Population pharmacokinetics of intravenously and orally administered docetaxel with or without co-administration of ritonavir in patients with advanced cancer

Abstract

AIM

Docetaxel has a low oral bioavailability due to affinity for P-glycoprotein and cytochrome P450 (CYP) 3A4 enzymes. Inhibition of the CYP3A4 enzymes by ritonavir resulted in increased oral bioavailability. The aim of this study was to develop a population pharmacokinetic (PK) model and to evaluate and quantify the influence of ritonavir on the PK of docetaxel.

METHODS

Data from two clinical trials were included in the data analysis, in which docetaxel (75 mg m⁻² or 100 mg) had been administered intravenously or orally (10 mg or 100 mg) with or without co-administration of oral ritonavir (100 mg). Population modelling was performed using non-linear mixed effects modelling. A three-compartment model was used to describe the i.v. data. PK data after oral administration, with or without co-administration of ritonavir, were incorporated into the model.

RESULTS

Gut bioavailability of docetaxel increased approximately two-fold from 19 to 39% (CV 13%) with ritonavir co-administration. The hepatic extraction ratio and the elimination rate of docetaxel were best described by estimating the intrinsic clearance. Ritonavir was found to inhibit in a concentration dependent manner the intrinsic clearance of docetaxel, which was described by an inhibition constant of 0.028 µg ml⁻¹ (CV 36%). A maximum inhibition of docetaxel clearance of more then 90% was reached.

CONCLUSIONS

A PK model describing both the PK of orally and intravenously administered docetaxel in combination with ritonavir, was successfully developed. Co-administration of ritonavir lead to increased oral absorption and reduced elimination rate of docetaxel.

WHAT IS ALREADY KNOWN ABOUT THIS SUBJECT

• Docetaxel is an approved drug for the treatment of cancer of various primary origins.

• An oral docetaxel regimen is warranted because of patient convenience and the opportunity to investigate more schedule intensive treatment regimens.

• Co-administration of ritonavir significantly enhanced the apparent oral bioavailability of docetaxel.

WHAT THIS STUDY ADDS

• This study demonstrates that ritonavir increased the absorption of docetaxel after oral administration.

• Furthermore, we showed that the clearance of docetaxel was inhibited in a concentration dependent manner.

• The developed model will be used for further development of an oral docetaxel regimen.

Introduction

Docetaxel has significant anti-tumour activity against a range of tumour types and is approved for the treatment of locally advanced or metastatic breast, non-small cell lung (NSCLC), head and neck, gastric and hormone refractory metastatic prostate cancer at doses ranging from 60 to 100 mg m⁻² administered as a 1 h i.v. infusion every 3 weeks. Currently, weekly schedules of docetaxel are increasingly used. The rationale behind weekly administration is that this schedule results in more frequent exposure of docetaxel to tumour cells while lower C_max values are reached [1].

Development of an orally available formulation of docetaxel is of interest because oral administration is preferred over i.v. administration, due to patient convenience, reduced administration costs and the opportunity to investigate more schedule-intensive treatment regimens [2].

The oral bioavailability of docetaxel is, however, limited due to P-glycoprotein (Pgp) and cytochrome P450 (CYP) metabolizing enzymes, mainly CYP3A4 [3]. The development of an oral docetaxel regimen started with the co-administration of cyclosporin A, a potent Pgp inhibitor, resulting in high systemic exposure to docetaxel [4]. However, further development of this combination was terminated because preclinical research showed that inhibition of CYP3A4 was even more effective in enhancing the systemic exposure of docetaxel [3]. Inhibition of CYP3A4 by ritonavir resulted in boosted systemic exposure of oral docetaxel in mice [3], and in cancer patients [5]. Although several studies have suggested that ritonavir may act as an inhibitor of Pgp [6], in vitro studies have shown that ritonavir is not a potent inhibitor of Pgp mediated transport of docetaxel [3]. The enhancement of the systemic exposure of CYP3A4 substrates by ritonavir is already standard practice in the treatment of HIV patients with protease inhibitors [7, 8]. The ritonavir dose used for boosting these agents is 100 mg, which is well below its therapeutic dose of 600 mg twice daily. In general these low doses show only limited side-effects [7, 8].

