Survival analysis using auxiliary variables via multiple imputation, with application to AIDS clinical trial data

We develop an approach, based on multiple imputation, to using auxiliary variables to recover information from censored observations in survival analysis. We apply the approach to data from an AIDS clinical trial comparing ZDV and placebo, in which CD4 count is the time-dependent auxiliary variable. To facilitate imputation, a joint model is developed for the data, which includes a hierarchical change-point model for CD4 counts and a time-dependent proportional hazards model for the time to AIDS. Markov chain Monte Carlo methods are used to multiply impute event times for censored cases. The augmented data are then analyzed and the results combined using standard multiple-imputation techniques. A comparison of our multiple-imputation approach to simply analyzing the observed data indicates that multiple imputation leads to a small change in the estimated effect of ZDV and smaller estimated standard errors. A sensitivity analysis suggests that the qualitative findings are reproducible under a variety of imputation models. A simulation study indicates that improved efficiency over standard analyses and partial corrections for dependent censoring can result. An issue that arises with our approach, however, is whether the analysis of primary interest and the imputation model are compatible.     

A mixture model for longitudinal data with application to assessment of noncompliance

In clinical trials of a self-administered drug, repeated measures of a laboratory marker, which is affected by study medication and collected in all treatment arms, can provide valuable information on population and individual summaries of compliance. In this paper, we introduce a general finite mixture of nonlinear hierarchical models that allows estimates of component membership probabilities and random effect distributions for longitudinal data arising from multiple subpopulations, such as from noncomplying and complying subgroups in clinical trials. We outline a sampling strategy for fitting these models, which consists of a sequence of Gibbs, Metropolis-Hastings, and reversible jump steps, where the latter is required for switching between component models of different dimensions. Our model is applied to identify noncomplying subjects in the placebo arm of a clinical trial assessing the effectiveness of zidovudine (AZT) in the treatment of patients with HIV, where noncompliance was defined as initiation of AZT during the trial without the investigators' knowledge. We fit a hierarchical nonlinear change-point model for increases in the marker MCV (mean corpuscular volume of erythrocytes) for subjects who noncomply and a constant mean random effects model for those who comply. As part of our fully Bayesian analysis, we assess the sensitivity of conclusions to prior and modeling assumptions and demonstrate how external information and covariates can be incorporated to distinguish subgroups.     

Contributions of CYP2C9/CYP2C19 genotypes and drug interaction to the phenytoin treatment in the Korean epileptic patients in the clinical setting

We examined the contribution of CYP2C9 and CYP2C19 genotypes and drug interactions to the phenytoin metabolism among 97 Korean epileptic patients to determine if pharmacogenetic testing could be utilized in routine clinical practice. The CYP2C9 polymorphism is a wellknown major genetic factor responsible for phenytoin metabolism. The CYP219 polymorphism, with a high incidence of variant alleles, has a minor influence on phenytoin treated Koran patients. Using a multiple regression model for evaluation of the CYP2C9 and CYP2C19 genotypes, together with other non-genetic variables, we explained 39.6% of the variance in serum phenytoin levels. Incorporation of genotyping for CYP2C9 and CYP2C19 into a clinical practice may be of some help in the determination of phenytoin dosage. However, because concurrent drug treatment is common in patients taking phenytoin and many environmental factors are likely to play a role in drug metabolism, these factors may overwhelm the relevance of CYP polymorphisms in the clinical setting. Further investigations with an approach to dose assessment that includes comprehensive interpretation of both pharmacogenetic and pharmacokinetic data along with understanding of the mechanism of drug interactions in dosage adjustment is warranted.     

