Investigating Clinically Adequate Concentrations of Oseltamivir Carboxylate in End-Stage Renal Disease Patients Undergoing Hemodialysis Using a Population Pharmacokinetic Approach

Abstract

End-stage renal disease (ESRD) patients receiving hemodialysis (HD) are at heightened risk for influenza, but the optimal oseltamivir dosage regimen for treating or preventing influenza in this high-risk population is still uncertain. Pharmacokinetic data for 24 adults with ESRD were pooled from a single-dose and a multiple-dose study to develop a population pharmacokinetic model using nonlinear mixed-effects modeling. The final model comprised five compartments, two each to describe the systemic pharmacokinetics of oseltamivir phosphate and its metabolite, oseltamivir carboxylate (OC), and a delay compartment to describe oseltamivir metabolism. Estimated OC clearance in the model was markedly faster during HD sessions (7.43 liters/min) than at other times (0.19 liter/min). Model simulations showed that 30 mg oseltamivir given after every HD session is the most suitable regimen for influenza treatment, producing trough OC concentrations above the median value achieved with the 75-mg twice-daily regimen in patients with normal renal function and peak concentrations below the highest oseltamivir exposures known to be well tolerated (median exposures after twice-daily dosing of 450 mg). Administration of the first dose following diagnosis of influenza need not wait until after the next HD session: addition of a single 30-mg dose during the 12 h before the next HD session raises OC exposures quickly without posing any safety risk. Further simulation showed that 30 mg oseltamivir given after every other HD session is the most suitable regimen for influenza prophylaxis.

INTRODUCTION

Patients with end-stage renal disease (ESRD) who undergo dialysis are at higher risk of infections because of a compromised immune system and the frequent use of catheters or insertion of needles to access the bloodstream. Infections such as influenza are a major cause of morbidity and mortality among ESRD patients, and preventative therapy and drug treatments are thus critical (1). Oseltamivir (Tamiflu) is a potent, stable, and selective inhibitor of influenza A and B virus neuraminidase enzymes. The prodrug, oseltamivir phosphate (OP), is administered orally and is rapidly absorbed from the gastrointestinal tract before being metabolized in the liver into the active form, oseltamivir carboxylate (OC). OC is excreted renally via glomerular filtration and active tubular secretion (2, 3).

Hemodialysis (HD) is the most commonly used form of dialysis in ESRD patients and is typically performed three times a week, with each session lasting 3 to 5 h (4). HD significantly alters the pharmacokinetic profile of drugs which are cleared renally, making it harder to predict a suitable dosing regimen. A recent population analysis of oseltamivir pharmacokinetics in patients with various degrees of renal impairment showed that those with creatinine clearance (CL_CR) rates of >60 ml/min do not require any dosage adjustments (5). In adult patients with severe renal impairment (CL_CR of 10 to 30 ml/min), recommended oseltamivir doses are 30 mg once daily for treatment and 30 mg every other day for prophylaxis, but the ideal treatment and prophylaxis dosages for patients with ESRD are less clear. In a study of 12 ESRD patients on HD, results of a noncompartmental analysis of oseltamivir pharmacokinetics after multiple doses suggested that a dose of 30 mg after alternate HD sessions would provide clinically adequate exposures for treatment (5-day course) and prophylaxis (6-week course) (6). However, the pharmacokinetic sampling in that study was done primarily after the first HD session, and the possibility remained that trough concentrations during the following alternate HD session would drop below target therapeutic levels.

The use of nonlinear mixed-effects modeling allows pooling of data from multiple studies of special populations, in which sample sizes are typically low, increasing the robustness of pharmacokinetic estimates and allowing simulation of alternative dosage regimens. The technique was recently applied to investigate oseltamivir posology in infants with influenza (7). The aim of the current study was to apply population pharmacokinetic modeling techniques to pooled data from two oseltamivir pharmacokinetic studies to determine the most suitable dosage regimen for treatment and prophylaxis of influenza in ESRD patients undergoing HD.

MATERIALS AND METHODS

Source data.

