Evaluation of a Pharmacology-Driven Dosing Algorithm of 3-Weekly Paclitaxel Using Therapeutic Drug Monitoring
- First Online:
- Cite this article as:
- Joerger, M., Kraff, S., Huitema, A.D.R. et al. Clin Pharmacokinet (2012) 51: 607. doi:10.1007/BF03261934
- 238 Views
Background and Objective
Severe neutropenia is the most frequent and important toxicity of 3-weekly paclitaxel and puts patients at substantial risk of infectious complications. It is well known that the time during which paclitaxel plasma concentrations exceed 0.05 μmol/L (TC>0.05) correlates with the extent of neutropenia. This study was initiated to develop a dosing algorithm that would be able to reduce severe neutropenia by targeting an individual paclitaxel TC>0.05 between 26 and 31 hours, and could be validated in a prospective randomized trial by comparing it to conventional dosing of paclitaxel.
Paclitaxel plasma concentration-time (n = 273) and absolute neutrophil count (ANC) data (152 of the 273 patients) were pooled from two previous studies and submitted to population pharmacokinetic and pharmacodynamic modelling using nonlinear mixed-effects modelling software NONMEM® version VII. To fit the data, we used a previously described 3-compartment model with saturable elimination and distribution, coupled to a semiphysiological model with a linear function to describe the myelotoxic effect of paclitaxel (Epaclitaxel) on circulating neutrophils (neutropenia). Patient age, sex, body surface area (BSA), bilirubin and renal function were tested as potential covariates on the maximum elimination capacity of paclitaxel (VMEL). Limited sampling strategies were tested on the pharmacokinetic model for their accuracy to predict paclitaxel TC>0.05. Subsequently, we proposed a first-cycle dosing algorithm that accounted for BSA, patient age and sex, while later cycles accounted for the previous-cycle paclitaxel TC>0.05 (target: 26 to 31 hours) and ANC nadir to adapt the paclitaxel dose for the next treatment cycle. To test the adequacy of the proposed dosing algorithm, we used extensive data simulations on the final pharmacokinetic/pharma-codynamic model, generating datasets of 1000 patients for six subsequent treatment cycles. Grade 4 neutropenia was tested as a potential endpoint for a prospective clinical trial and simulated for two scenarios, i.e. conventional dosing of paclitaxel 200 mg/m2 every 3 weeks, and personalized, pharmacology-driven dosing as outlined above.
Concentration-time data for paclitaxel were adequately described by the 3-compartment model. Also, individual ANC counts were adequately described by the semiphysiological model using a linear function to describe Epaclitaxel on neutropenia. Patient age, sex, bilirubin and BSA were significant and independent covariates on the elimination of paclitaxel. Paclitaxel VMEL was 16% higher in males than in female patients, and a 10-year increase in age led to a 13% decrease in VMEL. A single paclitaxel plasma concentration 24 hours after the start of infusion was adequate to predict paclitaxel TC>0.05 (root squared mean error [RSME] = +0.5%), and the addition of an end-of-infusion sample did not further improve precision (RSME = −0.6%). Data simulations on the final pharmacokinetic/pharmacodynamic model and using the proposed dosing algorithm resulted in a first-cycle paclitaxel dose ranging from 150 to 185 mg/m2 for women and from 165 to 200 mg/m2 for men. Dose adaptations for cycles two to six ranged from −40% to +30%, with a final median paclitaxel dose of 167 mg/m2 (range 76 to 311 mg/m2). When compared with conventional dosing (paclitaxel 200 mg/m2 every 3 weeks), personalized dosing reduced grade 4 neutropenia in cycle one from 15% to 7%, and further to 4% in cycle 2.
This study proposes a pharmacology-driven dosing algorithm of 3-weekly paclitaxel to reduce the incidence of grade 4 neutropenia. A randomized clinical trial comparing this dosing algorithm with conventional BSA-based dosing of paclitaxel in patients with advanced non-small cell lung cancer is currently ongoing.
