Pharmacokinetics and Pharmacodynamics of Tyrosine Kinase Inhibitors in the Treatment of Metastatic Renal Cell Carcinoma

Abstract

Tyrosine kinase inhibitors (TKIs) used in the treatment of metastatic renal cell carcinoma present high interindividual variability in clinical response and toxicity. In addition, high interindividual variability in pharmacokinetics (PK) of TKIs is observed which could explain therapeutic failure or elevated toxicity in some patients. Monitoring of plasma concentrations of TKIs and PK-guided dose adjustment could help to achieve plasma therapeutic levels in all the patients. The key step for applying PK-guided dosing is an established and stable over time exposure–response relationship. In this review, the current data about PK and PK/PD relationships of TKIs used in the treatment of metastatic renal cell carcinoma: axitinib, pazopanib, sunitinib and sorafenib are discussed with emphasis to reported factors of interindividual variability in plasma exposure (food intake, genetic background, biological and demographical variables and drug–drug interactions). In addition, this review provides recommendations about individualized dosing strategies based on exposure–response findings reported in the recent literature.

Keywords::

• pharmacodynamics/pharmacokinetics
• renal cell carcinoma
• therapeutic drug monitoring
• tyrosine kinase inhibitors

Combined targeting of VEGF and PDGF pathways in the treatment of metastatic renal cell carcinoma (mRCC) provides greater inhibition of tumor angiogenesis and lymphangiogenesis, which results in more effective inhibition of tumor growth and metastasis [1]. Tyrosine kinase inhibitors (TKIs) targeting VEGF and PDGF signaling pathways have significantly improved progression-free survival (PFS) and overall survival (OS) of patients with mRCC comparing to conventional chemotherapy. In addition, TKIs are accompanied with less adverse events (AEs) and oral route of administration constitutes a major improvement in the patients’ quality of life. However, high interindividual variability in clinical efficacy and safety of TKIs is observed. Treatment interruptions and discontinuations due to toxicity may decrease the clinical efficacy. Recent findings in the literature suggest relationship between plasma concentration of TKIs and clinical outcomes (efficacy and toxicity). Although TKIs are registered at fixed dosing regimens, high interindividual variability in plasma concentrations is observed which can be one of the factors of therapeutic failure or elevated toxicity in some patients. Therefore, monitoring of plasma concentrations and pharmacokinetically-guided dose adjustment can help to achieve plasma therapeutic levels in all the patients. Identification of factors explaining the interindividual variability in pharmacokinetics (PK) such as pharmacogenetic background, biological and demographic variables, concomitant intake of drugs interacting with the metabolism of TKIs may be useful to identify patients at risk of sub- or supra-therapeutic plasma concentrations and will help to adjust the dose in those patients.

The key step for applying PK-guided dose individualization is a clearly established and stable over time exposure–response relationship. Multiple pharmacokinetic/pharmacodynamic (PK/PD) studies have been performed in order to assess the PK target of plasma exposure accompanied with clinical efficacy and minimized toxicity. In this review, the current data about PK and PK/PD relationships of TKIs used in the treatment of mRCC: axitinib, pazopanib, sunitinib and sorafenib are discussed with emphasis to reported factors of interindividual variability in plasma exposure. In addition, we provide recommendations about individualized dosing strategies based on exposure–response findings reported in the recent literature.

Materials & methods

A literature search was performed (MEDLINE, PubMed) from January to April 2017. The primary search terms were: metastatic renal cell cancer, axitinib, pazopanib, sunitinib and sorafenib, PK and PD. All relevant English language articles that described clinical trials, PK and PD of TKIs in the treatment of renal cell carcinoma were revised in this article.

Axitinib

Pharmacodynamics

Axitinib (Inlyta^®) is a selective and potent inhibitor of kinase domain of VEGFR-1, -2 and -3 with eightfold lower inhibitory activity against PDGFR and KIT [2]. Axitinib blocks downstream signal transduction via the endothelial NO synthase (eNOS)/PI3K-Akt pathway implicated in the normal vascular homeostasis and in pathologic angiogenesis [2]. In various nonclinical models, axitinib showed dose-dependent inhibition of endothelial cell proliferation, angiogenesis and survival [2]. According to preclinical studies, the required in vivo pharmacologic concentration based on the inhibition of vascular VEGFR-2 phosphorylation and VEGF-mediated permeability is approximately 40 ng/ml (corresponding to 100 nmol/l, total concentration) in humans [2]. Axitinib is more potent than other TKIs targeting VEGFR with the concentration producing half maximal inhibitory concentration (IC50) of 0.1, 0.2, 0.1–0.3 nmol/l for VEGFR-1, -2 and -3, respectively [2].

Clinical data

In a Phase III trial, axitinib 5 mg twice daily (b.i.d.) demonstrated statistically significant improvement in PFS comparing to sorafenib (400 mg b.i.d.) in mRCC patients previously treated with one systemic therapy (median PFS: 6.7 vs 4.7 months with axitinib and sorafenib, respectively; hazard ratio, HR: 0.67; 95% confidence interval, CI: 0.54–0.81; p < 0.0001) [3]. Since 2012, axitinib is approved in the USA and in Europe for the second-line treatment of mRCC at the recommended dose of 5 mg b.i.d.

Safety profile

Axitinib presents similar safety profile than other TKIs with the most common AEs being diarrhea, hypertension and fatigue [3]. These AEs are usually managed by dose reductions and supportive drugs. In a Phase II study, the most common grade 3 AEs (according to the National Cancer Institute Common Terminology Criteria for Adverse Events) were hand-foot syndrome (16%), fatigue (16%), hypertension (16%) and diarrhea (14%). Nineteen percent of patients had to discontinue treatment due to a drug-related AEs. AEs resulted in temporary dose interruptions in 73% of patients and dose reductions in 45%. The most common AEs leading to dose interruption were dyspnea, nausea, fatigue, hypertension and vomiting.

Pharmacokinetics

Axitinib is administered orally at a recommended dose of 5 mg b.i.d. on a continuous daily dosing schedule with a possibility of dose titration (up to 10 mg b.i.d.) or reduction according to patient’s tolerance profile [4]. After oral administration, axitinib is rapidly absorbed with maximum plasma concentration (C_max) occurring within 3–6 h after administration (Table 1) [5]. The mean absolute bioavailability (F) of axitinib is 58% in the fasted state and 54% in the fed state [4]. The mean C_max and area under the concentration-time curve from time 0 to infinity (AUC_0–∞) increased proportionally over dosing range of 5–10 mg after single administration (mean estimated C_max 33.5, 51.1 and 69.4 ng/ml and geometric mean AUC_0–∞ 150, 251 and 321 ng.h/ml for doses of 5, 7 and 10 mg, respectively), indicating linear PK of axitinib in the therapeutic dose range [5,6]. In vitro studies revealed that axitinib binds to human plasma proteins in >99%, mostly to albumin and in a lesser extent to α₁-acid glycoprotein [7]. In vitro and in vivo studies demonstrated that axitinib is a substrate for the drug efflux transporters of the ATP-binding cassette (ABC) family: P-gp (encoded by the ABCB1 gene) and BCRP (encoded by the ABCG2 gene) [8]. P-gp is the main efflux transporter presented in the blood–brain barrier [9] and in vivo experimentations showed that P-gp strongly restricts axitinib brain accumulation [8]. This may contribute to therapeutic failure in patients with cerebral metastasis. Axitinib is a substrate for OATP1B1 and 1B3 which contribute to its hepatic absorption via active uptake [7]. The metabolism of axitinib in mediated mainly by cytochrome P450 (CYP) 3A4/5 isoform and to a lesser extent by CYP1A2, CYP2C19 and UDP-glucuronosyltransferase (UGT) 1A1 [10]. The major metabolites in plasma, sulfoxide metabolite and glucuronide conjugate, are pharmacologically inactive [7]. Mass-balance study confirmed that axitinib is mainly eliminated via hepatobiliary excretion into feces [11]. Renal excretion is negligible as only 1% of the administered dose is recovered in urine in the unchanged form [5]. Comparing to other TKIs used in mRCC, axitinib has a short half-life of 2.5–6.1 h, which results in steady-state achieved 2–3 days after start of treatment [4]. After continuous dosing of 5 mg b.i.d. of axitinib, mean (%CV) AUC_0–24 observed at steady-state is 265 ng.h/ml (77) [7]. Taking into account short half-life of axitinib and twice daily dosing schedule, the estimated accumulation ratio is 1.04–1.34 which is in accordance with the mean value of 1.40 reported in cancer patients [4].

Table 1. Pharmacokinetic parameters of tyrosine kinase inhibitors in the treatment of metastatic renal cell carcinoma.

	Axitinib	Pazopanib	Sorafenib	Sunitinib
Recommended daily dose	5 mg b.i.d.	800 mg q.d.	400 mg b.i.d.	50 mg q.d. (4/2)
F (%)	58	14–39	38–49	Unknown
Food effect	No	Yes	Controversed	No
Tmax (h)	3–6	3.5	2–14	6–12
Protein binding (%)	>99.0	>99.9	>99.5	95.0 sunitinib 90.0 SU12662
V/F (l)	160	25.2	213^†	2230
Metabolism	Major: CYP3A4 Minor: CYP1A2, CYP2C19, UGT1A1	Major: CYP3A4 Minor: CYP1A2, CYP2C8	Major: UGT1A9 Minor: CYP3A4	CYP3A4
Active metabolite	–	–	N-oxyde sorafenib	SU12662
Elimination	Hepatic	Hepatic	Hepatic + EC	Hepatic
CL/F (L/h)	38	0.605	8.13	34–62
t1/2(h)	2.5–6.1	31	25–48	40–60 sunitinib 80–110 SU12662
Transporter substrate	P-gp, BCRP, OATP1B1, OATP1B3	P-gp, BCRP, OATP1B1	P-gp, BCRP	P-gp, BCRP
Dose-exposure linearity	Yes	Up to 800 q.d.	Up to 400 mg b.i.d.	Yes
AUCss,0-τ	380 ng.h/ml	275 μg.h/ml	47.8 μg.h/ml	1592 ng.h/ml
Css,max	31.0 ng/ml	45.1 μg/ml	5.40 μg/ml	91.9 ng/ml
Ref.	[4–8,10]	[12–19]	[19–23]	[19,24–31]

^†Assuming median body weight of 80 kg.

4/2: 4-week dosing, followed by 2-week rest; AUC_ss,0-τ: Steady-state area under the concentration–time curve in the dosing interval; b.i.d.: Twice daily; CL/F: Apparent oral clearance; Css,_max: Steady-state maximum plasma concentration; EC: Enterohepatic circulation; F: Absolute bioavailability; NA: Not available; q.d.: Once daily; t_1/2: Elimination half-life; T_max: Time to peak concentration; V/F: Apparent volume of distribution.

In the population analysis, the PK of axitinib was described by a two-compartment model with first-order absorption and lag time [32,33]. Typical values (%CV) of systemic clearance (CL) and central volume of distribution (Vc) were 17.0 L/h (52.5) and 45.3 L (30.8), respectively [32]. The bioavailability after oral administration was estimated to be 46.5%.

