Human microdose evaluation of the novel EP1 receptor antagonist GSK269984A


Dr Thor Ostenfeld MBChB, PhD, Medicines Research Centre (European CEEDD), GlaxoSmithKline R&D Ltd., Gunnels Wood Road, Stevenage SG1 2NY, UK. Tel.: +44 14 3879 0466. Fax: +44 14 3876 2412. E-mail:



• The microdose administration of novel drug candidates to humans in early development is currently undergoing evaluation as a cost-efficient approach to the early assessment of pharmacokinetics (PK) before the commitment of resources required to support formal phase 1 studies.

• The microdose approach assumes that PK can be extrapolated linearly over the full range of exposures achieved with sub-pharmacological doses up to the therapeutic dose range.

• Few microdose studies have been undertaken in the context of pharmaceutical drug development and their precise role in the selection of candidates for further clinical evaluation at therapeutic doses has yet to be fully substantiated.


• The present study describes the elective application of a human microdose study with a novel EP1 receptor antagonist, GSK269984A, to address a critical development liability posed by uncertainty with respect to the predicted human PK profile.

• Microdose data revealed a favourable PK profile, consistent with a clinically acceptable dosing regimen. These data support the value of undertaking a microdose study early in the drug discovery process to facilitate risk evaluation and to enable decision-making.

AIM The primary objective was to evaluate the pharmacokinetics (PK) of the novel EP1 antagonist GSK269984A in human volunteers after a single oral and intravenous (i.v.) microdose (100 µg).

METHOD GSK269984A was administered to two groups of healthy human volunteers as a single oral (n= 5) or i.v. (n= 5) microdose (100 µg). Blood samples were collected for up to 24 h and the parent drug concentrations were measured in separated plasma using a validated high pressure liquid chromatography-tandem mass spectrometry method following solid phase extraction.

RESULTS Following the i.v. microdose, the geometric mean values for clearance (CL), steady-state volume of distribution (Vss) and terminal elimination half-life (t1/2) of GSK269984A were 9.8 l h−1, 62.8 l and 8.2 h. Cmax and AUC(0,∞) were 3.2 ng ml−1 and 10.2 ng ml−1 h, respectively; the corresponding oral parameters were 1.8 ng ml−1 and 9.8 ng ml−1 h, respectively. Absolute oral bioavailability was estimated to be 95%. These data were inconsistent with predictions of human PK based on allometric scaling of in vivo PK data from three pre-clinical species (rat, dog and monkey).

CONCLUSION For drug development programmes characterized by inconsistencies between pre-clinical in vitro metabolic and in vivo PK data, and where uncertainty exists with respect to allometric predictions of the human PK profile, these data support the early application of a human microdose study to facilitate the selection of compounds for further clinical development.


The EP1 receptor is one of four G-protein coupled receptor subtypes (EP1–4) activated by the arachidonic acid metabolite prostaglandin E2 (PGE2) [1]. Studies using prostaglandin E2 synthase (PGES) knock-out animals [2, 3] and PGES inhibitors [4, 5] provide evidence for the role of PGE2 as a pro-inflammatory mediator. Drugs that block prostanoid synthesis non-selectively, by inhibition of the cycloxygenase (COX) enzymes, have a well established place in the management of pain and inflammation. There is now a substantial body of pre-clinical data obtained using a variety of models that provides evidence for the relationship between EP1 receptor activation and hyperalgesia [6–12]. EP1-selective antagonists have been shown to be efficacious in models of inflammatory pain and they therefore offer the opportunity to develop a novel class of anti-inflammatory medicines with an improved benefit–risk profile over the existing COX inhibitors [13–15].

The medicinal chemistry and lead optimization programme leading to the discovery and selection of GSK269984A (sodium 6-[(5-chloro-2-{[(4-chloro-2-fluorophenyl)methyl]oxy}phenyl)methyl]-2-pyridinecarboxylate: see Figure 1) as a candidate for development as a treatment for inflammatory pain has been reported previously together with the key in vitro and in vivo characteristics at the time of candidate selection [16]. In vitro recombinant assay systems have shown GSK269984A to be a potent, functional competitive antagonist at the human EP1 receptor with nanomolar activity. The compound shows a 100–10 000 fold selectivity for EP1 over other key prostaglandin targets, although poor selectivity over the thromboxane A2 (TP) receptor. Further screening assays revealed no significant off-target activity for GSK269984A.

Figure 1.

Structure of GSK269984A

GSK269984A has been shown to possess analgesic efficacy in models of inflammation [16]. In the rat complete Freund's adjuvant (CFA) model of inflammatory pain, orally administered GSK269984A produced a dose-dependent reversal of hypersensitivity (ED50 2.6 mg kg−1). Full reversal of hypersensitivity was achieved at 10 mg kg−1, equivalent to the standard control (GW855454X, 30 mg kg−1 p.o [17]). Efficacy of GSK269984A was also demonstrated following oral dosing for 5 days (ED50circa 3 mg kg−1, orally) in a rat model of chronic inflammatory joint pain [18] and full reversal of hypersensitivity (equivalent to rofecoxib) was achieved at 10 mg kg−1.

