The Dose Makes the Poison

TL;DR. Paracelsus wrote in the sixteenth century that "all things are poison and nothing is without poison; only the dose makes a thing not a poison," and it remains the single most useful sentence in this subject. Water kills at a high enough dose; botulinum toxin is therapeutic at a low enough one. The distinction that resolves most public confusion is hazard versus risk: hazard is what something can do, risk is the chance it will do it at your actual exposure. Two exceptions genuinely break the dose rule and are worth knowing: allergy and, arguably, some effects of endocrine disruptors.

1. The principle

Every substance is toxic above some dose, and most are harmless below some dose.

Two crops of implications follow.

"Chemical-free" is meaningless. Everything material is made of chemicals. Water is a chemical. The word on a label conveys nothing.

"Natural" is not a safety category. The most acutely toxic substances known are natural products: botulinum toxin, ricin, tetrodotoxin, batrachotoxin, amatoxin. Conversely, many synthetic compounds are extremely safe. Origin tells you nothing about toxicity.

2. Hazard versus risk

This is the distinction that resolves most of the confusion in food and chemical scare stories, and it is worth internalising completely.

$$\text{Risk} = \text{Hazard} \times \text{Exposure}$$

• Hazard: the intrinsic capacity to cause harm. Can this cause cancer under some conditions?
• Risk: the probability of harm at the actual exposure. Will it, at the dose you receive?

A shark is a hazard. A shark in an aquarium behind glass is not a risk.

Why this matters in practice: the two main international bodies assessing carcinogens do different jobs, and their outputs are constantly conflated.

IARC Group 1 ("carcinogenic to humans") contains tobacco smoking, asbestos, plutonium, solar radiation, alcoholic beverages, processed meat, and outdoor air pollution. These are all in the same group and their risks differ by orders of magnitude. The group says how confident we are that it can cause cancer, not how much cancer it causes.

Two illustrations from this book:

• Processed meat is Group 1. 50 g a day raises lifetime colorectal cancer risk by roughly one case per hundred people. Smoking raises lung cancer risk by around 2,000 percent (Chapter 55).
• Aspartame was classified Group 2B in July 2023. On the same day, the joint FAO/WHO expert committee reaffirmed its acceptable daily intake unchanged. Both statements were correct because they answer different questions (Chapter 57).

3. How safety limits are actually set

This process is more structured than most people realise, and understanding it defuses a great many headlines.

Step 1: find the no-effect level. Animal studies, usually two species, lifetime exposure, plus reproductive and developmental studies, identify the NOAEL: the highest daily dose producing no observed adverse effect.

Step 2: divide by safety factors.

$$\mathrm{ADI} = \frac{\mathrm{NOAEL}}{100}$$

The standard factor of 100 is 10 for possible differences between species and another 10 for variation between humans. Additional factors are applied for particular concerns, such as effects on developing animals, which can take the total to 1,000 or more.

Step 3: check realistic exposure. Dietary surveys estimate how much people actually consume, including high consumers.

Step 4: set limits. MRLs for pesticides, maximum permitted levels for additives, tolerable weekly intakes for contaminants.

The practical consequences:

• A "safe limit" is typically 100 to 1,000 times below the level at which nothing happened in animals. There is very large headroom by design.
• Exceeding a limit does not mean harm. It means the product was made or used incorrectly and the margin has narrowed.
• An MRL for a pesticide is a good-practice standard, not a safety threshold (Chapter 6). Actual dietary exposure is typically well under 1 percent of the ADI.

4. Dose-response shapes

Not all responses are linear, and the shape determines whether a threshold exists.

The U-shape is the one most often forgotten. Every essential nutrient has one. Iodine deficiency causes goitre and intellectual disability; iodine excess causes thyroid dysfunction (Chapter 48). Selenium deficiency causes cardiomyopathy; excess causes selenosis. "More is better" is wrong for essentially every nutrient.

5. The exceptions that genuinely break the rule

Allergy. A true IgE-mediated allergy is not dose-dependent in the usual sense: trace exposure can trigger anaphylaxis in a sensitised person, and there is no safe dose for them (Chapter 98). This is why "may contain" labelling exists.

Genotoxic carcinogens. For substances that damage DNA directly, regulators assume no threshold, because in principle one mutation in one cell could initiate a cancer. This is a conservative assumption rather than a demonstrated fact, and it is why the guidance on things like aflatoxin and acrylamide is "as low as reasonably achievable" rather than a specific safe level.

