS w4w12 60 mg/m2 90 mg/m2 135 mg/m2180 mg/m2 Figure 15. Average AUC values of ddl per dose level and per evaluation when pharmacokinetic studies were performed in the ACTG #176 clinical trial. The AUC values were estimated either by PCNONLIN or a customized computer program for AUC using the trapezoidal rule. Bars: means ± SD. 85 Table 3: Statistical comparisons1 of AUC values of ddl, per dose level, on day 2 when ddl was administered alone and on the days (other days) when ddl and AZT were administered in combination. Dose level AUC on day 22 AUC on other days2 ,3 P-value 60 m g/m 2 4.81 (n=2) 3.04 (n=7) > 0.5 90 mg/m2 4.02 (n=4) 4.66 (n=8) > 0.5 135 m g/m 2 6.09 (n=10) 6.16 (n=21) 0.40 180 m g/m 2 - 10.68 (n=3) - 1 The statistical comparisons between AUC values were performed by the Wilcoxon Rank-Sum test for independent samples. ^The AUC values are represented as medians. ^The median AUC of the 60 m g/m 2 dose level was statistically significantly lower than the AUC value of the 90 m g/m 2dose level (P=0.0008), of the 135 m g/m 2 dose level (P=0.002), and of the 180 mg/m2 dose level (P=0.0006). No other statistical differences were observed between the AUC values of the four dose levels of ddl on the days when ddl and AZT were administered in combination. 86 j i m r i ‘ip p i o o n v 1 2 9 6 3 0 30 60 90 120 150 180 210 Dose level of ddl, mg/m2 Figure 16. Linear relationship between average AUC values of ddl in pediatric patients and d o se level of drug. 87 The variance-covariance matrix, n, from these analyses, was estimated using equation 9 (Materials and Methods, p.45): " 2 x « “ k .,a W k»,Vd 1.863 -0.404 0.411 n = w a .k » w2 a w a.v d = -0.404 0.417 0.104 w Vd.k* w v d ,a e 1 0.411 0.104 0.754 V.3. Population pharmacokinetic analysis of ddl In ACTG #176 study by the NONMEM program The ddl plasma d ata were analyzed by the program NONMEM using a simple structural regression model which did not include covariates (basic model) and more complex structural (expanded) models which included covariates. The structural model w as expressed in terms of the basic PK parameters clearance (Cl), volume of distribution (Vd), and first-order absorption rate constant (ka) or in terms of k,, Vd, and the elimination rate constant (k„). The OMEGA matrix of interindividual effects (Variance-Covariance matrix) was unconstrained. This means that covariances were included between all pairs of rjj elements. The average values of the PK parameters Cl and Vd (TRANS2) or Vd and ktl (TRANS 1) estimated by the STS method, were used as initial estimates in the NONMEM analysis of the data. The basic model parameters (Table 4, model 1) were: k, = 4.07 ± 1.74 h '\ Cl = 17.5 ± 2.38 L/h, and Vd 88 Table 4: Population pharmacokinetic models of ddl for pediatric patients with HIV-1 infection receiving a combination treatm ent of ddl and AZT. Model1 k.(h-’) Cl (L/h) Vd (L) (J 2 w2 c, W 2 v d OBF2 1. 4.07 17.5 18.0 0.29 1.23 0.32 0.28 177.00 (1.74) (2.4) (3.2) (0.14) (2.94) (0.20) (0.21) 2. 4.31 19.2 - 2.34xAZT0, 18.1 0.28 1.44 0.33 0.27 171.85 (2.00) (2.9) (1.21) (3.5) (0.15) (3.51) (0.23) (0.21) 3. 4.99 17.2 + 1.71xAZTP 18.2 0.35 0.46 0.27 0.21 175.27 (2.80) (2.4M2.9) (2.9) (0.16) (5.27) (0.18) (0.30) ’The numbers in parentheses represent the standard errors of the estimated parameters. 2 OBF represents the value of the minimum objective function (Log-likelihood function). 0 0 VO = 18.0 ± 3.23 L. The equation that described the average PK profile of ddl in pediatric patients using the NONMEM estimates of the PK parameters was: C„ = 0.07Xo (e-09* - e-4 0 7 * ) [Eq.34] The variance-covariance matrix, n, was: “ 2 k « wk*.a “ kt.Vd 1.28 0.204 0.255 n - " C U c . o > 2 a “ a.vd = 0.204 0.318 0.296 wva,k* wvd.a < * > 2Vd 0.255 0.296 0.279 Investigation of a possible PK interaction between ddl and AZT by the NONMEM program was performed using two types of analysis. The first type of analysis was b ased on a structural model which used an indicator variable to distinguish between the ddl PK data obtained after co-administration of ddl and AZT and those data obtained after administration of ddl alone. The indicator variable w as assigned the value 0 when ddl was administered alone and the value 1 w hen ddl was administered in combination with AZT (Table 4, model 2). Comparison of models 1 and 2 (Table 4) by the Likelihood ratio test indicated that the concurrent presence of AZT in plasma, as defined in model 2, did not alter the clearance of ddl. In the second type of analysis, the structural model expressed the clearance of ddl as a linear function of the 9 0 measured AZT plasma concentrations. NONMEM regression analysis with this model could not terminate successfully because no relationship could be determined between the clearance of ddl and the plasma concentrations of AZT. Using the TRANS1 parametrization option of the NONMEM program, estimates (mean ± SE) were obtained for ka, Vd, and ke|) as shown below: ka = 4.07 ± 1.74 h'1 (t*k a = 10.22 min) Vd = 17.9 ± 3.2 L ka l = 0.97 ± 0.09 h'1 (t*e l = 42.88 min) V.4. Pharmacokinetic analysis of AZT by STS method. A PK analysis was also performed for AZT in the same pediatric group. The one-compartment model with first-order absorption was adequate to describe the PK profiles of AZT alone and in combination with AZT (Figure 17). The AZT plasma levels increased with increasing AZT doses and the elimination half-life remained essentially unchanged, at the four AZT dose levels studied. On day 1, when AZT was administered alone, the maximum plasma AZT concentration, Cm ax, averaged from 4.19 ± 1.78 (range) to 8.60 ± 0.62 /uM (range) with increasing AZT doses. The elimination half-life, which was independent of the dose level, had an arithmetic mean of 38.03 ± 13.82 1 10 ^ A Z T Iolona ■ AZT'+ddl 9 1.0 < 0.1 B Q AZT a lo n a i ■ A Z T+ddl | 2 10 a . < 0.1 0 AZT a lo n a i ■ A Z T + ddl { 1 Time, hours Figure 17. Plasma concentration-time profiles of AZT in pediatric patients after oral administration of four dose levels (ACTG #176). Dose level # 1 :60mg/m , (A); dose level # 2 :9 0 mg/mr, (B); dose level # 3 :120 m g/m 2, (C); dose level #4: 180 mg/m2, (D). Solid circles depict ddl concentrations when the drug was administered alone and solid squares depict ddl concentrations when ddl and AZT were administered in combination. 92 minutes (mean ± SD, n=8) and a geometric mean of 35.92 minutes, suggesting that AZT obeyed linear kinetics. The arithmetic average half-life of absorption w as estimated to be 7.73 ± 9.50 minutes (mean ± SO, n= 8) and the geometric mean of the absorption half-life was 4.30 minutes (Cl=2.01 - 9.18). In the presence of ddl, AZT also obeyed a one-compartment model with first- order absorption (Figure 17). Table 5 shows the average values of the peak and trough concentrations of AZT, of Cm ax , and t ^ , per dose level of AZT. Table 6 shows the average values (geometric m eans and confidence intervals) of dose-independent PK parameters of AZT in the presence of ddl. The geometric mean was chosen over the arithmetic mean to provide average values for ka, kel, Cl, and V d because the frequency distributions of these parameters demonstrated a log-normal behavior (Figures 18-21). The geometric m eans of the absorption half-life, elimination half-life, clearance, and volume of distribution of AZT were 1.80 min (Cl = 1.15 - 2.81), 27.30 L/h (Cl=22.99 - 32.43), and 18.27 L (Cl = 14.08 - 23.69), respectively. The t-test for independent samples was used to provide comparisons between th e PK parameters of AZT on day 1, when AZT was administered alone, and o n the days when AZT was administered in combination with ddl. The t-test was performed on the log-transformed values of ka, kal, Cl, and Vd. Co- 93 Table 5: Mean PK parameter estimates of AZT when AZT was administered in combination with ddl. Dose (mg/m2) Peak [AZT] (mM) Trough [AZT] (m M ) ^m»x (M M) V n a x (min) 60 2.19±1.35 0.28±0.23 5.09±2.08 10.92±6.54 90 5.60±2.61 0.33±0.33 10.87±6.22 10.87±6.70 120 4.29±2.52 0.60±0.59 6.93±3.40 17.10±19.50 180 6.91 ±2.87 0.72±0.42 9.24±4.43 22.20±25.72 94 Table 6: Statistical comparison of dose-independent PK parameters of AZT on day 1, when AZT was administered alone, and on the days when AZT and ddl were administered in combination. Day k. (h ’j k., (h-1 ) Cl2 (L/h) Vd3 (L) day 14 9 67 (4.53-20.66) 1.16 (0.90-1.49) 31.31 (19.13-51.26) 26.87 (16.48-44.43) Other days5 23.10 (14.79-36.09) 1.49 (1.25-1.79) 27.30 (22.99-32.43) 18.27 (14.08-23.69) P-value6 0.934 0.908 0.299 0.249 1 The average values of the PK parameters represent the geometric means. The numbers in parentheses represent the 95% confidence intervals of the PK parameters. ^ e clearance was calculated by the equation CI=FX0/AUC based on the assumption that the mean oral bioavailability of AZT was F=0.6, as indicated in the literature (Yarchoan et al., 1986). ^ e apparent Vd was calculated by the equation Vd=CI/ka,. ^ e number of samples in the comparison was n=8. *The number of samples in the comparison was n=36. ®The P-values were obtained using the t-test on the log-transformed data. 95 >“ o Z > Or Ld Q d 15 12 9 6 3 0 100 0.1 1 10 - 1 >- u z o L U cc 14 12 10 8 6 4 2 0 40 30 20 10 -1 0 0 - 1 Figure 18. A. Log-normal frequency distribution of the absorption rate constant (ka) of AZT in pediatric patients with HIV-1 infection receiving a combination treatment with ddl and AZT. The histogram represents the distribution of k, in this pediatric population. The curve represents the theoretical normal distribution that could be fitted to this set of data. B. Normal frequency distribution of k,. 9 6 10 8 >- O o LlJ o n 0 / 10 - 1, kel of AZT (h ) Figure 19. Log-normal frequency distribution of the elimination rate constant (k„) of AZT in pediatric patients with HIV-1 infection receiving a combination treatment with ddl and AZT. The histogram represents the distribution of k., in this pediatric population. The curve represents the theoretical normal distribution that could be fitted to this set of data. 97 >- o or 8 6 4 2 0 100 10 AZT C le a r a n c e ( L / h ) Figure 20. Log-normal frequency distribution of the clearance (Cl) of AZT in pediatric patients with HIV-1 infection receiving a combination treatment with ddl and AZT. The histogram represents the distribution of Cl in this pediatric population. The curve represents the theoretical normal distribution that could be fitted to this set of data. 98 FREQUENCY 10 8 6 4 2 0 Vd of AZT (L) Figure 21. Log-normal frequency distribution of the volume of distribution (Vd) of AZT in pediatric patients with HIV-1 infection receiving a combination treatment with ddl and AZT. The histogram represents the distribution of Cl in this pediatric population. The curve represents the theoretical normal distribution that could be fitted to this set of data. 9 9 administration of ddl with AZT did not significantly alter the PK parameters of AZT (Table 6). Figure 22 depicts the average (median) AZT plasma AUC values on day 1, day 3, and on weeks 4 and 12. Statistical comparisons between AUC values w ere performed using the Wilcoxon Rank Sum test for independent samples (Table 7). The median AUC value of AZT on day 1 w as not statistically significantly different from the median AUC obtained on the days of AZT and ddl co administration, at any of the four dose levels of AZT. The AUC values of AZT in the presence of ddl increased linearly with dose level of AZT (Figure 23). In the presence of ddl, the median AUC of the 60 mg/m2 dose level w as statistically significantly lower than the AUC values of the 120 mg/m2 d o se level (P=0.025) and the 180 mg/m2 dose level (P=0.017). No other significant differences were observed between the AUC values of the four dose levels of AZT when AZT and ddl were administered in combination. Equation 35 describes the average PK profile of AZT in pediatric patients using the geometric means of the PK parameters of AZT: Cjj = 0.06X, (e 13 0 1 - e'1 1 2 4 t) [Eq.35] 100 O D < 2 6 3 dl d3 w4w12 dl d3 w4wt2 d1 d3 w4w12 al d3 w4w!2 60 mg/m2 90 mg/m2 120 mg/m2 180 mg/m2 Figure 22. Average AUC values of AZT per dose level and per evaluation when pharmacokinetic evaluations were performed. The AUC values were estimated either by PCNONLIN or a customized computer program for AUC using the trapezoidal rule. Bars: means ± SO. 101 Table 7: Statistical comparisons1 of AUC values of AZT, per dose level, on day 1, when AZT was administered alone, and on the days (other days) when AZT and ddl were administered in combination. Dose level AUC on day 12 AUC on other days2,3 P-value 60 mg/m2 3.51 (n=4) 3.31 (n=7) > 0.5 90 mg/m2 4.43 (n=4) 6.96 (n=16) > 0.099 120 mg/m2 6.28 (n=4) 7.14 (n=11) > 0.5 180 mg/m2 12.8 (n=2) 10.68 (n=8) > 0.5 1 The statistical comparisons between AUC values were performed by the Wilcoxon Rank-Sum test for independent samples. h 'h e AUC values are represented as medians. ^The median AUC of the 60 m g/m 2 dose level was statistically significantly lower than the AUC value of the 120 m g/m 2 dose level (P=0.025), and of the 180 mg/m2 dose level (P=0.017). No other statistical differences were observed between the AUC values of the four dose levels of AZT on the days when ddl and AZT were administered in combination. 