From DPYD genotype to CPIC activity score to 5-FU/capecitabine dose adjustment — preventing severe fluoropyrimidine toxicity before the first infusion
5-Fluorouracil (5-FU) and its oral prodrug capecitabine remain backbone agents in colorectal, breast, gastric, and head-and-neck cancer regimens, prescribed to roughly 2 million patients worldwide each year. Their narrow therapeutic index means that the enzyme responsible for degrading them — dihydropyrimidine dehydrogenase (DPD), encoded by DPYD — is the single most important determinant of whether a standard dose is safe or lethal. Before any infusion is scheduled, oncology teams must establish the patient's DPD capacity.
5-FU is a pyrimidine antimetabolite that is anabolized intracellularly to FdUMP, FdUTP, and FUTP, which inhibit thymidylate synthase and become misincorporated into RNA/DNA — killing rapidly dividing cells in tumor and normal tissue alike. Only ~1–3% of an administered dose reaches this anabolic (cytotoxic) pathway; the remaining >80% is shunted through the catabolic pathway in the liver, where three sequential enzymes degrade it to inert products:
• Dihydropyrimidine dehydrogenase (DPD) — reduces 5-FU to dihydrofluorouracil (DHFU); this is the rate-limiting, and by far most clinically important, step • Dihydropyrimidinase (DPYS) — converts DHFU to fluoro-ureidopropionic acid (FUPA) • β-ureidopropionase (UPB1) — converts FUPA to α-fluoro-β-alanine (FBAL), excreted renally
Because DPD sits at the very first and rate-limiting step, any reduction in its activity does not modestly slow clearance — it causes 5-FU plasma half-life and AUC to rise supralinearly. A patient with only 40% normal DPD activity does not simply retain 60% more drug; systemic exposure (AUC) can increase 2–3 fold because the enzyme is easily saturated even at partial capacity, pushing far more of the dose into the cytotoxic anabolic pathway in normal tissue.
Historically, 5-FU/capecitabine was dosed purely by body-surface area, with severe (Grade ≥3 CTCAE) toxicity — myelosuppression, mucositis, hand-foot syndrome, diarrhea — occurring in 10–40% of patients depending on regimen, and treatment-related death in an estimated 0.1–1% of all treated patients. Retrospective series consistently found that a disproportionate share of fatal and near-fatal cases occurred in patients with unrecognized partial or complete DPD deficiency.
Complete DPD deficiency is rare (~1 in 10,000), but partial deficiency attributable to the four CPIC core DPYD variants is carried by an estimated 3–8% of patients of European ancestry, and DPYD variant carriers account for a disproportionate ~25–30% of severe fluoropyrimidine toxicity cases despite being a small minority of all treated patients. This risk concentration is what makes pre-emptive genotyping cost-effective: identifying a small, high-risk subgroup before the first dose is given, rather than discovering deficiency after a life-threatening admission.
DPYD is a large gene (~950 kb, 23 exons) with hundreds of catalogued variants, but only a handful have been rigorously validated as clinically actionable. CPIC and the Dutch Pharmacogenetics Working Group (DPWG) restrict routine testing to four core variants with strong evidence of reduced or abolished DPD function, genotyped from a single blood draw before the first fluoropyrimidine dose is dispensed.
• DPYD*2A (c.1905+1G>A, rs3918290) — disrupts the canonical splice donor site of intron 14, causing exon 14 skipping and a truncated, non-functional protein. No-function allele (activity value 0). Allele frequency ~1% in Europeans; the single most extensively validated DPD-deficiency variant.
• DPYD*13 (c.1679T>G, rs55886062) — missense substitution (Ile560Ser) in a highly conserved FAD-binding domain, abolishing catalytic activity. No-function allele (activity value 0). Rare (~0.07% allele frequency) but high penetrance for severe toxicity.
• c.2846A>T (rs67376798) — missense substitution (Asp949Val) causing partial loss of function. Reduced-function allele (activity value 0.5). Allele frequency ~0.5–1%.
