Pharmacogenomic tests of oncology drugs at Instituto Nacional de Câncer (INCA)

Assembling collaborators from various backgrounds is widely recognized as a critical factor for the successful implementation of PGx programs.^[ 4 5 ^] At INCA, this was accomplished by creating a task force comprising professionals from various units of the institution (SupplementaryTable 1). This task force overlooked the implementation of the PGx testing program and designed the project's standard operating procedures (SOPs). Selection of drug-gene pairs

Selection was based on the availability of international, authoritative guidelines with clearcut recommendations for PGx-informed drug prescription based on clinically-validated PGx associations. Two sets of guidelines fulfilled this condition, namely CPIC (Clinical Pharmacogenetics Implementation Consortium)^[ 6 ^] and DPWG (Dutch Pharmacogenetics Working Group)^[ 7 ^] guidelines. These guidelines cover four chemotherapeutic druggene pairs, namely:

Fluoropyrimidines- DPYD, irinotecan- UGT1A1 and thiopurines- TPMT/NUDT15 were selected as initial targets of INCA's PGx program. Tamoxifen, which requires genotyping of several polymorphisms in CYP2D6, including copy number variation, will be eventually added to the PGx program.

Gastrointestinal cancer patients from INCA, potential candidates for chemotherapy with inrinotecan and/ or fluoropyrimidines were selected for genotyping DPYD and UGT1A1 variants, whereas ALL pediatric patients from INCA and seven other cancer hospitals from Rio de Janeiro and Sao Paulo, enrolled in a project focused on genomic deletions (principal investigator: Dr. Mariana Emerenciano), were genotyped for TPMT and NUDT15 polymorphisms.

Allele discrimination Taqman assays for the selected polymorphisms in DPYD, UGT1A1, TPMT and NUDT15 were validated at INCA's PGx lab. All genotyped polymorphisms affect the metabolism (biotransformation) of the target drugs and thereby modulate the metabolic phenotypes of the respective gene products. Assignment of metabolic phenotypes to individual diplotypes was done accordingly to the CPIC and/or DWPG guidelines.

The patient's genotype(s), inferred phenotype(s) and dosing recommendations according to the CPIC and/ or DPWG guidelines are inserted in a brief report (2 pages). The report includes also my institutional contacts - for consultation regarding the procedures, results and interpretation of the PGx tests - and the following disclosure statements: i) the dosing recommendations are based on the CPIC and/or DPWG guidelines according to the polymorphisms genotyped; ii) the possibility of influence of other, noninterrogated genetic variants cannot be excluded; iii) adherence to the dosing recommendations is a decision of the prescribing physician. For INCA's gastrointestinal cancer patients, reports are emailed to the head of the clinical oncology service and inserted in the electronic medical records, whereas in the case LLA children, the reports are forwarded to Dr. Emerenciano (see above).

The next sections present results from the implementation phase of the project. For each druggene pair, the following aspects will be covered: • Aims and rationale for the PGx tests;

Background information on the relevant pharmacogenes/polymorphisms, respective metabolic phenotypes, and evidence for their association with drug responses are beyond the scope of this article, but may be found in the CPIC and DPWG guidelines.^[ 8 9 10 13 14 ^] An overview of Brazilian studies related to PGx in oncology is also available.^[ 15 ^]

The PGx test targets UGT1A1 polymorphisms that encode UGT1A1 isoforms with reduced/null enzymatic activity. UGT1A1 converts the active metabolite of irinotecan (SN-38) into inactive SN38 glucuronide; consequently, UGT1A1 isoforms with reduced/null activity, lead to accumulation of SN-38, and increased risk of irinotecan-induced adverse effects. The DPWG guidelines are based on a UGT1A1 polymorphism (rs8173547), which modifies the number of TA repeats in the gene's promoter region, thus affecting the expression and activity of UGT1A1. The rs8173547 polymorphism is in virtual complete linkage disequilibrium with rs887829, a C>T transition,^[ 2 16 ^] which is genotyped at INCA in the PGx test for irinotecan. The variant T allele associates with reduced UGT1A1 expression and enzymatic activity, and consequently reduced SN-38 glucuronidation, SN-38 accumulation and increased risk of irinotecan-induced adverse effects.