OP and OC concentration data for the pooled analysis came from patients with ESRD who underwent HD in a single-dose study and a multiple-dose study (12 patients in each study; 24 patients in total). Both studies were conducted in accordance with good clinical practice guidelines and the principles of the Declaration of Helsinki. Each study protocol was approved by the ethics committee or investigational review board at participating study centers, and all participants gave written informed consent before study entry.

The single-dose open-label study (PP15974 [http://www.roche-trials.com/]) was conducted at three centers, in New Zealand, the United Kingdom, and the United States, to assess the pharmacokinetics, safety, and tolerability of oseltamivir in ESRD patients receiving HD and peritoneal dialysis (8). Of the 24 patients enrolled (12 male and 12 female), 12 were receiving HD and were included in the current analysis; these patients had CL_CR rates of 0 to 4 ml/min, had been on dialysis for at least 6 months, and were undergoing three 4-h HD sessions each week. Patients received 75 mg OP orally as a single dose 48 h before the start of their next HD session, 30 min after a standard meal. Serial blood samples for pharmacokinetic analysis were taken immediately before dosing and 0.5, 1, 1.5, 2, 3, 4, 5, 6, 8, 10, 12, 15, 24, 36, and 48 h postdose. During HD (at 48.5, 49, 49.5, 50, 51, and 52 h postdose), additional blood samples were drawn from the dialyzer inflow and outflow. Two more samples were collected at 53 h postdose and just before the next scheduled HD session (∼92 h postdose).

The multiple-dose open-label study (NP16472 [http://www.roche-trials.com/]) was conducted at a single center in New Zealand and evaluated the pharmacokinetic and safety profiles of OP and OC in adult patients with ESRD (CL_CR of <10 ml/min) who had been on HD or continuous ambulatory peritoneal dialysis for ≥3 months. The results of the study have already been published (6). Of the 24 patients enrolled, 12 were receiving HD and were included in the current analysis; these patients received 30 mg oseltamivir orally after the completion of alternate 5-h HD sessions during a 6.5-week study period (total of nine doses over a sequence of 19 sessions). Serial blood samples for pharmacokinetic analysis were taken on days 1 and 38 (after the 1st and 17th HD sessions), immediately before dosing, and 1, 2, 4, 8, 12, 20, 32, and 42 h after dosing. Only the inflow (arterial) samples were used for the current modeling analysis.

Samples were analyzed using a previously validated high-performance liquid chromatography–double mass spectrometry method (9). The lower limits of quantification (LOQs) for OP and OC were 1.0 and 10 ng/ml, respectively. Endpoints for both studies included systemic exposure, expressed as the maximum plasma concentration (C_max), the areas under the plasma concentration-time curves (AUC_0–last) for OP and OC, dialysis clearance (CL_D), and renal clearance.

Model development.

The population pharmacokinetic analysis was conducted using nonlinear mixed-effects modeling with NONMEM software, version 7.2.0 (Icon Development Solutions) (10). The first-order conditional estimation method was employed for all model runs, with the INTERACTION option. The base model was adapted from that described by Rayner et al. (9): this described the pharmacokinetics of oseltamivir with a multicompartment linear model, where OP kinetics were described by a two-compartment model with first-order absorption, OC kinetics were described by a two-compartment model, and metabolism was described by an intermediate compartment placed between the OP and OC central compartments. Complete (100%) metabolism was assumed, and the OC central volume was estimated. CL_D was described as an additional clearance component that was switched on at the start of the dialysis sessions and switched off at the end.

Interindividual random effects were described by log-normal parameter distributions. The residual error was initially described by combined proportional and additive error models. After further refinement, the residual error of the final model was described by the exponential error for OP and a maximum effect (E_max)-type function for OC in log-transformed variables (11), as follows:

$$\text{log}\left(C_{ij}^{\text{OP}}\right)=\log \left({\hat{C}}_{ij}^{\text{OP}}\right)+{\text{σ}}_{\text{OP}}{\text{ϵ}}_{ij}^{\text{OP}}$$ (1)

