Switching from body surface area-based to fixed dosing for the investigational proteasome inhibitor ixazomib: a population pharmacokinetic analysis

Abstract

Aims

This population pharmacokinetic analysis of the investigational oral proteasome inhibitor ixazomib assessed the feasibility of switching from body surface area (BSA)-based to fixed dosing, and the impact of baseline covariates on ixazomib pharmacokinetics.

Methods

Data were pooled from 226 adult patients with multiple myeloma, lymphoma or solid tumours in four phase 1 studies, in which ixazomib dosing (oral/intravenous, once/twice weekly) was based on BSA. Population pharmacokinetic modelling was undertaken using nonmem version 7.2.

Results

Ixazomib pharmacokinetics were well described by a three compartment model with first order absorption and linear elimination. Ixazomib was absorbed rapidly (K_a 0.5 h⁻¹), with dose- and time-independent pharmacokinetics. Estimated absolute bioavailability and clearance were 60% and 2 l h⁻¹, respectively. Although a small effect of BSA (range 1.3–2.6 m²) was observed on the peripheral volume of distribution (V₄), reducing the corresponding inter-individual variability by 12.9%, there was no relationship between BSA and ixazomib clearance (the parameter that dictates total systemic exposure following fixed dosing). Consistently, based on simulations (n = 1000), median AUCs (including interquartile range) were similar after BSA-based (2.23 mg m⁻²) and fixed (4 mg) oral dosing with no trend in simulated AUC vs. BSA for fixed dosing (P = 0.42). No other covariates, including creatinine clearance (22–213.7 ml min⁻¹) and age (23–86 years), influenced ixazomib pharmacokinetics.

Conclusions

This analysis supports a switch from BSA-based to fixed dosing, without dose modification for mild/moderate renal impairment or age, in future adult studies of ixazomib, simplifying dosing guidance and clinical development.

What is already known about this subject

• The development and clinical administration of many oncology drugs involves body surface area (BSA)-based dosing.

• The investigational oral proteasome inhibitor ixazomib has been investigated as a single agent for the treatment of multiple myeloma using BSA-based dosing.

• However, recent analyses have suggested that fixed dosing may, in many cases, be both more appropriate and more convenient.

What this study adds

• This population pharmacokinetic analysis showed that ixazomib pharmacokinetics are not impacted by BSA, creatinine clearance or age.

• This analysis was pivotal in transitioning the ixazomib clinical development programme from BSA-scaled to fixed dosing and illustrates the value of modelling and simulation to influence posology decisions in oncology drug development.

Introduction

The feasibility of proteasome inhibition as a therapeutic approach in cancer has been demonstrated with the first-in-class proteasome inhibitor bortezomib and, more recently, the second generation agent carfilzomib 1. Bortezomib is indicated as an intravenous (i.v.) or subcutaneous injection for the treatment of patients with multiple myeloma (MM) 2 based on its clinical activity in both previously untreated patients and those with relapsed disease 3–6. The drug is also approved for use in patients with relapsed mantle cell lymphoma 2, and has demonstrated efficacy in other haematological malignancies, such as follicular lymphoma and systemic amyloid light-chain (AL) amyloidosis 7,8. Like bortezomib, the second generation proteasome inhibitor carfilzomib is approved as an i.v. injection for the treatment of patients with relapsed/refractory MM 9 based on a favourable response rate reported in a pivotal phase 2 trial 10.

Novel proteasome inhibitors are currently being developed with the aim of improving dosing convenience, including through oral administration, and improving efficacy and safety, as compared with existing agents in the same class 1,11. Among these new drugs is ixazomib, the first oral proteasome inhibitor to enter clinical investigations for the treatment of haematological and non-haematological malignancies. Ixazomib (MLN2238) refers to the biologically active boronic acid form of ixazomib citrate (MLN9708). The drug substance is administered as a stable citrate ester, designated as ixazomib citrate. Under physiological conditions ixazomib citrate undergoes rapid hydrolysis to the biologically active boronic acid, ixazomib 12. While ixazomib shows comparable selectivity and potency to bortezomib, preclinical studies have suggested faster dissociation from the 20S proteasome and greater tissue penetration 12,13. Metabolism by multiple cytochrome P450 (CYP) enzymes, including CYP3A4, is expected to be the primary clearance mechanism for ixazomib, while renal elimination is a minor clearance pathway 14. An ongoing phase 1 study of ixazomib in patients with advanced solid tumours or lymphoma will provide a quantitative characterization of the mass balance, metabolism pathways and routes of excretion for ixazomib (NCT01953783). To date, ixazomib has been investigated as a single agent using body surface area (BSA)-based dosing in four phase 1 dose escalation studies, all of which have included collection of pharmacokinetic (PK) data, involving adult patients with MM (C16003 and C16004; NCT00932698 and NCT00963820), lymphoma (C16002; NCT00893464) or solid tumours (C16001; NCT00830869) 14–17. In a pooled analysis of PK data from these trials, plasma exposure of ixazomib increased proportionally with increasing oral dose over the dosing range of 0.48 to 3.95 mg m⁻² in patients with MM 18. A dose proportional increase in exposure was also observed with i.v. dosing over the dosing range of 0.5 to 3.11 mg m⁻². In all four of these trials, ixazomib was dosed according to BSA. This reflects the body size-based dosing paradigm that was introduced more than half a century ago and remains commonplace in clinical oncology practice 19. However, recent scientific analysis has questioned the use of BSA-based dosing and suggested that fixed dosing may, in many cases, be more appropriate 19–21. Specifically, there appears to be little evidence to support a reduction in inter-individual variation in exposure with BSA-based dosing among adults receiving anticancer drugs 19–21.