Factors of PK variability

Analysis of PK data from Phase I, II and III clinical trials showed high inter- and intra-individual variability in plasma exposure to axitinib (%CV: 17–94% and 20–33%, respectively) [4]. Different factors susceptible to influence PK of axitinib are discussed below.

Food intake

For the axitinib crystal polymorph form XLI formulation (commercial), high-fat, high-calorie meal increased AUC_0–∞ and C_max by 19 and 11%, respectively, comparing to overnight fasting. An opposite effect was observed when axitinib was administered with moderate-fat, standard-calorie meal which resulted in AUC_0–∞ and C_max decreased by 10 and 16% comparing to overnight fasting [34]. These changes are not considered clinically relevant; therefore, axitinib can be administered with or without food.

Hepatic impairment

The impact of hepatic impairment on axitinib PK was evaluated in a Phase I trial in subjects with normal hepatic function (n = 8) and mild (Child-Pugh A, n = 8) or moderate (Child-Pugh B, n = 8) hepatic impairment [35]. After a single oral 5 mg dose of axitinib, plasma drug exposure was similar in subjects with normal hepatic function and those with mild hepatic impairment, but approximately twice higher in subjects with moderate hepatic impairment (geometric mean (%CV) AUC_0–∞: 156 (63), 122 (167) and 304 (44) ng.h/ml, respectively). Therefore, altered axitinib elimination in patients with moderate hepatic impairment may require dose reduction (i.e., starting dose should be reduced from 5 mg b.i.d. to 2 mg b.i.d.). However, in this study, single 5 mg dose of axitinib was well tolerated in subjects with both mild and moderate hepatic impairment. Axitinib has not been studied in patients with severe hepatic impairment on axitinib PK.

Renal impairment

Since only 1% of axitinib dose is recovered in urines in the unchanged form, no Phase I study of axitinib PK has been conducted in patients with renal impairment. However, a population analysis was performed to evaluate the impact of renal function on axitinib PK and toxicity in 590 subjects [36]. The study reported mean axitinib systemic CL values of 14.0, 10.7, 12.3, 7.8 and 12.6 l/h in individuals with normal renal function (creatinine clearance, CL_CR ≥ 90 ml/min; n = 381), mild renal impairment (60–89 ml/min; n = 139), moderate renal impairment (30–59 ml/min; n = 64), severe renal impairment (15–29 ml/min; n = 5) and end-stage renal disease (ESRD < 15 ml/min; n = 1), respectively. This study highlights the lack of significant impact of renal impairment on axitinib elimination. Moreover, safety profiles were similar across studied groups, thus no starting dose adjustment is needed in patients with mild to severe renal impairment. Impact of ESRD on axitinib elimination should be considered with caution as only one patient was evaluated in that group.

Drug–drug interactions

The solubility of axitinib decreases with an increase in gastric pH, thus drugs raising gastric pH may decrease plasma exposure to axitinib. However, no clinically significant changes in absorption (AUC and C_max) were observed when axitinib was administered with rabeprazole, a proton pump inhibitor [5]. Therefore, drugs modifying gastric pH, including proton pump inhibitors and H₂-receptor antagonists, can be administered with axitinib [4].

Axitinib is mainly metabolized by CYP3A4 in liver. Concomitant administration of axitinib and ketoconazole, a strong CYP3A4 and P-gp inhibitor, resulted in increased plasma axitinib exposure (AUC_0–∞ increased by 2.06-fold; 90% CI: 1.84–2.30 and C_max by 1.50-fold; 90% CI: 1.33–1.70) (Table 2) [37]. This increase in exposure could be related to both inhibition of axitinib metabolism via CYP3A4 and inhibition of intestinal efflux mediated by P-gp which results in higher absorption of axitinib. In the presence of rifampicin, a strong CYP3A4 and P-gp inducer, AUC_0–∞ and C_max of axitinib were reduced by 79 and 71%, respectively [38]. According to the current recommendations, concomitant use of strong CYP3A4 inhibitors should be avoided and a therapeutic alternative with minimal potential of CYP3A4 inhibition should be chosen. However, if a strong CYP3A4 inhibitor must be co-administered, axitinib dose should be reduced by half as this dose is suggested to adjust plasma axitinib exposure to the range observed in patients not taking concomitant CYP3A4 inhibitors [4]. Concomitant use of strong CYP3A4 inducers should be avoided.

Table 2. Drug–drug interactions with tyrosine kinase inhibitors based on reported clinical data.

Drug	Type of interaction	Axitinib	Pazopanib	Sorafenib	Sunitinib
Ketoconazole	CYP3A4 inhibition	Increased AUC (2.06-fold) Increased Cmax (1.5-fold) [37]	Increased AUC (66%) Increased Cmax (45%) [68]		Increased composite AUC (51%) Increased composite Cmax (49%) [26]
Rifampicin	CYP3A4 induction	Decreased AUC (79%) Decreased Cmax (71%) [38]	Possible decrease in AUC and Cmax [68]		Decreased composite AUC (46%) Decreased composite Cmax (23%) [26]
Esomeprazole	Increase in gastric pH		Decreased AUC (40%) Decreased Cmax (40%) [68,69]
Felodipine	CYP3A4 competition			Increased Ctrough [143]
Prednisolone	CYP3A4 induction			Decreased Ctrough (>twofold) [142]
Nicardipine	CYP3A4 inhibition				Increased composite AUC (50%) [96]

AUC: Area under the concentration-time curve; C_max: Maximum plasma concentration level; C_trough: Minimum plasma concentration level; CYP3A4: Cytochrome P450 3A4.

Axitinib was identified as CYP1A2, CYP2C8 and P-gp inhibitor [7]. Therefore, concomitant administration of drugs metabolized by these isoenzymes or P-gp substrates may result in increased plasma levels of these drugs.

Biological & demographic variables

Two population analyzes evaluated impact of demographic covariates on axitinib PK. In the first study evaluating PK data from 10 Phase I trials including 337 healthy volunteers, systemic CL was not affected by any of the studied biological and demographic covariates (age, body weight, serum alanine aminotransferase (ALT), aspartate aminotransferase (AST), bilirubin, CL_CR, gender, race, smoking status) [32]. Vc was significantly influenced by body weight. In the second population approach analysis evaluating 590 subjects including both healthy volunteers and cancer patients, age ≥ 60 years and Japanese ethnicity were associated with decreased axitinib systemic CL whereas Vc increased with body weight [33]. However, changes in axitinib CL related to age and Japanese ethnicity were comprised within estimated interindividual variability in CL. Likewise, in both studies, Vc was influenced by body weight but the expected change in plasma exposure with change in body weight is of limited clinical relevance. Therefore, no axitinib dose adjustment based on demographic variables is needed. Of note the effect of smoking status could not be accurately assessed as the smokers constituted only 3% of the studied population [33]. Further investigations are needed in order to assess the impact of smoking status on PK of axitinib.

Genetic polymorphisms

Polymorphisms in genes encoding for CYPs implicated in the metabolism of axitinib may influence its PK. In a population PK study, the impact of single nucleotide polymorphisms in the CYP2C19 gene (*2, *3, *17) and of UGT1A1*28 genotype on axitinib systemic CL was evaluated in 337 healthy volunteers [32]. This study reported no impact of analyzed genetic polymorphisms on axitinib PK. Concerning genetic polymorphisms of ABC transporters, increased dose-adjusted steady-state AUC_0–6 of axitinib was reported in patients with ABCB1 haplotype (1236C > T, 2677G > T/A, 3435C > T) causing decreased expression of P-gp [39]. In a fixed effects meta-analysis performed using data from 11 clinical trials in healthy volunteers, no correlation between CYP3A4/5, CYP1A2, CYP2C19, UGT1A1 and the drug transporters P-gp and OATP1B1 (encoded by SLC01B1) polymorphisms and plasma exposure to axitinib was observed [40]. However, this meta-analysis evaluated axitinib exposure after single dose and did not take into account all existing genetic polymorphisms which could influence axitinib PK. Therefore, further investigations are warranted in order to clarify the importance of genotype-adjusted dose of axitinib.

PK/PD relationship

A Phase I study reported that increased axitinib plasma exposure was associated with higher efficacy, as indicated by decreased tumor perfusion and volume, specifically, a 50% decrease in volume transfer constant corresponded to plasma AUC_0–24 of 200 ng.h/ml [41]. In a Phase II dose titration study, patients with starting axitinib dose of 5 mg b.i.d. who did not present any AE and who were not taking more than two antihypertensive medications were randomized into two groups: either axitinib titration up to 10 mg b.i.d. or placebo titration. The number of patients achieving the objective response (OR) was higher in axitinib titration group suggesting a relationship between plasma axitinib exposure and clinical efficacy [42]. However, PFS was similar among patients who achieved the OR and those who did not. In addition, evaluation of PK data from this study showed that patients with steady-state AUC_0–24 ≥ median (261 ng.h/ml) and those with AUC_0–24 < median had similar clinical outcomes (Table 3) [43]. In a PK/PD population study, patients with steady-state AUC_0–24 greater or equal to median value (300 ng.h/ml) had significantly longer median PFS and OS than patients with AUC_0–24 < 300 ng.h/ml (Table 3) [33]. Finally, a retrospective analysis of 5-year survival in 52 mRCC patients showed that patients with axitinib 1–2 h postdose concentration at day 1 of treatment of 45.2–56.4 ng/ml (corresponding to third quartile; n = 12) had the best clinical outcome among other patients, as demonstrated by objective response rate (ORR) of 82%, median PFS of 28.3 months and median OS that was not reached after 5 years [44]. Patients in the fourth quartile (n = 13) had the highest incidence of grade ≥ 3 AEs. Kato et al. reported that high steady-state AUC_0–6 of axitinib (129.4 ng.h/ml) was associated with the incidence of fatigue during therapy (p = 0.013) [39]. Interestingly, the treatment period without discontinuation or dose reduction due to AEs in patients with high AUC_0–6 of axitinib (≥97.3 ng.h/ml; n = 9) was significantly shorter than for those with low AUC_0–6 (<97.3 ng.h/ml; n = 9, p = 0.024).

Table 3. Relationship between pharmacokinetic parameters, efficacy and safety of tyrosine kinase inhibitors in metastatic renal cell carcinoma patients.

