Pesticide Mixtures May Silently Amplify Liver Risk for Farm Workers

The protective suits, gloves and closed-cabin tractors that define modern pesticide handling can only do so much. For the millions of agricultural operators who mix, load and spray plant protection products around the world, exposure to these complex chemical formulations remains largely a matter of skin contact, and the regulatory frameworks designed to keep them safe still rest on assumptions that a growing body of toxicology research is beginning to challenge. A new proof-of-concept study from researchers at the German Federal Institute for Risk Assessment (BfR) and partner institutions, published in Discover Toxicology, offers one of the clearest demonstrations yet that what happens inside a pesticide bottle is just as important as what happens once it reaches a worker’s skin.

The team, led by Yemurai Musengi under the supervision of Denise Bloch, set out to answer a deceptively simple question: when a formulated pesticide product is applied to skin, do the co-formulants and secondary active substances within it change the toxicokinetics of the primary active ingredient? Conventional risk assessment, built around the orally derived acceptable operator exposure level (AOEL), largely evaluates active substances in isolation. The benchmark dose lower confidence limit from animal studies is divided by an uncertainty factor of 100, compared with exposure estimates from the agricultural operator exposure model, and regulators call it a day. What this tier-one approach glosses over, the researchers argue, is that operators never encounter an active substance alone. They encounter a formulation — a cocktail of active ingredients, solvents, surfactants and stabilizers that can interact in ways no single-substance assessment captures.

The study focused on a real commercial formulation, Revus top, manufactured by Syngenta Agro, which combines the fungicides difenoconazole (DIF) and mandipropamid (MDP) alongside undisclosed co-formulants. The choice was deliberate. Previous work by the same group had already flagged this product as a case study in kinetic interaction, showing that DIF drives liver steatosis — the abnormal accumulation of fat in liver cells — while MDP potently inhibits CYP3A4 and CYP2C9, the very cytochrome P450 enzymes responsible for breaking DIF down. If two active ingredients in the same product can sabotage each other’s metabolism, the researchers reasoned, then assessing DIF on its own could dramatically underestimate the systemic exposure of the person spraying it.

Because dermal contact is the dominant route of exposure for agricultural operators, the team engineered a testing strategy that mirrors the journey a pesticide droplet would actually take through a human body. They started with EpiDerm Full Thickness 400 (EFT-400), a three-dimensional reconstructed human skin model that replicates both the epidermis and dermis. Concentrated product and two field-relevant dilutions — 1:250 and 1:2000, matching standard spray-tank dilutions — were applied topically for 24 hours, with the applied doses of 2500, 10 and 1.25 micrograms of active ingredient per square centimeter verified as non-cytotoxic using an MTT viability assay. The receptor medium that accumulated beneath the skin was then treated as the systemically available dose and transferred to two downstream assays: human liver microsomes (HLM) to measure metabolic clearance, and rapid equilibrium dialysis (RED) to determine the fraction of DIF unbound to plasma proteins.

The analytical backbone of the study was a rigorously validated liquid chromatography–tandem mass spectrometry (LC-MS/MS) method, executed on Agilent 1290 and 6475 triple quadrupole systems. Calibration curves in both microsomal and plasma matrices spanned concentrations from 0.848 to 813 nanograms per milliliter with correlation coefficients above 0.98, while the lower limit of quantification reached an impressively low 0.270 nanograms per milliliter. Accuracy ranged from 95 to 119 percent, precision stayed under 20 percent coefficient of variation, and recoveries fell comfortably within the 70 to 120 percent window mandated by the SANTE/11,312/2021v2 guidance for pesticide residue analysis. This methodological rigor matters because the entire proof of concept hinges on detecting vanishingly small amounts of a highly lipophilic compound as it migrates from a formulated product through living skin and into enzymatic and dialysis assays.

The dermal absorption results revealed a strikingly nonlinear, concentration-dependent pattern. The concentrated product allowed only 0.183 percent of the applied DIF to penetrate the skin over 24 hours — a negligible amount consistent with the high viscosity and potential surface precipitation of suspension-concentrate formulations. Absorption peaked at 10.2 percent for the 1:250 dilution, likely because the aqueous vehicle improved solubilization of DIF and hydrated the stratum corneum, then fell back to 6.07 percent at 1:2000 as co-formulant effects diminished with further dilution. These values align with trends reported in large OECD-compliant datasets on pesticide dermal absorption and sit in a comparable range to the concentration-specific values of 2 and 4 percent established for difenoconazole in the 2011 EFSA peer review, though both are far below the conservative EFSA defaults of 10 and 50 percent.

