The samples were collected to the National Institute “and CETAC Research Center Policy” of Naples, Italy. This retrospective work was performed in compliance with the ethical values laid down by the Declaration of Helsinki, and informed consent documentation was reviewed and agreed by the independent ethics committee and CETAC Research Center policy. The study was planned to measure whether the PGx profile can affect taxane-induced neurotoxicity. In total, 76 cancer patients (29 male and 47 female) were enrolled, 35 of whom received adjuvant taxane chemotherapy (Table 1).

All patients had a diagnosis of carcinoma (primarily, breast, ovarian, genitourinary, etc.) and were treated with paclitaxel- or docetaxel-based therapy. The chemotherapy dose, schedule, and duration were as follows: for paclitaxel 175 mg/m² intravenously (IV) every 3 weeks for 4 cycles, (for adjuvant treatment for breast cancer) and/or 80 mg/m² weekly IV for 12 cycles, and for docetaxel IV 100 mg/m² for 4 cycles for firstline metastatic cancer.

The sample included 39 patients aged <60 years (51.3%) and 37 aged ≥60 years, who were separately analyzed and evaluated in relation to risk factors for neurotoxicity. The patients were separated into two arms: those with and those without neurotoxicity. Grade <2 and grade ≥2 neurotoxicity were also individually registered.

The inclusion criteria were: patients >18 years old, male and female, Eastern Cooperative Oncology Group performance status of 0–1, histologically proven cancer, and on a paclitaxel or docetaxel regimen. All enrolled patients were without comorbidity causative of peripheral neuropathy (i.e., diabetes). At the time of the neurotoxicity event, the patients were not receiving other anticancer agents.

The assessment of neurotoxicity was evaluated based on symptom narration, the presence of symmetrical “stocking-glove” numbness, loss of deep tendon reflexes, and burning and/or tingling after therapy. Baseline taxane-induced peripheral neurotoxicity was assessed according to the National Cancer Institute Common Toxicity Criteria for Adverse Events (NCI-CTCAV) version 4.0 grading scale from 0 to 4:

Genomic DNA was extracted with a mouth swab in accordance with the manufacturer's protocol using an Ampli-DNA extraction kit (Dia-Chem, srl, Naples, Italy).

The genotyping assay was performed using the TaqMan probe-based chemistry allelic discrimination assay in the OneStep platform (Life Technologies, Monza, Italy). The investigating panel test included the CYP2C8^*3, CYP3A4^*22, GSTP1, ERCC2, SLCO1B1, ABCG2, and XRCC3 polymorphisms. The reaction mix and temperature protocol (95°C for 15″ and 60°C for 1 min for 40 cycles) were performed in accordance with the manufacturer's protocol (Life Technologies, Monza, Italy). The primers and probe were designed by PrimerExpress 3.0 (Life Technologies) for the allelic discrimination assays (Supplementary Table 1). In addition, MDR1 and CYP3A4^*1B assay were performed using Ampli-MDR1 and Ampli-CYP3A4 kits (Dia-Chem, srl) in order to confirm the previously reported data (Bosch et al., 2006; Kus et al., 2016).

Differences according to age, gender, and adverse events, in particular for neutropenia and neuropathy, between taxane users and the control cohort were calculated using the Chi-square test. Univariate analyses were performed to match the two arms: the unadjusted logistic regression method was used to assess crude odds ratios (ORs) and 95% confidence intervals (CIs). Logistic regression models adjusted for major confounders like age and gender were used to calculate adjusted ORs and 95% CIs for each gene variants risk factors. Analyses were performed using SPSS for Windows, version 23.0 (IBM Corporation, NY, USA). A bilateral p < 0.05 was considered statistically significant.

Thirty-five cancer patients (26 female and 9 male) who received adjuvant taxane therapy were enrolled in this retrospective study. Of these, 23 (65.7%), had experienced an adverse event, and 5 (14.3%) of them were >grade 2 (Table 1). Six (17.1%) patients experienced neurotoxicity. In the sample, 22 subjects were under 60 years old, and 13 were over the age of 60 years. The sharing of genetic SNPs in agreement with risk factors is listed in Table 2. The control cohort was represented by 41 cases who did not receive a taxane; they were treated primarily with platinum and fluoropyrimidine-based chemotherapy. It is worth noting, is that the case cohort (taxane users) recorded more adverse events than the control cohort, with 23 (65.7%) and 12 (28.6%), respectively (p = 0.001). In addition, more taxane users experienced toxicity than the control, with 5 (14.3%) and 1 (2.4%), respectively (p = 0.003).

