Introduction

Top Abstract Introduction Experimental Procedures Results Discussion References

Cyclooxygenases (COXs) catalyze the conversion of arachidonic acid to prostaglandin H₂, which serves as the common precursor for the synthesis of prostaglandins, prostacyclins, and thromboxanes. Cyclooxygenases exist as two isoforms, which exhibit similar catalytic properties but differ in terms of regulation of expression (Jouzeau et al., 1997; Garavito and DeWitt, 1999; Hawkey, 1999). COX-1 is the major isoform expressed in healthy tissue and generates prostanoids for different functions such as gastric cytoprotection, maintenance of renal homeostasis, and platelet aggregation. The second isoform, COX-2, is also present at a basal level in certain tissues but its expression is up-regulated in response to inflammatory or mitogenic stimuli. The discovery of this second isoform of cyclooxygenase has provided the rationale for development of novel anti-inflammatory agents with improved gastrointestinal tolerability compared with nonselective NSAIDs, which inhibit both COX-1 and COX-2 (Jouzeau et al., 1997; Garavito and DeWitt, 1999; Hawkey, 1999). The efficacy of agents that selectively inhibit COX-2 in the treatment of the symptoms of osteoarthritis and the relief of acute pain, and their lower incidence of gastrointestinal-related adverse events compared with nonselective NSAIDs, have been demonstrated in several clinical studies (Ehrich et al., 1999a,b; Langman et al., 1999; Simon et al., 1999).

The difficulty of evaluating the selectivity of COX-2 inhibitors based on in vitro enzyme and cell-based assays is well recognized and has been extensively discussed (Jouzeau et al., 1997; Pairet and van Ryn, 1998; Brooks et al., 1999). Whole blood assays, based on the production of LPS-induced PGE₂ for COX-2 and on the synthesis of TXB₂ following clotting of the blood for COX-1, are considered to provide the most meaningful index of selectivity (Patrignani et al., 1994; Brideau et al., 1996; Brooks et al., 1999). In these assays, the selectivity of cyclooxygenase inhibition is measured in a physiological milieu taking into account the binding of the drugs to plasma proteins. These procedures have been adapted for ex vivo assays and have shown that rofecoxib and celecoxib have no significant effects on COX-1 activity at recommended therapeutic doses, providing the basis for specific COX-2 inhibition in humans (Brooks et al., 1999). Celecoxib has been reported to inhibit COX-1-mediated TXB₂ production when administered in excess of the therapeutic doses (McAdam et al., 1999).

Although selective COX-2 inhibitors do not affect COX-1 in most cell and enzyme assay systems, they can inhibit COX-1 under certain in vitro conditions. The inhibitory effects on COX-1 activity are strongly dependent on the concentration of free arachidonic acid that can compete with inhibitor binding at the active site. Assays at low substrate concentrations thus offer particularly sensitive prescreening assays to detect inhibitory effects on COX-1 (Riendeau et al., 1997a). Furthermore, compounds with low potencies in such assays will also show a reduced potential for the inhibition of COX-1 in tissues where the availability of arachidonic acid might be limiting. Although there is no evidence from clinical studies that agents that selectively inhibit COX-2 do affect COX-1, the possibility that COX-1 inhibition could occur in certain tissues, such as the kidney or in others under certain pathological conditions, cannot be totally excluded.

In the present study, we describe the properties of etoricoxib (MK-0663, L-791,456), a novel dipyridinyl inhibitor (Fig. 1) that has been developed for high in vitro selectivity of COX-2 inhibition using a combination of whole blood and sensitive COX-1 inhibition assays. The preclinical pharmacological profile of etoricoxib is reported, together with comparative data on the selectivity of other agents that have been reported to selectively inhibit COX-2.

View larger version (16K): [in this window] [in a new window]   Fig. 1.   Structure of etoricoxib (5-chloro-2-(6-methyl pyridin-3-yl)-3-(4-methylsulfonylphenyl) pyridine).

