A drum of astragalus extract comes back from the laboratory with a total plate count of 100,000 CFU/g, and the buyer's email asks one question: can you sterilize it?
Three of the four routes change the material. The fourth changes only the paperwork, and it is the cheaper conversation.
The count belongs to the field
Bioburden in a botanical extract is not created in the dryer. It arrives with the root: soil, dust, insects, the surface it was dried on, the sack it was stored in. Drying lowers the water activity and stops most organisms from multiplying. It does not remove the spores, and it does not remove the dead cells that a plate count will happily grow over.
Milling makes it worse in a specific way. A whole root has most of its contamination on the outside; grinding distributes that surface through the entire lot, so every gram now carries a share of what was on the skin.
The specification therefore has two different kinds of line. One is a count, in CFU per gram. The other is an absence, and an absence only means something with a sample size attached: absence in 10 g and absence in 1 g are different claims, and only one of them is usually what a customer thinks they bought.
Four treatments, four different bills
- Heat. Dry heat oxidises; steam denatures. Both work, and both are brutal on the compounds the extract exists for. Anthocyanins, carotenoids, volatile oils and several vitamins are the first to go, and the colour often shifts before the assay does. Moist heat is far more effective at a lower temperature, and water is exactly what a hygroscopic powder should not be given.
- Ionising radiation. The European Union maintains a positive list of foods that may be irradiated, and it contains one category: dried aromatic herbs, spices and vegetable seasonings, at a maximum overall average dose of 10 kGy. That is the whole list. An extract is not on it, so an irradiated extract is not a food ingredient that can be placed on the EU market, whatever the dose. Several member states authorise other foods nationally, which turns the question into a country-by-country one rather than a Union one. In the United States the same treatment is permitted for spices and dry vegetable seasonings up to 30 kGy under 21 CFR 179.26, on condition that the finished package says so, with the words treated with radiation or treated by irradiation. The obligation does not stay with the ingredient; it travels to the retail box.
- Ethylene oxide. It is a gas, and a gas does not respect a closed drum: a sealed fibre drum offers no protection at all, which is the first thing people get wrong. It also reacts with chloride in the material to form 2-chloroethanol, and that is the molecule laboratories actually look for, because ethylene oxide itself is gone by the time the sample arrives. The EU does not authorise it as a treatment for food, and since the recalls of 2021 results in the low hundredths of a milligram per kilogram have been enough to take lots out of the chain.
- Filtration. A 0.2 micrometre filter sterilises a liquid and cannot sterilise a powder. This is the reason a liquid extract line has an option that a spray-dried powder does not, and it is worth knowing before the powder is made rather than after.
What each of them costs is not only money. Heat costs marker, and sometimes solubility. Radiation costs a market and a sentence on the retail pack. Ethylene oxide costs a residue marker that then follows the material into every customer's specification for the next three years. Filtration costs almost nothing in quality and exists only for liquids. A quotation that compares three treatments on price per kilogram has compared the wrong column.

What each one does to the certificate
Every one of these treatments happens after the assay was run unless the process was designed otherwise, and that is where most of the trouble sits.
A certificate issued on the powder before treatment describes a material that no longer exists. If the marker falls 8% during a heat step, the buyer has a document for a lot with 8% more of the marker than the drum contains, and the discrepancy will surface at the customer's own incoming test, months later, as an accusation.
The fix is procedural rather than analytical. State the sampling point relative to the treatment, re-assay after it, and release on the post-treatment result. A specification that names the sampling point can be checked. One that leaves it out cannot.
Irradiation leaves evidence of its own. In fat-containing material it produces 2-alkylcyclobutanones, and EN 1785 is a standard method for detecting them, which means a statement that a material was not irradiated is testable rather than a promise. Ethylene oxide leaves 2-chloroethanol, which is the marker for the same reason: the parent compound does not survive the trip to the laboratory.
The last metre
A treatment that achieves a five-log reduction in a validated study can be undone by the room the powder is packed in.
Repacking is the usual point of failure. Material comes out of the treatment chamber, goes into an open bin, is weighed on a scale that also weighs raw herb, and is filled into a liner that was stored on the floor. Nothing in that sequence shows up on a plate count taken at the treatment step, and everything in it shows up three weeks later at the customer.
The liner is part of the process. A polyethylene liner inside a fibre drum, folded and tied, is a barrier. A stitched woven bag is a sieve with a label on it. And water activity decides what the survivors do: a treated powder held at a high water activity will grow what the treatment left behind, and the count that was compliant at release will not be compliant at the customer's door.

Validating a reduction rather than a result
A sterilisation claim is a reduction claim, and a reduction is only valid on the matrix it was measured on, at that moisture, in that packaging, at that depth.
The depth matters more than it looks. A powder bed shields: the surface of a drum receives the full dose, the centre receives less, and a validation done on a 2 cm layer says nothing about a 40 cm drum unless the process is designed to mix or to treat in thin layers.
Sampling is the other half. Twenty-five grams taken from a 25 kg drum is a probability statement about a population that was never measured. A sampling plan that takes increments from the top, middle and bottom, and from more than one drum per lot, costs an hour and changes what the result means.
What to write in the specification
Six lines make a sterilisation step auditable.
- The count, with a method. Total plate count, yeast and mould, each with its incubation and its limit.
- The organisms, with a sample size. Absence of Salmonella and of specified pathogens in 10 g or 25 g, named, because absence in 1 g is a weaker claim.
- The treatment, permitted or prohibited. Named, because a customer in one market may have to refuse a material that a customer in another will accept.
- The sampling point. Before or after treatment, stated, because a certificate without it can be true and useless at the same time.
- The re-test. The result that releases the lot, taken after treatment, with the interval.
- The packaging after treatment. Liner, seal, and the room, because that is where the reduction is usually lost.

The botanical extract production lines page shows where a treatment step can be inserted without breaking the flow, the stability article covers what the same heat does to the marker over a shelf life, and the contact page reaches the people who run the validation.
Ding Shaohua, Foreign Trade Manager, takes enquiries for this line. The same person answers on WhatsApp, TikTok, LinkedIn and Facebook.
FAQ
Can an extract be irradiated for the European market?
Not as a food ingredient. The Union positive list of foods authorised for ionising radiation contains a single category, dried aromatic herbs, spices and vegetable seasonings, at a maximum overall average dose of 10 kGy, and an extract is outside it. A few member states authorise other foods nationally, so the answer can differ by country, but it is never a decision a supplier can make on the buyer's behalf. In the United States the same treatment is allowed for spices and dry vegetable seasonings up to 30 kGy, provided the finished package carries the statement.
Does a heat step really change the assay?
It usually does, and it is worth measuring rather than assuming. Moist heat at a temperature high enough to reduce a spore load will also open glycosidic bonds and oxidise pigments, and the compounds most often standardised in botanical extracts are the ones most exposed to it. The practical rule is that the certificate should be issued on the material after the last treatment, not before it.
Is ethylene oxide still used on botanical materials?
It is used, and it is the reason 2-chloroethanol appears in so many residue reports. The EU does not authorise it as a treatment for food, and a gas penetrates a sealed drum completely, so a closed package is not evidence that the material was untreated. The marker is what the laboratory measures, because the parent compound does not survive storage and transport.
If the plate count passes, is the lot sterile?
No. A plate count under a limit describes the sample that was tested, and says nothing about the drum it came from. It says that the material is within a specification at the sampling probability the plan allowed, which is a different sentence from sterile, and mixing the two is how a customer ends up with a specification nobody can meet at the incoming test.