A proof-of-concept study [5] of oral docetaxel in combination with ritonavir was performed in patients with advanced solid tumours. The apparent bioavailability (ratio of area under the plasma concentration–time curve after oral and i.v. administration) of oral docetaxel (75 mg m⁻²) alone was approximately 14% [4]. The apparent bioavailability of 100 mg oral docetaxel in combination with 100 mg ritonavir was above 100% [5]. Considering that a standard weekly docetaxel dose is 35 mg m⁻²[9–11], systemic exposure to docetaxel needed for an effective weekly docetaxel regimen can be reached with the combination of both drugs. These results were considered promising and formed the basis for further clinical development of this combination.

The pharmacokinetics (PK) of this combination are critical for the further development, either for the evaluation of new formulations as well as for optimization of the design of oral docetaxel/ritonavir regimens (e.g. optimal dose, multiple ritonavir dosing, dosing interval). The PK are, however, not completely understood. The concentration–time curves of docetaxel show non-linear pharmacokinetics in the terminal part of the plasma concentration–time curve, suggesting a time and/or concentration dependent effect of ritonavir on the metabolism of docetaxel.

The primary objective of this study was to evaluate the influence of ritonavir on the absorption and elimination rate of docetaxel due to inhibition of Pgp and CYP3A4. Secondly, a population PK model, using nonlinear mixed effect modelling (NONMEM) was developed to assess simultaneously the PK of orally and intravenously administered docetaxel with or without co-administration of ritonavir. This model can be used for further development of the combination and to support future trials and schedules.

Materials and methods

Patients

Data were obtained from two clinical trials where the inclusion and exclusion criteria were similar [4, 5]. Patients with histological or cytological proof of cancer, for whom no standard of proven therapeutics existed, were included in the study. Eligibility criteria included a performance status ≤2 on the World Health Organization (WHO) scale, life expectancy of ≥3 months, adequate bone marrow (absolute leukocyte count ≥3.0 × 10⁹ l⁻¹, platelets >100 × 10⁹ l⁻¹), hepatic (serum bilirubin <20 µmol l⁻¹, aspartate amino transferase and alanine amino transferase ≤1.5 times the normal upper limit; in the case of liver metastases amino transferase and alanine amino transferase ≤3 times the normal upper limit) and renal function (serum creatinine ≤160 µmol l⁻¹ and/or clearance ≥50 ml min⁻¹), no radiotherapy (palliative limited radiation for pain reduction was allowed) or chemotherapy for at least 3 weeks prior to entry and able and willing to swallow oral medication. Exclusion criteria consisted of active bacterial or viral infections, clinical signs of active brain or leptomeningeal metastases, alcoholism, drug addiction, psychiatric disorders leading to inadequate follow-up, pregnancy, or breast feeding and chronic use of H₂-receptor antagonists or proton pump inhibitors. The studies were approved by the Medical Ethics Committee of the Netherlands Cancer Institute and written informed consent was obtained from all patients prior to study entry.

Drug administration

The i.v. formulation of docetaxel (Taxotere; Rhone-Poulenc Rorer/Aventis, Antony, France) was used for both i.v. and oral administration. This formulation contains 0.5 ml polysorbate 80 per 20 mg docetaxel. Commercially available ritonavir (Norvir; Abbott, Illinois, USA, 100 mg capsules) was used. Oral drugs were taken with 100 ml tap water after an overnight fast. Patients remained fasted until 1.5 h after docetaxel administration. Standard docetaxel pre-treatment consisting of oral dexamethsone 4 mg 1 h before drug administration and 4 mg every 12 h (two times) after drug administration and oral granisetron 1 mg 1 h before drug administration, was given during all cycles.

Study design

Patients in the first study were randomized into two groups. The first group received on day 1 ritonavir 100 mg and 60 min later 10 mg orally administered docetaxel, on day 8 they received 100 mg ritonavir and 10 mg oral docetaxel simultaneously. On day 22, patients received 100 mg i.v. administered docetaxel. Patients continued, if it was considered to be in their best interest, with 3-weekly docetaxel i.v. according to standard practice. The second randomization group followed the same schedule except that day 1 and day 8 were reversed [5].