Paclitaxel is a mitotic inhibitor and one of the most widely used anticancer drugs, approved for the treatment of advanced non-small cell lung cancer (NSCLC), early or advanced breast cancer, and ovarian cancer, and is either given after surgery or after treatment failure in the advanced setting. Paclitaxel exhibits nonlinear pharmacokinetics,[5,6] and population pharmacokinetic models of paclitaxel have usually implemented saturable elimination and distribution of the drug.[7,8] Paclitaxel undergoes hepatic metabolism and biliary excretion, which is why patients with liver function impairment or liver metastases have a slower elimination of the drug and experience more haematological and nonhaematological toxicity.[9–11] The exposure toxicity relationship of paclitaxel has usually been described by threshold models, whereas the time above a paclitaxel plasma concentration of 0.05 μmol/L (TC>0.05) is correlated with haematological toxicity, and the relationship between paclitaxel TC>0.05 and neutropenia is described by a sigmoid maximum effect (Emax) model.[9,12,13] Besides threshold models, semiphysiological models have been used since their introduction by Friberg et al.
The most important and dose-limiting toxicities in patients receiving 3-weekly paclitaxel are severe neutropenia and peripheral neuropathy, the latter being cumulative and potentially resulting in sensory and motor symptoms. In patients with advanced NSCLC receiving 3-weekly paclitaxel 200 mg/m2 in combination with a carboplatin area under the plasma concentration-time curve (AUC) of 6 mg · min/mL (AUC 6), the rate of severe neutropenia ranges between 13% and 39%, with the rate of febrile neutropenia (FN) ranging between 1% and 3%.[17–20] With paclitaxel at a 3-weekly dose of 225 mg/m2, FN was found in 3% of patients, but the rate of FN increased to 5% with the addition of bevacizumab to platinum/paclitaxel chemotherapy. In patients ≥70 years of age receiving paclitaxel 200 mg/m2, carboplatin AUC 6 and bevacizumab, severe neutropenia was found in 34% and FN in 6%. A substantial rate of FN of 12% was found in Japanese patients with advanced NSCLC receiving paclitaxel 225 mg/m2 and carboplatin AUC 6. Severe neutropenia is unpredictable, in part because of substantial inter-individual variability of paclitaxel pharmacokinetics. Importantly, the first-cycle neutrophil nadir is a good predictor for neutropenic complications in later chemotherapy cycles,[25,26] and the use of filgrastim to support bone marrow recovery reduces the risk of FN. Still, the use of neutrophil growth factors results in substantial costs. Therapeutic drug monitoring of paclitaxel may improve safety by reducing the rate of severe neutropenia.[6,28–30] Indeed, there is strong evidence for the relationship between paclitaxel plasma concentrations, neutropenia[5,6,13,28–35] and clinical outcome in patients with ovarian cancer and advanced NSCLC.[28,31] Paclitaxel TC>0.05 was the strongest predictor of neutropenia in patients with advanced solid tumours.[6,28–30,33,35] Although there is a fairly consistent association between paclitaxel TC>0.05 and neutropenia, only one study assessed Bayesian pharmacokinetic-guided dosing of paclitaxel in patients with lung cancer. The present study was initiated to develop a dosing algorithm for paclitaxel based on patient characteristics and paclitaxel plasma concentrations, to accommodate for inter-individual variability in the pharmacokinetics of the drug. The dosing algorithm is currently used in a prospective randomized study to evaluate the feasibility and efficacy of pharmacology-driven dosing of paclitaxel (EudraCT 2010-023688-16).
