Environmental Health

Two hundred and twelve chemicals

In 2001 the Centers for Disease Control and Prevention published the first National Report on Human Exposure to Environmental Chemicals. Rather than modeling what Americans were probably exposed to, it measured what was actually in them, in blood and urine samples from a representative national sample. The first report covered 27 chemicals. By the fourth report, in 2009, the list had grown to 212.

The results are worth sitting with. Bisphenol A was detectable in the urine of the overwhelming majority of people tested. So were several phthalate metabolites. So were per- and polyfluoroalkyl compounds, in nearly everyone. So was lead, in everyone, though at a small fraction of the level Americans carried in 1976.

None of those findings, by itself, means anyone was harmed. That is the first hard lesson of this course, and most public argument about chemicals founders on it. Detection is not dose. Dose is not effect. A modern analytical laboratory can find a few molecules of almost anything in almost anyone, and the interesting question is never whether a chemical is present. It is how much, for how long, at what age, and compared to what.

Key idea: Environmental health is the study of a chain, and finding a chemical in a person tells you that the chain reached its middle, not that it reached its end.

The chain itself

Every problem in this course sits somewhere on one continuum. Learn it now and the rest of the course becomes an exercise in locating things on it.

Link	Question it answers	Example: a coal-fired power plant
Source	Where does the agent originate?	Mercury present as a trace element in coal
Release	How does it enter the environment?	Vaporized mercury emitted from the stack
Transport and fate	Where does it go and what does it become?	Atmospheric travel, deposition into a lake, bacterial conversion to methylmercury
Exposure	Who contacts what concentration, for how long?	A family eating locally caught fish twice a week
Dose	How much enters the body?	Micrograms absorbed per kilogram of body weight per day
Internal and effective dose	How much reaches the target tissue?	Methylmercury crossing the placenta into fetal brain
Early effect	What changes first, before disease?	Altered neuronal migration and signaling
Outcome	What is the health consequence?	Deficits in attention, memory, and motor function in childhood

Interventions are possible at every link, and their cost and effectiveness differ enormously. You can scrub the stack, remediate the lake, advise people not to eat the fish, or treat the child. The first is expensive and protects everyone downwind forever. The last is cheap per case, arrives after the harm, and does not work well. Environmental health as a discipline is a standing argument for acting as far to the left of that table as you can afford.

Four words people mix up

Hazard is the intrinsic capacity of something to cause harm. A shark is a hazard. Botulinum toxin is a hazard.

Risk is the probability that harm occurs, which requires both hazard and exposure. A shark in an aquarium poses a hazard and almost no risk. This distinction resolves most of the confusion in chemical news coverage: a substance can be correctly described as a carcinogen, which is a hazard statement, while posing negligible risk at the doses anyone encounters.

Exposure is contact between a person and an agent at some concentration over some duration. It is measured outside the body: micrograms per cubic meter of air, milligrams per liter of water.

Dose is the amount that actually enters. Two people in the same room breathing the same concentration receive different doses because one is a child breathing faster relative to body size, or because one is exercising.

The distinction between exposure and dose is not pedantry. It explains why children are usually the most affected group in this field: relative to body weight, a child breathes more air, drinks more water, and eats more food than an adult, so an identical concentration delivers a larger dose. Add hand-to-mouth behavior, proximity to floors and soil, and organ systems still developing, and the same environment is a different environment for a two-year-old.

The point: Hazard describes a substance, risk describes a situation, exposure is measured outside the body and dose inside it, and a sentence that confuses any of these is usually wrong.

What happens between release and contact

Between a smokestack and a lung lies an enormous amount of physics and chemistry, collectively called environmental fate and transport. Four properties do most of the work.

Persistence is how long a compound lasts before breaking down. Some pesticides degrade in soil in days. Polychlorinated biphenyls and many fluorinated compounds effectively do not degrade at all on human timescales, which is why the latter are called forever chemicals.