	Experimental Procedures

Top Abstract Introduction Experimental Procedures Results Discussion References

Assays with Purified Recombinant Human COX-1 and COX-2. Recombinant human COX-1 and COX-2 were expressed in baculovirus-Sf9 cells and enzymes were purified as previously described (Percival et al., 1994; Cromlish and Kennedy, 1996). Enzymatic activity of the purified COX-2 was measured using a chromogenic assay based on the oxidation of N,N,N',N'-tetramethyl-p-phenylenediamine (TMPD) during the reduction of PGG₂ to PGH₂ (Copeland et al., 1994) with minor modifications (Chan et al., 1999). This assay includes a 15-min preincubation of the enzyme with inhibitor before the initiation of the reaction with 100 µM arachidonic acid and was performed in the presence or absence of the detergent Genapol X-100 (2 mM). Assays for the assessment of inhibition of purified COX-1 at 0.1 µM arachidonic acid substrate concentration, the determination of the stoichiometry of the COX-2-inhibitor complex, the rate constant of dissociation of the enzyme-inhibitor complex by recovery of enzymatic activity, and the recovery of intact inhibitor from that complex were all performed as previously described (Riendeau et al., 1997b). HPLC analysis of etoricoxib was performed on a Novapak C18 column (3.9 × 150 mm) (Waters, Milford, MA), using acetonitrile/water/acetic acid (50:50:0.1) as solvent, a flow rate of 2 ml/min, and monitoring at 275 nm (retention time = 6.1 min).

Mechanism of Inhibition of COX-2. The time-dependent inhibition of COX-2 was analyzed using the model developed for the inhibition of ovine cyclooxygenase (Rome and Lands, 1975). In this model (Scheme 1), an initial reversible binding of enzyme and inhibitor (characterized by the dissociation constant K_i), is followed by a conversion to a tight enzyme-inhibitor complex (characterized by a first order rate constant, k_on). The rate of reversal of this process (k_off) is considered to be negligible.
This model predicts that the observed rate constant (k_obs) for the exponential loss of cyclooxygenase activity during the preincubation of enzyme and inhibitor is given by the following equation:
The remaining enzyme activity after a range of preincubation periods with purified COX-2 was determined for a series of inhibitor concentrations (1-300 µM), and was computer fitted to a first order equation to give a series of first order rate constants (k_obs) for the onset of inhibition. The values of K_i and k_on were obtained by computer fitting the values of k_obs at each inhibitor concentration to the above equation. The k_off value was estimated from the recovery of activity following dialysis of the COX-2-etoricoxib complex. The overall dissociation constant (K_i*) for the purified COX-2-etoricoxib complex (EI*) was calculated using the kinetic constants determined above for the equilibrium defined in Scheme 1. 

Measurement of Dissociation Constants for Human COX-1. Human COX-1 was solubilized as previously described (Percival et al., 1994) except that phenylmethylsulfonyl fluoride and homovanillic acid were replaced by 50 µg/ml Pefabloc (Roche Molecular Biochemicals, Laval, Quebec, Canada) and 1 mM phenol, respectively, in the homogenization buffer. The enzyme was solubilized in homogenization buffer containing 1.5% decyl maltoside and was partially purified using a UNO Q1 anion exchange column (Bio-Rad, Richmond, CA) using a gradient of 0 to 400 mM KCl. The fractions were assayed using the TMPD assay and the active fractions were pooled and concentrated using a centriprep-30 (Amicon, Beverly, MA).

Whole Cell Assays with Transfected Chinese Hamster Ovary (CHO) Cells Expressing COX-1 and COX-2. Stably transfected CHO cells expressing human COX-1 and COX-2 were cultured and assayed for the production of PGE₂ following stimulation by arachidonic acid as previously described (Kargman et al., 1996). The cells were preincubated in Hanks' balanced salts solution containing 15 mM HEPES, pH 7.4, with the test drug or DMSO vehicle for 15 min at 37°C before a 15 min challenge with arachidonic acid (10 µM arachidonic acid for COX-2 and of 0.5 µM arachidonic acid for COX-1). Compounds were typically tested at 8 concentrations in duplicate using 3-fold serial dilutions in DMSO. Cyclooxygenase activity in the absence of test compounds is determined as the difference in PGE₂ levels of cells challenged with arachidonic acid versus the PGE₂ levels in cells mock-challenged with ethanol vehicle.