In drug metabolism and pharmacokinetic (DMPK) studies the in vitro metabolic stability of GSK269984A was profiled using microsomes derived from mouse, rat, dog, monkey and human liver. This revealed a low intrinsic clearance (CLi) across all species (≤ 0.7 ml min−1 g−1 liver) [16]. Further studies, undertaken with hepatocytes and S9 fraction to include phase 2 metabolic pathways, similarly provided evidence of low CLi and low metabolic turnover across all tested species. Interestingly, the CLi data for GSK269984A were found not to predict the in vivo PK profile observed across three pre-clinical species. These data showed GSK269984A blood clearance (CLb) to be high in the monkey and moderate in the rat and dog, which was reflected in the respective species terminal half-lives [16]. Oral bioavailability (relatively high in the rat (94%), moderate in the dog (39%) and low in the monkey (7%) [16]) was therefore considered to be limited by first-pass hepatic extraction in each of the species. Whilst the solubility and permeability data suggested that the compound should diffuse well across cell membranes, and there was no evidence that GSK269984A is a P-glycoprotein (P-gp) substrate, the steady-state volume of distribution (Vss) differed across the preclinical species being highest in the rat (circa 2.1 l kg−1) and similar in the dog and monkey (circa 0.6 l kg−1) even though the plasma protein binding was comparable across all three species (99.9% (rat and human). 99.8% (dog) and 99.7% (monkey). This would suggest that, in the rat at least, drug transporters (other than P-gp) may be involved in the distribution of GSK269984A.

Further profiling of GSK269984A in vitro revealed a low potential for inhibition of CYP1A2, 2C9, 2C19, 2D6 and 3A4 (IC50 51, 20, 85, 100 and >100 µm, respectively) [16]. However, the possible risk that drug−drug interactions might occur at GSK269984A concentrations relevant to efficacy was raised by data suggesting the potential for time-dependent inhibition of human CYP3A4 (approximately 2-fold reduction in IC50), and for inhibition of the OATP1B1 transporter (IC50circa 1 µm). Preliminary studies to investigate GSK269984A biotransformation in vitro (rat and human liver S9 fraction, as well as rat, monkey and human hepatocytes), revealed the formation of an acyl glucuronide in all species.

For the purposes of GSK269984A clinical dose predictions, three alternative scenarios were considered (Table 1). Collectively, they raised considerable uncertainty with respect to the likely human PK profile. Firstly, based on simple allometric scaling of in vivo PK parameters obtained from rat, dog and monkey, the human PK predictions indicated a high CLb (circa 90% of liver blood flow), low Vss (circa 0.3 l kg−1), low oral bioavailability (circa 10%), and a short terminal half-life (circa 0.2 h). Using these predictions, it was estimated that a daily dose of 11 g GSK269984A would be required to maintain efficacious plasma concentrations, and would necessitate an unacceptably frequent dosing regimen (circa 450 mg h−1) to accommodate the short terminal half-life.

Table 1. Predicted PK parameters and dose for GSK269984A; comparison with known PK parameters for marketed NSAIDs
Compound Clinical pharmacokinetic parameter Adult clinical dose (mg day−1)
CLb (ml min−1 kg−1) V ss (l kg−1) t 1/2 (h) F p.o. (%)
  • 1

    . Predictions based on simple allometric scaling of CLb and Vss using data from all pre-clinical species (rat, dog and monkey). Simple allometry was performed using the power equation:

  • image
  • where Y is the pharmacokinetic parameter of interest (CLb or Vss), W is the bodyweight, log a is the y-intercept and b is the slope obtained from the plot of log Y vs. log W [51].

  • Predicted terminal half-life (t1/2) was calculated using the predicted values of CLb and Vss from allometry according to the equation [52]:

  • image
  • Oral bioavailability (Fp.o.) was predicted using the following equation:

  • image
  • where Qh is liver blood flow (assumed to be 21 ml min−1 kg−1 for human).

  • 2

    . Prediction based on single species scaling from rat to man. In this case Vss rat =Vss human and CLb human were assumed to be proportionally equivalent to CLb rat based on respective liver blood flows (Qh) i.e.:

  • image
  • where Qh rat is assumed to be 90 ml min−1 kg−1. t1/2 and Fp.o. were calculated as described in (1).

  • 3

    . Predictions for GSK269984A assuming a PK profile comparable with common NSAIDs i.e. low CLb (20% of Qh); low Vss and reasonable Fp.o. (50%).

