Pharmacogenetics, enzyme probes and therapeutic drug monitoring as potential tools for individualizing taxane therapy

Abstract

The taxanes are a class of chemotherapeutic agents that are widely used in the treatment of various solid tumors. Although taxanes are highly effective in cancer treatment, their use is associated with serious complications attributable to large interindividual variability in pharmacokinetics and a narrow therapeutic window. Unpredictable toxicity occurrence necessitates close patient monitoring while on therapy and adverse effects frequently require decreasing, delaying or even discontinuing taxane treatment. Currently, taxane dosing is based primarily on body surface area, ignoring other factors that are known to dictate variability in pharmacokinetics or outcome. This article discusses three potential strategies for individualizing taxane treatment based on patient information that can be collected before or during care. The clinical implementation of pharmacogenetics, enzyme probes or therapeutic drug monitoring could enable clinicians to personalize taxane treatment to enhance efficacy and/or limit toxicity.

Taxanes are commonly used in the treatment of various solid tumors, such as breast cancer, non-small-cell lung cancer and prostate cancer. They work by binding to the microtubule β-tubulin subunit and interfering with microtubule depolymerization, which inhibits cell division [1,2]. The first taxane, paclitaxel, was approved by the US FDA in 1992 [201], followed by docetaxel 4 years later [202]. The search for improved formulations and new taxanes has resulted in FDA approval of albumin-bound paclitaxel [3] and cabazitaxel [4], and continues to this day [5,203].

Despite the initial success and continued optimization of taxanes, their use is associated with serious limitations. Taxanes have a narrow therapeutic window and a broad adverse event profile, of which hematopoietic and neurologic toxicities are most notable. Although the overall profiles are generally similar, the incidence of specific toxicities is quite different between the two agents and the dose-limiting toxicities of paclitaxel and docetaxel, neurotoxicity [6] and (febrile) neutropenia [7], respectively, are not the same.

Currently, taxane dosing is based primarily on body surface area (BSA) and in some cases diminished liver function is considered [8,9]. Using this approach to initial dose selection, large interindividual differences in pharmacokinetics (PK) are observed [10,11] and a subpopulation of the treated population experiences severe, treatment-limiting adverse events.

The taxanes are primarily eliminated via hepatic metabolism and biliary elimination (Figure 1) [12-14]. The proteins involved in hepatic uptake and intracellular metabolism influence taxane PK and may have an indirect role in determining taxane outcome. Taxanes enter the hepatocytes through rapid passive diffusion, due to their lipophilic character, and by hepatocellular uptake by OATP1B3 [15,16]. Both taxanes are substrates for the highly promiscuous CYP3A4/3A5 metabolic system [17,18], but the formation of paclitaxel metabolites is also highly dependent on CYP2C8 activity [19]. The efflux proteins P-gp and MRP2 are responsible for excreting taxanes and their metabolites into the bile [20-22].

Figure 1. Intracellular metabolism of docetaxel and paclitaxel.

Dashed lines indicate passive diffusion. The short series of arrows before the formation of the docetaxel metabolites (M1, M2, M3 and M4) designates the possible involvement of multiple enzyme reactions not presented in this figure.

6αOH-pac: 6α-hydroxypaclitaxel; diOH-pac: 6α-p-3′-dihydroxypaclitaxel; p3OH-pac: p-3′-hydroxypaclitaxel.

Currently, there are no methods for identifying patients who are at high risk of toxicity before treatment, thus all patients receiving taxanes need to be closely monitored by their physician. In this article we will describe the use of three potential strategies (pharmacogenetics, enzyme probes and therapeutic drug monitoring [TDM]), as tools to individualize treatment with paclitaxel or docetaxel.

The germline genome may be a principal source of variability in drug PK or patient susceptibility to adverse events. Discovery and validation of pharmacogenetic markers that modify drug PK or patient sensitivity could enable appropriate drug or dose selection before treatment initiation. A second approach to optimize initial dose selection is the use of enzyme probes. Enzyme probes may be useful for estimating the patient’s metabolic activity prior to treatment initiation, enabling selection of the optimal taxane starting dose for an individual patient. The final approach, TDM, would allow clinicians to optimize taxane doses during therapy. TDM is the systematic use of estimated drug exposure during previous cycles to select ideal doses in subsequent cycles. In this article we will summarize the rationale, describe preliminary attempts at implementation, discuss specific challenges to clinical translation, and make recommendations for reasonable next steps in the development of each of these three potential tools for individualizing taxane treatment.

Pharmacogenetics

Occurrence of adverse events and response to taxane treatment vary greatly among individuals [10,11]. These interindividual differences are in part a result of differences in drug exposure [23-25], but also reflect differences in patient sensitivity. Variation in the patient’s germline genome is a major factor that influences drug exposure and patient sensitivity. Genotyping specific SNPs can identify patients at elevated risk of adverse events or lack of efficacy, enabling clinicians to make more informed decisions for patients being treated with those chemotherapeutic agents [26]. Pharmacogenetics has been translated into clinical practice to enable individualization of therapy with 6-mercaptopurine and irinotecan based on TPMT [27] and UGT [28] polymorphisms, respectively. There may be individual SNPs that are valuable pretreatment biomarkers to individualize taxane therapy; however, no polymorphisms are currently being used to guide treatment decisions.

Candidate SNPs in genes relevant to taxane PK

Early pharmacogenetic studies typically used a candidate gene approach in which the investigator selects SNPs in genes that are likely to be important based on known biology of the drug or phenotype of interest. Because paclitaxel and docetaxel have comparable pharmacology and adverse event profiles it is likely that there is overlap between the polymorphisms that influence treatment outcomes. A few studies have been published that investigated the direct influence of SNPs on taxane PK, reporting inconsistent results [29-36]. For example, Bergmann et al. [34] reported lower paclitaxel clearance in individuals carrying the CYP2C8*3 variant and Fransson et al. [35] found a relationship between ABCB1 2677G>T/A polymorphisms and decreased clearance of the 6α-hydroxypaclitaxel metabolite. Henningsson et al. [33] and Marsh et al. [36] also investigated the influence of these polymorphisms on paclitaxel PK, but could not reproduce these findings.

At this time it is unclear whether there are any SNPs that could be useful for guiding taxane treatment decisions because the pharmacogenetic literature is highly inconsistent. This issue is caused by the frequent reporting of studies in small, heterogeneous cohorts of patients, and the accepted practice of evaluating multiple SNPs and end points without appropriate statistical correction. Thus, there is a high likelihood that published results reflect both false-positive and false-negative findings. This section will not be a comprehensive review of all SNPs or studies, we instead direct readers interested in that topic to previous reviews on taxane pharmacogenetics [37,38]. In this section we will highlight SNPs that are most likely to be associated with taxane treatment efficacy or toxicity based on multiple reports with consistent findings. In order to maximize reader interpretability, all results will be described as the effect of the variant compared with the wild-type genotype. An overview of successfully replicated associations between polymorphisms and taxane outcomes is displayed in Table 1.