102 15 N < O 3 < l 30 60 90 120 150 180 210 Dose level of AZT, mg/m2 Figure 23. Linear relationship between average AUC values of AZT in pediatric patients and dose level of the drug. 103 The variance-covariance matrix was: “ 2k« w k * .a “ kt.Vd 1.014 0.065 -0.038 n = “ a * . <i)2a “ a .v d s 0.065 0.489 -0.484 w Vd,k* “ v d .a -0.038 0.484 0.610 _ _ _ V.5. Population pharmacokinetic analysis of AZT by the NONMEM program The PK data of AZT were analyzed by the NONMEM program in a similar manner used to evaluate the ddl data. However, in this case, interindividual variabilities were determined only for the PK parameters Cl and Vd but not for ka, because the NONMEM method could not provide meaningful estimates for the interindividual variability associated with the parameter ka. More data are needed to estimate variance parameters with a precision comparable to that of structural PK model parameters (NONMEM Guide, Part V, p. 123). When NONMEM modeling has difficulty obtaining a meaningful value for a variance parameter, it usually provides a value for this parameter which is close to zero, as if there is practically no interindividual variability associated with this parameter. This is, in fact, an indication that the data can not support an elaborate statistical model and a simpler model should be used (NONMEM Guide, Part V, p. 123). Similar results of near zero interindividual variabilities were obtained in other studies (Williams et al., 1992). The OMEGA matrix of 104 random effects was unconstrained in terms of the parameters Cl and Vd, but the variance of k, was constrained to be zero. The basic regression model (Table 8, model 1) estimated a clearance of 35.2 L/h, a volume of distribution of 34.2 L, and a first-order absorption rate constant of 4.66 h'1 . The average PK profile of AZT using the NONMEM parameter estimates was: The presence of ddl in plasma did not influence estimation of Cl, Vd or ka of AZT when both types of analysis were employed (Table 8, models 2 & 3) as described in the above section. Although there was a reduction in the value of the minimum objective function in models 2 and 3 as compared to model 1, the difference was not significant at P < 0.01. Hence, the presence of ddl in plasma concurrently with AZT did not influence the pharmacokinetics of Using the TRANS1 parametrization option of the NONMEM program, the following population parameter estimates (mean ± SE) were obtained: C|j = 0.04X,, (e 103t - e * 46® * ) [Eq.36] The variance-covariance matrix was: W vd^Q * * * V d AZT. 105 Table 8: Population pharmacokinetic models of AZT in pediatric patients with HIV-1 infection receiving a combination treatment of ddl and AZT. Model1 k.(h M Cl (L/h) Vd (L ) a2 u 2 a w2 V d OBF2 1. 4.66 35.2 34.2 0.32 0.22 0.26 246.45 (1.65) (4.8) (5.0) (0.13) (0.15) (0.20) 2. 4.69 31.8 + 4.47xddl0, 33.4 0.32 0.24 0.28 243.01 (2.01) (4.5) (1.67) (5.3) (0.14) (0.16) (0.21) 3. 4.76 34.0 + 0.79xddlc 33.4 0.32 0.22 0.26 245.99 (2.01) (4.6) (0.86) (5.3) (0.13) (0.16) (0.21) ^he numbers in parentheses represent the standard errors of the estimated parameters. 2 OBF represents the value of the minimum objective function (Log-likelihood function). ka = 4.67 ± 1 .6 6 h'1 (t*K a = 8.91 min) ke, = 1.03 ± 0.12 h'1 (t*a, = 40.38 min) Vd = 34.2 ± 4.98 L V.6. Effect of prior ACT treatment on ddl pharmacokinetics The effect of ACT treatment prior to the beginning of a combination treatment of ddl and AZT, on ddl pharmacokinetics, was investigated using model 3 (Table 4), also shown as equation 25. The minimum objective function of model 3 was 180.54 and it was not significantly different from the corresponding value of model 1 (Likelihood ratio test). Therefore, treatment with AZT prior to a combination treatment with ddl and AZT did not affect the pharmacokinetics of ddl. V.7. Population pharmacokinetic analysis of ddl in the ACTG #144 study by NONMEM Figures 24 and 25 depict the plasma ddl concentration-time profiles of 90 pediatric patients with HIV-1 infection following oral administration of 50 and 150 m g/m 2 ddl under fasting conditions. Figures 26 and 27 depict the PK profiles of the two ddl dose levels in the presence food. It is clear from these plots that there is not a clear distinction between the PK profiles of the two ddl dose levels. A large number of the PK profiles of these two dose levels overlap. On average, however, the PK profiles of the high dose group 107 o> c •t c o c < D o c o CJ t 5 *D 1000 i '7rs " ^ ^ ;w ; ^?^SS 100 5 0 2 3 4 1 T i m e , h o u r s Figure 24. Plasma concentrations of ddl in patients receiving 50 m g/m 2 (dashed lines) and 150 mg/m2 (solid lines) ddl under fasting conditions. These plots indicate that there is no clear distinction between the PK profiles of these two dose levels. On average, however, the PK profiles of the high dose group appear to correspond to higher plasma concentrations than the profiles of the low dose group. 108 3000 E \ 1000 o> c ^ • : . . . - . . . . . : :!}<;--^Sb S ^ 2 m c o 100 o i_ C 0 ) o c o a -o • o 3 4 5 0 2 1 T im e , h o u r s 3000 \ 1000 o> c ~ 100 o u -♦ — c a) 0 1 10 O T 3 ■ O 1 0 1 2 3 4 5 T im e , h o u r s Figure 25. Plasma concentration-time profiles of ddl in patients receiving ddl under fasting conditions. A. 50 m g/m 2 ddl (dashed lines). B. 150 mg/m2 ddl (solid lines). 109 1 0 0 0 o> c c 100 o o k _ C 10 © o c o O 1 ■ O XJ 0.1 0 1 2 3 4 5 T i m e , h o u r s Figure 26. Plasma concentrations of ddl in patients receiving 50 mg/m2 (dashed lines) and 150 mg/m2 (solid lines) ddl under fed conditions. These plots indicate that there is no clear distinction between the PK profiles of these two dose levels. On average, however, the PK profiles of the high dose group appear to correspond to higher plasma concentrations than the profiles of the low dose group. 110 3 0 0 0 o > * 100 0 1 2 3 4 5 T im e , h o u r s _ 3 0 0 0 E ^ 1000 O ) c M c o 100 o u C Q > g 10 O o TJ - O 1 0 1 2 3 4 5 Time, hours Figure 27. Plasma concentration-time profiles of ddl in patients receiving ddl under fed conditions. A. 50 mg/m2 ddl (dashed lines). B. 150 mg/m2 ddl (solid lines). 111 appear to correspond to higher plasma concentrations than the profiles of the 50 m g/m 2 dose level, both under fasting and fed conditions. The PK profiles of ddl in the fasting state (Figure 24 & 25) indicated that some of them could best be described by the one-compartment model while others were b est described by the two-compartment model with first-order absorption. On the other hand, almost all of the PK profiles of ddl in the fed state (Figures 26 & 27) demonstrated monoexponential elimination of the drug. Although figures 24-27 suggested that on average, the one-compartment model could adequately describe the PK profiles of ddl, the pharmacokinetic data were fitted both by the one- and two-compartment models with first-order absorption in order to determine based on statistical methods which of these two structural PK models could describe the data most adequately. The minimum objective function (OBF) of the two-compartment model (both para- metrizations) had values smaller than the OBF of the one-compartment model. However, based on the Likelihood ratio test, the difference in OBF w as not significant at P < 0.01. Further evidence supporting the choice of the one- compartment model over the two-compartment model was provided by the fact that the elimination rate constant, /3, w as estimated with a standard error greater than the parameter estimate itself (Table 9). Therefore, the terminal elimination phase could not be estimated reliably by the two-compartment 112 I Table 9: Comparison of one- versus two-compartment model with first-order absorption. Model1 -2 - 3 k.lh1 ) k.flr1 ) CI(L/h/m 2 ) Vd (L/m2 ) A/B a(h’) jfffh1 ) Q(L/h/m2 ) O B F One-compartment (Ad van 2/Trans 1) 4.07 (0.49) 0.52 (0.03) “ 43.0 (3.8) - - “ “ 6072.15 One-compartment (Advan2/Trans2) 4.08 (0.60) - 22.4 (1.6) 43.0 (4.2) - - - - 6072.15 T wo-compartment (Advan4/Trans4) 2.91 (0.38) - 20.5 (1.7) Vc = 33.2 (3.5) Vp= 14.1 (2.3) - - - 13.6 (3.2) 6069.11 T wo-compartment (Advan4/Trans5) 1.17 (0.13) - - 13.5 (2.5) 9.62 (5.9) 2.89 (0.61) 0.19 (0.23) - 6066.23 1 k,, k„ C l, and V d are as defined in text, a is the distribution rate constant, P is the terminal phase elimination rate constant, and Q is the intercompartmental clearance in the two-compartment model. A/B is the ratio of the coefficients of the exponentials of a and P. respectively, in the two-compartment model. Vc and Vp represent the apparent volumes of distribution of the central and peripheral compartments, respectively. O B F is the value of the minimum objective function. 2 Numbers in parentheses represent the standard errors (SE) of the estimated parameters. u model. Based on these analyses, the one-compartment model with first-order absorption was chosen to perform all NONMEM analyses of ddl in this study. In the absence of any covariates, the basic PK parameters of ddl (mean ± SE) were estimated in this population as: ka = 4.08 ± 0.60 h'1 with a half-life of absorption, t*k a = 10.19 min, Cl = 22.4 ± 1.6 L/h/m2, Vd = 43.0 ± 4.2 L/m2, and kal = 0.52 h 1 with a half-life of elimination, t^k a , = 1.33 h. This is shown in model 1 of Table 10. Model 1 was taken as the basic model and it was expanded in a stepwise manner in order to investigate the effect of covariates such as food, age, weight, gender, and serum creatinine on the PK of ddl. One of the main questions that had to be answered w as whether food influences the absorption kinetics of ddl and/or the bioavailability (i.e., extent of absorption) of this drug. For this purpose, the structural PK model expressed ka as a linear function of the indicator variable FOOD. As already mentioned in the Methods section, the indicator variable FOOD allowed the regression model to distinguish between the PK profiles of the fasting and fed states and simultaneously provide estimates for the absorption rate constant of ddl in both states. Administration of ddl with food resulted in a significant decrease (P < 0.01) in the rate of absorption as compared to the fasting state (models 1 and 2 in Table 10). The relative bioavailability of ddl, however, was not different in the 114 Table 10: Com parison of population pharm acokinetic m odels of ddl in th e ACTG # 1 4 4 clinical trial. M odel12* k.(h'| C I(L /h /m 2 ) V d (L /m 2 ) F o2 C l ^ V d O B F 1. W ithout covariates 4.08 (0.60) 22.4 (1.6) 43.0 (4.2) 0.2 0.23 (0.04) 0.10 (0.23) 0.38 (0.20 0.31 (0.09) 6072.15 2. Influence of food on k. 5.81 - 4.14xF00D (1.53) (1.50) 21.6 (2.1) 39.9 (6.3) 0.2 0.17 (0.04) 1.48 (1.14) 0.36 0.13) 0.35 (0.09) 601550 3. Influence 4.08 22.5 43.1 0.2 (no food) 0.23 0.10 0.37 0.31 6072.14 of food on F (0.51) (2.4) (3.8) 0.201 (food) (0.003) (0.05) (0.32) (0.13) (0.15) 4. Influence 6.20 - 4.50xFOOD 20.9 39.6 0.2 (no food) 0.16 1.61 0.37 0.35 601334 of food on k. and on F (1.09) (1.05) (2.23) (3.51) 0.19 (food) (0.005) (0.03) (0.58) (0.11) (0.14) 5. Influence of food on k. and age on C l 5.74 - 4.07xFOOD (1.16) (1.08) 29.9- 0.1 O x A G E (4.5) (0.06) 40.4 (5.4) 0.2 0.16 (0.03) 1.91 (1.41) 0.38 (0.19) 0.36 (0.08) 5983.71 6. Influence of food on k„ age on C l, and weight on V d 4.92 - 3.52xFOOD (1.13) (1.11) 29.1 - 0.075xAGE (3.4) (0.037) 53.1 - 0.89xWT (7.8) (0.18) 0.2 0.16 (0.03) 0.64 (0.48) 0.36 (0.17) 0.36 (0.08) 593953 'Numbers in parentheses represent the standard errors (SE) of the estimated parameters. 