  • 4

    . Dose prediction estimates for GSK269984A assume an efficacious concentration (EC90) of 0.575 µg ml−1 (1.4 µm) and comparable brain: blood distribution of GSK269984A across species. The EC90 was estimated based on a simple pharmacokinetic (blood concentrations at 1 h post dose) and pharmacodynamic analysis (% reversal of hypersensitivity) from the rat joint pain model of inflammatory pain [16]. Daily required dose estimates for each scenario were calculated according to the equation CLb = R/Css corrected for oral bioavailability (Fp.o.); where CLb = blood clearance, Css= steady-state efficacious concentration at the EC90 (0.575 µg ml−1) and R = dosing rate (mg day−1) [53].

GSK269984A (rat, dog, primate)1 190.30.21011 0004
GSK269984A (rat)2 52.17.8803254
GSK269984A (NSAID like)3 40.22505504
Ibuprofen 0.750.152.0801 600–2 400
Naproxen 0.130.1614.099500–1 000

A second scenario was based on the assumption that the PK profile for GSK269984A in man would be similar to the most favourable preclinical PK profile, that seen in the rat [16]. Based on this single species scaling scenario the estimated human CLb for GSK269984A is low (circa 20% liver blood flow), with a larger Vss (circa 2.1 l kg−1), a longer terminal half-life of 7.8 h and an oral bioavailability of circa 80%. Using these PK estimates, a single once daily dose of circa 325 mg GSK269984A would be sufficient to maintain efficacious concentrations.

A third scenario was based on reference to the known human PK parameters for commonly available non-steroidal anti-inflammatory drugs (NSAIDs) containing the carboxylic acid moiety (akin to GSK269984A), exemplified by the propionic acid (ibuprofen, naproxen) and acetic acid class (diclofenac, indomethacin) ([19] and see Table 1). Such compounds have relatively low clearance and volumes of distribution, but good oral bioavailability and half-lives that would support an acceptable dosing regimen. If GSK269984A is assumed to have a comparably low human CLb (circa 20% liver blood flow), low Vss (circa 0.2 l kg−1), and a moderate t1/2 (circa 2 h) and oral bioavailability (circa 50%), then a total daily dose of circa 550 mg of GSK269984A can be predicted to maintain efficacious concentrations. Furthermore, a half-life of circa 2 h would support a dosage regimen of circa 200 mg three times a day.

In view of the absence of an in vitro/in vivo correlation of CLb across the pre-clinical species, the use of simple allometric scaling to facilitate human PK predictions for GSK269984A was recognized to be of questionable value, and posed a significant risk to further investment in the clinical development of the compound. Human microdose studies are currently being evaluated as a strategic (Phase 0) translational approach to the early clinical evaluation of compounds prior to the initiation of formal phase 1 studies and related enabling activities [20–24]. They have been the subject of recent regulatory guidance documents [25, 26] and enable the administration of a human dose that does not exceed 1/100th of that calculated to produce a pharmacological effect, with a maximum dose of 100 µg. Here we report on the PK evaluation of GSK269984A quantified in plasma using a highly sensitive high pressure liquid chromatography/tandem mass spectrometry (HPLC/MS/MS) method following oral and i.v. administration of sub-pharmacological doses to human subjects.


The study was performed under a GlaxoSmithKline protocol [EPB107015] from July to September 2006 at the James Lance GSK Medicines Research Unit, Sydney, Australia, in accordance with Good Clinical Practice (GCP), applicable regulatory requirements and the guiding principles of the Declaration of Helsinki. Ethical approval for the investigation was granted by the Human Research Ethics Committee of the South Eastern Sydney Area Health Service. Written informed consent was obtained from all subjects prior to their enrolment in the study.

Study participants

Prospective subjects were required to attend a screening medical visit within 28 days of dosing. Key eligibility criteria for inclusion were for subjects to be adult males, aged 18 to 45 years inclusive, in good health, body weight ≥50 kg, body mass index in the range 18.5–29.9 kg m−2, with normal 12-lead and 24 h Holter electrocardiogram (ECG) and normal clinical laboratory investigations. Key exclusion criteria were recent use of prescription drugs, positive pre-study urine screen for drugs of abuse, positivity for hepatitis B surface antigen, hepatitis C, or human immunodeficiency virus (HIV I or II), history of alcohol abuse or dependence, smokers or positive carbon monoxide breath test and recent clinical trial participation. The use of recreational drugs, alcohol and caffeine, as well as strenuous exercise, was restricted or prohibited for the duration of the study. For each treatment session, subjects were screened for drugs of abuse and alcohol prior to all other assessments.

Study design

The primary objective of the study was to assess the human pharmacokinetics of GSK269984A following single i.v. and oral microdoses of GSK269984A with a secondary objective to assess the safety and tolerability of the microdoses.

The study was conducted according to an open-label, single dose design on two separate groups of five healthy male subjects in the fasted state. The first group of five subjects received a single dose of GSK269984A (equivalent to 100 µg of the free acid) administered as an i.v. infusion over 1 h. Following satisfactory review of PK and safety data, a second group of five subjects received a single oral dose of GSK269984A (equivalent to 100 µg of the free acid). Both were administered as freshly prepared solutions in 5% w/v dextrose (20 µg ml−1).