Patients and Sampling Procedure
Paclitaxel concentration-time data were pooled from two previous clinical studies, including 145 patients with ovarian cancer, 102 patients with NSCLC and 26 patients with various solid tumours, resulting in a total study population of 273 patients.[28,38] Paclitaxel was given at a dose of 175 mg/m2 in 225 patients, and at various doses between 100 and 250 mg/m2 in 48 patients; 40 of the 273 patients received paclitaxel doses <175 mg/m2 or >200 mg/m2. The duration of the paclitaxel infusion was 3 hours in 261 patients and 24 hours in 12 patients. Paclitaxel was given as a single agent in 76 patients, and in combination with carboplatin in 197 patients. For patient characteristics, see table sI in the Supplemental Digital Content (http://links.adisonline.com/CPZ/A34). The samples for paclitaxel analysis were all collected in heparinized tubes. The blood samples for paclitaxel were taken before the start of the infusion, at 1.5 hours after the start, at the end of the infusion, and at 45 minutes and 2, 6, 9, 12 and 24 hours after the end of drug infusion. Whole blood was centrifuged immediately after withdrawal for 5 minutes (3000 rpm, 4°C), and the plasma fraction was stored at −20°C until analysis. Plasma concentrations of paclitaxel were determined by a validated isocratic high-performance liquid chromatographic (HPLC) method with solid-phase extraction as the sample pretreatment procedure, as has been described in detail elsewhere. The quantitation range of the HPLC assay was 10 to 10 000 ng/mL. Neutrophil counts were available in 152 of the 273 patients. In these patients, haematological toxicity was also assessed.
Basic and Covariate Pharmacokinetic Model
Population pharmacokinetic analysis of the concentration-time data of paclitaxel was performed using nonlinear mixed-effects modelling software NONMEM® version VII (UCSF and Globomax ICON, LLC, Hanover, MD, USA). NONMEM® uses a maximum likelihood criterion to simultaneously estimate population values of fixed-effects (e.g. drug clearance) and random-effects parameters (e.g. inter-individual, inter-occasion and residual variability). Standard errors were calculated for all parameters, and individual Bayesian pharmacokinetic parameters were obtained using the POSTHOC function. A previously described 3-compartment population model with saturable elimination and saturable distribution to peripheral tissues[28,38] was fitted to pooled paclitaxel plasma concentration-time data (figure 1). The following pharmacokinetic parameters were estimated: Volume of the central (V1; in L) and second peripheral (V2; in L) compartment, paclitaxel maximum elimination rate (VMEL; in μmol/h), Michaelis-Menten elimination constant (KmEL; in μmol/L), maximum transport capacity from the central to the first peripheral compartment (VMTR; in μmol/h), total plasma concentration of paclitaxel at half VMTR (KmTR; in μmol/L) and intercompartmental clearance between the central and second peripheral compartment (Q; in L/h). Inter-individual variability was estimated using a proportional error model (table I). Subsequently, the following covariates were tested on paclitaxel VMEL: body surface area (BSA), patient age, sex, creatinine clearance, total bilirubin and disease (lung cancer, ovarian cancer, other). Model fit was evaluated by comparing the difference in the objective function value (OFV) of hierarchical models, standard error of parameters, bootstrap analysis and the visual predictive check. The different covariates were sequentially added to the basic model (forward inclusion), and the significance level was set at p < 0.01, which corresponds to a decrease of OFV of >6.7. Significant covariates were included into an intermediate model followed by a stepwise backward elimination procedure. Covariates remained in the model when elimination of the covariate resulted in an increase of OFV of >7.9 (p < 0.005). Individual paclitaxel TC>0.05 was derived using Bayesian estimation on the final model.
Limited Paclitaxel Sampling Strategy
Limited sampling strategies were evaluated for their accuracy to predict individual paclitaxel TC>0.05. For this purpose, data simulations based on the final model were used to generate datasets of 1000 patients with the following specifications: paclitaxel 200 mg/m2 over a 3-hour infusion, random BSA (normal distribution, median 1.8 m2), age (normal distribution, median 56 years), sex (males : females 7 : 3) and serum bilirubin (normal distribution, median 7 μmol/L). In a first run, extensive blood sampling was simulated (at 1.5, 3, 3.75, 5, 9, 12, 15 and 24 hours after the start of a 3-hour infusion), and the generated individual TC>0.05 values were used as reference (regimen 1). Subsequently, the adequacy of Bayesian estimation using the POSTHOC function in NONMEM® to predict individual paclitaxel TC>0.05 from extensive pharmacokinetic sampling was evaluated when applying seven limited sampling strategies, based on the mean prediction error and the root squared mean error (RSME) [table II, regimens 2–8]. The seven limited sampling strategies tested were as follows: single 24-hour sample (regimen 2), 3- and 24-hour samples (regimen 3), 3.75- and 24-hour samples (regimen 4), 3, 3.75- and 24-hour samples (regimen 5), a single 12-hour sample (regimen 6), a single 5-hour sample (regimen 7) and a single 3-hour sample (regimen 8).