Table 1.

Replicated pharmacogenetic associations for taxane treatment outcomes from candidate gene studies.

Gene variant	SNP information	Study (year)	Total patients (n), M:F	Ethnicity	Cancer type	Taxane treatment and concomitant therapy	Outcome	Measure of association	Ref.
CYP3A5*3	6986A>G (rs776746, splicing defect)	Tsai et al. (2009)	59, 0:59	Taiwanese	BC	Docetaxel 75 mg/m2 + epirubicin + cyclophosphamide	↓ pleural effusion	p = 0.077 (trend)	[39]
							↓ febrile neutropenia	p = 0.030
		Gandara et al. (2009)	381, 252:129	Japanese and American	NSCLC	Paclitaxel 225 mg/m2 + carboplatin	↑ OS	HR: 1.64; p = 0.07 (trend)	[43]
							↑ PFS	HR: 1.56; p = 0.09 (trend)
		Gréen et al. (2011)	33, 0:33	NR	OC	Paclitaxel 135 or 175 mg/m2 + carboplatin	↓ leukopenia	p = 0.01	[40]
							↑ leukocyte nadir	p = 0.07 (trend)
		Leskelä et al. (2011)	118, 42:76	Caucasian	Solid tumors	Paclitaxel 80–90 or 150–175 mg/m2 + various agents	↓ neurotoxicity	HR per allele: 0.51; p = 0.012	[42]
CYP2C8*3	1196A>G (rs11572080, R139K)	Leskelä et al. (2011)	118, 42:76	Caucasian	Solid tumors	Paclitaxel 80–90 or 150–175 mg/m2 + various agents	↑ neuropathy	HR: 1.72; p = 0.032	[42]
	416G>A (rs10509681, K399R)	Hertz et al. (2012)	111, 0:111	Primarily Caucasian	BC	Paclitaxel 80–90 or 175 mg/m2 + various agents	↑ clinical complete response	OR = 3.92; p = 0.0066	[44]
							↑ neuropathy	OR = 3.13; p = 0.075 (trend)
ABCB1*2†	1236C>T (rs1128503, G412G); 2677G>T/A (rs2032582, A893S/T); 3435C>T (rs1045642, I1145I)	Tran et al. (2006)	58, 29:29	NR	Solid tumors	Docetaxel 75–100 mg/m2	↑ neutropenia	p = 0.046	[30]
		Sissung et al. (2008)	73, 73:0	NR	PC	Docetaxel 30 mg/m2 single agent or + thalidomide	↓ OS	p = 0.0017	[49]
							↑ neuropathy	p = 0.035
							↑ neutropenia	p = 0.053 (trend)
		Pan et al. (2009)	54, 38:16	Chinese	NSCLC	Docetaxel 75 mg/m2 + cisplatin	↓ clinical response	p = 0.015	[54]
		Kim et al. (2009)	118, 0:118	Korean	OC	Docetaxel 75 or paclitaxel 175 mg/m2 + cisplatin or carboplatin	↑ hematological and GI toxicity	p = 0.01	[50]
		Kim et al. (2012)	218, 0:218	Korean	BC	Docetaxel 100 mg/m2	↑ neutropenia	p = 0.015	[41]
		Sissung et al. (2006)	26, NR	NR	Solid tumors	Paclitaxel 100 mg/m2	↑ neuropathy	p = 0.09 (trend)	[53]
							Greater neutrophil decrease	p = 0.02
		Gréen et al. (2006)	53, 0:53	Caucasian	OC	Paclitaxel 175 mg/m2 or 135 mg/m2 + carboplatin	↑ clinical complete response	p < 0.05	[57]
		Johnatty et al. (2008)	309, 0:309	NR	OC	Paclitaxel 175 mg/m2 or 135 mg/m2 + carboplatin	↑ PFS	p = 0.001	[58]
		Grau et al. (2009)	47, 4:43	Caucasian	H&N	Paclitaxel 80 mg/m2	↑ response rate	p < 0.001	[56]
							↑ OS	p = 0.039
		Chang et al. (2010)	43, 26:17	Korean	Gastric cancer	Paclitaxel 175 mg/m2 + leucovorin + 5-FU	↓ PFS	p = 0.001	[51]
							↑ diarrhea	p = 0.034
							↑ mucositis	p = 0.004
		Bergmann et al. (2011)	92, 0:92	Caucasian	OC	Paclitaxel 175 mg/m2 + carboplatin	Greater neutrophil decrease	p = 0.02	[52]
CYP1B1*3	4326C>G (rs1056836, L432V)	Figg et al. (2007)	20, 20:0	Primarily Caucasian	PC	Docetaxel 30 mg/m2 + estramustine + thalidomide	↓ survival	p = 0.013	[69]
		Sissung et al. (2008)	52, 52:0	NR	PC	Docetaxel 30 mg/m2 ± estamustine and thalidomide or prednisone	↓ OS	p = 0.0004	[68]
		Rizzo et al. (2010)	95, 1:94	Caucasian	BC	Docetaxel 75 or 100 mg/m2 or paclitaxel 80 mg/m2	↓ hypersensitivity reactions	OR: 0.1361, p = 0.0008	[130]
		Pastina et al. (2010)	60, 60:0	NR	PC	Docetaxel 75 mg/m2 or 30 mg/m2	↓ response rate	p = 0.014	[70]
							↓ PFS	p = 0.032
							↓ OS	p < 0.001
		Marsh et al. (2007)	84, 0:84	Primarily Caucasian	BC	Paclitaxel 575–775 mg/m2 + doxorubicin + cyclophosphamide	↓ PFS	p = 0.037	[36]

^†Assessed any or all of the ABCB1*2 SNPs that are in linkage disequilibrium.

↑: Increase; ↓: Decrease; 5-FU: 5-fluorouracil; BC: Breast cancer; F: Female; GI: Gastrointestinal; H&N: Head and neck cancer; HR: Hazard ratio; M: Male; NR: Not reported; NSCLC: Non-small-cell lung cancer; OC: Ovarian cancer; OR: Odds ratio; OS: Overall survival; PC: Prostate cancer; PFS: Progression-free survival.