2 Bioavailability (F ) was fixed at the value of 0.2 in models 1, 4, 5, and 6. In models 2 and 3, F was fixed at 0.2 when ddl was administered without food and was estimated by N O N M EM in the presence of food. 3 Covariance matrix of interindividual effects was unconstrained. fasting or fed patients (Table 10, models 3 and 4). When the structural PK model related age to clearance and weight to volume of distribution a further improvement of the data fit w as observed (Table 10, model 6). Model 6, therefore, implied that the clearance of ddl decreases linearly with increasing age while the volume of distribution decreases linearly with increasing weight. Expansion of model 6 to evaluate the effect of the covariates gender and height did not result in a statistical improvement of the model. When model 6 was expanded to relate serum creatinine levels to clearance, a significant decrease in the value of the minimum objective function of model 6 (P < 0.01) was observed. However, the estimated coefficient of serum creatinine levels was associated with a large standard error and a 95% confidence interval of this parameter which included the null value (i.e. zero) of the parameter. The null value of a parameter is the value that causes the parameter to be effectively removed from the model. Model 6, therefore, was taken to be the model that best described the pharmacokinetic behavior of ddi and explained most of interindividual variability. Based on model 6, the coefficients of variation of ka, Cl, and Vd were 79.7%, 59.6%, and 60.3%, respectively. The intraindividual coefficient of variation was 40.0%. The variance-covariance OMEGA matrix was: “ 2 k » “ ka.a wk a . V d 0.636 -0.044 0.056" n = ^ C l.k a w 2 a wa ,V d = -0.040 0.355 0.158 “ v d .k a “ v d .C l < * > 2 V d 0.056 0.158 0.364 116 Figure 28 shows the predicted PK profiles of ddl, for a patient of average weight (19.2 kg) and ag e (77.7 months) in this study, when the drug was administered on an empty stomach and with food. This figure clearly indicates that the rate of absorption of ddl was slower in the presence of food. However, the area under the curve (AUC) and hence bioavailability, is approximately the same. The predicted A U C ^ values of the two plots in figure 28 w ere estimated by the trapezoidal rule using Sigma Plot 5.1. The average (predicted) A U C ^ and the 68% confidence intervals (Cl) of 50 m g/m 2 ddl administered under fasting and fed conditions were 391.60 (ng/ml)hr (01 = 137.29-1117.69) and 371.01 (ng/m l)hr (Cl=88.28-654.27), respectively. The average A U C ^ of the fasting and fed states of the 50 m g/m 2 dose level were 428.82 and 428.17 (ng/ml)hr, respectively. Figures 29 & 30 show the observed plasma concentrations of ddl, the predicted plasma concentrations for a patient of average age and weight, and the intraindividual and total (inter- and intraindividual) variabilities associated with each predicted concentration, with and without food, for the 50 and 150 m g/m 2 dose levels of ddl, respectively. In order to verify that the A U C ^ values which were calculated using the ddl PK parameters estimated by the NONMEM analyses of the entire data set, the data set w as separated into two subgroups according to dose level. Separate NONMEM analyses were then performed of the two subsets of data. This 117 0 1 2 3 4 5 T i m e , h o u r s Figure 28. Predicted concentration-time profiles of ddi for a patient of average age (77.7 months) and weight (19.2 kg) in the ACTG #144 clinical trial receiving 50 mg/m2 ddl under fasting conditions (solid line) and with food (dashed line). 118 8 0 0 Time, h ou rs 800 E \ CD _ c 600 c o % 400 i- c < 0 o c 200 o o "O ■o 0 0 1 2 3 4 5 Time, hou rs Figure 29. T he solid line represents the predicted concentrations of ddl (50 mg/m2 ) in a patient of average age and weight in this study fro m population P K param eters obtained w ith m odel 6. T he area bound by the dotted lines represents the 6 8 % confidence in terv a l due to in tra in d iv id u a l v a ria b ility (±SD ). T he area bound by the dashed lines represents the 68% confidence in te rv a l due to total (i.e., in tr a - and in terin d iv id u a l) v a ria b ility . A . ddl adm inistered on an em pty stom ach. B . ddl adm inistered w ith food. 119 o> c •t c o c 0 ) o c o o ■ D 2000 1500 1000 500 2 3 Time, h o u rs E 2000 \ cn c . 1500 c o 2 1000 c Q > O £ 500 o 0 1 2 3 4 5 Time, hours Figure 30. The solid line represents the predicted concentrations of ddl (150 mg/mz ) in a patient of average age and w eight in this study from population P K param eters obtained w ith m odel 6. The area bound by the dotted lines represents the 6 8% confidence in terv a l due to in tra in d iv id u a l v a ria b ility (±SD ). T he area bound by the dashed lines represents the 6 8 % confidence in terv a l due to total (i.e., in tra - and in terin d iv id u a l) v a ria b ility . A . ddl adm inistered on an em pty stom ach. B. d d l adm inistered w ith food. B 120 approach was not expected to provide the sam e estimates for the PK parameters as the NONMEM analysis of the entire data set due to the smaller number of data points included in these analyses and data restriction to one dose level. However, it was expected to provide AUC values in the range of those obtained by the NONMEM analysis of the entire data set. The mean A U C ^ values of 50 mg/m2 ddl and the 68% confidence intervals associated with these predictions, under fasting and fed conditions, were 452.52 (ng/ml)-hr (01 = 180.90 - 725.02) and 433.92 (ng/m l)hr (01=75.92 - 716.23), respectively. The predicted A U C ^ values and 68% confidence intervals of the 150 m g/m 2 ddl dose level were 1103.22 (ng/ml) hr (01=451.94 -1705.21) and 1112.54 (ng/ml)-hr (01=472.27 - 1687.64) in the fasting and fed states, respectively. The final equation that predicted the plasma concentration of ddl in this pediatric group, taking into account the conditions under which ddl was administered (i.e., with or without food), as well as a patient’s age and weight, is shown below: g 2X q(4 92 - 3 52xFOOD) (e” ^ 2 2 ,1 •0. 8f t t W T ) t _ q - ( 4 .9 2 - 3 . 5 2 x F O O D ) t j C,j = (4.92 - 3.52xFOOD)(53.1 - 0.89xWT) - (29.1 - 0.075xAGE) 121 V.8. Cellular anabolism of ddl in PBMC cells The anabolism of ddl was investigated in PBMC cells obtained from three healthy volunteers, as well as in separated lymphocytes (>95%) and monocytes (>95%). The ddl anabolites were quantitated by HPLC (Figure 31) and scintillation counting of the appropriate HPLC eluate fractions. The average percentage of monocytes in PBMC cells was 8.74% ± 1.49%. Activation of ddl to its mono*, di-, and triphosphates was highest in the separated lymphocytes than in the total PBMC cell population or monocytes (Table 11). The intracellular concentrations of ddl, ddAMP, and ddADP were not significantly different in the PBMC cells and lymphocytes (Table 11). However, concentrations of ddATP, the active anabolite of ddl against HIV-RT, were significantly higher in lymphocytes (32.31 ± 8.76 nM) than in PBMC cells (23.03 ± 5.20 nM). The concentrations of ddl and its mono-, di-and tri phosphates were statistically significantly higher in PBMC and T-cells than monocytes (Figure 32). V.9. Anabolism of AZT in the Jurkat/0 and Jurkat/AZT-10 T-cell lines. The anabolism of AZT was performed in the Jurkat/0 and Jurkat/AZT-10 cell lines in order to verify the reported differences in the activation of AZT to AZTTP in these two cell lines (Avramis et al., 1993). Figure 33 shows a typical HPLC chromatogram of AZT obtained for the determination of the specific activity of AZT. AZT eluted at 14.20 minutes. Figure 34 depicts the 122 22 % < 0 o c ( 0 J Q M s s 2 0 40 30 35 25 20 5 10 15 0 Time, m in u te s Figure 31. A typical HPLC chromatogram showing the separation of nucleotide triphosphates. ddATP was eluted at 28 minutes. The bars superimposed to this chromatogram represent the radioactivity counts (log scale) for the radioactive drug, ddl, and its anabolites ddAMP/ddIMP, ddADP, ddATP, and ddGTP. 123 Table 11: Anabolism of ddl in human PBMC cells, T-cells, and monocytes following a 1 hour incubation with a 1 //M mixture of 3H-ddl and ddl. Anabolite ddl (nM) ddAMP (nM) ddADP (nM) ddATP(nM) PBMC12 2276.01 ± 671.93 384.89 ± 106.36 18.10 ± 6.86 23.03 ± 5.20 (n = 8) T-cells13 2184.03 ± 506.73 406.39 ± 162.63 22.84 ± 5.99 32.31 ± 8.76 In = 17) M onocytes2 ,3 1163.48 ± 302.29 172.98 ± 85.66 10.93 ± 4.02 10.85 ± 1.72 (n = 6) ’The intracellular concentrations of ddl, ddAMP, and ddADP were not statistically different in PBMC and T-cells. The ddATP concentration was significantly higher in T-cells than in PBMC cells (P = 0.011). 2 The intracellular concentrations of ddl, ddAMP, ddADP, and ddATP were statistically higher in PBMC cells than in m onocytes with P-values P = 0.003, P = 0.002, P = 0 .0 1 9 , and P < 0.001, respectively. 3 The intracellular concentrations of ddl, ddAMP, ddADP, and ddATP were statistically higher in T-cells than in m onocytes w ith P-values P < 0 .0 0 1 , P = 0.003, P < 0 .0 0 1 , and P < 0 .0 0 1 , respectively. c o c Q > O c o o < ~o *D 4 5 40 35 30 25 20 15 10 5 I I PBM C V 7 7 X T -cells ■■M onocytes PBMC T-cells Monocytes Figure 32. Intracellular ddATP concentrations in PBMC cells, lymphocytes, and monocytes following a 1 hour incubation with a radioactive mixture of 1 mM ddl. The ddATP concentration in monocytes w as significantly lower than in PBMC cells (P < 0.001) and lymphocytes (P < 0.001). 125 1.50 4~ 1.00 • 0.50 ■ 0.00 0.00 0.50 1.00 1.50 Figure 33. A typical HPLC chromatogram showing separation of AZT. AZT was eluted at 13.18 minutes. 126 Concentration of AZT a n a b o lite s, nM 10000 1000 100 10 1 Figure 34. Comparative anabolism of AZT in the Jurkat/0 and Jurkat/AZT-10 cells. The intracellular AZTMP, AZTDP, and AZTTP concentrations were significantly higher in the Jurkat/0 cells than in the Jurkat/AZT-10 cells with P < 0.001, P =0.036, and P =0.003, respectively. ■ f i ! I lJu rk a t/0 V 7 7 X Jurkat/AZT-10 1 1 AZT AZTMP AZTDP AZTTP 127 intracellular levels of AZT, AZTMP, AZTDP, and AZTTP following a 2-hour incubation of 1x107 cells (Jurkat/0 and Jurkat/AZT-10) at 37 °C with a 2.4 nM radioactive mixture of AZT of specific activity 8610.24 ± 528.79 cpm/pmol. Intracellular activation of AZT to its anabolites was statistically greater in the Jurkat/0 than the Jurkat/AZT-10 cell lines (Table 12). The concentrations of the anabolites AZTMP, AZTDP, and AZTTP were 1630.01 ± 199.74 nM, 117.58 ± 33.02 nM, and 30.43 ± 4.87 nM, respectively, in the Jurkat/0 cell line and 580.52 ± 53.05 nM, 58.01 ± 3.14 nM, and 12.55 ± 1.06 nM, respectively, in the Jurkat/AZT-10 cell line. The differences in the concentrations of the AZT anabolites in the Jurkat/0 and Jurkat/AZT-10 cell lines could not be attributed to differences in the intracellular amount of the parent drug, AZT, in these two cell lines. Therefore, these results verified previous observations of reduced activation of AZT in the partially AZT resistant cell line Jurkat/AZT-10. Hence, the Jurkat/AZT-10 T-cell line is partially resistant to AZT. Both AZT and ddl anabolism were investigated in these two cell lines to directly compare the anabolism of these two drugs. The intracellular concentrations of the ddl anabolites, ddAMP, ddADP, and ddATP were 274.98 ± 40.52 nM, 28.08 ± 1.91 nM, and 54.17 ± 6.90 nM, respectively, in the Jurkat/0 cell line and 283.46 ± 44.10 nM, 29.24 ± 0.44 nM, and 47.23 ± 9.50 nM, respectively, in the Jurkat/AZT-10 cell line (Figure 35). There were no statistical differences between the corresponding ddl anabolite concentrations 128 T ab le 1 2 : C o m p arativ e a n a b o lism o f AZT in th e J u rk a t/0 a n d Ju rk a t/A Z T -1 0 cell lines. Anabolite AZT (nM) AZTMP (nM) AZTDP (nM) AZTTP(nM) Jurkat/0 6052.81 ± 2481.86 1630.01 ± 199.74 117.58 ± 33.02 30.43 ± 4.87 (n = 3) Jurkat/AZT-10 4638.28 ± 361.67 580.52 ± 53.05 58.01 ± 3.14 12.55 ± 1.06 (n = 3) P-value NS <0.001 0.036 0.003 Concentration of ddl a n a b o lite s, nM 1000 100 10 1 Figure 35. Comparative anabolism of ddl in Jurkat/0 and Jurkat/AZT-10 cell lines. There were no significant differences in ddl anabolism between Jurkat/0 and Jurkat/AZT-10 cells. I I J u rk a t/0 7 7 7 % Jurkat/AZT-10 i ddl ddAMP ddADP ddATP 130 in the Jurkat/0 and Jurkat/AZT- 10 cell lines (Figure 35). These results indicate that cellular resistance to AZT does not confer cross-resistance to ddl. V.10. Anabolism of ddl In vitro in Jurkat T-cell lines in the absence and presence of AZT The anabolism of ddl in Jurkat/0 and Jurkat/AZT-10 cells in the presense and absence of AZT was performed in two independent experiments. In