The low starting dose of docetaxel was selected for safety reasons because preclinical data in mice revealed that co-administration of ritonavir resulted in a 50-fold increase in systemic exposure to docetaxel [3, 5]. After the first PK interim analysis and clinical evaluation, the oral docetaxel dose was increased from 10 mg to 100 mg. In total 22 patients were included in this study.

The second study was a proof of concept study of oral docetaxel plus cyclosporin A [4]. In this study docetaxel was given in combination with cyclosporin A. Data of docetaxel administered simultaneously with cyclosporine A were not included in the current analysis. However, data of three patients receiving oral docetaxel 75 mg m⁻² alone and 14 patients receiving docetaxel i.v. 100 mg m⁻², were included.

PK sample collection, processing and storage procedures

Both studies applied extensive PK sampling during the first 48 h after administration for both docetaxel and ritonavir; predose, 0.25, 0.5, 0.75, 1, 1.25, 1.5, 2, 3, 4, 6, 8, 10, 24, 36, and 48 h [4].

Blood samples were collected in heparinized tubes and centrifuged; plasma was separated and stored at −20°C until analysis. Docetaxel plasma concentrations were measured in the first study with a validated liquid chromatography coupled with tandem mass spectrometry with a lower limit of quantification (LLQ) of 0.25 ng ml⁻¹[12]. In the second study [4], docetaxel concentrations were determined with a validated high-performance liquid chromatography assay (HPLC) with a LLQ of 10 ng ml⁻¹[13]. Plasma concentrations of ritonavir were determined using an isocratic reversed-phase ion-pair, HPLC assay with ultraviolet detection with a LLQ of 50 ng ml⁻¹[14]. The accuracies and precisions of both assays felt within ±15%. The quality control samples for the assays were prepared separately and therefore the results of different assays were considered interchangeable.

Data analyses

Data from the different administration routes were analyzed simultaneously using NONMEM software (version VI; Icon Development Solutions, Ellicot City, Maryland, USA) [15]. Piraña (an interface to NONMEM, and our cluster) was used for run deployment and analysis [16]. The First Order (FO) estimation method with natural logarithmically (Ln) transformed concentration–time data was used to develop the model and the First Order Conditional Estimation (FOCE) method was used in order to estimate the final parameter estimates. Standard errors for all parameters were calculated with the COVARIANCE option in NONMEM. The PK model of docetaxel i.v. has been extensively investigated by Bruno and co-workers [17, 18]. The first step was to fit this three-compartment model to the data of i.v. administered docetaxel.

Subsequently, a model with an additional depot compartment was fitted to data of oral and i.v. administered docetaxel (without co-administration with ritonavir). Different absorption models using a lag-time or a discrete number of transit compartments to mimic a gamma like, asymmetric S-shaped absorption profile, were investigated. This model was extended with the ritonavir pharmacokinetics. Maximum a posteriori Bayesian estimates of the PK parameters of ritonavir [clearance (CL) (l h⁻¹), distribution volume (Vd) (l), and absorption rate constant (Ka) (h⁻¹)] for each individual were calculated using the POSTHOC option of NONMEM and a previously developed population pharmacokinetic model of ritonavir [8]. Ritonavir data were lacking from a few patients; for these patients, the population parameters established with the ritonavir PK model were used [8].

Docetaxel CL and bioavailability (F) were subsequently modelled using a model previously applied by Lu et al. [19] according to a well-stirred liver model designed by Wilkinson et al. [20] This model assumes that the drug (in our case docetaxel) is exclusively metabolized by the liver, after possible loss in the gut:

(1)

(2)

(3)

where F_gut is gut bioavailability (=1 − extraction ratio across gut), F_hep is hepatic bioavailabilty (=1 − extraction ratio across liver (CL_i/(CL_i+ Q)), Q is hepatic blood flow, CL_i is intrinsic clearance, CL is the docetaxel clearance, CL_i,0 is the uninhibited CL_i of docetaxel, [RTV] is the plasma concentration of ritonavir, and K_i is the inhibition constant of ritonavir on docetaxel. This model was originally designed for total blood clearance [20]. Since we have only studied total plasma concentrations, we have assumed that the blood : plasma ratio is 1 and the free fraction is 1 or independent of the investigated concentration range.