Partitioning describes where a chemical prefers to be: dissolved in water, adsorbed to soil particles, volatilized into air, or dissolved in fat. A fat-soluble, persistent compound will leave water and enter organisms, and stay there.

Bioaccumulation follows: an organism takes in a compound faster than it excretes it, so tissue concentrations rise over a lifetime. Biomagnification is the further step where concentrations rise at each level of a food chain, because a predator eats many prey and keeps what they carried. This is why a swordfish carries far more methylmercury than the water it swims in, and why fish advisories name specific species.

Long-range transport completes the picture and produces one of the field's most unsettling findings. Persistent organic pollutants evaporate in warm regions, travel on air currents, and condense in cold ones. Communities in the Arctic, thousands of kilometers from any industrial source and eating a traditional marine mammal diet at the top of a long food chain, have carried among the highest measured body burdens of certain persistent pollutants in the world. Nobody there manufactured, used, or benefited from those chemicals.

Measuring what is in people

Biomonitoring measures chemicals or their metabolites in human tissue: blood, urine, hair, breast milk, cord blood, deciduous teeth. It has one enormous advantage over modeling exposure: it integrates every route and every source at once. You do not need to know whether a person's phthalate exposure came from food packaging, personal care products, or dust, because the urine reflects all of it.

It also has real limits, and three matter most. Short-lived compounds tell you about the last day or two, not the last decade, so a single urine sample is a snapshot of a moving target. A concentration in urine or blood does not by itself say whether that amount is harmful, which requires a separately established health-based reference value. And detection improves faster than toxicology, so every generation of instruments finds more compounds in people than the previous one, without any change in what those people were exposed to.

Used properly, biomonitoring has been transformative. The clearest demonstration is lead. National survey data captured a decline in mean blood lead levels of roughly three quarters between the late 1970s and the early 1990s, tracking the removal of lead from gasoline almost exactly. That single dataset did more to settle the argument about leaded fuel than a decade of debate.

How big is the problem

The World Health Organization has tried to quantify the whole field. Its 2016 assessment attributed about 12.6 million deaths in 2012, roughly 23 percent of all deaths worldwide, to modifiable environmental factors: air pollution, unsafe water and sanitation, chemical exposures, occupational risks, radiation, and the built environment. The largest single contributors were air pollution and, in lower-income settings, water, sanitation, and hygiene.

Two features of that estimate deserve attention. First, the word modifiable is doing real work: this is not a count of deaths caused by pollution in some metaphysical sense, but an estimate of what would not have happened under achievable conditions. Second, the burden is not distributed evenly, either between countries or within them, and that unevenness is the subject of the last module of this course.

Atlas's Global Health course covers the international picture in detail, including household air pollution and the water and sanitation burden, and the Epidemiology course covers the study designs and causal reasoning behind every number in this one. This course concentrates on the mechanisms and the regulation.

Common misconceptions

• If a chemical is detectable in your blood, you have been harmed. Detection limits have fallen by orders of magnitude; presence at parts per trillion is a measurement achievement, not a diagnosis.
• Natural chemicals are safer than synthetic ones. Arsenic, aflatoxin, and radon are entirely natural, and among the most potent environmental carcinogens people encounter.
• Hazard and risk are the same thing. A substance can be genuinely hazardous and pose negligible risk at realistic exposures, and the reverse is also possible.
• Children are just small adults. Per kilogram they breathe, drink, and eat more, behave differently, and have developing organ systems, so identical exposures produce larger and more consequential doses.
• Distance from industry means safety. Persistent compounds travel to cold regions and biomagnify, giving Arctic communities some of the highest body burdens measured anywhere.

Where this leaves us

• Every environmental health problem sits on one chain: source, release, transport and fate, exposure, dose, effective dose, early effect, outcome.
• Intervening early in the chain protects more people for longer than intervening late.
• Hazard is a property, risk is a probability, exposure is outside the body, and dose is inside it.
• Persistence, partitioning, bioaccumulation, and long-range transport determine where a released chemical ends up.
• Biomonitoring integrates all routes at once, and a measured concentration still needs a health-based reference value to interpret.
• WHO attributed roughly 23 percent of global deaths in 2012 to modifiable environmental factors, unevenly distributed.