the first experiment, the two cell lines were incubated with ddl alone or with a mixture of ddl and AZT. The average levels of ddATP in Jurkat/0 ceils were 51.95 ± 5.00 nM (mean ± SD, n=6) after one-hour incubation with ddl alone and 49.41 ± 4.53 nM (mean ± SD, n=6) in the co-presence of 1 mM AZT (Figure 36). The average ddATP levels in Jurkat/AZT-10 were 54.45 ± 12.98 nM (mean ± SD, n =6) in the presence of ddl alone and 53.88 ± 3.65 nM (mean ± SD, n=4) in the presence of both ddl and AZT (Figure 36). The concentrations of ddl anabolites in Jurkat/0 and Jurkat/AZT-10 cells were not statistically different in the presence of AZT (Table 13). In order to examine the effect that pretreatment with one drug could have on the anabolism of the other drug, I examined the anabolism of ddl after a two- hour pre-incubation of the Jurkat/0 and Jurkat/AZT-10 with 1 nM AZT. The average ddATP levels were 57.47 ± 7.08 nM and 55.19 ± 7.11 nM, respectively 131 Cone, of ddl anabolites, n M Cone, of ddl anabolites, nM 1000 100 ddl ddAMP ddADP ddATP 1000 100 ddl ddAMP ddADP ddATP Figure 36. Mono-, di-, and triphosphate levels of ddl anabolites in the Jurkat/0 (top graph) and Jurkat/AZT-10 (bottom graph) cell lines after a 1-hour incubation with 1 mM ddl, in the absence and presence of 1 mM AZT. The presence of AZT did not significantly alter the anabolism of ddl. 132 133 Table 13: Anabolism of ddl in human Jurkat/0 and Jurkat/AZT-10 cells using a radioactive mixture of 1//M 3H-ddl or a mixture of 1 jt/M 3H-ddl and 1 //M AZT. Anabolite1 ddl (nM) ddAMP (nM) ddADP (nM) ddATP(nM) Jurkat/0 (ddl alone) 1637.35 ± 135.27 244.55 ± 42.88 25.51 ± 3.42 51.95 ± 5.00 Jurkat/0 (ddl + AZT) 1705.70 ± 110.85 242.01 ± 32.95 23.94 ± 1.05 49.41 ± 4.53 Jurkat/AZT-10 (ddl alone) 1728.18 ± 236.20 281.26 ± 42.24 28.03 ± 1.68 54.45 ± 12.98 Jurkat/AZT-10 (ddl + AZT) 1877.48 ± 64.21 255.12 ± 17.39 25.00 ± 2.15 53.88 ± 3.65 'No statistical differences were observed between the anabolites of ddl in Jurkat/0 or Jurkat/AZT- 10 cell lines following incubation of these tw o cell lines with ddl alone or a mixture of ddl and AZT. Activation of ddl to its anabolites w as not statistically different in the Jurkat/0 and Jurkat/AZT-10 cell lines. (Table 14). Pre-incubation of these cells with AZT did not augment the activation of ddl to its anabolites (Figure 37). V.10. Development of a mathematical model to explain the synergistic inhibition of HIV-1 by ddl and AZT. Since my pharmacokinetic and pharmacodynamic results showed no apparent interactions between ddl and AZT either in plasma or in T-cells, the question remained as to where these two drugs might interact synergistically as the biology data suggested. Therefore, I proceeded to investigate whether Jackson’s equation describing the rate of DNA synthesis by DNA polymerase, using the four natural triphosphate substrates dCTP, dATP, dTTP, and dGTP, could be adapted to provide a mathematical model for the inhibition of HIV-RT by AZTTP and ddATP. Jackson’s equation was adapted to describe the kinetics of a DNA polymerase-like enzyme, such as the HIV-RT, in the presence and absence of one or more inhibitors, as in the Michaelis-Menten equation for competitive inhibition of an enzyme. In the Michaelis-Menten equation, a competitive inhibitor competes with the normal substrate for binding to the active site of the enzyme. In the presence of a competitive inhibitor, the maximum velocity, Vm ax, of the enzymatic reaction remains the same. However, the apparent Michaelis-Menten constant, K,,,, increases, in equation 3 7 ,1 considered AZTTP 134 Table 14: Anabolism of ddl in Jurkat/0 and Jurkat/AZT-10 cells following a 2 hour pre-incubation with 1 jvM AZT. Anabolite1 ddl (nM) ddAMP (nM) ddADP (nM) ddATP(nM) Jurkat/0 185B.94 ± 191.88 343.49 ± 27.08 36.83 ± 2.60 64.98 ± 5.60 Iddl alone) Jurkat/0 1776.07 ± 177.15 302.38 ± 65.17 36.48 ± 6.82 57.47 ± 7.08 (AZT pre incubation) Jurkat/AZT-10 1438.98 ± 137.49 194.29 ± 91.09 26.16 ± 5.20 56.73 ± 5.65 (ddl alone) Jurkat/AZT-10 1535.26 ± 293.72 181.05 ± 92.68 25.48 ± 10.48 55.19 ± 7.11 (AZT pre incubation) 1 No statistical differences were observed between the anabolites of ddl in Jurkat/0 or Jurkat/AZT-10 cell lines when the cells were pre-incubated with 1 //M AZT for 2 hours. C one, of ddl anabolites, n M C one, of ddl a n a b o lite s , nM I ino A Z T ( 2 2 A Z T pre- mcubation 1000 100 ddl ddAMP ddADP ddATP I Ino AZT £223 A Z T p re- incubation 1000 100 d d l ddAMP ddADP ddATP Figure 37. ddl anabolism in Jurkat/0 cells (top graph) and Jurkat/AZT-10 cells (bottom graph) following a 2 hour pre incubation with a 1 nM AZT. Pre-incubation with AZT did not significantly alter the anabolism of ddl. 1 36 acting as a competitive inhibitor of dTTP, thereby adversely influencing the rate of viral DNA synthesis, i.e., decreasing it. In equation 38, both AZTTP and ddATP were considered to act as competitive inhibitors of HIV-RT. The variables K , a n d K , d d A T P in equations 37 and 38 represent the inhibition constants for AZTTP and ddATP against RT, respectively. [ d T T P ] [ d A T P ] [ d C T P ] [ d G T P ] v K .., K.,. K.,c K,.t = ----------------------------------------x -----------------------x ------------------------ x ------------------------- [Eq.37] V „ , [ d T T P ] [ A Z T T P ] [ d A T P ] [ d C T P ] [ d G T P ] 1 + + 1 + 1 + 1 + ----------- K ... K.,. K,.c K... [ d T T P ] [ d A T P ] [ d C T P ] [ d G T P ] v K .., K ... K . , c K . , t ----------------------------------------x --------------------------------------- x -----------------------x ----------------------- [Eq.38] V „ , [ d T T P ] [ A Z T T P ] [ d A T P ] [ d d A T P ] [ d C T P ] [ d G T P ] 1 + + 1 + + 1 + 1 + ----------- K..T K1 iW T 1 P K 'tM A 1 p Kt> c K.,6 Equation 38 suggests that in the presence of both inhibitors, AZTTP and ddATP, the velocity of the reaction catalyzed by HIV-RT becomes much smaller than Vm ax, i.e, v < < Vm ax. On a biochemical basis, a synergistic inhibition of the HIV-RT by AZTTP and ddATP could occur if the drugs AZT and ddl were activated to their respective anabolites AZTTP and ddATP at concentrations higher than those obtained in the presence of either drug alone. However, the inhibition of HIV-RT by AZTTP and ddATP could be synergistic even in the absence of increased activation 137 of either AZT or ddl to their respective active anabolites as a result of the simultaneous presence of ddATP and AZTTP (Grindey et al., 1979, and Harrap et al., 1975). The pharmacodynamic studies of ddl in Jurkat/0 and Jurkat/AZT-10 cell lines indicated that the activation of ddl to ddATP was not altered by the simultaneous presence of 1 /jM AZT or following a 2 hour pre-incubation with 1 mM AZT. Hao et al. determined the K„, of the naturally occuring deoxynucleotide triphosphates as well as the K , constants of the inhibitors AZTTP and ddATP against human HIV-RT (Hao et al., 1988). In addition, Hao et al., investigated the effect of AZT and ddl on the intracellular levels of endogenous dNTPs and found that drug concentrations in the range of 0-50 mM did not cause significant alterations in the dNTP pool sizes in H9, MOLT4, CEM, and ATH8 cells. The values of K „, and K , reported by these investigators and the finding that the dNTP pools were not altered in the presence of pharmacologically effective drug concentrations, were utilized to predict the decrease in the rate of viral DNA synthesis in the presence of one and two HIV-RT inhibitors, by comparison to the rate of the uninhibited reaction. The K,,, values of HIV-RT for dTTP and dATP were 1.25 ± 0.06 mM and 11.2 ± 2.9 nM, respectively. The K j values of HIV-RT for the inhibitors AZTTP and ddATP were 0.10 ± 0.02 nM and 0.22 ± 0.12 mM , respectively (Hao et al., 1988). The 138 average intracellular concentrations of dATP and dTTP were 2 /xM and 4 /xM , respectively. In the presence of AZT alone, the rate of the inhibited HIV-RT (vjn h jb jted ) as compared to the rate of the uninhibited RT (vu n jn h ib ited ) was predicted by equation 39, after dividing equation 37 by the uninhibited rate and simplifying the solution: 1 + [dTTP] inhibited ^M ,T ------------ = ----------------------------------- [Eq.39] ^uninhibited 1 + IdH .Pl + [AZTTP] ^M .T ^ i,A Z T T P Similarly, in the presence of ddl alone, the decrease in the velocity of HIV-RT due to the inhibitor ddATP was: 1 + fdATPI inhibited ^M ,A ------------ = ----------------------------------- [Eq.40] ^uninhibited 1 + fdATPI + fddATP] ^M .A ^i, d d A T P In the presence of both AZT and ddl, the decrease in the velocity of the inhibited HIV-RT could be predicted by equation 41: 1 + fdTTPl 1 + fdATPI inhibited » < mj K m> a x [Eq 41] V u n in h ib ited 1 + WTTP] + [AZTTP] 1 + jdATP] + [dd_ATP] K m.t ^ ,A Z T T P ^M ,A ^i.ddA T P 139 Equations 39-41 indicate that the decrease in the velocity of HIV-RT becomes greater with increased concentrations of the inhibitors) and concurrent decreased concentrations of the natural substrates, the nucleoside triphosphates dATP and dTTP. According to equations 39 and 40 and using the reported K ,,, and K , values, the ICM values (i.e., inhibitor concentrations required to inhibit the enzyme RT by 50%) of AZTTP and ddATP were 0.42 nM and 0.26 mM, respectively. In the presence of both AZTTP and ddATP, at concentrations one-half of their respective IC^ values, vin h j b ite d was 44.4 % of V u n in h ib ited and not 50% as would be expected from independence, suggesting a synergistic interaction on HIV-RT by AZTTP and ddATP. Hence, this biochemical consideration could explain and calculate the synergistic interaction between AZTTP and ddATP against HIV-RT. Table 15 describes the reduction in the velocity of the RT-catalyzed reaction (vinhibned/vuninhibited)in relation to varying concentrations of the inhibitors ddATP and AZTTP, which had values 0.5, 1, 2, and 10 times their respective IC^ values. In the presence of one inhibitor, at a concentration equal to its ICg, value, vln h lb lt0 d was, by definition, 50% of vu n in h lb it# d . In the presence of both inhibitors, at concentrations equal to one-half their respective IC^ values, ^inhibited was 44.4% of vu n in h ib it# d , suggesting a synergistic inhibition of HIV-RT by AZTTP and ddATP. At an inhibitor concentration (AZTTP or ddATP) equal to 2 times its IC^ value, vin h ib lt# d was 33.3% of vu n in h ib itad . When both AZTTP ddATP Concentration Table 15: Extent of inhibition of HIV-RT (v,n h jb i,^ /v u n in h ib it^ ) by various concentration combinations of AZTTP and ddATP. AZTTP Concentration Cone.1 0 0 . 5 l C s o '<V _ 8 o C M 10ICM 0 1 0.667 0.500 0.333 0.091 0.5ICm 0.667 0.444 0.333 0.222 0.061 ic» 0.500 0.333 0.250 0.167 0.046 IO O 1 5 0.333 0.222 0.167 0.111 0.030 IO IC j o 0.091 0.061 0.046 0.030 0.008 ’The first row represents AZTTP concentrations equal to 0 ,0 .5 ,1 , 2 and 10 times the ICM value of AZTTP against HIV-RT. The first column represents ddATP concentrations equal to 0 ,0 .5 ,1 , 2, and 10 times the ICg, values of ddATP against HIV-RT. The numbers in this table represent values of v in h ib it.d /v uninhibit.d which correspond to the various AZTTP and ddATP concentrations. i 141 and ddATP were present, at concentrations equal to their respective ICg, values, vln h ib M d was 25% of vu n ln h ib ito d , which again indicated a synergistic interaction between AZTTP and ddATP against HIV-RT. The synthesis of viral DNA by RT could also decrease by reducing the dNTP pools, i.e., by decreasing the concentrations of the natural triphosphate substrates dATP, dGTP, dCTP, and dTTP. Table 16 shows the decrease in the velocity of the RT-catalyzed reaction in comparison to Vm ax , i.e., v/Vm ax. In order to calculate v/Vm ax, the intracellular concentrations of dGTP and dCTP were taken to be both equal to 4/xM while the K ^, values were 17.4 ± 0.6 nM and 6.5 ± 3.1 /jM, respectively (Hao et al., 1988). The concentrations and K ,,, values of dATP and dTTP were as described above. The ratio v/Vm ax was calculated to be 0.0083 when considering the normal intracellular dNTP concentrations, 0.373 when the dNTP concentrations were 10 times higher their normal levels, and 5.6x1 O '6 when the dNTP levels were 10 times lower than their normal levels. Thus, the mathematical models described in this section provided an explanation for the synergistic inhibition of HIV-RT by AZTTP and ddATP on a biochemical basis and suggested that greater inhibition of the HIV-RT could be achieved if AZT and ddl were administered concurrently with a drug that depletes the dNTP pools. 