Several criteria were considered for selecting gene variants for the pharmacogenomic panel tests: (i) search on the whole standardized polymorphisms acknowledged to influence the pharmacokinetics/pharmacodynamics of taxanes (www.pharmgkb.org); (ii) review of current researches, particularly trials including polymorphisms related to toxicity; (iii) identification of issues related to the impact of genotyping testing which might provide answers concerning the incorporation of PGx markers in clinical practice.

The genotypes of all genes analyzed in this study and their relation with any grade of neutropenia and neuropathy are summarized in Table 2. Distribution of genetic polymorphism according to neutropenia and neurotoxity, with the distinction for grading among 35 Taxane users only are provided (Supplementary Table 1).

The ABCB1 3435C>T rs1045642 genotype of taxane users was divided into two groups: TT allele vs. CT+CC alleles (4 cases, 11.5%). The OR for every toxicity grade was 1.67 (0.26–10.67, P = 0.50), when compared with CT+CC (medium and low risk, respectively) allele genotype.

The ABCB1 2677G>T/A rs2032582 genotype was divided into two groups: TT/AA allele vs. GT/A and GG alleles (2 cases, 6.10%). OR for every toxicity grade was 0.44 (95% CI: 0.07–2.76, P = 0.40), when compared with GT/A+ GG alleles (medium and low risk, respectively) allele genotype.

The CYP3A4^*1B−392A>G rs2740574 genotype was divided into two groups: AG+GG risk allele vs. AA alleles (9 cases, 25.7%). OR for every toxicity grade was 0.60 (95% CI: 0.20–1.83, P = 0.70), when compared with AA (low risk) allele genotype.

CYP2C8^*3 Lys399Arg rs10509681 genotype was divided into two groups: CT allele vs. TT alleles (10 cases, 28.6%). OR for every neuropathy grade was 1.62 (95% CI: 0.49–5.35, P = 0.19), when compare with CT (medium risk) allele genotype.

The CYP3A4^*22 intron 6 rs35599367 genotype was divided into two groups: CC allele vs. CT genotypes, No homozygous for TT were detected. The OR for any neuropathy grade was 2.17 (95% CI: 0.48–9.79, P = 0.30), in a boon of the CT genotype.

The GSTP1 Iso105Val rs1695 genotype was divided into two groups: GG allele vs. AG+AA genotypes. The OR for any grade neuropathy was 1.25 (95% CI: 0.44–3.60, P = 0.71).

The ERCC2 2251T>G Lys751Gln rs13181 (alias XPD) genotype was divided into three groups: TT vs. GT and GG genotypes. The OR for any grade neuropathy was 1.76 (95% CI: 0.66–4.71, P = 0.41), and the OR for GG was 1.97 (95% CI: 0.51–7.68, P = 0.41).

The SLCO1B1 Val174Ala rs4149056 genotypes was were separated into two groups: TT vs. CC+CT genotypes. The OR for any grade neuropathy was 1.03 (95% CI: 0.35–3.06, P = 0.62).

The ABCG2 G34A Val12Met rs2231137 genotype was divided into three groups: GG vs. GA and AA genotypes. The OR for any neuropathic grade was 1.00 (95% CI: 0.20–4.84, P = 0.41), for GA, and the OR for AA was 3.33 (95% CI: 0.61–18.5, P = 0.41).

The XRCC3 316A>G Thr241Met rs1799794 genotype was divided into two groups: AA vs. GA and GG genotypes. The OR for any neuropathy grading was 1.52 (95% CI: 0.51–4.91) for GA alleles, and the OR for GG+GA was 2.61 (95% CI: 0.91–7.61) P = 0.03.

Multiple genotyping methods have been validated for assessing the mutational profile of the mentioned SNPs, but no gold standard has been defined. Moreover, only a few studies have addressed the cost-effectiveness of pharmacogenomic testing in terms of the implications for clinical practice (Payne and Shabaruddin, 2010; Tirelli et al., 2011). For instance, Van den Akker-van Marle et al. (2006) integrated thiopurine S-methyltransferase (TPMT) genotyping prior to scheduling 6-mercaptopurine in pediatric Acute Lymphoblastic Leukemia (ALL); the mean calculated cost has been estimated to average around €150,00 (Van den Akker-van Marle et al., 2006).

