OECD TG 439 — What a Compliant RhE Model Must Demonstrate

The short answer: a compliant RhE model must be built from non-transformed human keratinocytes into a multi-layered epidermis with a functional stratum corneum, must demonstrate barrier function by a published ET-50 or IC50 against a benchmark chemical, must be released batch by batch against defined viability, barrier and morphology criteria, and must be read out by the MTT assay after a 42-hour post-treatment incubation. A chemical is classified as an irritant if mean tissue viability is at or below 50 percent of the negative control.

For a buyer, the practical consequence is narrower than the guideline is long: you cannot build your own RhE model and call the result TG 439 data. The guideline lists specific validated test methods. Using one of them is what buys you Mutual Acceptance of Data, and the model you buy comes with the batch release data you are required to report.

What the test actually is

The test chemical is applied topically to a three-dimensional reconstructed human epidermis, left for a short, method-specific exposure period, washed off, then incubated for a further 42 hours. Viability is measured by MTT. Irritants reduce viability; the cut-off is 50 percent.

The guideline is explicit that the RhE system, having no vasculature, is measuring the initiating events of the irritation cascade — cell and tissue damage — rather than the erythema and oedema that the in vivo endpoint describes. Cell viability is the readout that stands in for that damage.

TG 439 identifies chemicals requiring classification under UN GHS Category 2. It cannot distinguish Category 1 from Category 2 — a viability at or below 50 percent means “classify”, and corrosion data from elsewhere is needed to decide which category. It also cannot classify into the optional UN GHS Category 3 for mild irritants, which the guideline states as a limitation established by the original prospective validation study.

Requirement 1: general tissue conditions

The guideline’s construction requirements are specific and checkable.

Requirement	What the guideline specifies	Why it matters to a buyer
Cell source	Non-transformed human keratinocytes	Immortalised or transformed keratinocyte constructs do not qualify, regardless of how well they perform
Layer structure	Multiple layers of viable epithelial cells — basal layer, stratum spinosum, stratum granulosum — beneath a functional stratum corneum	This is the histology your supplier must be able to show for the model, not for your lot
Stratum corneum	Multi-layered, containing the essential lipid profile, with robustness to resist rapid penetration of cytotoxic benchmark chemicals such as SDS or Triton X-100	The lipid barrier is the whole point. A construct with keratinocyte layers and no functional barrier is not an RhE model
Containment	Containment properties must prevent passage of material around the stratum corneum to the viable tissue	Leakage around the insert edge models nothing and invalidates the result
Contamination	Free of bacteria, viruses, mycoplasma and fungi	Standard, and on the batch release documentation
Morphology evidence	Histological examination provided, demonstrating human epidermis-like structure including a multi-layered stratum corneum	Ask for it. It is a model-level document, and it exists

Requirement 2: barrier function, and how it is expressed

Barrier function is the property that distinguishes an RhE model from a keratinocyte monolayer, and the guideline requires it to be demonstrated quantitatively by one of two equivalent measures:

ET-50 — the exposure time required to reduce cell viability by 50 percent, when a benchmark chemical is applied at a fixed concentration. Models using 1% Triton X-100 are qualified this way.
IC50 — the concentration at which a benchmark chemical reduces viability by 50 percent after a fixed exposure time. Models using an 18-hour SDS treatment are qualified this way.

Each model developer sets an acceptability range, and only tissues within it may be used. These are published in the guideline.