Assay of Microsomal COX-1 at Low Arachidonic Acid Concentration. The assay of human COX-1 at low arachidonic acid concentration was performed using a microsomal preparation from U937 cells (Riendeau et al., 1997a). This assay system has been found to be more sensitive to inhibition than other COX-1 enzyme or cell assays and has been used to compare the relative potency of nonselective NSAIDs and agents that inhibit COX-2. Inhibitors were preincubated with the microsomal preparation for 15 min at room temperature in 0.1 M Tris-HCl, pH 7.4, 10 mM EDTA, 0.5 mM phenol, 1 mM reduced glutathione and 1 µM hematin. Arachidonic acid was added to a final concentration of 0.1 µM and the samples were further incubated for 40 min before quantitation of PGE₂ by radioimmunoassay. Cyclooxygenase activity is defined as the difference in PGE₂ levels of microsomes incubated with arachidonic acid versus the PGE₂ levels in microsomes incubated with ethanol vehicle.

TXB₂ Production by Calcium Ionophore-Activated Human Platelets. Human washed platelets in Hanks' balanced salts solution buffered with 15 mM HEPES, pH 7.4, were preincubated at a final concentration of 4 × 10⁷ cells/ml in the absence or presence of etoricoxib (from a 125-fold concentrated solution in DMSO). After 10 min, platelets were stimulated with 2 µM calcium ionophore A23187 for TXB₂ production (Riendeau et al., 1997b).

Human Whole Blood Assays for COX-2 and COX-1. The assays were done using identical procedures as reported previously (Brideau et al., 1996). Fresh human blood was collected from volunteers that had no apparent inflammatory conditions and had not taken and NSAID for at least 7 days before blood collection. Briefly, for the COX-2 assay, fresh heparinized whole blood from male volunteers (500 µl) was incubated with LPS at 100 µg/ml and with 2 µl vehicle (DMSO) or a test compound for 24 h at 37°C. Inhibitors were typically tested at five different concentrations using 3-fold serial dilutions of the highest tested concentration. PGE₂ levels in the plasma were measured after deproteination using derivatization to the methyl oxime and radioimmunoassay (Amersham, Oakville, Ontario, Canada). COX-2 activity is defined as the production of PGE₂ in the vehicle-treated and LPS-treated blood over that of background levels in unstimulated blood at time zero. For the COX-1 assay, an aliquot of fresh blood from male or female volunteers collected into vacutainers containing no anticoagulants was mixed with either DMSO or a test compound and was allowed to clot for 1 h at 37°C. TXB₂ levels in the serum were measured using an enzyme immunoassay (Cayman Chemicals, Ann Arbor, MI) after deproteination. IC₅₀ values were derived using a four-parameter fit of the titration with the software program SigmaPlot.

In Vivo Assays. All procedures used in the in vivo assays were approved by the Animal Care Committees or Institutional Animal Care and Use Committee at the Merck Frosst Centre for Therapeutic Research (Kirkland, Quebec, Canada); Merck, Sharp & Dohme Neuroscience Research Centre (Harlow, UK); Merck Research Laboratories (Rahway, NJ), and were performed according to guidelines established by the Canadian Council on Animal Care, the British Home Office, and the U.S. Department of Agriculture and National Institutes of Health, respectively. Etoricoxib was given orally as a suspension in 0.5% (adjuvant arthritis and hyperalgesia assay) or 1% (other assays) methocellulose in water. The suspensions were used at volumes of 10 ml/kg in rats and 1 ml/kg in squirrel monkeys.

