America manufactures or imports benzene at fifteen to twenty billion pounds a year. At the population median, it is not measurable in American blood. Lead is produced at roughly a sixth of that volume and is measurable in effectively everyone. We joined EPA's production data to CDC's biomonitoring for 43 chemicals, and the pattern is consistent: how much of something a country makes tells you very little about whether it ends up in people.

The join nobody had run

Two federal datasets sit a few clicks apart and are almost never put side by side.

EPA's Chemical Data Reporting programme collects how much of each chemical enters US commerce. Manufacturers report it every four years, and the 2024 cycle covers about 8,500 substances. Regulatory — EPA TSCA Chemical Data Reporting.

CDC's National Report on Human Exposure measures what is actually in Americans, through NHANES, with survey-weighted geometric means and percentiles by demographic group. Regulatory — CDC National Exposure Report.

One says how much exists. The other says how much is in you. Joining them requires a curated compound list to match against, which is the piece that does not exist in either agency. We matched them on CAS number through our Atlas: 43 chemicals carry both figures, and 22 of the 43 are measurable at the population median.

43 chemicals, ranked by volume

Production volumes are EPA's own bands, in pounds. EPA does not publish point estimates, so neither do we. Rows in grey are not measurable at the population median.

The "share detectable" column needs a word of explanation, because unlike the Canadian survey, CDC does not publish a detection frequency for these chemicals. What it publishes is a set of percentiles. Where the 50th percentile clears the detection limit, more than half the population is detectable; where only the 75th clears it, between a quarter and a half is. So the column is a bracket read off CDC's own percentiles, not a figure CDC states. Inferred We show it as a range because a range is what the published data supports.

US production volume against US body burden, 43 chemicals. Sources: EPA TSCA CDR (2024 cycle); CDC National Exposure Report.
ChemicalUS production, 2023 (lb)Share detectablePopulation geometric meanMeasured in
Benzene15,000,000,000 – <20,000,000,00025–50%below detection limitwhole blood
Ethylbenzene10,000,000,000 – <15,000,000,00025–50%below detection limitwhole blood
Toluene10,000,000,000 – <15,000,000,000>50%0.092 ng/mLwhole blood
Styrene8,500,000,000 – <10,000,000,00025–50%below detection limitwhole blood
Ethylene oxide7,000,000,000 – <8,500,000,000>50%20.8 pmol/g hemoglobinhemoglobin adduct
Copper4,000,000,000 – <5,500,000,000>50%115.0 µg/dLserum
Formaldehyde2,500,000,000 – <4,000,000,000>50%135.0 nmol/g hemoglobinhemoglobin adduct
Lead2,500,000,000 – <4,000,000,000>50%0.753 µg/dLblood
Bisphenol A1,000,000,000 – <2,500,000,000>50%1.1 µg/Lurine
n-Hexane1,000,000,000 – <2,500,000,000<5%below detection limitwhole blood
Acrylamide400,000,000 – <550,000,000>50%47.7 pmol/g hemoglobinhemoglobin adduct
Manganese400,000,000 – <550,000,000>50%9.52 µg/Lblood
Nickel250,000,000 – <400,000,000>50%1.11 µg/L
Chromium250,000,000 – <400,000,00010–25%below detection limitblood
Carbon tetrachloride100,000,000 – <250,000,000<5%below detection limitwhole blood
Chloroform100,000,000 – <250,000,000>50%below detection limit
Methylene chloride (dichloromethane)100,000,000 – <250,000,000<5%below detection limitwhole blood
Perchloroethylene (PERC)100,000,000 – <250,000,0005–10%below detection limitwhole blood
2,4-Dichlorophenoxyacetic acid75,000,000 – <150,000,000>50%0.349 µg/Lurine
o-Toluidine55,000,000 – <70,000,000>50%293.0 pg/mLurine
Alpha-pinene55,000,000 – <70,000,000>50%0.077 ng/mL
Trichloroethylene (TCE)50,000,000 – <100,000,000<5%below detection limitwhole blood
Cobalt25,000,000 – <40,000,000>50%0.173 µg/Lblood
1,4-Dichlorobenzene10,000,000 – <50,000,00025–50%below detection limitwhole blood
4-tert-Octylphenol10,000,000 – <25,000,00010–25%below detection limiturine
Decanal10,000,000 – <25,000,000<5%below detection limitserum
Nonanal (Aldehyde C-9)10,000,000 – <50,000,00025–50%below detection limitserum
Octanal10,000,000 – <25,000,00010–25%below detection limitserum
p-Phenylenediamine10,000,000 – <50,000,00010–25%below detection limit
Cadmium2,500,000 – <4,000,000>50%0.241 µg/Lblood
4,4′-Methylenedianiline (MDA)1,000,000 – <5,000,000>50%below detection limit
Arsenic850,000 – <1,000,000>50%6.32 µg As/Lurine
GenX (HFPO-DA)250,000 – <550,000<5%below detection limitserum
Barium250,000 – <400,000>50%1.07 µg/Lurine
Ethylparaben105,00025–50%below detection limiturine
Hexanal100,000 – <250,000>50%2.08 ng/mLserum
Methylparaben100,000 – <250,000>50%31.8 µg/Lurine
Selenium100,000 – <250,000>50%188.0 µg/Lblood
1,4-Dioxane75,000 – <100,000<5%below detection limitwhole blood
Mercury63,224>50%0.643 µg/Lblood
Hexachlorobenzene<50,000<5%below detection limitserum
Oxybenzone (Benzophenone-3)<25,000>50%19.3 µg/Lurine
Propylparaben<75,000>50%4.2 µg/Lurine

Why volume fails to predict

Look at the top of that table. The four largest-volume chemicals in the set are benzene, ethylbenzene, toluene and styrene. Together they account for more than 43 billion pounds a year, and three of the four are not measurable at the population median in blood.

