The 'Something From Nothing' Effect: Chemical Combination Exposure

That's the assumption underneath every consumer chemical on the shelf. Apply it on its own, in the dose the regulator approved, and it's safe. The trouble is that nobody uses one product. Nobody encounters one chemical. Your morning is a mixture, and the regulatory machine that signed off on each ingredient never tested the mixture you just wore on your skin. Chemical combination effects are the gap between the test design and the reality.

The most-cited demonstration of that gap is a 2002 paper with a deceptively casual title. Silva, Rajapakse and Kortenkamp called it 'Something from "Nothing"' — eight weak estrogenic chemicals, each present below the dose at which it would do anything measurable on its own, combining to produce a substantial estrogenic response. None of them, individually. All of them, together. The paper is the editorial spine of our endocrine disruptors guide — and of this article Silva, Rajapakse & Kortenkamp 2002.

A combination effect is what happens when chemicals that share a biological target — the same hormone receptor, the same step in fetal development, the same metabolic enzyme — act together at doses each individually too small to produce a visible response. The maths of toxicology assumes those effects add up. If Chemical A at a fifth of its threshold dose produces no measurable response, and Chemical B at a fifth of its threshold also produces nothing, then five chemicals each at a fifth of their respective thresholds should still produce something — because each contributes a fraction of the same biological lever.

The model is called concentration additionA toxicology framework that assumes the combined effect of chemicals acting on the same biological target equals the sum of their individual fractional effects, with each chemical's contribution scaled by its potency. The textbook null hypothesis for similarly acting mixtures since the 1930s. and it has been the textbook null hypothesis for similarly acting mixtures since the 1930s. The wrinkle is that classical safety testing wasn't built around concentration addition. It was built around the no-observed-effect concentration: the dose at which a single chemical produces no measurable response by itself. Cross that threshold and the chemical 'produces an effect.' Stay below it and it doesn't. The threshold is a binary line drawn one chemical at a time.

Combination effects are what happens when seven chemicals each sit just below their individual binary lines and the lines turn out to add up. The math has been there since the 1930s. The biology and the regulation never caught up to it together.

By the late 1990s a lab at the University of London — run by Andreas Kortenkamp — was working on the question with progressively tighter experiments. By 2002 the group had two papers ready, one in ES&TEnvironmental Science & Technology and a companion in EHPEnvironmental Health Perspectives, that turned the textbook prediction into a demonstration in living cells.

The first paper used a yeast estrogen screen. The researchers picked eight chemicals from the families that show up across consumer products — hydroxylated polychlorinated biphenyls, benzophenones used as UV filters, parabens used as preservatives, BPABisphenol A, the soy isoflavone genistein. Each had a known weak affinity for the estrogen receptor, and each had a known no-effect concentration in the assay. The team measured the single-chemical response curve for every one of them. Then they combined all eight at concentrations at or below each individual NOECNo-Observed-Effect Concentration — the doses where, individually, they did nothing — and dosed the yeast cells with the cocktail. The result, in the paper's own language, was 'substantial mixture effects even though each chemical was present at levels well below its NOEC,' with excellent agreement between the concentration-addition prediction and the observed response Silva et al. 2002.

The companion EHP paper, published five months later, took the same approach but with the body's own estrogen in the room. Eleven xenoestrogens were combined, each at a sub-NOEC level, in the presence of 17β-estradiol. The mixture dramatically enhanced the steroid hormone's action. The verbatim sentence from that paper became the most-quoted line of the field: 'not even sub-NOEC levels of xenoestrogens can be considered to be without effect on potent steroidal estrogens when they act in concert.' That single sentence broke the regulatory framing. The threshold, when you have eleven of them, isn't a threshold Rajapakse, Silva & Kortenkamp 2002.

Two years later the Brunel and Brunel-collaborator labs ran the same experiment in vertebrates. Brian and colleagues used male fathead minnows, the workhorse fish for vitellogeninAn egg-yolk protein produced in the liver of female fish in response to estrogen. When measured in male fish, it is the standard biomarker of environmental estrogenic exposure. biomonitoring, and worked through five well-known environmental estrogens one at a time: estradiol, the synthetic ethinylestradiol, the surfactant breakdown products 4-tert-nonylphenol and 4-tert-octylphenol, and BPA. Single-chemical concentration–response curves were determined first. Then the five chemicals were combined at equipotent fixed-ratio concentrations, the fish were exposed, and vitellogenin was measured. The mixture response was accurately predicted in advance by concentration addition. The maths from the yeast cells held in the freshwater fish Brian et al. 2005.

The natural objection at this point is that the synergy could in principle run wild — that mixtures might produce many-times-larger effects than the simple addition of fractions. A 2011 industry-led review by Boobis and colleagues went looking for it across the published literature on mixture toxicology and didn't find it. Synergy beyond concentration addition was bounded: in the cleanest analyses the magnitude of mixture effects did not exceed additive predictions by more than a factor of four at low doses. Which sounds like a defence of the regulatory status quo, until you notice it's what the Silva and Brian papers had already shown Boobis et al. 2011.