Sources

1. World Health Organization. (n.d.). Environmental health. who.int
2. Centers for Disease Control and Prevention. (n.d.). National Health and Nutrition Examination Survey. National Center for Health Statistics. cdc.gov
3. National Institute of Environmental Health Sciences. (n.d.). Environmental health topics. National Institutes of Health. niehs.nih.gov
4. Pruss-Ustun, A., Wolf, J., Corvalan, C., Bos, R., and Neira, M. (2016). Preventing disease through healthy environments: A global assessment of the burden of disease from environmental risks. World Health Organization.
5. Wild, C. P. (2005). Complementing the genome with an exposome. Cancer Epidemiology, Biomarkers and Prevention, 14(8), 1847-1850.

A radio contest in Sacramento

In January 2007 a Sacramento radio station ran a promotion called Hold Your Wee for a Wii. Contestants drank as much water as they could without using the bathroom. Jennifer Strange, a twenty-eight-year-old mother of three, drank roughly two gallons over about three hours. She complained of a severe headache, went home, and died that day of water intoxication. Her blood sodium had been diluted to the point that water moved into her brain cells and swelled them inside a rigid skull.

Water is not a poison. Water at that rate is. Paracelsus wrote the principle in the sixteenth century in a sentence that has survived five hundred years of chemistry: the dose makes the poison. Everything is toxic at some dose and nothing is toxic below some dose, with a set of exceptions important enough that the second half of this lesson is about them.

Why this matters: Toxicity is a property of a dose, not of a substance, which is why the only meaningful question about any chemical is how much, over how long, in whom.

The dose-response curve

Toxicology's central object is a graph. Put dose on the horizontal axis and some measure of response on the vertical, and for most agents you get an S-shaped curve: nothing happens, then effects appear and rise steeply, then the response saturates because everyone exposed is already affected.

Several landmarks on that curve have names you need.

Term	Definition	Used for
LD50	The dose lethal to half of an exposed animal population	Comparing acute potency between agents; a crude measure
NOAEL	No observed adverse effect level: the highest tested dose producing no significant adverse effect	The traditional starting point for setting a safe human exposure
LOAEL	Lowest observed adverse effect level: the lowest tested dose that did produce an effect	Used when no clean NOAEL exists, with an added safety factor
Benchmark dose	The dose producing a specified small response, estimated by fitting a model to all the data	Increasingly preferred, because it uses the whole curve rather than one tested dose

Notice a weakness in the NOAEL that took regulators decades to take seriously. A NOAEL is not a property of the chemical; it is a property of the experiment. It can only be one of the doses the researchers happened to test, and a study with few animals per group has low power, so it fails to detect real effects and thereby reports a higher NOAEL. A poorly designed study is rewarded with a more permissive answer. The benchmark dose approach fits a curve to all the data and reads off the dose corresponding to a defined response, which removes that perverse incentive.

Thresholds, and the agents that may not have one

For most toxic effects, the body has capacity to spare: enzymes that detoxify, cells that repair, reserves that absorb insult. Below some dose, that capacity is not exceeded and no adverse effect appears. That dose is a threshold, and threshold assumptions underlie essentially all non-cancer risk assessment.

Two categories are treated differently.

Genotoxic carcinogens damage DNA directly. In principle a single mutagenic event in a single cell could initiate a tumor, so regulators conventionally assume no threshold and model risk as proportional to dose all the way down to zero. This is a policy default rather than a demonstrated fact, since no experiment can measure a risk of one in a million, and it is deliberately conservative.

Developmental exposures are the other special case. During narrow windows of organ development, a dose that would be harmless to an adult can permanently alter structure or function. The effect is not merely larger; it is different in kind, and it is not reversible when exposure stops. This is why lead, methylmercury, and endocrine active compounds are regulated around pregnancy and childhood rather than around average adults.

