Caldwell.pdf

Compound optimization in early- and late-phase drug discovery: Acceptable
pharmacokinetic properties utilizing combined physicochemical, in vitro
and in vivo screens
Gary W Caldwell
Keywords β-adrenoceptor blockers, assays, bioavailability,
The RW Johnson Pharmaceutical Research Institute Drug Discovery DepartmentWelsh and McKean RoadsSpring House Introduction
The drug discovery and development process is scientifically complex and full of risk, and is therefore,expensive and time-consuming (Figure 1). Typically, a new Current Opinion in Drug Discovery & Development 2000 3(1):30-41
chemical entity (NCE) is promoted from discovery into preclinical development and if it succeeds in passing all New chemical entities (NCEs) are abandoned in development hurdles, it is submitted for an investigational new drug primarily because of insufficient efficacy, safety issues and for (IND) application and eventually enters phase I, II and III economic reasons. Since efficacy and safety deficiencies are related clinical development. If the compound passes all clinical in part to pharmacokinetics (PK), uncovering PK defects as early trials, it is submitted for a new drug application (NDA) and in drug discovery as possible would be highly valuable in reducing eventually enters the market-place. The average cost to NCE failures in preclinical and clinical development. In this discover and develop a NCE into a marketable drug in the review, a strategy is put forth to integrate drug USA, is typically hundreds of millions of dollars and metabolism/pharmacokinetics and toxicology functions into drug requires a decade or longer to reach the market-place [1]. It discovery. Compound optimization in early- and late-phase drug is clear that there is a critical need for pharmaceutical discovery is covered, emphasizing physicochemical properties, in companies to become more cost- and time-efficient in light vitro absorption, metabolism and in vivo animal PK of spiraling world healthcare expenditures. A significant methodologies, primarily from the period 1998 to 1999. The factor that governs the cost required for NCEs to become present study also illustrates the idea of sorting oral bioavailability marketed drugs is their high attrition rates in preclinical or data into high/intermediate/low categories based on combining clinical development [2]. The proportion of IND applications high/low rank-ordered information from physicochemical that fail has been estimated to be approximately 87% in properties and in vitro absorption, metabolism and serum binding phase I, 60% in phase II and 20% in phase III clinical studies assays. It is shown that by combining the results from solubility, [3•]. Coupling these attrition rates with the large stability, absorption and metabolism assays, the expenditures necessary for phase II and III clinical studies, high/intermediate/low human oral bioavailability for a series of β- produces the major financial problems associated with blockers can be approximately predicted. This method has a high pharmaceutical research. According to data compiled by sample throughput and should be useful in rank-ordering the DiMasi [2], 1943 INDs were filed in the USA between 1964 predicted oral bioavailability of large collections of compounds at and 1989; the total number of IND applications that were the lead optimization step of drug discovery. These results are dropped before reaching NDA status was 1613 or 83%. In useful for selecting compounds for future in vitro/in vivo other words, during this 25-year period, approximately 1 in correlation modeling or in vivo animal testing. This type of 6 NCEs nominated to IND status became a marketed drug.
approach will improve the decision making process of compound To make matters worse, only approximately 1 in 3 marketed drugs typically generates sufficient income to recover the costs Figure 1. The pharmaceutical discovery and development process.
Compound optimization in early- and late-phase drug discovery Caldwell 31
associated with its discovery and development [3].
studies [5]. If defects in pharmacokinetic properties could Therefore, it is generally true that the majority of be recognized and corrected at the drug discovery stage pharmaceutical discovery and development budgets and before they entered development, more time and time are spent on drug failures, and for NCEs that do go to resources could be allocated to projects with real market, only a few are genuinely profitable. Faced with this potential. A compound optimization group would low probability of success, the current trend of certainly enhance any early clinical studies such as the pharmaceutical companies merging is understandable.
drug selection interface group discussed above.
The abandonment of IND candidates is primarily In this review, the importance of a compound attributed to efficacy, safety and economic reasons. It has optimization group that is broadly integrated within a been reported for the period 1964 to 1989, that 46% of 1099 drug discovery department is discussed. Methods for IND applications were discontinued due to unacceptable compound optimization in early- and late-phase drug efficacy (eg, weak or lack of efficacy), 27% due to safety discovery are covered, emphasizing: (i) physicochemical issues (eg, toxicity), 23% due to economics (eg, limited properties (log P, pK , solubility, etc); (ii) in vitro and in market) and 5% for miscellaneous reasons (eg, vivo biophysical characteristics, such as absorption, unclassified reasons) [2]. This trend is consistent with a distribution, metabolism and excretion (ADME); (iii) smaller, but more recent study by Kennedy [3], where 46% pharmacokinetics; and (iv) toxicology. While the of 121 NCEs in clinical development were discontinued importance of human oral bioavailability data for due to unacceptable efficacy, 40% due to safety issues, 7% compound optimization in drug discovery is due to economics and 7% for miscellaneous reasons.
unquestioned, it is rarely available due to cost- and time- Therefore, insufficient pharmacological efficacy, human consuming experimental challenges. The present study adverse reactions and toxicity are estimated to account for 50% to 86% of the NCEs dropped from development. The high/intermediate/low human oral bioavailability based challenges of pharmacoeconomic predictions have on rank-ordered physicochemical properties and in vitro recently been addressed and will not be discussed here [3]. Human in vivo efficacy and safety predictions of NCEs are extremely difficult to forecast utilizing the presentpharmaceutical process, and the vast majority ofpharmaceutical companies are focused on reducing Terminology and definitions
clinical development attrition rates by attempting to Successful drug candidates typically have good biological properties such as potency, selectivity, efficacy and oral pharmacoeconomics at a much earlier stage than before.
bioavailability. It is important to have a reasonableunderstanding of these properties since some terms are To address efficacy and safety attrition rates, some used interchangeably in the literature, eg, potency for companies are reorganizing their traditional discovery efficacy and absorption for bioavailability. This section and development departments from two independent will briefly review their definitions. Potency refers to the functions, to organizations that have a drug selection amount of compound needed to produce a given interface group between these departments. There are biological effect and the terms activity and potency are several jargon phraseologies given to this type of interface used interchangeably in this review. Selectivity infers that group, namely, 'proof-of-concept' or 'proof-of-principle'.