In this paper, we report a population PK model-based analysis of patients enrolled in the four ixazomib phase 1 studies that used BSA-based dosing, the aim of which was to assess the feasibility of switching to fixed dosing for future clinical development. Additional objectives were to estimate the absolute bioavailability of ixazomib and to evaluate whether patient characteristics, such as age, gender and renal impairment, were likely to impact dosing in future trials. Renal dysfunction (reflected in a reduced creatinine clearance), in particular, is common among cancer patients, especially elderly patients, and can have a marked influence on the PK parameters of certain drugs and/or their active metabolite(s) 22–24, necessitating posology adjustments 25. This complication is especially common among patients with MM, with 30–40% of symptomatic patients having a serum creatinine concentration above the upper limit of normal at diagnosis and about 20% having a level exceeding 2 mg dl⁻¹ 26. Renal impairment has also been shown to develop or evolve over the course of the disease in up to 50% of MM patients 26,27.

Methods

Data collection

For the analysis, PK data were obtained for adult patients with advanced haematological or non-haematological malignancies (MM, lymphoma, or various solid tumours) who had participated in one of four open label, phase 1 dose-escalation studies of i.v. or oral ixazomib (Table 1) 14–17. For studies with oral dosing, ixazomib was available in capsule strengths of 0.2, 0.5, and 2 mg. The BSA-based administered dose was rounded up or down according to the available capsule strengths. In these studies, plasma concentrations of ixazomib were measured pre-dose and at various time points post-dose during cycle 1, following the first dose and after repeat dosing, using a validated liquid chromatography-tandem mass spectrometry method with a lower limit of quantification (LLOQ) of 0.5 ng ml⁻¹, an intra-run accuracy (% bias) of −9.8 to −3.5% and a precision (% CV) of 2.2–8.8%; <0.5% of the observations were below the LLOQ and were ignored. All patients with dosing, sampling time, and pre- and post-dose ixazomib plasma concentration data were included in the analysis.

Table 1.

Details of the studies included in the population pharmacokinetic analysis

Study	Ixazomib dose	Dose frequency	Route of administration	Dosing schedule	Cycle length	Type of malignancy	Number of patients included in present analysis
C16001 [17]	0.125–2.34 mg m−2	Twice weekly	Intravenous	Days 1, 4, 8 and 11	21 days	Solid tumours	88
C16002 [14]	0.125–3.11 mg m−2	Weekly	Intravenous	Days 1, 8 and 15	28 days	Relapsed/refractory lymphoma	30
C16003 [15]	0.24–2.23 mg m−2	Twice weekly	Oral	Days 1, 4, 8 and 11	21 days	Relapsed/refractory multiple myeloma	53
C16004 [16]	0.24–3.95 mg m−2	Weekly	Oral	Days 1, 8, and 15	28 days	Relapsed/refractory multiple myeloma	55

Patients in the four phase 1 trials provided written informed consent. Review boards at all participating centres approved the study protocols and protocol amendments, and the trials were conducted according to the stipulations set out in the Declaration of Helsinki and International Conference on Harmonization Guidelines for Good Clinical Practice.

Population pharmacokinetic modelling

Population PK models were built via non-linear, mixed effects modelling (Appendix 1) using nonmem® software version 7.2 (ICON Development Solutions, Dublin, Ireland) compiled with the Intel® Fortran 9.2 compiler 28. Compartmental PK models were coded using the ADVAN12 subroutine of nonmem and nonmem output was handled using TIBCO Spotfire S +® version 8.2 (TIBCO Software, Inc., Palo Alto, CA, USA). The first order conditional estimation method, with eta–epsilon interaction between inter- and intra-individual variability (FOCE INTER), was used to estimate PK parameters. Models were fitted to the log transformed i.v. and oral data simultaneously, and shrinkage on both eta and epsilon was reported 29.