You will also meet the word hormesis, the claim that some agents are beneficial at very low doses and harmful at higher ones, giving a J-shaped rather than a monotonic curve. It is real for some substances, essential nutrients most obviously, since selenium and copper are required and toxic. Whether it generalizes to industrial chemicals and radiation, and whether it should change regulation, is genuinely disputed, and the dispute is not resolved by asserting either that all chemicals are harmful at any dose or that all have safe low ranges.

In short: Most toxicity has a threshold, genotoxic carcinogens are modeled as though it does not, and developmental windows convert small doses into permanent effects.

What the body does with a chemical

Toxicokinetics traces four processes, usually abbreviated ADME.

Absorption depends on route. Inhalation delivers directly to a large, thin, highly vascular surface and bypasses the liver entirely. Ingestion passes through the gut and then through the liver's first-pass metabolism, which may destroy much of the dose or, occasionally, activate it. Dermal absorption is usually slow and depends heavily on whether a compound is fat soluble.

Distribution follows solubility. Water-soluble compounds stay in blood and extracellular fluid. Fat-soluble compounds accumulate in adipose tissue. Lead behaves like calcium and deposits in bone, where it can sit for decades and be remobilized during pregnancy, lactation, or bone loss in later life, which means an exposure at age five can deliver a dose at age forty.

Metabolism generally makes compounds more water soluble so they can be excreted, mostly through liver enzyme systems. Crucially, metabolism sometimes makes things worse: benzene is metabolized to reactive intermediates that damage bone marrow, and methanol is metabolized to formaldehyde and formic acid, which is why methanol poisoning is treated by giving ethanol to occupy the same enzyme.

Excretion is mostly renal, with biliary, exhaled, and, importantly for public health, lactational routes.

Two derived concepts matter for the rest of the course. Half-life is the time to eliminate half of what is present, and it varies over an extraordinary range: carbon monoxide clears from blood in hours, methylmercury in about two months, some fluorinated compounds in years. Body burden is the total amount present at a moment, and for anything with a long half-life it reflects years of accumulated intake rather than yesterday's exposure.

Why the same exposure is not the same exposure

Population risk assessment assumes a distribution of susceptibility, and it is worth knowing what drives it.

• Life stage. Fetuses and infants have immature metabolism, developing organs, and higher intake per kilogram; older adults have reduced renal clearance and often reduced reserve.
• Genetics. Enzyme variants change how fast a compound is activated or cleared, and the effect can be large. Slow acetylators clear certain aromatic amines differently, and variants in the enzymes that handle organophosphates change sensitivity to those pesticides.
• Existing disease. Asthma changes the meaning of an ozone exposure; kidney disease changes the meaning of a cadmium exposure.
• Nutrition. Iron deficiency increases lead absorption, which is one reason lead burdens and poor nutrition compound each other.
• Co-exposure. Smoking multiplies the risk from asbestos and from radon far beyond what either does alone.

That last point deserves a moment. Asbestos and smoking together produce lung cancer risk closer to the product of their separate effects than to their sum. When two agents multiply, a population that carries one of them experiences the other as a far worse hazard than population averages suggest.

Mixtures, which is what people are actually exposed to

Almost all toxicology tests one chemical at a time. Almost no human is exposed to one chemical at a time. Toxicologists describe three ways agents combine: additive, where effects sum, which is assumed for chemicals acting through the same mechanism; synergistic, where the combination exceeds the sum, as with asbestos and tobacco; and antagonistic, where one reduces the other's effect, as when selenium moderates some effects of mercury.

The regulatory response has been partial. Some frameworks group chemicals with a common mechanism and regulate them as a class, as the Food Quality Protection Act requires for organophosphate pesticides sharing a mode of action, and some drinking water standards for related compounds use a hazard index that sums the fractional contributions. Most regulation still handles chemicals individually, which is a known and unresolved gap, and it is the technical core of the cumulative impact argument you will meet in the environmental justice lesson.