Test method	Benchmark and exposure	Lower acceptance limit	Upper acceptance limit
EpiDerm SIT (EPI-200)	1% Triton X-100	ET-50 = 4.0 hr	ET-50 = 8.7 hr
SkinEthic RHE	1% Triton X-100	ET-50 = 4.0 hr	ET-50 = 10.0 hr
epiCS	1% Triton X-100	ET-50 = 2.0 hr	ET-50 = 7.0 hr
Skin+	1% Triton X-100	ET-50 = 4.0 hr	ET-50 = 9.0 hr
EpiSkin (SM) (no longer commercially available)	SDS, 18 hr	IC50 = 1.0 mg/mL	IC50 = 3.0 mg/mL
LabCyte EPI-MODEL24 SIT	SDS, 18 hr	IC50 = 1.4 mg/mL	IC50 = 4.0 mg/mL
KeraSkin SIT	SDS, 18 hr	IC50 = 1.5 mg/mL	IC50 = 4.8 mg/mL

How to read this table without over-reading it. These are release windows set by each developer, not measured performance. A wider window is not a worse model; it is a developer who has set a wider acceptable range. The one substantive comparison available here is that epiCS qualifies from an ET-50 floor of 2.0 hours against 4.0 hours for EpiDerm, SkinEthic and Skin+, which indicates a comparatively less resistant barrier at the permissive end of its range.

The buyer’s action item. The guideline requires the developer to provide these data to users so they can be included in the test report. That means you are entitled to the measured ET-50 or IC50 for the batch you are shipped, not merely a statement that it passed. Two lots sitting at opposite ends of a 4.0-to-8.7-hour window are not equivalent tissues for a borderline chemical.

Verify barrier function on receipt. The guideline recommends, as part of proficiency testing, that users verify the barrier properties of the tissues after receipt as specified by the model producer, and it is explicit that this is particularly important where tissues are shipped over long distances or long transit times. Once a laboratory has established the method and demonstrated proficiency, routine re-verification is not required. This is the most commonly skipped step in the workflow and it exists because transit degrades barrier function.

Requirement 3: the viability assay

The MTT assay is the required method. Not a substitute, not an equivalent — the guideline names it.

Parameter	Guideline requirement
Assay	MTT. Viable cells reduce MTT to a blue formazan precipitate, extracted from the tissue and quantified
MTT concentration and time	Tissue placed in MTT solution at an appropriate concentration, e.g. 0.3 – 1 mg/mL, for 3 hours
Extraction	Isopropanol or acidic isopropanol, or a similar solvent
Solvent blank	Optical density of the extraction solvent alone should be sufficiently small, i.e. OD < 0.1
Quantification	OD at 570 nm with a filter band pass of maximum ± 30 nm, or an HPLC/UPLC-spectrophotometry procedure
Post-treatment incubation	42 hours at 37°C. Standard across every method in the guideline
Replicates	At least three replicate tissues per test chemical and per control, in each run

The 42-hour incubation is not a convenience. The guideline states that the viability measurement must not be performed immediately after exposure, because the interval allows both recovery from weak cytotoxic effects and the appearance of clear ones. Reading early gives you a different assay.

Interference: the part that catches people out

MTT is a colorimetric readout, so anything coloured or anything that reduces MTT directly corrupts it. The guideline sets out a specific control structure:

Direct MTT reducers. Add the test chemical to fresh MTT solution. If it turns blue/purple, the chemical is presumed to reduce MTT directly, and a functional check on non-viable (killed) tissues is required. The correction control is NSMTT.
Coloured chemicals. Chemicals that are coloured, or become coloured during treatment, require a non-specific colour control on living tissues, NSCliving.
Both at once. A chemical that is both a direct reducer and colour-interfering, and that binds to tissue, risks a double correction. A third control on killed tissues, NSCkilled, is then required, and true viability is computed as the living-tissue value minus %NSMTT minus %NSCliving plus %NSCkilled.
Linearity. Each laboratory must determine the linear range of its spectrophotometer with commercial MTT formazan before testing for regulatory purposes. Results where %NSMTT or %NSCliving reach or exceed 50 percent of the negative control should be treated with caution, because 50 percent is the classification cut-off itself.
The escape route. For strongly coloured chemicals incompatible with plate-reader absorbance, HPLC/UPLC-spectrophotometry separates the formazan from the test chemical before quantification. When this is used, NSCliving and NSCkilled controls are never required regardless of the chemical — but the system must first be qualified against a set of parameters derived from FDA bioanalytical method validation guidance, listed in Annex 4 of the guideline.