Visual inspection of the plasma concentration–time profile for ixazomib and the objective function value (OFV) was calculated using likelihood ratio tests (P = 0.05) and used to investigate the base model. Combined oral and i.v. one, two, and three compartment models were tested, assuming first order absorption and linear elimination for all three model types. An exponential error model was used to describe the inter-individual variability for the PK parameters. The residual variability was modelled as an additive error model. Additive error on log transformed data corresponds to a proportional or exponential error on a normal scale. Inter-individual variability was tested on all PK parameters for which estimation of variability was supported by the data. Consideration was given to fitting a full block omega structure on the base model, followed by inspection of the correlations among the inter-individual variability to guide the development of a parsimonious omega structure.

Inter-occasion variability was also tested using two occasions. The first was defined as ≤168 h and second was for time >168 h 30. The three compartment model had the lowest OFV (459) compared with the one and two compartment models, based on the likelihood ratio test (P = 0.05). In addition, visual inspection of the plasma concentration–time log-linear profiles indicated that ixazomib has tri-phasic disposition. Based on these findings, the three compartment base model was selected.

Covariates (age, gender, race, weight, BSA, route, dosing regimen, creatinine clearance [estimated using the Cockcroft & Gault equation with total body weight 31], alanine aminotransferase [ALT], aspartate aminotransferase [AST], albumin, and bilirubin) were tested according to physiological relevance. Covariate selection for the final model was guided using likelihood ratio tests at the following significance levels: P = 0.05 for forward addition of covariates (ΔOFV = 3.84) and P = 0.01 for backward elimination of covariates (ΔOFV = 6.63). In each case, diagnostic plots and comparisons of changes in the minimum OFV between nested models were used to evaluate the covariates. The effects of a continuous covariate on a parameter were represented as a power function referenced to the median value in the population: , where TVP_i is the typical value of a PK parameter (P) for an individual i with a COV_i value of the covariate, θ₁ is the typical value for an individual with a reference covariate value of COV_ref and θ₂ is the exponent of the power function. In contrast, the effects of a categorical covariate on a parameter were represented through a binary relationship: TVP_i = θ₁(1 + θ_2*INDi), where θ₁ is the typical parameter value for an individual in the absence of the covariate (IND_i = 0) and θ₂ is the fractional change in the typical value for an individual if the covariate is present (IND_i = 1).

During development of the model, covariates were tested based on physiological relevance. Age, ALT, AST, albumin, bilirubin, weight, BSA, creatinine clearance, race, route, dosing regimen and gender, were tested on clearance (CL), and age, weight, BSA and gender were tested on central volume of distribution (V₂) and peripheral volume of distribution in compartment 3 (V₄). Final model selection was based on evaluation of goodness-of-fit plots, successful convergence and plausibility of parameter estimates, and the OFVs calculated using likelihood ratio tests. Shrinkage percent was calculated for each inter-individual variability 32 and for residual error. The adequacy of the final model was assessed using visual predictive check 33.

The final population PK model was used to simulate the ixazomib concentration–time profiles and exposure after BSA-based dosing (2.23 mg m⁻²) and fixed dosing (1.86 m² [mean patient BSA] × 2.23 mg m⁻² = ∼4 mg). These doses reflect the recommended weekly dose for ongoing studies of ixazomib. The virtual patients (n = 1000) in both groups had similar mean BSA of 1.86 m² (range 1.3–2.6 m²) based on the demographics of 2208 MM patients included in previous bortezomib clinical trials 4,5,34–36. Area under the plasma concentration–time curve (AUC)(0,∞) was calculated for both BSA-based and fixed dosing using the ratio of dose and simulated clearance values.

Results

PK data were available for 226 adult patients with haematological or non-haematological malignancies. These patients included: 88 with advanced solid tumours from the C16001 study (i.v. ixazomib 0.125–2.34 mg m⁻² twice weekly) 17, 30 with relapsed/refractory lymphoma from the C16002 study (i.v. ixazomib 0.125–3.11 mg m⁻² weekly) 14, 53 with relapsed/refractory MM from the C16003 study (oral ixazomib 0.24–2.23 mg m⁻² twice weekly) 15 and 55 with relapsed/refractory MM from the C16004 study (oral ixazomib 0.24–3.95 mg m⁻² weekly) 16. The demographics and baseline characteristics of patients included in the analysis corresponding to the investigated covariates are summarized in Table 2. Median (range) age and BSA among all 226 patients was 62.0 (23.0–86.0) years and 1.9 (1.3–2.6) m², respectively. Creatinine clearance ranged from 22 to 213.7 ml min⁻¹ (median, 88.0 ml min⁻¹).