Subjects were confined to bed rest for the duration of the dosing period and until 6 h post dose. All subjects remained resident in the clinical unit for no less than 24 h post dose.

Safety and tolerability evaluation comprised assessments of adverse events (AEs), vital signs (heart rate and blood pressure), electrocardiograms (12-lead ECG, lead-II ECG monitoring and Holter ECG) and clinical laboratory data (haematology and clinical chemistry).

Sample collection and bioanalysis

Blood samples for pharmacokinetic analysis were taken pre dose and post dose at 5, 10, 15, 30, 45 and 60 min, every 30 min between 1 and 4 h, every hour between 4 and 8 h, and thereafter at 10, 12, 18 and 24 h. Whole blood was collected into potassium EDTA tubes and plasma was separated by centrifugation at 3000 rev min−1 (4°C, 10 min) for storage at −20°C until analysis. Plasma samples were analyzed for GSK269984A free acid by Drug Metabolism and Pharmacokinetics, GSK, UK using a validated analytical method based on solid phase extraction (SPE), followed by HPLC/MS/MS.

GSK269984A free acid and its isotopically labelled internal standard ([13C6]-GSK269984) were extracted from human plasma by solid phase extraction (SPE) using an Oasis HLB µElution SPE plate conditioned with 200 µl of methanol and then 200 µl water. The plate was loaded with 100 µl 4% phosphoric acid to blank wells and 100 µl of internal standard working solution (1 ng ml−1 in dimethyl formamide) to all other wells, followed by 250 µl of the plasma samples. Wells were washed with 200 µl 50% methanol in water and the samples were eluted with a 50 µl aliquot of 50% methanol in acetonitrile. Extracts were analyzed by HPLC/MS/MS using an Agilent 1100 HPLC system (Waldbronn, Germany) with a CTC HTS Pal autosampler (CTC Analytics, Zwingen, Switzerland) coupled to an API-4000 triple quadrupole mass spectrometer controlled through Analyst™ and equipped with a TurboIonspray® (TISP) source (Applied Biosystems, Toronto, Canada). Aliquots of plasma extract were separated using a Hypersil GOLD, 5 µm column (50 × 4.6 mm i.d., ThermoHypersil, Cheshire, UK) using a mobile phase composition of 10 mm ammonium acetate at pH 2.5 (A) and acetonitrile (B) at 1 ml min−1 under isocratic conditions [A : B, 25:75 (v/v)]. The flow was directed to a TISP interface operating in the positive-ion mode and GSK269984A free acid and [13C6]-GSK269984 were detected using multiple reaction monitoring (MRM) of m/z 406 >143 and m/z 412 >143, respectively. Chromatographic peaks were integrated using Analyst™ (Applied Biosystems, Toronto, Canada). Calibration curves were constructed by fitting the GSK269984A free acid : [13C6]-GSK269984 peak area ratio to standard concentrations using the weighted (1/x2) linear regression model. Linearity was observed over the concentration range 0.02 to 20 ng ml−1 with a correlation coefficient >0.99. Concentrations of GSK269984A free acid in plasma and quality control samples were interpolated from this plot.

Based on quality control samples, the overall relative SD (an index of precision) was less than 12.9%. The overall relative bias (an index of accuracy) was within ± 14.2%. The lower limit of quantification (LLOQ) was validated at 0.02 ng ml−1 (circa 49 pm).

Pharmacokinetic analysis

PK analyses of plasma GSK269984A free acid concentration–time data were conducted using the non-compartmental Models 200 (for extravascular administration) and 202 (for constant infusion) of WinNonlin Professional Edition version 4.1.a (Pharsight Corporation, Mountain View, CA, USA). Actual elapsed time from dosing was used to estimate all individual plasma PK parameters. Actual observed values following administration of GSK269984A were used to estimate the maximum observed plasma concentration (Cmax), the first time to reach Cmax (tmax), and the absorption lag time (or time of sample preceding first quantifiable concentration; tlag). Where possible, the terminal plasma elimination rate constant (λz) was estimated from log-linear regression analysis of the terminal phase of the plasma concentration–time profile. The numbers of points included in the terminal phase were selected by visual inspection of a semi-log plot of the plasma concentration−time profile. The associated apparent terminal elimination half-life (t1/2) was calculated as t1/2= ln2/λz. The area under the plasma concentration−time curve for time zero to the last quantifiable time point [AUC(0,t)] and extrapolated to infinity [AUC(0,∞)] were calculated by a combination of linear and logarithmic trapezoidal methods. The linear trapezoidal method was used for all incremental trapezoids arising from increasing concentrations and the logarithmic trapezoidal method was used for those arising from decreasing concentrations. The percentage of AUC(0,∞) obtained by extrapolation (%AUCex) was calculated as [(AUC(0,∞) − AUC(0,t)]/AUC(0,∞) × 100]. Systemic clearance (CL), following i.v. administration was calculated by Dose/AUC(0,∞). Volume of distribution at steady-state was calculated by CL multiplied by mean residence time (MRT) following i.v. administration. Absolute bioavailability (F) was calculated as AUC(0,∞) divided by dose after oral administration, divided by AUC(0,∞) divided by dose after i.v. administration.