Development of a Target cell-Biologics-Effector cell (TBE) complex-based cell killing model to characterize target cell depletion by T cell redirecting bispecific agents

T-cell redirecting bispecific antibodies (bsAbs) or antibody-derived agents that combine tumor antigen recognition with CD3-mediated T cell recruitment are highly potent tumor-killing molecules. Despite the tremendous progress achieved in the last decade, development of such bsAbs still faces many challenges. This work aimed to develop a mechanism-based pharmacokinetic/pharmacodynamic (PK/PD) modeling framework that can be used to assist the development of T-cell redirecting bsAbs. A Target cell-Biologics-Effector cell (TBE) complex-based cell killing model was developed using in vitro and in vivo data, which incorporates information on binding affinities of bsAbs to CD3 and target receptors, expression levels of CD3 and target receptors, concentrations of effector and target cells, as well as respective physiological parameters. This TBE model can simultaneously evaluate the effect of multiple system-specific and drug-specific factors on the T-cell redirecting bsAb exposure–response relationship on a physiological basis; it reasonably captured multiple reported in vitro cytotoxicity data, and successfully predicted the effect of some key factors on in vitro cytotoxicity assays and the efficacious dose of blinatumomab in humans. The mechanistic nature of this model uniquely positions it as a knowledge-based platform that can be readily expanded to guide target selection, drug design, candidate selection and clinical dosing regimen projection, and thus support the overall discovery and development of T-cell redirecting bsAbs.     

Targeted drug delivery by gemtuzumab ozogamicin: mechanism-based mathematical model for treatment strategy improvement and therapy individualization

Gemtuzumab ozogamicin (GO) is a chemotherapy-conjugated anti-CD33 monoclonal antibody effective in some patients with acute myeloid leukemia (AML). The optimal treatment schedule and optimal timing of GO administration relative to other agents remains unknown. Conventional pharmacokinetic analysis has been of limited insight for the schedule optimization. We developed a mechanism-based mathematical model and employed it to analyze the time-course of free and GO-bound CD33 molecules on the lekemic blasts in individual AML patients treated with GO. We calculated expected intravascular drug exposure (I-AUC) as a surrogate marker for the response to the drug. A high CD33 production rate and low drug efflux were the most important determinants of high I-AUC, characterizing patients with favorable pharmacokinetic profile and, hence, improved response. I-AUC was insensitive to other studied parameters within biologically relevant ranges, including internalization rate and dissociation constant. Our computations suggested that even moderate blast burden reduction prior to drug administration enables lowering of GO doses without significantly compromising intracellular drug exposure. These findings indicate that GO may optimally be used after cyto-reductive chemotherapy, rather than before, or concomitantly with it, and that GO efficacy can be maintained by dose reduction to 6 mg/m(2) and a dosing interval of 7 days. Model predictions are validated by comparison with the results of EORTC-GIMEMA AML19 clinical trial, where two different GO schedules were administered. We suggest that incorporation of our results in clinical practice can serve identification of the subpopulation of elderly patients who can benefit most of the GO treatment and enable return of the currently suspended drug to clinic.     

In vitro ADME profiling using high-throughput rapidfire mass spectrometry: cytochrome p450 inhibition and metabolic stability assays

Early assessment of absorption, distribution, metabolism, and excretion (ADME) properties of drug candidates has become an essential component of modern drug discovery. ADME characterization is important in identifying compounds early that are likely to fail in later clinical development because of suboptimal pharmacokinetic properties or undesirable drug-drug interactions. Proper utilization of ADME results, meanwhile, can prioritize candidates that are more likely to have good pharmacokinetic properties and also minimize potential drug-drug interactions. By integrating a RapidFire system with an API4000 mass spectrometer (RF-MS), we have established a high-throughput capability to profile compounds (>100 compounds/wk) in a panel of ADME assays in parallel with biochemical and cellular characterizations. Cytochrome P450 inhibition and time-dependent inhibition assays and microsomal stability assays were developed and fully optimized on the system. Compared with the classic liquid chromatography-mass spectrometry method, the RF-MS system generates consistent data with approximately 20-fold increase in throughput. The lack of chromatographic separation of compounds, substrates, and metabolites can complicate data interpretation, but this occurs in a small number of cases that are readily identifiable. Overall, this system has enabled a real-time and quantitative measurement of a large number of ADME samples, providing a rapid evaluation of clinically important drug-drug interaction potential and drug metabolic stability.     