TKI	N	PK parameter	Outcome	Results	p-value	Ref.
Efficacy
Axitinib	112	Dose titration vs placebo titration	OR	54.0 vs 34.0%	0.019	[42]
		AUC0–24 ≥ vs < 261 ng.h/ml	PFS	13.0 vs 16.5 months	0.981
	49	C1–2: 45.2–56.4 ng/ml	OS	Not reached vs 20.3–27.7 months for patients in 3rd quartile with C1–2: 45.2–56.4 ng/ml compared with other quartiles	NA	[44]
		AUC0–12: 154–620 ng.h/ml	PFS	28.3 vs 7.5–11.8 months	NA
			ORR	81.8 vs 16.7–53.8%	NA
	168	AUC0–24 ≥ vs < 300 ng.h/ml	OS	37.4 vs 15.8 months	<0.001	[33]
			PFS	13.8 vs 7.40 months	0.003
Pazopanib	10	Ctrough ≥ vs < 15 μg/ml	PR + SD	83 vs 0%	NA	[12]
	205	Ctrough ≥ vs < 20.5 μg/ml	PFS	52.0 vs 19.6 weeks	0.0038	[45]
			RR	45.0 vs 19.0%	<0.0001
			Tumor shrinkage	37.9 vs 6.90%	<0.0001
Sunitinib	146	Composite AUC0–24 ≥ vs < 800 ng.h/ml	TTP	Longer TTP with increasing AUC0–24	0.001	[46]
			OS	Longer OS with increasing AUC0–24	0.010
			ORR	ORR increased with increasing AUC0–24	<0.001
	55	AUC ≥ vs < 1973 ng.h/ml	OS	35.2 vs 16.7 months	0.0051	[47]
Sorafenib		No study in mRCC
Safety
Axitinib	20	AUC0–6	Fatigue	129.4 vs 87.1 ng.h/ml for patients with and without fatigue	0.013	[39]
			Hoarseness, diarrhea Hypertension	No correlation	>0.05
		AUC0–6 ≥ vs < 97.3 ng.h/ml	Treatment period without discontinuation or dose reduction due to AE	Median: 35 days vs not reached	0.024
Pazopanib	54	Ctrough 12.6–46 μg/ml	Hypertension	77.0 vs 39.0%	NA	[12]
	205	Ctrough ≥ vs < 15 μg/ml	Diarrhea, hair color change, ALT increase, HFS and stomatitis	Increased incidence of AEs by twofold	NA	[48]
	205	Ctrough	HFS	Positive relationship	<0.01	[49]
Sorafenib	51	AUCcum	Any grade ≥3 toxicity	Odds ratio = 1.07	0.03	[50]
	52	Ctrough	Grade ≥2 HFS	Ctrough higher in patients exhibiting grade ≥2 HFS	0.0045	[51]
			Grade ≥2 hypertension	Ctrough higher in patients exhibiting grade ≥2 hypertension	0.0453
	72	AUC0–12	Any grade 3–4 toxicity	AUC0–12: 61.9 μg.h/ml vs 53.0 μg.h/ml for patients with and without AE	0.017	[52]
	22	Ctrough	Any grade ≥3 toxicity	7.70 vs 4.40 μg.ml for patients with and without AE	0.0083	[53]
	94	Ctrough >10.0 μg/ml	Any grade 3–4 toxicity		<0.042	[54]
Sunitinib	28	Composite Ctrough >100 ng/ml	DLT	More frequent DLT	NA	[29]
	443	Composite AUC0–24	Fatigue	Positive relationship	NA	[46]
		Composite AUCcum	Neutropenia	Negative relationship	NA
		Composite Ctrough	dBP	Positive relationship	NA
	36	Composite Ctrough	Grade ≥2 mucositis	126.5 vs 84.8 ng/ml for patients with ≥ or < grade 2 mucositis	0.04	[55]
			Altered taste	159.9 vs 105.2 ng/ml for patients with ≥ or < grade 2 altered taste	0.05
	21	Composite Ctrough ≥ vs < 100 ng/ml	Any grade ≥3 toxicity	75.0 vs 23.1%	NA	[56]
		Composite Ctrough	Grade ≥3 thrombocytopenia, anorexia	Concentration-dependent severity of AEs	< 0.05
	92	Sunitinib AUC0–24	Any grade ≥3 toxicity	Odds ratio = 1.16	0.005	[47]
		SU12662 AUC0–24	Grade ≥ 2 thrombocytopenia	Odds ratio = 1.27	0.028
	19	Ctrough ≥ vs < 90 ng/ml	Grade ≥2 thrombocytopenia	100 vs 55.6%	0.033	[57]
			Grade ≥2 hypertension	90.0 vs 22.2%	0.0055
	24	Composite Ctrough ≥ vs < 180 ng/ml	QT interval	15.4 vs 9.60 ms	NA	[58]
	52	CL/F	Any grade 3 toxicity	34.4 vs 41.4 L/h for patients with and without grade 3 AE	0.025	[59]
		Composite Ctrough	Fatigue	Positive relationship	0.007

AE: Adverse event; AUC: Area under the concentration-time curve; AUC_cum: Cumulative AUC; C_1–2: Concentration 1–2 h post-dose; CL/F: Apparent oral clearance; C_trough: Minimum plasma concentration level; Composite C_trough: C_trough sunitinib + metabolite; Composite AUC: AUC sunitinib + metabolite; dBP: Diastolic blood pressure; DLT: Dose-limiting toxicity; HFS: Hand-foot syndrome; NA: Not available; OR: Objective response; ORR: Objective response rate; OS: Overall survival; PD: Progressive disease; PFS: Progression-free survival; PR: Partial response; RR: Response rate; SD: Stable disease; sVEGFR2: Soluble VEGF receptor 2; TKI: Tyrosine kinase inhibitor; TTP: Time to progression.

Individualized dosing recommendations

Although multiple studies demonstrated exposure–response relationship for axitinib, no plasma target exposure has been established to perform individualized dosing. Therefore, based on the current knowledge, toxicity-driven axitinib dose titration seems the best option to improve plasma drug exposure in mRCC patients. According to manufacturer label, the recommended starting dose of axitinib is 5 mg b.i.d. with subsequent dose increase to 7 mg b.i.d. and further to 10 mg b.i.d. in patients who did not experience grade >2 AE, significant BP increase and are not concomitantly treated with antihypertensive agents [4]. However, in routine practice, dose titration is often performed in patients concomitantly treated with antihypertensive agents and axitinib is used at doses which are not listed in manufacturer label (e.g., 6 mg b.i.d.). These approaches may be beneficial to optimize axitinib exposure in all the patients especially in those who are at increased risk of toxicity.

Pazopanib

Pharmacodynamics

Pazopanib (Votrient^®) is an angiogenesis inhibitor with a highly potent and selective activity against VEGFR-1, -2 and -3, PDGFR-α and -β, and c-Kit tyrosine kinases [60]. It shows modest activity against FGFR-1 and -3, and c-fms receptor tyrosine kinases. In vivo studies demonstrated that pazopanib inhibits angiogenesis and tumor growth in a dose-dependent manner in many human tumor xenografts in mice. Preclinical PK/PD studies showed that plasma steady-state concentration of ≥17.5 μg/ml would be optimal for in vivo activity [60].

Clinical data

In a randomized, double-blind Phase III trial, pazopanib showed significant improvement in PFS compared with placebo (9.2 vs 4.2 months; HR = 0.46; 95% CI: 0.34–0.62; p < 0.0001) in treatment-naive and cytokine-refractory patients with advanced RCC [61]. In addition, the OR rate was 30% with pazopanib comparing to 3% with placebo (p < 0.001). Pazopanib was approved in 2009 by US FDA for the first-line treatment of mRCC patients and for the second-line treatment of patients who have been previously treated with cytokine therapy, at the recommended dose of 800 mg once daily (q.d.).

Safety profile

In a Phase III clinical trial, the most common AEs were diarrhea (52%), hypertension (40%), hair color changes (38%), nausea (26%), anorexia (22%) and vomiting (21%) [61]. Hutson et al. reported that the most frequent grade 3 or 4 treatment-related AEs in RCC patients treated with pazopanib 800 mg q.d. were hypertension (8%), increased ALT (6%) and AST (4%), diarrhea (4%) and fatigue (4%) [62]. These AEs led to dose reductions to 400 mg in 31% of patients. Approximately 50% of these patients could undergo subsequent dose re-escalations. Fifteen percent of patients discontinued pazopanib as a result of an AE, especially related to an elevation of liver enzymes.

Pharmacokinetics

In a Phase I study evaluating pazopanib PK in the dose range of 50–2000 mg taken p.o. once daily, steady-state AUC and C_max of pazopanib did not increase in dose-proportional manner [12]. A plateau of steady-state exposure was observed for doses above 800 mg q.d. Mean pazopanib (%CV) AUC_0–24, plasma through concentration (C_trough) and time to reach maximum concentration (T_max) at day 1 of treatment were 275.1 μg.h/ml (203), 9.4 μg/ml (240) and 3.5 h, respectively, for the recommended dose of 800 mg q.d. The absolute oral bioavailability of pazopanib at dose 800 mg ranged from 14 to 39% [13]. Administration of 400 mg of pazopanib as crushed tablet and oral suspension resulted in increased rate and extent of oral absorption comparing to whole tablet (AUC_0–72 and C_max increased by 46% and approximately twofold, respectively, and T_max decreased to 2 h when administered as crushed tablet whereas administration of oral suspension led to increase in AUC_0–72 and C_max by 33 and 29%, respectively, and decrease in T_max to 1 h) [14]. Pazopanib binds to plasma proteins in >99.9%, mostly to albumin and to a lesser extent to α₁-acid glycoprotein [15]. According to in vitro studies, pazopanib is a substrate for P-gp, BCRP and OATP1B1 [13]. The metabolism of pazopanib is mediated by CYP3A4 and to a lesser extent by CYP1A2 and CYP2C8. Four minor metabolites of pazopanib have been identified and they account for 6% of the total exposure in plasma. One of pazopanib metabolites (GSK1268997) showed similar potency against VEGF-stimulated human umbilical vein endothelial cells than pazopanib, the others are 10- to 20-folds less active [16]. However, given the low-circulating concentrations of the active metabolites of pazopanib, their contribution to pharmacological effect is negligible. The mean elimination half-life of pazopanib is 31.1 h [12]. Pazopanib is mainly excreted as unchanged drug in feces (67% of the administered dose), which most likely corresponds to the unabsorbed dose. Less than 4% of the administered dose is recovered in urine. On day 22 of daily administration of 800 mg of pazopanib, mean AUC_0–24 was 743.3 μg.h/ml [12]. Upon multiple dose administration, no time-dependent PK was observed as indicated by accumulation ratio of 1.45 (90% CI: 0.75–2.79) which is an expected value for a compound administered on a q.d. dosing schedule and a half-life of approximately 30 h [16].

In a population analysis reported by Imbs et al. [17], the PK of pazopanib was described by a one-compartment model with first-order absorption and lag time. The interindividual variability of 40% on apparent clearance (CL/F) was observed and this value was larger than the intraindividual variability (27%) which emphasizes that therapeutic drug monitoring (TDM) might be useful to optimize pazopanib treatment. Yu et al. described a complex two-compartment model with combined fast and slow absorption processes [18]. The nonlinear dose–concentration relationship (plasma exposure to pazopanib decreased over time) was observed which is due to saturated bioavailability of pazopanib. High inter- and intra-individual variability of relative bioavailability was observed (36 and 75%, respectively).