CYP3A4 & CYP3A5

Both taxanes are substrates for the highly promiscuous CYP3A4/3A5 metabolic system [17,18]. There is a strong linkage between the functional CYP3A5*1 allele and CYP3A4*1B allele, and presence of these alleles together has been related to increased docetaxel clearance [30,31]. A few studies have reported associations between the low-activity CYP3A5*3 (6986A>G, rs776746) SNP and a lower risk of hematological toxicity in patients treated with docetaxel [39] or paclitaxel [40], however, the opposite finding has also been reported [41].

Paclitaxel is only partially metabolized by CYP3A4/3A5, so it is somewhat less likely that this polymorphism will have adequate predictive power to be clinically useful in guiding paclitaxel treatment decisions. Regardless, it is interesting that the low-activity CYP3A5*3 variant has been demonstrated to decrease neurotoxicity [42] and increase survival [43] in paclitaxel-treated patients.

CYP2C8

Systemic elimination of paclitaxel occurs primarily by CYP2C8-mediated hepatic metabolism [27]. Secondary analyses of three small studies [32,40,44] suggested an increase in neurotoxicity risk of patients carrying the low-activity CYP2C8*3 variant, which consists of two polymorphisms that occur in almost complete linkage disequilibrium (416G>A, R139K, rs11572080; and 1196A>G, K399R, rs10509681). This has recently been replicated in analyses of independent patient cohorts [42,45]. Hertz et al. also reported a corresponding increase in neoadju-vant clinical complete response from 23 to 55% in breast cancer patients who carried the *3 variant [44]. These reports of increased toxicity and efficacy are consistent with the finding that patients carrying the *3 variant have 11% lower paclitaxel clearance and, consequently, greater drug exposure [34].

ABCB1

The ABCB1 gene encodes P-gp (also known as MDR1), expressed in the liver and intestine. Three polymorphisms (3435C>T, I1145I rs1045642; 2677G>T/A, A893S/T, rs2032582; and 1236C>T, G412G, rs1128503) in moderate linkage disequilibrium in the ABCB1 gene have been extensively studied for an association with taxane treatment outcomes, however, it is unclear which, if any, SNP is causing the observed effect on outcome [46]. These three SNPs have been looked at as a haplotype, sometimes referred to as ABCB1*2, as well as being studied individually. It was originally hypothesized that the nonsynonymous triallelic (2677G>T/A) SNP is causative but more recent data suggest that the two silent polymorphisms (3435C>T and 1236C>T) may alter protein function [47,48].

The ABCB1 variant alleles have been associated with a higher risk of neutropenia in multiple studies with single-agent docetaxel [30,41,49]. In addition, an association between an ABCB1 variant allele and a higher risk of hematological and gastrointestinal toxicities has also been reported [50]. These findings correspond with the study of Bosch et al., in which patients homozygous for the variant alleles had higher area under the curve (AUC) due to decreased docetaxel elimination [29]. Consistent with their shared pharmacology, multiple studies have reported an increase in hematological or gastrointestinal toxicities for paclitaxel-treated patients carrying ABCB1 variants [50-53].

The three frequently studied polymorphisms in ABCB1 (3435C>T, 2677G>T/A and 1236C>T) have also been associated with taxane response or survival. The previously mentioned study of Sissung et al. found that docetaxel-treated patients carrying variant alleles had decreased overall survival (p = 0.0017) [49] and Pan et al. observed a poorer response in these patients [54]. The association with efficacy of paclitaxel treatment is somewhat less clear; ABCB1 variant alleles have been associated with shorter progression-free survival, overall survival and lower disease control rates in Koreans [51,55], but greater response rate [56,57] and progression-free survival [58] in other populations.

Other SNPs in genes relevant to taxane PK

Associations with taxane treatment outcomes have been reported for other SNPs in a variety of transporters or phase II metabolic enzymes, such as the hepatocellular uptake and efflux proteins OATP1B3 and MRP2 [59]. The GST system has also been extensively investigated, but in most cases these patients were treated with taxane–platinum combination therapy and the findings are most likely relevant to the pharmacogenetics of the platinum compound [30,50,60-62].

Candidate SNPs in genes relevant to taxane pharmacodynamics

CYP1B1

CYP1B1 is an enzyme that is not expressed in normal tissue and does not contribute to taxane PK, however, it has been detected in various tumors [63]. The *3 variant of the CYP1B1 enzyme (4326C>G, L432V, rs1056836) has shown increased catalytic activity toward its substrates, including endogenous hormones such as 17β-estradiol [64-66]. This altered enzyme activity may influence taxane efficacy by binding directly to taxanes or producing an estrogen metabolite that antagonizes the taxane mechanism of action [49,67], providing a plausible explanation for the multiple studies that have demonstrated inferior survival or response to taxanes for patients carrying the CYP1B1*3 genotype [36,68-70].

β-tubulin

Taxanes exert their cytotoxic effect by binding to β-tubulin in the cellular microtubules. Recently, Leandro-García et al. identified two linked polymorphisms in the proximal promoter of TUBB2A (-101T>C and -112A>G) that increase gene transcription. Patients carrying these variants may be at decreased risk of developing paclitaxel-induced neurotoxicity [71]. Similarly, a polymorphism in TUBB1 may modulate a patient’s risk of taxane-induced thrombocytopenia [72], however, replication of these associations has not yet been attempted to our knowledge.

It is worth pointing out that β-tubulin somatic mutations have also been investigated for an influence on paclitaxel response [73], however, the importance of these variations has been controversial. The β-tubulin genes are known to be highly conserved across species [74] and genotyping of this gene has been problematic owing to interference of β-tubulin pseudogenes [75].

Other SNPs in genes involved in taxane pharmacodynamics

Testing the hypothesis that taxane sensitivity may be related to the activity of proteins involved in the repair of DNA damage, one study reported a significant association with a tagSNP in the FANCD2 gene that increased gene expression and neuropathy risk [76]. This association awaits replication in an independent patient cohort.

Genome-wide & noncandidate gene association studies

An alternative method for discovering polymorphisms that influence treatment outcome is the genome-wide association study (GWAS). This approach enables the simultaneous interrogation of a huge amount of the known genetic variation in humans. Baldwin et al. published the first GWAS of a taxane clinical end point, reporting an association between an FGD4 polymorphism and the onset of peripheral sensory neuropathy in a large paclitaxel-treated discovery cohort and two independent replication cohorts [77]. This polymorphism in FGD4 was found to increase the risk of neuropathy by 57% in the discovery cohort (hazard ratio [HR]: 1.57; 95% CI: 1.30–1.91; p = 2.6 × 10^-6) and an even larger increase in risk was detected in independent cohorts of European (HR: 1.72; 95% CI: 1.06–2.80; p = 0.013) and African–American (HR: 1.93; 95% CI: 1.13–3.28; p = 6.7 × 10^-3) patients. FGD4 encodes for the protein Frabin, a widely expressed guanine nucleotide exchange factor for Cdc42, a small rhoGTPases that regulates cellular morphogenesis, including myelination. FGD4 has previously been linked with the congenital Charcot–Marie–Tooth disease, a condition that resembles taxane-induced sensory peripheral neuropathy, providing a plausible biological explanation for their finding. Other intriguing SNPs related to the onset or severity of neuropathy in this study were found in EPHA5 (rs7349683) and FZD3 (rs10771973).

No GWAS of docetaxel outcomes has been published, however, one study used the Affymetrix DMET 1.0 platform (Affymetrix Inc., CA, USA) to simultaneously interrogate nearly 2000 variants in 225 genes that may be relevant to drug PK [78]. Interesting associations of SNPs with docetaxel response (PPAR-δ, SLUT1C2 and CHST3) and toxicity (SPG7, CHST3, CYP2D6, NAT2, ABCC6, ATP7A, CYP4B1 and SLC10A2) were reported, however, replication of these findings has yet to be presented so it is unclear whether these variants have a true effect on docetaxel treatment outcome.

Challenges of pharmacogenetics

The primary challenge to the field of pharmacogenetics at this time is the difficulty of validating associations. Even the most highly studied and biologically reasonable effects, such as the relationship between CYP2D6 genotype and tamoxifen outcomes, continue to be debated in the literature [79]. Validation of pharmacogenetic associations requires carefully planned studies in large, independent patient cohorts. After validation, prospective genotype-guided studies are required to demonstrate a clinical benefit of modifying treatment in patients who carry a specific SNP. This potential benefit will need to be weighed against the additional cost of genotyping patients to identify potentially rare variants of interest. Finally, variability in treatment outcome may be the result of the interplay of many genetic factors, including modifications in the epigenome or the proteome. Thus, the explanatory value of any one SNP may be limited, and assessment of a set of SNPs may be required to to guide taxane treatment decisions.

Future work for pharmacogenetics

Several associations have been reported between SNPs and taxane treatment outcomes, although none have been consistently demonstrated in multiple independent cohorts to warrant clinical implementation. Still, there are a few promising gene variants, such as the polymorphisms in ABCB1, CYP2C8, CYP1B1, and FGD4, which based on the available data should be prioritized for replication in large, independent, prospectively collected patient cohorts. After replication, these markers should be tested in prospective, genotype-guided studies to ascertain whether the clinical implementation of pharmacogenetics would improve patient outcomes at a reasonable cost to the system. Finally, future work should seek to understand the influence of dynamic changes upstream or downstream of the genome on the static changes found within the genome.

Enzyme probes

Activity of metabolic enzymes is a major determinant of drug exposure. There are many factors that could influence enzymatic activity for a given patient, such as drug interactions, comorbidities, or genetic mutations. Instead of attempting to account for each factor individually, it may be best to directly measure the metabolic activity for the patient using a probe marker for the enzyme of interest. The ideal probe would measure enzyme activity in a patient sample ex vivo, however, for taxane metabolism only in vivo probes have been investigated. An ideal in vivo enzymatic probe is a safe, conveniently administered and quickly interpretable marker agent that shares the metabolic pathway of the drug of interest, enabling estimation of the expected rate of metabolism of the drug [80]. Use of a pretreatment probe could help clinicians select a more appropriate dose in order to target a desired level of exposure, potentially preventing toxicity or ineffectiveness.

The FDA maintains a list of substrates that can be used in drug development to measure in vivo activity of the major CYP450 enzymes involved in drug metabolism, including CYP2D6, CYP3A4, and CYP2C9 [204]. Currently, the use of enzyme probes in clinical practice is limited; one example is the use of an enzyme probe to predict TPMT activity ex vivo in patients who will receive thiopurine drugs. However, there could be many more applications, particularly for CYP3A4, which is involved in the metabolism of the majority of drugs used clinically. In this article we will refer to CYP3A because CYP3A4 and CYP3A5 have highly overlapping substrate specificity [81], which makes it difficult to determine their individual contribution to drug metabolism. CYP3A is involved in the metabolism of both docetaxel and paclitaxel [82], and paclitaxel is also metabolized by CYP2C8, therefore, this section will focus on the clinical translation of probes for CYP3A and CYP2C8 activity.

Docetaxel probes

Erythromycin breath test

The erythromycin breath test (ERMBT) is a validated in vivo assay for CYP3A activity that has been studied as a potential probe for docetaxel PK. For this test ¹⁴C-labelled erythromycin is administered intravenously and the amount of ¹⁴C-labelled CO₂ exhaled is measured at specific time points, yielding an estimate of CYP3A activity [83]. Several studies have reported a significant correlation between the ERMBT and docetaxel clearance [31,84], with ERMBT explaining as much as 67% of the docetaxel PK variability [85]. The utility of the ERMBT has also been demonstrated in groups of patients in whom docetaxel dose selection is particularly challenging or would be particularly beneficial, such as the elderly [86] or patients with impaired liver function [87]. However, not all studies of the ERMBT were able to detect a significant correlation with docetaxel PK [88].

Alternative probes for CYP3A activity

Several CYP3A probes other than the ERMBT have been investigated for a relationship with docetaxel PK and toxicity (Table 2). In the study from Michael et al., which did not detect an association with the ERMBT, the clearance of anti-pyrine was correlated with docetaxel clearance (r² = 79.49%; p = 0.007) [88]. Dexamethasone plasma clearance has also been studied, though the correlation was unexpectedly limited to females only [89,90]. The utility of midazolam plasma clearance as a CYP3A probe is uncertain based on several conflicting reports [91-93]. Finally, the urinary excretion of metabolites from exogenous cortisol has shown promise; a correlation between urinary 6-β-hydroxycortisol (6-β-OHF) and docetaxel clearance has been reported (r = 0.867; p < 0.001) [94]. Yamamoto et al. published the first attempt to prospectively use an enzyme probe to guide taxane dosing. In this study the docetaxel dose was selected based on 6-β-OHF, resulting in a significant decrease in docetaxel PK variability when compared with standard BSA-based dosing [95]. Prospective validation of this approach makes a compelling case for further research into the use of 6-β-OHF to guide docetaxel dosing to improve therapeutic outcomes.

Table 2.

Summary of probe studies estimating taxane pharmacokinetic parameters.

Probe	Study (year)	Total patients (n), M:F	Cancer type	Taxane dose and infusion time	Probe parameter	Taxane pharmacokinetic parameter	Correlation	p-value	Ref.
ERMBT	Baker et al. (2009)†	92, 51:41	Solid tumors	20–75 mg/m2, various infusion times	ERMBT C20 min	CL (l/h)	r2 = 0.072	p = 0.036	[31]
	Slaviero et al. (2004)†	54, 31:23	Solid tumors	40 mg/m2, 2 h	ERMBT C20 min	CL (l/h)	r2 = 0.019	p = 0.0005	[84]
	Hirth et al. (2000)†	21, 12:9	Bone or soft tissue sarcoma	100 mg/m2, 1 h	Ln ERMBT C20 min	BSA-adjusted CL (l/h/m2)	r2 = 0.67	p = 0.0001	[85]
	Hurria et al. (2006)†	19‡, 8:12	Breast, prostate or lung cancer	35 mg/m2, 30 min	ERMBT C20 min	AUCinf (μg/ml × h)	NR	p = 0.02	[86]
						AUCu inf (μg/ml × h)	NR	p = 0.01
						CLu (l/h)	NR	p = 0.04
	Hooker et al. (2008)†	77, 41:36	Solid tumors	40–75 mg/m2, 1 h	ERMBT C20 min	Ln CLu	r2 = 0.035	p = 0.177	[87]
		22§, NR	Solid tumors	40–75 mg/m2, 1 h	ERMBT C20 min	Ln CLu	r2 = 0.603	p < 0.0001
	Hertz et al. (2012)¶	14, 3:11	Solid tumors	75–90 mg/m2, 1 h	Ln ERMBT AUC0–1h	BSA-adjusted Ln AUC0–6 h	NR	NS	[96]
Dexamethasone	Puisset et al. (2004)†	21, 10:11	Solid tumors	75–100 mg/m2, 1 h	CLdexamethasone (l/h)	CLdocetaxel (l/h)	r2 = 0.34	NR	[89]
	Puisset et al. (2007)†	17, 10:7	Solid tumors	75–100 mg/m2, 1 h	CLdexamethasone (l/h)	CLdocetaxel (l/h)	r2 = 0.78	p = 0.01	[90]
		38, 20:18	Solid tumors	75–100 mg/m2, 1 h	CL (l/h)	CL (l/h)	NR	p = 0.001
Hydrocortisone	Yamamoto et al. (2000)†	29, 10:19	Advanced NSCLC	60 mg/m2, 1 h	Total urinary 6-β-OHF over 24 h	CL (l/h)	r2 = 0.752	p < 0.001	[94]
Antipyrine	Michael et al. (2012)†	19‡, 12:8	Advanced breast or NSCLC	75 or 36 mg/m2, 1 h	CL (ml/min)	CL (l/h)	r2 = 0.795	p = 0.007	[88]
Midazolam	Goh et al. (2002)†	31‡, 16:16	Solid tumors	75 or 100 mg/m2, 1 h	CL (ml/min)	CL (l/h)	r2 = 0.36	p = 0.005	[91]
	Hilli et al. (2011)†	20, 0:20	High-risk breast cancer	80 mg/m2, 1 h	Midazolam-1-OH AUCinf	CL (l/h)	r2 = 0.012	NS	[92]
					Midazolam AUC ratio	CL (l/h)	r2 = 0.005	NS
	Zamboni et al. (2011)†	29, 0:29	Ovarian or peritoneal cancer	75 mg/m2, 1 h	CL (ml/min)	CL (l/h)	r2 = 0.05	NS	[93]
Quinidine	Gréen et al. (2009)¶	38, 0:38	Ovarian or peritoneal cancer	175 or 135 mg/m2, 3 h	MR (quinidine/3-hydroxyquinidine	AUC0–24 h (μg/ml × h)#	r2 = 0.450	p < 0.001	[32]
Rosiglitazone	Hertz et al. (2012)¶	14, 3:11	Solid tumors	75–90 mg/m2, 1 h	BSA-adjusted Ln 3 h concentration	BSA-adjusted Ln AUC0–6 h	r2 = 0.38	p = 0.018	[96]

^†Docetaxel studies.

^‡One patient of unknown gender was excluded from the pharmacokinetic analyses.

^§Subanalysis in patients with liver impairment.

^¶Paclitaxel studies.

^#AUC_{0–24 h} refers to the AUC of the metabolite 6α-hydroxypaclitaxel.

6-β-OHF: 6-β-hydroxycortisol; AUC: Area under the concentration–time curve of total drug; AUC_inf: Area under the concentration–time curve from zero to infinity; AUC_{u inf}: Area under the concentration–time curve of unbound drug from zero to infinity; BSA: Body surface area; C_{20 min}: Flux of ¹⁴CO₂ at 20 min after administration; CL: Plasma clearance; CL_u: Plasma clearance of unbound drug; ERMBT: Erythromycin breath test; F: Female; Ln: Natural log; M: Male; Midazolam AUC ratio: Area under the concentration–time curve midazolam-1-OH: area under the concentration–time curve midazolam; MR: Metabolic ratio; NR: Not reported; NS: Not significant; NSCLC: Non-small-cell lung cancer.

Paclitaxel probes

The use of a probe for paclitaxel dosing, as compared with docetaxel, is complicated by the multiple routes of metabolism. The first attempt to predict paclitaxel clearance with a probe was performed by Gréen and colleagues, who attempted to use quinidine as a CYP3A probe, but no correlation was observed with paclitaxel clearance (Table 2) [32]. However, low enzyme activity was correlated with high 6α-hydroxypaclitaxel AUC, the metabolite formed by CYP2C8. This suggests that when CYP3A enzyme activity is diminished, CYP2C8 may compensate by increasing its metabolism of paclitaxel.

Our group attempted to explain the variability in paclitaxel exposure by utilizing separate probes for each enzyme. We reported for the first time that rosiglitazone may have value as an in vivo probe of CYP2C8 activity; a single concentration of rosiglitazone at 3 h explained approximately 38% of the variability in paclitaxel AUC_{0–6 h} (p = 0.018), however, the inclusion of ERMBT as a CYP3A probe did not meaningfully contribute to the explanation of the variability in paclitaxel exposure [96].

Relationship between probes & clinical outcome

There is reasonable evidence of a relationship between CYP3A probes and docetaxel PK, however, the association between the probe and clinical outcome has not been as comprehensively documented [86,95,97]. In one study the antipyrine disappearance rate correlated with neutrophil nadir and risk of grade 3+ neutropenia [97]. In the previously mentioned prospective study of cortisol metabolite-based dosing, nominally fewer patients in the individualized dosing arm experienced grade 3–4 neutropenia when compared with the BSA-based dosing arm (86 vs 93%), though the small magnitude of effect suggests that this may not be a clinically useful approach for preventing neutropenia [95].

Challenges of enzyme probes

In most cases the PK of a drug is determined not by a single enzyme, but by the interplay of a variety of factors such as enzymes, transporters and formulation vehicles. For instance, both taxanes are substrates for a variety of uptake (OATP1B3) [15,16] and efflux (P-gp) transporters [20,21]. The influence of these transporters on the results of probe assays has only been recently recognized and is not well characterized [31,98]. In addition, the vehicle Cremophor^® EL (BASF Corp., Ludwigshafen, Germany) might interfere with the ability of probes to estimate paclitaxel metabolism since Cremophor EL influences the unbound paclitaxel concentration [99], but further research is required to assess this possible effect.

Another challenge for the development of enzyme probes is substrate dependence, the idea that a change in the enzyme–substrate interaction or the system within which the interaction takes place will not have a consistent effect on the probe and the drug of interest. For example, imatinib, a CYP3A4 inhibitor, decreases ERMBT-estimated CYP3A activity but does not change docetaxel clearance [100]. Substrate dependency has also been reported in relation to certain genotypes, such as CYP2C8*3, which was described as a low-activity variant in the previous section, but seems to have increased activity toward other substrates, including rosiglitazone [101] and pioglitazone [102]. The influence of transporters, supplementary metabolic pathways, and substrate dependence may explain the lack of correlation seen when patients are administered multiple probes of the same enzyme [103].

If an enzyme probe that accurately predicts in vivo taxane exposure is validated, logistical challenges to clinical implementation will still exist. Sample collection and analysis, particularly for the ERMBT which requires specialized collection equipment, will introduce new costs to the healthcare system and potential inconvenience to patients. Each potential probe has specific challenges but the overall benefit of a validated probe that could accurately predict taxane exposure is likely to outweigh these limitations and have widespread clinical utility.

Future work for enzyme probes

The existing studies suggest that an in vivo probe may explain a significant portion of the interindividual variability in taxane PK. There is more work to be done to translate enzyme probes to clinical application, starting with determination of the probe that best reflects the PK of each taxane. Once the optimal probe is identified, prospective studies are needed to verify the a priori improvement in taxane dose selection from probe usage. Then, prospective studies comparing therapeutic outcomes using probe versus BSA-based dosing are needed to validate the clinical utility of enzyme probes, and the costs of this approach must be weighed against these benefits, particularly in subgroups of patients in whom empirical dose selection is challenging.

TDM

TDM is an approach to individualizing therapy that employs systematic drug-level monitoring to adjust future dosages. TDM enables targeting of an exposure level in order to improve the likelihood of response, minimize the probability of toxicity, or both. In general, TDM is considered useful for drugs that have shown extensive PK variability, a narrow therapeutic window and a well-defined relationship between systemic exposure (PK) and toxicity or response (pharmacodynamics [PD]) [104,105]. Most of the current anticancer drugs, including taxanes, meet the first two criteria [104], but it is challenging to define a drug’s PK–PD relationship. Over the last 10 years, tremendous progress has been made in describing the PK–PD for taxanes. In this section we will briefly review these PK–PD relationships and describe the potential and challenges of using TDM as a tool for individualizing docetaxel or paclitaxel therapy.

Docetaxel

PK–PD relationship

The toxicity profile of docetaxel is predominately hematological, with other less common toxicities including fluid retention and neuropathy. Several studies have reported a relationship between docetaxel PK parameters, particularly clearance or AUC, and these toxicities [23,25,106-111]. Our understanding of the relationship between docetaxel PK and treatment efficacy is more limited with only a few studies demonstrating an association between PK and progression or survival (Table 3) [23,25,106].

Table 3.

Summary of studies on docetaxel pharmacokinetic–pharmacodynamic relationships.

Study (year)	Total patients (n), M:F	Cancer type	Docetaxel dose, infusion duration and schedule	PK parameter	Clinical outcome	Statistical association	Ref.
Bruno et al. (1998)	640, 270:370	Solid tumors	75 or 100 mg/m2, 1 h, 3-weekly	50% lower CL	4.3-fold increase in odds of experiencing neutropenia	p < 0.0001	[106]
				Plasma concentration >0.20 μM for over 4 h	40% greater risk of fluid retention	p = 0.0029
				AUC	Risk of progression decreased as AUC increase in NSCLC patients	p = 0.0232
Charles et al. (2006)	68, 39:29	Solid tumors	40 mg/m2, 1 h, weekly	CL < 30 l/h	Odds ratio of hematological toxicity = 9.1	p = 0.008	[107]
				AUC > 2.3 μM	Odds ratio of hematological toxicity = 6.9	p = 0.02
Bruno et al. (2003)	180, 118:62	NSCLC	100 mg/m2, 1 h, 3-weekly	AUC	Predictor of severe toxicity during first cycle	p < 0.0001	[25]
Ozawa et al. (2008)	200, 86:114	Solid tumors	60 mg/m2, 1 h, 3-weekly	AUC	Predictor of febrile neutropenia	p = 0.001	[108]
Baker et al. (2005)	55, 26:29	Solid tumors	75 or 50 mg/m2, 1 h, 3-weekly	AUCu and AUC	Correlation with percentage decrease in ANC	p = 0.002 and 0.029	[111]
				AUCu	Correlation with occurrence of grade 4 neutropenia	p = 0.013
Minami et al. (2006)	69, 12:57	Solid tumors	20–60 mg/m2, 1 h, 3-weekly	AUCu	Correlation with occurrence of grade 4 neutropenia	p = 0.05	[109]
				Cmaxu	Correlation with occurrence of grade 4 neutropenia	p = 0.01
Sandstrom et al. (2005)	44, 0:44	Breast cancer	Docetaxel 70 mg/m2, 1 h, 3-weekly + epirubicin	AUC	Correlation with white blood cell survival fraction	NR	[110]

ANC: Absolute neutrophil count; AUC: Area under the concentration–time curve of total drug; AUC_u: Area under the concentration–time curve of unbound drug; CL: Plasma clearance; C_maxu: Peak plasma concentration of unbound drug; F: Female; M: Male; NR: Not reported; NSCLC: Non-small-cell lung cancer; PK: Pharmacokinetics.

Clinical attempts at TDM

Only one published study has explored the potential for individualizing docetaxel dosing through TDM [112]. Engels et al. utilized a limited PK sampling strategy and a validated population PK model with a predefined target AUC of 4.9 mg/l × h. Fifteen patients were treated with at least one course of PK-guided docetaxel and were compared with a group of 15 patients receiving conventional (BSA-based) docetaxel therapy. TDM-guided dosing resulted in a decrease in interindividual variability (standard deviation of natural log-transformed AUC) of 39%; somewhat lower than the study’s predefined objective of a 50% decrease. Even though the differences in percentage decrease in white blood cell or absolute neutrophil count were also not statistically significant, PK-guided dosing successfully decreased the interindividual variability in these measures by approximately 50%. Despite the lack of statistical significance, this small study suggests that AUC-targeted dosing may attenuate the interindividual variability in drug exposure and hematological toxicity, and demonstrates the clinical feasibility of docetaxel TDM with limited PK sampling.

Paclitaxel

PK–PD relationships

Numerous studies have found a relationship between paclitaxel exposure and toxicity [113-122] or efficacy (Table 4) [116,118,123]. Based on these studies the most important parameter of paclitaxel PK seems to be the time that the systemic concentration remains above a threshold of 0.05 or 0.1 μM. This parameter has been associated with a variety of adverse events, such as neutropenia and neuropathy, and measures of treatment effectiveness.

Table 4.

Summary of studies on paclitaxel pharmacokinetic–pharmacodynamic relationships.

Study (year)	Total patients (n), M:F	Cancer type	Paclitaxel dose, infusion duration and schedule	PK parameter	Outcome	Statistical association	Ref.
Gianni et al. (1995)	30, 0:30	OC and BC	135 or 175 mg/m2 (OC) or 225 mg/m2 (BC), 3 h, 3-weekly	T > 0.05 μM > 24 h	↑ neutropenia	p < 0.05	[115]
Ohtsu et al. (1995)	27, 18:9	Solid tumors	105–270 mg/m2, 3 or 24 h, 3-weekly	T > 0.05 μM = 14.3 h†	Corresponds to a 50% ↓ in granulocyte count	H = 1.88	[117]
						p < 0.05
Mielke et al. (2005)	24, 12:12	Solid tumors	100 mg/m2, 1 or 3 h, weekly	T > 0.05 μM ≥ 10.6 h	↑ neuropathy	p = 0.023	[119]
				AUC	Correlation with neuropathy	p = 0.002
				AUCu	Correlation with neuropathy	p = 0.003
Joerger et al. (2007)	105, 0:105	OC	175 mg/m2, 3 h, 3-weekly + carboplatin	T > 0.05 μM (continuous)	Correlation with tumor complete response	p = 0.02	[116]
				T > 0.05 μM (continuous)	Correlation with tumor partial response	p = 0.05
				T > 0.05 μM > 61.4 h	↑ severe neutropenia	p = 0.01
				T > 0.05 μM > 61.4 h	↑ time to progression	p = 0.05
Miller et al. (2004)	29, 0:29	Solid tumors	360 mg, 3 h, 3-weekly	T > 0.05 μM (continuous)	Inverse correlation with nadir ANC	p = 0.04	[120]
Huizing et al. (1993)	18, 0:18	OC	175 or 135 mg/m2, 3 or 24 h, 3-weekly	T > 0.1 μM = 11.16 h†	Corresponds to a 50% ↓ in ANC	H = 2.73	[113]
				T > 0.1 μM = 15.16 h†	Corresponds to a 50% ↓ in WBC count	H = 2.16
Huizing et al. (1997)	55, 41:14	NSCLC	100–250 mg/m2, 3 h + carboplatin	T > 0.1 μM > 15 h	↑ survival	p = 0.05	[123]
Jiko et al. (2007)	16, 16:0	Urogenital cancers	175 or 150 mg/m2, 3 h, 3-weekly + carboplatin (n = 10) or gemcitabine (n = 6)	T > 0.1 μM (continuous; in patients receiving gemcitabine)	Correlated with percentage ↓ in neutrophils and platelets	p = 0.018	[114]
Mould et al. (2006)	160, 0:160	Endometrial cancer	Paclitaxel 150 mg/m2, 24 h, 3-weekly + doxorubicin	Linear binned AUC	Correlation with granulocytopenia	p < 0.001	[118]
				AUC	Inverse correlation with overall survival	p = 0.055
Nakajima et al. (2005)	23, 0:23	OC	180 mg/m2, 3 h, 3-weekly + carboplatin	AUC	Correlation with leukocytopenia	p < 0.05	[121]
Kobayashi et al. (2007)	33, 22:11	Gastric cancer	60–90 mg/m2, 1.5 h, weekly + 5-FU or cisplatin	AUC (in patients receiving 90 mg/m2)	Correlation with hematological toxicity	p = 0.01	[122]
					Correlation with dose-limiting toxicity	p = 0.03

^†Value calculated in a sigmoid E-max model representing the time above threshold corresponding to a 50% decrease in the pharmacodynamic parameter.

↑: Increase; ↓: Decrease 5-FU: 5-fluorouracil; ANC: Absolute neutrophil count; AUC: Area under the concentration–time curve; AUC_u: Area under the concentration–time curve of unbound drug; BC: Breast cancer; F: Female; H: Hill constant; M: Male; NSCLC: Non-small-cell lung cancer; OC: Ovarian cancer; PK: Pharmacokinetics; T > 0.05 μM: Amount of time paclitaxel plasma concentration exceeds 0.05 μM; T > 0.1 μM: Amount of time paclitaxel plasma concentration exceeds 0.1 μM; WBC: White blood cell.

Clinical attempts at TDM

Woo et al. reported the initial attempt at utilizing paclitaxel TDM, targeting a range for systemic exposure in seven children with recurrent acute leukemia receiving 24-h infusions [124]. In this single-arm study the infusion rate was adjusted based on clearance estimated 8 h into infusion. Five of the seven children reached AUCs between 75 and 125% of the target, whereas none were projected to reach target AUC without adjustment, demonstrating that adjusting the paclitaxel dose based on a target AUC range is clinically feasible for a 24-h infusion. In another study de Jonge et al. targeted a paclitaxel threshold of >0.1 μM for ≥15 h in 25 non-small-cell lung cancer patients [125]. In the first cycle a standard regimen of 175 mg/m² was given to all patients and subsequent doses were adjusted based on their time above threshold. In the first cycle more than one-third of the patients (nine out of 25; 36%) had suboptimal exposure, but using TDM this decreased to 23% (five out of 22) and 11% (one out of nine) by cycles 2 and 6, respectively.

Challenges of TDM

The previously described studies demonstrate that taxane TDM is clinically feasible. However, there are serious challenges to its widespread implementation. TDM requires the development of an assay for the drug of interest that quickly returns accurate results. As with the other techniques described, TDM will introduce new healthcare costs by requiring collection and analysis of patient samples, and highly trained staff to interpret the PK results and recommend appropriate dose adjustments. It could also be burdensome on the patient, who will need to remain at the treatment center or return to the facility for sample collection. However, these challenges have been overcome in other therapeutic areas such as infectious disease or neurology, both of which employ TDM in certain treatment situations.

Future work for TDM

Pilot studies of the use of TDM suggest that this approach is clinically feasible and can successfully limit taxane interindividual variability or enable clinicians to target specific parameters of drug exposure. Though the connection between taxane exposure and treatment outcome has been an area of rich investigation, additional research is needed to more explicitly define these PK–PD relationships. Application of this developing knowledge will facilitate the design and execution of prospective TDM studies. One approach to this, reported recently by Joerger et al., is to perform PK–PD simulations to establish TDM-based dosing algorithms and estimate the clinical benefit of implementation [126]. Their proposed algorithm was simulated in 1000 patients and resulted in a reduction of grade 4 neutropenia in cycle 1 from 15 to 7%, and a further reduction to 4% in cycle 2. These simulations could assist in the design of large, prospective studies to compare TDM and empirical dosing with the primary aim of demonstrating an improvement in relevant clinical outcomes. The cost–effectiveness of TDM will also need to be considered before clinical uptake of TDM for taxane therapy is realized.

Conclusion

This article examined the potential use of pharmacogenetics, enzyme probes and TDM as tools for individualizing taxane therapy. Of the three approaches, TDM may, when ready for implementation, have the greatest value for clinical practice. Until then, pharmacogenetics and enzyme probes could assist in treatment decision-making in the near future. All three tools require continued validation and prospective studies with predefined, clinically relevant end points and robust statistical power, in addition to recognition of the added costs and potential inconvenience of clinical implementation.

It is important to note that these tools are not mutually exclusive; ultimately taxane treatment may incorporate elements of each approach into an adaptive treatment paradigm centered on the individual patient (Figure 2). Pharmacogenetics, in parallel with our rapidly advancing ability to select drugs based on tumor characteristics [127,128], could be used to select patients most likely to benefit or least likely to experience severe toxicity for treatment with a taxane. An enzyme probe could then be combined with other patient characteristics known to influence taxane exposure, such as age, gender and bilirubin levels, in order to design an initial dosing regimen for that patient [11,129]. Finally, TDM could guide iterative dose adjustment to ensure the patient’s taxane exposure is optimized. After 20 years of experience with taxanes in clinical practice it is time to look beyond BSA-based dosing and utilize the available tools to establish a revolutionary, individualized taxane treatment paradigm, ushering in the long-awaited era of personalized cancer therapy.

Figure 2. Integrated approach to taxane treatment that utilizes pharmacogenetics, enzyme probes and therapeutic drug monitoring to enhance efficacy and limit toxicity of therapy for cancer patients.

These tools can help screen out patients who are not good candidates for taxane treatment, assist in estimating a starting dose for taxane-treated patients and enable ongoing optimization of taxane dosing throughout treatment.

BSA: Body surface area in mg/m².

Future perspective

The one-size-fits-all approach to taxane dosing, which almost exclusively uses the patient’s BSA to determine what dose they will receive throughout their course of treatment, will eventually be replaced by an individualized, adaptive dosing scheme. The SNPs that appreciably influence a patient’s likelihood of response or toxicity will be used in conjunction with somatic genetic information to screen out patients who are likely to experience suboptimal treatment outcomes. Initial doses of taxanes will be estimated using complex algorithms that combine standard patient factors with in vivo assays that predict drug exposure. Actual exposure will be monitored during treatment to enable adjustment of subsequent doses to achieve target concentrations that maximize patient outcomes. Initial validation and translation of these tools will likely be limited to specific treatment settings, but ultimately these incremental steps will coalesce into a revolutionary treatment paradigm that deploys these complementary tools to maximize efficacy and minimize toxicity of taxane treatment.

Executive summary.

Background

• ■Taxane dosing is primarily based on body surface area, but this does not reduce the large interindividual variability in pharmacokinetics and unpredictability of treatment response and toxicity.
• ■Tools that have been used to individualize treatment with other drugs, such as pharmacogenetics, enzyme probes and therapeutic drug monitoring, are being investigated for their potential role in taxane treatment.
• ■This article highlights the progress and potential for each tool and the challenges and next steps for clinical translation.

Pharmacogenetics

• ■Variants in genes relevant to pharmacology or biology may influence a patient’s response to therapy or risk of side effects.
• ■Numerous studies have been published on associations between pharmacogenetic variants and taxane treatment outcomes, with highly inconsistent results.
• ■Some associations have been consistently replicated in multiple independent patient populations (CYP3A5*3, CYP2C8*3, ABCB1*2 and CYP1B1*3), suggesting that there may be a true association between the variant and treatment outcome.
• ■These variants should be prioritized for validation in large patient cohorts, followed by prospective evaluation of genotype-guided taxane treatment.

Enzyme probes

• ■Drug exposure is largely determined by the activity of metabolic enzymes.
• ■Enzyme probes may be able to estimate the in vivo activity of the enzyme of interest, enabling selection of an appropriate dose to achieve a target exposure.
• ■Multiple CYP3A probes (erythromycin breath test, antipyrine and cortisol) have demonstrated an in vivo correlation with docetaxel pharmacokinetics.
• ■Rosiglitazone is the only CYP2C8 probe that has been reported to have an in vivo correlation with paclitaxel pharmacokinetics.
• ■Translation of a probe into clinical practice will require prospective confirmation that probe-based dosing confers a clinically relevant benefit.

Therapeutic drug monitoring

• ■The taxanes have high interindividual pharmacokinetic variability and a narrow therapeutic window, making them ideal for therapeutic drug monitoring.
• ■Pharmacokinetic–pharmacodynamic studies have demonstrated that certain parameters of drug exposure are associated with likelihood of toxicity or efficacy from taxane therapy.
• ■Therapeutic drug monitoring has been piloted for both docetaxel and paclitaxel, demonstrating clinical feasibility and success in achieving targets of drug exposure.
• ■Prospective evaluation of the benefit of therapeutic drug monitoring is necessary before widespread adoption into clinical use.

Conclusion

• ■The three tools outlined in this article have great potential for improving taxane treatment; however, none are ready for clinical implementation at this time.
• ■Each will need to improve patient outcomes in large, well-designed prospective studies before integration into standard clinical practice.
• ■The benefit to patients will need to be balanced with the additional expense and potential for patient inconvenience when determining the clinical role for each tool.
• ■In the future these three tools will be used complementarily to individualize each patient’s taxane treatment to maximize efficacy and limit toxicity.