Table 2.

Key covariates in the population pharmacokinetic analysis (n = 226)

Continuous covariates	Median (range)
Age (years)	62.0 (23.0–86.0)
Body surface area (m2)	1.9 (1.3–2.6)
Weight (kg)	78.0 (35.5–132.3)
Albumin (g l−1)	38.0 (23.0–48.0)
Alanine aminotransferase (U l−1)*	20.0 (7.0–100.0)
Aspartate aminotransferase (U l−1)	24.0 (9.0–82.0)
Bilirubin (μmol l−1)	7.0 (1.7–39.3)
Creatinine clearance (ml min−1)	88.0 (21.9–213.7)

Categorical covariates	n (%)
Race (Caucasian / other)	196 (87) / 30 (13)
Gender (male / female)	130 (58) / 96 (42)

^*Three patients had missing alanine aminotransferase measurements.

Plasma concentration–time profiles following a single dose of i.v. or oral ixazomib (weekly dosing regimens; studies C16002 and C16004, respectively) are shown in Figure 1. Plasma concentration–time profiles for the twice weekly i.v. or oral ixazomib dosing schedules (C16001 and C16003) are shown in Supplementary Figure 1.

Figure 1.

Mean plasma concentration–time profiles following single dose administration of (A) intravenous ixazomib, , 0.125 mg m⁻² (n = 1); , 0.25 mg m⁻² (n = 7); , 0.5 mg m⁻² (n = 10); , 1.4 mg m⁻² (n = 4); , 1.76 mg m⁻² (n = 1); , 2.34 mg m⁻² (n = 1); , 3.11 mg m⁻² (n = 5) and (B) oral ixazomib (weekly dosing regimens), , 0.24 mg m⁻² (n = 1); , 0.48 mg m⁻² (n = 1); , 0.80 mg m⁻² (n = 2); , 1.20 mg m⁻² (n = 1); , 1.68 mg m⁻² (n = 3); , 2.23 mg m⁻² (n = 2); , 2.97 mg m⁻² (n = 24); , 3.95 mg m⁻² (n = 4)

A combined oral and i.v. three compartment model assuming first-order absorption and linear elimination was selected, with inter- and intra-individual variability described by an exponential function and additive error model, respectively. All model parameters were estimated with adequate precision when i.v. and oral data were fitted together. Random effects characterizing inter-individual variability were added on CL, V₂, V₄, absorption rate constant (K_a) and bioavailability (F).

Estimated ixazomib PK parameters from the base model and final model are shown in Table 3 and goodness-of-fit plots for the final model are shown in Figure 2. In the final model, the standard error obtained from the covariance step was <25% for all fixed effect parameters. Final estimates of ixazomib CL, K_a, and V₂ were 2.0 l h⁻¹, 0.5 h⁻¹ and 14.3 l, respectively, while absolute oral F was estimated as 60%. Shrinkage for CL and V₂ was <30%, but was higher for all other parameters tested (V₄, K_a and F, range 41–59%). Shrinkage on the residual error model was 6.7%. As shown in Figure 2A and 2B, data (population and individual predictions vs. observations) were evenly and randomly distributed across the line of identity, indicating no major bias and that the model was appropriate for the population and each individual. Examination of final model goodness-of-fit plots revealed conditional weighted residuals were evenly distributed about zero (Figure 2C and 2D), indicating no major bias in the structural and residual error models.

Table 3.

Estimated pharmacokinetic parameters in the base model and final model

Parameter	Base model estimates			Final model estimates
	Mean (%SE)	%BSV* (%SE)	% shrinkage	Mean (%SE)	%BSV* (%SE)	% shrinkage
CL (l h−1)	2.0 (4.9)	42.3 (16.9)	28.3	2.0 (4.8)	42.3 (17.2)	28.0
V2 (l)	15.2 (9.5)	96.2 (14.7)	20.3	14.3 (10.0)	100.5 (13.5)	20.2
Q3 (l h−1)	9.5 (14.1)	–	–	9.7 (15.3)	–	–
V3 (l)	410.0 (6.7)	–	–	412.0 (7.0)	–	–
Q4 (l h−1)	22.6 (5.6)	–	–	22.3 (5.7)	–	–
V4 (l)	86.2 (15.5)	58.1 (24.2)	45.0	83.4 (17.7)	45.2 (32.4)	51.3
Ka (h−1)	0.5 (7.2)	57.4 (21.6)	58.6	0.5 (7.4)	58.1 (21.8)	58.7
F	0.6 (6.3)	43.9 (16.9)	41.1	0.6 (6.0)	42.5 (16.8)	41.1
BSA on V4†	–	–	–	2.3 (18.9)	–	–
Residual error
Additive	0.3 (6.1)	–	7.0	0.3 (6.1)	–	6.7

^*% coefficient of variance.