undesirable side effects are minimized or eliminated, The primary mission of a drug selection interface group is while efficacy refers to the maximum level of biological to increase the quality and quantity of human data prior effect a compound can produce. The oral bioavailability of to full phase I clinical development. Typically, a drug compounds is primarily dictated by the following serial candidate is tested in a limited group of human subjects rates: liberation, absorption, metabolism and elimination (eg, 1 to 15) together with formal pharmacokinetic and [6]. The simple diagram shown in Figure 2 can be used to toxicology analysis. It is expected that this type of define these terms. Oral ingestion is generally the safest, information will indicate potential development problems, most convenient and most economical method for and thereby, significantly reduce the overall attrition rate compound administration, and when a compound is of NCEs in future clinical studies. Another strategy is to given orally, the liberation rate is defined as the net integrate drug metabolism/pharmacokinetics and transfer of compound from the mouth and stomach to the toxicology functions into a compound optimization group small intestine. Typically, compounds are released from either partially or entirely within drug discovery. It has formulations that may depend upon disintegration, been argued that uncovering human pharmacokinetic dissolution, solubility, surface area and chemical or defects (eg, oral bioavailability, half-life, metabolites, enzymatic stability. In other words, the liberation rate drug-plasma protein binding, etc) in drug discovery represents the total amount of intact compound available would be highly valuable in reducing NCE failures in at the small intestine after oral dosing. The absorption rate preclinical and clinical development, since efficacy and involves the net transfer of compounds from the safety deficiencies are related in part to pharmacokinetic gastrointestinal fluid across primarily the small intestine problems [4]. The jargon phraseology applied to this type into the portal blood system. There are several mechanisms of compound uptake; the main processes pharmacokinetics' or 'just-in-time pharmacokinetics'. The available are passive diffusion and carrier transport.
goal of this compound optimization group is to eliminate Intestinal P-glycoprotein (P-gp) is another transporter compounds with pharmacokinetic defects utilizing human system that needs to be considered. P-gp is expressed at the tissues, human-derived cell lines and/or in vivo animal lumenal surface of the intestinal epithelium, and therefore, 32 Current Opinion in Drug Discovery & Development 2000 Vol 3 No 1
Figure 2. Routes from the gastrointestinal tract into the systemic circulation.
acts to oppose the uptake of compounds into the portal This fraction or a percentage is normalized for the different blood system [7]. The amount of compound that passes compound doses (D) given by the two routes. Alternately, F through the intestinal tissues must pass through the liver and may be subjected to first-pass metabolism effects. Attimes, first-pass effects can prevent effective concentrations of compounds from reaching the hepaticblood system and eventually the general systemic Where f represents the net fraction of an oral dose liberated circulation. The cytochrome P450 (CYP) enzyme system, from the formulation that reaches the small intestine, the net which is primarily located in the smooth endoplasmic fraction absorbed across the apical membrane of the reticulum of liver cells, and in smaller quantities in the epithelial cell is denoted by f , and f represents the net kidney, lung and gastrointestinal epithelium, is fraction escaping the first-pass hepatic effect.
responsible for the monooxygenase metabolism ofcompounds [8]. Sometimes drug metabolites formed in After introduction into the portal circulation system, the liver are excreted back into the intestinal tract via the compounds can bind to various constituents such as tissue bile. These metabolites are either excreted in the feces or proteins, cell proteins and blood proteins. Compound reabsorbed into the portal blood system and ultimately binding to various blood proteins and tissue proteins is excreted in the urine. Compounds that reach the systemic important because it can influence the therapeutic, circulation either unchanged or as metabolites are pharmacodynamic and toxicological action of certain excreted by the urinary system. Sublingual and rectal drugs. Competition for binding to tissues and blood administration of compounds have the advantage that the proteins is likely to occur between different compounds if compound is somewhat protected from rapid first-pass present at the same time. Human blood consists of three metabolism by the liver, however, these routes are not as major systems: (i) formed elements (ie, erythrocytes, leukocytes and platelets); (ii) a fluid portion; and (iii) largeamounts of various salts [9]. The major cell body in the The amount of orally-administered compound reaching the blood is the erythrocyte (ie, red blood cell), which systemic circulation is measured from the ratio of the area comprises approximately 95% of the total cellular fraction under the plasma-concentration versus time curve after oral in the blood. There are three major components in the administration (AUC) to that after intravenous (AUC) erythrocyte capable of binding compounds - hemoglobin, administration. Thus, the oral bioavailability (F) is defined carbonic anhydrase and the cell membrane. If blood is allowed to naturally coagulate, a clear straw-colored fluid(ie, serum) can be separated from the cellular fraction by F = [(AUC)oral /( AUC)iv]*[ iv centrifugation. In contrast, plasma is obtained by Compound optimization in early- and late-phase drug discovery Caldwell 33
centrifugation of blood to which an anticoagulant was relevant physicochemical, ADME, pharmacokinetics and added immediately after removal from the body. Thus, toxicology criteria in the decision making process of drug serum contains water (90 to 92%), all blood proteins (6 to discovery. In Figure 3, the organizational format of a 8%) and various salts (eg, 0.1 M NaCl), while plasma broadly-based drug discovery department that is used to contains water, proteins minus the clotting factors and illustrate this idea is shown. Compound hit generation, salts. The concentration of various serum proteins can lead optimization and candidate selection steps are the vary individually and on a daily basis by as much as 10% three main decision points used to drive the overall of their average value. It is also interesting to note that process to produce drug development candidates. The serum protein concentration levels can be highly affected points have input from informatics (ie, bio- and by certain physiological and pathological conditions.
chemoinformatics, molecular modeling, and computer Human serum albumin (HSA), α -acid glycoprotein and automation science), biology (ie, biochemistry, (AGP), the high-density lipoproteins (HDL) and the low- molecular biology, cell biology and pharmacology), density lipoproteins (LDL), are the most important serum chemistry (ie, combinatorial, parallel and scale-up proteins responsible for the binding of compounds in synthesis) and drug metabolism/toxicology (ie, serum. Typically, HSA is largely responsible for serum pharmacy, pharmacokinetics). This type of format binding of acidic and neutral compounds, whereas AGP naturally allows the composition of the department to and lipoproteins bind mainly basic compounds. HSA and change as the discovery process advances, ie, early drug AGP serve as depots and transport proteins for numerous discovery emphasizes more basic research in biology and endogenous and exogenous compounds. Among the chemistry, while late stage discovery shifts to more endogenous substances bound with high affinity to blood applied research. It allows for the elimination of proteins are long chain fatty acids, L-tryptophan and compounds utilizing a combination of activity and bilirubin. The blood protein-exogenous complex (ie, blood biophysical data at each decision point. This type of protein-drug complex) acts as a transport mechanism to organization also provides a maximum feedback loop carry compounds to the sites of action; this transport is extremely important for compounds that exhibit low metabolism/toxicology scientists. In other words, solubility in the water portion of the serum. In some cases, structural modifications suggested by metabolic and the circulating protein-compound complex also serves to toxicity data are incorporated into the synthetic plan. We replenish the free compound that is removed by various will describe each decision point emphasizing the input distribution and elimination processes. Thus, it maintains from physicochemical, ADME and in vivo free compound concentration at its therapeutic level and pharmacokinetic assays for the selection of drug provides a mechanism that prolongs the duration of development candidates. It is not the intention of this compound action. Therefore, determination of the report to exhaustively review the array of important drug concentration of unbound or bound compound with blood metabolism/toxicology and absorption assays that proteins is an important parameter to measure to establish abound in the literature [11,12•,13-16], rather, it is to the importance of oral bioavailability values.
highlight some important assays in the drug discoverysetting and make general comments concerning their use.