142 Table 16: Effect of dNTP pools on the velocity of the RT-catalyzed reacton. dNTP Cone. 0.1 1 10 v/Vm ax 5.6x10-® 0.0083 0.373 143 VI. DISCUSSION The complexity of HIV-1 infection and the inability, thus far, to find a curative therapy against HIV-1 has prompted many scientific efforts to focus on the development of antiretroviral therapies which increase the survival and improve the quality of life of HIV-infected individuals. In this project, I investigated the individual and population pharmacokinetics of ddl in pediatric patients enrolled on two clinical trials. The main objectives of the project were to examine whether there is a drug-drug interaction in plasma between ddl and AZT when administered concurrently, whether the presence of food and differences in the developmental stages of pediatric patients influence the PK behavior of ddl in this patient population, and whether there is a synergistic augmentation in the activation of ddl to ddATP in the presence of AZT, in human T-cells. Three years ago when these studies were initiated, the clinical trials evaluating the efficacy of ddl against HIV-1 were conducted in adult patients who differ from pediatric patients in terms of body water/fat compartments, plasma protein binding, activity of hepatic enzymes, and the kidney’s ability for eliminating processes. Furthermore, the clinical trials designed to investigate ddl, until that time, did not extensively study the influence of food or the effect of different developmental stages of the patients, on the pharmacokinetic behavior of ddl. In addition, none of the studies with ddl and AZT had 144 explained the observed synergism between these two drugs, in HIV-infected patients and T-cells. The pharmacokinetics of ddl were evaluated in a phase I clinical trial (ACTG #176) in pediatric patients with HIV-1 infection, receiving a concurrent combination treatment with ddl and AZT. This study also evaluated the pharmacokinetics of AZT in combination with ddl. Prior to analysis of study samples, assay methods of ddl had to be evaluated. The HPLC assay which had been previously used, had significant limitations in the lower ranges of detection of drug in plasma and in the quantity of plasma necessary for the assay. Since study subjects were children, the availability of plasma was limited. The RIA method produced impressive accuracy and reproducibility requiring only 10 nL of plasma to assay ddl. This method was later used by the ACTG #144 phase ll/lll clinical trial to assay plasma ddl. AZT was also assayed by a radioimmunoassay method which has been used for many years (Henry et al., 1988). Patients enrolled on the ACTG #176 clinical trial were administered a single dose of AZT on day 1, a single dose of ddl on day 2, and single doses of AZT and ddl on day 3. Starting on day 4, AZT was administered every 6 hours and ddl was administered every 12 hours due to the much longer half-life of ddATP 145 as compared to that of AZTTP. The PK data of ddl and AZT from this clinical trial were analyzed by the conventional Standard Two-Stage (STS) method and by the population pharmacokinetics program NONMEM. The computer software PCNONUN was used in the STS method to provide PK parameter estimates for each patient in the study. The pharmacokinetic profiles of both drugs were adequately described by the one-compartment model with first- order absorption. Earlier studies of ddl (Hartman et al., 1991) and AZT (Gitterman et al., 1990) also described the plasma elimination of these drugs as monoexponential while other studies suggested a biexponential elimination (Balis et al., 1989, Balis et al., 1991). Biexponential elimination of ddl was observed primarily in patients who received a high dose of ddl, which w as 6.4 mg/kg or 320 mg for an adult patient of 50 kg and 448 kg for a patient of 70 kg (Hartman et al., 1990). The estimated PK parameters of ddl and AZT were characterized by a wide variability, as indicated by the large values of the standard deviations. This could be due to the variable absorption kinetics of ddl and/or due to differences in the developmental stages of the pediatric population studied (Pizzo, 1990). The absorption rate constant was the most variable PK parameter of ddl. The frequency distribution of this parameter could not be described either by the normal or the log-normal distribution. Absorption of ddl was very fast in some patients and very slow in others. The variability in 146 the absorption rate constant of ddl and AZT among patients could probably be attributed to interindividual differences in gastric transit time. The inability to provide a reliable estimate for this parameter by the STS method might have been due to the limited number of PK samples obtained in this study (3 PK samples + pre-treatment sample). The average values of the absorption half- lives on day 2, when ddl w as administered alone, and on the days when ddl w as administered in combination with AZT were expressed both a s arithmetic and geometric means, as shown in the results section. The t-test performed on the actual data or th e log-transformed data indicated no statistical differences between the absorption half-lives of ddl in the absence and presence of AZT. Thus, there was no interaction between ddl and AZT at the absorption site. The frequency distributions of the other PK parameters of ddl, such as elimination rate constant, clearance, and volume of distribution were log- normally distributed. Thus, the average values of these parameters were expressed a s geometric means. Application of the t-test on the log- transformed values of these PK parameters indicated that the presence of AZT did not significantly alter th e pharmacokinetic profiles of ddl. These results, therefore, indicate that AZT does not influence the pharmacokinetic behavior of ddl in pediatric patients who receive ddl and AZT concurently. 147 Statistical comparisons between plasma ddl AUC values were performed using the Wilcoxon Rank Sum test for independent samples. The median AUC of ddl on day 2 (ddl alone) was not statistically significantly different from the median AUC of the days of drug co-administration, at any dose level of ddl. The 60 m g/m 2 ddl dose level had a statistically significantly lower AUC value than the 90 mg/m2 (P=0.008), the 135 mg/m2 (P =0.002), and the 180 m g/m 2 (P=0.0006) dose levels. No other significant differences w ere found between the AUC values of other dose levels (Table 3). The inability to detect any statistical differences among the AUC values of the other ddl dose levels was probably due to the large variability associated with the pharmacokinetics of this drug in pediatric patients. This became more apparent when a few patients showed no ddl concentrations in the plasma specimens even with the very sensitive RIA assay. As determined earlier, the absence of significant differences between the AUC, absorption half-life, and elimination half-life of day 2 and on the days of ddl and AZT co-administration indicated that AZT does not influence the pharmacokinetics of ddl in pediatric patients with HIV-1 infection. Similar results of no interaction between AZT and ddl have been reported in jn vivo studies in animal models (Wientjes et al., 1992) and in adults receiving a combination treatment of ddl and AZT (Collier et al., 1993, Yarchoan et al., 148 1994). T hus, it can b e concluded that ddl and AZT do not interact pharmacokinetically in mammalian system s. The pharmacokinetics of ddl in HIV-1 infected children receiving a combination treatment of ddl and AZT were also investigated by the computer program NONMEM. As in the STS method, the data were described by the one- compartment model with first-order absorption. The interindividual variabilities of the param eters ka, Cl, and Vd were described by the constant coefficient of variation model (CCV) which is appropriate for skewed or log-normal parameter distributions. The CCV model was chosen over the additive type error model based on the results of the STS analysis of the pharmacokinetic data, i.e., th e frequency distributions of the PK parameters. The choice of the CCV model w as verified by performing two separate NONMEM analyses of the data, using th e additive type error model in one analysis, and the CCV model in the other. The NONMEM analysis of the data w as statistically better (P < 0.01) with th e CCV error model than with the additive type model. Intra individual variability was also described by the CCV model. A ll pharmacokinetic profiles were regarded to be obtained from separate individuals although more than one PK evaluation was performed for m ost individuals. This was necessary in order to stabilize the NONMEM analyses and provide meaningful estimates for the elements of the OMEGA matrix, i.e., 149 the random interindividual variabilities associated with the PK (fixed) parameters, due to the limited number of subjects in this study. The NONMEM parameter estimates of ddl were similar overall to those obtained by the STS analysis (Table 4). The estimates of volume of distribution and elimination half-life of ddl by NONMEM were greater than those obtained by the STS method, but not different. In addition, the PK (fixed) parameters of ddl were estimated with greater confidence by NONMEM than by the STS method as indicated by the narrower confidence intervals of the parameters obtained by NONMEM. This was probably due to the fact that the ACTG #176 clinical trial was a limited pharmacokinetic study. The STS method is known to provide good estimates for fixed effects parameters when the number of data per individual is large, but not when the PK data are limited. In fact, a simulation PK study comparing the STS method to NONMEM showed that the STS estimates of fixed effects parameters are as good as, or even slightly better, than the NONMEM estimates, if the number of data per individual is large, because this method does not use approximations (i.e., linearization of the model). However, regardless the number of data per individual, the STS method usually overestimates random interindividual variability with an upward bias because each individual parameter is estimated from the original drug plasma-time data with some error which is not biological in origin. On the other hand, NONMEM provides estimates of random interindividual effects which are relatively unbiased and more precise than the STS estimates, though they can be highly imprecise if the number of individuals in the study is small (Sheiner et al., 1981, Sheiner et al., 1983, Ludden, 1988). The findings of this simulation study support the results obtained by the STS and NONMEM analyses of the ddl data provided by the ACTG #176 clinical trial. The fixed effects parameters ka, Cl, and Vd were estimated with greater confidence by the NONMEM method than by the STS method due to the limited number of PK samples per individual in this study. The random interindividual effects were also estimated with greater precision by NONMEM. Interindividual variability of the parameter ka was very large by both methods. The coefficient of variation (%CV) of ka was 136.5% and 113.1% by the STS and NONMEM analyses, respectively. The coefficients of variation of clearance and volume of distribution determined by the STS method were 64.6% and 86.8%, respectively. The NONMEM estimates of %CV of these two parameters were 56.4% and 52.8%, respectively, which are smaller in magnitude than the corresponding STS estimates. One pharmacokinetic study of ddl in adult patients, performed by the STS method, provided values for the elimination half-life of 37 ± 4 min when the one-compartment model was used to fit the data and a terminal half-life of 71.8 ± 4.5 minutes using the two-compartment model (Hartman et al., 1991). A NONMEM analysis of ddl administered in adult patients estimated a terminal elimination half-life for ddl of 29.71 minutes using the two-compartment model with first-order absorption (Drusano et al., 1992). The NONMEM analysis of ddl by Drusano et al. was performed by describing inter- and intraindividual variabilities by the constant coefficient of variation model, as with the ACTG #176 data. The volume of distribution of ddl in adult clinical trials ranged from 18.0 ± 2.1 L (Drusano et al., 1992) to 0.83 ± 0.06 L/kg or 58.1 L for a 70-kg individual (Hartman