Guidelines for the quality control of population pharmacokinetic-pharmacodynamic analyses: an industry perspective.

Quality population modeling and simulation analyses and reports are something every modeler desires. However, little attention in the literature has been paid to what constitutes quality regarding population analyses. Very rarely do published manuscripts contain any statement about quality assurance of the modeling results contained therein. The purpose of this manuscript is to present guidelines for the quality assurance of population analyses, particularly with regards to the use of NONMEM from an industrial perspective. Quality guidelines are developed for the NONMEM installation itself, NONMEM data sets, control streams, output listings, output data files and resultant post-processing, reporting of results, and the review processes. These guidelines were developed to be thorough yet practical, though are not meant to be completely comprehensive. It is our desire to ensure that what is reported accurately reflects the collected data, the modeling process, and model outputs for a modeling project.

Population pharmacokinetics-pharmacodynamics of alemtuzumab (Campath) in patients with chronic lymphocytic leukaemia and its link to treatment response.

AIMS: To characterize alemtuzumab pharmacokinetics and its exposure-response relationship with white blood cell (WBC) count in patients with B-cell chronic lymphocytic leukaemia (CLL). METHODS: Nonlinear mixed effects models were used to characterize plasma concentration-time data and WBC count-time data from 67 patients. Logistic regression was used to relate summary measures of drug exposure to tumour response. RESULTS: Alemtuzumab pharmacokinetics were best characterized by a two-compartment model with nonlinear elimination where V(max) (microg h(-1)) was [1020 x (WBC count/10 x 10(9) l(-1))(0.194)], K(m) was 338 microg l(-1), V(1) was 11.3 l, Q was 1.05 l h(-1) and V(2) was 41.5 l. Intersubject variability (ISV) in V(max), K(m), V(1) and V(2) was 32%, 145%, 84% and 179%, respectively. The reduction in WBC over time was modelled by a stimulatory loss indirect response model with values of 18.2 for E(max), 306 microg l(-1) for EC(50), 1.56 x 10(9) cells l(-1) h(-1) for K(in) and 0.029 per h for K(out). The probability of achieving a complete or partial response was >/=50% when the maximal trough concentration exceeded 13.2 microg ml(-1) or when AUC(0-tau) exceeded 484 microg h(-1) ml(-1). CONCLUSIONS: Alemtuzumab displayed time- and concentration-dependent pharmacokinetics with large interpatient variability, both in pharmacokinetics and pharmacodynamics, which was probably reflective of differences in tumour burden among patients. A direct relationship between maximal trough concentrations and clinical outcomes was observed, with increasing alemtuzumab exposure resulting in a greater probability of positive tumour response.

A brief introduction to Monte Carlo simulation.

Simulation affects our life every day through our interactions with the automobile, airline and entertainment industries, just to name a few. The use of simulation in drug development is relatively new, but its use is increasing in relation to the speed at which modern computers run. One well known example of simulation in drug development is molecular modelling. Another use of simulation that is being seen recently in drug development is Monte Carlo simulation of clinical trials. Monte Carlo simulation differs from traditional simulation in that the model parameters are treated as stochastic or random variables, rather than as fixed values. The purpose of this paper is to provide a brief introduction to Monte Carlo simulation methods.

Prospective allometric scaling: does the emperor have clothes?

We are not suggesting that an allometric relationship between pharmacokinetic parameters in animals and humans does not exist; we believe that it does. We are suggesting that using prospective AS to select doses in an FTIM study may lead to a false sense of security given the large publication bias in the literature. There are a number of unrecognized pitfalls to this approach, including (1) prediction intervals so wide as to be of limited use, (2) prediction error is often no better than arbitrarily chosen constants, and (3) it is not possible to determine which drugs will fail a priori. We encourage journals to publish studies in which prospective AS has failed so as scientists we may begin to see what makes these compounds different, with a goal toward better prediction of human pharmacokinetics.

Rank power of metrics used to assess QTc interval prolongation by clinical trial simulation.