Endocrine disruptors, arguably. Some researchers argue that hormone-mimicking compounds can show non-monotonic dose-response curves, with effects at low doses that do not appear at high ones, on the grounds that hormone systems themselves work at very low concentrations with feedback loops. This is genuinely contested; regulators have generally not accepted it as a basis for policy, and it remains an active scientific argument rather than a settled one.

Bioaccumulation and timing. Substances that accumulate (methylmercury, lead, PFAS, dioxins) mean a tiny daily dose can build to a harmful body burden over years, so "the dose" must be understood as cumulative. And timing can matter more than dose: thalidomide's effect depended on a specific window in early pregnancy, and lead exposure in early childhood does damage that the same exposure in an adult does not.

6. Toxicology's other vocabulary

LD50 comparisons are widely misused. Botulinum toxin has an extraordinarily low LD50 and is injected into faces weekly, because the therapeutic dose is a tiny fraction of it and it stays local. Acute toxicity and everyday risk are different questions.

7. Your body's own defences

The reason low doses of most things are harmless is that you have a substantial and continuously running detoxification system.

This system evolved because plants and microbes have been producing toxic compounds for hundreds of millions of years. You are equipped for a continuous low-level chemical assault, which is exactly what an ordinary diet is.

This is also why "detox" products are unnecessary. Your liver and kidneys detoxify continuously and cannot be accelerated by juice, foot pads, or supplements. No commercial detox product has ever identified which toxin it removes, demonstrated its removal, or shown a health outcome. When a detox product does something measurable, it is usually a laxative or a diuretic.

8. Applying it

A checklist for any chemical scare story:

1. What is the dose? In what quantity, how often, over how long?
2. How does that compare with the ADI or tolerable intake? Usually a fraction of a percent.
3. Hazard or risk? Is this "can cause harm" or "does cause harm at this exposure"?
4. Route? Injected into rodents at a hundred times human exposure is not the same as eaten.
5. What species, and at what dose? Rodent studies routinely use doses many multiples of any plausible human intake, deliberately, to detect effects in small groups.
6. Compared with what? Every choice has a counterfactual. Not using a preservative means more food poisoning; not using a pesticide means less food.
7. What is the absolute risk? Not the relative one.

9. The bottom line

• Every substance is toxic at some dose and almost every substance is harmless below some dose. Water, oxygen, salt, and every vitamin included.
• "Natural" describes origin, not safety. The most acutely toxic substances known are natural products.
• Hazard is what something can do; risk is what it will do at your exposure. IARC assesses hazard, which is why processed meat and plutonium appear in the same group.
• Safety limits are set 100 to 1,000 times below the level at which nothing happened in animals. Exceeding one indicates bad practice, not harm.
• The genuine exceptions to the dose rule are allergy, genotoxic carcinogens (where a no-threshold assumption is used), bioaccumulating substances, and critical developmental timing.
• Your liver, kidneys, glutathione, and DNA repair systems detoxify continuously. Detox products add nothing to a system that never stops.

Sources and notes

The Paracelsus formulation is from Die dritte Defension, 1538. Toxicological terminology (NOAEL, LOAEL, ADI, TDI, LD50) and the standard 100-fold safety factor follow WHO/IPCS Environmental Health Criteria and EFSA guidance. IARC's hazard-based classification system and its explicit statement that groups do not indicate risk magnitude follow the IARC Monographs preamble. The aspartame case, with IARC Group 2B and JECFA's simultaneous reaffirmation of the 40 mg/kg ADI in July 2023, is documented in the joint WHO release. Processed meat absolute risk translation follows Cancer Research UK's published analysis of IARC Monograph 114. Dose-response shapes, including U-shaped nutrient curves and hormesis, follow Calabrese's work and standard nutrition references. The linear no-threshold assumption for genotoxic carcinogens follows ICRP and EFSA policy positions. Non-monotonic dose responses for endocrine disruptors follow Vandenberg et al., Endocrine Reviews, 2012, and the regulatory responses that have not adopted it. Human detoxification systems follow standard biochemistry. The absence of evidence for commercial detox products follows Klein and Kiat's review, Journal of Human Nutrition and Dietetics, 2015.

Open questions. Whether non-monotonic dose responses are a real and general phenomenon for endocrine-active compounds, and if so how regulation should handle them, is genuinely unsettled and is the most substantive current challenge to the classical dose-response framework.

👉 Next: caffeine, the world's most-used drug.