Pharmacokinetics of High-Dose Oseltamivir in Healthy Volunteers

Abstract

The effects of loading doses and probenecid coadministration on oseltamivir pharmacokinetics at four increasing dose levels in groups of eight healthy adult Thai volunteers (125 individual series) were evaluated. Doses of up to 675 mg were well-tolerated. The pharmacokinetics were dose linear. Oseltamivir phosphate (OS) was rapidly and completely absorbed and converted (median conversion level, 93%) to the active carboxylate metabolite. Median elimination half-lives (and 95% confidence intervals [CI]) were 1.0 h (0.9 to 1.1 h) for OS and 5.1 h (4.7 to 5.7 h) for oseltamivir carboxylate (OC). One subject repeatedly showed markedly reduced OS-to-OC conversion, indicating constitutionally impaired carboxylesterase activity. The coadministration of probenecid resulted in a mean contraction in the apparent volume of distribution of OC of 40% (95% CI, 37 to 44%) and a reduction in the renal elimination of OC of 61% (95% CI, 58 to 62%), thereby increasing the median area under the concentration-time curve (AUC) for OC by 154% (range, 71 to 278%). The AUC increase for OC in saliva was approximately three times less than the AUC increase for OC in plasma. A loading dose 1.25 times the maintenance dose should be given for severe influenza pneumonia. Probenecid coadministration may allow considerable dose saving for oseltamivir, but more information on OC penetration into respiratory secretions is needed to devise appropriate dose regimens.

Oseltamivir has been stockpiled by many governments across the world for treatment and prophylaxis in the event of an influenza pandemic (36). Oral oseltamivir is safe and generally well-tolerated in healthy persons, except for dose-related gastrointestinal disturbance, at doses of up to 1,000 mg per day. Efficacy in the prevention and treatment of uncomplicated seasonal influenza in previously healthy persons has been assessed (2, 9, 10, 11, 16, 17, 24, 33). Doses of 75 mg (the currently approved dose) or 150 mg given twice daily are effective in treating uncomplicated acute H1N1 and H3N2 influenza in adults (9, 10, 33). Avian influenza, especially H5N1 infection, causes severe disease with very high mortality in humans. Over 60% of the patients with confirmed H5N1 influenza in the recent epizootic have died since November 2003 (37). Higher doses than those currently recommended and rapid attainment of therapeutic concentrations may improve the outcome.

Oseltamivir phosphate (OS) is an orally bioavailable ester prodrug form of the active carboxylate metabolite, a potent and selective inhibitor of influenza A and B virus neuraminidase (3, 35). After oral administration, OS is rapidly hydrolyzed to its active metabolite oseltamivir carboxylate (OC), which is then excreted by glomerular filtration and renal tubular secretion without further metabolism (1, 5, 12, 13, 23, 25, 30). The OC elimination half-life is approximately 6 to 8 h in adults and up to 13 h in children (25) and 36 h in patients with renal failure (29). The pharmacokinetic properties of OS and OC in Caucasian and Japanese subjects are similar (30) and are unaffected by coadministration with amantadine (23). In the elderly, clearance (CL) is reduced, so plasma OS and OC concentrations are higher than those in younger subjects (1). There is no parenteral preparation, and oseltamivir pharmacokinetics in severely ill patients have not been assessed. One proposed strategy to increase OC levels, and thereby reduce dose requirements, is concurrent administration with probenecid, a potent competitive inhibitor of the renal tubular secretion of weak organic acids (27). Probenecid has a marked effect on OC pharmacokinetics in healthy adult volunteers (13, 14, 15, 28, 30), reducing renal CL and increasing the area under the concentration-time curve (AUC), but whether this effect increases concentrations of OC in respiratory secretions has not been determined.

Patients with H5N1 influenza usually develop severe pneumonia. The clinical and virological responses to oseltamivir in patients with avian H5N1 influenza are slower than those in patients with seasonal influenza infections (6, 38). Resistance emerges readily (6). In animal models, higher oseltamivir doses and longer treatment courses than those required for other influenza virus infections are needed to prevent and treat H5N1 influenza (4, 8, 40). The target OC concentrations in the treatment of avian influenza need to be defined, and thus, the optimum dose regimen remains uncertain. H5N1 infections are associated with high and sustained levels of viral replication, so it would seem best to reach target steady-state concentrations as early as possible in the course of infection (i.e., within 4 h). The objective of this study was to assess high-dose regimens as potential treatments for H5N1 influenza and to characterize the effects of probenecid on oseltamivir pharmacokinetics. Salivary concentrations were measured as a surrogate for concentrations in pulmonary secretions.