Semiphysiological Modelling of Neutropenia
A semi-physiological model was used to link paclitaxel pharmacokinetics with neutropenia. The model includes a compartment representing proliferating cells linked to a compartment representing the systemic circulation through three transit compartments, mimicking precursor cell maturation within the bone marrow. The transition rate constant (ktr) between the compartments was supposed to be first-order and equal for all transitions. The chain of transit compartments allows the description of the time delay between drug exposure, impaired cell proliferation or cell killing, and the resulting effect on circulating neutrophils. The average maturation time or mean transition time (MTT) represented the time a cell took to pass from the proliferation stage to the circulation pool. The status of the proliferation compartment is dependent on the number of cells in that compartment and on the proliferation rate constant (kprol). The disappearance of peripheral neutrophils from the circulation pool is given by the first-order rate constant kcirc. At steady-state conditions, both kprol and kcirc equal ktr. A feedback mechanism imitated the effect of the release of endogenous growth factors as a response to the decrease of cells in the circulation pool. This leads to increased cell proliferation and was modelled by a power function of the ratio between the baseline absolute neutrophil cell count (ANCbase) and the cell count at time t (ANCt) according to (ANCbase/ANCt)γ, where γ constitutes the feedback constant. A linear function was used to model the myelotoxic effect of paclitaxel (Epaclitaxel) on the proliferation rate (kprol) of circulating neutrophils according to equation 1:
where FB represents the feedback parameter (ANCbase/ANCt)γ and C represents the plasma concentration. The ‘Slope’ parameter represents individual sensitivity of the bone marrow to paclitaxel. Logarithmic transformation of neutrophil counts was used throughout model-building together with the first-order conditional estimation method. Individual ANC curves and nadir were derived using Bayesian estimates.
Paclitaxel Dosing Algorithm
We evaluated a pharmacology-driven dosing algorithm of paclitaxel, in which the first-cycle dose accounted for individual BSA, patient age and sex. For cycles two to six, previous-cycle paclitaxel TC>0.05 (target: 26 to ≤31 hours) and ANC nadir were used to adapt the paclitaxel dose for the next cycle. Target paclitaxel TC>0.05 (26 to ≤31 hours) was defined based on previous data.[6,28–30] To test the adequacy of the proposed dosing algorithm, we used extensive data simulations on the final pharmacokinetic/pharmacodynamic model, generating datasets of 1000 patients for six subsequent treatment cycles. We simulated quantitative neutrophil toxicity and evaluated grade 4 neutropenia as a potential endpoint for a prospective clinical trial. For this purpose, grade 4 neutropenia was simulated for two scenarios, i.e. conventional dosing of paclitaxel 200 mg/m2 every 3 weeks, and dosing according to the proposed algorithm as outlined above and in figure 2. Besides grade 4 neutropenia, the following parameters were additionally calculated for subsequent treatment cycles to assess the accuracy of the proposed dosing algorithm: frequency of grade 3 and 4 neutropenia, paclitaxel TC>0.05, median ANC nadir and administered paclitaxel dose. In addition to inter-individual and residual (unidentified) variability for paclitaxel pharmacokinetics and ANC, simulations also accounted for an inter-occasion variability of 15.2% for paclitaxel pharmacokinetics, as described previously.
Basic and Covariate Pharmacokinetic Model
Concentration-time data for paclitaxel were adequately described by the 3-compartment pharmacokinetic model, with separate compartments for the two metabolites 6-hydroxy-paclitaxel and 3-hydroxy-paclitaxel. Paclitaxel VMEL was 35.8 μmol/h, with an inter-individual variability of 37% and a residual unidentified variability of 18% (table I). Patient sex, age, serum bilirubin (BILI) and BSA were significant and independent covariates on VMEL, but the type of disease was not. This resulted in equation 2:
In equation 2, SEX is 0 for females and 1 for males. According to equation 2, paclitaxel maximum elimination capacity was 16% higher in males than in female patients, and a 10-year increase of age led to a 13% decrease of VMEL. The validity of the covariate model was supported by the standard errors of the main pharmacokinetic parameters (table I), the visual predictive check (figure 3) and bootstrap analysis (table I). With regards to the visual predictive plot, the 95% confidence interval of model predictions includes the observed data with few outliers (figure 3). Median paclitaxel TC>0.05 was 24.8 hours for a paclitaxel dose of 175 mg/m2.