The drug discovery process
Traditionally, drug discovery groups have focused
The goal of the 'hit generation' step is to screen large primarily on compound synthesis and target screening. In compound libraries in a relatively short amount of time in this type of environment, medicinal chemists synthesize an attempt to find compounds that cause a specific compounds in order to maximize in vitro or in vivo biological response, ie, 'hits'. This step includes target potency, selectivity and efficacy for a relevant biological identification, selection, validation and high-throughput target. Optimized structure-activity relationships (SARs) screening (HTS) of large structurally-diverse compound are generated based on in vitro potency versus structural libraries. The productivity in target identification and modifications; however, in many cases, all other selection has improved with the development of physicochemical properties (log P, pK , solubility, etc), automated DNA sequencing, genomics databases and ADME, pharmacokinetic and toxicological properties have bioinformatic tools [17]. However, target validation been ignored until a later time. As mentioned earlier, remains a time-consuming process where assays are uncovering pharmacokinetic defects is highly valuable in typically performed without automation. HTS operations reducing NCE failures since efficacy and safety are highly automated to handle sample preparation, assay deficiencies are related in part to pharmacokinetics [4].
procedures and large volume data processing. After each The consequence of the exclusion of these properties in step is optimized to operate efficiently, it is common to drug discovery has been to waste time and resources in screen 100,000 compounds in a 1- to 6-month period development groups by putting them in a 'patch and [18,19]. As new 'mix and read' detection assays are mend' strategy. It is clear that the way to avoid this developed, HTS is moving toward ultra-HTS which will situation is to perform many of the traditional screen over 100,000 compounds per day [20,21•].
development physicochemical, ADME, pharmacokinetics Compound libraries are typically created by combinatorial and toxicological studies in drug discovery [10•].
and parallel synthesis paradigms [22•], and HTS can Unfortunately, many of these traditional in vitro and in identify thousands of hits or only a few from these vivo assays are not well adapted for the higher-throughput libraries. The hit rate for HTS bioassays is set somewhat screening that is necessary in drug discovery. Using a arbitrarily, that is, the activity cut-off is lowered until an hierarchical organizational approach and with recent adequate number of hits is obtained. Cluster analyses can methodological advances, it is feasible to add some also be performed on thousands of hits in order to reduce 34 Current Opinion in Drug Discovery & Development 2000 Vol 3 No 1
Figure 3. The organizational format of a broadly-based drug discovery department.
the number of hits to an adequate number [23]. It is approaches typically use QSAR techniques, knowledge- interesting to note that libraries of either virtually based systems or neural network modeling to predict assembled compounds or actual compound libraries have physicochemical properties and relevant biological been computationally screened utilizing three- (3-D) or parameters. While these calculations are well established, four-dimensional (4-D) quantitative structure-activity the accurate prediction of reliable data is still debatable.
relationship (3-D/4-D-QSAR) techniques to generate hits There is a serious disadvantage in selecting or eliminating [24,25]. Virtual screening is a promising new technique for hits based on calculated or screening data of the discovery of high-affinity hits [26]. Hits can also be physicochemical, ADME, pharmacokinetics or toxicology produced from 'me-too' compounds (close structural data at this point. It should be remembered that it is rare derivatives of a known active drug). The de novo design of for a hit compound to have all of the desirable properties new ligands bound in 3-D receptor protein structures necessary for it to proceed directly to phase I clinical using molecular modeling methods represents another studies without structural modifications, furthermore, it is source of hits [27]. The exact number of hits chosen is generally true that any single property of a compound can largely dictated by the in vitro potency, the compound be optimized; however, this optimized property is often patentability, the complexity of the analog chemistry, achieved at the expense of other properties. If substantial chemistry head count and the duration of the drug structural manipulation to the prototype hit is required to discovery program. In addition to these criteria, it could obtain the desired potency, selectivity and efficacy, then be of interest for certain projects to incorporate all favorable physicochemical, ADME, pharmacokinetics physicochemical, ADME, pharmacokinetics and and toxicology properties obtained in the original hit may toxicology selection criteria at this decision point.
be reduced or forfeited in the process. Therefore, it is Unfortunately today, there are no in vitro or in vivo HTS conceivable that nothing will be gained by screening for methods capable of assaying 100K compound libraries in a biophysical characteristics at this point in drug discovery.
It is more advantageous to screen the analogs for pharmacokinetics and toxicology properties. DNA favorable physicochemical, ADME, pharmacokinetics and microarray technology or DNA chips is a promising toxicology properties during the lead optimization step.
method used to monitor changes in expression at themRNA level [28]. While unproven at this time, it is Once useful structural prototypes have been obtained from conceivable that the DNA chip will discover pertinent the hit generation step, analog construction around these toxicodynamic markers [29] that can be used in reporter templates utilizing traditional solution- and solid-phase gene, branched-DNA or other HTS assays for early chemistries are initiated so as to improve, primarily, in vitro toxicology assessment [30]. There has been considerable potency [22]. This process is referred to as the 'hit-to-lead effort made in computer-aided physicochemical property optimization step' or 'lead optimization step'. The types of [31], solubility [32,33], intestinal uptake [33], permeability structural modifications incorporated into the hit are usually [34], metabolism [35] and toxicity [36] predictions. These governed by in vitro potency data and by the wisdom of the Compound optimization in early- and late-phase drug discovery Caldwell 35
medicinal chemist. If the hit has a 3-D X-ray crystallographic pK [45], octanol-water partition coefficients [46], compound-receptor protein structure, computational thermodynamic solubility-pH profiles [47], and liposomal chemistry (ie, molecular modeling) techniques are an membrane-water partition coefficients [48] utilizing important method for lead optimization [27]. Typically, potentiometric techniques have been developed.
hundreds to thousands of analogs are required in the lead Immobilized artificial membrane (IAM) chromatography optimization step to select compounds worthy of can be used to model membrane partitioning of advancement to the candidate selection step and receive compounds [49]. IAMs are solid-phase membrane more extensive in vivo animal testing. The exact number of mimetics that are prepared by covalently immobilizing analogs synthesized per biological target is largely dictated monolayers of cell membrane phospholipids to silica by the potency, efficacy and selectivity deemed necessary particles at high molecular surface densities, and are used for the target, the personnel available, the complexity of the as a chromatographic stationary-phase to mimic the lipid analog chemistry and the duration of the drug discovery environment of a fluid cell membrane [50]. The analyte program. Since the total number of analogs for any retention (capacity) factors on IAM chromatographic particular target is synthesized over some time period, an columns have been shown to predict drug permeability iterative strategy can be used to organize the workload. For across the blood-brain barrier [51]. An alternative to IAM example, if 1000 analogs per target are synthesized over a 1- has been suggested recently utilizing lysophospholipid year period, then for 30 targets, roughly 2500 samples per micellar electrokinetic chromatography [52]. The Caco-2 month need to be tested. There are only a few pertinent (immortalized human colon adenocarcinoma) and the physicochemical properties, ADME and toxicology assays MDCK (Madin-Darby canine kidney) cell lines have been that have medium sample throughput capable of handling used as an in vitro model for assessing membrane 2500 samples per month. Turbidimetric methods for permeability [53,54]. These cell lines differentiate on measuring kinetic solubilities of compound have been microporous filter membranes into columnar epithelium estimated to handle 6000 samples per month [33].