$$\log (C_{ij}^{\text{OC}})=\log \left({\hat{C}}_{ij}^{\text{OC}}\right)+\left({\text{σ}}_{\text{L}}-({\text{σ}}_{\text{L}}-{\text{σ}}_{\inf \hspace{0.17em}})\frac{{\hat{C}}_{ij}^{\text{OC}}-10}{{\hat{C}}_{ij}^{\text{OC}}+10}\right){\text{ϵ}}_{ij}^{\text{OC}}$$ (2)

where σ_OP is the standard deviation of oseltamivir residual error in log-transformed variables and σ_L and σ_inf are standard deviations of OC residual error in log-transformed variables at 10 ng/ml and at high (infinite) concentrations, respectively. A value of 10 ng/ml (the LOQ value for OC) was selected as the definition of σ_L. The 50% effective concentration (EC₅₀) for this function was estimated to be close to 10 ng/ml and was therefore set to a value of 10 without compromising the model fit.

Model evaluation.

Endpoints explored in the model were as follows: apparent plasma clearances of OP (CL/F) and OC (CL_M/F), apparent dialysis clearance of OC (CL_D/F), apparent central compartment volume of OP (V/F) and OC (V_M/F), peripheral compartment volume (V_P/F), OP apparent intercompartmental clearance (Q/F), OC intercompartmental rate constants k₄₆ and k₆₄, and metabolism rate constant k_met.

A nonparametric bootstrap technique (12, 13) was used to evaluate the uncertainty of the parameter estimates by obtaining 95% confidence intervals (CIs) for the parameter estimates, computed as the 2.5th and 97.5th quantiles of the bootstrap parameter distributions. The extensive model evaluation process included evaluation of diagnostic plots, prediction of time courses of OP and OC concentrations for 1,000 simulated trials, and visual predictive checks (VPCs) (14, 15). The degree of regression to the mean was evaluated by computing shrinkage for random effects and residual error (16).

Model-based simulations.

The final population pharmacokinetic model was used to simulate OP and OC concentration-time courses for a variety of dosage regimens in patients undergoing HD and to evaluate exposure spreads in the study population. Two types of simulation were performed, using pharmacokinetic parameters that were (i) estimated from the population model for the 24 patients in the pooled sample and (ii) sampled from the model-predicted distributions for a hypothetical cohort of 2,400 patients. Results of these simulations were consistent with each other; therefore, only the conditional simulation results are presented.

Simulated C_max and C_min (minimum plasma concentration) distributions for OC were computed for several possible dosage regimens (using commercially available capsule strengths), including those recommended by Robson et al. (6) on the basis of their multiple-dose study. Some of the scenarios modeled the effect of an additional oseltamivir dose between 1 and 12 h before the first HD session, both with and without a second dose after the first session. The regimens selected were guided by pharmacokinetic bridging considerations, with the aim of identifying a regimen that provides trough OC concentrations similar to those in patients with normal renal function treated with 75 mg twice daily (treatment indication) or 75 mg once daily (prophylaxis indication). Data are presented for treatment regimens of a 30-mg dose after every other HD session and a 30-mg dose after every HD session (with and without an initial dose 6 h before the next planned HD session) and for the prophylaxis regimen of a 30-mg dose after every other HD session. Simulations were also compared with historical data for patients with normal renal function treated with oseltamivir at the highest dosage accepted as well tolerated (450 mg twice daily). The respective reference limits were as follows: median C_min of 168 ng/ml (95% CI, 56.1 to 315 ng/ml) after 75 mg twice daily, median C_min of 40 ng/ml (90% CI, 14 to 75 ng/ml) after 75 mg once daily, and median C_max of 2,390 ng/ml (95% CI, 1,470 to 3,930 ng/ml) after 450 mg twice daily (2, 5, 17).

RESULTS

Analysis data set.

The demographic and clinical characteristics of the 24 patients analyzed are shown in Table 1. The data set for analysis contained 210 and 645 quantifiable plasma samples for OP and OC, respectively (see Data Set S1 in the supplemental material). Samples with concentrations below the LOQ were excluded, as was a sample with a quantifiable OC concentration that was taken before any drug was administered. For one patient, data recorded after a nominal time point of 500 h postdose were excluded, because the actual sampling times were inconsistent with the specified nominal times and the values recorded were inconsistent with the expected concentration-time profile. The individual profiles of plasma OP and OC concentrations versus nominal time in each study are shown in Fig. 1.

