DPYD Testing Before 5-FU and Capecitabine: How Genotype-Guided Dosing Prevents Severe Toxicity

A single gene test, read before the first cycle of 5-FU or capecitabine, can identify many of the patients most likely to suffer life-threatening chemotherapy toxicity, so their starting dose can be cut or the drug avoided. Prospective data show this lowers the rate of severe reactions. What the test cannot do is promise you safety, because a normal result still leaves a large share of serious toxicity unexplained.

Key points#

• Fluorouracil (5-FU) and its oral form capecitabine anchor treatment for colorectal, breast, gastric, and other common cancers.
• The DPD enzyme, encoded by the DPYD gene, clears most of the drug. When it works poorly, the drug builds up and can be fatal.
• CPIC turns a person's DPYD result into an activity score that maps directly to a starting dose.
• A 2018 Dutch trial is the reason pretreatment testing became standard rather than optional.
• A normal genotype is reassuring but not exonerating; clinical monitoring stays essential for everyone.

Why the labels changed#

The clearest sign that this test moved from research into routine care is on the drug labels themselves. The US Food and Drug Administration issued a safety labeling update for fluorouracil in 2024 and for capecitabine in 2025, adding boxed-warning language about serious or fatal reactions in people with dihydropyrimidine dehydrogenase (DPD) deficiency. The revised labels advise testing for DPYD variants before starting treatment unless therapy cannot wait, direct that patients with complete deficiency should not receive these drugs, and call for a reduced starting dose when deficiency is partial.

Two things about that language are worth reading carefully. The label positions testing as a step taken beforehand, and it stops short of claiming that a normal genotype rules DPD deficiency out. In other words, the regulator is describing a risk-reduction step, not certifying a test or a laboratory and not issuing a safety guarantee.

The enzyme standing between the drug and the patient#

Roughly four-fifths of an administered fluoropyrimidine dose is broken down by DPD, the enzyme the DPYD gene builds. Picture it as the body's disposal system for these drugs. When it runs at full capacity, active drug is cleared on schedule. When capacity is low or absent, the drug lingers and produces the toxicities that define these regimens at their worst: severe mouth sores, heavy diarrhea, dangerously low neutrophil counts, hand-foot syndrome, and nerve damage. In a person with complete DPD deficiency, a standard dose can be lethal.

Inherited variants in DPYD are the most common identifiable reason the disposal system underperforms. Someone carrying a single non-working copy of the gene faces a sharply higher chance of grade 3 or worse toxicity at a full dose, and the risk rises again for the rarer patients who carry two damaging copies.

From a genotype to a starting dose#

The Clinical Pharmacogenetics Implementation Consortium (CPIC) gives this a workable arithmetic. It assigns each of a person's two lowest-scoring DPYD copies a functional value and adds them into an activity score:

• Score 2.0, two normal copies. Normal metabolizer, standard dose.
• Score 1.0 to 1.5, intermediate metabolizer. CPIC advises starting at about half the usual dose, then titrating up as tolerance allows.
• Score 0 to 0.5, poor metabolizer. CPIC advises avoiding these drugs and choosing another agent.

The variants that drive the score fall into two groups. Non-function variants such as c.1905+1G>A (the classic DPYD2A) and c.1679T>G (DPYD13) yield a protein with essentially no activity. Decreased-function variants such as c.2846A>T and the intronic c.1129-5923C>G (part of the HapB3 haplotype) leave some activity intact. The CPIC guideline hosted through PharmGKB lays out this logic, and the commercial tests exist to put it into practice.

The evidence that made testing the default#

The dosing rules would be theory without outcome data behind them. The strongest support comes from a Dutch prospective study published in The Lancet Oncology in 2018. Investigators genotyped patients for four DPYD variants before treatment and gave carriers a reduced dose. Doing so brought the toxicity risk in carriers much closer to that of non-carriers, and the authors found the approach practical in routine oncology care. That trial, rather than any single laboratory's marketing, is why pretreatment testing shifted from a nice-to-have to an expectation.

One panel everyone can read the same way#

A test result only means something if you know which variants it looked for, and for years laboratories checked inconsistent sets. In 2024 a coalition of professional bodies, including the Association for Molecular Pathology, the American College of Medical Genetics and Genomics, CPIC, the College of American Pathologists, and European and Dutch pharmacogenetics groups, published a joint consensus in the Journal of Molecular Diagnostics naming a minimum panel that every DPYD test should cover. The four core alleles are c.1905+1G>A (DPYD2A), c.1679T>G (DPYD13), c.2846A>T, and c.1129-5923C>G (HapB3). A shared panel means a normal result from one laboratory carries the same meaning as a normal result from another.

What a normal result does not promise#

This is where the caveats deserve as much of your attention as the recommendations, and it is a useful case study in how to read any screening test. Think of the four-variant panel as a smoke detector that catches many fires but not all of them. Routine testing of the common actionable variants predicts only a minority of severe toxicity cases, and studies have found that a large fraction of patients who develop serious reactions carry none of the tested variants. Reduced DPD activity can also come from rarer DPYD variants outside the panel, from variants poorly characterized in some populations, and from causes that are not genetic at all and that no genotype can reveal.

Ancestry sharpens the point. The four core variants were characterized mostly in people of European ancestry and are uncommon or negligible in African and East Asian populations, which lowers the test's sensitivity for those patients rather than reassuring them. Some health systems add an enzyme-based check, such as measuring uracil in plasma, though that approach has its own handling pitfalls. The practical takeaway is that watching for early signs of toxicity stays essential no matter what the genotype says. A test that removes some of the highest-risk cases up front is valuable precisely because everyone understands what it leaves behind.