^†Adding BSA as a covariate on V₄ decreased BSV from 58.1% in the base model to 45.2% in the final model (difference: −12.9%). BSA, body surface area; BSV, between-subject variability; CL, log-normally distributed central clearance; F, bioavailability; K_a, rate constant; Q₃, Q₄, inter-compartmental clearance terms; SE, standard error; V₂, central volume of distribution; V₃, peripheral volume of distribution in compartment 2; V₄, peripheral volume of distribution in compartment 3.

Figure 2.

Final model goodness-of-fit plots: (A) log-transformed observed vs. log-transformed population-predicted ixazomib plasma concentration (PRED), (B) log transformed observed vs. log transformed individual predictions of ixazomib plasma concentration (IPRED), (C) conditional weighted residuals (CWRES) vs. PRED and (D) CWRES vs. time after first dose of ixazomib. In A and B, the green line is a line with a slope of 1. In C and D, it is a line with a slope of 0. The blue line in all four panels represents the LOWESS (locally weighted scatterplot smoothing) line. The plots in A and B demonstrate a lack of bias in the model and that the model is appropriate for the population and each individual. The plots in panels C and D indicate a lack of bias in the structural and residual error models

Both weight and BSA were tested as covariates on CL, V₂ and V₄ in covariate analysis. BSA but not weight was significant on CL, neither BSA nor weight were significant on V₂, and V₄ was the only parameter on which an effect was identified for both BSA and weight. On V₄, both weight and BSA showed a similar drop in OFV (35 points). Inter-occasion variability was tested on CL but it resulted in an increase in OFV of 9 points and hence was not included in the model. In multivariate analysis, adding both BSA and weight did not result in a significant drop in OFV since both BSA and weight are correlated. Hence, BSA was included as a covariate on V₄ because dosing in the phase 1 studies was BSA-based and weight was not superior to BSA. In the final model, the optimal set of covariates selected by forward addition/backward elimination included only BSA on V₄. Forward selection included BSA on V₄ and CL, while backward selection included only BSA on V₄. Adding BSA as a covariate on V₄ decreased inter-individual variability from 58.1% in the base model to 45.2% in the final model (i.e. a relative decrease of 22.2%).

A plot of individual values of ixazomib CL across the range of BSA (1.3–2.6 m²) showed no apparent relationship between CL and BSA (Figure 3A). Similarly, analysis of individual ixazomib CL values showed no associations between CL and patient age (n = 226; Figure 3B), creatinine clearance (n = 226; Figure 3C), ixazomib dose (n = 226; Figure 3D), patient gender (n = 226; Figure 3E) or race (n = 226; Figure 3F). Additionally, neither route of administration nor dosing regimen was identified as a statistically significant covariate on CL.

Figure 3.

Relationship between inter-individual variability (IIV) in clearance (CL) and (A) body surface area (BSA), (B) patient age, (C) creatinine clearance (CL_cr), (D) ixazomib dose, (E) patient gender and (F) patient race (n = 226). A–D are scatter plots of IIV in CL vs. the covariate. The symbols indicate observed values and the solid blue line represents the LOWESS (locally weighted scatterplot smoothing) line. E and F are box plots of the distribution of IIV in CL by the covariate.The black line within the box is at the median, the box is bounded by the 25th and 75th percentiles of the distribution and the whiskers represent 1.5 times the interquartile range. There was no apparent relationship between ixazomib CL and any of the covariates tested

Predictive performance of the final model was assessed by visual predictive check. One thousand replicates of the dataset were simulated using the final population PK model following i.v. or oral dosing with ixazomib 4 mg, the starting dose selected for ongoing phase 3 clinical evaluations. The simulated data were then plotted with the observed data from phase 1 dose escalation studies, in which various ixazomib doses were administered, normalized to 4 mg dose (Figure 4). The median, and 5th and 95th percentiles (90% prediction interval) of the simulated data are shown. For the i.v. data, 3.2% of the observed data was below the 5th percentile while 6.2% of the data was above the 95th percentile. For oral dosing, 4.1% and 4.7% of observed concentrations were below the 5th and above the 95th percentiles, respectively.