Modeling multi-drug chemotherapy: tailoring treatment to individuals

BACKGROUND: Predicting and tailoring optimal cancer treatments presents a major challenge. METHODS: A computational model (kinetically tailored treatment, or KITT model) is developed to predict drug combinations, doses, and schedules likely to be effective in reducing tumor size and prolonging patient life. Treatment strategies may be tailored to individuals based on tumor cell kinetics. The model incorporates intra-tumor heterogeneity and evolution of drug resistance, apoptotic rates, and cell division rates. Tumor growth may follow an exponential or a Gompertzian trajectory. Drug pharmacodynamic and pharmacokinetic models are used. Toxicity is modeled in several ways. RESULTS: A key prediction of KITT is that including cytostatic drugs like tamoxifen and herceptin during treatment with cytotoxic drugs substantially increases the probability of cure and prolongs patient life. Results also suggest that altering drug scheduling may be more effective but not more toxic than dose escalation. CAF chemotherapy (cyclophosphamide, adriamycin, and 5-fluorouracil) is predicted to be more effective than CMF (cyclophosphamide, methotrexate, and 5-fluorouracil). KITT also suggests that tumors with a high proliferative index (PI) may respond better to drug combinations incorporating two cell-cycle phase-specific drugs than do tumors with a low PI. Tumors with a low PI, in contrast, are predicted to respond better to regimens involving two cell-cycle phase-non-specific drugs than do tumors with a high PI. These predictions are borne out by clinical trial results published in the literature, which are discussed.Simulated predictions of the model match well with results from a clinical trial by Silvestrini et al. (2000. Int. J. Cancer 87, 405). The results of simulating the growth of 26896 tumors are used to construct a decision tree for prognosis to identify the key tumor and treatment variables. CONCLUSION: Additional tests of the model are needed in which physicians collect information on apoptotic and proliferative indices, cell-cycle times, and drug resistance from biopsies of each individual's tumor. Computational models may become important tools to help optimize and tailor cancer treatments.     

On Assessment of Bioequivalence for Drugs with Negligible Plasma Levels

In bioavailability studies, bioequivalence between drug products is usually determined based on some pharmacokinetic responses such as area under the blood or plasma concentration-time curve and maximum concentration. For some drug products, however, we may have negligible plasma levels because their intended routes of administration. In this case, assessment of bioequivalence between drug products of this kind may be established using clinical endpoints such as therapeutic response and time to the onset of a therapeutic response. In this paper, we propose two procedures which modify the method of generalized estimating equations (Liang and Zeger, 1986) and the proportional hazard models for paired failure times to assess bioequivalence between two drug products under the structure of a standard two-sequence, two-period crossover design. An example concerning a bioequivalence trial for albuterol metered dose inhaler indicated for acute bronchospasm (Herson, 1991) is used to illustrate the proposed procedures.     

Development of a Drug-Response Modeling Framework to Identify Cell Line Derived Translational Biomarkers That Can Predict Treatment Outcome to Erlotinib or Sorafenib

Development of drug responsive biomarkers from pre-clinical data is a critical step in drug discovery, as it enables patient stratification in clinical trial design. Such translational biomarkers can be validated in early clinical trial phases and utilized as a patient inclusion parameter in later stage trials. Here we present a study on building accurate and selective drug sensitivity models for Erlotinib or Sorafenib from pre-clinical in vitro data, followed by validation of individual models on corresponding treatment arms from patient data generated in the BATTLE clinical trial. A Partial Least Squares Regression (PLSR) based modeling framework was designed and implemented, using a special splitting strategy and canonical pathways to capture robust information for model building. Erlotinib and Sorafenib predictive models could be used to identify a sub-group of patients that respond better to the corresponding treatment, and these models are specific to the corresponding drugs. The model derived signature genes reflect each drug’s known mechanism of action. Also, the models predict each drug’s potential cancer indications consistent with clinical trial results from a selection of globally normalized GEO expression datasets.     