MATERIALS AND METHODS

Groups of eight subjects per dose level were monitored in a four-period randomized open-label crossover study using a balanced four-by-four Latin square design with two individuals per sequence. Standardized differences (standard deviation [SD]/difference ratios) were likely to be less than 3 for dose and probenecid effects and 1.5 for the loading dose (13), so this design had >90% power to detect differences of this magnitude. The four groups of eight healthy Thai volunteers (four male and four female) were randomized within each block to one of four study sequences, as follows: (i) O, oseltamivir only, administered in two equal doses 12 h apart; (ii) OP, same as the O regimen but with the addition of probenecid in eight 500-mg doses given every 6 h; (iii) OL, an oseltamivir loading dose and then one smaller maintenance dose at 12 h; and (iv) OPL, same as the OL regimen but with the addition of probenecid in eight 500-mg doses given every 6 h.

Oral probenecid was started together with the oseltamivir. Four study blocks were conducted sequentially with escalating oseltamivir doses: 75, 150, 300, and 450 mg. The corresponding loading doses evaluated were 150, 225, 450, and 675 mg, respectively. Volunteers were allowed to participate at each level to explore inter- and intrasubject variation further.

Volunteers were healthy nonsmokers aged 18 to 45 years, were not pregnant or lactating, were taking no other drugs, and had normal baseline hematology, biochemistry, and urine screening results and no drug allergies. The study protocol was approved by the ethical committee of the Faculty of Tropical Medicine, Mahidol University, and the network steering committee of the Southeast Asia Influenza Clinical Research Network. A full explanation of the study was provided to all potential volunteers, who gave informed written consent to participation.

Volunteers were admitted to the volunteer facility at the Hospital for Tropical Diseases the evening before the study began and thereafter remained in bed. They were given a standard American breakfast estimated at 440 kcal (37% fat) 30 to 45 min before each initial drug administration and were allowed to eat freely from 1 h after drug administration. The evening meal given 30 to 45 min before the second drug administration was a standard Thai meal comprising rice, vegetables, and a small portion of meat (circa 700 kcal). The oral study medications, OS (Tamiflu; Roche Pharmaceuticals, Switzerland) and probenecid (Benuryl; Valeant Pharmaceuticals), were taken with water. Volunteers were observed for 1 h after drug administration to ensure that the medications were retained. Blood samples (4 ml) were taken at 0, 30, 60, 90, 120, and 180 min and then at 4, 5, 6, 7, 8, 10, 12, 14, 16, 24, and 48 h. An important source of variation, not addressed in previous pharmacokinetic studies with humans, is rapid ex vivo hydrolysis of the OS to OC (18, 19). This process was prevented by the use of fluoride-oxalate blood collection tubes, which were immediately subjected to centrifugation at 2,000 × g for 10 min. Plasma was stored at −80°C. The washout period between doses was ≥5 days after the 2-day sampling period. Saliva samples were obtained at 0, 6, 8, 10, and 12 h. Complete 24-h urine collections were made from 0 to 24 h and from 24 to 48 h.

Analytical methods.

OS and OC in plasma, urine, and saliva were measured by high-throughput liquid chromatography (LC)-tandem mass spectrometry. Plasma samples were analyzed using mixed-mode solid-phase extraction with mixed-phase cation-standard density plates (3M Empore, Bracknell, United Kingdom) and hydrophilic interaction LC-tandem mass spectrometry (20). Deuterated OS and OC internal standards (50 μl) were added to 50-μl samples in a 96-well plate. The pHs of the samples were adjusted with 500 μl of ammonium acetate buffer (pH 3.5; 5 mM). The whole sample volume was loaded for plasma and saliva analyses, while one-third of the sample volume was loaded for urine analysis. OS and OC levels were quantified using an API 5000 triple quadrupole mass spectrometer (Applied Biosystems/MDS SCIEX, Foster City, CA) with a TurboV ionization source interface operated in the positive-ion mode. Quantification was performed using selected reaction monitoring for m/z transitions from 313 to 225 and 316 to 228 for OS and deuterated OS, respectively, and 285 to 197 and 288 to 200 for OC and deuterated OC, respectively. The LC system was an Agilent 1200 system (Agilent Technologies, Santa Clara, CA). Data acquisition and quantification were performed using Analyst 1.4 (Applied Biosystems/MDS SCIEX, Foster City, CA). The coefficient of variation for the measurement of OS in plasma at 3 ng/ml was 5.3%, and that for OC at 30 ng/ml was 4.0%.

