Eniluracil, 24 years later: a target validated for drug handling, retested for disease biology

What was tried

NCT06568692, registered by Processa Pharmaceuticals as PCS6422-BC-01, was an open-label, randomised Phase 2 study of PCS6422 with capecitabine in advanced or metastatic breast cancer. The registry lists eniluracil as the other name for PCS6422. Three arms: PCS6422 40 mg with capecitabine at 150 mg twice daily, PCS6422 40 mg with capecitabine at 225 mg or 75 mg twice daily, and an active comparator of capecitabine at the standard 1000 mg/m2 twice daily. Co-primary outcomes were RECIST 1.1 objective response rate and adverse events.

Eligibility covered unresectable triple-negative or hormone receptor positive, HER2-negative disease where other therapies were not indicated. It ran from 2 October 2024 to 29 July 2026 with 20 patients enrolled. ClinicalTrials.gov records it as terminated for "Business Reason", with no results posted and the record updated on 6 October 2026.

The biological hypothesis

Dihydropyrimidine dehydrogenase, encoded by DPYD, catalyses the rate-limiting step of pyrimidine catabolism and degrades 5-fluorouracil. Published pharmacology on inactivating it is unambiguous: with eniluracil, oral 5-FU bioavailability rises to approximately 100%, half-life extends to 4 to 6 hours, systemic clearance falls more than twentyfold to values comparable to glomerular filtration rate at 46 to 58 mL/min/m2, and renal excretion rather than enzymatic catabolism becomes the principal route of elimination at roughly 45% to 75%.

Target engagement is equally documented. In resected colorectal tissue, eniluracil drove DPD activity below the limit of detection at 0.0 pmol/min per mg protein, against placebo values of 57 in primary tumour, 119 in metastatic tumour and 243 in normal liver, with mononuclear cell activity returning to baseline within six days of the last dose.

The hypothesis for breast cancer was that this control over exposure would convert into a better therapeutic index for capecitabine. That is the step the evidence has never supported. Open Targets scores the DPYD to breast cancer association at 0.116, built from a literature component of 0.304 and a clinical precedence component of 0.176, with no genetic association evidence at all. The same platform carries 263 pharmacogenomic annotations for DPYD whose top entries concern fluorouracil toxicity rather than efficacy, and its recorded safety liabilities are diarrhoea, drug toxicity, hand-foot syndrome and hyperbilirubinaemia. DPYD is a drug-disposition gene that sits in an oncology pathway, not an oncology target.

What actually happened

The mechanism had already been tested at scale. In FUMA3008, a randomised open-label Phase 3 study in previously untreated metastatic colorectal cancer, 981 patients were randomised and 964 treated, 485 to eniluracil plus oral 5-FU and 479 to intravenous 5-FU with leucovorin. Overall survival was the primary endpoint and the result did not meet the protocol-specified criteria for equivalence, with a hazard ratio of 0.880, 95% confidence interval 0.75 to 1.03, and median survival of 13.3 against 14.5 months. Median progression-free survival was statistically inferior at 20.0 against 22.7 weeks, p = 0.01.

The toxicity data from that trial show the mechanism doing what it was designed to do. Grade 3 or 4 granulocytopenia occurred in 5% of eniluracil patients against 47% of controls, while grade 3 or 4 diarrhoea ran at 19% against 16%, p = 0.354. The enzyme block reshaped the toxicity profile decisively and left efficacy slightly worse than the comparator.

For the breast cancer attempt, the only public readout is a sponsor update dated 17 December 2025, reporting 19 patients randomised and a preliminary look at the first 16. It described exposure to the catabolite FBAL as up to ten times lower with the combination, a greater proportion of patients with metabolite-related side effects at similar severity, and hand-foot syndrome limited to grade 1 in the combination arm against up to grade 2 on capecitabine alone. No efficacy figures were given. The update expected an interim analysis of the first 20 patients in early 2026 and final enrolment by the end of that quarter. Enrolment closed at 20 and the study was terminated for business reasons with no posted results.

openFDA holds 62,647 FAERS reports naming capecitabine, 51,460 serious and 5,863 with a fatal outcome, led by diarrhoea at 8,754 and palmar-plantar erythrodysaesthesia at 4,396. That is the toxicity burden the program was trying to improve on.

Failure mechanism, best guess

The posted reason is a business decision, and nothing in the record contradicts it. A small sponsor reached the enrolment count it had designated for an interim look and stopped, which is why the archetype recorded here is sponsor_decision rather than an efficacy or safety call.