The effect of ritonavir on F_gut was modelled in an non ritonavir concentration manner [19].

(4)

where F_doc is the gut bioavailability of docetaxel without booster (this value is estimated based on three patients who received oral docetaxel without ritonavir), θ(x) is the fixed effect to calculate the increase in gut bioavailability when ritonavir is co-administrated, and RTV is an indicator given the value 1 when ritonavir is given and 0 otherwise.

For the NONMEM analyses, subroutine ADVAN6 TRANS1 was used.

Interindividual and interocassional variability for different PK parameters was estimated using an exponential model. Different models were evaluated for their adequacy to estimate the residual variability. Discrimination between models was based on both graphical and statistical methods. The significance of an increase in the goodness-of-fit between hierarchical models was tested using the log-likelihood ratio test [15]. A minimal difference of the objective function value (OFV) of more than 10.83, corresponding to a significance level P < 0.001 was used for discrimination between two hierarchical models differing in one parameter.

Furthermore, the model was evaluated by visual inspection of the goodness-of-fit plots. Xpose (version 4.0), an R based (version 2.9.0) model building aid, was used for the graphical goodness-of-fit analyses [21].

For model evaluation, the shrinkage of eta and epsilon were calculated [22, 23].

The adequacy of the final model was evaluated by a visual predictive check [24]. The data set was simulated 1000 times. The observed data, the median and the 95% confidence area of the predicted median, and the 5^th and 95^th percentile of the prediction interval were plotted using PsN (versison 3.0) and Xpose (version 4.0) [21, 25].

Additionally, plots of two individuals were depicted showing the observed, predicted and individual predicted concentrations.

The final model was used for simulating different ritonavir dosing strategies in combination with 100 mg docetaxel. The combinations investigated were 50, 100, and 200 mg ritonavir administered simultaneously or 1 h prior to docetaxel, and two doses of 100 mg ritonavir administered simultaneously or 1 h prior and 2 h after docetaxel intake. The AUC(0,∞) for these dosing strategies was calculated for 1000 simulated patients using NONMEM. The geometric mean and the 90% confidence intervals were calculated using the log-transformed data. The aim of this simulation study was to estimate the expected mean exposure to docetaxel with different ritonavir schedules, and to estimate whether the exposure to docetaxel can be increased by optimizing the ritonavir schedule.

Results

Thirty-six patients were included in the two studies, 20 males and 16 females with median age 54 years (range 31–73). PK data were assessed from 72 treatment courses, which are specified in Table 1. In total, the data set comprised 1025 docetaxel plasma concentrations and 276 ritonavir plasma concentrations.

Table 1.

Number of patients per treatment

Dose and administration route	Number of patients with doc. data	Number of patients with RTV data
doc. i.v. 100 mg m−2	15
doc. i.v. 100 mg	17
doc. p.o. 75 mg m−2	3
RTV p.o. 100 mg + doc. p.o. 10 mg after 1 h	6
RTV p.o. 100 mg + doc. p.o. 100 mg after 1 h	15	8
RTV p.o. 100 mg + doc. p.o. 10 mg	5
RTV p.o. 100 mg + doc. p.o. 100 mg	11	8

doc., docetaxel; RTV, ritonavir; p.o., oral; i.v., intravenous.

Model development

Figure 1 provides an overview of the developed model. The three-compartment model developed by Bruno and co-workers [17, 18] was fitted to our i.v. data. The observed concentrations were well predicted and the estimated fixed effects were not relevantly different compared with the estimates obtained by Bruno et al. The parameter estimates are given in Table 2.

Figure 1.

Differential equations of the final model. k_tr: transit rate constant (2/(mean absorption time; MAT)), CL: docetaxel clearance, V3: volume of the central compartment, k_xy: rate constant between compartment x and y

Table 2.

