The study, led by Julia C. Stingl of the Department of Clinical Pharmacology and Pharmacoepidemiology in Heidelberg, set out to measure two distinct burdens. The first, which the authors call the victim drug burden, is the degree to which systemic anticancer therapies depend on enzymes whose activity can be predicted from germline genetics and for which dosing recommendations based on metabolizer status already exist. The second, the pathway overlap burden, is how often the other medications a patient takes occupy, inhibit, or induce those same enzymes.

Both were assessed using structured, reproducible methods drawn from regulatory product information and a large real-world medication cohort. To establish the victim burden, the team extracted all small-molecule drugs recommended in the Onkopedia guidelines of the German Society for Hematology and Medical Oncology for gastrointestinal malignancies, excluding monoclonal antibodies and other protein-based therapeutics whose clearance does not rely on cytochrome P450 enzymes. Thirty-five drugs made the cut, ranging from fluoropyrimidines such as 5-fluorouracil and capecitabine to tyrosine kinase inhibitors including imatinib, regorafenib, selpercatinib, and tucatinib.

For each compound, the metabolic enzymes and transporters listed in the German Summary of Product Characteristics were recorded, along with a semiquantitative estimate of how much each pathway contributes to clearance. The resulting map was dominated by a single enzyme family. Across the 35 drugs, 16 distinct metabolic and transport pathways emerged, but CYP3A4 and its close relative CYP3A5 accounted for the overwhelming majority of pathway involvement, serving as major routes of elimination for agents such as apatinib, avapritinib, cabozantinib, encorafenib, erlotinib, everolimus, irinotecan, larotrectinib, lenvatinib, and regorafenib.

Other contributors included CYP1A2, CYP2C8, CYP2C9, CYP2C19, and CYP2D6, while the phase II enzyme UGT1A1 figured mainly for irinotecan and DPYD for the fluoropyrimidines. Crucially, however, only a minority of these drugs depend substantially on enzymes for which actionable pharmacogenetic guidelines exist. The Clinical Pharmacogenetics Implementation Consortium and the Dutch Pharmacogenetics Working Group provide genotype-based prescribing recommendations for CYP2D6, CYP2C19, CYP2C9, DPYD, and UGT1A1, but not for CYP3A4, whose known variants are rare and do not produce clearly separated metabolizer phenotypes.

That asymmetry is where the study's second analytical layer becomes important. The authors argue that a genetically normal metabolizer can be pushed toward an intermediate or poor metabolizer phenotype by concomitant drugs, a phenomenon known as phenoconversion. Enzyme inhibition by an interacting medication can shift effective enzyme activity by an amount comparable to the differences seen between pharmacogenetic genotype groups, and the direction and size of the shift depend on the patient's genetically determined baseline.

The team had previously developed a quantitative model that translates substrate-mediated pathway sharing into estimates of actual enzyme activity, and the new analysis extends that framework to oncology, where many drugs sit in a narrow therapeutic window and small changes in clearance can translate into meaningful differences in toxicity and efficacy. To estimate how often such overlap occurs in practice, the researchers turned to the ADRED cohort, a prospective multicenter study of more than 8,000 patients who presented to emergency departments with adverse drug reactions. The dataset covered 7,967 analyzable cases with a median of seven concomitant medications per patient, spanning 839 distinct active substances.

One study center, at Ulm University Hospital, specializes in gastrointestinal and oncological care and contributed an enriched subgroup of 567 cases, 47 percent of them involving cancer patients. Every recorded medication was annotated as a substrate, inhibitor, or inducer of relevant drug-metabolizing enzymes, drawing on the FDA list of drugs that interact with CYP enzymes and transporter systems and on a two-step, large language model-assisted extraction pipeline applied to the German product information. The extraction workflow was validated against a manual reference standard assembled by two blinded reviewers who agreed on 49 of 50 assessments, an inter-reviewer agreement of 98 percent.