Acute Models of Efficacy. The efficacy of etoricoxib and other compounds was evaluated in rat models, including carrageenan-induced paw edema assay, carrageenan-induced paw hyperalgesia assay, and endotoxin-induced pyresis using previously described procedures (Chan et al., 1995). A nonhuman primate model of endotoxin-induced pyresis using telemetric body temperature measurement was also used (Chan et al., 1997). Test compounds were given 2 h after injection of endotoxin in conscious, unrestrained squirrel monkeys. The reversal of increases in body temperature was measured 2 h after administration of tested compounds.

Chronic Model of Efficacy. The efficacy of etoricoxib in chronic inflammatory conditions was evaluated in adjuvant-induced arthritis in rats as previously described (Fletcher et al., 1998). Test compounds were given orally twice daily starting on day 0 and continued until euthanasia 21 days later (Chan et al., 1999). In this study, body weight, paw volume, thymus, and spleen weights were measured. In addition, radiographs were taken from adjuvant injected and noninjected hind paws on day 0 and day 21.

Models of Gastrointestinal Integrity. The effect of etoricoxib on the integrity of the gastrointestinal mucosa was tested using a ⁵¹Cr-EDTA urinary excretion assay in rats. This assay has been used in rats (Davies et al., 1994) and in humans (Bjarnason et al., 1986) and was found to be more sensitive to detect effects of NSAIDs on gastrointestinal integrity compared with the ⁵¹Cr-fecal excretion assay used previously (Chan et al., 1995). ⁵¹Cr-EDTA is an inert compound that is not taken up by extravascular tissue following its absorption and that is excreted completely via the kidney. Thus, urinary excretion of ⁵¹Cr can be used as an index for gastrointestinal permeability. Male Sprague-Dawley rats (~300-360 g) were dosed orally with a test compound either once or twice daily for 3 to 10 days. Immediately after the administration of the last dose, the rats were given orally 2 ml of water containing 10 µCi of ⁵¹Cr-EDTA. The animals were placed individually in metabolism cages with food and water ad libitum. Urine was collected for a 24-h period and the amount of radioactivity was determined. Urinary ⁵¹Cr excretion was calculated as a percentage of the total administered dose. In the 10-day chronic studies with etoricoxib, an interim urinary ⁵¹Cr excretion readout was obtained after 5 days of dosing. In addition, the effect of etoricoxib on gastrointestinal integrity was also examined in nonhuman primates in a model where ⁵¹Cr fecal excretion is measured following i.v. administration of ⁵¹CrCl₃ in saline (Chan et al., 1995).

Statistics. Results are expressed as means ± S.E. Unless otherwise specified, differences between vehicle control and treatment groups were tested using one-way ANOVA followed by multiple comparison by the Dunnett's test. A p value of <0.05 was considered statistically significant. Dose-response curves for percentage of inhibition were fitted by a four-parameter logistic function using a nonlinear least-squares regression. IC₅₀ or ID₅₀ was derived by interpolation from the fitted four-parameter equation.

Materials. The following compounds were synthesized by the Medicinal Chemistry Department at Merck Frosst Centre for Therapeutic Research: etoricoxib (Friesen et al., 1998), rofecoxib (Prasit et al., 1999), celecoxib (Penning et al., 1997), meloxicam (Engelhardt et al., 1996), JTE-522 (Wakitani et al., 1998), valdecoxib (Talley et al., 2000), and 6-methoxy-2-naphthaleneacetic acid. Sources of other compounds were diclofenac, nimesulide, etodolac, ibuprofen, nabumetone, lipopolysaccharide (Sigma-Aldrich, Oakville, Ontario, Canada); indomethacin (Merck, Sharp & Dohme Canada, Kirkland, Quebec, Canada), and M. Butyricum (Difco, Detroit, MI).