Pharmacokinetic analysis.

In compartmental modeling (using WinNonlin; Pharsight Co., Mountain View, CA), each pair of concentration profiles (for OS and OC) were modeled simultaneously, assuming a one-compartment model with first-order absorption and elimination for both profiles and complete OS-to-OC conversion. Incomplete conversion was not supported by the data. Except for the first-order absorption rate constant (k_a) for OS, pharmacokinetic parameters for the first and second doses were assumed to be the same. Noncompartmental pharmacokinetic analysis (NCA) was conducted using STATA (release 9.0; StataCorp, TX). Models were compared using the Akaike information criterion, analysis of variance for nested models, and visual examination of residuals. A model with a lag phase for OS absorption was selected if it offered a significant increase in the likelihood function and reduced the standard errors of estimates. The effects of probenecid administration and the oseltamivir dose on the pharmacokinetic parameters were examined using mixed-effects modeling. Effects of covariates such as age and sex on the pharmacokinetic parameters were examined in the same model by the log likelihood ratio test.

As oseltamivir was given orally, the fraction of the drug absorbed (F) was not known, so the volume of distribution (V) and CL estimates are V/F and CL/F. For OC, these values can be estimated if the fraction converted is assumed to be as follows: 1 − the fraction of OS eliminated in the urine. This approach was supported by the balance estimations. The fraction of the dose excreted was calculated as a ratio of the total amount of the drug excreted (as OC or OS) in the urine during the period from 0 to 48 h to the orally administered dose in molar units.

RESULTS

Subjects.

Studies of four dose levels with 32 individual studies per level (n = 128) were planned, but one subject withdrew after one series because of a skin reaction, so there were a total of 125 individual pharmacokinetic series. Fifty-two volunteers were screened, and 21 participated in the study: 11 males and 10 females, with a median age of 25 years and a range of 21 to 42 years. There were a total of 32 individual blocks comprising four dose levels; 10 volunteers participated in one block and 11 participated in two nonconsecutive blocks.

Adverse effects.

Adverse events, graded as mild, moderate, or severe, were reported on 33 occasions—11 (18%) following the administration of oseltamivir alone and 22 (35%) following administration with probenecid. These were generally mild (two were of moderate severity, and none were severe) and were related mainly to nausea, vomiting, mild abdominal discomfort, or nasal congestion. A 27-year-old female subject in the OPL arm developed a self-limiting fixed drug eruption on the left hand which was considered most likely to represent a probenecid allergy. In 24 cases (7 with oseltamivir alone and 17 with oseltamivir and probenecid), the adverse events were considered to be probably or possibly related to the drugs. Adverse events were associated with higher levels of OC in plasma (P, 0.006 for the OC AUC) and with taking probenecid (P, 0.024). After adjustment for the plasma OC levels, probenecid had no effect on the rate of adverse events. This conclusion remained true when adverse events only possibly or probably related to the study medication were considered.

Plasma drug concentration profiles.

Oseltamivir was rapidly absorbed and converted to OC (Fig. 1). One-compartment models for OS and OC with complete first-order interconversion provided good fits to the data (Table 1). A 21-year-old male (enrolled twice, in blocks 2 and 4) in whom levels of OS and OC were repeatedly similar (Fig. 2) had very high OS peaks, indicating considerably reduced biotransformation. In the other subjects, the OS levels were approximately 10-fold lower than the OC levels.

FIG. 1.

Concentrations of OS (A) and OC (B) in plasma (median values) for all subjects by study block.

TABLE 1.