Summary statistics (n, arithmetic mean, 95% confidence interval about the arithmetic mean, SD, minimum, median, maximum) were calculated for all PK parameters. For AUC(0,t), AUC(0,∞), Cmax, CL, Vss and t1/2, data were loge transformed and the geometric mean, the 95% confidence interval about the geometric mean and the between subject coefficient of variation (CVb%) were calculated. CVb(%) was determined as: CVb(%) = 100 × square root (exp(SD2) − 1), where SD is the standard deviation of the loge transformed data.


Five male subjects were enrolled into each group in the study. Subjects were aged 25–44 [mean (SD), 34.8 (7.19)] years and 20–32 [26.6 (4.45)] years for the i.v and oral groups, respectively. The mean weights were 81.6 (SD 10.3) kg and 76.5 (SD 9.6) kg, respectively with mean BMI of 26.4 (SD 2.2) kg m−2 and 24.0 (SD 1.5) kg m−2, respectively. Excluding one subject in the oral group who was of mixed race, all other subjects were white Caucasians. All 10 subjects completed the study as planned and were included in the safety population and in the PK analyses (Table 2). Mean plasma concentration–time profiles for GSK269984A following single i.v and oral dose administration are shown in Figure 2 and the PK parameters are summarized in Table 3.

Table 2. Demographic characteristics of clinical participants
  i.v. Oral
Number of subjects completed as planned 55
Number of subjects included in safety analysis 55
Number of subjects included in PK analysis 55
Age (years) Mean, SD (range) 34.8, 7.19 (25–44)26.6, 4.45 (20–32)
Males (n) 55
Weight (kg) Mean, SD 81.6, 10.376.5, 9.6
BMI (kg m−2) Mean, SD 26.4, 2.224.0, 1.5
 White – White/Caucasian/European heritage:54
 Mixed race01
Figure 2.

Mean plasma GSK269984A concentration-time profiles following single oral (n= 5) or i.v. (n= 5) microdoses of GSK269984A (100 µg, free acid). inline image100 µg i.v. infusion (1 h); inline image100 µg oral

Table 3. Summary of human GSK269984A pharmacokinetic parameters following single oral and intravenous doses (100 µg, free acid) of GSK269984A (n= 5 per dosing regimen)
Parameter Regimen Geometric mean %CVb
  1. *Absolute bioavailability estimate based on geometric mean of AUC(0,∞) from each dose. †Between subject coefficient of variation.

AUC(0,∞) (ng ml−1 h) i.v.10.2022.9
C max (ng ml−1) i.v.3.228.7
t 1/2 (h) i.v.8.1526.5
CL (l h−1) i.v.9.8122.9
V ss (l) i.v.62.7823.8
F (%)* Oral94.9NA


For both the oral and the i.v. doses, GSK269984A free acid concentrations remained quantifiable in the plasma for the 24 h post dose period. Following the single i.v. dose systemic concentrations of GSK269984A free acid were detectable at 5 min after the start of the infusion. Plasma concentrations increased rapidly and reached a peak (3.2 ng ml−1, circa 7.9 nm) at the end of the 1 h i.v. infusion. The elimination phase was described by a mono-exponential function with a half-life of 8.2 h. GSK269984A was found to have a moderate steady-state volume of distribution (62.8 l, circa 0.8 l kg−1) and a relatively low clearance (9.8 l h−1). Following the single oral dose, GSK269984A free acid was detectable in the plasma after 5 min and peak concentrations were seen between 0.5 and 1.5 h post dose, thereafter decreasing in a mono-exponential fashion with an elimination half-life of 9.9 h. Absolute oral bioavailability was estimated to be in excess of 94%.

Safety and tolerability

GSK269984A in a dose equivalent to 100 µg of free acid was found to be well tolerated when administered by either the oral or i.v. route. Four AEs were reported and all were mild. The most frequently reported AE was headache, reported by three of the 10 participants. One subject reported a skin reaction to an ECG electrode. There were no changes in the safety laboratory values, vital signs or ECG recordings that were considered to be clinically important by the investigator.