	Results

Top Abstract Introduction Experimental Procedures Results Discussion References

Selective Inhibition of COX-2 by Etoricoxib in Whole Cell, Enzyme, and Whole Blood Assays. The IC₅₀ values for the inhibition of COX-1 and COX-2 by etoricoxib in various in vitro assays are summarized in Table 1, together with those of indomethacin as a reference for a nonselective inhibitor. Etoricoxib was a potent inhibitor of the production of PGE₂ in CHO cells expressing human COX-2 (IC₅₀ = 79 ± 12 nM) with a high selectivity compared with its effects on COX-1 (IC₅₀ > 50 µM in CHO cells). In addition, etoricoxib did not affect the synthesis of TXB₂ mediated by COX-1 in calcium ionophore-stimulated human platelets (IC₅₀ > 100 µM). Etoricoxib inhibited the cyclooxygenase activity of purified human COX-2 (spectrophotometric assay), with IC₅₀ values of 5.0 ± 1.1 µM (n = 8) and 4.1 ± 0.6 µM (n = 8) for the assays performed either in the absence or presence of the detergent Genapol X-100, respectively.

Potency and Selectivity of Inhibitors in Whole Blood Assays. The results are summarized in Table 2. Among the compounds showing more than a 5-fold selectivity in whole blood assays, all showed IC₅₀ values against COX-2 within a 2-fold range (0.5-1.1 µM). Using the ratio of COX-1 IC₅₀/COX-2 IC₅₀, selectivity ratios for the inhibition of COX-2 of 106, 35, 30, 7.6, and 7.3 were obtained for etoricoxib, rofecoxib, valdecoxib, celecoxib, and nimesulide, respectively. Lower selectivity ratios were observed for diclofenac, etodolac, and meloxicam (2- to 3-fold). JTE-522, nabumetone, and 6-methoxy-2-naphthaleneacetic acid (the active metabolite of nabumetone) were not potent inhibitors of either COX-2 or COX-1 in whole blood assays (Table 2).

Effects of Inhibitors on COX-1 Activity. Under certain conditions, selective COX-2 inhibitors show weak but detectable effects on COX-1. The production of PGE₂ by U937 microsomes incubated with a low concentration of arachidonic acid (0.1 µM) represents the most sensitive assay to detect inhibitory effects on COX-1 (Riendeau et al., 1997a). The data for this assay are summarized in Table 3 for etoricoxib and other inhibitors that have shown at least a 2-fold selectivity for the inhibition of COX-2 in the whole blood assay. On average, the inhibitors were found to be 10 to 100 times more potent in the COX-1 microsomal assay compared with the COX-1 whole blood assay. The U937 microsomal assay indicates significant differences in the ability of COX-2 selective agents to inhibit COX-1 under these assay conditions (Fig. 2) with IC₅₀ values of 12, 2, 0.25, and 0.05 µM for etoricoxib, rofecoxib, valdecoxib, and celecoxib, respectively. To further confirm the differences detected by the U937 COX-1 assay, K_i values were estimated with human recombinant COX-1 using an assay based on the ability of the inhibitors to compete with the time-dependent inhibition of indomethacin. Figure 3 shows representative data for the effects of inhibitors on the time-dependent inhibition by 2 µM indomethacin, using a fixed concentration of 10 µM of each inhibitor. Celecoxib and valdecoxib showed direct inhibitory effects at time zero and also decreased the inhibitory effects of indomethacin, whereas etoricoxib and rofecoxib had little or no effect. Using additional data at different inhibitor concentrations, K_i values for COX-1 (Table 4) of 167, 18, 1.6, and 0.3 µM for etoricoxib, rofecoxib, valdecoxib, and celecoxib, respectively, were obtained. This rank order of the K_i values is completely consistent with the inhibitory data of the microsomal assay and further confirms the lowest ability of etoricoxib to interact with COX-1 among this group of COX-2 selective agents.

View larger version (21K): [in this window] [in a new window]   Fig. 2.   Inhibitory effects of COX-2-selective agents in the sensitive COX-1 microsomal assay at low arachidonic acid concentration.