If your chemical is coloured or reducing, budget for the extra tissues. The controls consume tissue, and tissue is the cost.

Requirement 4: controls and run acceptance criteria

Every run carries a concurrent negative control and a concurrent positive control. Their job is to demonstrate that the tissue’s viability, barrier function and sensitivity all sit inside the historical acceptance range.

Negative control: water or phosphate-buffered saline.
Positive control: 5% aqueous SDS.

The acceptance criteria are method-specific and published in the guideline.

Method	Negative control OD, acceptable range	Positive control viability (5% SDS), as % of negative control	SD between tissue replicates
EpiDerm SIT (EPI-200)	≥ 0.8 and ≤ 2.8	< 20%	≤ 18%
SkinEthic RHE	≥ 0.8 and ≤ 3.0	< 40%	≤ 18%
epiCS	≥ 0.8 and ≤ 2.8	< 20%	≤ 18%
Skin+	≥ 0.8 and ≤ 2.5	< 40%	≤ 18%
LabCyte EPI-MODEL24 SIT	≥ 0.7 and ≤ 2.5	< 40%	≤ 18%
KeraSkin SIT	≥ 0.7 and ≤ 1.6	≤ 40%	≤ 18%
EpiSkin (SM) (no longer commercially available)	≥ 0.6 and ≤ 1.5	≤ 20%	≤ 18%

Two consequences worth planning around.

EpiDerm and epiCS require the positive control to fall below 20 percent viability, against 40 percent for the others. That is a tighter run acceptance criterion. Tighter is better for data quality and worse for throughput, because a marginal positive control invalidates the run.

The negative control OD is a tissue-quality readout. The guideline states these ODs should reflect the quality of the tissues after shipment and receipt, and should not fall below historically established boundaries. A negative control drifting toward the lower limit is telling you something about your cold chain before it tells you anything about your chemical.

Users of the HPLC/UPLC route must use the same negative control OD ranges as their acceptance criterion.

Requirement 5: batch release, and what your supplier owes you

This is the requirement that most directly concerns a buyer, and it is stated plainly: the RhE model may only be used if the developer or supplier demonstrates that each batch meets defined production release criteria — most relevantly viability, barrier function and morphology.

The guideline then adds the sentence that matters: these data should be provided to the test method users, so that they are able to include this information in the test report.

So the batch data is not a favour. It is a documented input to your own report, and the test report contents specified by the guideline include, for the specific RhE model used: viability, barrier function, morphology, the model’s quality controls, and reference to historical data including acceptability of the QC data against historical batch data.

Only results produced with qualified tissues can be accepted. A run on a batch outside its release window is not a failed run; it is not a run at all.

What to ask for with every delivery

The batch number of the tissues shipped.
The measured ET-50 or IC50 for that batch, with the benchmark chemical and exposure stated, and the acceptance window it sits in.
The batch viability release data.
The morphology documentation for the model, showing the multi-layered stratum corneum.
Confirmation of freedom from bacterial, viral, mycoplasma and fungal contamination.
The model SOP for the version you are running.
The producer’s instructions for verifying barrier function on receipt.

Requirement 6: proficiency, before routine use

Before routine use of any method under the guideline, a laboratory must demonstrate technical proficiency using ten specified proficiency substances — five UN GHS No Category and five Category 2.