Pharmacokinetic and pharmacodynamic modelling of the novel human granulocyte colony-stimulating factor derivative Maxy-G34 and pegfilgrastim in rats

Objectives:  This study aims to compare pharmacokinetics and pharmacodynamics of pegfilgrastim, a pharmaceutical recombinant human granulocyte colony-stimulating factor (rhG-CSF), with that of a newly developed reagent, Maxy-G34. This comparison was performed using rat experiments and biomathematical modelling of granulopoiesis.
Methods:  Healthy rats and those with cyclophosphamide-induced neutropenia were treated with either pegfilgrastim or Maxy-G34 under various schedules. Time courses of absolute neutrophil count (ANC) and G-CSF serum level were measured and we constructed a combined pharmacokinetic/pharmacodynamic model of both drugs. Neutropenic episodes were assessed by experimental data and model simulations.
Results:  Both Pegfilgrastim and Maxy-G34 showed strong dose-dependent efficacy in reducing neutropenic episodes. However, time courses of ANC and G-CSF serum levels were markedly different. The biomathematical model showed good agreement with these data. We estimated that differences between the two drugs could be explained by lower bioavailability and reduced elimination of Maxy-G34. Based on the data and model interpolations, we estimated that Maxy-G34 is superior in reducing neutropenic episodes. Also, we predicted that G-CSF administration 48 h after cyclophosphamide would be superior to its administration after 2 or 24 h, for both derivatives.
Conclusion:  Maxy-G34 is a highly potent drug for stimulation of neutrophil production in rats. By our modelling approach, we quantified differences between Maxy-G34 and pegfilgrastim, related to pharmacokinetic parameters. Model simulations can be used to estimate optimal dosing and timing options in the present preclinical rat model.     

Practical perspectives of personalized healthcare in oncology

There is an increasing prevalence of drug-diagnostic combinations in oncology. This has placed diagnostic stakeholders directly into the complex benefit-risk, cost, value and uncertainty-driven development paradigm traditionally the preserve of the drug development community. In this review we focus on the delivery of the clinical data required to advance such drug-diagnostic combination development programmes and ultimately satisfy regulators and payors of the value of contemporaneous changes in diagnostic and treatment practice. Ideally all stakeholders would like to initially estimate, and ultimately specify, the comparative benefit-risk for a new treatment option with and without changing diagnostic practice. Hence, in an ideal world clinical trial design is focused on acquiring biomarker treatment interaction data. In this review we describe the key scientific and feasibility inputs required to design and deliver such trials and the drivers, advantages and disadvantages associated with departing from this model. We do not discuss the discovery of new biomarkers nor the analytical validation and marketing of diagnostic products. Following on from trial design we describe how subsequent success then depends upon the concepts that guide trial design being driven into the complex world of large, multinational clinical trial delivery. For every aspect of a traditional clinical drug trial such as supply, recruitment and adherence, there is a corresponding concept for the diagnostic element. In practice, this means that each patient's contribution to the decision making data-set is subject to double jeopardy (attrition on clinical outcome and biomarker status). Historically, this has led to significantly reduced power for detecting biomarker-treatment interactions, reduced decision making confidence and a waste of valuable human and financial resources. We describe recent practice changes and experience that have led to the successful delivery of such trials focusing on both pre- and on trial aspects. The former includes the pivotal role of tissue banks in accurate estimation of evaluability and prevalence for biomarker assays and the latter several practices designed to engage and incentivize key stakeholders particularly CRAs and pathologists. The result is that in the new world of developing personalized treatments for cancer patients the real-time acquisition and monitoring of biomarker data receives similar support to that traditionally reserved for clinical outcome data and far more patients contribute to the testing of personalized medicine hypotheses.     