Factors of PK variability

Pazopanib has a complex PK profile characterized by pH-dependent absorption and nonproportional dose–exposure relationship. As a result, high inter- and intraindividual variability in systemic exposure is observed (19.0–77.0% and 27.3%, respectively) [12,63]. Multiple factors influencing the PK of pazopanib have been identified of which the most important are factors influencing gastric pH (food intake, antacids) and drug–drug interactions in CYP-mediated metabolism.

Food intake

Pazopanib is highly lipophilic and practically insoluble in pH > 4, thus its absorption in gastrointestinal tract may be strongly influenced by food intake. In a Phase I study evaluating the effects of food on pazopanib PK, patients were administered either 400 mg or 800 mg with a low-fat, high-fat meal or in the fasted state. Pazopanib AUC_0–72 and C_max were increased by approximately twofold when taken with either low-fat or high-fat meal comparing with the corresponding values in patients in the fasted state [64]. In addition, food intake showed similar effect on several pazopanib metabolites with approximately twofold increase in mean AUC_0–24 and C_max, however the exposure to pazopanib active metabolite remained <5% of the corresponding mean value for parent compound. According to current recommendations, pazopanib should be administered in the fasted state (1 h before or 2 h after a meal) which would minimize intraindividual variability in the systemic exposure to pazopanib [64].

Hepatic impairment

The impact of hepatic impairment on pazopanib PK and the maximum tolerated dose in patients with liver dysfunctions were evaluated in 89 patients with solid tumor in a Phase I study [65]. Patients with mild hepatic impairment (according to National Cancer Institute Organ Dysfunction Working Group) tolerated the recommended dose of 800 mg q.d. whereas patients with moderate liver dysfunction tolerated 200 mg q.d. According to these findings, FDA recommends no dose adjustment in patients with pre-existing mild hepatic impairment and for patients with moderate hepatic impairment, a starting dose of 200 mg is recommended. In this study, patients with severe hepatic impairment tolerated the dose of 200 mg q.d., even though pazopanib is not recommended in those patients [16].

Renal impairment

No study evaluating impact of renal impairment on PK of pazopanib was conducted as pazopanib is excreted in urine to a very low extent (<4% of the administered dose). A population PK study showed that CL_CR (calculated using Cockcroft-Gault formula) did not influence CL of pazopanib, thus PK of pazopanib seems not to be modified by renal impairment. Moreover, Shetty et al. analyzed retrospectively nine mRCC patients with ESRD treated with pazopanib in routine clinical practice [66]. Dose reductions (to 400 mg and 600 mg q.d.) due to AEs were required in two patients treated with 800 mg q.d. Four patients treated with 600 mg q.d. required dose reduction to 400 mg. Finally, Czarnecka et al. reported that pazopanib treatment of patients with ESRD undergoing hemodialysis is safe and effective based on experience from clinical practice [67]. According to these findings, no dose adjustment is recommended in patients with mild or moderate renal impairment, however, still little is known about use of pazopanib in patients with severe renal impairment (CL_CR < 30 ml/min).

Drug–drug interactions

Concomitant administration of ketoconazole, a strong CYP3A4 and P-gp inhibitor, increased mean AUC_0–24 and C_max of pazopanib by 66 and 45%, respectively (Table 2) [68]. Thus the increase in pazopanib exposure may be a result of both CYP3A4 and P-gp inhibition. According to current recommendations, use of strong CYP3A4 inhibitors should be avoided or pazopanib dose should be reduced to 400 mg if another treatment option cannot be used. Similarly, concomitant use of CYP3A4 inducers such as rifampicin could decrease plasma exposure to pazopanib, therefore concomitant administration is not recommended [16].

The absorption of pazopanib is pH-dependent and it may be decreased by drugs raising gastric pH. Administration of esomeprazole, a proton pump inhibitor, in the evening and pazopanib in the morning resulted in mean pazopanib AUC_0–24 and C_max decreased by 40% [68,69]. Therefore, use of drugs increasing gastric pH (proton pump inhibitors and H2 receptor agonists) may be a factor of subtherapeutic exposure to pazopanib and treatment failure. Use of antacids should be avoided; however, if no other therapeutic option can be chosen, then short acting antacids should be used or pazopanib and antacids should be administered with several hours of interval [16].

Concerning the impact of pazopanib on other drugs, pazopanib is a weak CYP3A4 and CYP2D6 inhibitor and has no effect on CYP1A2, CYP2C9 and CYP2C19 [70]. The concomitant use of drugs metabolized by CYP3A4 and CYP2D6 isoenzymes may increase their plasma exposure, thus in case of drugs with narrow therapeutic index, an alternative therapy should be considered. Indeed, as reported in clinical trials, increased incidence of liver enzymes elevation was reported in patients treated with pazopanib who were concomitantly receiving simvastatine, a known substrate for CYP3A4 [16]. Inhibition of CYP3A4 by pazopanib likely contributed to increased exposure to simvastatine.

Biological & demographic variables

In the population PK study reported by Imbs et al., none of the tested biological and demographic covariates (age, gender, body weight, nephrectomy status, CL_CR, serum albumin, serum α₁-acid glycoprotein, AST and ALT) had a significant impact on pazopanib CL/F [17]. These results were confirmed in another population analysis [71]. However, the number of patients analyzed in these two studies was low (n = 25 and 32, respectively).

Genetic polymorphisms

To our best knowledge, no data about the influence of pharmacogenetic determinants on the pazopanib PK are currently available in the literature.

PK/PD relationship

Considering the relationship between plasma exposure to pazopanib and toxicity, Imbs et al. reported that patients who discontinued pazopanib short after the start of treatment had higher plasma C_trough at the beginning of treatment than patients who continued the therapy (mean [±SD] C_trough of 28.6 ± 15.3 mg/l and 20.1 ± 10.4 mg/l, respectively) [71]. It suggests that subjects with higher plasma exposure at the beginning of treatment are at increased risk of toxicity. Indeed, the occurrence of hypertension in patients treated with pazopanib has been associated with pazopanib steady-state plasma C_trough ≥ 15 μg/ml (Table 3) [12].

The association between plasma exposure to pazopanib and clinical outcomes was first observed in Phase I study in which 83% of RCC patients (n = 5) having either partial response or stable disease as the best response, achieved a steady-state C_trough > 15 μg/ml whereas all patients (n = 4) with C_trough < 15 μg/ml had disease progression [12]. Moreover, this result is in accordance with previous findings from preclinical studies reporting a threshold steady-state concentration of 17.5 μg/ml needed for the inhibition of VEGFR-2 [60]. Another PK/PD study reported that patients with a steady-state C_trough > 20.5 μg/ml had more than fivefold greater median observed tumor shrinkage than patients with steady-state C_trough ≤ 20.5 μg/ml (Table 3) [45]. In addition, the median PFS for patients C_trough ≤ 20.5 μg/ml was 19.6 weeks comparing to 52.0 weeks for patients with C_trough > 20.5 μg/ml (p = 0.0038). On the basis of these findings, PK-guided dose adjustment of pazopanib seems of clinical interest to optimize efficacy and decrease toxicity of pazopanib. Two prospective trials were performed in order to evaluate the feasibility of PK-guided dose individualization. In a randomized, two-sequence, three-periods, crossover study [63], plasma pazopanib concentrations were measured at days 14, 28 and 42 and AUC_0–24 was calculated to guide individualized dosing based on the predefined target exposure (median AUC_0–24 from two Phase I studies: 805 μg.h/ml). Briefly, all the patients were treated with pazopanib at recommended dose of 800 mg q.d. for 14 days. During the second period, half of the patients received pazopanib individualized dose whereas the second half continued with pazopanib 800 mg q.d. During the third period, the two regimens switched. Pazopanib dose individualization did not show any benefit in decreasing the interindividual variability in exposure which argues for TDM of pazopanib. However, a small number of patients (n = 13) and the intraindividual variability on AUC_0–24 observed in this study was of the same magnitude as the interindividual variability (%CV 24.8 and 27.3, respectively) which made the individualized dosing difficult. Interestingly, de Wit et al. reported the correlation between C_trough (concentration taken exactly 24 h after pazopanib intake) and AUC_0–24 which shows that TDM could be guided by evaluation of pazopanib C_trough [63]. Verheijen et al. carried out a study evaluating the feasibility and utility of individualized pazopanib dosing in which the PK target reported by Suttle et al. was used (steady-state C_trough > 20.0 μg/ml) [72]. In that study, starting dose for all the patients was 800 mg q.d. and the dose adjustments at weeks 3, 5 and 7 were performed according to the following algorithm: patients with a C_trough < 15.0 μg/ml had a dose increase of 400 mg daily in case of no grade > 2 AEs or 200 mg daily when grade 2 but not grade >3 AEs occurred. Patients with a C_trough of 15.0–19.9 μg/ml had a dose increase of 200 mg if toxicity was grade <3. In patients who presented at least one C_trough < 20.0 μg/ml over the study course, the dose escalation led to significant increase in exposure (mean [%CV] C_trough increased from 13.2 [38.0] μg/ml to 22.9 μg/ml [44.9]). In addition, an association between high steady-state C_trough and toxicity was observed. Nine patients with a mean (%CV) C_trough of 51.3 μg/ml (45.1) experienced grade ≥3 toxicity and subsequently required a dose reduction to 600 or 400 mg daily, even though the C_trough in these patients remained above the threshold (28.2 μg/ml, %CV 25.3). In addition, a decrease in plasma pazopanib exposure over time was observed which was already described for other TKIs [73,74]. This time-dependent decrease in pazopanib plasma exposure could be likely due to saturated absorption process and could be managed by dose escalation in later weeks as proposed in the algorithm. However, Mir et al. discussed that patients treated with pazopanib 800 mg q.d. with subtherapeutic exposure would benefit more if the twice daily regimen (400 mg b.i.d.) instead of dose increment was applied [75]. According to model-based simulations, twice daily regimen would result in increased AUC and C_trough by 59 and 75%, respectively [18]. Therefore, Yu et al. proposed that in patients with subtherapeutic exposure at 800 mg q.d., a twice daily regimen (400 mg b.i.d.) might potentially lead to higher exposure. However, it would require further validation in prospective clinical trials [18]. Finally, the feasibility of PK-guided dose individualization of pazopanib was demonstrated but still little is known about its clinical benefit; therefore, further investigations are needed in order to address this issue [18].

Individualized dosing recommendations

Recent studies reporting feasibility of PK-guided dosing strategy for pazopanib support its increasing use in routine practice. On the basis of the current data, we propose that PK-guided individualized dosing of pazopanib is based on the threshold steady-state C_trough of ≥20.5 μg/ml (Table 4) [45]. The algorithm proposed by Verheijen et al. seems the most adequate for dose adjustment and could be applied in patients with subtherapeutic concentrations of pazopanib. According to that algorithm, patients with a C_trough < 15.0 μg/ml could follow a dose escalation of 400 mg daily in case of no grade >2 AEs or 200 mg daily when grade 2 but not grade >3 AEs occur. In patients with a C_trough of 15.0–19.9 μg/ml a dose increase of 200 mg could be applied if no grade >3 toxicity occurs [72]. Blood samples should be drawn exactly 24 h after last pazopanib intake [63]. In addition, time-depended decrease in plasma exposure to pazopanib should be taken into account when performing TDM and should be managed by dose escalation in later weeks of the therapy [72]. As the twice daily dosing regimen was not evaluated in prospective clinical trials, its superiority over once daily dosing regimen is not certain. Finally, dried blood spot sampling could be used to accurately assess plasma concentrations of pazopanib which constitutes a more patient-friendly approach [76].