Monte Carlo simulation was used to assess the type I error rate and rank order of power for six different metrics using linear mixed-effect models, including two variables recommended by the European Agency for the Evaluation of Medicinal Products (EMEA) in the analysis of QTc interval data. The metrics analyzed were maximal change in QTc interval from baseline, maximal QTc interval, area under the QTc interval-time curve (AUC), average QTc interval, maximal QTc interval with baseline QTc interval as covariate, and AUC with baseline QTc interval as covariate. Two dosing regimens were studied: multiple-dose oral and multiple-dose continuous intravenous infusion. Both regimens were designed to produce similar maximal plasma concentrations, albeit with the infusion regimen maintaining maximal plasma concentrations for a longer period of time. The ability of the metrics to detect a drug effect was examined, assuming drug effect followed either an Emax or linear model. All statistics had a type I error rate near the nominal value. Regardless of pharmacokinetic or pharmacodynamic model, AUC with baseline QTc interval as a covariate had greater power than any other metric examined. The simulations also suggest that mean QTc interval data not be used.

Clinical trial simulation in drug development.

Clinical trial simulation is the application of old technologies, e.g., Monte Carlo simulation, to a new problem, that problem being how to maximize the information content obtained during the drug development process with an intent to have the greatest chance of "success" in a clinical trial. When the information content of the drug is high, then simulation provides a method to synthesize that information into a coherent package that indicates the sponsor has good control over the pharmacology of the drug. From a purely financial point of view, what simulation offers pharmaceutical companies is the possibility of reducing the number of required studies, maximizing the chances for success in a clinical trial, and possibly shortening development time; all outcomes which will reduce drug development costs. The purpose of this paper is to introduce clinical trial simulation to the reader by discussing its potential in drug development, to briefly review the literature, and to make recommendations and caveats regarding its use.

Prospective allometric scaling: does the emperor have clothes?

We are not suggesting that an allometric relationship between pharmacokinetic parameters in animals and humans does not exist; we believe that it does. We are suggesting that using prospective AS to select doses in an FTIM study may lead to a false sense of security given the large publication bias in the literature. There are a number of unrecognized pitfalls to this approach, including (1) prediction intervals so wide as to be of limited use, (2) prediction error is often no better than arbitrarily chosen constants, and (3) it is not possible to determine which drugs will fail a priori. We encourage journals to publish studies in which prospective AS has failed so as scientists we may begin to see what makes these compounds different, with a goal toward better prediction of human pharmacokinetics.

Clonazepam and sertraline: absence of drug interaction in a multiple-dose study.

Thirteen subjects (seven men, six women) completed a placebo-controlled, randomized, double-blind, crossover study to determine whether an interaction occurs between clonazepam and sertraline. Ten days of once-daily doses of either clonazepam 1 mg and placebo (CZ + PL) or clonazepam 1 mg and sertraline 100 mg (CZ + SR) were administered; there was an 11-day washout period. Sertraline did not significantly affect the pharmacokinetics of clonazepam (p > 0.13). Clonazepam apparent oral clearance, volume of distribution, and half-life were 3.9 +/- 0.2 L/hr, 233 +/-11 L, and 40.5 +/- 0.3 hours, respectively. The kinetics of the inactive metabolite 7-aminoclonazepam were marginally affected by sertraline, with a 21% decrease in the elimination half-life (p = 0.03) relative to CZ + PL and no significant difference between treatments in area under the curve or metabolite ratio. Card sorting (CS), digit-symbol substitution test (DSST), nurse-rated sedation scale (NRSS), and self-rated sedation scores were assessed four times daily on days -1 (PL + PL), 1, 4, 7, and 10. There were no differences between treatments in area under the effect curve or maximum observed effect for CS, DSST, or NRSS. Maximum impairment on all assessment days was low, with a less than 10% change from the drug-free values for CS and DSST. Despite higher clonazepam concentrations, predose (time 0) psychomotor and sedation scores did not differ among days -1, 1, 4, 7, and 10 or between treatments. These results in healthy volunteers indicate that sertraline does not affect the pharmacokinetics or pharmacodynamics of clonazepam.

The effect of collinearity on parameter estimates in nonlinear mixed effect models.