et al. 1991). The clearance of ddl in adult patients with HIV-1 infection was reported to be 35.2 ± 4.2 L/h (Drusano et al., 1992) and 0.75 ± 0.06 L/kg/h or 52.5 L/h for a 70 kg individual (Hartman et al., 1991). In pediatric clinical trials with ddl i.v. monotherapy, the elimination (terminal) half-life of ddl averaged 1 hour, the steady-state volume of distribution averaged 24 ± 14 L/m2, and th e total clearance averaged 30.6 ± 10.8 L/h/m 2 (Balis et al., 1992). Absorption of orally administered ddl was fast in all clinical trials with ddl; the time to reach peak concentration was less than 0.5 h. The pharmacokinetics of AZT were also adequately described by th e one- compartment model with first-order absorption. A study by Balis et al. evaluating the pharmacokinetics of intravenously and orally administered AZT, in pediatric patients with HIV-1 infection, reported a biexponential plasma elimination for AZT probably due to the larger number of pharmacokinetic samples per individual in this study and more frequent blood sampling (Balis et al.1989). On the other hand, a NONMEM PK evaluation of orally administered AZT in adult patients modeled the kinetics of this drug by the one-compartment model with first-order absorption, because the rapid oral absorption of AZT and its short distribution (a) phase made estimation of the a-phase by the two-compartment model unreliable (Gitterman et al. 1990). The AZT PK data of the ACTG #176 clinical trial were limited in number (3 samples + pre-treatment sample per individual), thus, the limited sampling could not support a pharmacokinetic model more complicated than the one- compartment model with first-order absorption. Twelve AZT PK profiles from a total of 56 could not be modeled by PCNONLIN due to the limited number of data per individual and the poor quality of these data. The profiles that could not be fitted by PCNONLIN were associated with a much slower absorption of the drug. These data, however, were included in the NONMEM analysis of AZT pharmacokinetics, since NONMEM does not require estimation of individual PK parameters. Estimation of the absorption rate constant of AZT by the STS method was highly imprecise with an upward bias. The upward bias could partly be attributed to the fact that most of the PK profiles, which could not be modeled by PCNONLIN, were characterized by slower absorption kinetics than the rest of the PK profiles. These profiles were not included in the estimation of the average ka in this population. As a result, the absorption rate constant of AZT was estimated to be greater than its true value, i.e., with an upward bias. The frequency distributions of the other pharmacokinetic parameters of AZT, i.e., elimination rate constant, plasma clearance, and apparent volume of 153 distribution were lognormal. Thus, the average values of these parameters were expressed as geometric means. Co-administration of ddl with AZT did not significantly alter the PK parameter estimates of AZT as indicated by the t-test on the log-transformed parameter estimates. In addition, the Wilcoxon Rank-sum test indicated no significant differences between the AUC values of AZT, per dose level, on day 1 when AZT w as administered alone and on the days when AZT and ddl were administered in combination. The PK data of AZT were also analyzed by NONMEM. The NONMEM population estimates of the basic PK parameters ka, Cl, and V d (mean ± SE) were described in Table 8. The NONMEM estimates for the fixed effects parameters of AZT were closer in magnitude to values reported in the literature than the STS estimates. A phase I clinical trial of AZT in pediatric patients with HIV-1 infection reported values (mean ± SD) for the clearance and steady-state volume of distribution of AZT of 38.5 ± 9.7 L/m2 and 45 + 28 L/m2, respectively (Balis et al., 1989). The study by Balis et al. did not provide an estimate of the absorption rate constant of AZT. A different study, evaluating the pharmacokinetics of AZT in adult patients, estimated the clearance and volume of distriburion of AZT to be 1.3 ± 0.1 L/kg/hr and 3.0 ± 0.6 L/kg, respectively. Absorption of AZT in this study w as very rapid with an absorption rate constant of 6.3 ± 2.67 h'1 . The estimate of ka by the STS method appears to be biased upward, for the sam e reasons as described previously for ddl. The estimates of clearance and volume of distribution were smaller than those reported in the literature. The precision of the estimated interindividual variabilities was greater by NONMEM than by the STS method. The coefficients of variation of clearance and volume of distribution were 69.9% and 78.1% by the STS method and 46.7% and 50.6% by NONMEM, respectively. Investigation of a possible PK interaction between ddl and AZT in plasma by the NONMEM method showed that the concurrent presence of AZT in plasma or the simultaneous plasma levels of AZT did not alter the pharmacokinetic behavior of ddl. In addition, the pharmacokinetic parameters of ddl in patients who had been on AZT monotherapy prior to their entry onto the ACTG #176, a combination study of ddl and AZT, were not statistically different from those in patients who had never received prior AZT monotherapy. The population pharmacokinetic studies on ddl and AZT, both with the STS and NONMEM methods, during a combination treatment with these two drugs lead to the conclusion that the clinical improvement observed in the pediatric patients on the ACTG #176 clinical trial could not be attributed to increased plasma concentrations of either ddl and AZT or changes in any other PK parameter. 155 The NONMEM analysis of the pharmacokinetic data of the ACTG #144 clinical trial provided estimates of PK parameters which were in the same order of magnitude as reported values in the literature for pediatric patients. The data set could be described adequately by the one-compartment model with first- order absorption. There was no statistical difference between the one- and two-compartment models. The concentration-time scatterplots (figures 29 and 30) indicated that the two-compartment model could better describe the PK profiles of ddl in the fasting state. This was especially obvious in the scatterplot of the 150 mg/m2 dose level of ddl under fasting conditions. However, the one-compartment model with first-order absorption appeared to be the best model for fitting the ddl PK data of the fed state. The presence of food delayedthe absorption of ddl, thereby making the a-absorption phase indistinguishable. The graphic representation of the data was in agreement with the results of NONMEM analyses which indicated that the absorption rate constant of ddl was much smaller in the presence of food than in the fasting state. It is clear from the NONMEM analyses and the predicted average A U C ^ values that the extent of absorption, i.e. bioavailability of ddl, did not depend on whether this drug was administered with or without food. Balis et al. (Balis et al., 1992) reported that the A U C ^ of ddl in pediatric patients following administration of 60 m g/m 2 ddl was 1.5 juM-hr or 354.3 (ng/mQ-hr. This value is similar to the average predicted AUC value of ddl by the NONMEM analyses which was 428.82 (ng/ml)tir following the administration of 50 mg/m 2 ddl under fasting conditions. The decrease in the absorption rate constant of ddl could be due to a delay in gastric emptying caused by the presence of food. A decrease in the absorption rate constant, but not the extent of absorption, had been observed during administration of other drugs with food, such as digoxin (Gibaldi et al., 1983). Previous reports in the literature, on the pharmacokinetics of ddl in adult patients, indicated that the extent of absorption and not the rate of absorption of ddl decreases in the presence of food (Shyu et al., 1991). The disagreement between the reported results and the results of this analysis could be due to the fact that different drug formulations were used in the two studies, the patient populations in the two studies were different, and the disease status of the patients in these two studies might had been different. In the study by Shyu et al. (Shyu et al., 1991), ddl was administered as a chewable tablet, whereas in the ACTG #144 study ddl was administered as a buffered solution. It is possible that food interfered with tablet disintegration (dissolution), therefore, preventing complete release of ddl from the formulation. The fact that in the study by Shyu et. al. no difference was observed between the absorption rate constants of ddl in the fasting and fed states is probably because this study evaluated only 8 patients. The ACTG #144 clinical trial evaluated 90 patients, therefore, the results of this trial are 157 more statistically significant, especially for a drug such as ddl which is characterized by a wide pharmacokinetic variability. The final model which described the pharmacokinetics of ddl in the ACTG #144 clinical trial indicated that the absorption rate constant of ddl decreased in the presence of food, that plasma clearance decreased linearly with increasing age, and that the apparent volume of distribution decreased linearly with increasing weight. The estimated volume of distribution of ddl had a value that is similar to total body water (»42 L). Pharmacokinetic studies with ddl in rats also indicated that the estimated volume of distribution of this drug had a value close to that of total body water in rats (Wientjes et al., 1992). The decrease in clearance with age and volume of distribution with weight indicates that ddl plasma concentrations are higher in older than in younger children. This is probably related to the fact that total body water decreases with increasing age and weight (Friis-Hansen, 1961). According to the final model (Table 10, model 6), the predicted concentrations ± SD for a patient of average weight and age in this study, included the majority of the observations. This was graphically depicted in figures 29 and 30. For patients of greater age and weight, model 6 predicted a shift in the predicted concentration-time profiles towards higher concentrations. A small number of observations lied outside the 68% confidence intervals of the graphs in figures 29 & 30. In figure 29A, the PK profile of one individual revealed the presence of enterohepatic 158 recycling which tends to increase drug plasma concentrations at times when the drug should be eliminating. Two other individual PK profiles with plasma concentrations higher than predicted by the 68% confidence interval belong to two male patients. One of these two patients was of age and weight above average while the other patient was of age and weight below average. The serum creatinine levels of these three patients were within the normal range. Therefore, the higher observed plasma concentrations in these patients could not be attributed to differences in renal function. Figure 29B indicates that the observed ddl plasma concentrations were much higher than predicted in four patients who had been administered 50 mg/m2 ddl with food. Assuming that these patients were in fact administered the correct ddl dose, the higher plasma concentrations in these patients could not be attributed to differences in gender, weight, age, and serum creatinine. Two of these patients were male and two were females. The serum creatinine levels of all four patients were very similar and within the normal range. Three of the four patients were of age and weight above average. One of these patients was of age and weight below average. Therefore, it can be concluded that individual differences in the extent of absorption of ddl contribute greatly to the wide range of observed ddl plasma concentrations. Overall, NONMEM analyses provided meaningful estimates for the PK parameters of ddl and for the inter- and intraindividual variabilities of the 159 plasma concentrations. In addition, the NONMEM analyses of the PK data from the ACTG #176 clinical trial verified reports in the literature regarding the superiority of this method over the STS method in cases with limited PK sam ples per individual. A comparison of the NONMEM PK parameters of ddl in the ACTG #176 and the ACTG #144 studies indicates that the estimates of the absorption rate constant and clearance were similar in the two studies (Tables 4 and 10). The estimated volume of distribution, however, was larger in the ACTG #144 study (Vd = 43.0 