and form tight cellular junctions. The utility of liquid Permeation properties are related to transcellular compound chromatography/mass spectrometric (LC/MS) methods absorption. The use of artificial membranes in a 96-well to measure the apparent permeability (P ) coefficients of plate-based format to predict permeation properties of compounds from a Caco-2 cell culture intestinal model compounds has been estimated to handle 10,000 to 20,000 samples/month [37]. Hepatotoxicity studies, which give electrophoresis/frontal analysis has been used to screen information on the maximum compound concentration drugs interacting with human serum and human serum compatible with cell survival, can be performed using the 3- proteins [56]. The rate of metabolism utilizing rat, monkey (4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide and human primary hepatocytes has been demonstrated (MTT) colorimetric method [38]. The assay is based on the using a sample handling system for 96-well plates directly ability of mitochondrial enzymes in live cells to reduce a coupled to a LC/MS, ranking compounds based on their tetrazolium salt into a colored formazan dye. A 96-well resistance towards metabolism [39]. The rationale behind plate-based assay using a multi-well scanning these experiments is that compounds that are resistant to spectrophotometer has been described [39]. Recently, an metabolism are more likely to exhibit low in vivo first-pass HTS technique utilizing rat liver microsomes and a pulsed effects. Reverse transcription-polymerase chain reaction ultrafiltration mass spectrometer has been developed (RT-PCR) assays have been developed to study the [40•,41]. Pulsed ultrafiltration-electrospray mass induction of human cytochromes P450 (1A1, 2A6, 2E1 and spectrometry is an attractive method for in vitro formation 3A3/4) mRNA in cell culture systems [57]. The various and mass spectrometric characterization of metabolites in an assays outlined above can be used to simply rank-order automated fashion. It was stated that the method had the compounds or to select compounds based on a validated potential for automated HTS of up to 60 samples (ie, in vitro/in vivo correlation. The ranking of compounds profiles) per hour. Fluorometric assays for assessing based on these assays should be considered as a first step cytochrome P450 drug-drug inhibition for the five principal to reduce the number compounds for future in vitro/in vivo drug-metabolizing enzymes, CYP1A2, CYP2C9, CYP2C19, correlation modeling or in vivo testing. To be successful, CYP2D6 and CYP3A4 have been developed in a 96-well compounds with the desired activity and the fewest in plate-based assay [42,43,44••]. The judicious application of vitro biophysical defects or the best in vitro/in vivo these techniques in rank-ordering compounds for more correlations must be selected. Based on our experience, extensive in vitro/in vivo correlation modeling or in vivo the combination of a thermodynamic solubility assay [47], animal testing may be useful; however, one should be a Caco-2 permeability assay [55], a hepatocyte or cautious since the data will contain many false-positive and microsome assay to estimate hepatic extraction ratios [39], false-negative results. Experience indicates that a drug-blood protein binding assay [56] and a P450 drug- physicochemical, ADME, pharmacokinetic and toxicology drug inhibition assay [42], gives the most pertinent information is only essential for the subset of analogs that information. Valuable information is also gained from demonstrate the correct in vitro potency range deemed studying the profile of metabolites produced in rat and necessary for the project to be viable. Using the example human hepatocytes or microsomes. It should be noted that above, for 2500 samples per month, we have found it interpretation of results from all of the assays listed above necessary to assay approximately 10% of these could be hotly debated, since it will be necessary to compounds or 250 compounds per month. If we only eliminate compounds with activity at this point. In a later assay those compounds that are considered to have the section, it is shown how a combination of rank-order correct activity, then there are many low-throughput in screening assays can provide a high/intermediate/low vitro assays at our disposal. Approaches for measuring prediction of human oral bioavailability.
36 Current Opinion in Drug Discovery & Development 2000 Vol 3 No 1
Once the selection criteria have been defined in the lead Intuitively, one would suspect that maximum oral optimization step, compounds are promoted to the bioavailability (eg, ≥ 90%) should occur when a compound 'candidate selection' step where the goal is to assay has a high liberation rate (eg, the net transfer of compounds utilizing primarily in vivo animal models.
compound from the mouth and stomach to the small Using in vivo data obtained from animal species and intestine), a high absorption rate (eg, net transfer of allometric scaling, human pharmacokinetic parameters compounds from the gastrointestinal fluid across can be predicted, such as volume of distribution, primarily the small intestine into the portal blood system) clearance, half-life and oral bioavailability [58]. Note that and low metabolism rate (eg, first-pass effect). Of course at this stage, the function of the chemistry group has the minimum concentration of compound in the systemic shifted to scale-up chemistry (Figure 3) that is necessary, system (eg, low oral bioavailability, ≤ 10%) should occur since larger quantities of compounds are required for in when the compound has a low liberation rate, a low vivo studies. The number of candidates at this point is absorption rate and a high metabolism rate. All other totally arbitrary. For example, assume we have rank- combinations of liberation, absorption and metabolism ordered 250 compounds from one analog series, our can be considered as having intermediate oral experience indicates that we only need to assay bioavailability values using the above arguments. Thus, if approximately 10% of these compounds or 25 compounds we could devise liberation, absorption and metabolism per target. There are again several low-throughput in vivo assays that ranked compounds in a high/low manner, it assays at our disposal. An effective approach for screening should be possible to predict into which oral a large number of samples is to simultaneously measure bioavailability range a compound would fall by utilizing multiple analytes in a single analysis. In vivo cassette- the scheme in Table 1. The liberation rate of a compound dosing pharmacokinetic studies utilizing LC/MS have can be approximated by examining the dissolution rate been successful via the oral [59] and intravenous routes and the compound's chemical stability. Since compounds [60]. Alternatively, by using automated procedures to are primarily solutions or suspensions in drug discovery, handle plasma preparation, and rapid LC/MS techniques the dissolution rate can be approximately assessed by the to perform quantitation, we measure the oral solubility of the compound assuming that the surface area bioavailability of singly-dosed compounds in a rapid of the compound is not a factor. Thus, the dissolution rate manner. With this technique, many hundreds of will not be a limitation for compounds having reasonable compounds per year can be measured. Utilizing these fast aqueous solubility (> 0.1 mg/ml) [33]. The chemical quantitation techniques, it is possible to perform short- stability of compounds at low gastric pH values can be term in vivo toxicity experiments in order, for example, to approximately assessed by stability studies at pH 2 over a study exposure levels. Microdialysis sampling is a flexible time period equivalent to the compound's mean residence technique for the study of in vivo pharmacokinetics of the time in the stomach (human ≈ 75 min). That is, we can extracellular fluid in the brain [61]. In recent years, measure the disappearance of compounds at pH 2 for 75 microdialysis with LC/MS detection has emerged as an min. Thus, compounds with aqueous solubility > 0.1 important tool in biochemical research [62]. The various in mg/ml and chemical stability > 50% (arbitrarily selected) vivo assays outlined above can be used to identify potential can be considered to have high liberation rates, while pharmacokinetic defects, such as low oral bioavailability, those with solubilities < 0.1 mg/ml and chemical stability short half-life, active metabolites, etc. Based on in vivo and in < 50% can be considered to have low liberation rates vitro data, attempts can be made to correct these problems, (Table 1). All other combinations of solubility and stability and if successful, these compounds are promoted to drug will be considered to have a low liberation rate. The Caco- development candidacy and enter preclinical development.