Table 4. Evidences and guidelines for therapeutic drug monitoring of tyrosine kinase inhibitors in metastatic renal cell carcinoma patients.

TKI	Interindividual variability (CV%)		Intraindividual variability (CV%)		Target exposure	Guidelines	Ref.
	Ctrough	AUC0–24	Ctrough	AUC0–24
Axitinib	NA	17–94%	20–22%	20–33%	NA	No guidelines based on pharmacological evidences
Pazopanib	11–90%	19–77%	NA	NA	Ctrough ≥20.5 μg/ml	Ctrough <15.0 μg/ml without grade >2 AE: ↗ 400 mg q.d./step Ctrough <15.0 μg/ml with grade 2 AE: ↗ 200 mg q.d./step 15 μg/ml > Ctrough < 19.90 μg/ml without grade > 3 AE: ↗ 200 mg q.d./step	[45,72,157]
Sorafenib	25–104%	12–117%	NA	31–47%	NA	No guidelines in mRCC patients
Sunitinib	34–59%	13–49%	NA	NA	50 ng/ml > composite Ctrough <100 ng/ml	Composite Ctrough <50 ng/ml without grade ≥3 AE: ↗ 12.5 mg q.d./step Composite Ctrough ≥50 ng/ml with grade ≥3 AE: ↗ 12.5 mg q.d./step	[29,56,124,157]

AE: Adverse event; AUC: Area under the concentration–time curve; b.i.d.: Twice daily; Composite C_trough: C_trough sunitinib + metabolite; C_trough: Minimum plasma concentration level; CV: Coefficient of variation; mRCC: Metastatic renal cell carcinoma; NA: Not available; q.d.: Once daily.

Sunitinib

Pharmacodynamics

Sunitinib (Sutent^®) was identified as a multiple receptor TKI targeting VEGFR-1, -2 and -3, PDGFR-α and -β, c-KIT, FLT3, CSF-1R and the glial cell-line derived neurotrophic factor receptor (RET) [77]. In vivo study showed that sunitinib had direct antitumor activity by inhibiting proliferation and survival of several human tumor xenografts in addition to antiangiogenic activity through its potent inhibition of VEGFR-2 and PDGFR-β signaling pathway. Inhibition of these two receptors was achieved with a sunitinib plasma level of 125–250 nmol/l (corresponding to 50–100 ng/ml) after a single oral dose of sunitinib [78].

Clinical data

A multicenter, randomized, Phase III clinical trial assessed clinical efficacy and safety of sunitinib (50 mg q.d., 4-week on, 2-week off) compared with IFN-α as first-line treatment in mRCC patients (n = 750 in each arm). Sunitinib treatment was associated with a longer median PFS than IFN-α (11 vs 5 months, respectively; HR: 0.42; 95% CI: 0.32–0.54; p < 0.001) [79]. Based on RECIST (Response Evaluation Criteria In Solid Tumors), patients treated with sunitinib had a significant higher ORR (47 vs 12%, in the sunitinib group and IFN-α group, respectively, p < 0.001). Median OS was 26.4 months in the sunitinib arm compared with 21.8 months in IFN-α arm (HR: 0.821; 95% CI: 0.673–1.001; p = 0.051) [80]. Since 2006, sunitinib is the standard first-line treatment for mRCC at a recommended daily dose of 50 mg during 4 weeks, followed by 2 weeks’ rest.

Safety profile

The Phase III trial reported that serious AEs (grade 3 and 4) occurred in 23.7% of patients under sunitinib therapy compared with 6.9% in patients receiving IFN-α. The most frequent sunitinib-related grade 3 AEs comprised: hypertension (12%), fatigue (11%), hand–foot syndrome (9%) and diarrhea (9%) [80]. These toxicities can led to treatment interruption or a dose reduction in 38 and 32% of patients, respectively, and to sunitinib discontinuation in 8% of patients [79].

Pharmacokinetics

Following oral administration, sunitinib is slowly absorbed [24] and C_max is observed after 6–12 h. Sunitinib bioavailability remains unknown in humans [25]. Its apparent volume of distribution (V/F) is large (2230 l), suggesting tissue distribution. Sunitinib is metabolized by CYP3A4 to its active metabolite N-desethyl sunitinib, known as SU12662, which is also metabolized by the same enzyme to inactive compounds [26,27]. SU12662 represents 23–37% of the total exposure. In vitro assays and studies in mice reported that these drugs were moderate affinity substrates of P-gp and BCRP [28]. Besides, sunitinib is also an inhibitor of P-gp, BCRP, MRP2 (encoded by ABCC2) and MRP4 (encoded by ABCC4) [81,82]. In vivo study showed that both P-gp and BCRP restrict brain accumulation of sunitinib. Concomitant administration of the dual P-gp/BCRP inhibitor elacridar could increase this accumulation [28]. However, different case reports showed response to sunitinib in RCC patients who had developed brain metastases [83,84]. Sunitinib and SU12662 are highly bound to plasma protein (95 and 90%, respectively), mostly to albumin. They exhibit linear PK since the C_max and AUC increase proportionally with sunitinib dose in the dose range of 25–100 mg [26,29]. Repeated daily administration leads to plasma accumulation of sunitinib and SU12662 of 3- to 4.5-fold and 7- to 10-fold, respectively [29,30]. They are primarily eliminated via feces with a limited renal excretion of unchanged drug and metabolites (<16%). Elimination half-lives are, respectively, 40–60 h and 80–110 h for sunitinib and SU12662, resulting in the achievement of steady-state within 14–21 days [26]. The PK characteristics of sunitinib are affected by circadian rhythms. A study carried out in 12 patients showed that C_trough after sunitinib administration at 1 or 6 p.m. were higher than when the drug was taken at 8 a.m. (66.0, 58.9 and 50.7 ng/ml, respectively; p = 0.006). However, no significant difference in the AUC was observed regardless the administration times [31]. Since sunitinib and SU12662 present similar pharmacological activity [26], the composite exposure (sunitinib plus SU12662 concentrations) reflects the total active drug exposure in plasma.

Factors of PK variability

Sunitinib exhibits large interindividual variability in composite exposure with %CV of 13–49 and 34–59% in AUC and C_trough, respectively [63]. Recently, Yu et al. developed a semiphysiological PK model to describe sunitinib and SU12662 PK. They reported an interindividual variability in CL/F for sunitinib and its metabolite of 34 and 42%, respectively [85].

Food intake

A Phase I study conducted in 16 healthy volunteers receiving a single dose (50 mg) of sunitinib after 10-h fast and after a high-fat, high-calorie meal showed that sunitinib exposure was slightly increased by food (ratios of fed/fasted geometric least-squares means: C_max: 104% and AUC_0-∞: 112%). Additionally, food decreased SU12662 C_max by 23% comparing to 10-h fasting, suggesting a delay in the formation and/or absorption of the metabolite, while AUC_0–∞ remained unaffected. Food has no significant effect on the bioavailability of sunitinib since the 90% confidence intervals were within the 80–125% bioequivalence range for both C_max and AUC. Thus sunitinib can be administered with or without food [86].

Renal impairment

PK studies did not report a relationship between CL/F and CL_CR for either sunitinib or SU12662 [26,87]. In a Phase I single-dose study, the PK of sunitinib was evaluated in subjects with normal renal function (CL_CR > 80 ml/min; n = 8), severe renal impairment (CL_CR < 30 ml/min; n = 8) and ESDR (CL_CR < 15 ml/min; n = 8) requiring hemodialysis. PK of sunitinib and SU12662 was not significantly affected by severe renal impairment. Although sunitinib and SU12662 exhibited lower C_max (by 10 and 20%, respectively) and AUC_0–∞ (by 14 and 24%, respectively) compared with subjects with normal renal function, these differences were not clinically relevant. Finally, no statistical difference was observed in elimination half-lives between the three groups. Respective C_max and AUC_0–∞ were lower by 38 and 47% for sunitinib and by 30 and 31% for SU12662 in ESRD patients compared with subjects with normal renal function, respectively (p < 0.05). However, exposure to sunitinib and SU12662 in hemodialysed patients was similar after hemodialysis, indicating that these compounds were not eliminated by the hemodialysis [88] which is consistent with a large V/F combined to high protein binding of sunitinib and SU12662. Four case reports and a study conducted in six mRCC patients under sunitinib with ESDR and severe renal impairment, showed similar safety than that observed in patients with normal renal function [89,90]. Based on all these data, no dose adjustment is recommended in patients with renal impairment or ESRD.

Hepatic impairment

A Phase I trial evaluated the effect of hepatic impairment on sunitinib and SU12662 PK. No significant difference in plasma composite exposure was reported in subjects presenting mild (Child-Pugh A; n = 8) or moderate (Child-Pugh B; n = 8) hepatic impairment compared with patients with normal liver function (n = 8) treated with a single dose 50 mg of sunitinib. Composite C_max was 26.0, 27.3, and 26.7 ng/ml and composite AUC_0–∞: was 1938, 2002, and 1999 ng.h/ml, in subjects with normal hepatic function and mild or moderate hepatic impairment, respectively [91]. No study has been carried out in subjects with severe hepatic impairment. Based on these data, no dose adjustment is needed in patients with mild or moderate liver impairment. Sunitinib is not recommended in patients with severe hepatic dysfunction.

Drug–drug interactions

Sunitinib exhibits pH-dependent solubility and is soluble over pH range of 1.2–6.8. Its solubility, as well as its absorption, decreases rapidly when pH exceeds 6.8. Little evidence has been published concerning potential interactions between sunitinib and proton pump inhibitors, but no effect is expected [92]. Nevertheless, a retrospective study conducted in 231 mRCC patients, showed a significant difference in PFS and OS between patients receiving gastric acid suppressing agents and patients not taking any, suggesting an effect of acid suppressing agents on sunitinib bioavailability [93]. Further investigations are warranted to confirm this potential interaction. Considering that sunitinib is mainly metabolized by CYP3A4/5, drug interactions with inhibitors and inductors of CYP3A4 were widely studied [94]. The effect of ketoconazole and rifampicin on PK of sunitinib was evaluated in healthy volunteers. After a single dose of sunitinib co-administrated with ketoconazole, composite C_max and AUC_0–∞ were increased by 49 and 51%, respectively (Table 2). Concomitant administration of rifampicin resulted in a decrease in the composite C_max and AUC_0–∞ values of 23 and 46%, respectively [26]. In two patients co-treated with sunitinib and mitotane, a very strong CYP3A4 inducer, sunitinib exposure was decreased compared with seven patients who did not receive mitotane (267 and 268 μg.h/l vs 1344 μg.h/l) [95]. In a case report, a concomitant administration of nicardipine, a moderate CYP3A4 inhibitor, resulted in an increase in composite C_trough by approximately 50% [96]. Grapefruit consumption led to a negligible increase in the relative bioavailability of sunitinib compared with midazolam (11 vs 50%, respectively) and, therefore, was considered not clinically relevant [97]. Docetaxel did not significantly affect sunitinib exposure, irrespective of docetaxel dose and administration schedule [98]. Additionally, mTOR inhibitors (sirolimus and everolimus) did not change plasma exposure of sunitinib (AUC and C_max) [99,100]. Yet, the association of sunitinib with everolimus was responsible of severe toxicities requiring dose reduction of both drugs [100]. In a study combining sunitinib and gefitinib, no PK drug–drug interactions were reported but the maximum tolerated dose of sunitinib was 37.5 mg/day [101].