Limited Paclitaxel Sampling Regimen
A single paclitaxel plasma concentration 24 hours after the start of infusion was adequate to predict observed paclitaxel TC>0.05 (RSME = +0.5%), and also adequately predicted paclitaxel AUC (RSME = +0.8%). The addition of an end-of-infusion sample did not further improve precision (RSME = −0.6%) [table II, regimen 3]. Single paclitaxel sampling times at 12 and 5 hours and at the end of paclitaxel infusion were not adequate to predict observed paclitaxel TC>0.05, as the precision markedly decreased and bias increased (table II, regimens 6–8).
Semiphysiological Modelling of Neutropenia
The semiphysiological model with a linear function to describe Epaclitaxel could adequately be fitted to the individual ANC counts. MTT for neutrophils was 141 hours (5.9 days), with an inter-individual variability of 27%. The slope parameter was 2.6 μmol/L, with an inter-individual variability of 45%, the coefficient of variation for neutrophils was 31.6%, and γ as a component of the FB (equation 1) was 0.2 (residual error 8%). With conventional dosing of paclitaxel at 200 mg/m2, grade 4 neutropenia was found in 15.3% of treatment cycles, and grade 3 and 4 neutropenia in 40.3% (table III). The average population ANC nadir was found to be 1100/μL at 11.5 days, and the median ANC on day 14 was found to be 1200/μL. Goodness-of-fit plots between model-predicted and observed cell counts for neutrophils support the accuracy of the model (plots not shown). The model was successfully validated by the posterior predictive check, as the median day 14 ANC fell into the 50% confidence interval of 1000 simulation runs (1100–1300/μL neutrophils).
Paclitaxel Dosing Algorithm
Data simulations on the final pharmacokinetic/pharmacodynamic model and using the proposed dosing algorithm (figure 2) resulted in a first-cycle paclitaxel dose ranging from 150 to 185 mg/m2 for women and from 165 to 200 mg/m2 for men. Dose adaptations for cycles two to six ranged from −40% to +30%. When compared with conventional dosing of paclitaxel (200 mg/m2 every 3 weeks), pharmacology-driven dosing reduced grade 4 neutropenia in cycle one from 15% to 7%, and further to 4% in cycle 2 (table III). At the same time, the average predicted ANC nadir increased from 1200/μL with conventional dosing to 1500/μL in cycle one and to 1400/μL in cycle two with the proposed dosing algorithm. Over the cycles, paclitaxel dose adaptations resulted in an increased range of paclitaxel dose per m2 (figure 4a), less variability of paclitaxel TC>0.05 (figure 4b), less variability and a moderate increase of the ANC nadir (figure 4c). The median paclitaxel dose decreased from 200 mg/m2 with conventional dosing to 185 mg/m2 with pharmacology-driven dosing in cycle one and further to 167 mg/m2 in cycle six (range 76 to 311 mg/m2). Twenty-one percent of all patients received a final paclitaxel dose >200 mg/m2 (figure 5). The proposed paclitaxel dosing algorithm allowed for dose reductions in all patients experiencing grade 4 neutropenia, and for dose escalation in patients with no severe neutropenia (Common Toxicity Criteria [CTC] grades 3 and 4) and a paclitaxel TC>0.05 <26 hours.