Atlas's Epidemiology and Public Health course covers how human studies establish these relationships, including the confounding problems that make low-dose environmental epidemiology hard. This course takes those methods as given and concentrates on what they have found.

Common misconceptions

• Some substances are toxic and others are safe. Toxicity is a property of dose; water, oxygen, and table salt all kill at sufficient exposure.
• A NOAEL is the highest safe dose. It is the highest dose tested that produced no detected effect, so a small, weak study can produce a higher NOAEL than a good one.
• Metabolism always detoxifies. Benzene, methanol, and several pesticides become more toxic after metabolism, not less.
• Once exposure stops, the exposure is over. Lead stored in bone and persistent compounds in fat continue delivering internal dose for years or decades.
• Testing chemicals individually is adequate. Real exposures are mixtures, and interactions can be additive, synergistic, or antagonistic, with asbestos and smoking the clearest multiplicative example.

Putting it together

• The dose makes the poison: a fatal water-drinking contest and a therapeutic drug illustrate the same curve.
• NOAEL, LOAEL, and benchmark dose are three ways to summarize a dose-response study, and the benchmark dose uses all the data.
• Most effects have thresholds; genotoxic carcinogens are modeled without one, and developmental windows are a separate case entirely.
• ADME determines internal dose, and half-life determines whether body burden reflects yesterday or a decade.
• Susceptibility varies by life stage, genetics, disease, nutrition, and co-exposure, and some co-exposures multiply.
• People encounter mixtures while toxicology mostly tests single agents, a gap regulation has only partly closed.

Sources

1. Britannica. (2024). Toxicology. britannica.com
2. Agency for Toxic Substances and Disease Registry. (n.d.). Toxicological profiles. Centers for Disease Control and Prevention. atsdr.cdc.gov
3. National Institute of Environmental Health Sciences. (n.d.). Environmental health topics. National Institutes of Health. niehs.nih.gov
4. Klaassen, C. D. (Ed.). (2019). Casarett and Doull's Toxicology: The Basic Science of Poisons (9th ed.). McGraw-Hill Education.

Twenty-one thousand tons under a school

Between 1942 and 1953 the Hooker Chemical Company buried roughly 21,800 tons of chemical waste in an abandoned canal excavation in Niagara Falls, New York. In 1953 it covered the site with clay, sold the land to the local school board for one dollar, and included a deed clause disclaiming liability. A school was built on it. Houses went up around it.

By the mid-1970s, after wet years raised the water table, residents were finding oily black liquid in basements, chemical burns on children's feet, and corroding drums surfacing in yards. Lois Gibbs, a homeowner whose son had begun having seizures after starting at the school, organized her neighbors. On 2 August 1978 the New York State health commissioner declared a health emergency, and five days later President Carter declared a federal emergency, the first ever for a non-natural disaster. Two hundred and thirty-nine families were relocated, and about seven hundred more followed in 1980. In December of that year Congress created the Superfund program.

Here is the part that matters for this lesson. When state and federal officials were asked how much of these chemicals a person could safely be exposed to, they did not have a defensible way to answer. There was no agreed method. Five years later a National Research Council committee published one, and the framework in its report is still, with modifications, how the United States decides what a number in a regulation should be.

Key idea: Quantitative risk assessment exists because a specific political crisis exposed that agencies were making numerical decisions with no stated method.

Four steps

The framework has four parts, and each answers a distinct question.

Hazard identification asks what an agent can do at any dose. Evidence comes from human epidemiology, animal bioassays, mechanistic and cell-based studies, and structural comparison to known toxicants. This is where cancer classifications live, and where the most common public misreading happens. When an agency classifies a substance as a known human carcinogen, that is a statement about the strength of the evidence that it can cause cancer under some conditions. It says nothing about potency or about the risk at any actual exposure. Sunlight and processed meat are in that top evidence category alongside plutonium.