Estimations of the basic i.v. model and the final model

			I.v. model		Final model
			Typical value (CV%)	ÌIV (CV%)	Typical value (CV %)	IIV (CV%)	IOV (CV%)
Pharmacokinetic model
CL	Total plasma clearance	(l h−1)	44.1 (6%)	29.4% (15%)
CLi,0	Intrinsic clearance	(l h−1)			113 (19%)	60% (26%)	22% (127%)
MAT	mean absorption time	(h)			1.3 (21%)	87% (35%)	52% (29%)
V2iv	Volume central compartment (i.v.)	(l)	8.9 (9%)	37.8% (14%)	9.8 (10%)	45% (27%)
V2po	Volume central compartment (p.o.)	(l)			44.0 (17%)	35% (27%)
Q1	Intercompartmental CL (C2-C3)	(l h−1)	6.1 (7%)		6.9 (12%)
V3	Volume peripheral compartment	(l)	7.3 (9%)		7.5 (16%)
Q2	Intercompartmental Cl (C3-C4)	(l h−1)	14.4 (12%)	20.3% (20%)	15.7 (14%)
V4	Volume peripheral compartment	(l)	388 (11%)		376 (15%)
F1	Gut bioavailability	(%)			19% (21%)	72% (45%)
FRTV	Gut bioavailability + RTV	(%)			39% (13%)	72% (45%)	44% (35%)
Ki	Inhibition constant: ritonavir – docetaxel	(µg ml−1)			0.028 (36%)	122% (33%)
Q	Hepatic blood flow (fixed)	(l h−1)			80
CL∼V2	Correlation CL∼V2					44.6%
Residual error
P	Proportional error	(%)	23% (8%)		32% (14%)
P	Proportional error first four hours after oral administration	(%)			63%

CV, covariance; IIV, inter individual variability.

A depot compartment was added to the model and fitted to the oral data. Different absorption models were investigated for their significance. In an early model development stage, an absorption lag-time or a transit compartment did not relevantly improve the model. In a final stage, however, it was found that a model with a single transit compartment described the data best.

The next step was to add the ritonavir data to the model. Gut bioavailability was modelled according to equation 4. The bioavailability without and with ritonavir was found to be 19% and 39%, respectively (coefficient of variation (CV) 21% and 13%, respectively). Separate estimates of the bioavailability for sequential and simultaneous administration of docetaxel and ritonavir did not result in a significant model improvement.

The hepatic bioavailability was modelled by calculating the hepatic extraction ratio according to equations 1 and 3. The hepatic blood flow was fixed at 80 l h⁻¹, since estimation of this parameter resulted in a poor precision and a large estimation confidence interval determined by log-likelihood profiling. The estimated uninhibited intrinsic clearance of docetaxel was 119 l h⁻¹ (CV 19%) and the inhibition constant of docetaxel and ritonavir was 0.028 µg ml⁻¹ (CV 36%).

The hepatic bioavailability was incorporated in the differential equation of the first compartment by multiplying it with the input amount of docetaxel and by extracting the remaining fraction.

Elimination

The elimination was modelled using similar equations and parameters as for the hepatic bioavailability. The CL of docetaxel was modelled according to equation 2. The effect of this equation is graphically depicted in Figure 2. This figure shows that the docetaxel clearance decreased almost instantaneously after administration of ritonavir. Thus, ritonavir inhibited CYP3A4, thereby reducing the clearance of docetaxel by approximately 90%, after which it gradually returned to baseline level.

Figure 2.

The estimated clearance of docetaxel vs. time (h) in individual patients receiving 100 mg docetaxel and 100 mg ritonavir (RTV) simultaneously and the average concentration–time curve of 100 mg ritonavir (▴)

Volume of distribution

Separate analyses of the volume of distribution of the central compartment for orally and intravenously administered docetaxel resulted in different estimates of 44 l (CV 17%) and 9.8 l (CV 10%), respectively. It was hypothesized that these differences were caused by the main excipient of the formulation, polysorbate 80. The assumption that the initial volume after i.v. administration increased after elimination of polysorbate 80 (or its micelles) to a volume similar to that for oral administration was evaluated. The data did not support such a shift in this PK parameter. Therefore, different estimates for the distribution volume were implemented in the final PK model.

Error model

Visual inspection of the concentration–time curves showed higher inter-patient variability for orally, compared with intravenously administered docetaxel. This was mainly attributed to large interindividual variability (IIV) in the bioavailability and absorption rate of orally administered docetaxel.