2 cell line is used as an in vitro model to study drugtransport in the intestinal epithelium [55]. Compounds Prediction of human oral bioavailability at the
that are completely absorbed in humans have lead optimization step
permeability rates typically > 1.0 x 10-6 cm/s, compounds Based on equation 2, if accurate human in vitro/in vivo that are absorbed > 1%, but ≤ 100%, have permeability correlation models to predict the f , f and f were rates of 0.1 x 10-6 to 1.0 x 10-6 cm/s and compounds that are available, human oral bioavailabilities (F) could be absorbed < 1% have permeability rates ≤ 1.0 x 10-7 cm/s.
estimated. There are several good reviews outlining Thus, compounds with permeability rates > 1.0 x 10-6 cm/s methods to predict human oral bioavailability based on in can be considered to have high absorption rates, while vitro (human/animal)/in vivo (animal) metabolism data those with < 1.0 x 10-6 cm/s can be considered to have low [63,64]. The individual variation in oral bioavailability absorption rates (Table 1). It should be emphasized, that measurements can be very large due to genetic and these permeability cut-offs are somewhat variable environmental factors, and thus, cannot be accurately or depending upon the experimental condition (eg, stirred readily predicted. To take into account this natural versus unstirred water layers) [55], however, the separation variability, our strategy is simply to sort oral of the Caco-2 data into high/low categories is typically not a bioavailability data into high/intermediate/low categories problem. Once the compound has passed from the small based on combining rank-ordered information from several intestine to the portal system, the liver may metabolize a in vitro assays. This method has a higher sample throughput considerable portion of the compound prior to entering the than in vitro/in vivo correlation modeling, and therefore, systemic circulation. The prediction of metabolism rates can should be useful in sorting large collections of compounds at be studied using isolated human hepatocyte cells and/or the lead optimization step for future in vitro/in vivo various liver subcellular fractions [39,65]. The rate of first- correlation modeling or in vivo animal testing.
pass metabolism can be roughly estimated into high/low Compound optimization in early- and late-phase drug discovery Caldwell 37
Table 1. Oral bioavailability classified into high/intermediate/low categories based on in vitro rank-ordered liberation, absorption
and metabolism data.
Liberation ranking
Absorption ranking
Metabolism ranking
Oral bioavailability
categories by examining the rate of drug disappearance (or the vivo first-pass effect data, we observe that the in vitro human rate of appearance of the metabolites) from human hepatocyte hepatocyte cell data could approximately differentiate the in or microsomal suspensions. Thus, compounds with hepatocyte vivo first-pass effect data in a high/low manner. That is, stability > 50% (arbitrarily selected) can be considered to have propranolol, alprenolol and oxprenolol were highly low metabolism rates, while those with hepatocyte stability < metabolized in vitro and also exhibited a high in vivo first-pass 50% can be considered to have high metabolism rates (Table 1).
effect. Timolol (3), pindolol, acebutolol and atenolol were
Thus, a simple high/low strategy could be used to roughly resistant towards metabolism by human hepatocytes and also estimate first-pass metabolism effects. Alternately, when exhibited a low in vivo first-pass effect. However, it should be concentration studies and scaling techniques are performed, it noted that metoprolol (6) did not fit this pattern, suggesting in
has been found that drugs that had high in vivo extraction vitro human hepatocyte cell data obtained at a single ratios had in vitro hepatocyte clearance values > 30 µl/min/106 concentration and time point do not completely correlate with cells, while drugs that had low in vivo extraction ratios had in in vivo first-pass effect data. Finally, if we combine the vitro hepatocyte clearance values < 10 µl/min/106 cells [65].
high/low liberation, absorption and metabolism result fromTable 2 with the scheme outlined in Table 1, we can predict the An homologous series of β-adrenoceptor blocking drugs (β- high/intermediate/low oral bioavailability for the β-blockers.
blockers, Figure 4 and Table 2) were chosen as model Note that timolol and pindolol were correctly chosen as having compounds to illustrate the above scheme, since their high bioavailability, however, metoprolol was predicted to pharmacokinetic parameters display a wide variability [66,67].
have a high oral bioavailability, while the experimental data These compounds have molecular weights that range from 248 suggests its value is closer to an intermediate value. Clearly, Da (4, pindolol) to 336 Da (1, acebutolol), and dissociation
the in vitro methods discussed here are able to distinguish β- constants (pK ≈ 9) which minimize the influence of these blockers with favorable characteristics. We believe the above parameters on solubility, stability, absorption and metabolism.
procedure will work reasonably well for other analog series The pseudo-thermodynamic solubility was measured using a and probably worse for structurally-diverse sets of simple LC/MS technique. Briefly, the compound was saturated at pH 7.4 (ionic strength = 0.15), shaken for 30 min, As mentioned earlier, the prediction of drug-serum binding is filtered through a nylon filter and assayed by LC/MS. The an important factor to understand along with oral stability was measured by preparing the sample at pH 2 and bioavailability. It has been reported that capillary measuring the disappearance over a 75 min period. Data for electrophoresis/frontal analysis (CE/FA) can be used in an the β-blockers are listed in Table 2 and indicate that these automated manner to estimate the blood protein-drug binding drugs are all highly liberated. For the Caco-2 transport in the association constants [56]. The percentage of drug bound in apical to basolateral direction, 50 µM of the β-blocker of serum samples or artificial mixtures of serum proteins samples interest was placed in the insert (apical side) [55]. The insert (eg, HSA + AGP) can also be determined. Briefly, CE/FA is was moved seven times over a 1 h period to wells containing used to determine the free drug concentration in a drug- fresh buffer. These well (basolateral) samples were spiked with protein binding equilibrium. In CE/FA, sample volumes that an internal standard and analyzed by LC/MS. The Papp represent approximately 5 to 7% of the total volume of the coefficients obtained from this experiment are listed in Table 2.