They have something in common. They are volatile. A person breathing them in clears them within hours, through exhalation and metabolism, so a blood sample drawn at a random moment usually finds nothing at the median. Volume of production says nothing about how long a molecule stays.

Now look at what is measurable. Lead, manganese and copper are metals, and metals accumulate. Formaldehyde, ethylene oxide and acrylamide are measured a different way again: as adducts bound to haemoglobin, which persist for roughly the lifespan of a red blood cell rather than for hours.

The variable that predicts body burden is not how much is made. It is how long the molecule, or the mark it leaves, survives inside a person. Inferred — this is our reading of the pattern across the 43 compounds, not a finding either agency publishes.

What "not measurable" does not mean

This is the sentence most likely to be misread, so we will be blunt about it.

Not measurable in blood does not mean not exposed, and it certainly does not mean safe. Benzene is a recognised human carcinogen. Its absence from the median blood sample is a statement about how fast the body clears it, not about whether it causes harm. A chemical that passes through quickly can still do damage on the way through, and repeated short exposures do not become harmless because each one is brief.

If anything, the volatiles are the harder case. A persistent chemical leaves a record that a laboratory can find years later. A volatile one leaves almost nothing, which makes exposure harder to study, not less real.

Why we cannot rank the body burdens

A reader will reasonably want the next table: the chemicals ranked by how much of each is in people. We cannot honestly produce it.

The 43 measurements arrive in eight different units, across four different matrices. Some are micrograms per litre of urine, some nanograms per millilitre of whole blood, some picomoles per gram of haemoglobin, one is nanograms per gram of blood lipid. A number in one of those units cannot be placed above or below a number in another. Doing so would produce a ranking that looks authoritative and means nothing.

Production volume can be ranked, because it is all in pounds. Body burden cannot. That asymmetry is why the table above shows presence rather than a league table.

Method

Production figures come from EPA's 2024 Chemical Data Reporting cycle, reported as bands and copied verbatim. 504 substances in that dataset report under confidential accession numbers rather than CAS numbers and cannot be joined by anyone; they are excluded here rather than guessed at.

Body-burden figures are CDC's published survey-weighted statistics. We do not compute them from NHANES microdata: doing that correctly requires the survey's sampling weights, strata and primary sampling units, and doing it incorrectly produces a plausible-looking wrong number.

Eight of the underlying CDC tables cover subpopulations rather than the whole country, six of them non-smokers, and those are labelled as such in the source data. Where CDC reports a value below the laboratory's limit of detection, we record no geometric mean rather than a zero. Benzene is exactly that case: it has real 90th and 95th percentile values while the median sits below detection.

The detection brackets are derived, not quoted. CDC publishes percentiles rather than detection frequencies for this series, so we infer the bracket from which percentiles clear the laboratory's limit. That is weaker than a stated figure and is labelled as inferred throughout. It also carries the same caveat as any detection statistic: it describes what the instrument can see, not a dose and not a harm.

The joined dataset is published in the Embr Exposure Ledger, free to reuse with attribution.

Frequently asked questions

Does high production mean high exposure?

Not reliably. Across 43 chemicals with both US production and US body-burden data, the highest-volume substances include several that are not measurable at the population median in blood. Persistence in the body predicts measurability far better than manufacturing volume does.

Why is benzene not measurable if so much is produced?

Benzene is volatile. The body clears it within hours through exhalation and metabolism, so a blood sample taken at a random moment usually finds nothing at the median. That is a statement about clearance speed, not about safety: benzene is a recognised human carcinogen.

Which chemicals do show up in people?

In this set, metals such as lead, manganese and copper, which accumulate; and chemicals measured as haemoglobin adducts such as formaldehyde, ethylene oxide and acrylamide, where the mark persists for roughly the lifespan of a red blood cell.

Can you rank chemicals by how much is in people?

Not honestly, at least not across this set. The measurements use eight different units across four biological matrices. Micrograms per litre of urine cannot be compared with picomoles per gram of haemoglobin. Production volume can be ranked because it is all in pounds; body burden cannot.

Where do the numbers come from?

Production volumes from EPA's TSCA Chemical Data Reporting 2024 cycle, reported as bands rather than point values. Body burdens from CDC's National Report on Human Exposure to Environmental Chemicals, which publishes survey-weighted statistics from NHANES.

Citations
  1. EPA Toxic Substances Control Act Chemical Data Reporting, 2024 reporting cycle. Source Regulatory
  2. CDC National Report on Human Exposure to Environmental Chemicals, biomonitoring data tables. Source Regulatory