Despite the estimation of the IIV in these two parameters, the residual error remained larger for orally administered docetaxel, especially for the ascending part of the curve. Separate estimates of the proportional error for the first 4 h after oral administration resulted in an improved fit of the model.

Model evaluation

Goodness-of-fit plots from the final model showed that both the population and individual predicted concentrations are equally distributed around the line of identity (figures not included).

The eta-shrinkage of the random effects on the interindividual variability were in the range of 9.2–22.5%, the eta-shrinkage of the random effects on the intra-individual variability were in the range of 22.5–70.8% and the epsilon-shrinkage was 7.5%. The shrinkage of the intra-individual variability was fairly high, but considered adequate for further simulation studies.

The visual predictive check plots (Figure 3A–C) showed that the measured concentrations were well distributed within the 90% confidence interval. Furthermore, the 90% confidence interval was relatively large for orally, and small for i.v. administered docetaxel.

Figure 3.

Visual predictive check (VPC) plots: The grey surface is the 90% confidence area of the predicted median, and the 5^th and 95^th percentile of the prediction interval. The black line is the observed median and the grey dotted lines are the 5^th and 95^th percentile of the observed data. In A) oral docetaxel 100 mg and ritonavir 100 mg are given simultaneously, in B) oral docetaxel 100 mg and ritonavir 100 mg are given 1 h sequentially and in C) docetaxel 100 mg is given intravenously. On the y-axis the concentration of docetaxel in ng ml⁻¹

Model performance was further investigated by inspecting the observed and predicted concentrations of six curves (see Figure 4).

Figure 4.

The observed, predicted (Pred) and individual predicted (Ipred) concentrations vs. time for two individuals (1 and 2) on three occasions: oral docetaxel 100 mg in combination with 100 mg ritonavir (A), oral docetaxel 100 mg given 1 h after 100 mg ritonavir (B) and 100 mg docetaxel given intravenously (C). Observed (); Ipred (); Pred ()

Simulation of different ritonavir regimens

The simulations of different ritonavir treatment strategies showed that the apparent oral bioavailability can be marginally increased by increasing the ritonavir dose, or by giving multiple ritonavir doses. The geometric mean of the AUC(0,∞) and the 90% confidence interval of docetaxel in combination with different RTV treatment regimes are given in Table 3. The highest exposure of 100 mg oral docetaxel was seen when 200 mg ritonavir was administered 1 h prior to docetaxel intake.

Table 3.

Simulation of 1000 patients treated with 100 mg docetaxel in combination with 10 different ritonavir (RTV) treatment regimens. The geometric mean of the AUC(0,∞) (in mg l–1 h) with the 90% confidence interval are given

	Simultaneous with docetaxel	1 h prior to docetaxel
Single dose 50 mg RTV	1.07 (1.00, 1.13)	1.87 (1.77, 1.97)
Single dose 100 mg RTV	1.36 (1.28, 1.45)	2.56 (2.41, 2.71)
Single dose 200 mg RTV	1.71 (1.61, 2.15)	3.32 (3.14, 3.50)
50 mg RTV −1 or 0 h + 2 h post-dose	1.27 (1.19, 1.35)	2.18 (2.06, 2.29)
100 mg RTV −1 or 0 h + 2 h post-dose	1.63 (1.53, 1.74)	2.95 (2.79, 3.12)

Discussion

The amount of docetaxel absorbed increased from 19% without to 39% when co-administrated with ritonavir. In vitro results showed that the Pgp mediated transport is hardly influenced by ritonavir [3]. Thus, the reduced pre-systemic clearance by inhibition of CYP3A4 is probably the major determinant in the enhancement of the absorption of docetaxel. The estimated amount of docetaxel absorbed is much lower than the apparent bioavailability (AUC_oral/AUC_i.v.), which was above 100% [5]. This apparent discrepancy can be explained by reduced elimination due to inhibition of CYP3A4 mediated elimination. The exposure to docetaxel can be increased by both improving the passage through the gastrointestinal barrier by inhibiting CYP3A4 and P-glycoprotein in the enterocytes and furthermore by inhibition of the first pass metabolism by the liver. Additionally, the elimination of docetaxel can be decreased as well by reduced metabolism by CYP3A4 enzymes.