PURPOSE: To demonstrate how correlations among predictor variables in a population pharmacokinetic model affect the ability to discern which covariates should enter into the structural pharmacokinetic model. METHODS: Monte Carlo simulation was used to generate multiple-dose concentration-time data similar to that seen in a Phase III clinical trial. The drugs' pharmacokinetics were dependent on two covariates. Five data sets were simulated with increasing correlation between the two covariates. All data sets were analyzed using NONMEM both with and without inclusion of the covariates in the structural pharmacokinetic model. Summary measures for ill-conditioning and sensitivity analysis were used to examine how increasing correlation among covariates affects the accuracy and precision of the parameter estimates. RESULTS: When covariates were included in the structural pharmacokinetic model and the correlation between covariates increased, the standard error of the parameter estimates increased and the value of parameter estimates themselves became increasingly biased. When the correlation between predictor variables was 0.75, the standard errors of the parameter estimates were too large to declare statistical significance. CONCLUSIONS: Correlations among predictor variables greater than 0.5 when entered into the model simultaneously should be a warning to researchers because the (1) the accuracy of the parameter estimates themselves may be biased and (2) the precision of the estimates may be inflated due to ill-conditioning.

Assessment of QTc prolongation for non-cardiac-related drugs from a drug development perspective.

It is increasingly likely that non-cardiac-related drugs will need to be routinely analyzed for their ability to prolong cardiac repolarization. This leads to potential problems for drug companies in general and statisticians in particular. A number of issues immediately arise in regard to answering the question, "Does drug X prolong cardiac repolarization?" These include identifying what is the desired outcome, what is the dependent variable, and what analysis method should be used. The purpose of this article is bring to to light some of the issues regarding the analysis of QTc data, the advantages and disadvantages of these analysis methods, and some general recommendations.

Drug interactions at the renal level. Implications for drug development.

The kidney plays a major role in the elimination of drugs. The purpose of this paper is to: (i) review the mechanisms of renal elimination; (ii) identify potential mechanisms for renal drug interactions; (iii) review in vitro and in vivo animal models for studying renal elimination mechanisms and identifying potential drug-drug interactions; (iv) review experimental designs used in identifying drug-drug interactions in humans with an emphasis on gaining information regarding the mechanism of the interaction; and (v) make recommendations regarding the potential for renal drug interactions in drug development. It is concluded that clinically significant drug interactions resulting in toxicity because of some mechanism at the renal level appear to be relatively rare and that in vitro screening should not be done on all drugs during drug development. Five potential mechanisms exist for drug interactions at the renal level: (i) a displacement of bound drug resulting in an increase in drug excretion via an increase in glomerular filtration; (ii) competition at a tubular secretion site resulting in a decrease in drug excretion; (iii) competition at the tubular reabsorption site resulting in an increase in drug excretion; (iv) a change in urinary pH and/or flow that may increase or decrease drug excretion depending on the pKa of the drug; and (v) inhibition of renal drug metabolism. The most well known renal drug interaction is competitive inhibition of tubular secretion, ultimately leading to an increase in plasma drug concentration. Only when renal clearance is a major contributor to the total clearance (> 30%) and plasma concentrations are greater than the Michaelis-Menten transport constant does the potential exist for clinically significant renal drug-drug interactions because only then does nonlinear pharmacokinetics become evident. The potential for drug interactions is small when renal clearance is less than 20 to 30% of the total clearance and/or when plasma concentrations are less than the Michaelis-Menten transport constant, unless the drug has a narrow therapeutic window.

Coverage and precision of confidence intervals for area under the curve using parametric and non-parametric methods in a toxicokinetic experimental design.

PURPOSE: The coverage and precision of parametric Bailer-type confidence intervals (CIs) for area under the curve (AUC) was compared to nonparametric bootstrap confidence intervals. METHODS: Concentration-time data was simulated using Monte Carlo simulation under a toxicokinetic paradigm with sparse (SSC) and dense sampling (DSC) conditions. AUC was calculated using the trapezoidal rule and 95% CIs were computed using various parametric and nonparametric methods. RESULTS: Under SSC, the various parametric CIs contained the true population AUC with coverage probabilities ranging from 0.77 to 0.95 with low inter-subject variation (coefficient of variation (CV) = 15%) and from 0.82 to 0.95 with high inter-subject variation (CV = 50%). The nominal value should be close to 0.95. DSC tended to increase coverage by about 0.05. Bailer's method always produced the lowest coverage of all parametric CIs examined. Under SSC, bootstrap CIs had coverage probabilities ranging from 0.62 (CV = 15%) to 0.68 (CV = 50%). DSC increased coverage to 0.77. Parametric CIs were wider than their nonparametric counterparts, often giving lower CI estimates less than zero. Bailer's method and Bailer's method using the jackknife estimate of the standard error were the worst in this respect. Bootstrap CIs never had lower CI estimates less than zero. However, SSC tends to produce bootstrap distributions that are not continuous which, if used, may produce biased CI estimates. CONCLUSIONS: Bootstrap CI estimates were judged to be the "best". However, the limitations of the bootstrap should be clearly recognized and it should not be used indiscriminately. Examination of the bootstrap distribution for its degree of discreteness must be part of the statistical process.