Simulation of differential drug pharmacokinetics under heat and exercise stress using a physiologically based pharmacokinetic modeling approach

Under extreme conditions of heat exposure and exercise stress, the human body undergoes major physiological changes. Perturbations in organ blood flows, gastrointestinal properties, and vascular physiology may impact the body's ability to absorb, distribute, and eliminate drugs. Clinical studies on the effect of these stressors on drug pharmacokinetics demonstrate that the likelihood of pharmacokinetic alteration is dependent on drug properties and the intensity of the stressor. The objectives of this study were to use literature data to quantify the correlation between exercise and heat exposure intensity to changing physiological parameters and further, to use this information for the parameterization of a whole-body, physiologically based pharmacokinetic model for the purposes of determining those drug properties most likely to demonstrate altered drug pharmacokinetics under stress. Cardiac output and most organ blood flows were correlated with heart rate using regression analysis. Other altered parameters included hematocrit and intravascular albumin concentration. Pharmacokinetic simulations of intravenous and oral administration of hypothetical drugs with either a low or high value of lipophilicity, unbound fraction in plasma, and unbound intrinsic hepatic clearance demonstrated that the area under the curve of those drugs with a high unbound intrinsic clearance was most affected (up to a 130% increase) following intravenous administration, whereas following oral administration, pharmacokinetic changes were smaller (<40% increase in area under the curve) for all hypothetical compounds. A midazolam physiologically based pharmacokinetic model was also used to demonstrate that simulated changes in pharmacokinetic parameters under exercise and heat stress were generally consistent with those reported in the literature.     

Preliminary results in HIV-1-infected patients treated with transfer factor (TF) and Zidovudine (ZDV)

The efficiency of HIV-1 specific transfer factor (TF) administration, combined with Zidovudine (ZDV), in asymptomatic persistent generalised lymphadenopaty, or AIDS related complex (ARC) patients was evaluated. Twenty patients were randomly assigned to receive only ZDV (1st group) or ZDV together with HIV-1-specific TF (2nd group). HIV-1-specific TF was administered orally at 2 × 10⁷ cell equivalent daily for 15 days, and thereafter once a week for up to 6 months. There were no significant differences between the two groups in clinical evolution, red blood cells, haemoglobin, lymphocytes, CD20 subset, transaminases,β-2-microglobulin, p24 antigen. White blood cells, CD8 lymphocytes as well as IL-2 levels increased in the second group, while the CD4 subset increased in the first group. The combination treatment with ZDV and TF appeared to be safe and well tolerated. Furthermore, levels of serum cytokines were investigated in 10 patients (8 asymptomatic and 2 ARC) treated with ZDV, and compared with 5 patients of the 2nd group (3 asymptomatic and 2 ARC) treated with ZDV plus HIV-1-specific TF. Peripheral lymphocytes, CD4, CD8 subsets, IL-2, TNFα, IL-6, p24 antigen, IL-2 soluble lymphocyte receptors (sR), CD4sR, CD8sR and ß-2-microglobulin were evaluated at the baseline and at the 3rd month. The CD4 subset was not significantly different in the two groups, whilst IL-2 increased in the 2nd group receiving ZDV plus TF, suggesting an activation of the Th1 secretion pattern.     

Design and characterization of MP0250, a tri-specific anti-HGF/anti-VEGF DARPin® drug candidate

MP0250 is a multi-domain drug candidate currently being tested in clinical trials for the treatment of cancer. It comprises one anti-vascular endothelial growth factor-A (VEGF-A), one anti-hepatocyte growth factor (HGF), and two anti-human serum albumin (HSA) DARPin® domains within a single polypeptide chain. While there is first clinical validation of a single-domain DARPin® drug candidate, little is known about DARPin® drug candidates comprising multiple domains. Here, we show that MP0250 can be expressed at 15 g/L in soluble form in E. coli high cell-density fermentation, it is stable in soluble/frozen formulation for 2 years as assessed by reverse phase HPLC, it has picomolar potency in inhibiting VEGF-A and HGF in ELISA and cellular assays, and its domains are simultaneously active as shown by surface plasmon resonance. The inclusion of HSA-binding DARPin® domains leads to a favorable pharmacokinetic profile in mouse and cynomolgus monkey, with terminal half-lives of ～ 30 hours in mouse and ～ 5 days in cynomolgus monkey. MP0250 is thus a highly potent drug candidate that could be particularly useful in oncology. Beyond MP0250, the properties of MP0250 indicate that multi-domain DARPin® proteins can be valuable next-generation drug candidates.     