The competitive inhibition of hepatic CYP3A4 was assumed to be ritonavir concentration dependent resulting in non-linear elimination of orally administered docetaxel. The total plasma clearance of docetaxel decreased almost instantaneously and gradually returned back to the initial clearance in parallel with declining concentrations of the inhibitor, ritonavir.

An effect of polysorbate 80 on the distribution of docetaxel was previously suggested by Loos et al. [26]. Due to the high molecular weight of polysorbate 80 (1310 Da) and the formation of large micelle complexes, absorption of polysorbate 80 after oral administration is not to be expected. Together with a fast elimination [27], plasma concentrations of polysorbate 80 will be negligible after oral administration, which is in contrast to i.v. administration. I.v. administered docetaxel is present in the circulation in at least three distinguishable forms, polysorbate-bound, protein-bound and free. Non-linear pharmacokinetics of docetaxel were seen in mice at doses above the therapeutic range [27]. Van Tellingen et al. showed that immediately after infusion of a dose of 100 mg m⁻² of docetaxel, concentrations of polysorbate 80 were found above the critical micelle concentrations (0.009% v : v [28]). However, soon after administration these concentrations dropped below the detection limit (0.01% v : v) of the assay and most likely under the critical micelle concentration [27]. The authors concluded that polysorbate 80 can result in non-linear pharmacokinetics but not at dose levels relevant to the clinical situation. The results of the simultaneous analysis of orally and i.v. administered docetaxel showed, however, that volume of the central compartment is small for i.v. administered docetaxel and much higher after oral administration. This indicates that polysorbate 80 micelles might be the cause of a low distribution volume for i.v. administered docetaxel. Micelle breakdown occurs instantaneously and completely. Thus, an instantaneous increase in volume of distribution for i.v. administered docetaxel can be expected after micelle breakdown. This immediate increase was, however, not identifiable in the analysis, probably because a large volume of distribution at steady state was found for docetaxel.

Docetaxel given as a standard 1 h infusion shows high interindividual variability, which is associated with variability in efficacy and toxicity [29]. This variability is higher for orally administered docetaxel, which is mainly due to a high variability in absorption. This can clearly be seen in the visual predictive check plots (Figure 4). The initial part of the visual predictive check plots show a relatively large 90% confidence interval for oral, in contrast with i.v. administered docetaxel. The main risk of a high variability is under-, and over-dosing of patients. Further research is needed to investigate strategies for reducing the interindividual variability. A possible method could be increasing the ritonavir dose. This is expected to result in a more complete and sustained inhibition of CYP3A4. Within the proof of concept study, patients were given a flat dose of docetaxel. However, when the variability was too high, dose individualization based on one or more covariates proven to have a significant influence on the pharmacokinetics, for instance body surface area, hepatic function, age, α₁-acid glycoprotein [18], C1236T mutation in the ABCB1 gene [30] and CYP3A4*1B polymorphism [31] could be considered. However, it is questionable whether these covariates are of relevance considering the high inter-patient variability caused by the oral route of administration. Another option, on the condition that the inter-occasion variability is low, is to individualize the dose by therapeutic drug monitoring.

Simulations of different combination regimes have shown that a regimen with 200 mg ritonavir dosed 1 h prior to docetaxel intake, results in the highest exposure. These results should be evaluated cautiously. The PK model was developed without prior knowledge concerning the dose effect of ritonavir on the oral bioavailability of docetaxel, and data concerning multiple ritonavir doses were not available. However, these results point out the direction for further clinical development of the combination.

Oral administration of docetaxel in combination with ritonavir has shown to be a promising mode of administration. The PK profile is markedly different compared with intravenously administered docetaxel. The hypothesized differences were quantified in the current analysis. The current model forms a suitable tool for further development of this combination and will support design of further studies and schedules.

In conclusion, co-administration of ritonavir lead to improved oral absorption and a ritonavir concentration dependent inhibition of CYP3A4, which resulted in a reduced elimination rate for docetaxel. A PK model of docetaxel in combination with ritonavir was successfully developed and will be used to establish optimal combination regimens.

Competing interests

None declared.