The minor allele frequency (MAF) of rs887829 in samples from INCA's gastrointestinal cancer patients (n=150) was 34.1% (Table 1), in excellent agreement with the MAF (35.2%) reported in the ABraOM repository of genomic variants for 1,171 unrelated Brazilians from the Southeast region.^[ 17 ^]

Based on the rs887829C>T genotypes, 40.1% (CC), 49.2% (CT), and 9.7% (TT) of tested patients were assigned to normal, intermediate and poor UGT1A1 phenotypes, respectively. According to the DPWG guidelines, which we adopted (Table 2), only poor metabolizers require adjustment of the initial irinotecan dose. The recommendation is to “start treatment with 70% of the standard dose; if the patient tolerates this initial dose, the dose can be increased, guided by the neutrophil count”.

The PGx tests implemented at INCA target four DPYD polymorphisms (Table 3), which encode isoforms of the enzyme dihydropyrimidine dehydrogenase (DPD) with reduced (rs67376798 and rs75017182) or null activity (rs391820 and rs55886062). DPD accounts for ~85% of 5-fluoropyrimidine (5-FU) elimination in humans, and consequently, reduction or loss of DPD activity leads to 5-FU accumulation and increased risk of toxicity. This affects patients treated with either 5-FU or capecitabine, since 5-FU is the active metabolite of capecitabine.

The targeted DPYD polymorphisms were rare (MAF<0.5%) or absent in gastrointestinal cancer patients from INCA (n=150;Table 3). These findings are consistent with data from the ABraOM repository,^[ 17 ^] as well as from previous studies of Brazilian patients with severe fluoropyrimidine-induced toxicity.^[ 18 19 ^]

The CPIC and DWPG guidelines for fluoropyrimidines attribute “activity scores” (AS) to the DPYD variants (Table 3) and DPYD diplotypes (Table 4). The DPD metabolic phenotypes - normal, intermediate and poor metabolizers - are then assigned according to this AS system. The vast majority (98.7%) of samples from INCA was assigned the normal metabolizer phenotype; poor metabolizers were not identified, whereas intermediate metabolizers ammounted to 1.3% of the samples (Table 4). The CPIC guidelines, adopted at INCA, recommend avoidance of fluoropyrimidines in DPD poor metabolizers and dose reduction (25-50%) for intermediate metabolizers; the extent of the dose reduction depends on the AS of the DPYD diplotypes, as shown inTable 4 . It is suggested that if no, or clinically tolerable, toxicity occurs in the first two cycles, the fluoropyrimidine dose may be increased in subsequent cycles.

The combined frequency of intermediate and poor metabolizers in the INCA samples (1.3%), is considerably lower than the frequency of fluoropyrimidine-induced severe adverse effects, which may reach 35% in Brazilian patients with digestive cancer.^[ 19 ^] This discrepancy indicates that only a proportion of the adverse effects may be attributed to the interrogated DPYD variants, and has prompted adoption of phenotyping methods to assess, directly or indirectly, DPD enzymatic activity. These methods have been explored in Brazilian studies^[ 18 19 ^] and are listed in the DPWG guidelines as an alternative to guide fluoropyrimidine dosing when the DPD phenotype cannot be inferred from DPYD diplotypes.^[ 9 ^] The CPIC fluoropyrimidine guideline, however, does not include DPD phenotypic tests, and they are currently not implemented at INCA.

Mercaptopurine and thioguanine, commonly prescribed for lymphoid malignancies and myeloid leukemias, respectively, are prodrugs that require conversion into active metabolites to exert their clinical effects. These thiopurines and their metabolites are substrates for several enzymes, of which TPMT and NUDT15 are the most relevant for PGx testing. The PGx tests implemented at INCA target TPMT (rs1142345, rs1800460 and rs1800462) and NUDT15 (rs116855232) polymorphisms, associated with reduced/null activity of the respective enzymes, leading to accumulation of active thiopurine metabolites and increased risk of toxicity.^[ 13 ^]

Table 5 shows the frequency distribution of the TPMT and NUDT15 polymorphisms in 162 pediatric LLA patients genotyped at INCA. All variant alleles were identified in the cohort, their MAF's ranging from 0.3 to 4.9%, in good agreement with the ABraOM repository^[ 17 ^] and a previous study from our group.^[ 20 ^]

Thiopurine dosing recommendations are based on TPMT and NUDT15 compound (combined) metabolic phenotypes, inferred from the respective diplotypes, as proposed in the CPIC guidelines.^[ 13 ^] Accordingly, there are 5 possible compound phenotypes (Table 6) with distinct dosing recommendations: i) normal metabolizer phenotype, assigned to 87.5% of the samples, no change in thiopurine initial dose is required; ii) intermediate metabolizers of either enzyme (11.3%), consider starting the treatment with reduced doses; iii) intermediate metabolizers of both enzymes (“compound intermediate metabolizers”), dose reduction is recommended in view of the considerably increased risk of toxicity; iv and v) poor metabolizers of one (0.6%) or both enzymes (none observed in the tested samples), drastic dose reduction of the initial dose and reduced frequency of administration is strongly recommended. Following the initial reduction, thiopurine doses should be adjusted based on the degree of myelosuppression and the disease-specific guidelines.