It happens before you've left the bathroom. Concentration-addition mathematics doesn't require the chemicals to come from the same product, the same exposure route, or even the same time of day — only that they share a biological target. A daily routine that contains three sources of phthalates (a fragranced lotion, a perfume, a flexible-PVC item used in passing) and three sources of bisphenols (a thermal receipt, a polycarbonate water bottle, the lining of last night's tin) doesn't deliver a single named dose. It delivers an additive load on the same set of hormone receptors. The shorthand for this in the literature is the EDCEndocrine-Disrupting Chemical burden — the total combined hormonal pressure across all sources, which the regulatory model never sums Kortenkamp 2007.

The economic estimate of the load is the most-cited single number in the European endocrine-disruptor literature. In 2015, Leonardo Trasande and a Brunel-led team ran a Monte Carlo expert-elicitation simulation across the conditions plausibly attributable to EDC exposure in the European Union. The median estimate, across 1,000 simulations, was €157 billion annual EU disease cost attributable to EDC exposure — Trasande et al. 2015 JCEM, median Monte Carlo result across 1,000 simulations of attributable disease cost per year — about 1.23% of EU gross domestic product — to put a number alongside something physical, roughly the entire annual budget of NHS England of EU gross domestic product. The lower-bound sensitivity analysis was €109 billion. The estimate was endorsed by the Endocrine Society. That methodology assumed cumulative, not single-chemical, exposure. The single-chemical regulatory approach gives lower numbers because it leaves the combinations unweighted Trasande et al. 2015.

The companion paper by Bellanger broke the cost out by chemical class. The largest contribution wasn't bisphenols or phthalates. It was organophosphate pesticides — 13.0 million IQ points lost annually in the European Union attributable to prenatal organophosphate-pesticide exposure — Bellanger et al. 2015 JCEM IQ points lost annually, with an attributable cost of €146 billion. PBDEPolybrominated Diphenyl Ether — a class of brominated flame retardant flame retardants accounted for a separate 873,000 IQ points and €9.59 billion. Two chemical classes, two routes, the same brain Bellanger et al. 2015. The figure has been mis-quoted as '11 million IQ points from flame retardants' in secondary sources for a decade — it's actually thirteen million from organophosphates, with the flame-retardant figure a tenth of that. Different products, different routes, conflation matters.

For two decades the cocktail-effect literature lived mostly in cells and fish. The mechanism was solid; the human prospective data was harder. In February 2022 the SELMA pregnancy cohort published a paper in Science that bridged the gap — and validated the same mixture in three independent biological models.

The 2022 Science paper, led by Nicolò Caporale with a consortium spanning the Karolinska Institute, the SELMA cohortSwedish Environmental Longitudinal, Mother and child, Asthma and allergy cohort — a Swedish prospective birth cohort following ~2,000 mother–child pairs from early pregnancy through childhood, used to study environmental influences on child development. group at Karlstad University, and Barbara Demeneix's mixture-toxicology lab in Paris, started with biomonitoring data from Swedish pregnant women. The team identified an EDC mixture — the paper calls it Mixture N1 — that recurred in maternal urine and serum and that varied across the cohort. They then did something the previous mixture literature hadn't: they took the empirical mixture, scaled to the concentration ratios actually measured in the mothers, and tested it in three independent biological models. Human brain organoids grown in culture. Frog tadpoles (Xenopus laevis). Zebrafish larvae (Danio rerio). The same mixture, the same ratios, three biological substrates.

All three responded. Gene-expression patterns associated with neurodevelopmental risk converged across organoids, tadpoles, and zebrafish. Then the team went back to the children. Up to 54% of children in the SELMA cohort had prenatal EDC mixture exposures above experimentally derived levels of concern — Caporale et al. 2022 Science of the children in the cohort had prenatal exposures above the experimentally derived levels of concern. Comparing the upper decile to the lowest decile of exposure, the upper decile had a 3.3× times higher risk of language delay in offspring. The same mixture in a Swedish mother's urine, in cultured human neurons, in a tadpole, in a zebrafish, in a child's vocabulary at thirty months Caporale et al. 2022.

The cleanest place to see the gap is in US pesticide law. The FQPAFood Quality Protection Act of 1996, Public Law 104-170, signed 3 August 1996 of 1996 amended the federal pesticide-residue framework to require EPAEnvironmental Protection Agency to consider 'the cumulative effects on infants and children of such residues and other substances that have a common mechanism of toxicity' when setting tolerances. EPA built common mechanism groupsCohorts of pesticide active ingredients that share the same biological mode of toxicity — e.g. organophosphates inhibiting acetylcholinesterase, N-methyl carbamates reversibly inhibiting the same enzyme, pyrethroids targeting the voltage-gated sodium channel. for organophosphates, N-methyl carbamates, triazines, chloroacetanilides, and pyrethroids. The framework works. It has also never been extended beyond pesticides. Cosmetics, food contact materials, plastics, household chemicals, and industrial substances are still assessed one substance at a time.