OS and OC pharmacokinetic parameter estimates

Parametera	Median (IQR) and range for:
	OSb	OCb
ka (h−1)	1.011 (0.785-1.699), 0.414-9.552
F	3.439 (1.939-7.318), 0.178-33.425
Tlag (h)	0.466 (0.308-0.603), 0.173-1.187
Cmax (nmol/ml/mg/kg)	0.153 (0.115-0.214), 0.049-0.666	0.662 (0.543-0.716), 0.280-1.126
Tmax (h)	1.0 (1.0-1.5), 0.5-3.0	5.0 (4.0-5.0), 1.0-7.0
V (liters/kg)	11.019 (7.846-15.855), 1.858-24.575	3.309 (2.776-3.943), 1.621-6.232
kf (h−1)	0.684 (0.554-0.860), 0.270-4.288
kel (h−1)		0.136 (0.113-0.160), 0.064-0.262
t1/2 (h)	1.013 (0.806-1.252), 0.162-2.566	5.080 (4.345-6.142), 2.649-10.870
CL (liters/kg/h)	8.286 (6.354-10.742), 1.106-15.203	0.417 (0.370-0.484), 0.313-1.017
AUC0-12 (nmol/ml·h)/(mg/kg) 2	0.324 (0.260-0.403), 0.194-2.508	5.076 (4.252-5.713), 2.206-7.432
AUC0-∞ (nmol/ml·h)/(mg/kg) (estimated by NCA)	0.325 (0.262-0.404), 0.194-2.547	5.872 (5.414-6.617), 2.617-7.829
AUC0-∞ (nmol/ml·h)/(mg/kg) (estimated by compartmental modeling)	0.386 (0.298-0.504), 0.211-2.894	7.685 (6.618-8.646), 3.147-10.235

^aParameters estimated by compartmental modeling were as follows: k_a, F (change in k_a after the second dose such that k_a after the second dose equals k_a after the first dose divided by F), V, k_f (first-order rate constant for OS-to-OC conversion), k_el (elimination rate constant), t_1/2 (elimination half-life), CL, AUC_0−∞ for plasma, and T_lag (lag time). Parameters estimated by NCA were as follows: C_max, T_max (time to maximum concentration), AUC_0-12 for plasma, and AUC_0-∞ for plasma. AUC_0-12 was calculated using the trapezoid method. AUC_0-∞ was estimated using an exponential fit (C_t [concentration at time t] = A·e^−αt, where α is the elimination rate constant, A is the maximum plasma concentration, and t is time) to the last three drug levels before the second dose (i.e., the levels at 8, 10, and 12 h), as implemented in STATA.

^bMolecular weights: OS base, 312.41; OC, 284.35.

FIG. 2.

Relationship between the AUC_0-∞ for plasma and the actual dose of OS (A) or OC (B) received. Data are from O and OL regimens, all doses. The filled circles are for the one subject with consistently impaired conversion of OS to OC.

Pharmacokinetics.

For 100 profiles (47 for oseltamivir alone and 53 for oseltamivir plus probenecid), the addition of an OS absorption lag phase improved model fits. The OS k_a was reduced in 90% of cases after the second dose by a median factor of 3.4 (95% confidence interval [CI], 2.3 to 5.2). There was no evidence of dose dependence in either the OS elimination kinetics, the conversion of OS to OC, or the elimination of OC (Table 2). The OS dose-AUC and maximum concentration (C_max)-AUC relationships were linear over wide ranges (Fig. 2). Median elimination half-lives (and 95% CIs) were 1.0 h (0.9 to 1.1 h) for OS and 5.3 h (4.7 to 5.9 h) for OC. There was no relationship between the half-life and the C_max. There was also a linear relationship between the oseltamivir dose and the OC AUC, as follows: log (AUC) = 1.039 log (dose) + 1.944 (Fig. 2). This equation translates into a 152% AUC increase for a 1.5-fold dose increase and a 205% increase for a doubling of the dose. These values are not significantly different from the dose ratios and confirm dose linearity in the pharmacokinetics of OS and OC. In all analyses, no sequence effect was observed.

TABLE 2.