capillary are injected onto the capillary to result in a frontal The results for the β-blockers indicated that acebutolol and peak shape for the drug. Based on the frontal theory, the free atenolol (2) had a low absorption rate, while all of the other
drug concentration is directly determined from the height of drugs were highly absorbed. If we compare the P coefficients the frontal peak. In the normal CE polarity mode, the basic β- to the in vivo human absorption data of the β-blockers (Table blockers elute first and are resolved from the negatively 2), we observe that the P coefficients correctly distinguish the charged serum proteins. We have shown that the binding poorly absorbed β-blockers (acebutolol and atenolol) from the capacity of the β-blockers with human protein mixtures of highly absorbed β-blockers. After known drug concentrations AGP, HSA, HDL and LDL has the same high/low ranking (5 µM) were incubated with hepatocytes for 6 h, LC/MS order as that obtained with human serum samples. The results, procedures were developed to determine the amount of drug from this experiment indicated that propranolol had which was metabolized (ie, % metabolized) [39]. These results significantly higher binding to these proteins than the other β- are listed in Table 3 and indicate that propranolol (8),
blockers. Therefore, timolol and pindolol have higher oral alprenolol (7) and oxprenolol (5) were highly metabolized. If
bioavailabilities and lower serum binding profiles than we compare the in vitro human hepatocyte cell data with the in 38 Current Opinion in Drug Discovery & Development 2000 Vol 3 No 1
Figure 4. β-adrenoceptor blocking agents.
1 acebutolol
2 atenolol
3 timolol
4 pindolol
(Sandoz)
5 oxprenolol
6 metoprolol
7 alprenolol
8 propranolol
(Wyeth-Ayerst)
Table 2. In vitro and in vivo biophysical data for a series of β-adrenoceptor blocking drugs.
Solubility1
Stability2
Liberation
In vivo
Absorption
Hepatocytes5
In vivo
Metabolism
acebutolol
atenolol
pindolol
oxprenolol
metoprolol
alprenolol
propranolol
1The drugs were prepared at pH 7.4 (ionic strength = 0.15) in an aqueous solution, shaken for 30 min, filtered through a nylon filter andassayed by LC/MS. The units are mg/ml.
2The stability (% remaining) was measured by preparing the sample at pH 2 and measuring the disappearance over a 75 min period. Thedata are normalized to the starting concentration at time = 0.
3Caco-2 P coefficients (see reference [55] for experimental details). The data are normalized to propranolol (7.5 ± 1.2 x 10-5 cm/s).
4In vivo human absorption (%) (see references [66] and [67]). The data are normalized to propranolol (90%). The net fraction absorbedacross the apical membrane of the epithelial cell is denoted by f .
5In vitro human hepatocyte cell data (% metabolized in 6 h at 5 µM of drug). See reference [39] for experimental details. The data arenormalized to the starting concentration at time = 0 for propranolol. In vivo human first-pass effect (%) (see references [66] and [67]). Thedata are normalized to propranolol (60%).
6The notation f represents the net fraction escaping the first-pass hepatic metabolism effect.
Compound optimization in early- and late-phase drug discovery Caldwell 39
Table 3. Comparison of the predicted and experimental oral bioavailability values for a series of β-adrenoceptor blocking drugs.
Liberation
Absorption
Metabolism
Predicted oral
Experimental oral
ranking1
ranking1
ranking1
bioavailability2
bioavailability3
acebutolol
atenolol
pindolol
oxprenolol
metoprolol
alprenolol
propranolol
1The ranking of the drugs was based on results from Table 2.
2The ranking of the oral bioavailability was based on the system used in Table 1.
3The in vivo human oral bioavailability (%) values are from references [66] and [67].
Conclusion
References
This review has focused on integrating drugmetabolism/pharmacokinetics and toxicology functions into drug discovery in order to reduce the attrition rates in clinical development. Three main decision points are used in drug 1. DiMasi JA: Risks, regulation, and rewards in new drug
discovery to drive the overall process to produce superior development in the United States. Regul Toxicol Pharmacol
drug development candidates. These include hit generation, (1994) 19:228-235.
lead optimization and candidate selection steps. Theelimination of compounds occurs at each decision point 2. DiMasi JA: Success rates for new drugs entering clinical
utilizing a combination of activity, physicochemical properties, testing in the United States. Clin Pharmacol Ther (1995)
absorption, metabolism, in vivo pharmacokinetics and toxicity 58:1-14.
assays. This type of organization provides a maximum 3. Kennedy T: Managing the drug discovery/development
feedback loop between medicinal chemists and drug interface. Drug Disc Today (1997) 2:436-444.
metabolism/toxicology scientists. Literature has been cited • A good review of the skill sets necessary to manage an interface primarily between 1998 to 1999. At the lead optimization step, between discovery and development. a combination of a thermodynamic solubility assay [47], aCaco-2 permeability assay [55], a hepatocyte or microsome 4. Caldwell J: The importance of drug metabolism studies for
assay to estimate hepatic extraction ratios [39], a drug-blood efficient drug discovery and development. Yakubutsu Dotai
protein binding assay [56], and a cytochrome P450 drug-drug (1996) 11(1):119-126.
inhibition assay [42] gives the most pertinent information. The 5. Caldwell GW: Novel biophysical analytical methods: The
profile of metabolites produced in rat and human hepatocytes utility of combining physicochemical and in vitro
or microsomes is also suggested at this step. These higher- physiological proprieties of drug compounds for estimating
throughput assays can be rank-ordered to prioritize in vivo oral bioavailability. Second International Symposium on
compounds for more extensive in vitro/in vivo correlation Drug-Drug Interaction Potential. Baltimore, MD, USA (1997).
modeling or in vivo animal studies. A simple example was 6. Benet LZ, Mitchell JR, Sheiner LB: Goodman and Gilman's - used to illustrate this idea. Utilizing a series of β-adrenoceptor The Pharmacological Basis of Therapeutics. 8th Edition.
blocking drugs, a strategy was developed to sort oral Gilman AG, Rall TW, Nies AS, Taylor P (Eds), Pergamon bioavailability data into high/intermediate/low categories based on combining rank-ordered information from several invitro assays. This oral bioavailability ranking procedure 7. Spahn-Langguth H, Baktir G, Radschuweit A, Okyar A, Terhaag B, Ader P, Hanafy A, Langguth P: P-glycoprotein transporters
worked reasonably well for the β-adrenoceptor blocking and the gastrointestinal tract: Evaluation of the potential in
drugs, and therefore, should work for other analog series. The vivo relevance of in vitro data employing talinolol as model
predictions will probably be worse for structurally-diverse sets compound. Int J Clin Pharmacol Ther (1998) 36:16-24.
of compounds. The candidate selection step is used to assaycompounds utilizing primarily in vivo animal models to gain 8. Lewis DF (Ed): Cytochromes P450: Structure, Function and information on key pharmacokinetic parameters and drug Mechanism. Taylor & Francis Ltd, London (1996):1-348.
exposure levels [60]. While the above procedure appears to be 9. Reif OW, Lausch R, Freitag R: High-performance capillary
a reasonable start, clearly much remains to be done to improve electrophoresis of human serum and plasma proteins. Adv
drug selection in discovery and development.