Biological & demographic variables

Age does not affect PK of sunitinib and its active metabolite [26,102]. A meta-analysis using a population approach, based on data from 14 clinical studies (n = 590 subjects), identified significant covariates contributing to variability in PK and thus in plasma exposure to sunitinib and SU12662. Among the different demographic parameters analyzed, tumor type, gender, ethnicity, total body weight and Eastern Cooperative Oncology Group (ECOG) scores have been reported as factors of variability in CL/F, whereas gender and body weight described in part the interindividual variability in Vc for both sunitinib and SU12662 [87]. However, these covariates have a negligible influence on composite plasma exposure. More recently, it has been shown that lean body mass (LBM) could be responsible for 23.5% of the interindividual variability in composite AUC. Indeed, patients with a low LBM present higher plasma exposure, which supports more frequent sunitinib-related dose-limiting toxicities in patients with sarcopenia [103]. Finally, Chae et al. showed that body surface area affected clearance of sunitinib in 31 Asian mRCC patients [104].

Genetic polymorphisms

Polymorphisms in genes encoding for metabolizing enzymes (CYP3A4, CYP3A5) or efflux transporters (P-gp and BCRP) could affect absorption, metabolism or elimination of sunitinib and SU12662 [105]. CYP3A5 expression is different between and within ethnic groups. CYP3A5 is expressed in approximately 10–25% of Caucasians, 30–50% of Asian and South Americans and 55–95% of African–Americans [106,107]. The presence of allele CYP3A5*1 (presence of an A allele on rs776746) results in expression of the CYP3A5 enzyme, which converts sunitinib to SU12662. A study conducted in 114 patients with solid tumors treated with sunitinib investigated the relationship between SNPs in genes encoding for CYP3A4 and CYP3A5 and clearance of sunitinib and SU12662 [108]. A 22.5% decrease in the CL/F of sunitinib was reported in CYP3A4*22 (rs35599367) carriers corresponding to decreased enzyme activity, compared with wild type patients (p = 0.01). Inversely, a 22.5% increase in CL/F of sunitinib (p = 0.05) and a 21.5% higher CL/F of SU12662 was observed in carriers of a CYP3A5*1 allele compared with GG carriers with entire absence of CYP3A5 (p = 0.04). However, among the 14 tested SNPs neither reached the expected significance threshold (p < 0.0005), hence none was considered to affect significantly CL/F of sunitinib or SU12662 [108]. In a study conducted in Asian mRCC patients, CC genotype for ABCB1 (3435 C > T) was associated with an increased sunitinib exposure (76.81 vs 56.55 ng/ml; p = 0.03), higher risk of all-grade rash (Relative risk = 3.00, 95% CI: 1.17–7.67) or mucositis (Relative risk = 1.60, 95% CI: 1.10–2.34) and disease progression comparing to the CT/TT genotype, responsible for increased P-gp activity [109]. However, no such relationships were observed with the CYP3A5 polymorphism. Moreover, ABCB1 CT/TT genotype affected sunitinib clearance in 31 Asian mRCC patients (effect size 31.14%; p = 0.006) [104]. Finally, different studies showed that ABCG2 (421 C > A) was responsible for increased composite exposure (expressed as AUC) in cancer patients [47,57].

Van Erp et al. investigated the association between SNP of genes encoding for metabolizing enzymes or transporters and the development of toxicities in 219 patients with solid tumors receiving sunitinib as monotherapy [110]. The presence of a copy of TT allele in the ABCG2 haplotype (15622 C > T, 1143 C > T) was associated with emergence of any grade ≥ 2 toxicity (Odds ratio 2.63; p = 0.016) and subjects with ABCB1 TTT haplotype (3435 C > T, 1236 C > T, 2677 G > T) had higher prevalence of hand-foot syndrome (Odds ratio 2.56; p = 0.035). The G allele in CYP1A1 (2455 A > G) resulted an increased risk for both leukopenia and mucosal inflammation (Odds ratio 6.24; p = 0.029 and 4.03; p = 0.021, respectively). Van der Veldt et al. reported in another study that both CYP3A5*1 allele and ABCB1 TGC haplotype were associated with longer PFS (HR: 0.27; p = 0.032 and HR: 0.52; p = 0.033, respectively) [111]. The CYP3A5*1 allele was also reported to be associated with an increased risk of dose reductions due to toxicity (HR: 3.75; p = 0.022) [112]. Some of these results were confirmed in a large cohort of 333 mRCC patients: CYP3A5*1 allele was associated with dose reductions (Odds ratio = 2.0; p = 0.039) due to toxicity and ABCB1 CGT haplotype was associated with an increased PFS (HR: 1.9; p < 0.001) [113].

PK/PD relationship

Based on preclinical and retrospective clinical trials, a target of composite C_trough in the range of 50–100 ng/ml was suggested. Indeed, Mendel et al. showed in vivo that an inhibition of VEGFR-2 and PDGFR-β was obtained when plasma concentration of sunitinib reached or exceeded 50–100 ng/ml [78]. In a Phase I study including 28 patients with advanced cancer, Faivre et al. reported that most of OR were observed with a composite C_trough ≥ 50 ng/ml [29]. Narjoz et al. demonstrated in 55 mRCC patients that a composite AUC > 1973 ng.h/ml after 21 days of sunitinib treatment was associated with twice longer OS (Table 3) [47]. Additionally, a meta-analysis based on six different clinical trials showed that a composite AUC ≥ 800 ng.h/ml at steady-state was associated with improved clinical outcomes such us longer time to progression and a longer OS (Table 3) [46]. Finally, according to a prediction model, patients with composite AUC ≥ 1500 ng.h/ml would have 70% probability to achieve an OR [46].

A Phase I study recorded more frequent dose-limiting toxicities with a composite C_trough > 100 ng/ml at steady state in mRCC patients [29]. A retrospective study of 21 patients with RCC reported more frequent grade ≥ 3 sunitinib-induced toxicities (thrombocytopenia, anorexia and fatigue) in patients with a composite C_trough ≥ 100 ng/ml than those with composite C_trough < 100 ng/ml (Table 3). In patients who developed grade ≥ 3 AE a dose reduction or a treatment discontinuation was performed, resulting in a shorter time to treatment failure and a shorter PFS (median time to treatment failure, 71 vs 590 days; p = 0.04; median PFS, 238 vs 748 days, p = 0.02 in patients with composite C_trough ≥ 100 ng/ml vs those with composite C_trough < 100 ng/ml, respectively) [56]. Houk et al. showed in a meta-analysis a relationship between sunitinib exposure (composite AUC_0–24) and AEs such as fatigue, hypertension and neutropenia [46]. In 36 Asian mRCC patients, receiving attenuated dosing of sunitinib (37.5 mg q.d. in a 4/2 schedule), occurrence of grade ≥ 2 mucositis and altered taste were associated with higher composite exposure compared with occurrence of grade 1 or no toxicity (Table 3) [55]. Narjoz et al. reported that composite AUC_0–24 at day 21 was independently associated with any grade ≥ 3 acute toxicity (Odds ratio 1.16; 95% CI: 1.05–1.28; p = 0.005) [47]. Grade ≥2 thrombocytopenia was associated with higher SU12662 exposure (Odds ratio 1.27; 95% CI: 1.03–1.57; p = 0.028) in mRCC patients [47] and was also more frequent in Japanese patients with a sunitinib C_through ≥ 90 ng/ml (Table 3) [57]. Moreover, QT interval prolongation was associated with composite exposure in 24 patients [58]. Finally, Kloth et al. reported that patients with any grade 3 AE had a significantly lower CL/F of sunitinib than patients without grade 3 toxicities and that there was a positive relationship between composite C_trough levels and the occurrence of fatigue (Table 3) [59].

Individualized dosing recommendations

The actual dose regimen for sunitinib is an oral daily dose of 50 mg during 4 weeks, followed by a 2-week rest period (4/2 dosing schedule) [26]. However, considering that sunitinib-related toxicities are responsible for frequent dose reductions or discontinuations, alternative dosing schedules were evaluated in order to decrease sunitinib toxicity without affecting treatment efficacy. Continuous 37.5 mg daily dosing schedule showed similar results when compared with classic schedule with a tendency of longer time to progression for the latter [114]. Furthermore, 2/1 schedule (50 mg q.d., 2-week on, followed by 1-week off treatment) has been assessed in various studies [115–122]. In total, patients under 2/1 schedule presented lower risk of AE and dose reductions than patients treated with the standard dose regimen, with similar efficacy. These results were supported by a PK/PD modelling study which predicted that efficacy would be comparable on both dosing schedules, though a lower incidence of grades 3 and 4 thrombocytopenia would be observed with 2/1 dosing schedule. The model also predicted that plasma exposure (C_trough) to sunitinib and its active metabolite would be similar with the 2/1 and 4/2 schedules [123]. The 2/1 schedule is increasingly used in clinical practice and several prospective randomized trials are ongoing in order to determine its safety and efficacy in mRCC patients (NCT02689167 and NCT02060370).

To date, amongst TKIs indicated for treatment of mRCC, sunitinib presents the highest level of evidence for TDM. The target composite C_trough concentration should be comprised between 50 and 100 ng/ml for the recommended sunitinib dosing with 4-weeks on/2-weeks off schedule (Table 4). Furthermore, Lankheet et al. showed in a pilot study the feasibility of safe PK-guided dose escalation of sunitinib [124]. For a continuous daily dosing of 37.5 mg sunitinib, a composite C_trough ≥ 50 ng/ml should be achieved at steady-state. The dose can be lowered by 12.5 mg if any grade ≥3 toxicity appears, and sunitinib dose can be increased by 12.5 mg if the C_trough target is not achieved and the patient does not suffer from any grade ≥3 AE [124].

Sorafenib

Pharmacodynamics

Sorafenib (Nexavar^®) is a multiple TKI targeting Raf kinase, VEGFR-1, -2 and -3, PDGFR-β, FLT-3, c-Kit and RET receptor tyrosine kinases [125]. In a number of xenograft models, sorafenib inhibited tumor cell proliferation and angiogenesis [126].