The deeper issue is a validated pharmacological hypothesis carrying an unvalidated therapeutic one. Every confirmed result sits on the disposition side: enzyme inactivation, exposure, half-life, catabolite suppression, redistributed toxicity. The one time the mechanism was asked for a survival benefit in a powered trial, it returned a hazard ratio whose interval spanned 1.0 and a progression-free survival result that was significantly worse. Reopening it in a different tumour type with a different prodrug does not retire that result, and a 20-patient response rate study was not sized to overturn it.

How to prevent this next time

No patient-level data from the terminated study exists and the sponsor update reported no efficacy numbers, so no quantitative efficacy modelling is defensible here. Three qualitative levers remain specific.

First, separate pharmacokinetic success from therapeutic success. Tenfold catabolite suppression is a pharmacodynamic confirmation, not evidence of benefit, and should not serve as a program-advancing milestone.

Second, apply a base-rate adjustment anchored on the prior Phase 3. A mechanism that failed an equivalence endpoint in 964 treated patients begins the next attempt with a low prior, and the design should state what new fact is supposed to change that prior and how the trial would detect it.

Third, use the target's evidence profile as a gate. A target with no genetic association to the indication and a safety profile built from pharmacogenomic toxicity annotations is a modifier of drug handling, and programs built on one belong in therapeutic index studies.

The single highest leverage change would have been to make the trial's primary endpoint the therapeutic index gain the mechanism actually produces, a prespecified reduction in grade 3 or higher fluoropyrimidine toxicity at matched systemic exposure, rather than an objective response rate that 20 patients could never resolve.

What this means for similar programs

DPYD enters the Claidex failure graph at a Mutational Risk Score of 34, in the yellow band. Almost all of it is genetic deficit, 13.26 of a possible 15, reflecting the 0.116 association with breast cancer, against phase burden of 8.05 and saturation of 6.77. The pattern inverts a crowded oncology target: few competing programs, little late-phase history, and almost no evidence linking the gene to the disease treated.

The read-across is to any program whose target is a metabolising enzyme, transporter or other pharmacokinetic modifier. Such targets reliably produce clean mechanistic data and are structurally prone to advancing on that data alone. A recent preprint on DPYD and MTHFR polymorphisms against 5-fluorouracil pharmacokinetics and toxicity in breast cancer patients frames the gene the way the clinical record always has, as a determinant of toxicity.

Open questions

Will the 20 enrolled patients' response and progression-free survival data ever be posted, and if not, what obligation covers results reporting for a terminated Phase 2 of this size?

Is there any tumour context in which DPD inactivation changes a fluoropyrimidine's efficacy ceiling rather than only its tolerability?

Sources

1. ClinicalTrials.gov. NCT06568692, PCS6422-BC-01, PCS6422 with capecitabine in advanced or metastatic breast cancer. https://clinicaltrials.gov/study/NCT06568692 Schilsky RL, et al. Randomized, open-label, phase III study of a 28-day oral regimen of eniluracil plus fluorouracil versus intravenous fluorouracil plus leucovorin as first-line therapy in patients with metastatic/advanced colorectal cancer (FUMA3008). J Clin Oncol 2002;20(6):1519-1526.Baker SD. Pharmacology of fluorinated pyrimidines: eniluracil. Invest New Drugs 2000;18(4):373-381.Dihydropyrimidine dehydrogenase (DPD) rapidly regenerates after inactivation by eniluracil (GW776C85) in primary and metastatic colorectal cancer. Cancer Chemother Pharmacol 2003;52(5):399-404.Amstutz U, et al. Clinical Pharmacogenetics Implementation Consortium (CPIC) Guideline for Dihydropyrimidine Dehydrogenase Genotype and Fluoropyrimidine Dosing: 2017 Update. Clin Pharmacol Ther 2018;103(2):210-216.Henricks LM, et al. DPYD genotype-guided dose individualisation of fluoropyrimidine therapy in patients with cancer: a prospective safety analysis. Lancet Oncol 2018;19(11):1459-1467.Open Targets Platform. DPYD (ENSG00000188641) and breast cancer (MONDO_0007254), association scores, pharmacogenomics and safety liabilities, accessed 2026-10-10. openFDA FAERS. Reports naming capecitabine, accessed 2026-10-10. Processa Pharmaceuticals. Clinical update on Phase 2 study in metastatic breast cancer, 17 December 2025. https://www.globenewswire.com/news-release/2025/12/17/3206931/0/en/Processa-Pharmaceuticals-Provides-Clinical-Update-on-Phase-2-Study-in-Metastatic-Breast-Cancer.html Preprint. Role of DPYD and MTHFR gene polymorphism on 5-fluorouracil pharmacokinetics and treatment-related toxicity in breast cancer patients.