Don Mycotoxin Wheat

Illustrative case study. Company names and identifying details are fictional. Technical details reflect real industry practice and current regulation.

Deoxynivalenol (DON) — vomitoxin — is the mycotoxin wheat millers fear most. Produced by Fusarium molds in the field, it survives milling, baking, and most processing. When a wet harvest pushed DON levels in incoming wheat to 3.2 ppm — against the FDA advisory level of 1 ppm for finished wheat products — the mill faced a choice: reject the grain and lose the supply, or find a way to manage it.

A mill in a bad harvest year

Red River Milling (fictional) milled hard red spring wheat in a 70-employee plant in North Dakota — 1.5 million bushels a year, sourced from 200 farms. The 2025 harvest was wet: Fusarium head blight swept the region, and DON levels in the grain supply ran two to three times normal. The mill’s mycotoxin testing program — every incoming load, rapid DON test, 1 ppm action limit — started rejecting trucks in July. By August, 30% of deliveries were failing.

The supply crisis

Rejecting 30% of the wheat supply wasn’t sustainable — the mill had flour contracts to fill, and the alternative supply (from unaffected regions) cost 40% more with freight. The operations team proposed blending: mixing high-DON wheat with clean wheat to bring the average below 1 ppm. It’s a common industry practice, legal under FDA’s advisory framework (which sets levels for finished products, not for blending itself) — but it’s also a practice that demands rigorous control, because the average can hide dangerous variation.

QA pushed back with data. DON is not uniformly distributed — it concentrates in damaged kernels, which concentrate in certain grain fractions. Blending 3-ppm wheat with 0.2-ppm wheat doesn’t reliably produce 1-ppm flour; it produces flour with hot spots — pockets of high DON from the damaged kernels that survive blending. The mill’s own testing confirmed it: blended flour showed a coefficient of variation of 45% in DON levels across samples, meaning some portions were well above the 1-ppm target even when the average looked fine.

Cleaning as the real solution

The investigation turned to grain cleaning — the physical removal of damaged kernels before milling. Fusarium-damaged kernels (FDK, “tombstones”) are lighter, shriveled, and visually distinct. The mill’s cleaning line — aspirator, gravity table, optical sorter — was designed for general cleaning, not mycotoxin reduction.

Trials showed the opportunity. Running high-DON wheat through an intensified cleaning sequence (double aspiration, gravity-table adjustment for density separation, optical sorter set to reject discolored kernels) reduced DON from 3.2 ppm to 1.1 ppm — a 65% reduction, achieved by physically removing the damaged kernels where the toxin concentrated. The rejected material (about 8% of the grain) went to animal feed, where the FDA guidance allows higher levels with species-specific limits.

But 1.1 ppm was still above the 1-ppm advisory for finished flour. The final step was strategic blending — but now blending cleaned high-DON wheat (1.1 ppm, low variation) with clean wheat, under a statistical blending protocol: every blend tested with a 10-sample composite, released only if all samples were below 0.8 ppm (building in a safety margin below the 1-ppm advisory).

What actually caused it

1. A harvest-year hazard. Fusarium pressure varies by year and region. The mill’s mycotoxin program was designed for normal years; the wet harvest was an abnormal hazard level the program wasn’t built to manage. Hazard analysis has to consider worst-case years, not average ones.

2. Blending without understanding variation. The initial blending proposal treated DON as uniformly distributed. It’s not — it’s concentrated in damaged kernels. Blending averages the numbers but not the risk, unless the variation is controlled first.

3. Cleaning equipment underutilized. The mill owned the tools to reduce DON (aspirator, gravity table, optical sorter) but ran them at standard settings. The equipment’s mycotoxin-reduction capability was an untapped control waiting for someone to ask the question.

What changed on the floor

Immediate: the intensified cleaning protocol became standard for any wheat testing above 1.5 ppm DON — a written procedure, not a trial. The statistical blending protocol (10-sample composite, 0.8-ppm release limit) was implemented for all blends. Rejected grain went to feed channels with proper documentation.

Within 30 days: the mycotoxin program was rewritten around harvest-year risk — pre-harvest Fusarium forecasting (using regional disease models) now triggers enhanced testing protocols before the first truck arrives. The rapid-test program expanded from DON to multi-mycotoxin screening (DON, zearalenone, ochratoxin A) on every tenth load.

Within 90 days: the mill invested in an upgraded optical sorter ($120,000) with mycotoxin-specific sorting algorithms — it detects the subtle discoloration of Fusarium-damaged kernels at 99% accuracy. The sorter paid for itself in one harvest season by recovering grain that would otherwise have been rejected.

What the numbers showed after

The wet-harvest season finished with zero flour lots above the 1-ppm advisory, zero customer complaints for mycotoxins, and a 92% acceptance rate on incoming wheat (up from 70% at the crisis peak) — because intensified cleaning recovered grain that simple rejection would have lost. The mill’s flour contracts were all filled, and the cleaning protocol became a competitive advantage the sales team now mentions in bids.

What you’d do Monday morning

Check whether your mycotoxin program has a bad-year plan — not just a testing procedure, but a defined response when 30% of supply fails. If the plan is “reject and hope,” you don’t have a plan.

Then look at your grain cleaning equipment and ask what it’s actually set to achieve. If it’s running standard settings in a high-mycotoxin year, you’re leaving your best control on the table.

Yeah, but actually — blending gets a bad reputation, and sometimes it deserves it. But the problem here wasn’t blending — it was uncontrolled blending. Done with cleaned grain, statistical verification, and a safety margin, blending is a legitimate tool. The lesson isn’t “never blend.” It’s “never blend blind.”