TABLE 1.

Baseline demographics and clinical characteristics of patients

Parametera	Value
	Single-dose study (n = 12)	Multiple-dose study (n = 12)
Mean age (SD) (yr)	49.9 (9.6)	48.0 (13.2)
No. (%) of patients in age group
18 to 65 yr	12 (100)	11 (92)
>65 yr	0 (0)	1 (8)
No. (%) of males	8 (67)	11 (92)
Mean body mass index (SD) (kg/m2)	28.3 (6.8)	25.4 (4.5)

^aSD, standard deviation.

FIG 1.

Observed individual plasma concentrations versus nominal time (semilogarithmic scale) in a single-dose study (a) and a multiple-dose study (last dosing interval) (b). Left and right panels show data for oseltamivir and oseltamivir carboxylate, respectively.

Model development and evaluation.

After refinement of the base model described previously (9), the final model was produced as shown in Fig. S1 in the supplemental material. Parameter estimates for the final model are presented in Table 2. In the final model (see the supplemental material for the final model code), the systemic pharmacokinetics of OP were described by two compartments, and oral OP absorption was described by a first-order rate constant of 2.94 h⁻¹. The systemic pharmacokinetics of OC were described by two compartments, and oseltamivir metabolism to OC was described by a delay compartment to capture the formation-rate-limited kinetics of OC. In the final model, elimination of OP was almost complete before the start of the next HD session. Nondialysis (residual renal) clearance of OC (CL_M/F) was very slow (0.189 liter/h), but dialysis clearance (CL_D/F) was much faster (7.43 liters/h), resulting in an estimated total OC clearance of 7.62 liters/h. The median CL_D/F value from the bootstrap runs was 7.28 liters/h (95% CI, 5.74 to 9.34 liters/h).

TABLE 2.

Parameter estimates for the final model^a

Parameter	Type of parameter	Model estimate (CV [%])	Bootstrap median (95% CI)	Median % CV (95% CI) for bootstrap median	Shrinkage (%)
ka (h−1)	θ1	2.94	2.66 (1.32–21.0)
CL/F (liters/h)	θ2	385	373 (294–457)
V/F (liters)	θ3	1,650	1,630 (1,110–2,530)
Q/F (liters/h)	θ4	126	155 (75.8–324)
VP/F (liters)	θ5	4,950	5,240 (2,100–107,000)
CLM/F (liters/h)	θ6	0.189	0.238 (0.0383–0.484)
CLD/F (liters/h)	θ7	7.43	7.28 (5.74–9.34)
VM/F (liters)	θ8	16.1	15.6 (11.1–20)
k46 (h−1)	θ9	0.213	0.226 (0.131–0.453)
k64 (h−1)	θ10	0.512	0.545 (0.366–0.849)
kmet (h−1)	θ11	0.218	0.234 (0.17–0.302)
σOP	θ12	0.844 (84.4)	0.826 (0.647–1.01)	82.6 (64.7–101)	1.4
σOC − l	θ13	1.3 (130)	1.92 (1.03–2.62)	192 (103–262)	6.9
σOC − inf	θ14	0.0735 (7.4)	0.0578 (0.0335–0.0888)	5.8 (3.4–8.9)	6.9
ω2ka	Ω(1,1)	0.127 (116)	1.19 (0.493–6.12)	109 (70.2–247)	25.9
ω2V	Ω(2,2)	1.34 (59.4)	0.357 (0.16–0.566)	59.7 (40.0–75.2)	9.0
ω2CLM	Ω(3,3)	0.353 (68.1)	1.77 (0.255–4.37)	133 (50.5–209)	7.1
ω2CLD	Ω(4,4)	0.464 (25.9)	0.061 (0.0309–0.0991)	24.7 (17.6–31.5)	7.4
ω2VM	Ω(5,5)	0.0673 (27.2)	0.0673 (0.0374–0.101)	25.9 (19.3–31.8)	1.6