View larger version (20K): [in this window] [in a new window]   Fig. 3.   Effect of etoricoxib and other inhibitors on the time-dependent inhibition of human COX-1 by indomethacin. COX-1 was preincubated with 2 µM indomethacin in the absence (control) or in the presence of the indicated inhibitors at a final concentration of 10 µM for different periods of time before the measurement of remaining COX-1 activity.

Mechanism of Inhibition of COX-2 by Etoricoxib. The kinetic mechanism of inhibition of COX-1 and COX-2 by etoricoxib was investigated using purified recombinant human enzymes. The inhibition of COX-2 by etoricoxib was time-dependent and slowly reversible. Analysis of the time-dependent inhibition data gave a value of 67 ± 22 µM for the initial reversible binding constant K_i, and a value of 0.015 ± 0.002 s¹ for the rate constant of isomerization to the tightly bound enzyme-inhibitor complex k_on (n = 3; data not shown). The second order rate constant for the onset of inhibition (k_on/K_i) of COX-2 by etoricoxib was 2.2 ± 1.0 × 10⁴ µM¹ s¹, compared with a value of 3.6 ± 1.0 × 10³ µM¹ s¹ for rofecoxib.

1

Effect of Etoricoxib in Models of Inflammation, Hyperalgesia, and Pyresis. As summarized in Table 5, etoricoxib was effective in models of carrageenan-induced paw edema, carrageenan-induced paw hyperalgesia, and endotoxin-induced pyresis in rats with a potency (0.3-0.9 mg/kg) comparable to that of rofecoxib and indomethacin. It should be noted that dose-dependent inhibition was observed with etoricoxib in these models at a dosing range of 0.1 to 30 mg/kg. In the rat hyperalgesia study, a complete reversal of hyperalgesia response was observed with etoricoxib at doses of 10 mg/kg or above (data not shown).

View larger version (95K): [in this window] [in a new window]   Fig. 4.   Effect of etoricoxib in the rat adjuvant arthritis model. a, normal tarsus, control rat. T, tarsus; c, calcaneus; l, talus; m, metatarsus; arrows, cuboidal bones (central and third tarsal bones). Note well defined joint spaces, sharp cortical and trabecular bony margins, and smooth periosteal borders. b, vehicle, tarsus with severe arthrosis. There is marked periarticular soft tissue swelling (broad white arrows). The tarsocrural joint is collapsed (black arrows) and the intertarsal articulations are narrow. There are multiple sites of cortical and trabecular osteolysis (long white arrow). Fluffy periosteal reaction can be seen along the cortices of the tibia (wavy white arrow) and tarsal bones. c, indomethacin at 1 mg/kg, tarsus with moderate arthrosis. There is moderate soft tissue swelling joint space narrowing. The osteolysis and periosteal reaction are less severe than in b. d, etoricoxib at 3 mg/kg/day (administered b.i.d.), tarsus with mild arthrosis. There is mild periarticular soft tissue swelling and slight trabecular bone loss in the cuboidal bones (compare with b and c).

Gastrointestinal Tolerability of Etoricoxib. Etoricoxib was tested in a model of urinary excretion of ⁵¹Cr following oral administration of ⁵¹Cr-EDTA in rats to evaluate its effects on the integrity of the gastrointestinal mucosa. In this model, acute oral dosing of indomethacin (3 mg/kg) or diclofenac (3 mg/kg) resulted in a significant increase in urinary ⁵¹Cr excretion in rats. The total urinary ⁵¹Cr excretion in the vehicle-, indomethacin-, and diclofenac-treated groups was 1.3 ± 0.1% (n = 11), 2.6 ± 0.1% (n = 5, p < 0.001 versus control), and 6.2 ± 1.5% (n = 6, p < 0.005 versus control), respectively. In a 3 day multiple dosing study, oral administration of indomethacin or diclofenac b.i.d. caused a significant dose-dependent increase in urinary ⁵¹Cr excretion (historical controls, Fig. 5). In contrast, multiple oral dosing of etoricoxib at a dose of 200 mg/kg/day (administered with b.i.d. dosing) for up to 10 days had no effect on urinary ⁵¹Cr excretion (Fig. 5).