Dose-response assessment converts the study data into a quantitative relationship, and the method splits between cancer and non-cancer endpoints, as you will see below.

Exposure assessment estimates how much people actually contact, through which routes, for how long. It is usually the weakest link and the most decision-relevant one.

Risk characterization combines the previous two and states the result with its uncertainties, including who the estimate applies to.

From an animal study to a reference dose

For non-cancer effects, the output is a reference dose: an estimate of a daily exposure, in milligrams per kilogram of body weight per day, likely to be without appreciable risk of harm over a lifetime, including for sensitive groups. Its inhalation equivalent is a reference concentration in milligrams per cubic meter.

The calculation starts from a point of departure, typically a NOAEL or a benchmark dose lower bound, and divides by uncertainty factors. Each factor has a stated justification.

Factor	Typical value	What it covers
Interspecies	10	Extrapolating from test animals to humans
Intraspecies	10	Variation in sensitivity among humans, including children and the ill
Subchronic to chronic	1 to 10	Extrapolating a short study to lifetime exposure
LOAEL to NOAEL	1 to 10	Applied when no dose in the study was free of effects
Database	1 to 10	Gaps such as a missing developmental or reproductive study

Work an example. A well-conducted chronic rat study finds a NOAEL of 50 milligrams per kilogram per day for liver effects. The database is complete and the study is chronic, so only the two standard factors apply: 10 for animal to human and 10 for human variability, giving 100.

The reference dose is 50 divided by 100, or 0.5 milligrams per kilogram per day.

Now suppose the developmental toxicity study is missing. A database factor of 3 is added, giving a total of 300, and the reference dose drops to about 0.17. Notice what happened: the chemical did not change, and the number did. Reference doses encode how much we know as well as what we know.

Estimating what people actually get

Exposure assessment uses an intake equation. For drinking water the chronic daily intake, in milligrams per kilogram per day, is the concentration times the intake rate times the exposure frequency times the exposure duration, divided by body weight times the averaging time.

Use standard defaults: intake 2 liters per day, exposure frequency 350 days per year, exposure duration 30 years at one residence, body weight 70 kilograms, and for a non-cancer assessment an averaging time equal to the exposure duration, which is 30 times 365, or 10,950 days.

Suppose the contaminant is present at 5 milligrams per liter. The numerator is 5 times 2 times 350 times 30, which is 105,000. The denominator is 70 times 10,950, which is 766,500. The chronic daily intake is about 0.137 milligrams per kilogram per day.

The hazard quotient is that intake divided by the reference dose: 0.137 divided by 0.5, which is about 0.27. A hazard quotient below 1 indicates exposure below the level of concern.

Now change one input. Suppose sampling finds 20 milligrams per liter instead. The numerator becomes 420,000, the intake becomes about 0.548, and the hazard quotient becomes about 1.1. Crossing 1 does not mean people will be harmed, because the reference dose already contains a hundredfold margin. It means the exposure is no longer clearly within the protective envelope, which is the trigger for action rather than a prediction of disease.

What matters here: A hazard quotient compares an estimated intake to a deliberately conservative benchmark, so exceeding 1 signals that the margin has been eaten into, not that harm has occurred.

Cancer risk, which uses different arithmetic

For carcinogens modeled without a threshold, the output is not a safe dose but an estimated probability. The dose-response step produces a cancer slope factor, expressed in inverse milligrams per kilogram per day, and risk is the lifetime average daily dose multiplied by that slope.

The averaging time changes: for cancer, dose is spread over a nominal 70-year lifetime, which is 25,550 days, whether or not exposure lasted that long.

Suppose a contaminant is present at 0.002 milligrams per liter, with the same intake assumptions and a 30-year exposure. The numerator is 0.002 times 2 times 350 times 30, which is 42. The denominator is 70 times 25,550, which is 1,788,500. The lifetime average daily dose is about 0.0000235 milligrams per kilogram per day.