• c.1129-5923C>G / HapB3 (rs75017182) — a deep-intronic variant in linkage with a synonymous exon 10 change; creates a cryptic splice site producing partially aberrant transcripts alongside some normal transcript. Reduced-function allele (activity value 0.5). Allele frequency 3–5% in Europeans — the most common of the four, though individually lower-risk than *2A or *13.
Genotyping platforms (TaqMan 5′-nuclease allelic discrimination, or a broader pharmacogenomic NGS/SNP-array panel) call each position with >99.5% accuracy; any ambiguous or novel-variant heterozygous call triggers confirmatory Sanger sequencing before results are released to the ordering oncologist.
The core CPIC panel is deliberately conservative — it includes only variants with consistent, reproducible association with severe toxicity across multiple independent cohorts. This specificity comes at the cost of sensitivity: an estimated 30–50% of patients who present with clinically severe DPD-deficiency-driven toxicity carry no detectable variant on the 4-variant panel, either because of rare/private DPYD variants not on the panel, non-genetic causes of low DPD activity (hepatic impairment, drug interactions), or genuinely normal genotype with toxicity from other causes. This is the central rationale for pairing genotyping with a complementary phenotypic assay (Stage 4).
A raw genotype call is not directly actionable — it must be translated into a predicted metabolic phenotype using a standardized, reproducible algorithm. CPIC's Gene Activity Score (AS) system assigns each DPYD allele a numeric function value and sums the two allele values per patient, producing a continuous 0–2 scale that maps directly onto dosing recommendations.
Each DPYD allele is assigned a function value: 1.0 for a normal/functional allele (e.g. the reference *1 allele), 0.5 for a reduced-function allele (c.2846A>T, HapB3), and 0 for a no-function allele (*2A, *13). The patient's diplotype Activity Score is simply the sum of the two allele values.
Worked example for this patient: genotype = *1 (functional, 1.0) / *2A (no-function, 0) → AS = 1.0 + 0 = 1.0.
CPIC phenotype bins: • AS 0: Poor Metabolizer — complete or near-complete DPD deficiency; fluoropyrimidines generally contraindicated • AS 0.5–1.0: Intermediate Metabolizer — 50–75% dose reduction recommended, titrate by tolerance • AS 1.5: Intermediate Metabolizer (mild) — ~50% dose reduction, less pronounced deficiency • AS 2.0: Normal Metabolizer — standard full dose, no DPYD-based adjustment
When a patient carries two reduced- or no-function alleles in combination (e.g. HapB3/HapB3, AS = 1.0 via two 0.5 alleles), the clinical phenotype is treated identically to a single no-function heterozygote at the same AS value — the algorithm is allele-value additive, not variant-identity specific, which is what makes it generalizable to any diplotype combination without an exhaustive lookup table.
CPIC assigns "Strong" evidence to the *2A, *13, and c.2846A>T associations with severe toxicity, based on multiple concordant cohort studies and a 2015 meta-analysis (Meulendijks et al., Lancet Oncology, >7,000 patients) showing DPYD*2A carriers have a relative risk of Grade ≥3 toxicity of approximately 2.85–4.4 compared to non-carriers. HapB3 evidence is rated "Moderate-Strong" — individually lower-penetrance but the most population-prevalent reduced-function allele, meaning it contributes the largest absolute number of preventable severe-toxicity events. CPIC, DPWG, and the French GPCO/RNPGx guidelines are concordant in their AS-to-dose mapping, giving oncology pharmacists a single harmonized recommendation regardless of which guideline the treating center follows.
The Activity Score converts directly into a prescribing action: a percentage reduction of the standard starting dose, followed by cycle-by-cycle titration guided by observed tolerance and, where infrastructure permits, direct pharmacokinetic or phenotypic monitoring. Because genotyping alone misses a substantial fraction of at-risk patients, many centers now pair it with a plasma-based DPD phenotyping assay.
