For most health hazard classes under OSHA’s Hazard Communication Standard, mixture classification follows a defined sequence. You use test data on the complete mixture where available, apply bridging principles where the evidence supports them, and otherwise classify from ingredient data using the rules for that hazard class. At that ingredient-based step, concentrations, the toxicity data available and the ingredients included in the assessment can each move a product into or out of a hazard category.
The examples below use OSHA’s Hazard Communication Standard, the US implementation of GHS, and its Appendix A health hazard criteria. They show why classification depends not just on the calculation, but on the inputs behind it.
The order is test, bridge, then calculate, with exceptions
For most health hazard classes, Appendix A starts with data on the mixture itself. If the complete mixture has been tested for the relevant hazard, those results provide the starting point for classification. Where suitable whole-mixture data are unavailable, bridging principles may allow data from tested mixtures and their ingredients to support classification. Where neither approach applies, classification relies on ingredient data, using additivity formulas or concentration limits specific to the hazard class.
Carcinogenicity, germ cell mutagenicity and reproductive toxicity are exceptions. Under A.0.4.2, mixtures in those classes are generally classified from their ingredients, unless a case-by-case justification supports using data on the complete mixture.
Physical hazards follow separate criteria. Flammability, for example, depends on properties of the mixture such as flash point. This post focuses on health hazards and the ingredient data used to assess them.
Bridging depends on knowing how formulations differ
Appendix A describes six bridging principles, each with conditions and limits on where it applies.
Dilution covers a tested mixture diluted with an ingredient of equivalent or lower toxicity than its least toxic original ingredient, provided the diluent is not expected to affect the toxicity of the other ingredients. Batching allows the classification of one production batch to support another batch of the same mixture made by, or under the control of, the same manufacturer, unless there is reason to expect a significant change in toxicity.
For specified hazard classes, increasing the concentration of Category 1 ingredients in a tested Category 1 mixture can keep the mixture in Category 1 without additional testing. Interpolation can apply when the concentrations in an untested mixture sit between those of two tested mixtures with the same hazardous ingredients and hazard category. The substantially similar mixtures principle can support replacing one ingredient with another when the required concentration and toxicity conditions are met. The aerosol principle can carry certain classifications from a tested non-aerosol form to an aerosol form when the added propellant does not affect the relevant hazard.
These are not blanket permissions to copy a classification from one product to another. Each requires evidence that the formulations meet the relevant conditions. A team needs to retrieve the tested formula, compare the compositions and understand which differences matter before it can use bridging with confidence.
Acute toxicity uses an additivity formula
For acute toxicity, Appendix A combines ingredient acute toxicity estimates, or ATEs, into an estimate for the mixture:
100ATEmix=∑iCiATEi\frac{100}{ATE_{\text{mix}}}=\sum_i\frac{C_i}{ATE_i}ATEmix100=i∑ATEiCi
Here, CiC_iCi is each ingredient’s concentration in percent. For the oral toxicity examples below, ATE is expressed in mg/kg body weight. Relevant ingredients generally include those present at 1% or more, plus ingredients below 1% where there is reason to expect that they could affect the classification.
Under OSHA, an oral ATE above 300 and up to 2,000 mg/kg falls in Category 4. A value above 2,000 mg/kg does not trigger acute oral toxicity classification; OSHA has not adopted GHS Category 5.
Consider a product containing ingredient A at 20% and ingredient B, with an oral LD50 of 1,000 mg/kg, at 5%. The remaining 75% consists of water and ingredients that do not contribute to acute oral toxicity.
- With a measured LD50 of 400 mg/kg for A at 20%: 20/400+5/1,000=0.055020/400 + 5/1{,}000 = 0.055020/400+5/1,000=0.0550. The mixture ATE is approximately 1,818 mg/kg: Category 4.
- With the same LD50 and A reduced to 17%, replacing the difference with water: 17/400+5/1,000=0.047517/400 + 5/1{,}000 = 0.047517/400+5/1,000=0.0475. The mixture ATE is approximately 2,105 mg/kg: not classified for acute oral toxicity.
- With A at 20%, but only a Category 4 classification available rather than a measured LD50, Table A.1.2 supplies a converted point estimate of 500 mg/kg: 20/500+5/1,000=0.045020/500 + 5/1{,}000 = 0.045020/500+5/1,000=0.0450. The mixture ATE is approximately 2,222 mg/kg: not classified for acute oral toxicity.