Behavioral sensitization to cocaine in the absence of altered brain cocaine levels.

We conducted experiments investigating the role of altered cocaine distribution in behavioral sensitization. The first was designed to determine whether carry-over from one injection to the next occurs after acute cocaine administration. Female, Sprague-Dawley rats were administered 5 mg/kg 3H-cocaine and 24 h later were challenged with either 5 mg/kg unlabeled cocaine or saline. Animals were sacrificed 15 min after drug administration. There was no difference between groups in cocaine levels in brain, liver, or plasma, thus indicating that carry-over did not occur following acute cocaine administration. The second experiment was designed to determine whether bound cocaine could be released following acute or multiple dose cocaine administration. In the acute dose study, animals were administered either 20 mg/kg cocaine or saline, challenged 24 h later with 5 mg/kg 3H-cocaine, and sacrificed 5 min after drug administration. Animals with previous cocaine experience exhibited a significant increase in the number of rearings. The groups did not differ in brain or plasma cocaine levels. In the multiple dose study, animals were injected daily for 4 days with 20 mg/kg cocaine or saline, challenged with 5 mg/kg 3H-cocaine on day 5, and sacrificed 10 min after drug administration. Animals with previous cocaine experience exhibited significantly greater locomotor activity and number of rearings. There was no difference between groups in cocaine levels in various brain regions, plasma, or liver. Brain cocaine content in various regions was significantly correlated, though heterogeneously distributed within the various regions. The highest cocaine levels were found in hippocampus, striatum, thalamus/hypothalamus, and cortex. These results provide further evidence that behavioral sensitization is not the result of cocaine redistribution following repeated administration.

Context-dependent cross-sensitization between cocaine and amphetamine.

We investigated the effect of amphetamine pretreatment on the locomotor response to subsequent cocaine challenge. Rats were administered either 0.75 mg/kg amphetamine in a testing environment and saline in their home cage, saline in the testing environment and 0.75 mg/kg amphetamine in their home cage, or saline in both the testing environment and home cage. After 5 pairings of drug to environment, conditioning was tested by administration of a saline injection. Both amphetamine treated groups exhibited increased locomotion in response to saline injection. After the eighth session of the pairing regimen, all animals were administered 5 mg/kg cocaine in the test environment and their behavior measured for 30 min. Sensitization to cocaine was observed only in rats with previous amphetamine exposure in the testing environment. There was no difference between groups in whole brain, striatal, or plasma levels of cocaine. The data support the hypothesis that sensitization is independent of brain cocaine levels.

Preliminary physiologically based pharmacokinetic model for cocaine in the rat: model development and scale-up to humans.

A physiologically based multicompartmental model has been developed to describe the concentration-time course of cocaine in plasma and tissues in the rat. The compartments included in the model were brain, heart, gut, liver, muscle, fat, venous blood, arterial blood, and a mass-balance compartment. Drug delivery to the tissues was assumed to be flow limited. The model incorporated a nonsaturable binding site for cocaine in the liver. Elimination occurred via both blood and hepatic elimination. The model was validated using independently derived data. The model was scaled to humans and accurately predicted the cocaine levels following intranasal and inhalation administration. However, a poor fit was observed following intravenous administration. Future models incorporating non-constant blood flow and pharmacodynamics need to be developed.

High-performance liquid chromatographic assay for xanomeline, a specific M-1 agonist, and its metabolite in human plasma.