Clinical data

In a Phase III trials evaluating clinical efficacy of sorafenib 400 mg b.i.d. versus placebo in patients with RCC resistant to previous standard therapy, patients in sorafenib group had significantly longer PFS than patients in placebo group (median PFS: 5.5 vs 2.8 months; HR: 0.44; 95% CI: 0.35–0.55; p < 0.01 for sorafenib and placebo group, respectively) [127]. Sorafenib is approved in the USA and in several European countries for the treatment of advanced RCC in patients who progressed on prior IFN-α- or IL-2-based therapy at the recommended dose of 400 mg b.i.d.

Safety profile

As reported in Phase III clinical study, diarrhea, rash, fatigue and hand–foot skin reactions are the most common AEs associated with sorafenib [127–129]. Toxicity led to sorafenib discontinuation in 10% of patients, 21% of patients had to interrupt sorafenib for a median duration of 7 days and 13% of patients had a dose reduction. The reasons of treatment discontinuations were mostly constitutional, gastrointestinal, dermatologic or pulmonary–upper respiratory tract AEs.

Pharmacokinetics

Sorafenib is administered orally at the recommended dose of 400 mg b.i.d. Its mean relative bioavailability is 38–49% when compared with oral solution, the absolute bioavailability is unknown [130]. Plasma exposure to sorafenib increased less than proportionally with increasing doses in the range 50 mg every fourth day to 600 mg b.i.d. [20]. Mean AUC_0–12, C_max and T_max of sorafenib 400 mg b.i.d. observed at day 1 were 21.8 μg.h/ml, 2.9 μg/ml and 2.9 h, respectively, and they increased to 47.8 μg.h/ml, 5.4 μg/ml and 12.1 h, respectively, at day 28 [20]. The elimination half-life is 25–48 h, therefore steady-state plasma concentrations are achieved approximately 7 days after start of treatment. After multiple dosing at 400 mg b.i.d., the accumulation ratio of sorafenib in plasma was 2.87. As indicate data from in vitro studies, sorafenib is highly bound to plasma proteins (>99.5%), mainly to albumin. Using equilibrium dialysis method, the unbound plasma fraction (f_u) of sorafenib in both cancer patients and healthy volunteers was found to be 0.3% [131]. Sorafenib is a substrate for BCRP and P-gp [132] which could contribute to its decreased intestinal absorption. In vivo experiments in mice showed that BCRP plays the major role in efflux of sorafenib at the blood–brain barrier. Use of dual P-gp and BCRP inhibitors could increase brain accumulation of sorafenib in patients with cerebral metastases. Sorafenib undergoes two metabolic pathways: oxidation by CYP3A4 and glucuronidation by UGT1A9 [133]. The main metabolite in plasma, N-oxyde sorafenib, accounting for 9–16% of circulating analytes at steady-state, showed similar pharmacologic activity as the parent compound [130]. The main route of sorafenib elimination is hepatobiliary excretion as approximately 50% of the sorafenib oral dose in recovered in the feces in unchanged form. Approximately 20% of the dose was recovered in urine as metabolites. Renal excretion is negligible. Time-dependent decrease in sorafenib plasma exposure was first observed in 25 patients treated for hepatocellular carcinoma (HCC) (median dose-normalized AUC_0–12: 102 μg.h/ml on day 15 vs 63 μg.h/ml on day 120) [52]. Two other studies, both in HCC and RCC patients, reported similar findings [51,73]. The reason of this time-dependent decrease in plasma exposure remains currently unknown. However, the hypothesis of autoinduction of CYP3A4-mediated metabolism of sorafenib can be excluded because a significant decrease in plasma exposure appears 3 months after start of treatment.

In the population analysis, the PK of sorafenib was described by a one-compartment model with saturated absorption, first-order intestinal loss and linear elimination [134]. Wide dose range analyzed in this study (400–2400 mg/day) allowed estimating the median bioavailability which was approximately 2.5-fold lower for 1600 mg/day than for 400 mg/day (51 and 20%, respectively). Model-based simulations showed that dividing daily dose (>800 mg/day) into three or four administrations would result in higher plasma exposure in nonresponders treated with 800 mg/day (400 mg b.i.d.). In another study, PK of sorafenib was described by a one-compartment model with four transit absorption compartments and enterohepatic circulation [21]. The model estimated moderate-to-high interindividual variability on CL/F, V/F and absorption rate constant (ka) (%CV 18–68%). The interoccasion variability on CL/F was 48% which suggests that CL and F may vary between occasions due to drug–drug interactions, food intake or changes in solubility.

Factors of PK variability

Similar to other TKIs, the interindividual variability in sorafenib PK is high (12–117% for AUC and 25–104% for C_trough) [63] and the intraindividual variability is moderate (31–47% for AUC), which supports the application of TDM to guide sorafenib therapy. Multiple factors influencing the PK of sorafenib are discussed below.

Food intake

In a Phase I study, food intake had no relevant impact on sorafenib PK [22]. However, conflicting results have been reported and according to the EMA, sorafenib should be administered without food or with a low-fat or moderate-fat meal [130].

Hepatic impairment

The impact of hepatic impairment on PK and tolerability of sorafenib was evaluated in 72 cancer patients with normal liver function or mild (bilirubin > ULN but ≤ 1.5× ULN and/or AST > ULN), moderate (bilirubin > 1.5× ULN to ≤ 3× ULN and any AST), severe (bilirubin > 3–10× ULN and any AST) and very severe liver dysfunction (albumin <2.5 mg/dl, any bilirubin and any AST) [135]. No significant association between plasma exposure (AUC) to sorafenib or its N-oxyde metabolite and Child-Pugh score or bilirubin was found. The recommended dose of 400 mg b.i.d. was not tolerated in patients with moderate and severe hepatic dysfunction. Therefore, in patients with moderate hepatic dysfunction a starting dose of 200 mg b.i.d. is recommended and in patients with very severe liver impairment, 200 mg/day. Patients with severe liver dysfunction did not tolerate sorafenib at 200 mg every third day.

Renal impairment

Urinary excretion constitutes minor route of sorafenib elimination with <20% of the dose recovered in urine [130]. Nevertheless, a Phase I study was carried out to investigate the impact of renal impairment on plasma sorafenib exposure in patients with normal renal function (CL_CR ≥ 60 ml/min), pre-existing mild (CL_CR 40–59 ml/min), moderate (CL_CR 20–39 ml/min) and severe renal impairment (CL_CR < 20 ml/min) and in patients undergoing hemodialysis [135]. There was no association between CL_CR and AUC of sorafenib or N-oxyde sorafenib in 54 studied patients. Evaluation of the dose-limiting toxicity allowed proposing starting dose of sorafenib according to renal function: 400 mg b.i.d. is recommended in patients with mild renal impairment, 200 mg b.i.d. in patients with moderate renal impairment, 200 mg q.d. in patients undergoing hemodialysis. The recommended dose could not be evaluated in patients with severe renal impairment because of insufficient number of patients in that group (n = 4). However, authors emphasize that subsequent dose escalation can be performed according to individual tolerance profile. In addition, based on clinical data from 32 mRCC patients including 14 patients with renal impairment (CL_CR 32–60 ml/min) treated with sorafenib at 400 mg q.d., the incidence of diarrhea and hand-foot syndrome were higher in patients with renal dysfunction (57 vs 33% and 86 vs 56%, respectively) [136]. Dose reductions and dose interruptions were also more frequent in patients with renal dysfunction group (57 vs 28% and 43 vs 22%, respectively). Finally, clinical reports confirmed that sorafenib can be used at a reduced dose in mRCC patients with ESRD [137] and in patients undergoing hemodialysis [138–140]. However, careful monitoring of these patients is needed as they present higher incidence of AEs.

Drug–drug interactions

Concomitant administration of CYP3A4 inducers such as rifampicin or enzyme inducing antiepileptics has been reported to significantly decrease plasma exposure to sorafenib [141]. In a case report, Noda et al. documented an increase in sorafenib CL under treatment with prednisolone (Table 2) [142]. Once prednisolone dose was tapered, patient developed grade 3 AE due to an increase in sorafenib plasma exposure which finally required a dose reduction from 400 mg b.i.d. to 200 mg b.i.d. Lathia et al. reported a study evaluating impact of ketoconazole on plasma exposure of single 50-mg dose of sorafenib in 16 healthy volunteers [133]. Mean plasma sorafenib AUC_0–∞ was similar when administered with or without ketoconazole (geometric mean AUC_0–∞: 9.82 and 11.04 μg.h/ml, respectively). The mass–balance study demonstrated that the glucuronidation pathway via UGT1A9 plays a significant role in sorafenib elimination. Therefore, the authors conclude that the risk of PK interactions with CYP3A4 inhibitors seems to be unlikely in clinical settings. However, Gomo et al. documented a twofold increase in sorafenib exposure after the concomitant administration of CYP3A4 substrate felodipine in a 80-year-old patient with HCC [143]. This interaction may be due to a deficiency in sorafenib metabolic pathways related to patient’s age and his HCC and not to the inhibitory effect of felodipine. Concomitant use of neomycin decreased plasma exposure of sorafenib by 54% in healthy volunteers [130]. Gastrointestinal flora with glucuronidase activity contribute to hydroxylation of conjugated forms of sorafenib and its subsequent enterohepatic recycling. Therefore, use of antibiotics eradicating gastrointestinal flora could decrease the efficacy of sorafenib.

In vitro studies showed that sorafenib is a moderate inhibitor of CYP3A4, CYP2C19 and CYP2D6 [144] as well as UGT1A1 inhibitor [145]. A PK drug–drug interaction study was carried out in 28 cancer patients in order to evaluate the impact of sorafenib on PK of midazolam, omeprazole and dextromethorphan, substrate for CYP3A4, CYP2C19 and CYP2D6 enzymes, respectively [144]. No significant changes in their plasma exposure were observed when sorafenib was co-administered. Therefore, concomitant use of CYP3A4, CYP2C19 and CYP2D6 substrates with sorafenib is possible. Because of possible inhibition of glucuronidation via UGT1A1 and UGT1A9 by sorafenib, caution should be taken when sorafenib is administered with drugs predominantly metabolized by these enzymes [130].

Biological & demographic variables

In two population analyses, the tested covariates (age, gender, body weight, body surface area, serum creatinine, AST, bilirubin, kidney function parameters, cholesterol and blood urea nitrogen) did not explain the interindividual variability in PK of sorafenib [21,134]. Moreover, no association between sorafenib AUC and bilirubin was found in the study of Miller et al. [135].

CL of sorafenib and other TKIs is restrictive, thus it depends on the unbound fraction (f_u) of the drug in plasma. Increased f_u of sorafenib would increase its CL and, therefore, result in decreased total plasma concentration. The unbound sorafenib concentration, which is responsible for the pharmacological activity, would remain unchanged. Therefore, the interindividual variability in plasma exposure to sorafenib may be a result of variations in plasma protein binding. Hypoalbuminemia related to severe renal and hepatic impairment or denutrition is often observed in cancer patients. Multivariate analysis carried out in 54 cancer patients showed that albuminemia was the single parameter independently associated with sorafenib CL (p = 0.0036) [146]. According to in vitro studies, sorafenib f_u would increase 1.7-fold as albuminemia decreased from 45 g/l to 30 g/l. On the contrary, another study reported that neither pretreatment albumin nor α₁-acid glicoprotein was predictable of sorafenib f_u [131]. Overall, hypoalbuminemia should be taken into account as a possible factor of low total plasma exposure to sorafenib.