The first two calculations show a three-percentage-point concentration change crossing the 2,000 mg/kg boundary. The third shows the consequence of having less detailed toxicity data. An appropriate measured LD50 and a category-derived point estimate are not interchangeable options: the criteria use the measured value where available and provide conversion values where needed. The classification record should show the source of each ATE and how it was derived.
Unknown toxicity changes the calculation too. If ingredients with unknown acute toxicity total more than 10% of the mixture, A.1.3.6.2.4 replaces 100 on the left side of the formula with 100 minus their total percentage. In the first scenario, replacing 12% of the non-acutely-toxic balance with ingredients of unknown acute toxicity changes the estimate to 88/0.05588/0.05588/0.055, or 1,600 mg/kg.
The disclosure requirement is separate from that greater-than-10% correction. Where an ingredient of unknown acute toxicity is present at 1% or more and classification is not based on testing the complete mixture, Appendix A requires a statement on the label and SDS identifying the percentage of the mixture consisting of ingredients of unknown acute toxicity.
Skin corrosion and irritation use concentration limits
For mixtures assessed using the additivity approach, Table A.2.3 sets concentration limits for skin corrosion and irritation. A mixture is Skin Category 1 when its Category 1 ingredients total 5% or more. It is Skin Category 2 when Category 1 ingredients total at least 1% but less than 5%, when Category 2 ingredients total 10% or more, or when ten times the Category 1 total plus the Category 2 total reaches 10%.
Assuming that approach applies and no other relevant evidence changes the classification, a mixture containing 3% of a Skin Category 1 ingredient and no other skin-corrosive or skin-irritant ingredients is classified as Skin Category 2. At 5%, it becomes Skin Category 1. That two-percentage-point change brings the corrosion pictogram, the signal word “Danger” and different precautionary statements.
Those generic additive thresholds do not apply to every formulation. Table A.2.4 sets separate limits where the additivity approach does not apply, including for acids with pH ≤2, bases with pH ≥11.5 and certain other corrosive or irritant ingredients. In those cases, a corrosive ingredient at 1% can trigger Skin Category 1. The calculation therefore starts with deciding which rules apply to the ingredients, not simply adding their concentrations.
Where spreadsheet classifications go wrong
The arithmetic is usually straightforward. The harder part is maintaining the composition, source data and assumptions behind it.
Premixes are one example. For acute toxicity, OSHA allows a classified premix’s actual or derived ATE to be used when calculating the final mixture. If the calculation instead uses the premix’s individual components, their concentrations must be carried through to the final product correctly. The record needs to show which approach was used so components are neither omitted nor counted twice.
Supplier concentration ranges introduce another uncertainty. An SDS listing an ingredient at 10–30% does not establish its exact concentration in the purchased material. The classification needs a defensible basis that accounts for the hazard the mixture could present across the applicable range, with the source information and assumptions recorded.
As the acute toxicity examples show, a measured LD50 and a converted point estimate can produce different results. The source of every ATE therefore needs to remain visible. Supplier data can also become outdated when a supplier revises its SDS but the classification spreadsheet retains the old information, a trigger explored in why an SDS can go out of date without a formula change.
Unit mismatches create further problems: weight percent and volume percent used interchangeably, for example, or a recipe recorded in grams per batch that was never converted to the concentration basis required for classification.
To test the process, rebuild one product’s acute toxicity calculation from its source data. Can you reproduce the category and identify the basis for every input? Check whether each ATE came from measured data or a converted category, how each premix was handled and whether the concentrations match the current formula.
Then follow a concentration change through the workflow. When a formulator changes a development batch, does the person responsible for classification receive the revised composition, or does someone have to remember to send it?
Keeping classification inputs connected to the recipe
Uncountable’s SDS authoring software puts SDS work inside the R&D platform, connecting safety documentation with formulation data rather than maintaining it in a separate workflow. Formula-linked SDS records give teams a shared starting point for drafting and reviewing safety information.
That connection does not replace the classification rules or the judgment needed to apply them. It helps address the problem underneath the examples: whether the person classifying a product can identify the formula, ingredient data and assumptions that support the result.
.png)
.png)