Results of mixed-effects modeling^a of OS and OC pharmacokinetic parameters

Drug form and pharmacokinetic parameterb	Estimate (95% CI)c and P value for:			P value for dose-probenecid interaction	P value for period	σb	σ	P = σb2/ (σb2 + σ2)
	Constantd	Dosee	Probenecid
Oseltamivir
t1/2	0.998 (0.924 to 1.072); P < 0.001	P = 0.094	0.111 (0.043 to 0.178); P = 0.001	0.140	0.926	0.127	0.192	0.303
CL	8.520 (7.344 to 9.696); P < 0.001	P = 0.059	−0.742 (−1.457 to −0.027); P = 0.042	0.106	0.596	2.442	2.035	0.590
V	3.451 (3.167 to 3.735); P < 0.001	P = 0.214	P = 0.113	0.251	0.760	0.596	0.656	0.452
Log (AUC)	−0.775 (−1.003 to −0.547); P < 0.001	0.750 (0.702 to 0.797); P < 0.001	0.119 (0.040 to 0.198); P = 0.003	0.332	0.687	0.471	0.225	0.814
Log (Cmax)	−1.730 (−1.954 to −1.504); P < 0.001	0.772 (0.704 to 0.841); P < 0.001	0.924	0.996	0.453	0.414	0.329	0.613
Oseltamivir carboxylate
t1/2	2.294 (2.161 to 2.427); P < 0.001	P = 0.694	0.514 (0.402 to 0.627); P < 0.001	0.918	0.197	0.243	0.319	0.366
CL	0.664 (0.639 to 0.688); P < 0.001	P = 0.060	−0.253 (−0.268 to −0.238); P < 0.001	0.106	0.597	0.051	0.043	0.581
V	1.816 (1.732 to 1.900); P < 0.001	P = 0.096	−0.437 (−0.494 to −0.380); P < 0.001	0.178	0.863	0.169	0.162	0.521
Log (AUC)	2.175 (2.084 to 2.267); P < 0.001	0.720 (0.694 to 0.746); P < 0.001	0.945 (0.901 to 0.989); P < 0.001	0.222	0.347	0.167	0.124	0.643
Log (Cmax)	−0.354 (−0.476 to −0.232); P < 0.001	0.749 (0.714 to 0.783); P < 0.001	0.535 (0.477 to 0.594); P < 0.001	0.906	0.565	0.221	0.167	0.638

^aRandom-effects model. For any parameter a (after transformation, if necessary to provide normality in the distribution), the following mixed-effects model was fitted: a_ijk = a + probenecid_j + dose + period_k + v_i + ɛ_ijk [v_i ∼ N(0, σ_b²); ɛ_ijklm ∼ N(0, σ²); i = 1,…, 21; j = 0, 1; k = 1, 2, 3, 4], where a is a population estimate of the parameter for participants who received 75 mg of oseltamivir alone in the first period of the first block, dose represents the change in the parameter's value resulting from dose increase (modeled as a continuous variable), probenecid_j represents the change in the parameter value associated with adding probenecid, period_k represents the change in the parameter value associated with period k, v_i represents the ith subject effect, ɛ_ijklm represents the error term or residual, σ_b² denotes between-participant variability, and σ² denotes within-participant variability.

^bEstimated by noncompartmental modeling (C_max) or compartmental modeling (V, CL, AUC, and t_1/2 [half-life]). The AUC and C_max parameters for OS and OC were modeled after log transformation, while the V and half-life for OS and the V, half-life, and CL for OC were modeled after square-root transformation.

^cThe estimate and 95% CI are given if the P value is <0.05.

^dThe constant corresponds to the average value of the transformed parameter for a subject who received oseltamivir only at a dose of 1.25 mg/kg. To calculate the parameter estimate, this value needs to be back transformed [using exp(x) or x²] to the original scale.

^eThe dose in milligrams per kilogram was modeled after a logarithmic transformation (to the base of 2), so this coefficient corresponds to the increase in the parameter value associated with the doubling of the dose.

Conversion of OS to OC.

Excluding the poor converter, the median OS-to-OC AUC ratio was 5% (range, 3 to 9%). The ratio was lower following probenecid coadministration: 2% (1 to 6%). This was because of the increased OC AUC resulting from the reduced renal elimination. The poor converter had a median OS-to-OC AUC ratio of 65% (range, 63 to 77%) with oseltamivir alone (n = 4), compared to 22% (21 to 27%) with probenecid coadministration (n = 4). Estimates of the OS V and OS elimination rate constant for this subject were among the lowest in the data set, and the total OS CL was at least three times lower than that for any other participant. For OC, estimates of V and CL were among the highest in the data set but were not distinct outliers.