Figure 4.

Dose-normalized ixazomib concentration over time and model-simulated median and 90% prediction interval following (A) intravenous and (B) oral dosing. Observed data (open circles) are from various doses tested in the dose escalation studies and scaled to a dose of 4 mg. Simulations were done for ixazomib 4 mg. Visual predictive checks showed that the model was able to characterize both intravenous and oral data. , median; , high; , low; ○, observations (doses >1 mg)

As shown in Figure 5A, the distributions of values of AUC were similar after BSA-based dosing and fixed dosing at the same ixazomib dose. The median AUCs (5th and 95th percentiles) after BSA-based and fixed dosing were 1503 (492–3531) ng ml⁻¹h and 1417 (479–3346) ng ml⁻¹h, respectively. There was also no observable trend in simulated AUC vs. BSA for fixed ixazomib dosing (P = 0.42; r² = 0.001; Figure 5B). No other covariates tested (race, ALT, AST, albumin and bilirubin) were clinically meaningful in relation to ixazomib CL or AUC (data not shown). AUC after BSA-based (2.23 mg m⁻²) and fixed dosing (4 mg) for 107 patients who received oral dosing in the C16003 and C16004 studies are shown in Supplementary Figure 2. The median AUC values were similar after BSA-based and fixed dosing (1294 vs. 1223 ng ml⁻¹h).

Figure 5.

Simulated area under the plasma concentration–time curve (AUC) for ixazomib (A) after body surface area (BSA)-based (2.23 mg m⁻²) and fixed (4 mg) dosing (n = 1000) and (B) vs. BSA for fixed dosing. Both simulations in A showed no significant difference in AUC between fixed dosing and BSA-based dosing of ixazomib. The symbols in B indicate simulated values and the solid grey line represents the linear regression line

Discussion

This is the first report on the population PK parameters of ixazomib, which is the first oral proteasome inhibitor to undergo clinical evaluation in patients with advanced haematological malignancies and solid tumours. In this analysis of phase 1 data, ixazomib population PK parameters in an adult cancer patient population were well described by a three compartment, mixed effects model with a first order absorption and linear elimination process. Even though none of the patients received both i.v. and oral drug, we were able to estimate absolute bioavailability with adequate precision using i.v. and oral PK data from two separate groups using a model-based approach in lieu of a dedicated study. However, the limited number of PK samples collected during the absorption phase may have contributed to the high shrinkage value for k_a. Furthermore, only plasma PK samples were collected and ixazomib has tri-exponential disposition without clear separation between the distribution and elimination phases. This may have contributed to the high shrinkage value in V₄. Population PK analysis showed that ixazomib had dose and time independent PK. These findings are consistent with previously reported results from non-compartmental analysis 18,37. The analysis also clearly showed that BSA over the range 1.3–2.6 m² does not significantly impact on ixazomib CL or systemic exposure. In addition, none of the other covariates tested, including creatinine clearance (≥22 ml min⁻¹), age, gender and race, had any clinically meaningful influence on ixazomib PK.

The lack of a significant effect of BSA on ixazomib population PK parameters suggests that a fixed dose is appropriate for future studies of ixazomib in adult patient populations. Indeed, based on the results of this analysis, patients have already been switched from BSA-based to fixed dosing in all ongoing and planned clinical studies of ixazomib. The starting dose of ixazomib in two phase 3 studies in MM (NCT01564537 and NCT01850524) is a fixed dose of 4 mg, which approximately corresponds to the BSA-based dose of 2.23 mg m⁻² (based on a mean BSA of 1.86 m² from bortezomib clinical trials in MM). Consistent with previous reports 19–21, these results support the premise that anticancer drug development should start with fixed rather than BSA-based dosing, thereby simplifying dosing for patients and clinical development. Fixed dosing also has the potential to reduce errors in dose calculations 19. However, there may be drugs, such as cisplatin, for which BSA-based dosing remains more appropriate than alternative dosing methods 38–40.