A reversed-phase high-performance liquid chromatographic assay (HPLC) was utilized for monitoring xanomeline (LY246708/NNC 11-0232) and a metabolite, desmethylxanomeline, in human plasma. Xanomeline, desmethylxanomeline and internal standard were extracted from plasm with hexane at basic pH. The organic solvent extract was evaporated to dryness with nitrogen and the dried residue was reconstituted with 0.2 M HCl-methanol (50:50, v/v). A Zorbax CN 150 x 4.6 mm I.D., 5-microns column and mobile phase consisting of 0.5% (5 ml/l) triethylamine (TEA) adjusted to pH 3.0 with concentrated orthophosphoric acid-tetrahydrofuran (THF) (70:30, v/v) produced consistent resolution of analytes from endogenous co-extracted plasma components. Column effluent was monitored at 296 nm/0.008 a.u.f.s. and the assay limit of quantification was 1.5 ng/ml. A linear response of 1.5 to 20 ng/ml was sufficient to monitor plasma drug/metabolite concentrations during clinical trials. HPLC assay validation as well as routine assay quality control (QC) samples indicated assay precision/accuracy was better than +/- 15%.

Factors affecting species differences in the kinetics of metabolites of trichloroethylene.

The hepatocarcinogenicity of trichloroethylene (TRI) in mice has been attributed to a metabolite, trichloroacetate (TCA). Rats of various strains appear to be resistant to TRI-induced hepatocarcinogenesis and produce lower peak concentrations of TCA. Mice, however, also form significant amounts of another carcinogenic metabolite, dichloroacetate (DCA). The present study was conducted to investigate the interspecies differences in the metabolism of TRI between the mouse, rat, and dog and to gain further insight into the role metabolic factors may play in the apparent species specificity of liver tumor induction by TRI. Fischer 344 rats and beagle dogs were dosed orally with TRI and blood was analyzed for TRI, DCA, TCA, and trichloroethanol (TCE). Data on the metabolism of TRI in mice have been previously published. Limited data are available on the metabolism of TRI in humans. Dogs produce higher peak concentrations and have a larger area under the concentration-time curve (AUC) for TCA as compared to rats given similar doses of TRI. Dichloroacetate was not found in measurable concentrations, that is, above 4 nmol/ml, the minimal quantifiable concentration, in the blood of either rats or dogs. Appreciable concentrations of DCA were found in the blood of mice administered TRI in previous studies. Trichloroethanol was found to be present in the blood, urine, and bile, primarily as the glucuronide conjugate. In all species, peak TCA concentrations were observed beyond the disappearance of TRI. The AUC for TCE glucuronide is consistent with its acting as a precursor for TCA and probably contributes to the continued increase in TCA concentration after TRI disappears from the system. Investigations into the binding of TCA to plasma constituents in the rat, dog, mouse, and human suggest that binding also plays a role in species differences in the distribution and elimination of TCA.

Uptake and biodisposition of 2-ylcyanamide-1,3,4-thiadiazole (LY217896) by red blood cells.

In man, 14C-2-ylcyanamide-1 3 4-thiadiazole (LY217896) accumulates into red blood cells (RBCs) where it is rapidly metabolized. Both in man and ex vivo, within a few hours of administration of 14C-LY217896 at least two intracellular metabolites were detected within the RBCs using HPLC. These metabolites were never detected extracellularly. After 24 h no detectable radioactivity was found in the plasma and all the radioactivity was detected within the cellular fraction. All radioactivity was identified as a single peak within the RBCs, indicating the metabolite(s) to be highly polar compared to LY217896. Parent LY217896 was never detected within the RBCs at any time point, suggesting transport, either by diffusion or a carrier mediated mechanism, was the rate limiting step. Due to the nature of the preparation it was impossible to separately characterize uptake and biotransformation. Nevertheless, uptake/biotransformation was found to be temperature sensitive, sodium independent, and energy dependent. Both niacin and vitamin B6, but not nicotinamide, competitively blocked the uptake and subsequent intracellular metabolism of LY217896.

Animal models for studying transport across the blood-brain barrier.

There are many reasons for wishing to determine the rate of uptake of a drug from blood into brain parenchyma. However, when faced with doing so for the first time, choosing a method can be a formidable task. There are at least 7 methods from which to choose: indicator dilution, brain uptake index, microdialysis, external registration, PET scanning, in situ perfusion, and compartmental modeling. Each method has advantages and disadvantages. Some methods require very little equipment while others require equipment that can cost millions of dollars. Some methods require very little technical experience whereas others require complex surgical manipulation. The mathematics alone for the various methods range from simple algebra to complex integral calculus and differential equations. Like most things in science, as the complexity of the technique increases, so does the quantity of information it provides. This review is meant to serve as a starting point for the researcher who wishes to study transport and uptake across the blood-brain barrier in animal models. An overview of the mathematical theory, as well as an introduction to the techniques, is presented.