We used a large dataset of paclitaxel concentration-time data in patients with various solid tumours receiving 3-weekly paclitaxel predominantly in combination with carboplatin to assess the potential efficacy of a pharmacology-driven dosing algorithm to improve the benefit-risk ratio of this frequently used chemotherapy doublet. The 3-compartment pharmacokinetic model used to describe paclitaxel concentration-time data supported the concept of both nonlinear distribution of the drug to the peripheral compartment and nonlinear elimination. The maximum elimination capacity of paclitaxel as found in the present study (36 μmol/h) is in rather good accordance with our two previous publications[28,38] (37 μmol/h and 29 μmol/h) and the study by de Jonge et al. (47 μmol/h). The present model is still empirical, and a more mechanism-based population model using plasma and blood concentrations of total and unbound paclitaxel, as well as Cremophor EL concentrations, has been described. This model by Henningsson et al. supported the hypothesis that Cremophor EL is mainly responsible for the nonlinear pharmacokinetics of paclitaxel, but it comes at the cost of additional bioanalysis in plasma and blood, and increased complexity of the model. In the present study, we found an MTT for neutrophils of 5.9 days, similar to the previous analysis in ovarian cancer patients and the study by Friberg et al. (5.3 days). Accordingly, the neutrophil nadir for the 3-weekly paclitaxel schedule varies between 11.5 days in the present study, 11.7 days in the study by Minami et al. and roughly 12 days in a multiple-pool cell lifespan model for neutropenia following Cremophor-based and Cremophor-free formulations of paclitaxel.
The proposed dosing algorithm for 3-weekly paclitaxel used first-cycle dose adjustments for BSA, patient sex and age, and added paclitaxel TC>0.05 and neutropenia for dose adjustments in cycles two to six. When using extensive data simulations to assess the potential use of the proposed dosing algorithm, pharmacology-driven dosing of paclitaxel resulted in a marked reduction of the incidence of grade 4 neutropenia from 15% with conventional dosing to 7% in cycle one and to 4% in cycle two. Neutrophil toxicity in the present study was in close agreement with data in the literature. For example, Sandler et al. found grade 4 neutropenia in 17% of patients with advanced NSCLC receiving 3-weekly paclitaxel and carboplatin, whereas our model found grade 4 neutropenia in 15% of patients. This supports the external validity of the model used. Importantly, pharmacokinetic/pharmacodynamic modelling accounted for inter-individual, intra-individual and inter-occasion variability, and results of grade 4 neutropenia with individualized dosing were compared with conventional dosing of paclitaxel at 200 mg/m2, as used in large clinical studies in patients with advanced NSCLC.[21,22,43,44]
The proportion of FN was not simulated, as FN is dependent on ANC nadir, non-haematological toxicity such as mucositis, patient predisposition and unknown covariates, and cannot be predicted with adequate precision even in large datasets. However, there is good evidence showing the association between the patient’s early haematological response (neutropenia) and the risk for later neutropenic complications. For example, clinical studies have shown the predictive value of the first-cycle leukocyte nadir for predicting neutropenic complications in later cycles.[25,26] Importantly, the greatest risk for FN is in the earliest cycles of chemotherapy. For example, in elderly non-Hodgkin’s lymphoma patients receiving CHOP chemotherapy (cyclophosphamide, doxorubicin, vincristine, prednisone), 63% of toxic deaths (mostly neutropenia-related) occurred in the first treatment cycle. According to a population-based patterns-of-care study, 65% of the hospitalizations due to FN occurred in the first two cycles of chemotherapy in lymphoma patients. Similarly, 75% of the episodes of FN in patients with advanced breast carcinoma receiving chemotherapy with docetaxel and doxorubicin occurred in the first treatment cycle. These data strongly support the significance of avoiding severe neutropenia in the first two treatment cycles to reduce the risk for FN. FN is a substantial burden on cancer patients receiving chemotherapy, affecting healthcare resources and driving costs in lung cancer. When 3-weekly paclitaxel 200 mg/m2 is given concurrently with carboplatin AUC 6, the rate of FN was 2% in the study of Sandler and colleagues, increasing to 5.2% when given concurrently with bevacizumab. Without the use of supportive granulocyte colony-stimulating factors, the rate of FN was 10% in chemotherapy-naive patients receiving single-agent, three-weekly paclitaxel 200 mg/m2 for advanced NSCLC.