View larger version (27K): [in this window] [in a new window]   Fig. 5.   Effects of multiple dosing of etoricoxib, diclofenac, or indomethacin on gastrointestinal integrity in rats. Test compound was administered orally b.i.d., for 3, 5, or 10 days at the indicated doses, followed by oral administration of 51Cr-EDTA (10 µCi/rat). Urine was collected for 24 h and 51Cr excretion was calculated as a percentage of total dose (±S.E). *p < 0.05 versus vehicle control.

View larger version (35K): [in this window] [in a new window]   Fig. 6.   Effects of chronic dosing of etoricoxib on gastrointestinal integrity in squirrel monkeys. Etoricoxib (50 mg/kg b.i.d.), diclofenac (1 mg/kg b.i.d.), or naproxen (5 mg/kg b.i.d.) were administered to squirrel monkeys for 4 to 5 days. Fecal 51Cr (as a percentage of injected dose) was measured (±S.E.) in groups of three to six animals after treatment with vehicle, diclofenac, naproxen, or etoricoxib. *p < 0.05 versus vehicle control.

	Discussion

Top Abstract Introduction Experimental Procedures Results Discussion References

Etoricoxib is a novel dipyridinyl agent that selectively inhibits COX-2 and that shows an efficacy similar to traditional NSAIDs in various rodent models of inflammation, pain, and fever and also in a primate model of pyresis. Etoricoxib is more than 100-fold selective for COX-2 versus COX-1 in various cell and whole blood assays. In addition, no detectable effects were noted against human prostacyclin synthase (IC₅₀ > 100 µM; data not shown) or in a diverse array of receptor and enzyme assays performed by a contract laboratory (MDS Panlabs, Bothell, WA).

Various inhibitors that demonstrate selectivity for COX-2 have been shown to inhibit COX-2 and COX-1 by different mechanisms, apparently due to an Ile-Val difference at the active site (Gierse et al., 1996; Wong et al., 1997). The weak inhibition of COX-1 is of a simple competitive nature, whereas COX-2 shows an additional time-dependent component with formation of a tight enzyme-inhibitor complex (Ouellet and Percival, 1995; Gierse et al., 1999). Because of this difference, the selectivity observed in a particular assay is strongly dependent on the concentration of arachidonic acid used as substrate and on the preincubation time of the inhibitor with the enzyme before the initiation of the reaction. Thus, a large number of articles have reported different COX-1/COX-2 selectivity ratios for the same compounds using different test systems and assay conditions (Jouzeau et al., 1997; Pairet and van Ryn, 1998). Whole blood assays have the advantage of being performed under conditions where arachidonic acid is generated endogenously in the presence of the test compound. In this system, the potency of drugs is often significantly reduced, due to binding of the drug to plasma proteins and the presence of arachidonic acid at the time of COX-2 induction. This is in contrast to most other in vitro assays, which are typically performed with preincubation of the enzyme and drug in the absence of arachidonic acid, allowing formation of the tight inhibitor-COX-2 complex and resulting in a strong inhibition of COX-2 and high selectivity (Ouellet and Percival, 1995; Riendeau et al., 1997b; Pairet and van Ryn, 1998). For example, shifts in IC₅₀ values of 12-, 20-, and 1,000-fold were observed for etoricoxib, rofecoxib, and celecoxib in the whole blood assay compared with the values for the COX-2-transfected CHO cells. Similarly, valdecoxib has been reported to show a 28-fold selectivity for COX-2 in whole blood assays, in close agreement with the present results, but much lower than the 28,000-fold selectivity observed in enzyme assays (Talley et al., 2000). Meloxicam, etodolac, and nimesulide showed a 2- to 7-fold selectivity in COX-2 whole blood assays, which is considerably lower than in other assays (Riendeau et al., 1997b; Cullen et al., 1998; Brooks et al., 1999; Tustin, 1999), but in agreement with values reported for similar whole blood assays (Glaser et al., 1995; Panara et al., 1999; Warner et al., 1999).