Bioavailability.

The amounts of OS converted to OC and excreted in the urine add up to the dose administered (Fig. 3). The OS converted to OC accounted for a median of 93% (range, 76 to 113%) of the total dose. This finding indicates nearly complete absorption and no other significant routes of OS metabolism or excretion.

FIG. 3.

Relationship between the estimated total amount of drug excreted in the urine and the dose of OS given.

Sources of pharmacokinetic variation.

More than half of the variation in pharmacokinetic parameter estimates resulted from intersubject variation (Table 2). These proportions were as follows: OS CL, 59%; OC CL, 58%; OC V, 56%; OS AUC, 81%; OC AUC, 64%; OS C_max, 61%; and OC C_max, 64%. Changes in OC kinetics and OS kinetics were not correlated. Age, weight, and sex had no effect on OC pharmacokinetics, but the mean OS CL for women was approximately 25% greater than that for men, 9.86 liters/kg/h (95% CI, 8.51 to 11.20 liters/kg/h) versus 7.33 liters/kg/h (5.80 to 8.87 liters/kg/h; P = 0.007), and the mean V in women was approximately 30% larger than that in men, 10.09 liters/kg (95% CI, 7.76 to 12.71 liters/kg) versus 14.27 liters/kg (12.16 to 16.55 liters/kg; P = 0.02). Excluding the male poor converter, men had larger OS AUC values than women, by a mean of 28% (95% CI, 5 to 57%), and higher OS C_max values, by a mean of 35% (5 to 73%).

Loading dose.

The loading doses produced predicted concentrations and therefore would be expected to attain average steady-state concentrations within 4 h of administration. The predicted average steady-state OC concentrations achieved with multiple doses of 75, 150, 300, and 450 mg (for a 60-kg subject) are 227, 454, 908, and 1,362 ng/ml, respectively. To achieve these concentrations after the first dose, the ratio of the loading dose to the maintenance dose was estimated to be 1.24 (range, 1.04 to 1.86). Thus, the required loading and maintenance doses would be 1.68 and 1.35 mg/kg of body weight, 3.35 and 2.70 mg/kg, 6.71 and 5.40 mg/kg, and 10.06 and 8.09 mg/kg, respectively, assuming the same absorption rate for all doses. The minimum steady-state concentration would be achieved in 2 h with these loading doses.

Effects of coadministered probenecid.

The administration of probenecid had small effects on OS pharmacokinetics. On average, probenecid decreased the OS CL by approximately 13% (95% CI, 2.4 to 23%; P = 0.042) but did not affect the OS V significantly. The OS AUC increased 15% (95% CI, 2.5 to 30%; P < 0.001), but the OS C_max was not affected significantly (P = 0.99). There was no dose effect. In contrast, probenecid had a profound effect on OC pharmacokinetics, markedly increasing the C_max and AUC. The median AUC increase was 154% (range, 71 to 278%). Intrasubject variation in the probenecid effect was greater than intersubject variation; the SD of the intersubject AUC increases was 24%, while the intrasubject SD was 40%. Thus, intersubject variation explained 26% of the overall variance. The mean reduction in OC CL was 61% (95% CI, 58 to 62%). The SD of the intersubject change in OC CL was 3.4%, compared with 6.2% for intrasubject changes, so that intersubject variation explained 23% of the overall variance. Probenecid consistently contracted the OC V, causing a mean reduction of 40% (95% CI, 37 to 44%). None of the total variation change in V was explained by intersubject variation (between-participant SD [σ_b] = 0%; within-subject SD σ = 14%). There was no correlation between the probenecid effects on CL and V (P = 0.28). The effects of probenecid on OS and OC pharmacokinetics were independent of the OS dose. Modeling suggested that target concentrations of OC in plasma could be obtained with only 46% of the oseltamivir loading dose and 36% of the maintenance dose with probenecid coadministration, saving 63% of the oseltamivir dose over the 5-day course.

Urinary OS and OC excretion.