The number of patients needed to be genotyped (NNG) in order to prevent one drug-induced adverse event is a validated metric to assess the potential population impact of PGx tests.^[ 21 22 ^] It appears intuitive that the lower the NNG, the greater the perceived clinical utility and cost-effectiveness of the PGx test. NNG for the PGx tests are estimated from the odds ratio (OR) of PGx associations with drug-induced adverse effects, the frequency of the adverse reactions (denoted q ) and the frequency ( p) of the PGx variant(s) linked to the adverse effects.^[ 21 22 ^] Using OR and q values from published studies and p- values for the relevant metabolic phenotypes in the INCA samples, the NNG values plotted inFigure 1 are obtained.^[ 22 ^] The figure shows also the number of patients needed to be treated (NNT = NNG x p ) to avoid one adverse effect. The NNG values ranged widely, from 14 for UGT1A1 irinotecan in high doses (>300mg.m² ) to 312 for DPYD- fluoropyrimidines, whereas the NNTs for the four gene-drug pairs varied over a considerably narrower range (2-4). Using the TPMT/NUDT15 thiopurines as an example, the results inFigure 1 imply that it is necessary to genotype 38 LLA patients for TPMT and NUDT15 polymorphisms (NNG) and adjust the initial thiopurine dosing in four patients (NNT, i.e., those who tested positive for variants in either or both genes) in order to prevent one additional adverse effect to thiopurines.

The INCA PGx testing program had been successfully implemented by the end of 2019. Early in 2020, an expansion of this program, to provide PGx tests for irinotecan and fluoropyrimidines to patients treated at three cancer centers in the states of São Paulo and Rio Grande do Sul were approved by CEP-INCA. Other developments of the PGx program were also envisaged:

Unfortunately, the outbreak of the COVID-19 pandemic imposed a postponement, but certainly not suppression, of these developments; of note, testing for thiopurines TPMT/NUDT15 was not interrupted and a total of 278 children have been tested at the time of this writing. We are aware that continuity and, especially, expansion of the PGx program will encounter challenges not only of assuring financial support but also of availability of trained personnel, adoption of the PGx-informed prescription by the clinical staff, updating of the dosing recommendations as novel evidence of PGx associations emerge, and, ultimately, evidence of clinical utility and cost-effectiveness in a realistic scenario.

Cost-effectiveness may appear as of especial concern for fluoropyrimidines, in view of the rarity of the targeted DPYD polymorphisms, and, consequently the large NNG estimated from the available data (Figure 1). However, the previously mentioned overview of PGx tests in oncology^[ 2 ^] identified 72 published PGx studies in high-impact journals, which reported clinical associations of DYPD variants or DPD phenotypes with outcomes of treatment with fluoropyrimidines. Accordingly, preemptive PGx testing for fluoropyrimidines is now required by the Swiss Agency of Therapeutic Products and recommended by EMA, the European Medicines Agency (Table 7), whereas in France, health authorities recently made mandatory the routine determination of DPD deficiency prior to administration of fluoropyrimidine-based chemotherapy.^[ 23 ^] Other regulatory bodies, however, acknowledge the “actionability” of DPYD and DPD polymorphism but do not recommend or require PGx tests to guide fluoropyrimidine dosing (Table 7). Heterogeneity/ discordance among international regulatory bodies with respect to PGx tests extends to the other drug/ gene pairs interrogated at INCA, as shown inTable 7 . To my knowledge, there are no recommendations from ANVISA (Brazilian Health Regulatory Agency) or Brazilian professional organizations, regarding PGx tests for germline variants associated with clinical response to oncology drugs. Writing for the Brazilian Journal of Oncology, I would like to challenge the SBOC (Brazilian Society of Clinical Oncology) to take a leading role in the assessment of the clinical utility of PGx tests of germline variants for oncology drugs prescribed to Brazilians.