Substance	CAS	In vivo score (TG 404)	Physical state	UN GHS category
Naphthalene acetic acid	86-87-3	0	Solid	No Category
Isopropanol	67-63-0	0.3	Liquid	No Category
Methyl stearate	112-61-8	1	Solid	No Category
Heptyl butyrate	5870-93-9	1.7	Liquid	No Category (optional Cat. 3)
Hexyl salicylate	6259-76-3	2	Liquid	No Category (optional Cat. 3)
Cyclamen aldehyde	103-95-7	2.3	Liquid	Category 2
1-bromohexane	111-25-1	2.7	Liquid	Category 2
Potassium hydroxide, 5% aqueous	1310-58-3	3	Liquid	Category 2
1-methyl-3-phenyl-1-piperazine	5271-27-2	3.3	Solid	Category 2
Heptanal	111-71-7	3.4	Liquid	Category 2

The guideline permits substitution where a listed substance is unavailable or cannot be used for another justified reason, provided the replacement has adequate in vivo and in vitro reference data and the same selection criteria are applied — and the substitution must be justified.

Two of these are flagged in the guideline itself: 1-methyl-3-phenyl-1-piperazine and 1-bromohexane can give variable results between laboratories depending on the supplier of the substance. If your proficiency run wobbles on those two specifically, that is a known effect and not necessarily a problem with your technique.

Annex 3 Table 3 of the guideline publishes indicative cell viability ranges for all ten substances across all methods. These are explicitly not part of the guideline and are provided for information, but they are useful when setting the method up for the first time.

Requirement 7: the prediction model

The classification rule is short.

Mean tissue viability after exposure and 42-hour post-treatment incubation	Result
≤ 50% of the negative control	Requires classification and labelling under UN GHS — Category 2 or Category 1. Further information on corrosion is needed to decide which. If separately shown to be non-corrosive (e.g. by TG 430, 431 or 435), the chemical is irritant, UN GHS Category 2
> 50% of the negative control	May be considered non-irritant, UN GHS No Category, depending on the regulatory framework in the member country

Runs and repeats. A single run of three replicate tissues is sufficient when the classification is unequivocal. For borderline results — non-concordant replicates, or mean viability at 50 ± 5 percent — a second run should be considered, and a third if the first two disagree. Budget tissues for that possibility on any chemical you expect to sit near the line.

Which commercial models hold validation

TG 439 lists seven validated test methods. Six remain commercially available; EpiSkin was removed in the 2024/25 update because it is no longer commercially available from December 2024, though data generated with it remains covered by Mutual Acceptance of Data.

Every method in the guideline cleared the same minimum predictive capacity following independent peer review: 80 percent sensitivity, 70 percent specificity, 75 percent accuracy.

Model	Status in TG 439	Validation route	Also in TG 431 (corrosion)	Commercial status
EpiSkin (SM)	Validated Reference Method; used to define the Performance Standards	Full prospective validation 2003–2007	Yes	Not commercially available from December 2024. Existing data remains valid under MAD
EpiDerm SIT (EPI-200)	Validated Reference Method	Full prospective validation 2003–2007; the modified EPI-200 validated against the original ECVAM PS in 2008	Yes (EpiDerm SCT)	Available
SkinEthic RHE	Validated	Validation against the original ECVAM Performance Standards, 2008	Yes	Available
LabCyte EPI-MODEL24 SIT	Validated	Validation study 2011–2012 against the PS of TG 439	Yes (SCT version)	Available
epiCS	Validated	Performance-standards-based validation per GD 220, ESAC opinion 2016, independent peer review 2018	Yes	Available
Skin+	Validated	Performance-standards-based validation per GD 220, ECVAM opinion 2016, independent peer review 2018	Not in the TG 431 method set we reviewed	Available
KeraSkin SIT	Validated	Performance-standards-based validation per GD 220, independent peer review 2020	Not in the TG 431 method set we reviewed	Available

EPiTRI, developed at the Industrial Technology Research Institute in Taiwan, was validated against the TG 439 performance standards in a published four-laboratory study reporting 96 percent sensitivity, 70 percent specificity and 83 percent accuracy overall, and correctly identified di-n-propyl disulphide as an irritant where most validated reference methods called it a non-irritant. Confirm its current listing in the guideline version your regulator accepts, and confirm commercial supply for your region, before specifying it. The head-to-head detail is in EpiDerm vs SkinEthic RHE vs EPiTRI.