The metabolic clearance findings delivered the study’s central punchline. Pure DIF incubated in human liver microsomes was rapidly depleted, with less than 1 percent remaining after 45 minutes and an intrinsic clearance of 105 microliters per minute per milligram of microsomal protein. Within the concentrated formulated product, however, clearance plummeted to 45.7 — roughly half the rate — and the 1:250 dilution still showed depressed clearance at 66.2, evidence that even after dilution, MDP’s inhibitory grip on CYP enzymes had not fully relaxed. At the 1:2000 dilution, clearance dropped sharply to 15.1, a deviation from classical Michaelis-Menten expectations that the authors attribute largely to assay artifacts rather than true biology. Taken together, the pattern shows that the co-application of MDP — whether in concentrate or spray dilution — slows the hepatic breakdown of the liver-toxic DIF, prolonging and amplifying its systemic residence.

Plasma protein binding added another layer of complexity. In both its pure form and within the concentrated product, DIF was bound to plasma proteins at more than 99 percent, reflecting its inherent lipophilicity and the strong affinity of azole fungicides for serum albumin. But as the formulation was diluted, the fraction unbound rose significantly — to 0.29 percent at 1:250 and 4.83 percent at 1:2000. The researchers propose an intriguing mechanism: at high concentrations, surfactants in the product sit above their critical micelle concentration and sequester DIF inside micelles, hiding it from both plasma proteins and analytical detection. Below that threshold, sub-micellar surfactant concentrations instead interact with plasma proteins themselves, altering their conformation and displacing DIF from binding sites. Because only unbound chemical is available for metabolism and tissue distribution, this shift has direct consequences for how much biologically active DIF circulates in a worker’s bloodstream.

The authors are candid about the limitations that must temper interpretation. The 24-hour exposure window exceeded the 6 to 10 hours typical of occupational scenarios, potentially inflating absorption and availability, and DIF retained within the skin layers was conservatively excluded from the absorbed dose. The cell culture medium used as the carrier in downstream assays contains amino acids and salts that may themselves bind DIF, explaining why the measured fraction unbound came in five times lower than the 0.4255 percent reported by Wetmore and colleagues in 2012 using DMSO as solvent. Buffer chamber concentrations in the dialysis assay fell below the limit of quantification, forcing the team to substitute conservative values — the LOQ, half the LOQ, and 1 percent of the LOQ — to bracket the fraction unbound, though the relative ranking across treatments held firm regardless of which substitution was used.

What elevates this study beyond a technical curiosity is its positioning within the broader shift toward new approach methodologies in regulatory toxicology. The in vitro clearance and fraction-unbound values generated here are not endpoints in themselves; they are the quantitative raw material for physiologically based kinetic (PBK) models that use differential equations to describe absorption, distribution, metabolism and excretion across the human body. By feeding formulation-specific parameters into PBK models and applying in vitro-in vivo extrapolation (IVIVE), researchers can perform reverse dosimetry — translating in vitro points of departure into in vivo health-based guidance values without generating new animal data, in line with the 3Rs principle. The whole-mixture approach the team champions sidesteps the combinatorial explosion that plagues bottom-up mixture modeling, treating the entire formulation as a single kinetic entity while still acknowledging that not all interaction mechanisms are known. The team plans to integrate these parameters into multiple PBK models, layer in skin permeability inputs, run sensitivity and uncertainty analyses, and benchmark predictions against in vivo data submitted for pesticide authorization. If the proof of concept holds, the era of assessing pesticide safety one active ingredient at a time may finally be drawing to a close — and with it, a quieter, more honest accounting of the chemical reality that farm workers face every time they step into a sprayed field.

Subject of Research: Physiologically based kinetic modeling of plant protection product mixtures for agricultural operator risk assessment

Article Title: A proof-of-concept study towards physiologically based kinetic (PBK) modeling in agricultural operator risk assessment

Article References: Musengi, Y., Kumar, V., Deepika, D., D’Ambrosio Melendrez, A. P., Marx-Stoelting, P., & Bloch, D. (2026). A proof-of-concept study towards physiologically based kinetic (PBK) modeling in agricultural operator risk assessment. Discover Toxicology, 3(1), Article 8. https://doi.org/10.1007/s44339-026-00052-2

Image Credits: AI Generated

DOI: 10.1007/s44339-026-00052-2

Keywords: plant protection products, PBK modeling, difenoconazole, mandipropamid, dermal exposure, IVIVE, CYP inhibition, fraction unbound, agricultural operators, risk assessment, new approach methodologies, mixture toxicity