The present study builds upon existing evidence for using Bayesian pharmacokinetic-guided dosing of paclitaxel to attain prespecified plasma concentrations of the drug.[36,49] Woo and colleagues used a Bayesian approach in seven children with refractory leukaemias receiving single-agent paclitaxel 315 mg/m2 over 24 hours. Elimination of paclitaxel was estimated after the first 8 hours of infusion, and the infusion rate was adjusted 12 hours after the start of infusion to achieve a target AUC between 31.5 and 45 μmol · h/L. In fact, target AUC was achieved in five of seven children, and personalized dosing reduced inter-individual variability of paclitaxel AUC. In another clinical study, 40 tumour patients were treated with either irinotecan at a conventional dose of 350 mg/m2 or doses based on an equation consisting of midazolam clearance, γ-glutamyl-transferase and height. Individualized dosing of irinotecan resulted in a substantial reduction of severe neutropenia from 45% to 10%, and a total avoidance of FN. In the study by van der Bol et al., cytochrome P450 3A4 phenotyping using midazolam as a probe drug was performed 1 week before the start of irinotecan treatment, and midazolam clearance was used to guide the irinotecan dose. In a further study in 25 patients with stage IIIB or IV NSCLC, de Jonge and colleagues used a Bayesian approach to adapt 3-weekly paclitaxel 175 mg/m2 to achieve a paclitaxel TC>0.1 ≥15 hours. In this study, the paclitaxel dose of subsequent treatment courses was increased to the lowest dose for which the predicted paclitaxel TC>0.1 was ≥15 hours. The authors showed that it was possible to decrease the proportion of patients with paclitaxel TC>0.1 <15 hours from 36% in the first (unadapted) cycle to 11% in the fifth and sixth treatment courses. Dose increases ranged from 5 to 65 mg/m2 in 29 of the 67 personalized treatment courses. This was achieved without any increase in paclitaxel-related adverse events. The study of de Jonge and colleagues is based on the results of an earlier dose-escalation trial showing improved clinical outcome in chemotherapy-naive patients with advanced NSCLC if they had a paclitaxel TC<0.1 ≥15 hours. In more detail, median overall survival was 4.8 months in patients with paclitaxel TC>0.1 <15 hours compared with 8.2 months in patients with paclitaxel TC>0.1 ≥15 hours. Similarly, in a study in ovarian cancer patients receiving postoperative paclitaxel 175 mg/m2 and carboplatin AUC 5 mg · min/mL, patients with paclitaxel TC>0.05 ≥61.4 hours (mean value) had a longer time to disease progression than patients with paclitaxel TC>0.05 <61.4 hours (89 vs 61.9 weeks). Indeed, paclitaxel TC>0.05 or TC>0.1 (time during which paclitaxel plasma concentrations exceed 0.1 μmol/L) were repeatedly shown to be the most relevant predictors for haematological toxicity,[28,34,51] which is why this is the most promising pharmacological parameter for personalized dosing of paclitaxel.
The strengths of this study include the availability of a population model that had been validated in various patient populations,[10,28,38] the simulation of large groups of patients (n = 1000) and the integration of pharmacokinetic and pharmacodynamic aspects by applying the semiphysiological model, as defined above. The study is limited by its retrospective character and the fact that time-varying covariates such as co-medication and performance status of the patients were not taken into account, which may not reflect clinical reality. Furthermore, weekly administration of paclitaxel has recently gained popularity, especially in breast cancer, where it has proven to be superior to 3-weekly administration of paclitaxel.[2,52,53] However, weekly paclitaxel administration is not superior to 3-weekly regimens in patients with advanced NSCLC.
This study proposes a pharmacology-driven dosing algorithm to reduce the incidence of grade 4 neutropenia after the administration of 3-weekly paclitaxel. A randomized clinical trial comparing this dosing algorithm with conventional BSA-based dosing of paclitaxel in patients with advanced NSCLC is currently ongoing.
Dr Feiss is an employee of Saladox Biomedical, Inc. No sources of funding were used to conduct this study or prepare this manuscript. The authors have no conflicts of interest that are directly relevant to the content of this study.