The pipeline produced three errors among the 50 evaluated outputs, an error rate of 6.0 percent with a 95 percent confidence interval of 2.1 to 16.2 percent, and the discrepancies reflected incomplete or misclassified individual pathway annotations rather than any systematic bias toward particular CYP enzymes. The authors caution that this validation was a pragmatic quality-control exercise on a small sample, so the confidence interval is wide and the figure should be read as preliminary rather than a universal performance estimate. The overlap results were striking.

In the overall cohort, 84 percent of patient medication profiles contained at least one CYP3A substrate and 64 percent contained two or more. For the polymorphic enzymes that anchor classical pharmacogenetics, at least one substrate appeared in 57 percent of profiles for CYP2D6, 54 percent for CYP2C19, and 40 percent for CYP2C9. The oncology-enriched Ulm subgroup showed equal or higher burdens, with 93 percent of profiles carrying a CYP3A substrate and 69, 72, and 44 percent for CYP2D6, CYP2C19, and CYP2C9 respectively, differences that remained statistically detectable for CYP3A, CYP2D6, and CYP2C19 after adjustment for multiple testing.

Notably, most CYP3A, CYP2D6, and CYP2C19 substrates were major substrates metabolized predominantly by that single enzyme, whereas CYP2C9 and CYP2C8 involvement was usually minor and distributed across enzymes. By contrast, recognized pharmacological modifiers were far less common. FDA-labeled enzyme inducers appeared in roughly one percent or fewer of profiles regardless of enzyme, and moderate or strong inhibitors stayed below ten percent for most pathways: in the general cohort, the highest proportions were 12.9 percent for CYP1A2 inhibitors, 9.0 percent for CYP2C8, 8.1 percent for CYP2D6, and 6.8 percent for CYP3A, with descriptively higher figures in the oncology subgroup, reaching 17.6 percent for CYP2C8 and 15.9 percent for CYP3A.

Substrate sharing, in other words, dwarfs classical inhibition as a source of potential pathway interaction. Consistent with that, CYP3A was significantly more prevalent among the anticancer victim drugs than among co-medication substrates, at 68.6 percent versus 29.4 percent, an odds ratio of 5.23, while the other six CYP enzymes showed no statistically significant differences between the two drug sets. The authors are careful about what these numbers do and do not show.

The ADRED cohort consists exclusively of patients who experienced adverse drug reactions, so the estimates describe pathway overlap in a clinically enriched population rather than prevalence in unselected cancer care, and individual medication profiles could not be linked to specific anticancer regimens. No genotype data, enzyme activity measurements, pharmacokinetic parameters, or prospective outcomes were available, meaning the study cannot confirm that phenoconversion occurred in any individual patient or quantify its effect on drug exposure or toxicity. Recent separate analyses suggest a major substrate can shift CYP2C19 enzyme activity by about 30 percent, but the clinical consequences of such shifts in oncology remain to be tested.

The authors frame their results as descriptive, exploratory estimates of the metabolic pathway-overlap burden and argue that pharmacogenetic implementation strategies in gastrointestinal oncology should weigh both genotype and medication-related pathway overlap when estimating actual drug exposure. Future studies combining pharmacogenetic data, detailed treatment information, and longitudinal medication records will be needed to determine how often this overlap changes outcomes in defined patient populations.

Subject of Research: Metabolic pathway overlap between pharmacogenetically relevant enzymes and polypharmacy in gastrointestinal oncology

Article Title: Estimating metabolic pathway overlap in precision oncology: integrating pharmacogenetics and polypharmacy

Article References: Stingl, J. C., Krzoska, D. M., Neiss, J. A., Sommer, J., Ettrich, T., Seidling, H. M., & Seufferlein, T. (2026). Estimating metabolic pathway overlap in precision oncology: integrating pharmacogenetics and polypharmacy. Frontiers in Pharmacology, 17, Article 1947359. https://doi.org/10.3389/fphar.2026.1947359

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1947359

Keywords: pharmacogenetics, phenoconversion, polypharmacy, gastrointestinal oncology, CYP3A4, CYP2D6, CYP2C19, CYP2C9, drug-drug interactions, precision oncology, drug metabolism, ADRED cohort