In a retrospective analysis of phase 1 PK studies of investigational oral and/or i.v. anticancer agents, BSA-based dosing reduced inter-individual PK variability for only five of the 33 agents analyzed (15%) 20. Furthermore, the relative reduction in variability in drug clearance for these five drugs was between 15% and 35%, which indicates that, even in the minority of cases in which BSA may be a potentially important consideration for dosing, only up to one-third of the variability can be explained by BSA. In the present study, however, BSA did not explain any variability in drug clearance. Interestingly, the feasibility of using fixed dosing for administration of oral capecitabine (a prodrug of the cytotoxic agent 5-fluorouracil) has been demonstrated in a PK and pharmacogenetic study of 26 women with metastatic breast cancer 41. While inter-individual variability in capecitabine and metabolite PK parameters was high following fixed dosing, it was not attributable in any way to observed differences in BSA, thus indicating the viability of the fixed dosing approach. Taken together, these findings suggest that BSA-based dosing may not be optimal for many anticancer agents, particularly oral drugs, and that fixed dosing may be more appropriate, at least until more individualized, patient-tailored strategies (based on genotyping, phenotyping and/or therapeutic drug monitoring) 19 become available, where appropriate, for use in clinical practice.

The early development, phase 1 studies of ixazomib were conducted in patients with different types of advanced cancer (including solid tumours, lymphoma and MM), which resulted in heterogeneous patient level data in the present analysis. However, this provided an opportunity to explore covariate/PK relationships across parameters such as age, baseline bilirubin and creatinine clearance over a wide range of data. Consequently, the results of the present analysis have additional important implications for the further clinical development of ixazomib. For example, there is a high prevalence of renal impairment in patients with MM 42, yet the lack of influence of creatinine clearance (range 22–213.7 ml min⁻¹) on ixazomib PK parameters suggests that there may be no need for starting dose adjustment in such patients with mild or moderate renal impairment (defined as creatinine clearance 30–90 ml min⁻¹). Due to the lack of impact of mild or moderate renal dysfunction on ixazomib PK parameters, two ongoing phase 3 studies in patients with relapsed/refractory MM (NCT01564537) or newly diagnosed, transplant-ineligible MM patients (NCT01850524) – both comparing the efficacy and safety of ixazomib plus lenalidomide and dexamethasone vs. placebo plus lenalidomide and dexamethasone – required no starting dose adjustment for patients with a creatinine clearance ≥30 ml min⁻¹. Furthermore, since renal elimination is a minor clearance pathway for ixazomib, renal dysfunction (creatinine clearance ≥30 ml min⁻¹) was not expected to have any meaningful impact on ixazomib PK. However, a phase 1 study to assess the PK of ixazomib in relapsed/refractory MM patients with severe renal impairment (defined as a calculated creatinine clearance of <30 ml min⁻¹), including those on haemodialysis, is also ongoing (NCT01830816) to provide dosing guidance in this patient population. Additionally, the absence of impact of age (over the range 23–86 years) on ixazomib PK parameters suggests that dose adjustments of ixazomib will not be required for elderly patients. This is an important consideration as, with a median age of onset of 70–74 years, the incidence of MM is highest in elderly individuals 43.

Lastly, some of the PK findings for ixazomib from the present analysis contrast with those reported for the first generation proteasome inhibitor bortezomib. Bortezomib displays accumulation in plasma following twice weekly repeat dosing that is greater than expected from its single dose PK profile 44. However, this is not seen for ixazomib, which has time-independent PK parameters. These differences may reflect the faster dissociation of ixazomib from the proteasome and greater tissue penetration, as compared with bortezomib, which has been reported in preclinical experiments 19 and may increase the drug availability at the site of action, such as in the bone marrow or tumours 45.

In conclusion, the population PK parameters of ixazomib can be well described by a three compartment model with first order absorption and linear elimination in patients with haematological and non-haematological malignancies. The model showed that ixazomib has dose and time independent PK. The model was also able to capture data for both i.v. and oral administration and for both twice weekly and weekly dosing regimens. The PK parameters of ixazomib were not affected by BSA, creatinine clearance, gender or age, implying that fixed dosing (starting dose of 4 mg), without dose modification for specific patient characteristics, such as mild/moderate renal dysfunction, was appropriate for ongoing and future clinical trials. Overall, these results provide an example of the application of modelling and simulation to influence posology decisions in oncology drug development. The ability to use a fixed dosing paradigm without the need for dose modifications for mild/moderate renal impairment may be a particular benefit for oral administration of ixazomib due to its potential to simplify dosing for patients and clinical development.

Financial support

This work was funded by Takeda Pharmaceuticals International Co.

Competing Interests

All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare NG, YZ, A-MH, D-LE and KV had support from Takeda Pharmaceuticals International Co. for the submitted work. NG, YZ, A-MH, D-LE, and KV have been employed for Takeda Pharmaceuticals International Co. in the previous 3 years. There are no other relationships or activities that could appear to have influenced the submitted work.