In a further report, an early outline of the genotyping costs for “home-made tests” using allele discrimination on the fluorescent-based platform, was calculated at about €20,00 per SNP (Di Francia et al., 2011). Realistic selection of the optimal method in terms of costs per samples is dependent on the specific test provided by laboratories (Di Francia et al., 2010). Our PGx panel assay identifies 5 polymorphisms, and it cost is averaged to €100,00.

Moreover, the major issues to be considered by clinical laboratories providing genotyping services, are: (i) the ease of use of FDA-approved tests; (ii) the lack of government reimbursement; (iii) the need for genotyping accuracy; and (iv) the requirement to find expert clinicians who areable to correctly understand the PGx results (Di Francia et al., 2012b).

The clinical effectiveness of the polymorphism described here, could help in developing new diagnostic tool for driving treatment decisions (Frederiks et al., 2015). In particular, molecular testing for a mutation in the ABCB1 (alias MDR1), CYP3A4^*1B, CYP2C8^*3, and XRCC3 genes will possibly help oncologists to select subjects who are most expected to avoid taxane neurotoxicity. For assessing a basic profile of patients responding well/poorly, a panel test of five genetic variants is planned (Table 3).

The aspects addressed here could help clinicians to stratify patients' profiles from genotype A (the most likely responders to treatment) to genotype C (bad responders with higher odds risk of acute and cumulative neurotoxicity). The PGx profile defined as low risk for toxicity showed wild-type expression of ABCB1 (3435CC and 2677GG), conferring the normal intrusion/extrusion of taxane and active metabolites. In addition, the regular CYP2C8 and CYP3A4 392GG allotypes ensure appropriate metabolic activity. Also, XRCC3 316AA exhibits regular expression, as well as normal DNA repair activity.

Pharmacogenomic profiles can show a predisposition to a higher neutropenia and neuropathy risk by reveling higher transmembrane expression of the ABCB1 (3435TT and 2677TT/AA), variant phenotype, conferring the excessive extrusion of taxane from neoplastic cells and causing high plasma concentration. Also poor metabolic activity due to CYP2C8^*3 (399R) and CYP3A4^*1B (392 AA/AG) causes pharmacokinetic problems, and lower expression of the XRCC3 316AG/GG phenotype probably interferes with DNA replication of neoplastic cells and less likely with that of hematopoietic cells, resulting in severe neutropenia, as previously observed in a Caucasian population (Qiu et al., 2013).

There have been certain restrictions in our projected panel tests: (i) these PGx signatures need to be validated in multiple clinical trials with a larger number of patients; (ii) our genotyping data are limited to a Caucasian population; (iii) we did not adjust our data for multiple comparisons (i.e., type of cancer) due to a low number of cohort samples; (iv) the selection of the gene variants was made on the basis of recent findings in clinical trials with significant correlations between the PGx profile and taxane treatments. However, with regard to the gene variants analyzed in this pilot study, the single endpoint was to evaluate the usefulness and cost-effectiveness of a PGx panel assay suitable for application in clinical practice, with particular attention to so-called “frail patients” (Berretta et al., 2013a, 2016).

Furthermore, defining an individual PGx profile does not afford a unique target to assess the optimal strategic approach for the management of taxane-induced neuropathy; thus, it is necessary to seek complementary and alternative medicines (Berretta et al., 2017), as well as to look at nutrition (Berretta et al., 2013b).

In the next few years, it can be expected that there will be links between pharmaceutical and biotechnology companies to undertake larger and broader studies validating tests available for routine diagnostics in pharmacogenomics concerning paclitaxel and docetaxel. Currently, our proposed pharmacogenomic panel assay is useful because it is low cost (about €100,00/genotype/patient) and it is suitable for most clinical laboratory with real time-PCR equipment. In addition, high genomic expertise is not needed to interpret genotype results (Table 3).

In summary, clinicians and laboratory managers should join in evaluating the benefits and limitations, particularly regarding costs and applicability, of the pharmacogenomic tests that are likely suitable for routine clinical practice integration.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.