The 2008 US National Research Council report on phthalates pushed for the framework to broaden. The committee recommended that EPA conduct cumulative risk assessment for all chemicals affecting male reproductive development through a shared anti-androgenic mode of action, citing the Howdeshell rat-mixture data showing a five-phthalate cocktail inhibited fetal testicular testosterone in a cumulative, dose-additive way National Research Council 2008. Howdeshell's mixture comprised butylbenzyl, di-n-butyl, di-2-ethylhexyl, diisobutyl, and dipentyl phthalate — five distinct products of distinct supply chains, one converging biological target Howdeshell et al. 2008. The NRC pivot didn't move the regulator off the pesticides-only ceiling.

The EFSAEuropean Food Safety Authority published the EU's harmonised methodology in 2019 — the agency calls it MIXTOX. The guidance defines a tiered framework for combined-exposure risk assessment using the hazard index, point-of-departure index, and margin of exposure for the total mixture. It applies to food and animal-feed contaminants where the European Food Safety Authority's remit reaches. The wider-chemical version is supposed to come through REACHRegistration, Evaluation, Authorisation and Restriction of Chemicals — Regulation (EC) No 1907/2006, the EU's principal chemicals regulation EFSA Scientific Committee 2019.

The EU's Chemicals Strategy for Sustainability, adopted as Commission Communication COM(2020) 667 final on 14 October 2020, explicitly committed to introducing a Mixture Assessment FactorA multiplier — typically of 10 or similar magnitude — applied to single-substance risk assessment under REACH to account for unmeasured combined exposure. Proposed in EU CSS 2020 §2.2.2; not yet adopted into binding law as of 2026. into REACH. Verbatim from the strategy: 'The Commission will assess how to best introduce in REACH (a) mixture assessment factor(s) for the chemical safety assessment of substances.' Six years on, that proposal is still not law. The Commission's REACH revision package was set back in September 2025 by a negative RSBRegulatory Scrutiny Board — the Commission's internal quality-control body for impact assessments opinion on the impact assessment, and as of April 2026 the formal legislative proposal had still not been tabled. Andreas Kortenkamp and Michael Faust, who started this whole programme back in the late 1990s, ran a Science policy piece in 2018 arguing that REACH needs a generic Mixture Assessment Factor on grounds the empirical literature had already been making for fifteen years Kortenkamp & Faust 2018. Sixteen years and counting since the 'Something from Nothing' paper. The EU is still drafting the multiplier.

The UK position post-Brexit is that UK REACH inherits the same single-substance framework. The Drinking Water Inspectorate reports near-perfect compliance against individual-parameter limits — 99.997% of the 18,483 samples taken in 2024 met every chemical-by-chemical standard — but doesn't publish a routine combined-exposure analysis for the mixture as a whole. The Pesticide Residues in Food committee does, in specific cases, assess combined residues for substances with similar effects (the most-cited recent example being triazole fungicides in olive oil and grapes), and finds those combined assessments acceptable on a case-by-case basis. The point isn't that nobody is looking. It's that one-substance-at-a-time is the default and combined assessment is the exception.

The practical move follows the maths. If the issue is additive load on shared receptors, the lever you have is the count, not just the dose. Cutting the number of distinct chemical-bearing products in your routine reduces total mixture load more reliably than swapping one ingredient for a 'free-from' version of the same product. The Eso-Friendly approach is to check the categories where the load comes from — fragrance, plastics, sunscreen filters, preservatives — and to thin the deck.

The instinct people sometimes have at this point — that the only meaningful intervention is regulatory and individual choice is futile — isn't right. The regulatory framework genuinely is slow, and the Mixture Assessment Factor genuinely is still not law. But concentration-addition mathematics also says the load is the sum of the components: remove one consistent contributor and the load drops by that contributor's fractional share. The arithmetic is on your side. The reason we'd take fragrance out of a routine first isn't that fragrance is the worst chemical on the shelf. It's that fragrance is the single category most likely to be three sources at once.

Twenty-three years after the 'Something from Nothing' paper, the maths the field has converged on is the maths the textbooks always had: concentration addition is the workable default, synergy beyond additivity is bounded, and 'safe doses' set one chemical at a time underestimate combined exposure by exactly the amount the regulatory model leaves out. The yeast cells in 2002 said it. The fathead minnows in 2005 said it. The Swedish mothers, brain organoids, frogs, and zebrafish in 2022 said it. The European Commission proposed the multiplier in 2020 and is still drafting the legislation.

The next time you stand at the bathroom counter and line up the seven products, the question worth asking isn't which one of them is dangerous. It's which one of them you could remove. The threshold isn't a threshold when there are seven of them. The count is the lever you have. The deeper methodology question — why low-dose effects can run the wrong way altogether, with curves that peak below the regulator's testing range Vandenberg 2014 — is the next article in this thread.