Overall, 64% (range, 24 to 89%) of the OS dose was excreted in urine as either OC or OS. The urinary excretion of OS increased with the dose (P < 0.001) and was unaffected by probenecid (P = 0.35). OC and OS renal CL values were correlated well. OS renal CL was approximately 4.4% (range, 1.2 to 31.3%) of the total CL without probenecid and 3.1% (0.7 to 25%) with probenecid. Corresponding fractions for OC were 64% (18 to 92%) and 73% (25 to 108%) of the total CL. OC CL increased by approximately 3 ml/min (range, 0.7 to 5.4 ml/min) for each 100-mg increase in dose (P = 0.010). The ratio of OC CL to creatinine CL correlated with the dose; for OS, this ratio was 3.85 (range, 3.33 to 4.38) for the 150-mg total dose and increased by 0.1 (0.04 to 0.2) for each additional 100 mg of the dose. For OC, the ratio was 2.79 (2.53 to 3.05) for the 150-mg total dose and increased by 0.07 (0.04 to 0.11) for each additional 100 mg of the dose. Probenecid coadministration markedly reduced urinary CL of OC and OS (P < 0.001). The ratio of urinary drug CL to creatinine CL was reduced by 23% for OS, to 0.89 (0.45 to 1.34), and by 54% for OC, to 1.52 (0.29 to 1.75). No sequence effects were observed, except in the ratio of drug CL to creatinine CL (P, 0.002 for OC and 0.044 for OS), which was higher at the first time point than at the subsequent time points.

Salivary OS and OC concentrations.

The median ratio of OS concentrations in saliva and plasma was 92% (range, 17 to 1,300%; interquartile range [IQR], 68 to 128%), and the ratio for OC was 4.7% (range, 0 to 61%; IQR, 2.2 to 9.2%). The median ratio of the OS AUCs from 0 to 12 h (AUC_0-12) for saliva and plasma was 92% (range, 24 to 276%), compared with 4.5% (range, 0.8 to 30%) for OC. The relationship between AUCs for OC in plasma and saliva was linear (Fig. 4A). The administration of probenecid increased the OC AUC for saliva much less than the AUC for plasma, by 34% (−50 to 207%) for saliva compared with 104% (−45 to 210%) for plasma. There was a significant negative relationship (P < 0.001) between the salivary AUC increase and saliva OC levels without probenecid (Fig. 4B).

FIG. 4.

(A) Relationship between AUCs (log scales) for OC in plasma and saliva. (B) Relationship between the relative increase in the AUC_0-∞ for OC in saliva with the administration of probenecid [(OP group AUC_0-∞ − O group AUC_0-∞)/O group AUC_0-∞)] and the AUC_0-∞ for OC in saliva without the administration of probenecid. All doses were evaluated.

DISCUSSION

This, the most detailed examination yet of the pharmacokinetic properties of OS and its active metabolite (OC), provides a sound basis for high-dose treatment regimens for H5N1 influenza. The absorption of OS was almost complete but was usually slower after the second dose, which may be related to different food contents of the preceding meals. On average, 93% of OS was converted to OC, and the remainder was excreted in the urine, indicating no other significant routes of elimination. OS concentrations generally peaked around 1 h after drug administration, and OC concentrations peaked between 4 and 5 h. OC levels were approximately 10 times higher than OS levels, and AUCs for OC in plasma were 20 times higher than those for OS, reflecting a considerably smaller OC V. The OS elimination half-life averaged around 1 h, compared to 4.5 to 6.5 h for OC. These values are approximately similar to previous estimates (1, 5, 12, 13, 23, 25, 30). There was no evidence of dose dependency in any of the pharmacokinetics over an eightfold dose range, so increasing dosing should lead to predictable increases in plasma drug concentrations. The administration of a loading dose 25% higher than the maintenance dose would seem appropriate for severely ill patients. This regimen would achieve average steady-state trough levels within 2 h of the loading dose with normal rates of conversion. In cases of life-threatening illness, achieving therapeutic concentrations as rapidly as possible may save lives.