Domain-based assays of individual antibody concentrations in an oligoclonal combination targeting a single protein

Quantitation of individual monoclonal antibodies (mAbs) within a combined antibody drug product is required for preclinical and clinical drug development, including pharmacokinetic (PK), toxicology, stability, and biochemical characterization studies of such drugs. We have developed an antitoxin, XOMA 3AB, consisting of three recombinant mAbs that potently neutralize the known subtypes of type A botulinum neurotoxin (BoNT/A). The three mAbs bind nonoverlapping BoNT/A epitopes with high affinity. XOMA 3AB is being developed as a treatment for botulism resulting from BoNT/A. To develop antibody-specific assays, we cloned, expressed, and purified BoNT/A domains from Escherichia coli. Each mAb bound only to its specific domain with affinity comparable to the binding to holotoxin. mAb-specific domains were used to develop an enzyme-linked immunosorbent assay (ELISA) for characterization of the integrity and binding activity of the three mAbs in the drug product. An electrochemiluminescence bridging assay that is robust to interference from components in serum was also developed, and we demonstrate that it can be used for PK assays. This type of antigen engineering to generate mAb-specific domains is a general method allowing quantitation and characterization of individual mAbs in a mAb cocktail that binds the same protein and is superior to anti-idiotype approaches.     

Generation by phage display and characterization of drug-target complex-specific antibodies for pharmacokinetic analysis of biotherapeutics

Anti-idiotypic antibodies play an important role in pre-clinical and clinical development of therapeutic antibodies, where they are used for pharmacokinetic studies and for the development of immunogenicity assays. By using an antibody phage display library in combination with guided in vitro selection against various marketed drugs, we generated antibodies that recognize the drug only when bound to its target. We have named such specificities Type 3, to distinguish them from the anti-idiotypic antibodies that specifically detect free antibody drug or total drug. We describe the generation and characterization of such reagents for the development of ligand binding assays for drug quantification. We also show how these Type 3 antibodies can be used to develop very specific and sensitive assays that avoid the bridging format.

Abbreviations: BAP: bacterial alkaline phosphatase; CDR: complementarity-determining regions in VH or VL; Fab: antigen-binding fragment of an antibody; HRP: horseradish peroxidase; HuCAL®: Human Combinatorial Antibody Libraries; IgG: immunoglobulin G; LBA: ligand binding assay; LOQ: limit of quantitation; NHS: normal human serum; PK: pharmacokinetics; VH: variable region of the heavy chain of an antibody; VL: variable region of the light chain of an antibody.     

The quality of registration of clinical trials

Background

Lack of transparency in clinical trial conduct, publication bias and selective reporting bias are still important problems in medical research. Through clinical trials registration, it should be possible to take steps towards resolving some of these problems. However, previous evaluations of registered records of clinical trials have shown that registered information is often incomplete and non-meaningful. If these studies are accurate, this negates the possible benefits of registration of clinical trials.

Methods and Findings

A 5% sample of records of clinical trials that were registered between 17 June 2008 and 17 June 2009 was taken from the International Clinical Trials Registry Platform (ICTRP) database and assessed for the presence of contact information, the presence of intervention specifics in drug trials and the quality of primary and secondary outcome reporting. 731 records were included. More than half of the records were registered after recruitment of the first participant. The name of a contact person was available in 94.4% of records from non-industry funded trials and 53.7% of records from industry funded trials. Either an email address or a phone number was present in 76.5% of non-industry funded trial records and in 56.5% of industry funded trial records. Although a drug name or company serial number was almost always provided, other drug intervention specifics were often omitted from registration. Of 3643 reported outcomes, 34.9% were specific measures with a meaningful time frame.

Conclusions

Clinical trials registration has the potential to contribute substantially to improving clinical trial transparency and reducing publication bias and selective reporting. These potential benefits are currently undermined by deficiencies in the provision of information in key areas of registered records.