It is of interest to compare the level of COX-2 selectivity as measured with the in vitro whole blood assays to clinical data, where selectivity has been assessed with ex vivo readouts. Meloxicam, which showed a 2-fold selectivity for COX-2 inhibition in the in vitro whole blood assay (Table 2), caused significant inhibitory effects of COX-1-mediated platelet TXB₂ production in volunteers ex vivo at clinically relevant doses, although to a lesser extent than diclofenac (Panara et al., 1999; Tegeder et al., 1999). Celecoxib and nimesulide showed about a 7-fold selectivity in the present whole blood assays, slightly higher than the ratios of 1.4 and 5.3, respectively, reported for these inhibitors by Warner et al. (1999), but lower than the 40-fold value reported for celecoxib by Talley et al. (2000). In humans, nimesulide had no effect on serum TXB₂ levels (Cullen et al., 1998), but caused a 50% reduction of platelet TXB₂ production in ex vivo assays after administration of a single dose of 100 mg (Panara et al., 1998). Modest, but significant effects on platelet TXB₂ production have been reported for celecoxib at a dose of 800 mg (McAdam et al., 1999), without effect on agonist-induced platelet aggregation or bleeding time at a dose 600 mg b.i.d. (Leese et al., 2000). Rofecoxib showed a 35-fold selectivity for COX-2 in the current assay and a 77-fold selectivity in the assay described by Warner et al. (1999). The higher selectivity detected in the latter assay is due to a lower potency against COX-1, which might be related to the use of calcium ionophore to stimulate thromboxane production. For rofecoxib, no effect on platelet function or TXB₂ production has been observed, even at supratherapeutic doses (up to a single dose of 1000 mg, 20- to 80-fold the recommended dose) (Ehrich et al., 1999a).

The clinical data summarized above indicate a clear trend toward the reduction of effects on COX-1-mediated platelet TXB₂ synthesis as the selectivity of the drug increases in the whole blood assays. A selectivity ratio of 6- to 7-fold for COX-2 appears to be sufficient to spare agonist-induced platelet functions, but not so completely as to eliminate their effects on COX-1-mediated TXB₂ production. In contrast, rofecoxib, with a 35-fold selectivity, does not affect platelet function or thromboxane production. Whole blood assays, however, may not reflect the situation in other tissues where a lower arachidonic acid availability may exacerbate COX-1 inhibition. The present assays for the determination of COX-1 K_i values and IC₅₀ values for the inhibition of microsomal COX-1 were designed to compare the potential of the various inhibitors to inhibit COX-1 at low arachidonic acid concentration. The two assays gave the same rank order of potency of the various inhibitors and showed that etoricoxib is significantly less potent against COX-1 than the other agents that selectively inhibit COX-2. Whether the higher selectivity of etoricoxib in vitro will confer therapeutic advantages in normal or pathological situations remains to be fully evaluated. Clinical studies in progress have shown that etoricoxib is well tolerated and efficacious in the treatment of the symptoms of osteoarthritis (Gottesdiener et al., 1999) and is selective for COX-2 in vivo even at single doses more than 8-fold the clinically effective dose in osteoarthritis (Dallob et al., 2000). In addition, its major metabolites have no significant effects on COX-1 activity (N. Chauret, J. A. Yergey, C. Brideau, R. W. Friesen, J. Mancini, D. Riendeau, J. Scheigetz, J. Silva, A. Styhler, L. A. Trimble, and D. A. Nicoll-Griffith, submitted). Further clinical studies with etoricoxib should contribute to fully explore the therapeutic potential and tolerability of selective inhibitors of COX-2 in inflammatory conditions, cancer, and neurological diseases.