Chromatogr (1994) 34:1-56.
Acknowledgements
10. Heykants J, Meuldermans W: Nonclinical kinetics and
metabolism studies in support of the safety assessment of
The author would like to acknowledge the contribution of drugs. Drug Inf J (1994) 28:163-172.
Per A Peterson, Michael Ernest, Keith Demarest, Joseph • A good review of the studies necessary for preclinical and clinical Gunnet, William Hageman, Eric Ericson, John A Masucci, Patrick Sasso, William Jones, Amy Maden, MaryEvangelisto, Zhengyin Yan, Scott M Easlick and Patricia A 11. Rodrigues AD: Use of in vitro human metabolism studies in
McDonnell. Allen Reitz and Edward L Tolman are gratefully drug development. An industrial perspective. Biochem
acknowledged for reviewing the manuscript.
Pharmacol (1994) 48(12):2147-2156.
40 Current Opinion in Drug Discovery & Development 2000 Vol 3 No 1
12. Rodrigues AD: Preclinical drug metabolism in the age of
27. Bohm H-J: Computational tools for structure-based ligand
high-throughput screening: An industrial perspective.
design. Prog Biophys Mol Biol (1996) 66(3):197-210.
Pharm Res (1997) 14(11):1504-1510.
• A good review of some of the potential problems encountered in 28. Gerhold D, Rushmore T, Caskey CT: DNA chips: promising
preclinical drug metabolism and rational screening strategies. toys have become powerful tools. Trends Biochem Sci
(1999) 24(5):168-173.
13. Rodrigues AD: Rational high-throughput screening in
preclinical drug metabolism. Med Chem Res (1998)
29. Johnson DE, Braeckman RA, Wolfgang GH: Practical aspects
8(7/8):422-433.
of assessing toxicokinetics and toxicodynamics. Curr Opin
Drug Discovery Dev (1999) 2(1):49-57.
14. Tarbit MH, Berman J: High-throughput approaches for
evaluating absorption, distribution, metabolism and
30. Todd MD, Ulrich RG: Emerging technologies for accelerated
excretion properties of lead compounds. Curr Opin Chem
toxicity evaluation of potential drug candidates. Curr Opin
Biol (1998) 2:411-416.
Drug Discovery Dev (1999) 2(1):58-68.
15. Smith DA, van de Waterbeemd H: Pharmacokinetics and
31. Wildman SA, Crippen GM: Prediction of physicochemical
metabolism in early drug discovery. Curr Opin Chem Biol
parameters by atomic contributions. J Chem Inf Comput Sci
(1999) 3:373-378.
(1999) 39(5):868-873.
16. Sinko PJ: Drug selection in early drug development:
32. Huuskonen J, Salo M, Taskinen J: Aqueous solubility
Screening for acceptable pharmacokinetic properties
prediction of drugs based on molecular topology and
using combined in vitro and computational approaches.
neural network modeling. J Chem Inf Comput Sci (1998)
Curr Opin Drug Discovery Dev (1999) 2(1):42-48.
38(3):450-456.
17. Benton D: Integrated access to genomic and other
33. Lipinski CA, Lombardo F, Dominy BW, Feeney PJ: bioinformation: an essential ingredient of the drug
Experimental and computational approaches to estimate
discovery process. SAR QSAR Environmental Research
solubility and permeability in drug discovery and
(1998) 8(3-4):121-155.
development setting. Adv Drug Deliv Rev (1997) 23:3-25.
18. Broach JR, Thorner J: High-throughput screening for drug
34. Stenberg P, Luthman K, Ellens H, Lee CP, Smith PL, Lago A, discovery. Nature (1996) 384:14-16.
Elliott JD, Artursson P: Prediction of the intestinal
absorption of endothelin receptor antagonists using three
19. Silverman L, Campbell R, Broach JR: New assay
theoretical methods of increasing complexity. Pharm Res
technologies for high-throughput screening. Curr Opin
(1999) 16(10):1520-1526.
Chem Biol (1998) 2:397-403.
35. Erhardt PW: Drug metabolism data: Past and present
20. Mere L, Bennett T, Coassin P, England P, Hamman B, Rink T, status. Med Chem Res (1998) 8:400-421.
Zimmerman S, Negulescu P: Miniaturized FRET assays and
microfluidics: key components for ultra-high-throughput
36. Cronin MT: Computer-aided prediction of drug toxicity in
screening. Drug Disc Today (1999) 4(8):363-369.
high throughput screening. Pharm Pharmacol Commun
(1998) 4:157-163.
21. Winkler T, Kettling U, Koltermann A, Eigen M: Confocal
fluorescence coincidence analysis: an approach to ultra
37. Kansy M, Senner F, Gubernator K: Physicochemical high
high-throughput screening. Proc Natl Acad Sci USA (1999)
throughput screening: Parallel artificial membrane
96(4):1375-1378.
permeation assay in the description of passive absorption
• An interesting paper outlining an approach to screen 100,000 processes. J Med Chem (1998) 41(7):1007-1010.
38. Plumb JA: Cell sensitivity assays: the MTT assay. Methods
22. Fecik RA, Frank KE, Gentry EJ, Menon SR, Mitscher LA, Mol Med (1999) 8(Cytotoxic drug resistance mechanisms):25-30.
Telikepalli H: The search for orally active medications through
combinatorial chemistry. Med Res Rev (1998) 18:149-185.
39. Caldwell GW, Masucci JA, Chacon E: High throughput liquid
• A good review of combinatorial chemistry with ideas concerning chromatography-mass spectrometry assessment of the
the incorporation of ADME properties in the design of libraries. metabolic activity of commercially available hepatocytes
from 96-well plates. Comb Chem High-Throughput Screen
23. Stanton DT, Morris TW, Roychoudhury S, Parker CN: (1999) 2(1):39-51.
Application of nearest-neighbor and cluster analyses in
pharmaceutical lead discovery. J Chem Inf Comput Sci
40. Van Breemen RB, Nikolic D, Bolton JL: Metabolic screening
(1999) 39(1):21-27.
using on-line ultrafiltration mass spectrometry. Drug Metab
Dispos (1998) 26(2):85-90.
24. de Julian-Ortiz JV, Galvez J, Munoz-Collado C, Garcia- • An interesting approach for doing online drug metabolism. Domenech R, Gimeno-Cardona C: Virtual combinatorial
syntheses and computational screening of new potential
41. Nikolic D, Fan PW, Bolton JL, Van Breemen RB: Screening
anti-herpes compounds. J Med Chem (1999) 42(17):3308-
for xenobiotic electrophilic metabolites using pulsed
ultrafiltration-mass spectrometry. Comb Chem High-
Throughput Screen (1999) 2(3):165-175.