L/m2 ) than in the ACTG #176 study (Vd = 18.0 L). Consequently, the ddl plasma concentrations were higher in the combination study. It is possible that bioavailabilty (extent of absorption) of ddl was greater in the combination study than in the ACTG #144 study. In the ACTG #176 study, antacid was administered two minutes prior to the oral administration of ddl. In the ACTG #144 study, ddl was dissolved in the appropriate amount of antacid, hence, ddl and antacid were administered simultaneously. Administration of antacid prior to the administration of ddl, might have resulted in more efficient neutralization of the gastic juice in the stomach. Differences in plasma protein binding of ddl are unlikely to be responsible for the different estimates of volume of distribution obtained in the two clinical trials because it has been shown that only 5-6% of ddl binds to plasma protein (Anderson et 160 al., 1990). It is also possible that the patient populations of the two studies were different in terms of clinical status. Pharmacodynamic studies of ddl were performed in human PBMC cells, separated lymphocytes, and monocytes from healthy volunteers. The aim of these studies was to determine whether ddl could be converted to its active anabolite, ddATP, at sufficient levels to inhibit HIV-1 replication in these groups of cells. One of the disadvantages of HIV-1 treatment with AZT is that activation of this drug to AZTTP in monocytes does not produce sufficient AZTTP levels to inhibit HIV-RT. As a result, many investigators have hypothesized that monocytes may serve as a "reservoir" for the propagation of HIV-1 (Ho DD et al., 1986). The pharmacodynamic studies with ddl in PBMC cells, lymphocytes, and monocytes demonstrated that the ddATP concentrations in monocytes were significantly lower than in PBMC cells and lymphocytes. The lower levels of ddATP and of the other anabolites of ddl, ddAMP and ddADP, in monocytes could be a result of reduced uptake of the drug by monocytes rather than due to reduced efficiency of the enzymes in monocytes to anabolize the drug. As shown in Table 11, the intracellular concentration of ddl in monocytes was significantly lower than in PBMC cells or lymphocytes. 161 The absence of a PK interaction between ddl and AZT in pediatric patients with HIV-1 infection, receiving a simultaneous treatment with ddl and AZT (ACTG #176) suggested that the clinical improvement observed in the patients on this study could not be attributed to increased plasma concentrations of either AZT or ddl. Since these drugs are not the active species against HIV-1 RT, the improved clinical status of these pediatric patients might have been due to augmentation of the intracellular levels of the active anabolites of AZT and ddl, AZTTP and ddATP, when both drugs were present. Therefore, pharmaco dynamic studies of ddl were performed in T-cell lines, sensitive (Jurkat/O) and partially resistant to AZT (Jurkat/AZT-10), in the presence and absence of 1 m M AZT or following a 2-hour pre-incubation of the cells with 1 nM AZT. These studies demonstrated that AZT did not modulate the anabolism of ddl when administered simultaneously or prior to ddl. Since the pharmacokinetic and pharmacodynamic studies indicated that ddl and AZT did not interact in plasma and that AZT did not augment activation of ddl to ddATP, a mathematical model was developed to explain the synergistic inhibition of HIV-RT by ddl and AZT based on biochemical considerations. This model was based on Jackson’s equation which describes the rate of DNA synthesis by DNA polymerase using the four natural triphosphate substrates dCTP, dATP, dTTP.and dGTP. The mathematical model describes the rate of viral DNA synthesis by HIV-RT, a DNA 162 polymerase-like enzyme, in the presence and absence of one or more Michaelis-Menten competitive inhibitors (Equations 37 and 38). According to this model, the rate of DNA synthesis by RT, in the presence of both AZTTP and ddATP at concentrations half of their ICg, values, was 44.4% of the uninhibited reaction indicating a synergistic interaction on HIV-RT by AZTTP and ddATP. This model, therefore, provided a rational explanation for the synergistic interaction between ddl and AZT against HIV-1 based on the principles of enzyme kinetics and the specific enzymatic properties of HIV-RT with regard to AZZTP and ddATP. Since the mode of inhibition by AZTTP and ddATP (ddNTPs) involves competition with normal 2’-deoxynucleoside-5’- triphosphates (dNTP) for binding to RT, the mathematical model that has been developed suggests that inhibition of HIV-RT could become even greater if the levels of dATP and dTTP were decreased. Experiments by several investigators indicated that the ratios of AZTTP/dTTP and ddATP/dATP were critical in determining the antiviral activity of ddNTPs (Gao W-Y et al., 1993). In the experiments performed by Gao et al., AZT was most active in protecting activated PBMC cells than in resting cells because the ratio of AZTTP to dTTP was more than 10 times higher in the activated than in the resting PBMC cells. On the other hand, ddl was less active against activated PBMC cells because the ratio ddATP/dATP was more than 14 times lower in activated than in resting PBMC cells (Gao W-Y et al., 1993). The 163 dNTP pools in PBMC cells could be depleted by drugs that inhibit the enzyme ribonucleotide reductase (RR), such as hydroxyurea or analogs of this compound. Then, based on this mathematical model it might be beneficial to design M vitro experiments and clinical trials to evaluate the efficacy of drug combinations which use antiretroviral drugs, such as ddl and AZT, in addition to an inhibitor of RR which depletes the dNTP pools. 164 VII. CONCLUSIONS - The RIA assay of ddl w as much more sensitive than the HPLC assay. - The PK profiles of both ddl and AZT were adequately described by the one-compartment model with first-order absorption in pediatric patients with HIV-1 infection. - There was significant variability in PK parameters of ddl and AZT in pediatric patients with HIV-1 infection. - The frequency distibutions of the PK parametes of ddl, kel, Cl, and V„ were log-normal. Average values for these parameters were expressed as geometric means. - There were no statistically significant differences between the PK parameters of ddl when ddl was administered alone and in combination with AZT. - Similar PK estimates were obtained by the STS and NONMEM methods. - The volume of distribution and the elimination half-life were greater by NONMEM than by the STS method. 165 - PK parameters o f ddl were estimated with greater confidence by the NONMEM method than the STS method. - Food delayed absorption of ddl but did not prolong the half-life of elimination. - The extent of absorption of ddl or bioavailability was not altered by food. - The final population model of ddl concluded: - decrease in ka in presence of food • linear decrease in Cl with increasing age - linear decrease in Vd with increasing weight - Activation of ddl to ddATP was significantly higher in T-cells than in whole PBMC cells or monocytes. - The intracellular ddATP concentrations were significantly lower in monocytes than in T-cells or PBMC cells. - There were no statistically significant differences in the intracellular ddATP levels between Jurkat/0 (sensitive to AZT) and Jurkat/AZT-10 (partially resistant to AZT) cells. 166 - The simultaneous presence or pre-incubation with 1 A ZT did not alter the anabolism of ddl in the Jurkat cell lines. - A mathematical model based on a biochemical rationale was developed which provided an explanation for the synergistic inhibition of HIV-RT by ddATP and AZTTP. 167 VIII. REFERENCES Ahluwalia G, Cooney DA, Mitsuya H, et al. Initial studies on the cellular pharmacology of 2’3’-dideoxyinosine, an inhibitor of HIV infectivity. Biochem Pharmacol 1987; 26: 3797-3800. Anderson BD, Wygant MB, Xiang T-X, et al. Preformulation solubility and kinetic studies of 2’,3’-dideoxypurine nucleosides: potential anti-AIDS agents. Intern J Pharm 1988; 45: 27-37. Anderson BD, Hoesterey BL, Baker DC, et al. Uptake kinetics of 2\3’- dideoxyinosine into brain and cerebrospinal fluid of rats: Intravenous infusion studies. J Pharmacol Exper Ther 1990; 255: 113-118. Avramis VI and Plunkett W. Metabolism and therapeutic efficacy of 9-/9-D- arabinofuranosyl-2-fluoroadenine (F-ara-A) against murine leukemia P338. Cancer Res 1982; 42: 2587-2591. Avramis VI, Huang SH, and Holcenberg JS. Drug synergism, antagonism, and collateral sensitivity involving genetic changes. In Chou T-C and Rideout DC (eds), Synergism and antagonism in chemotherapy, Academic Press, Inc., San Diego, 1991, p. 585-619. Avramis VI, Kwock R, Solorzano M, and Goldsmith J. Synergistic effect between antiviral nucleoside analog drugs. Proc. "3rd European Conference on clinical aspects and treatment of HIV infection" p. 168, Abstr. pl37, Paris 1992. Avramis VI, Kwock R, Solorzano MM, and Gomperts E. Evidence of in vitro development of drug resistance to azidothymidine (AZT) in T-lymphocytic leukemia cell lines (Jurkat E6-1/AZT-100) and in pediatric patients with HIV-1 infection. AIDS 1993;4:1287-1296. Antal EJ, Grasela TH Jr., and Smith RB. An evaluation of population pharmacokinetics in therapeutic trials. Part III. Prospective data collection versus retrospective data assembly. Clin Pharmacol Ther 1989; 4£: 552-559. Back DJ, Ormesher S, Tjia JF, and Macleod R. Metabolism of 2’3’- dideoxyinosine (ddl) in human blood. Br J Clin Pharmacol 1992; 25: 319-322. Balis FM, Pizzo PA, Eddy J, et al. Pharmacokinetics of zidovudine administered intravenously and orally in children with human immunodeficiency virus infection. J Pediatr 1989:114: 880-884. 168 Balis FM, Pizzo PA, Butler KM, et al. Clinical pharmacology of 2’,3’- dideoxyinosine in human immunodeficiency virus-infected children. J Infect Dis 1992: 165: 99-104. Beal SL. Population pharmacokinetic data and param eter estimation based on their first two statistical moments. Drug Metab Rev 1984; 15(1&21: 173-193. Boucher CAB and Larder B A Viral variation and therapeutic strategies in HIV infection. MediTech Media Ltd, London, 1994. Butler KM, Husson RN, Balis FM, et al. Dideoxyinosine in children with symptomatic human immunodeficiency virus infection N Engl J Med. 1991:324: 137-144. Callebaut C, Krust B, Jacotot E, and Hovanessian AG. T cell activation, CD26, as a cofactor for entry of HIV in CD4+ cells. Science 1993; 262: 2045-2050. Clapham PR, W eber JN, Whitby D, et al. Soluble CD4 blocks the infectivity of diverse strains of HIV and SIV for T cells and monocytes but not for brain and muscle cells. Nature 1989; 337: 368-370. Cohen J. HIV ’Cofactor’ comes in for more heavy fire. Science 1993; 2 6 2 " 1971. Collier AC, Coombs RW, Fischl M A et al. Combination therapy with zidovudine and didanosine compared with zidovudine alone in HIV-1 infection. Ann Intern Med 1993; 119: 786-793. Cox SW, Albert J, Ljungdahl-Stahle E, et al. Effect of resistance on combination chemotherapy for human immunodeficiency virus infection. Advan Enz Regul 1993; 2 2 " 27-36. Dalakas MC. Retroviruses and inflammatory myopathies in humans and primates. Baillieres Clin Neurol 1993; 2 " 659-91. Drusano GL, Yuen GJ, Morse G, et al. Impact of bioavailability on determination of the maximal tolerated dose of 2’,3’-dideoxyinosine in phase I trials. Antimicrob Agents Chemother 1992; 2 6 " 1280-1283. Eales L-J and Iarkin JM. Current concepts in the immunopathogenesis of AIDS and HIV infection. Br Med Bull 1988; 44 (1): 38-55. 