Reading the evidence: labels, numbers, and study design
Pest control generates confident claims. This page gives a non-scientist the working tools to test them: what the label legally says, what registration does and does not mean, and why one testimonial is not evidence.
How to read a pesticide label
Every registered pesticide label carries an EPA Registration Number (for example, in the format company number–product number), which identifies the exact registered product and formulation. Signal words indicate the product's acute toxicity category, and the highest category triggered by any route of exposure sets the word:
| Signal word | Toxicity category | What triggers it |
|---|---|---|
| CAUTION | Lower (Categories III–IV) | The least acutely hazardous of the labelled tiers; still a pesticide, still label law. |
| WARNING | Moderate (Category II) | Moderate acute effects by oral, dermal, inhalation, eye, or skin routes. |
| DANGER | Highest (Category I) | Category I effects — classically corrosive, irreversible eye damage or skin corrosion. DANGER on its own says: the worst endpoint is corrosive or otherwise Category I. |
| DANGER–POISON, with skull-and-crossbones | Highest (Category I, systemic) | A separate designation: Category I acute systemic toxicity by the oral, dermal, or inhalation route. A corrosive-DANGER product carries POISON and the skull only if it also meets these acute-systemic Category I criteria — the two DANGERs are not the same statement. |
The label also states where and how the product may legally be used, at what rate, and with what protective equipment. The governing principle in US law and on every label is: the label is the law — using a product inconsistently with its labelling is a violation, not a shortcut.
What EPA registration does and does not establish
Registration means the EPA has reviewed the product's data package and determined that, used as directed, it meets the statutory standard for registration, including that its labelling is not false or misleading and it will not generally cause unreasonable adverse effects when used per directions. Registration is not a government endorsement, not a ranking against competing products, and not proof that a product is the best choice for a given infestation. Efficacy data requirements also vary by product type and claim. “EPA-registered” on an advertisement answers a legal question, not a performance question.
LD50, explained — and its limits
LD50 is the dose, in a standardised test, at which 50% of test animals die, usually expressed as milligrams per kilogram of body weight. Lower LD50 means higher acute toxicity per unit weight. Its limits matter: it measures one acute endpoint in test animals, not chronic effects, not allergenicity, not exposure in real rooms, and not risk — risk depends on exposure as well as hazard. Comparing two products by LD50 alone is a category error that marketing sometimes invites.
Lab efficacy vs field efficacy
A laboratory assay forces contact: insects on a treated surface, no alternative food, no harborage choice, a susceptible lab strain. The field offers untreated refuges, competing food, resistant field strains, dust, heat, and decayed residues. Products and strategies therefore routinely perform worse in buildings than in assays. Good evidence reports the setting honestly; this desk always asks which setting a claim comes from.
Study-design basics
- Controls: a comparison group without the intervention (or with a standard one). Without a control, improvement may be seasonal, behavioural, or due to sanitation done at the same time.
- Replication: the result repeated across multiple independent units — apartments, colonies, sites. One treated kitchen is an anecdote with a sample size of one unit.
- Identification: the species confirmed. Studies and treatments built on a misidentified pest do not generalise.
- Follow-up: measured over a period long enough to cover hatch cycles and reinfestation, not just 24-hour knockdown.
- Testimonials: a testimonial has no control, no replication, no species confirmation, and a selection filter (happy customers are quoted). It is marketing material, never evidence.
How this desk writes summaries
We cite journal, year, and authors only where the citation is certain; otherwise we describe findings generally and say so. We do not publish invented DOIs or percentages we cannot verify. Charts that illustrate a concept are labelled schematic. When we find an error, we correct the page and log the correction on the About page — silently rewriting history is how misinformation desks operate, and we decline to.