As part of a wider strategy to address the current high rate of attrition for new chemical entities that proceed to clinical development, both the European Agency for the Evaluation of Medicinal Products (EMEA) and the US Food and Drug Administration (FDA) have issued guidance documents to support the use of exploratory (Phase 0) studies in the early evaluation of the PK and PD properties of such compounds in humans [25, 26, 28]. In this regard, an emerging literature reveals how microdose studies using highly sensitive analytical methods, namely LC/MS/MS and accelerator MS (AMS) can be used to extrapolate the PK profile of compounds and metabolites (assuming linearity of exposure with dose) to higher clinically relevant doses [22, 27, 29–31]. However, few microdose investigations have been reported specifically in the context of drug development and their value to the overall process has yet to be fully established [20, 22, 23]. Madan and colleagues have reported the first example of the use of the microdosing technique in man to enable ‘go/no-go’ decision making for compound selection [32]. In that study, an evaluation of the PK profile for four novel histamine H1-receptor antagonists and one reference compound in a two period crossover study (oral vs. i.v. single microdoses) was used to advance the compound with the most favourable profile. More recently, microdosing data were used to support the decision to terminate progression of the HIV integrase inhibitor PF-4776548 [24]. Here, we have shown how the incorporation of a microdose study into the planned clinical development of a novel EP1 receptor antagonist GSK269984A, through the administration of a sub-pharmacological dose to a small number of subjects, provided an opportunity to address a critical developmental risk posed by allometric predictions of a sub-optimal human PK profile.

Following i.v. administration, GSK269984A was rapidly distributed with a moderate steady-state volume of distribution (circa 63 l, circa 0.8 l kg−1) and a relatively low clearance (circa 10 l h−1, circa 11% of liver blood flow). The decline in GSK269984A concentration could be described by a single exponential function and an elimination half-life of approximately 8–9 h. The PK profile for GSK269984A following oral administration was consistent with rapid absorption, Cmax being observed between 0.5 and 1.5 h post dose. Elimination following the oral dose was very similar to that observed following i.v. administration, with a mono-exponential decline in plasma concentration and an elimination half-life of 9–10 h. It is recognized, however, that the true elimination phase may not have been fully defined as the blood sampling regimen was limited to the 24 h post dose period. A further limitation is that the microdoses were not administered in a crossover paradigm, each subject receiving only one dose (oral or i.v.) of GSK269984A (in keeping with EU guidance), which thus precluded calculation of individual estimates of absolute bioavailability. Nevertheless, the observed values for systemic exposure after oral dosing approached those seen with the i.v. dose, suggestive of high oral bioavailability for GSK269984A. A comparison of mean AUC(0,∞) for both regimens, provides an estimated bioavailability in excess of 94%. In this study, GSK269984A was administered orally as a solution and most likely proved optimal for absorption of GSK269984A. Administration of solid dosage forms may not result in such rapid absorption of GSK269984A.

There was a notable lack of agreement between the observed human PK parameters for GSK269984A and the predictions based on simple allometric scaling of in vivo PK data from three pre-clinical species (rat, dog and monkey). Our data therefore raise well recognized issues relating to the choice of pre-clinical species for undertaking human PK predictions as well as the most appropriate allometric method for interspecies scaling [33–37]. The tendency for predictions of human CL to be over estimated by allometric scaling has been noted previously for several established drugs [34] and more recently for new chemical entities in microdose investigations [32]. In the present study, estimates for human CL were markedly overestimated by the allometric predictions (almost by a factor of 10), suggesting fundamental species differences in the disposition and elimination of GSK269984A.

Based on an analysis of PK data for 103 non-peptide compounds administered intravenously to the rat, dog, monkey and human, Ward & Smith have indicated that it is the likelihood of accurately predicting CL (rather than distributional volume, Vd) that should be the main consideration upon which to select the most appropriate preclinical species for the purposes of human PK projections [38, 39]. Their analysis indicated that the most accurate approach for human CL projections would be based on simple allometric scaling of PK data from the monkey, in comparison with extrapolation of PK from rat or dog alone, or scaling from any combination of pre-clinical species [38–40]. For that series of compounds, it has been noted that the use of monkey data for human CL prediction carries an overall success rate of around 70% and that further improvements in the predictive success rate might be achieved through the integration of calculated molecular properties for a given compound into the extrapolative model [39, 41]. A combination of molecular property descriptors as part of the extrapolation exercise has been shown to improve the predictability of human PK, not only in terms of the overall prediction, but also for aiding the selection of the most appropriate preclinical species to use [41]. Of the 103 compounds studied in the analysis approximately 30 ‘acid-like’ compounds containing a –COOH structural motif were identified. A retrospective analysis (data not shown), comparing measured human CLb with measured preclinical CLb, showed a poor correlation with both dog and rat (r2 values of 0.188 and 0.120, respectively), whilst a much improved correlation with primate (r2= 0.735), suggesting that primate would be a more reliable species for predicting the pharmacokinetic properties for acid-like compounds in man.