With a cancer slope factor of 1.5, the excess lifetime cancer risk is about 0.0000352, or roughly 35 in a million, about one in twenty-eight thousand.

Is that acceptable? That question is not answered by the arithmetic. In the Superfund program, EPA generally treats a range from one in a million to one in ten thousand as the zone within which cleanup decisions are made, with one in a million as the point of departure. Our result sits inside that band, which means the science has done its job and a policy judgment now has to be made.

Assessment, management, and the argument about the wall between them

The 1983 framework's most influential recommendation was conceptual: keep risk assessment, which asks what the science shows, separate from risk management, which asks what to do about it given cost, technology, law, and public values. The point was to prevent desired conclusions from bending the science.

That separation has been criticized from two directions, and both criticisms are serious.

The practical criticism is that a rigid wall produces assessments that answer questions nobody asked. If assessors never talk to decision makers, they may spend three years characterizing a pathway that is irrelevant to the actual choice. A 2009 National Research Council report on the framework recommended adding an explicit planning and scoping phase in which managers and assessors jointly define the decision the assessment must inform, and it also recommended treating cancer and non-cancer endpoints more consistently rather than with two disconnected methods.

The technical criticism concerns conservatism. Each uncertainty factor is individually defensible, and they multiply. Combine a hundredfold factor with a reasonable maximum exposure scenario built from upper-percentile intake assumptions and you can produce a number several orders of magnitude below any dose at which an effect has been observed. As a policy choice that is defensible: society may prefer to err toward protection. As a scientific claim it is not, and describing a reference dose as the level at which harm begins is simply wrong. Honest practice presents both a central-tendency and a high-end estimate, and says which is which.

Common misconceptions

• A carcinogen classification tells you how dangerous something is. It grades the strength of evidence that the agent can cause cancer, not potency or risk at real exposures.
• A hazard quotient above 1 means people will get sick. It means estimated intake exceeds a benchmark that already includes large protective factors.
• A reference dose is the dose at which harm begins. It is a point of departure divided by uncertainty factors, often a hundredfold or more below any observed effect level.
• Risk assessment decides policy. It produces an estimate; whether a risk of one in twenty thousand is acceptable is a management judgment involving cost, law, and values.
• The numbers are precise. Exposure assessment is usually the weakest link, and the output is an estimate with uncertainty that should be stated alongside it.

The short version

• Love Canal exposed the absence of any agreed method, and the 1983 framework supplied one.
• The four steps are hazard identification, dose-response assessment, exposure assessment, and risk characterization.
• A reference dose is a point of departure divided by justified uncertainty factors, so missing data lowers the number.
• Chronic daily intake divided by the reference dose gives a hazard quotient, with 1 as the action trigger.
• Cancer risk is a lifetime average daily dose times a slope factor, judged against a management range from one in a million to one in ten thousand.
• Separating assessment from management protects the science and, taken too rigidly, produces answers to the wrong questions.

Sources

1. U.S. Environmental Protection Agency. (n.d.). Risk assessment. epa.gov
2. U.S. Environmental Protection Agency. (n.d.). Integrated Risk Information System (IRIS). epa.gov
3. National Research Council. (1983). Risk assessment in the federal government: Managing the process. National Academies Press. nap.nationalacademies.org
4. Wikipedia contributors. (n.d.). Love Canal. en.wikipedia.org
5. National Research Council. (2009). Science and decisions: Advancing risk assessment. National Academies Press.

Five days in a river valley

Donora, Pennsylvania sits in a horseshoe bend of the Monongahela River, ringed by hills, with a zinc smelter and a steel wire works along the water. On 26 October 1948 a temperature inversion settled over the valley: a layer of warm air above cold, which acts as a lid and stops anything below from rising. The mills kept running. The smoke had nowhere to go.

By the third day people could not see across the street. Firefighters went house to house with oxygen. Twenty people died during the episode, in a town of about fourteen thousand, and roughly six thousand, nearly half the population, were made ill. The mills shut down on the morning of 31 October. Rain came that afternoon and the air cleared.