^aCI, confidence interval; CL/F, apparent clearance of oseltamivir; CL_D/F, apparent dialysis clearance of oseltamivir carboxylate; CL_M/F, apparent clearance of oseltamivir carboxylate; CV, coefficient of variation (100 × standard deviation [%]); k_a, absorption rate constant; k_met, metabolism rate constant; k₄₆ and k₆₄, oseltamivir carboxylate intercompartmental rate constants; Q/F, oseltamivir apparent intercompartmental clearance; σ_OP, standard deviation of oseltamivir residual error in log-transformed variables; σ_{OC − l}, standard deviation of OC residual error in log-transformed variables at 10 ng/ml; σ_{OC − inf}, standard deviation of OC residual error in log-transformed variables at high (infinite) concentrations; V/F, apparent central compartment volume of oseltamivir; V_P/F, apparent peripheral compartment volume of oseltamivir; V_M/F, apparent central compartment volume of oseltamivir carboxylate; ω², interindividual variance; θ, NONMEM fixed-effects parameter; Ω, interindividual covariance matrix.

The intraindividual (residual) variability was high for OP concentrations (coefficient of variation [CV] = 84.4%) and for OC concentrations near the LOQ of 10 ng/ml (CV = 130%); variability at high OC concentrations, however, was very low (CV = 7.4%), indicating a good fit of the observed OC concentrations in the range of greatest interest. Interindividual variability of CL_D/F was related mainly to differences in CL_D/F estimates between the two studies. When the fixed effect was used instead of the random effect, CL_D/F was estimated at 9.29 liters/h (95% CI, 0.69 to 17.88 liters/h) for the single-dose study and 6.27 liters/h (95% CI, 3.92 to 8.61 liters/h) for the multiple-dose study. The difference can probably be explained by the imprecise estimate of CL_D/F for the single-dose study, as the 95% confidence interval for CL_D/F for this study completely overlapped the 95% confidence interval for CL_D/F for the multiple-dose study.

Diagnostic plots and VPCs indicated a good fit for the observed OC data (Fig. 2). As shown, distributions of interindividual random effects differed between the two studies, and these differences were most pronounced for V/F and CL_D/F. Model evaluation (bootstrap and VPC evaluations) indicated that simulated data were in good agreement with observed OC data. Various model evaluation techniques (including bootstrap and VPC evaluations) indicated the ability of the model to predict the central tendency and spread of OC concentrations.

FIG 2.

Model diagnostics for the final model. (a) Goodness-of-fit plots. The gray solid lines (y = x or y = 0) are included for reference. The bold red lines are the LOWESS (local regression smoother) trend lines. (b) Visual predictive checks (VPCs) showing logarithmic concentrations of oseltamivir carboxylate in plasma versus nominal time. From left to right, the data are for the single-dose study and the multiple-dose study (first dose and last dose). Circles show observed concentrations, and lines show medians (red) and 5th and 95th percentiles (blue) for simulated concentrations. Simulated values were computed from 1,000 trials simulated using dosing, sampling, and covariate values from the analysis data set.

Model-based simulations.

The predicted OC plasma concentration-time curves resulting from two of the treatment regimens evaluated are illustrated in Fig. 3a and b for the 24 patients in the analysis data set. Lines on each graph show OC concentrations in adults with normal renal function who received the standard therapeutic dose and the maximum well-tolerated dose (using previously described median and 95% CI values [see Materials and Methods]). With the previously recommended dosage regimen of 30 mg after every other session, some C_min values dipped below the lower 95% confidence limit of the 75-mg dose range for a few hours before the next dose (Fig. 3a), whereas a 30-mg dose after every HD session produced exposures that remained in the target range for the entire dosing interval (Fig. 3b). Mean C_max and C_min values, together with systemic exposure (AUC) values for these two simulations, are shown in Table 3.

FIG 3.

Conditional predictions of oseltamivir carboxylate plasma concentrations after two different oseltamivir regimens: 30 mg after every other session (a) and 30 mg after every session (b). bid, twice daily.

TABLE 3.