However, in this instance, it is unlikely that GSK269984A would have been selected as a clinical candidate based on an extrapolation of data from the monkey alone. Interestingly, it was predictions based on the PK data for GSK269984A achieved in the rat alone that appear to support the best in vivo pre-clinical approximation in terms of the directional magnitude of the human PK parameters (low CL, moderate Vss, moderate t1/2, high Fp.o.) (Tables 1 and 3). In order to test further for the consistency of PK across the species, a plot of plasma concentration (normalized for dose) against ‘pharmacokinetic time’ (converted from chronological time) to include the observed individual data for all species, according to the principles of Dedrick [42], further illustrates the proximity of the rat PK profile to the human (Figure 3). For GSK269984A PK at least, man would appear to be more like a rat than a monkey. Following the availability of human microdose data, [14C]-GSK269984A was synthesized to investigate the metabolism and clearance of the compound using a variety of in vitro and in vivo systems across several species, enabling further comparison of rat with human. Consistent with the earlier in vitro data, subsequent oral excretion balance studies and isolated perfused liver studies in the rat with [14C]-GSK269984A confirmed that acyl glucuronidation was the predominant route of metabolism. The conjugates accounted for circa 41% of the administered dose in vivo, with unchanged GSK269984A in bile representing circa 15% of the dose. The pattern of drug-related material in faeces from intact rats likely reflects biliary secretion of parent compound and metabolites (with subsequent breakdown of any glucuronide conjugates by gut microflora), together with the presence of unabsorbed material and parent compound possibly resulting from secretion across the gut wall. For the rat, the major route of elimination of radiolabelled drug-related material was via the faeces (circa 97% of the dose) with minor urinary elimination (circa 1% of the dose, data not shown).

Figure 3.

Complex Dedrick plot for normalization of interspecies exposure to GSK269984A reveals the closer allometric proximity of humans to rats, when compared with dogs or monkeys. Data are for individual animals. inline image, Rat 1; inline image, Rat 2; inline image, Rat 3; inline image, Human 1; inline image, Human 2; inline image, Human 3; inline image, Human 4; inline image, Human 5; inline image, Monkey 1; inline image, Monkey 2; inline image, Monkey 3; inline image, Dog 1; inline image, Dog 2; inline image, Dog 3

In contrast to the discordant in vivo PK data observed across the three pre-clinical species tested, it was the in vitro metabolic data for GSK269984A that provided both cross-species agreement on a low value for CLi, as well as consistency with the relatively low human CL observed following microdose administration. Many studies have shown how in vivo CL in the rat or dog can be predicted successfully from the in vitro data for the same species [34]. A similar approach can be undertaken for human CL predictions, although for certain classes of compound there is a lack of correlation between the CLi value derived in vitro using human microsomes or hepatocytes and that calculated in vivo[43]. Ostensibly, the present data would lend support to other studies that have shown how the prediction of human CL in vivo can be improved by the appropriate incorporation of in vitro data. For example, predictions of human CL based on allometric scaling from pre-clinical in vivo data can be improved by first normalizing in vivo CL values using the respective in vitro CLi values [44], while other investigators have proposed the initial correction of in vitro human CLi using a scaling factor derived from animal studies [45, 46]. However, in this current study, the values obtained for GSK269984A CLi in microsomes (rat, dog, primate and human), S9 fractions (rat, dog, primate and human) and hepatocytes (rat, primate and human) were essentially equal across all species [16]. A further consideration relates to the level of uncertainty introduced into the CL predictions with respect to incorporated estimations of the free fraction (fu). This may be particularly relevant for acidic compounds with appreciable plasma protein binding, such as GSK269984A, for which an approximate 3-fold variation in fu was observed across the species.

The present investigation revealed a favourable human PK profile for GSK269984A that was consistent with a clinically acceptable oral dosing regimen and which provided a rational basis for progression of the compound into formal clinical development. In this case, the microdose strategy enabled the key liability surrounding PK predictions to be discharged at an earlier developmental stage, with reduced cost and resource allocation than would otherwise have been required to initiate a conventional first-into-human (FIH) study. Other recognized advantages of the microdose study included the minimal requirements for drug substance (around 50 g), reduced animal use and toxicology requirements in keeping with current regulatory guidance for phase 0 investigations [25], as well as reduced safety risk in humans owing to the low dose of administered GSK269984A. The estimated all inclusive pre-clinical costs (chemical development, safety assessment and DMPK activities) to the point of microdosing with GSK269984A amounted to <£50k. While the pre-clinical time frame was marginally shorter than that required to support equivalent activities for a conventional FIH study, the cost for the latter (now also including 28 day toxicology studies in two species) could have been expected to exceed >£400k. Furthermore, additional cost saving was realized by the development of a highly sensitive analytical assay based on HPLC/MS/MS methodology, which avoided the need to synthesize a radiolabelled compound for quantitative measurement using accelerator mass spectrometry (AMS). Other recent studies have similarly demonstrated the convenience and sensitivity of LC/MS/MS methodology for the quantitative analysis of low molecular weight compounds following microdose administration [30, 31, 47, 48]. Another advantage of using LC/MS/MS over AMS was the ability to have rapid turnaround (within a few days of the microdose sessions) for analysis of plasma samples and the resulting data. It should be noted that the present study was designed to enable sequential i.v. and oral dosing using two groups of subjects. As such, the approach would have enabled early termination of the study in the event of a sub-optimal PK profile being observed following the i.v. dose, thereby avoiding any further unnecessary human exposure.