It has been reported that HCC patients with severe depletion of skeletal muscle (sarcopenia) had significantly higher steady-state AUC_0–12 of sorafenib and higher incidence of AEs [147]. These findings were confirmed in 55 mRCC patients treated with sorafenib where sarcopenia both with BMI were predictors of dose-limiting toxicities [148]. The evidence of exposure-toxicity relationship for sorafenib suggests that sarcopenia and BMI < 25 kg/m² could be related to high plasma sorafenib concentrations. Finally, a study carried out in 54 solid cancer patients revealed that median dose-normalized AUC_0–12 was 2.1-fold greater in female than in male patients (136.0 vs 64.8 μg.h/ml, respectively; p = 0.0008) [50]. However, the greater exposure in female patients may be due to lower LBM in these patients comparing to male patients (45.6 vs 55.6 kg, respectively; p = 0.0001) which supports a higher incidence of dose-limiting toxicities in patients with sarcopenia.

Genetic polymorphisms

Although the population PK analysis did not report any correlation between CYP3A4*1B, CYP3A5*3C, UGT1A9*3 and UGT1A9*5 genotypes and sorafenib PK in 111 cancer patients [21], the impact of UGT1A1*28 allele cannot be excluded. Peer et al. reported that patients carrying UGT1A1*28 allele, and possibly UGT1A9*3 allele, are at an increased risk of high plasma sorafenib concentrations, as well as a greater incidence of hand-foot syndrome [145]. Boudou-Rouquette et al. reported no association between genetic variants of CYP3A5, UGT1A9, ABCB1 and ABCG2 [50]. However, UGT1A9 polymorphism (2152 C > T), known to induce UGT1A9 expression, was associated with grade ≥2 diarrhea (p = 0.015). Lower plasma sorafenib exposure was observed in HCC patients heterozygous for ABCB1 (3435 C > T) (P-gp genotype) and ABCG2 (34 G > A, 1143 C > T) (BCRP genotype) polymorphisms comparing to homozygous patients (wild-type or mutant) [149]. These findings demonstrate that the functional polymorphisms of P-gp and BCRP affect their expression and activity in cancer patients. In a recent study carried out in Chinese mRCC patients, ABCB1 (2677 G > T/A, 3435 C > T) genetic variants were significantly associated with increased risk of hand-foot syndrome and hypertension and UGT1A1*6 allele was associated with hypertension [54].

PK/PD relationship

To date, little is known about the association of sorafenib PK with clinical efficacy in mRCC. The single association between plasma sorafenib exposure and clinical efficacy (disease control rate and PFS) was documented in melanoma patients [150]. Multiple studies reported exposure-toxicity relationship of sorafenib (Table 3). In a study evaluating 83 patients with solid tumors treated with sorafenib in routine practice, median AUC_0–12 of sorafenib was significantly higher in patients developing grade 3–4 AEs than in other patients (61.9 vs 53.0 μg.h/ml; p = 0.017, respectively) [52]. The incidence of hypertension and hand-foot syndrome significantly decreased over time which might be related to time-dependent decrease in exposure to sorafenib. Another study also reported an association between the severity of skin rash and plasma AUC_0–12 of sorafenib (p = 0.02) [151]. Fukudo et al. demonstrated that patients developing grade ≥2 hand-foot skin reaction and hypertension had significantly higher sorafenib concentrations than remaining patients (p = 0.0045 and p = 0.0453, respectively) [51]. The optimal cut-off concentrations estimated by receiver operating characteristic curve for the prediction of grade ≥2 hand-foot syndrome and hypertension in Japanese patients were 5.78 μg/ml and 4.78 μg/ml, respectively. In addition, higher risk of developing AEs was observed in female patients which might be related to higher plasma exposure in those patients. The authors hypothesize that TDM might be useful to prevent grade ≥2 hand-foot skin reactions by maintaining plasma sorafenib concentrations below 5.78 μg/ml in Japanese patients. These threshold values cannot be extrapolated to Caucasian patients. On the other hand, plasma concentrations above 4.78 μg/ml associated with grade ≥2 hypertension might result in better clinical efficacy. Cumulated sorafenib AUC between day 0 and day 30 of 3161 μg.h/ml was a significant threshold for developing any grade ≥3 toxicity in 54 solid tumors patients (p = 0.018) [50]. Finally, Mai et al. reported that plasma sorafenib C_trough of >10.0 μg/ml in Chinese mRCC patients was associated with severe AEs [54].

Individualized dosing recommendations

According to current data in the literature, no PK target to guide sorafenib therapy in patients with mRCC is available. However, the relationship between PK of sorafenib and incidence of AEs and the time-dependent decrease in plasma exposure suggest that the monitoring of plasma concentrations is of potential interest in optimizing sorafenib therapy. Boudou-Rouquette et al. reported that TDM in nonselected patients with differentiated thyroid cancer contributed to increased PFS in those patients comparing to the Phase III DECISION trial (26.0 vs 10.8 months, respectively) [152]. The nonlinear PK of sorafenib above the recommended dose of 400 mg b.i.d. due to saturated absorption in the intestinal gut, might be a limiting factor of the dose individualization. Hornecker et al. reported that patients treated with 400 mg b.i.d. who could undergo dose escalation would benefit more from the increased plasma exposure if the daily dose (>800 mg/day) was divided into three or four administrations [134]. Henin et al. showed that the risk of hand–foot syndrome was increased with daily dose fractioning [153], which supports the significant increase in sorafenib plasma exposure and, therefore, a potential clinical interest of a fractioned dosing regimen [154]. This option should be taken into account when considering sorafenib dose increase over 800 mg/day.

Conclusion

Recent findings about PK/PD relationships of TKIs suggest that achieving a target plasma exposure would improve the clinical outcome which further supports use of TDM to individualize the therapy. Regarding TKIs used in the treatment of mRCC, plasma exposure targets have been reported for pazopanib (steady-state C_trough > 20.0 μg/ml) and sunitinib (composite steady-state C_trough > 50 ng/ml for continuous dosing at 37.5 mg q.d.).

The feasibility of TDM in patients treated with sunitinib was demonstrated in a prospective trial reported by Lankheet et al. [124] and was further confirmed in another study evaluating feasibility of TDM for both pazopanib and sunitnib in clinical practice [155]. Considering pazopanib, a prospective, randomized crossover study performed by de Wit et al. [63] showed that application of TDM did not result in a decrease in interpatient variability in plasma concentrations of pazopanib and that the percent of patients achieving target plasma exposure in PK-guided dosing group was similar to that in fixed dosing group. However, a recent prospective study reported by Verheijeen et al. [72] demonstrated the feasibility of TDM for pazopanib in routine practice. Further PK/PD studies in mRCC patients are needed for axitinib and sorafenib in order to establish a target exposure which would allow performing TDM in the future.

Before TDM is performed routinely in hospital laboratories, its utility in optimizing the therapy should be confirmed in prospective randomized trials comparing patients undergoing PK-guided doing with those on standard dosing. Considering the clinical benefit of TDM for TKIs in mRCC, it has not yet been evaluated in prospective trials. The benefice of TDM for TKIs was only demonstrated for imatinib in patients with chronic myeloid leukemia in the OPTIM study [156]. Nevertheless, application of PK-guided dosing of TKIs in mRCC patients could be relevant at the moment of disease progression, apparition of a severe AE or in order to monitor the compliance. Similarly, TDM could be particularly helpful in the treatment of patients presenting factors of increased risk of developing severe toxicities (elderly, sarcopenia, renal or hepatic impairment, treatment with strong CYP-inhibitors or inducers).

Future perspective

To date, most of the data on exposure–response relationships of TKIs come from retrospective analyzes with limited number of patients. The increasing evidence for potential use of PK targets to optimize the therapy encourages performing PK/PD studies during drug development. Therefore, PK-guided dosing could be implemented in clinical practice shortly after the drug is commercialized. Moreover, more patient-friendly approaches could be developed such as dried blood spot sampling. The increasing interest of performing TDM for TKIs among clinicians will contribute to implementation of this approach into standard clinical procedures in the hospital Table 2 laboratoriesTable 4.

Executive summary

• Development of oral targeted therapies such as tyrosine kinase inhibitors (TKIs) highly improved treatment of various cancers providing better clinical efficacy with less toxic effects comparing to conventional chemotherapy.

• High interindividual variability in terms of efficacy and toxicity in patients treated with TKIs is observed. Evidence of relationship between plasma exposure (defined as area under the concentration-time curve [AUC] or trough concentration [C_trough]) and clinical efficacy has been reported for many TKIs, and therapeutic drug monitoring (TDM) is now increasingly used in clinical practice to guide dose individualization of certain TKIs which improves the therapeutic response and decreases the toxicity in cancer patients.

• Main TKIs used in the treatment of metastatic renal cell carcinoma – axitinib, pazopanib, sunitinib and sorafenib – as other TKIs, exhibit high interpatient variability in pharmacokinetics (PK) which could be due to altered absorption process after food intake, hepatic or renal impairment, genetic polymorphisms of metabolizing enzymes or drug–drug interactions. High interpatient variability of plasma exposure together with narrow therapeutic window, dose-proportional PK in the range of recommended doses and repeated dosing make of these TKIs candidate drugs for TDM. However, before TDM is applied in clinical practice, an exposure–response relationship has to be demonstrated and a target of exposure needs to be validated in prospective clinical trials.

• To date, an exposure–response relationship has been demonstrated for pazopanib and sunitinib and feasibility of TDM for these drugs has been demonstrated in prospective trials. Therefore, TDM is increasingly used in hospital laboratories to perform PK-guided dosing for these drugs based on threshold values of steady-state C_trough > 20.0 μg/ml for pazopanib and composite (sunitinib + SU12662) steady-state C_trough > 50 ng/ml for continuous dosing at 37.5 mg q.d. for sunitinib. The level of evidence to apply TDM in clinical practice is still lacking for axitinib and sorafenib. Prospective trials are warranted in order to establish a PK target for axitinib and sorafenib to perform PK-guided dose individualization.

• A key step before TDM is applied in clinical practice is to demonstrate its feasibility and clinical benefice in prospective randomized trials in which patients undergoing PK-guided dosing are compared with those under standard dosing regimen. Nevertheless, monitoring of plasma concentrations of TKIs may be of interest in patients at the moment of disease progression, apparition of a severe AE and in order to verify the compliance. Finally, TDM could be particularly helpful in the treatment of patients presenting factors of increased risk of developing severe toxicities (elderly, sarcopenia, renal or hepatic impairment, treatment with strong CYP-inhibitors or inducers).

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.