Predicted oseltamivir and oseltamivir carboxylate steady-state C_max and C_min for patients in the analysis data set^a

Drug and dosing regimen	Cmax (ng/ml)	Cmin (ng/ml)	AUCss,168 h (h · ng/ml)	AUCss,48 h (h · ng/ml)
Oseltamivir
30 mg after every HD session	12.1 (6.02–24.3)	0.167 (0.15–0.217)	223 (167–230)	63.8 (47.8–65.8)
30 mg after every other HD session	11.9 (5.76–24.1)	0.0484 (0.0432–0.0647)	110 (82.7–114)	31.6 (23.6–32.6)
Oseltamivir carboxylate
30 mg after every HD session	1,170 (850–1,930)	221 (57.6–396)	81,300 (52,800–134,000)	23,200 (15,100–38,300)
30 mg after every other HD session	903 (634–1,540)	42.1 (8.91–102)	39,200 (25,900–65,700)	11,200 (7,410–18,800)

^aAUC_{ss,168 h}, area under the plasma concentration-time curve at steady state to 168 h; AUC_{ss,48 h}, area under the plasma concentration-time curve at steady state to 48 h; C_max, peak plasma concentration; C_min, minimum plasma concentration; HD, hemodialysis.

To identify a suitable regimen for oseltamivir prophylaxis, the individual distributions of OC concentrations were replotted using a once-daily 75-mg regimen as the lower limit of the target range. Model simulations showed that the dosage regimen of 30 mg oseltamivir after every other session was the most suitable one (see Fig. S2 in the supplemental material).

To investigate a possible scenario in which a physician wishes to start treatment for influenza immediately and not wait until after the next planned HD session (referred to here as the “first” HD session), additional model simulations were run to evaluate whether a single initial dose of oseltamivir could be given at various times before the first HD session without exceeding target exposures when the first post-HD-session dose was given. Simulations were based on a regimen of 30 mg after every HD session and showed that an initial oseltamivir dose administered 1 h before the first HD session produced OC concentrations within the target range, whether or not this was followed by a second dose after the first HD session. When the initial oseltamivir dose was administered 6 h before the first HD session, with no post-HD dose, some individual OC concentrations briefly dipped below the target range during the second HD session, but adding a dose after the first session avoided this (see Fig. S3 in the supplemental material); results were similar when the initial dose was given 12 h before the first HD session (data not shown). Thus, administration of a second dose after the first HD session maintains concentrations within the target range, irrespective of when the first presession dose is administered.

Simulated distributions (median and 95% prediction intervals) of the model-predicted OC concentration-time courses for 2,400 simulated subjects (not shown) were consistent with the conditional simulations.

DISCUSSION

We report the development of a population pharmacokinetic model to describe the kinetics of oseltamivir in patients with ESRD who are undergoing HD and the effect that HD has on oseltamivir pharmacokinetics. The model also allowed us to identify suitable dosing regimens for treatment and prophylaxis through the use of simulations. The final model for ESRD patients was more complex than a previously published model for other patient populations (18). Although HD sessions are characterized by sharp decreases in plasma drug concentrations (6), there is a small increase in plasma OC concentrations shortly after the HD session resulting from equilibration of OC within body compartments after the session, as seen in the observed and modeled concentration-time profiles (e.g., see Fig. 1 and 3). A second compartment was included in the model to describe this effect.

Since our model showed that OP was mostly eliminated before the start of HD, it was not possible to deduce from the available data whether HD had any effect on the pharmacokinetics of OP. However, because the renal clearance of OP is negligible, neither ESRD nor HD was expected to affect OP concentrations. As expected, the residual renal clearance of OC during the periods between HD sessions was very low in patients with ESRD but increased almost 40-fold during the sessions. The estimated value of 7.43 liters/h for CL_D/F was similar to that reported for the multiple-dose study (6).