Pharmacokinetic-pharmacodynamic relationships for this important antiviral are not well-characterized, so the optimum plasma OC concentration profile for the treatment of H5N1 influenza is not known. This is because viral replication takes place principally in the respiratory mucosa and perhaps, in H5N1 influenza pneumonia, in the pulmonary parenchyma and extrapulmonary sites. OC concentrations in the plasma and bronchoalveolar fluids of rats were reported to be similar, with a slower decline in bronchoalveolar fluid than in plasma (7), but recent studies with mice (21) and the considerably smaller apparent V for OC than for OS suggest a more limited tissue distribution of OC than of OS. Respiratory secretions could not be measured directly in this volunteer study, but saliva was sampled as a surrogate. OS concentrations in saliva and plasma were similar, but OC concentrations averaged less than 5% of those in the plasma. Clearly, obtaining further information on OC concentrations in respiratory secretions and the lung in influenza pneumonia is an urgent priority.

Changes in OS and OC kinetics did not correlate with each other. Over half the variation was accounted for by intersubject differences. Women had lower concentrations of OS in plasma than men, as a result of both larger V and higher CL values. There were no differences in OC kinetics between men and women.

Probenecid had a marked effect on OC pharmacokinetics, as reported previously (13, 15, 30), but not only through the inhibition of renal tubular secretion. Probenecid inhibits two polymorphic drug transporters: human organic anion transporter 1, present on the renal tubular basolateral membrane, and the multidrug transporter MRP2 (or ABCC2), present on the renal and enterocyte brush border membranes and the liver canalicular membrane (21). MRP2 is also expressed in lung cells in culture (32). The coadministration of probenecid increased the AUC for OC in plasma by a median of 154% because it reduced urinary CL and also independently contracted V by 40%. This finding suggests that transport to some or all tissue compartments was reduced. The increase in the OC AUC for saliva with probenecid administration was only one-third of the increase in the AUC for plasma, and there was a significant negative relationship between the salivary AUC increase and the saliva OC levels without probenecid administration. MRP2 is expressed in salivary glands (34), so this outcome may result from variable levels of baseline transporter activity in the salivary glands or the saturation of back transport at high drug concentrations. This finding suggests that OC transport into different tissues, such as the bronchoalveolar compartment, is not simply by passive diffusion and provides an important warning that the probenecid effect needs to be investigated directly with OC measurements in respiratory secretions before the drug can be recommended. With these important provisos, modeling suggests that to attain concentrations in plasma similar to those obtained with oseltamivir alone, the dose of oseltamivir can be reduced by about two-thirds if coadministered with probenicid.

One subject, studied twice, had consistent markedly reduced conversion of OS to OC. The principle enzyme responsible for this conversion is human carboxylesterase 1, which has polymorphic variants and also functional promoter polymorphisms (31, 41). This low level of conversion is relevant both to antiviral efficacy, which would be significantly reduced, allowing greater selective pressure for the emergence of resistance, and to toxicity. Oseltamivir has been linked previously with neuropsychiatric reactions and suicides in adolescents and is potentially neuroexcitatory (15). OS is actively exported from the central nervous system compartment by P-glycoprotein, thereby limiting brain uptake, whereas OC penetrates the blood-brain barrier poorly (22, 26). High concentrations of OS in plasma may result in increased concentrations in the central nervous system. Individuals with reduced P-glycoprotein activity or reduced carboxylesterase activity, such as infants and the volunteer in this study, may be at increased risk for neurotoxicity (15). Studies are under way to investigate the pharmacogenetic basis of this impaired OS conversion. The conversion of OS to OC may also be reduced in cases of severe influenza; human carboxylesterase 1 expression is reduced by interleukin-6 (39). In this extensive volunteer study, oseltamivir doses of up to 675 mg were well-tolerated. The only significant adverse effect in the study, a localized cutaneous fixed drug eruption, may well have resulted from probenecid. Although all but two adverse events were considered minor, these events were positively associated with the OC concentrations in plasma.

Clinical and experimental evidence suggests that high doses of oseltamivir will be needed for the optimal treatment of H5N1 influenza (4, 6, 8, 38, 40). This study provides reassurance that adult maintenance doses of up to 450 mg are reasonably well-tolerated, with dose linearity in pharmacokinetics. A loading dose should be used. Probenecid may allow a dose reduction of approximately two-thirds, but as there remains important uncertainty about its effects on OC penetration into respiratory secretions, this issue will need to be investigated directly before probenecid can be generally recommended.