Performance standards: can you validate a new model?

Yes, and the route is defined. Performance Standards exist to facilitate validation and assessment of similar and modified RhE-based test methods, in accordance with the principles of GD 34, and the guideline confirms that the existing Performance Standards remain valid after the removal of EpiSkin.

The commercial caveat is the important one: Mutual Acceptance of Data is only guaranteed for test methods validated according to the Performance Standards if those methods have been reviewed and adopted by OECD. A model that meets the performance standards on paper but has not been through OECD adoption does not carry MAD. That gap is the difference between a scientifically validated model and a regulatorily usable one, and it is where a procurement decision can quietly go wrong.

Where the guideline does not apply

Worth knowing before you commit tissues.

Gases and aerosols. Not assessed in any validation study; the guideline does not permit testing them, though it acknowledges it is conceivable with RhE technology.
UN GHS Category 3 (mild irritants). Cannot be classified by this guideline.
Category 1 versus Category 2. Cannot be resolved. You need corrosion data.
Agrochemical formulations. The guideline cites a study of 65 agrochemical formulations giving overall accuracy of 54 percent, sensitivity 44 percent and specificity 60 percent, and states plainly that this indicates a lack of applicability of the RhE-based in vitro skin irritation test for agrochemical formulations. That is an unusually direct statement for a test guideline and worth heeding.
Mixtures generally. Applicable, but the guideline advises upfront consideration of whether the result will be scientifically meaningful, and says the TG should not be used for a specific category of mixtures where evidence of non-applicability can be demonstrated.

What it does cover: solids, liquids, semi-solids and waxes; aqueous and non-aqueous liquids; water-soluble and insoluble solids. Solids should be ground to a fine powder where possible, and the epidermis surface moistened before application to improve contact.

How TG 431 relates: corrosion on the same tissue

TG 431 is skin corrosion — irreversible damage, visible necrosis through the epidermis into the dermis — and it uses the same RhE test system with a different protocol. TG 439 states this directly: it does not itself provide adequate information on skin corrosion, and TG 431 is based on the same test system under another protocol.

The differences that matter operationally:

	TG 439 — irritation	TG 431 — corrosion
Endpoint	Reversible damage; UN GHS Category 2	Irreversible damage; UN GHS Category 1, with partial sub-categorisation
Exposure	One exposure, 15–60 min depending on method	Two or three timepoints: 3 min and 60 min (EpiDerm SCT, SkinEthic RHE, epiCS, LabCyte); EpiSkin additionally used 240 min
Post-treatment incubation	42 hours	None equivalent; viability read after the exposure sequence
Replicates	At least 3 per chemical per run	At least 2 per exposure time
Positive control	5% aqueous SDS	8N KOH (glacial acetic acid for the EpiSkin method)
Cut-off	≤ 50% viability = classify	Two-step model, see below
Sub-categorisation	Not applicable	Supports optional Sub-category 1A, and a combination of 1B-and-1C. Cannot separate 1B from 1C

The TG 431 prediction model for EpiDerm SCT, SkinEthic RHE, epiCS and LabCyte EPI-MODEL24 SCT runs in two steps:

Step 1 — corrosive or not. Viability below 50 percent after 3 minutes is corrosive. Viability at or above 50 percent after 3 minutes and below 15 percent after 60 minutes is corrosive. Viability at or above 50 percent after 3 minutes and at or above 15 percent after 60 minutes is non-corrosive.

Step 2 — sub-category, for chemicals called corrosive in step 1, using the 3-minute value, with a threshold that differs per model: below 25 percent for EpiDerm SCT, below 18 percent for SkinEthic RHE, and below 15 percent for both epiCS and LabCyte EPI-MODEL24 SCT indicates optional Sub-category 1A; at or above indicates a combination of 1B-and-1C.