The results of the analysis and model simulations indicate that the previously recommended regimen for ESRD patients on HD, i.e., 30 mg after every other HD session (6), results in the C_min of OC falling below the level where a therapeutic effect can be expected, particularly just before the HD session preceding the next oseltamivir dose (Fig. 3b). The simulations show that a more suitable dosing schedule for the treatment of influenza is 30 mg oseltamivir after every HD session, as the C_min values during each session are adequate to provide a therapeutic effect. At this dosing frequency, OC C_max values are lower than the median value recorded for patients with normal renal function receiving oseltamivir at 450 mg twice daily; this is the highest dose at which tolerable clinical exposures have been demonstrated (17). The simulated AUC_0–48 over the dosing period produced by a 30-mg dose after every HD session (23,200 ng · h/ml) is larger than that produced by the standard 75-mg twice-daily dosage in healthy volunteers (≈12,000 ng · h/ml) and equivalent to that produced by a 150-mg twice-daily dosage shown to be well tolerated (19, 20). Reducing the dose to less than 30 mg after every HD session would risk the minimum concentration falling below the putative target concentration during the course of therapy.

The dosing simulations did show that the regimen recommended for prophylactic use in ESRD patients, i.e., 30 mg oseltamivir after every other HD session (6), produced a C_min equivalent to the steady-state C_min produced by the recommended prophylaxis regimen of 75 mg once daily for 10 days in adults with normal renal function (with the possible exception of brief periods after the end of every second HD session). Hence, this regimen may be suitable for influenza prophylaxis in patients with ESRD on HD. Because starting oseltamivir treatment sooner after illness onset is associated with improved therapeutic outcomes (21), physicians who diagnose influenza in a patient with ESRD may not want to delay starting treatment until the next HD session is complete. The simulation results show that giving an initial single 30-mg dose at any time before a planned HD session provides therapeutic OC exposures quickly without posing any safety risk. The simulations also support the recommendation that a post-HD dose should always be given irrespective of the timing of this initial pre-HD dose; this ensures that therapeutic concentrations remain within the target range and, importantly, does not result in exposures exceeding tolerable levels even if the pre-HD dose is given only 1 h prior to the HD session.

Two recent studies investigated the pharmacokinetics/pharmacodynamics and dose-response relationship of oseltamivir based on data from phase II dose-ranging inoculation studies. Rayner et al. (22) reported a 3-group AUC relationship with time to cessation of viral shedding and resolution of symptoms. Kamal et al. (23) also reported a drug-disease model describing the effect of oseltamivir on viral shedding and showed that the 75-mg twice-daily standard dose in otherwise healthy subjects with influenza was at or near the plateau of the dose-response relationship. Although we bridged the data by using C_min, it is not fully clear which exposure metric (C_min or AUC) is optimal for oseltamivir pharmacokinetic bridging for special populations. A hollow-fiber study by Drusano and colleagues (24) suggests AUC/EC₅₀ to be the pharmacodynamically linked variable. One pragmatic argument is that oseltamivir is generally safe and well tolerated, with a wide safety margin (2). The risk of treatment failure in a high-risk population would exceed the risk of developing any serious adverse effect of oseltamivir. We therefore conservatively aimed to ensure that all exposure measures (particularly AUC and C_min) were at or above the exposure of the standard 75-mg twice-daily regimen which was shown to be effective in phase III studies, provided they also fell within the safety margin (2). In the current analysis, C_min was used as the bridging metric because it was the most sensitive to dosing frequency relative to the hemodialysis schedule.

To develop the current population model for ESRD patients, we pooled pharmacokinetic data from two studies to create a data set from 24 patients—this is equivalent to the normal sample size for a phase I study and is sufficient to study interindividual variability in pharmacokinetics without the need for rigorous investigation of covariate effects. Moreover, each study participant had densely sampled kinetics for OP and OC (>10 samples each) during on- and off-dialysis periods. We conclude that for ESRD patients undergoing HD, the dosing schedule for oseltamivir for the treatment of influenza should be to initiate therapy as soon as diagnosis is made and then to administer a dose of 30 mg after every HD session thereafter. Treatment duration is 5 days, the same as for otherwise healthy adults (25, 26). For influenza prophylaxis, a dose of 30 mg after every other HD session is adequate. The exact duration of prophylaxis should be decided by the prescribing clinician; however, the recommended duration of postexposure prophylaxis is at least 10 days, and during community outbreaks, preexposure prophylaxis may be continued for up to 6 weeks (6, 25, 26). These recommended dosage regimens are reflected in the current product label.