Although administration of GSK269984A microdoses to humans has offered specific reassurance in terms of an acceptable PK profile for humans, potential limitations to the approach should also be appreciated. From a safety perspective, and with respect to the existing regulatory requirements for the non-clinical studies required to support safe human microdose administration, a truncated package of toxicology data may not be similarly reflected in the subsequent safety assessments required for conventional FIH studies. Indeed, for GSK269984A, subsequent 28 day repeat dose toxicology studies in the dog revealed unexpected CNS histopathology that precluded further clinical evaluation of the compound at higher exposures and for longer dosing duration. This outcome could not have been foreseen on the basis of the limited pre-clinical data available at the time of the microdose study. A second consideration is that there may be a lack of agreement between the PK profile seen following microdose administration and that seen at higher doses, which may reflect non-linearity in absorption, distribution, metabolism and/or excretion over the dose range [20, 22]. To date, a formal comparison of microdose with therapeutic dose PK data has only been undertaken in the clinical setting for a few established drugs of limited chemical diversity [27, 29, 49], so the predictive value of the methodology has yet to be fully defined. It also remains to be seen whether the scope of application of microdose data, when combined with human in vitro data, might be usefully broadened to update the assumptions underpinning physiologically-based PK (PBPK) models and thereby enhancing their predictive value [50]. A final consideration relates to the economics of integrating a microdose study into early clinical development when compared with a conventional plan that incorporates a traditional FIH trial alone. On the one hand, assuming an eventual successful outcome, the total cost of drug development in terms of both time and resource allocation to regulatory approval will necessarily be greater. Indeed, in the case of GSK269984A, having made the decision to progress to a microdose study, all activities to support a default FIH study were intentionally placed on hold pending the availability of human PK data. Following the successful microdose readout, there was then a requirement for additional compound synthesis and toxicology resource to backfill studies in support of a conventional phase 1 programme, such that the overall time to achieve phase II completion would have been longer than for a default plan. On the other hand, for an unsuccessful outcome, a cost-saving can be realized since the decision to terminate further compound development based on microdose data can be achieved at an earlier time point with fewer resources as described by Harrison et al. [24]. Nevertheless, the question of whether the inclusion of a microdose study might present a universally cost-efficient approach for a given drug development programme irrespective of the outcome remains a complex one. In our view, the decision will be a function of the risk analysis, the cost, the availability and timing of resource allocation, as well as the development strategy and the desired outcomes for enabling compound progression. In terms of timing, Harrison and colleagues make a compelling argument, based on their case studies of PF-184298 and PF-4776548, for the early acceptance of PK prediction uncertainty (where it exists) and the immediate diversion of resources into an exploratory human PK study [24]. Such an approach, it is argued, may be preferable to the pursuit of costly and lengthy preclinical investigations in an attempt to gain further understanding of discordance in PK predictions, and where the impact of such activities on the success of additional drug optimization is likely to remain uncertain. Our experience with the GSK269984A programme would be supportive of such an approach. An appreciable level of in vitro metabolic stability had not been a common feature across all the EP1 antagonists represented in the chemical series [16]. Furthermore, a recognized inconsistency with respect to in vivo PK for GSK269984A across three species, along with mismatch between in vitro and in vivo data, presented a threshold level of uncertainty that was sufficient to trigger the decision (in the discovery phase) to undertake an exploratory human PK study.

In conclusion, the present data, acquired following oral and i.v. microdose administration of GSK269984A to humans, is indicative of a PK profile that can be described in terms of rapid absorption, high oral bioavailability, moderate Vss and low CL. The profile, although consistent with a clinically acceptable dosing regimen, would require confirmation at higher doses. These microdose data helped to address a risk of failure that was unquantifiable on the basis of the observed in vivo PK profile across three pre-clinical species. Our study therefore supports the early application of a microdose strategy, perhaps during the lead optimization phase and prior to formal selection of the clinical candidate, as a means to facilitate decision making and to discharge risk. This may be particularly relevant for highly protein bound acidic compounds and for drug development programmes that encounter lack of consistency between in vitro and in vivo pre-clinical data, with ensuing uncertainty around the human PK predictions.

Competing Interests

There are no competing interests to declare.


The authors acknowledge the contributions of other members of the GSK269984A project team [Matt Davies, Tina Panchall, Stephanie North, Mike Briggs and Paul Goldsmith (Drug Metabolism & Pharmacokinetics), Colin Fish, Paul Giffen, Steve Polley, Nicola Robertson, Louise Rutherford and Elizabeth Soames (Safety Assessment), Adrian Hall (Medicinal Chemistry), Andy Billinton (Biology), Maria Davy (Discovery Medicine) and Jeffrey Price (Clinical Operations)] and the clinical investigation team [Malcolm Handel (Investigator) and staff at the James Lance GSK Medicines Research Unit, Sydney, Australia]. The study was sponsored by the Neurology Centre of Excellence for Drug Discovery, GlaxoSmithKline R&D, Harlow, UK and the authors are employees of GSK R&D.