The authors would like to thank all patients, their families and the physicians who participated in the C16001, C16002, C16003, and C16004 studies. The authors acknowledge the help of Shripad Chitnis and Mohammad Saleh of Takeda Pharmaceuticals International Co. The authors would also like to acknowledge Jane Saunders, a medical writer with Firekite (part of the KnowledgePoint360 Group, an Ashfield company), for writing support during the development of this manuscript, which was funded by Millennium: The Takeda Oncology Company.

Contributors

NG, A-MH, D-LE, and KV designed the research; NG, YZ, and KV performed the research and collected the data, NG, YZ, and KV analyzed the data, all authors interpreted the data, NG drafted the manuscript and all authors reviewed the draft manuscript and approved the final version for submission.

Appendix 1. Control stream corresponding to the final nonmem model

$PROBLEM MLN9708 ORAL AND IV BASE MODEL /additive error model

;-------------------------

; UNITS

; Time – h

; Dose – mg

; Cp – ng ml⁻¹ = μg l⁻¹

; Clearances – l h⁻¹

; Volumes – l

;-------------------------

;C 3 compartment model with administration into either oral or the central compartment

$DATA NM_061813_race.csv IGNORE=@ BLANKOK

$INPUT COMMENT SID = DROP ID STDY SUB AMT NAMT TIME UN DV MDV CMT RTE RACE AGE HT BWT BSAC SEX SRCR CRCL ALT AST ALBUMIN BILIRUBIN DOSECAT = DROP

$SUBROUTINES ADVAN12;TRANS4

$PK

V0 =(BSAC/1.90)**THETA(9)

IF (V0.LE.0) V0 = 0.1

CL = THETA(1)*EXP(ETA(1)) ; CL clearance

V2 = THETA(2)*EXP(ETA(2)) ; V2 central volume

Q3 = THETA(3)*EXP(ETA(3)) ; Q3 intercompartmental clearance (central and periph 1)

V3 = THETA(4)*EXP(ETA(4)) ; V3 peripheral 1 volume

Q4 = THETA(5)*EXP(ETA(5)) ; Q4 intercompartmental clearance (central and periph 2)

V4 = (THETA(6)*V0)*EXP(ETA(6)); V4 peripheral 2 volume

KA = THETA(7)*EXP(ETA(7)) ; KA absorption rate constant

F1 = THETA(8)*EXP(ETA(8)) ; F1 Bioavailability

S2 = V2/1000

; Relationship:

K = CL/V2

K23 = Q3/V2

K32 = Q3/V3

K24 = Q4/V2

K42 = Q4/V4

$ERROR

IPRED = 0

IF(F.GT.0) IPRED = LOG(F)

Y = IPRED+ERR(1)

$THETA

(0.01 3) ; Theta (1) is CL clearance

(0.01 110) ; Theta (2) is V2 central volume

(0.01 13) ; Theta (3) is Q3 intercompartmental clearance (central and periph 1)

(0.01 590) ; Theta (4) is V3 peripheral 1 volume

(0.01 22) ; Theta (5) is Q4 intercompartmental clearance (central and periph 2)

(0.01 135) ; Theta (6) is V4 peripheral 2 volume

(0.01 1) ; Theta (7) is KA absorption rate constant

(0.01 0.5 1) ; Theta (8) is F1 Bioavailability

(0.01,1) ; THETA(9)

$OMEGA

0.34 ;FIXED ; ETA(1) applies to CL clearance

0.62 ;FIXED ; ETA(2) applies to V2 central volume

0 FIXED ; ETA(3) applies to Q3 intercompartmental clearance (central and periph 1)

0 FIXED ; ETA(4) applies to V3 peripheral 1 volume

0 FIXED ; ETA(5) applies to Q4 intercompartmental clearance (central and periph 2)

0.64 ;FIXED ; ETA(6) applies to V4 peripheral 2 volume

0.4 ;FIXED ; ETA(7) applies to KA absorption rate constant

0.16 ;FIXED ; ETA(8) applies to F1 bioavailability

$SIGMA

0.61 ; EPS(1)

$ESTIMATION METHOD = 1 PRINT = 5 MAX = 9999 NOABORT NSIG = 3 SIGL = 8 POSTHOC INTER MSFO = msfo.outputfile

$COVA

$TABLE ID STDY SUB AMT NAMT TIME UN MDV RTE RACE SEX AGE BWT BSAC SRCR CRCL ALT AST ALBUMIN BILIRUBIN CWRES IPRED

KA CL V2 V3 Q3 Q4 V4 F1 ETA1 ETA2 ETA3 ETA4 ETA5 ETA6 ETA7 ETA8 NOPRINT FILE =./BSA_V4.tab

Supporting Information

Additional Supporting Information may be found in the online version of this article at the publisher's web-site: