CYP2C8*3 increases risk of neuropathy in breast cancer patients treated with paclitaxel

Abstract

Background

Paclitaxel-induced neuropathy is an adverse event that often leads to therapeutic disruption and patient discomfort. We attempted to replicate a previously reported association between increased neuropathy risk and CYP2C8*3 genotype.

Patients and methods

Demographic, treatment, and toxicity data were collected for paclitaxel-treated breast cancer patients who were genotyped for the CYP2C8*3 K399R (rs10509681) variant. A log-rank test was used in the primary analysis of European-American patients. An additional independent replication was then attempted in a cohort of African-American patients, followed by modeling of the entire patient cohort with relevant covariates.

Results

In the primary analysis of 209 European patients, there was an increased risk of paclitaxel-induced neuropathy related to CYP2C8*3 status [HR (per allele) = 1.93 (95% CI: 1.05–3.55), overall log-rank P = 0.006]. The association was replicated in direction and magnitude of effect in 107 African-American patients (P = 0.043). In the Cox model using the entire mixed-race cohort (n = 411), each CYP2C8*3 allele approximately doubled the patient's risk of grade 2+ neuropathy (P = 0.004), and non-Europeans were at higher neuropathy risk than Europeans of similar genotype (P = 0.030).

Conclusions

The increased risk of paclitaxel-induced neuropathy in patients who carry the CYP2C8*3 variant was replicated in two racially distinct patient cohorts.

introduction

Paclitaxel is one of the most highly effective therapies in the treatment of breast cancer, improving disease-free survival when added sequentially to anthracycline-based combination therapy in the adjuvant setting [1]. Many patients, however, are unable to tolerate the full course of paclitaxel therapy due to the appearance and progression of sensory neurotoxicity. In large trials, the rate of grade 2 or higher (grade 2+) and grade 3 or higher (grade 3+) sensory neuropathy is 15%–20% and 5%–10%, respectively [2, 3]. Sensory neuropathy typically manifests as a tingling or burning sensation in the extremities that may progress to loss of function that can be irreversible if treatment is continued [4]; thus, paclitaxel therapy is often discontinued once a patient experiences grade 2+ neurotoxicity.

There are known risk factors for the development of paclitaxel-induced neuropathy. Patients who have prior neuropathy, either from diabetes [5] or neurotoxic chemotherapeutic treatment [6], or patients who are older [7] or African American [8], have been reported to be at increased risk. The progressive nature of paclitaxel-induced neuropathy suggests toxicity development may be attributed to cumulative drug exposure. Indeed, increased cumulative dose [9] and an increase in the time that the drug concentration remains above a threshold for a given dose [10] are both associated with increased neuropathy risk.

Paclitaxel is primarily metabolized by CYP2C8 [11], with a contribution from CYP3A4 [12], and exposure to paclitaxel in cancer patients is correlated with CYP2C8 activity [13]. Thus, any factor which modulates CYP2C8 activity is likely to influence paclitaxel exposure, which is highly variable between patients even after accounting for baseline factors such as body size and bilirubin [14]. CYP2C8 has few known inhibitors and inducers, limiting concerns for typical drug interaction. However, single-nucleotide polymorphisms (SNPs) in the CYP2C8 gene, such as the *3 (rs11572080 R139K and rs10509681 K399R) variant, have decreased paclitaxel metabolic activity [15] leading to increased drug exposure [16]. Based on data from the International HapMap Project, the *3 variant is more common in Caucasian individuals [Utah residents with northern and western European ancestry (CEU) allele frequency (AF) = 0.14] than African Americans [African Ancestry in Southwest United States (ASW) AF = 0.04] [17, 18].

Gréen et al. [19] were the first to suggest a potential increase in neuropathy risk for patients who carried the CYP2C8*3 variant. We recently reported results from a small pharmacogenetic study demonstrating that breast cancer patients treated with neoadjuvant paclitaxel who were carriers of the *3 allele were more likely to achieve clinical complete response from paclitaxel treatment (55% versus 23%; OR = 3.92, 95% CI: 1.46–10.48, corrected P = 0.046), but tended to have higher incidence of grade 3+ neuropathy (22% versus 8%; OR = 3.13, 95% CI: 0.89–11.01, uncorrected P = 0.075) [20]. Leskela et al. [21] also reported a significant increase in neuropathy risk for patients who were homozygous for the *3 allele. Therefore, we hypothesized that the increase in neuropathy risk associated with CYP2C8*3 could be replicated separately in independent cohorts of European-American and African-American breast cancer patients treated with paclitaxel.

materials and methods

patients and treatments

CYP2C8*3 K399R (referred to as CYP2C8*3 from now on) was genotyped in a cohort of patients treated between 2005 and 2011 and derived from the University of North Carolina Lineberger Comprehensive Cancer Center Breast Cancer Database, which collects patient data, including self-reported race, treatment details, and toxic effects. Eligible women received neoadjuvant and/or adjuvant paclitaxel-containing regimens and enrolled in an IRB-approved clinical trial that collected genomic DNA from all newly diagnosed patients. In most cases, patients received paclitaxel on a standard neoadjuvant or adjuvant treatment regimen, with a predefined dose, schedule, and duration. Some patients received additional biologic treatment concurrent with paclitaxel, most commonly HER2-targeted therapy for the subset of HER2-overexpressing tumors. Neuropathy descriptions were recorded in the patient record based on clinician judgment and patient reported symptoms. A study coordinator blinded to patient genotype translated the physician description to a grade based on National Cancer Institute Common Terminology Criteria for Adverse Events [22]. Use of supplemental neuropathy prevention (glutamine, vitamin B complex, or vitamin B6) or treatment (gabapentin or amitriptyline) was at the discretion of the treating clinician. All patients signed informed consent to participate and agreed to allow DNA to be collected for additional pharmacogenetic studies. This study adhered to the declaration of Helsinki and the study protocol was approved by the UNC Institutional Review Board.

SNP genotyping

A 30 ml blood sample was collected from each subject at the time of study enrollment. DNA used for genotyping was extracted by the UNC BioSpecimen Processing Facility and plated at 60 ng/µl. Genotyping was carried out blinded to clinical data using the Affymetrix DMET™ Plus Chip (Affymetrix, Inc., Santa Clara, CA) at Gentris Corporation (Gentris Corporation, Morrisville, NC) following the manufacturer's protocol with known genomic DNA controls provided by Affymetrix to monitor inter- and intra-assay performance. Any sample with call rate <98% was excluded from analysis. CYP2C8*3 K399R (rs10509681) (AM_10125) was the only SNP analyzed for this replication study; all non-*3 loci are assumed to be wild-type (*1), enabling classification of each subject as CYP2C8 homozygous variant (*3/*3), heterozygous (*1/*3), or homozygous wild-type (*1/*1).

statistical analysis

The primary analysis was carried out in a cohort of self-reported European-American patients who were not analyzed in the previous neoadjuvant study [20]. African-American patients were analyzed separately in a cross-race replication. These two groups were then combined with patients of other races and previously reported patients to create a large mixed-race cohort. CYP2C8*3 genotype frequencies were assessed for concordance with expectations under Hardy–Weinberg equilibrium (HWE), using Fisher's exact test. The primary toxicity end point was the cumulative paclitaxel dose at which grade 2 or higher (grade 2+) neuropathy was first reported; any patient not experiencing grade 2+ toxicity was censored at their cumulative dose received. The primary analysis plan was to use the log-rank test to determine whether there is a difference in risk of grade 2+ neuropathy across European-American patients classified by the CYP2C8*3 genotype. A standard α = 0.05 was utilized due to the single SNP–phenotype association tested in the primary analysis.

Following log-rank analysis, additional covariates [age (continuous variable), prior diagnosis of diabetes (yes versus no), taxane schedule (80–90 mg/m² weekly versus 175 mg/m² every 2 or 3 weeks), use of prophylactic or therapeutic neuropathy treatment (yes versus no)] were included in a multiple Cox proportional hazards model with CYP2C8*3 genotype assuming an additive genetic effect. Backward elimination using the Akaike information criterion (AIC) was used to select the final model. AIC balances model goodness of fit and complexity by penalizing the inclusion of extra covariates; it has been shown to be an effective model selection tool [23]. An additional replication of the positive findings in the European-American cohort was attempted via log-rank analysis in the self-reported African-American patients. Finally, the entire patient cohort was analyzed in a multiple Cox proportional hazards model which included self-reported race (European-American versus Non-European-American) in addition to the previously described covariates. All statistical analyses were carried out in the R statistical software, version 2.13.0 (R Development Core Team, Vienna, Austria).

results

patient population

A total of 411 paclitaxel-treated patients were eligible for analysis and successfully genotyped for CYP2C8*3 by DMET™ Plus. Demographic data including patient and treatment characteristics for the European-American (n = 209), African-American (n = 107), and the combined mixed-race cohort (n = 411) can be found in Table 1.

Table 1.

Characteristics of patient cohorts: European-American cohort (n = 209), African-American cohort (n = 107), and the entire mixed-race cohort (n = 411)

	European-American (n = 209)	African-American (n = 107)	Mixed-race (n = 411)
Age (years)
Median	51	46	50
Range	24–84	25–68	22–84
Self-reported race
European	209	0	287 (70%)
African-American	0	107	107 (26%)
Other	0	0	17 (4%)
Neuropathy
Grade 2+	35 (17%)	23 (21%)	76 (18%)
CYP2C8*3 (K399R) genotypea
Wild-type (*1/*1)	155 (74%)	101 (94%)	330 (80%)
Heterozygous (*1/*3)	51 (24%)	6 (6%)	76 (18%)
Variant (*3/*3)	3 (1%)	0	5 (1%)
Treatment before paclitaxel
AC (doxorubicin/cyclophosphamide)	154 (74%)	87 (81%)	316 (77%)
AC + bevacizumab	2	0	2
A (doxorubicin)	2	0	2
AC + docetaxel	1	0	1
None	50 (24%)	20 (19%)	90 (22%)
Treatment concurrent to paclitaxel
Trastuzumab	33 (16%)	21 (20%)	71 (17%)
Bevacizumab	4 (2%)	4 (4%)	9 (2%)
Carboplatin	1	1	2
Carboplatin + bevacizumab	1	2	3
Trastuzumab + lapatinib	4	0	4
Trastuzumab + cyclophosphamide	0	0	1
None	166 (79%)	79 (74%)	321 (78%)
Paclitaxel schedule and dose
80–90 mg/m2 weekly	64 (31%)	34 (32%)	131 (32%)
175 mg/m2 every 2 weeks	130 (62%)	64 (60%)	235 (57%)
175 mg/m2 every 3 weeks	15 (7%)	9 (8%)	45 (11%)
Total paclitaxel received
Median (mg/m2)	700	700	700
Range (mg/m2)	80–1280	160–1280	80–1280
Diabetes
Prior diagnosis	16 (8%)	24 (22%)	48 (12%)
No prior diagnosis	193 (92%)	83 (78%)	363 (88%)
Neuropathy prophylaxis or treatment
Gabapentin	6 (3%)	7 (7%)	15 (4%)
Amitriptyline	5 (2%)	1	9 (2%)
Glutamine	62 (30%)	22 (21%)	112 (27%)
Vitamin B complex	1	0	1
Vitamin B6	4 (2%)	2	7 (2%)
Total	78 (37%)	32 (30%)	144 (35%)
Treatment modalityb
Neoadjuvant	58 (28%)	47 (44%)	188 (46%)
Adjuvant	153 (73%)	61 (57%)	226 (55%)

Counts and percentages (in parentheses) are presented for categorical data. Medians and ranges are presented for quantitative data.

^aDifferences expected due to cohort composition and known differences in allele frequency between races.

^bTwo patients in the European-American and one in the African-American cohort were treated with paclitaxel neoadjuvantly and adjuvantly.

CYP2C8*3 K399R

The distribution of the CYP2C8*3 variant conformed to HWE separately in the European-American (P = 0.79) and African-American (P = 0.77) cohorts [24]. Allele frequencies in the European-American (AF = 0.14) and African-American (AF = 0.03) patients were consistent with that previously reported for each reference population in the International HapMap Project: CEU AF = 0.14 and ASW AF = 0.04 [17, 18]. The number of *1/*1, *1/*3, and *3/*3 women in each cohort is displayed in Table 1 along with the frequency of grade 2+ neuropathy (17%–21%), which was similar to that reported in prior paclitaxel studies.

neuropathy by genotype

analysis in European-American cohort

In the primary analysis of the European-American cohort, the log-rank test demonstrated a difference in risk of grade 2+ neuropathy across genotype groups, as displayed in the inverted Kaplan–Meier curves in Figure 1. As expected, risk of neuropathy was highest in patients who were homozygous for the *3 variant and lowest in patients homozygous for the wild-type allele (log-rank P = 0.006). Assuming an additive genetic effect, each variant allele a patient carried (0, 1, or 2) approximately doubled their neuropathy risk [HR (per allele) = 1.93, 95% CI: 1.05–3.55, P = 0.032].

Figure 1.

Incidence curve for grade 2+ neuropathy across genotype groups in the European-American cohort (n = 209). The highest risk was seen in the variant homozygotes and the lowest risk in the wild-type homozygotes.

Clinically relevant covariates (age, prior diabetes diagnosis, use of neuropathy prophylaxis or treatment, and paclitaxel schedule) were included in a multiple Cox proportional hazards model. The only covariate that was kept in the backwards selection procedure was diabetes history, which increased neuropathy risk as expected (Table 2). Similar to the unadjusted analysis, in the final model the association between CYP2C8*3 genotype and risk of grade 2+ neuropathy was significant [HR (per allele) = 1.95, 95% CI: 1.06–3.58, P = 0.031].

Table 2.

Multiple Cox proportional hazards model for grade 2+ neuropathy in the European-American cohort (n = 209) and the entire mixed-race cohort (n = 411)

	European-American cohort (n = 209)			Mixed-race cohort (n = 411)
	Hazard ratio	95% confidence interval	P-value	Hazard ratio	95% confidence interval	P-value
CYP2C8 additive genetic model	1.95	1.06–3.58	0.031a	1.98	1.25–3.13	0.004a
Diagnosis of diabetes	2.25	0.87–5.81	0.093	Excluded from final model
Age	Excluded from final model			1.00	1.00–1.04	0.102
Self-reported raceb	Not applicable			1.76	1.06–2.93	0.030a

CYP2C8*3 was assumed to have an additive genetic effect in which the estimated hazard ratio represents the increase in neuropathy risk per variant allele.

^aStatistically significant.

^bNon-European-American (African-American + other) versus European-American.

replication in African-American cohort

A total of 107 self-reported African-American individuals were assessable in the cross-race replication. As expected, the variant allele was substantially less common in this cohort (AF = 0.03) and there were no *3 homozygous individuals. Patients carrying one CYP2C8*3 allele had greater neuropathy risk than wild-type homozygous patients in the log-rank analysis (HR = 3.30, 95% CI: 1.04–10.45, P = 0.043) (Figure 2).

Figure 2.

Incidence curve for grade 2+ neuropathy across genotype groups in the African-American cohort (n = 107). Higher risk was seen in the carriers of the CYP2C8*3 variant when compared with the wild-type homozygotes (P = 0.043). There were no homozygous variant individuals in this cohort.

Cox proportional hazards model in a mixed-race cohort

A total of 411 paclitaxel-treated women were assessable in a Cox proportional hazards model that included self-reported race (European-American versus non-European-American). In the final model, the association between increased risk of grade 2+ neuropathy and CYP2C8*3 was significant [HR (per allele) = 1.98, 95% CI: 1.25–3.13, P = 0.004] (Figure 3). There was also a higher risk of grade 2+ neuropathy in non-European women (HR = 1.76, 95% CI: 1.06–2.93, P = 0.030) (Figure 4) and a non-significant, negligible influence of patient age (HR = 1.00, 95% CI: 1.00–1.04, P = 0.102), but not of the other covariates, including diabetes (Table 2).

Figure 3.

Incidence curve for grade 2+ neuropathy across genotype groups in the mixed-race cohort (n = 411) indicating an approximate doubling of grade 2+ neuropathy risk for each *3 variant a patient carries, supporting an additive genetic effect.

Figure 4.

Incidence curve for grade 2+ neuropathy across racial groups in the entire mixed-race cohort (n = 411). After adjusting for CYP2C8*3 and age, non-European Americans (n = 124) were at higher risk of grade 2+ neuropathy than European Americans (n = 287) (P = 0.030).

discussion

CYP2C8*3 has previously been implicated as a potential risk factor for paclitaxel-induced peripheral neuropathy [20, 21]. The current study, which analyzed a single candidate SNP in a large cohort of patients, had adequate statistical power to determine the effect of CYP2C8*3, and other potentially relevant clinical covariates, on neuropathy risk. We successfully replicated an association between the CYP2C8*3 K399R (rs10509681) variant and increased risk of grade 2+ paclitaxel-induced neuropathy in both European-American and African-American patient cohorts. Finally, a model was built that included all patients and relevant clinical covariates, and after adjusting for age and race, the risk of grade 2+ neuropathy approximately doubled for each CYP2C8*3 allele a patient carried.

The prior report of an association between CYP2C8*3 and paclitaxel-induced neuropathy from Leskela et al. was carried out in a cohort of Spanish paclitaxel-treated cancer patients. Using a similar analysis plan, end point, and covariates, the current study confirmed the impact of CYP2C8*3 in a larger cohort with robust statistical power (n = 411 versus n = 118). An important aspect that differentiates this study from that of Leskela et al. [21] and most other pharmacogenetic studies is the single genotype–phenotype association that was selected a priori for analysis. The vast majority of pharmacogenetic studies are hypothesis-generating, discovery phase studies with multiple variants and clinical end points tested. Based on data from our group and others, this study exclusively tested the hypothesis that the CYP2C8*3 variant increases risk of paclitaxel-induced neuropathy. Confirming this hypothesis in two patient cohorts that are independent from the discovery cohort is strong evidence of the clinical validity of this association [25].

Other attempts at demonstrating this association may have suffered from suboptimal patient inclusion [26–29]. The study from Rizzo et al. [26] combined patients on either paclitaxel (24%) or docetaxel (76%) and analyzed these groups together. While docetaxel has both structural and mechanistic similarities with paclitaxel, it has a lower incidence of neurotoxicity and is not metabolized by CYP2C8. The other three studies utilized patient cohorts that were concomitantly treated with the neuropathic agent carboplatin [30]. Systematic concomitant treatment with another neurotoxic agent that is not interacting with CYP2C8 could dilute the apparent estimate of the effect of CYP2C8 genotype on neuropathy risk, leading to false-negative findings. In the present study, all patients were treated with paclitaxel and only six patients (1.5%) were treated previously or concurrently with a neuropathic chemotherapeutic agent (docetaxel or carboplatin) (Table 1). Post hoc exclusion of these subjects for sensitivity testing had a negligible influence on our results (data not shown).

Based on HapMap reference populations, the *3 variant is not found in patients of African descent [Yoruban in Ibadan, Nigeria (YRI) AF = 0.00]. The AF in our African-American patients (AF = 0.03) was consistent with that reported for individuals of African ancestry living in the United States (ASW AF = 0.04), likely reflecting the known admixture found in African-American patients [31]. Based on our findings, not only is the *3 variant found in individuals from a wide range of self-reported races, the approximate doubling of neuropathy risk it confers is consistent across racial groups.

Beyond CYP2C8 genotype, non-European individuals were at an increased risk of neuropathy (HR = 1.76, 95% CI: 1.05–2.93, P = 0.031), supporting a recent publication [8] and a preliminary report [32] from two separate large clinical trials of paclitaxel use in breast cancer patients. These findings suggest that while CYP2C8*3 is one factor that influences a patient's risk of paclitaxel-induced neuropathy, perhaps along with age and diabetes history, there are other currently unappreciated factors at work. It is noteworthy that the risk of HIV-associated distal neuropathy [33] and diabetes-related neuropathy [34] is greater in African Americans than in European Americans; thus, there seems to be a general predisposition to neuropathy in African Americans irrespective of etiology. We hypothesize that there are inter-race differences in frequencies of genetic loci responsible for this phenotype.

The major limitation of this study is the retrospective use of a clinical registry instead of a prospective clinical study. This manifests in a number of ways, most notably the differences in paclitaxel treatment and schedule, the use of neuropathy prophylaxis or treatment, and the potential non-uniformity in toxicity collection. We have attempted to adjust for these factors where possible. In meta-analyses comparing the risk of neuropathy for the weekly versus tri-weekly schedules, the dose intensity was found to be a more important factor than schedule itself [35]. The patients in this analysis were treated with one of three standard paclitaxel regimens: 3-h infusion of 175 mg/m² every 3 weeks (58.3 mg/m²/week) or every 2 weeks (87.5 mg/m²/week) or a 1-h weekly infusion (80–90 mg/m²/week). Despite attempts to include these data in Cox models, we could not detect a significant influence on neuropathy risk for paclitaxel dose, schedule, or infusion time, all of which are highly collinear in this dataset.

In conclusion, we have replicated in two racially homogenous populations our previous finding that CYP2C8*3 carriers are more likely to experience peripheral neuropathy when treated with paclitaxel. Analysis of the influence of a single SNP (CYP2C8*3) on the risk of a defined adverse event (grade 2+ neuropathy) in a large cohort of paclitaxel-treated patients enabled a definitive assessment of this association, which remained significant after adjustment for clinical covariates that are thought to modify risk of neuropathy. In our mixed-race cohort, the risk of neuropathy approximately doubled for each *3 variant a patient carried, indicating a clear additive or gene-dose effect (Figure 3). Patients who are known to carry CYP2C8*3, and particularly those known to be homozygous, should be assumed to be at increased risk of peripheral neuropathy during paclitaxel treatment. Additional work is needed to translate these findings to clinical practice, such as a genotype-guided dosing study. A prospective study could demonstrate the clinical utility of individualizing paclitaxel therapy based on CYP2C8 genotype, enabling optimization of the risk/benefit profile of this essential chemotherapeutic agent.

funding

This work was supported by the Breast Cancer Research Foundation (to LAC); by Lineberger Comprehensive Cancer Center; by a Clinical and Translational Science Award 5UL1RR025747-04; by the NIH-National Institute of General Medical Sciences T32GM081057; and by the NIH SPORE in Breast Cancer 5P50CA058223.

disclosure

The authors have declared no conflicts of interest.