The guideline is candid about the cost of that sub-categorisation: it reports that roughly 29, 31, 33 and 30 percent respectively of the Sub-category 1A calls by EpiDerm SCT, SkinEthic RHE, epiCS and LabCyte EPI-MODEL24 SCT may actually be 1B or 1C substances — that is, over-classifications.

TG 431 also publishes the model surface area per test method, which TG 439 does not: EpiSkin 0.38 cm², EpiDerm SCT 0.63 cm², SkinEthic RHE 0.5 cm², epiCS 0.6 cm², LabCyte EPI-MODEL24 SCT 0.3 cm². Tissue quality control uses the same benchmarks and essentially the same windows as TG 439.

How they are used together. Run TG 439 and get a viability at or below 50 percent, and you know the chemical classifies, but not whether it is Category 1 or 2. TG 431 (or TG 430 or TG 435) resolves that. Many laboratories run the corrosion test first when corrosivity is plausible, because a corrosive result makes the irritation test redundant for classification purposes. The integrated approach is set out in OECD Guidance Document 203 on IATA for skin corrosion and irritation, which TG 439 points to for a full evaluation of local skin effects after a single dermal exposure.

Procurement notes

Four things that are guideline-adjacent but decide whether a study runs on time.

The models are not interchangeable in practice. Total application time across the guideline ranges from 15 minutes for LabCyte to 60 minutes for EpiDerm, and application volumes, mesh use and acceptance criteria all differ. These exposure periods are optimised per method to compensate for each model’s intrinsic barrier properties. A laboratory with a validated EpiDerm SOP cannot drop SkinEthic tissues into it. Changing model means revalidating the protocol.

The shipping cadence sets your calendar. One major supplier publishes a fixed weekly production and shipping cycle — dispatched Mondays, delivered Tuesday morning by priority courier in the US, with a four-day shelf life that includes transit. That is a hard constraint you plan against rather than negotiate. See lead times and shipping live cells and tissue.

Media variants are scheduled production runs. Phenol-red-free, antibiotic-free, anti-fungal-free and hydrocortisone-free tissues are produced with the agent removed days before shipment. That is a production decision, not a picking decision, and it changes both price and lead time.

Nobody publishes a price. RhE suppliers route to contact, and the unit that gets quoted is the kit rather than the tissue. Ask for a quote against a specific model, format, kit size, media variant and delivery week, or you will not get two comparable answers. The method for comparing what comes back is in how to compare two custom biology quotes.

Checklist before you commit a study

Which guideline version your regulator or client accepts, and whether your chosen model is listed in it.
Whether the model is commercially available in your region on your timeline.
Whether your chemical falls inside the applicability domain — not a gas or aerosol, not an agrochemical formulation, and if a mixture, whether the result will be meaningful.
Whether your chemical is coloured or a direct MTT reducer, and therefore needs NSMTT, NSCliving and possibly NSCkilled controls — and the extra tissues those consume.
Whether you have completed proficiency testing on the ten substances for this method.
Your spectrophotometer’s linear range with commercial MTT formazan, determined before regulatory testing.
Whether you will receive the batch ET-50 or IC50 and the rest of the release data, in writing, with the shipment.
Whether you have booked barrier verification on receipt, particularly for a long transit.
Whether you have budgeted tissues for a second and third run on any chemical likely to sit near 50 percent viability.
Whether you also need TG 431 to resolve Category 1 versus Category 2.

Where we fit

We do not run this assay and we are not a distributor for any model named here. What we do is put one written specification — model, format, kit size, media variant, delivery week, and the batch release documentation you require — to the suppliers who can fill it, and return the responses on one basis. If you already know which model you need, that is a fast quote. If you do not, the model comparison is the place to start.

OECD TG 439: what a compliant RhE model must demonstrate