CPIC translates Activity Score directly into a starting-dose percentage of the standard fluoropyrimidine regimen:
• AS 0 (Poor Metabolizer): avoid 5-FU/capecitabine entirely; select an alternative non-fluoropyrimidine regimen if clinically feasible • AS 0.5: reduce starting dose by ≥50%, titrate cautiously; consider alternative agent if feasible • AS 1.0 (this patient): reduce starting dose by 50%, titrate to cycle 2+ based on toxicity, pharmacokinetics if available • AS 1.5: reduce starting dose by 25–50%, titrate by tolerance • AS 2.0 (Normal Metabolizer): standard full dose, no DPYD-based adjustment
Dose titration in subsequent cycles is not automatic — it requires a documented absence of Grade ≥2 toxicity at the reduced dose before escalating, typically in 25% increments, guided by CBC, mucositis/diarrhea grading, and hand-foot syndrome assessment at each cycle.
Because the 4-variant CPIC panel explains only an estimated 10–20% of clinically severe DPD-deficiency toxicity, an increasing number of centers add a direct phenotypic assay: measurement of the pre-treatment plasma uracil (U) to dihydrouracil (UH2) ratio, or plasma uracil (PreU) concentration alone, by LC-MS/MS.
• PreU <16 ng/mL: normal DPD activity • PreU 16 ng/mL and above: partial DPD deficiency — dose reduction indicated • PreU >150 ng/mL: profound/complete DPD deficiency — fluoropyrimidines contraindicated
A validation cohort (Meulendijks et al., JAMA Oncology 2020) established this cutoff's predictive value for Grade ≥3 toxicity independent of DPYD genotype, catching partial-deficiency patients who carry no detectable variant on the standard panel. Combining genotype-first triage with reflex phenotypic testing in genotype-negative but clinically suspicious patients is now considered best practice at high-volume oncology pharmacogenomic services.
The final test of any pharmacogenomic dosing algorithm is prospective clinical outcome data: does upfront genotyping and dose adjustment actually reduce severe and fatal toxicity, without unacceptably compromising anti-tumor efficacy through under-dosing? Multiple prospective cohorts now answer this question affirmatively, and DPYD testing has moved from experimental to guideline-mandated standard of care in much of Europe.
Toxicity is graded prospectively using CTCAE v5.0 criteria at day 8 and day 15 of each cycle: absolute neutrophil count, platelet count, diarrhea frequency/severity, oral mucositis, and hand-foot syndrome are the five domains most tightly linked to DPD deficiency. In this patient, cycle 1 at the 50%-reduced dose (1200 mg/m²) produced only Grade 1 mucositis (mild oral discomfort, normal diet) and ANC 3100/µL (within normal range) at the day-11 check — a marked contrast to the projected 60–70% chance of Grade ≥3 hematologic or gastrointestinal toxicity that population data predict for an AS-1.0 intermediate metabolizer receiving the unadjusted full dose. Based on this favorable tolerance, the oncology team escalates the cycle-2 dose to 75% of standard (1800 mg/m²) per the CPIC titration pathway.
The strongest prospective evidence comes from a multicenter Dutch cohort (Henricks et al., Lancet Oncology 2018, n=1103) that implemented upfront DPYD genotyping with CPIC/DPWG-guided dose reduction in routine practice. Compared to historical controls treated at standard dose without genotyping, DPYD variant carriers receiving genotype-guided reduced dosing experienced severe (Grade ≥3) treatment-related toxicity in ~28% of cycles, versus ~73% in historical variant-carrier controls treated at full dose — without a statistically significant reduction in treatment efficacy, since dose is subsequently titrated upward once tolerance is confirmed.
On the strength of this and concordant evidence, the European Medicines Agency (2020) and multiple national guideline bodies now mandate DPYD phenotype or genotype testing before any fluoropyrimidine prescription; several health systems report cost-effectiveness driven primarily by avoided hospitalizations for severe toxicity, which can each cost more than the genotyping assay for an entire treated cohort.
A single DPYD*2A/*2A homozygous (complete-deficiency) patient given a standard, untested 5-FU dose can develop fulminant, multi-organ, CTCAE Grade 5 toxicity — profound pancytopenia, hemorrhagic enterocolitis, and multi-organ failure — within 7–10 days, a preventable death recurring in case reports for decades before upfront genotyping became routine. The four-variant CPIC panel, a single blood draw processed in under a week, converts this into a fully anticipated, dose-adjusted, safely monitored treatment course.