169 Fischl MA, Richman DD, Greico MH, et al. The efficacy of azidothymidine (AZT) in the treatm ent of patients with AIDS and AIDS-related complex: A double-blind placebo controlled trial. N Engl J Med 1987; H Z : 185-191. Fischl MA, Richman DD, Hansen N, et al. The safety and efficacy of zidovudine (AZT) in the treatment of subjects with mildly symptomatic human immunodeficiency virus type 1 (HIV) infection: a double blind placebo controlled trial. A nn Intern M ed 1990; 112: 727-737. Friis-Hansen B. Body water compartments in children: changes during growth and related changes in body composition. Pediatrics 1961; 2& 169-171. Fujiwara T and Mizuuchi K. Retroviral DNA Integration: Structure of an integation intermediate. Cell 1988; 54: 497-504. Furman PH, Fyfe JA, St Clair MH, et al. Phosphorylation of 3’-azido-3’- deoxythymidine and selective interaction with human immunodeficiency virus reverse transcriptase. Proc Natl Acad Sci USA 1986; 55: 8333-8335. Gallo RC. The AIDS virus. Scientific American 1987; 256: 47-56. Gallo RC. HIV-The cause of AIDS: An ovreview on its biology, mechanisms of disease induction, and our attempts to control it. J AIDS 1988; 1: 521-535. Gao WY, Shirasaka T, Johns DG, et al. Differential phosphorylation of azidothymidine, deoxycytinine, and deoxyinosine in resting and activated blood mononuclear cells. J Clin Invest 1993; 51: 2326-2333. G artner S, Markovits P, Markovitz DM, et al. The role of mononuclear phacocytes in HTLV-III/LAV infection. Science 1986; 222: 215-221. Gibaldi M and Prescott L (eds). Handbook of clinical pharmacokinetics. Adis Health Science Press, Balgowlah, 1983. Gitterman SR, Drusano GL, Egorin MJ, et al. Population pharmacokinetics of zidovudine. Clin Pharmacol T her 1990; 4g: 161-167. Grasela TH Jr., A ntal EJ, Townsend RJ, et al. An evaluation of population pharmacokinetics in therapeutic trials. Part I. Comparison of methodologies. Clin Pharmacol T her 1986; 23' 605-612. 170 Grasela TH Jr. Antal EJ, Ereshefsky L, et al. An evaluation of population pharmacokinetics in therapeutic trials. Part II. Detection of a drug-drug interaction. Clin Pharmacol Ther 1987; 42: 433-441. G reene WC. AIDS and the immune system. Scientific American, 1993; 269(31: 99-105. Grevel J, Whiting B, Kelman AW, et al. Population analysis of the pharmacokinetic variability of high doese metochlopramide in cancer patients. Clin Pharmacokinet, 1988; 14: 52-63. Grindey GB and Cheng YC. Biochemical and kinetic approaches to inhibition of multiple pathways. Pharmac Ther 1979; 4: 307-327. Groopman JE . Antiretroviral therapy and immunomodulators in patients with AIDS. Am. J. Med. 1991; 5Q: 4A/18S-4A/29S. H ao Z, Cooney DA, H artm an NR, et al. Factors determining the activity of 2’,3’-dideoxynucleosides in suppressing human immunodeficiency virus i s vitro. Mol Pharmacol 1988; 14: 431-435. H arrap KR and Jackson RC. Enzyme kinetics and combination chemotherapy: an appraisal of current concepts. Adv Enzyme Regul 1975; 11: 77-96. Hartman NR, Yarchoan R, Pluda JM, et al. Pharmacokinetics of 2’3’- dideoxyinosine in patients with severe human immunodeficiency infection.II. The effects of different oral formulations and the presence of other medications. Clin Pharmacol Ther 1991; IQ: 278-285. Hartman NR, Yarchoan R, Pluda JM, et al. Pharmacokinetics of 2’,3’ dideoxyinosine in patients with severe human immunodeficiency virus infection. Clin Pharmacol Ther 1990; 42: 647-654. Haseltine WA. Replication and pathogenesis of the AIDS virus. J AIDS 1988; 1: 217-240. Henry K, Chinnock BJ, Quinn RP, et al. Concurrent zidovudine levels in semen and serum determined by radioimmunoassay in patients with AIDS or AIDS- related complex. JAMA 1988; 259(20): 3023 Hirsch MS. Chemotherapy of human immunodeficiency virus infections: Current practice and future prospects. J Inf Dis 1990; 161: 845-857. 171 H o DD, Rota TR, and Hirsch MS. Infection of monocyte/macrophages by human T lymphotropic virus type III. J Clin Invest 1988; 11: 1712-1715. Hobbs G, Keilbaugh SA, Prusoff WA, and Simpson MV. Mechanism of toxicity of A ZT and other anti-HIV-1 dideoxynucleosides. Int Conf AIDS, 1991:2:215. Jackson RC. The role of ribonucleotide reductase in regulation of deoxyribonucleoside triphosphate pool composition: Studies with a kinetic model. In Cory JC and Cory AH (eds), Inhibitors of ribonucleoside triphosphate reductase activity, Pergamon Press, Inc., N. York, 1989, P. 127-150. Johnson VA, Merrill DP, Videler JA et al. Two drug combinations of zidovudine, didanosine, and recombinant interferon-a inhibit replication of zidovudine-resistant human immunodeficiency virus type 1 synergistically in vitro. J Infect Dis 1991; 164: 646-655. Kahn JO, Lagakos SW, Richman DD e t al. A controlled trial comparing continued zidovudine with didanosine in human immunodeficiency virus infection. N Engl J Med 1992; 327: 581-587. Kameoka J, Tanaka T, Nojima Y, et al. Direct association of adenosine deaminase with a T cell activation antigen, CD26. Science 1993; 261: 466-469. Knupp CA, Shyu WC, Dolin R, et al. Pharmacokinetics of didanosine in patients with acquired immunodeficiency syndrome or acquired immunodeficiency syndrome-related complex. Clin Pharmacol Ther 1991; 42: 523-535. Larder BA, Darby G, Richman DD. HIV with reduced sensitivity to zidovudine (A ZT) isolated during prolonged therapy. Science 1989; 242: 1731-1734. Larder BA and Kemp SD. Multiple mutations in HIV-1 reverse transcriptase confer high-level resistance to zidovudine (AZT). Science 246: 1155-1158. Ludden TM. Population pharmacokinetics. J Clin Pharmacol 1988; 2g: 1059- 1063. Macleod R, Eagling VA, Sim SM, and Back DJ. In vitro inhibition studies of the glucuronidation of 3’-azido-3’-deoxythymidine catalysed by human liver UDP- glucuronosyl transferase. Biochem Pharmacol 1992; 42: 382-386. M aitre PO, Vohez S, Heykants J, et al. Population pharmacokinetics of alfentantil: the average dose-plasma concentration relationship and interindividual variability in patients. Anesthes 1987; j&: 3-12. 172 McKinney R E Jr. Antiviral therapy for human immunodeficiency virus infection in children. Ped Clin N Am 1991; 28(1): 133-151. Merigan TC. Treatment of AIDS with combinations of antiretroviral agents. Am J Med 1991; 3Q (supp4A): 4A/1S-4A/29S. Mitsuya H, Weihold KJ, Furman PA, et al. 3’-azido-3’-deoxythymidine (BW A5090): An antiviral agent that inhibits the infectivity and cytopathic effectof human T-lymphotropic virus type III/lymphadenopathy-associated virus in vitro. Proc Natl Acad Sci USA 1985; S Z : 7096-7100. Mitsuya H and Broder S. Inhibition of the in vitro infectivity and cytopathic effect of human T-lymphotropic virus type III/lymphadenopathy-associated virus (HTLV-III/LAV) by 2’,3’-dideoxynucleosides. Proc Natl Acad Sci USA 1986; 22: 1911-1915. Mitsuya H and Broder S. Strategies for antiviral therapy in AIDS. Nature, 1987; 325: 773-778. Mitsuya H, Yarchoan R, Kageyama S, and Broder S. Targeted therapy of human immunodeficiency virus-related disease. FASEB 1991; 2 ‘ * 2369-2381. Nair V and Sells TB. Interpretation of the roles of adenylosuccinate lyase and of AMP deaminase in the anti-HIV activity of 2’,3’-dideoxyadenosine and 2’,3’- dideoxyinosine. Biochim Biophys Acta 1992; 1119: 201-204. NONMEM 77 Version III Level 1. NONMEM Project Group, University of California at San Francisco. NONMEM Users Guide-Part V. NONMEM Project Group, University of California at San Francisco. Peck CC and Rodman JH. Analysis of pharmacokinetic data for individualizing drug dosage regimens. In Evans WE, Schentag JJ and Jusko WJ (eds), Applied pharmacokinetics. Principles of therapeutic drug monitoring, Applied Therapeutics, Inc., WA, 1986. Periclou AP and Avramis VI. Development of a population pharmacokinetic model (PPM) to describe didanosine (ddl) and azidothymidine (AZT) interaction after a combination treatment in pediatric patients with HIV-1 infection. Proc. 4th European Conf Clin Aspects & Treat of HIV-1 Infect, Milan Italy 1994a, p. 180, Abstr. # P 259. 173 Periclou AP, Solorzano MM, and Avramis VI. Comparative anabolism of azidothymidine (AZT) and didanosine (ddl) in human T-cell lines and PBMC cells sensitive and resistant to AZT. Proc 4th European Conf Clin Aspects & T reat of HIV-1 Infect., Milan Italy 1994b, p.217, Abstr. # 0 51. Periclou AP, Solorzano MM, and Avramis VI. Comparative anabolism of Azidothymidine (AZT) and didanosine (ddl) in human T-cell lines sensitive and resistant to A ZT and human PBMC cells. Proc Am Assoc Cancer Res 1994c, 25: 333, Abstr. #1982. Periclou AP, Solorzano MM, and Avramis VI. Drug synergism by nucleoside analog drugs against HIV-RT based on multiple inhibitor kinetics. Am. Assoc. Cancer Res., March 1995, Toronto Canada, Abstr. #1720 (accepted). Pizzo PA, Eddy J, Falloon J, et al. Effect of continuous intravenous infusion of zidovudine (AZT) in children with symptomatic HIV infection. N Engl J Med 1988: 319: 889-896. Pizzo PA. Considerations for the evaluation of antiretroviral agents in infants and children infected with human immunodeficiency virus: A perspective from the national cancer institute. Rev Infect Dis 1990; 12: S561-S569. Racine-Poon A and Smith AF. Population models. In Berry (ed), Statistical methodology in the Pharmaceutical sciences, Marcel Dekker, Inc, New York, 1990. Rawn JD. Biochemistry. H arper & Row, Publishers, New York, 1983; p.331 Richman DD, Fischl MA, Grieco MH, et al. The toxicity of azidothymidine (A ZT) in the treatment of patients with AIDA and AIDS-related complex. N Engl J Med 1987; 212: 192-197. Richman DD. The treatment of HIV infection. Azidothymidine (AZT) and other new antiviral drugs. Infect Dis Clin North Am 1988; 2: 397-407. Roberts JD, Bebenek K, and Kunkel TA. The accuracy of reverse transcriptase from HIV-1. Science 1988; 242: 1171-1173. Sheiner LB and Beal SL. Evaluation of methods for estimating population pharmacokinetic parameters. II. Bioexponential model and experimental pharmacokinetic data. J Pharmacokinet Biopharm 1981; 635-651. 174 Sheiner LB and Beal SL. Evaluation of methods for estimating population pharmacokinetic parameters. III. Monoexponential Model: Routine clinical pharmacokinetic data. J Pharmacokinet Biopharm 1983; 11: 303-319. Sheiner LB. Analysis of pharmacokinetic data using parametric models-1: Regression models. J Pharmacokinet Biopharm 1984; 12: 93-118. Shelton MJ, O’Donell AM, and Morse GD. Didanosine. Ann Pharmacother 1992; 26: 660-669. Shyu WC, Knupp CA, Pittman K-A, et al. Food-induced reduction in bioavailability of didanosine. Clin Pharmacol Ther 1991; 2Q: 503-507. Singlas E, Pioger J-C, Taburet A-M, et al. Zidovudine sidposition in patients with severe renal impairment: Influence of hemodialysis. Clin Pharmacol Ther 1989; 4fc 190-197. Steimer LB, Mallet A, and Mentrfe. Estimating interindividual pharmacokinetic variability. In Rowland M, Sheiner LB, and Steimer JL (eds), Variability in drug therapy, Raven Press, New York, 1985. Tateno M, Gonzalez-Scarano F, and Levy JA. Human immunodeficiency virus can infect CD4-negative human fibroblastoid cells. Proc Natl Acad Sci USA 1989; M : 4287-4290. Valentine FT, Seidlin M, and Hochster H. Phase I study o f 2\3’-dideoxyinosine: Experience with 19 patients at New York University Medical Center. Rev Inf Dis, 1990; i2 : S534-S539. Valentine C and Foulis A. Alpha trial: pancreatitis on didanosine (ddl) risk factors and pathological features. The M RC/ANRS Coordinating Committee of the European/Australian Alpha Trial. Int Conf AIDS 1993; 5: 477. Volberding PA, Lagakos SW, Koch MA, et al. Zidovudine in asymptomatic human immunodeficiency virus infection. N Engl J Med 1990; 222: 941-949. Vozeh S, Maitre PO, and Stanski DR. Evaluation of population (NONMEM) pharmacokinetic param eter estimates. J Pharmacokinet Biopharm 1990; iS : 161- 173. W eber JN and Weiss RA. The virology of human immunodeficiency viruses. Br Med Bull 1988; 44(1): 20-37. 175 Weinberg K. Personal Communications. Whiting B, Kelm an AW, and Grevel J. Population pharmacokinetics. Theory and clinical application. Clin Pharmacokinet 1986; H : 387-401. Wientjes M G and Au JL-S. Lack of pharmacokinetic interaction betw een 2\3’- dideoxyinosine and 3’-deoxythymidine in rats. Antim Agents Chem 1992; 36: 665-668. Wientjes M G, Mukheiji E , and Au Jl^S . Nonlinear disposition of intravenous 2 \3 ’-dideoxyinosine in rats. Pharm Sci 1992; j) :1070-1075. Williams PJ, Lane J, Murray W, et al. Pharmacokinetics of the digoxin-quinidine interaction via mixed-effect modelling. Clin Pharmacokinet 1992; 2 2 0 ) : 66-74. Yarchoan R, Weinhold K J, Lyerly HK, et al. Administration of 3’-Azido-3’- deoxythymidine, an inhibitor of HTLV-HI/LAV replication, to patients with AIDS or AIDS-related complex. Lancet 1986; 1:575-580. Yarchoan R, Mitsuya H, Myers CE, and Broder S. Clinical pharmacology of 3’- azido-2’,3’-dideoxythymidine (zidovudine) and related dideoxynucleotides. N Engl J Med 1989; 221: 726-738. Yarchoan R, Lietzau JA, Nguyen B-Y, et al. A randomized pilot study of alternating o r simultaneous zidovudine and didanosine therapy in patients with symptomatic human immunodeficiency virus infection. J Inf Dis 1994; _1£2:9-17. 176
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
Asset Metadata
Core Title
00001.tif
Tag
OAI-PMH Harvest
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC11256543
Unique identifier
UC11256543
Legacy Identifier
9617135