25. Ekins S, Bravi G, Binkley S, Gillespie JS, Ring BJ, Wikel JH, Wrighton SA: Three and four dimensional-quantitative
42. Crespi CL, Miller VP, Penman BW: Microtiter plate assays
structure activity relationship (3D/4D-QSAR) analyses of
for inhibition of human, drug-metabolizing cytochromes
CYP2D6 inhibitors. Pharmacogenetics (1999) 9(4):477-489.
P450. Anal Biochem (1997) 248(1):188-190.
26. Wang J, Ramnarayan K: Toward designing drug-like
43. Crespi CL: Higher-throughput screening with human
libraries: A novel computational approach for prediction of
cytochromes P450. Curr Opin Drug Discovery Dev (1999)
drug feasibility of compounds. J Combinatorial Chem (1999)
2(1):15-19.
1(6):524-533.
Compound optimization in early- and late-phase drug discovery Caldwell 41
44. Crespi CL, Miller VP, Penman BW: High throughput
56. McDonnell PA, Caldwell GW, Masucci JA: Using capillary
screening for inhibition of cytochrome P450 metabolism.
electrophoresis/frontal analysis to screen drugs
Med Chem Res (1998) 8(7/8):457-471.
interacting with human serum proteins. Electrophoresis
•• An outstanding paper for screening the inhibition of P450 (1998) 19(3):448-454.
57. Mattes WB, Li AP: Quantitative reverse transcriptase/PCR
45. Avdeef A, Box KJ, Comer JE, Gilges M, Hadley M, Hibbert C, assay for the measurement of induction in cultured
Patterson W, Tam KY: pH-Metric log P 11. pK determination
hepatocytes. Chem Biol Interact (1997) 107:47-61.
of water-insoluble drugs in organic solvent-water
mixtures. J Pharm Biomed Anal (1999) 20(4):631-641.
58. Sanwald-Ducray P, Dow J: Prediction of the
pharmacokinetic parameters of reduced-dolasetron in man
46. Takacs-Novak K, Avdeef A: Interlaboratory study of log P
using in vitro/in vivo and interspecies allometric scaling.
determination by shake-flask and potentiometric methods.
Xenobiotica (1997) 27(2):189-201.
J Pharm Biomed Anal (1996) 14(11):1405-1413.
59. Olah TV, McLoughlin DA, Gilbert JD: The simultaneous
47. Avdeef A: pH-Metric solubility. 1. Solubility-pH profiles
determination of mixtures of drug candidates by liquid
from Bjerrum plots. Gibbs buffer and pK in the solid state.
chromatography/atmospheric pressure chemical
Pharm Pharmacol Comm (1998) 4(3):165-178.
ionization mass spectrometry as an in vivo drug screening
procedure. Rapid Commun Mass Spectrom (1997) 11:17-23.
48. Avdeef A, Box KJ, Comer JE, Hibbert C, Tam KY: pH-Metric
logP 10. Determination of liposomal membrane-water
60. Bayliss MK, Frick LW: High-throughput pharmacokinetics:
partition coefficients of ionizable drugs. Pharm Res (1998)
Cassette dosing. Curr Opin Drug Discovery Dev (1999)
15(2):209-215.
2(1):20-25.
49. Yang CY, Cai SJ, Liu H, Pidgeon C: Immobilized artificial
61. Wong PS, Yoshioka K, Xie F, Kissinger PT: In vivo
membranes - screens for drug-membrane interactions. Adv
microdialysis/liquid chromatography/tandem mass
Drug Deliv Rev (1997) 23(1-3):229-256.
spectrometry for the on-line monitoring of melatonin in rat.
Rapid Commun Mass Spectrom (1999) 13(5):407-411.
50. Caldwell GW, Masucci JA, Evangelisto M, White R: Evaluation
of the immobilized artificial membrane
62. Masucci JA, Ortegon ME, Jones WJ, Shank RP, Caldwell GW: phosphatidylcholine. Drug discovery column for high-
In vivo microdialysis and liquid chromatography/
performance liquid chromatographic screening of drug-
thermospray mass spectrometry of the novel
membrane interactions. J Chromatogr A (1998) 800(2):161-169.
anticonvulsant 2,3:4,5-bis-O-(1-methylethylidene)-β-D-
fructopyranose sulfamate (topiramate) in rat brain fluid. J
51. Ducarne A, Neuwels M, Goldstein S, Massingham R: IAM
Mass Spectrom (1998) 33(1):85-88.
retention and blood brain barrier penetration. Eur J Med
Chem (1998) 33:215-223.
63. Obach RS, Baxter JG, Liston TE, Silber BM, Jones BC, MacIntyre F, Rance DJ, Wastall P: The prediction of human
52. Masucci JA, Caldwell GW, Foley JP: Comparison of the
pharmacokinetic parameters from preclinical and in vitro
retention behavior of β-blockers using immobilized
metabolism data. J Pharmacol Exp Ther (1997) 283(1):46-58.
artificial membrane chromatography and lysophospholipid
micellar electrokinetic chromatography. J Chromatogr A
64. Iwatsubo T, Hirota N, Ooie T, Suzuki H, Shimada N, Chiba K, (1998) 810(1-2):95-103.
Ishizaki T, Green CE, Tyson CA, Sugiyama Y: Prediction of in
vivo drug metabolism in the human liver from in vitro
53. Yee S: In vitro permeability across Caco-2 cells (colonic)
metabolism data. Pharmacol Ther (1997) 73(2):147-171.
can predict in vivo (small intestinal) absorption in man -
fact or myth. Pharm Res (1997) 14(6):763-766.
65. Houston JB, Carlile DJ: Prediction of hepatic clearance from
microsomes, hepatocytes, and liver slices. Drug Metab Rev
54. Irvine JD, Takahashi L, Lockhart K, Cheong J, Tolan JW, (1997) 29(4):891-922.
Selick HE, Grove JR: MDCK (Madin-Darby canine kidney)
cells: A tool for membrane permeability screening. J
66. Meier J: Pharmacokinetic comparison of pindolol with
Pharm Sci (1999) 88(1):28-33.
other beta-adrenoceptor-blocking agents. Am Heart J
(1982) 104(2):364-373.
55. Caldwell GW, Easlick SM, Gunnet J, Masucci JA, Demarest K: In vitro permeability of eight β-blockers through Caco-2
67. Hinderling PH, Schmidlin O, Seydel JK: Quantitative
monolayers utilizing liquid chromatography/electrospray
relationships between structure and pharmacokinetics of
ionization mass spectrometry. J Mass Spectrom (1998)
beta-adrenoceptor blocking agents in man. J
33(7):607-614.
Pharmacokinet Biopharm (1984) 12(3):263-287.

Source: http://gmp.asso.fr/Documents/Biblio/caldwell.pdf