Four years later, in December 1952, a similar inversion held coal smoke over London for five days. The initial official count was about four thousand excess deaths; later analyses put the figure far higher. Britain passed a Clean Air Act in 1956. The United States passed its first federal air pollution law in 1955.

Nobody now dies in five days from an American air pollution episode. What replaced those episodes is subtler, larger, and harder to see: a continuous exposure at concentrations nobody notices, producing deaths that appear in mortality statistics rather than in newspapers.

Key idea: Air pollution stopped being an acute disaster and became a chronic one, which made it far less visible and considerably more deadly in total.

Six pollutants and their standards

The Clean Air Act directs EPA to identify pollutants that endanger public health and come from numerous sources, and to set national standards for them. Six are on the list, and they are called criteria pollutants because the standards must be based on published criteria documents summarizing the science.

Pollutant	Main sources	Primary standard
Fine particles, PM2.5	Combustion: vehicles, power plants, wood and wildfire smoke; also formed from gases	9.0 micrograms per cubic meter annual, 35 over 24 hours
Coarse particles, PM10	Dust, construction, agriculture, grinding	150 micrograms per cubic meter over 24 hours
Ozone	Not emitted; formed from nitrogen oxides and volatile organics in sunlight	0.070 parts per million over 8 hours
Nitrogen dioxide	Vehicles, power plants, gas appliances indoors	100 parts per billion over 1 hour, 53 annual
Sulfur dioxide	Coal and oil combustion, smelting	75 parts per billion over 1 hour
Carbon monoxide	Incomplete combustion: vehicles, heaters, generators	9 parts per million over 8 hours
Lead	Metals processing, piston-engine aviation fuel	0.15 micrograms per cubic meter, rolling three-month average

Two features of this system matter more than the numbers. First, each pollutant has a primary standard, set at a level requisite to protect public health with an adequate margin of safety including sensitive groups, and a secondary standard, protecting welfare: crops, forests, visibility, buildings. Second, the standards are reviewed on a five-year cycle with advice from an independent scientific committee, and they have generally tightened. The annual fine particle standard, for instance, was lowered to 9.0 micrograms per cubic meter in 2024, having previously been 12.0 and before that 15.0. The World Health Organization's 2021 guideline is 5.

That gap between an American standard and a WHO guideline is not a disagreement about the science. WHO guidelines are health-based recommendations with no legal force. American primary standards are enforceable requirements that trigger mandatory state planning, and, as you will see in the final lesson, the statute forbids EPA from considering the cost of meeting them.

Why particles do the most damage

Size determines fate. Particles larger than about 10 micrometers are mostly caught in the nose and upper airway. Particles below 10 reach the thoracic airways. Particles below 2.5 micrometers, which is roughly one thirtieth the width of a human hair, penetrate to the alveoli where gas exchange happens and there is no mucociliary escalator to remove them. Ultrafine particles below 0.1 micrometers can cross into the bloodstream.

Deposited particles provoke local inflammation and oxidative stress, and that inflammation does not stay local. Systemic inflammatory mediators, autonomic nervous system effects on heart rate variability, and endothelial dysfunction combine to accelerate atherosclerosis and to trigger acute cardiovascular events. This is why the largest share of deaths attributed to fine particles is cardiovascular rather than respiratory, which surprises most students.

The evidence comes in three layers, and their convergence is what makes the case strong.

Cohort studies. The Harvard Six Cities study followed just over eight thousand adults in six American cities for more than a decade and reported an adjusted mortality rate ratio of 1.26 between the most polluted city, Steubenville, Ohio, and the least polluted, Portage, Wisconsin. The American Cancer Society cohort, with roughly half a million participants, found that each additional 10 micrograms per cubic meter of fine particles was associated with about a 4 percent increase in all-cause mortality, 6 percent in cardiopulmonary mortality, and 8 percent in lung cancer mortality.

Natural experiments. These answer the obvious objection that polluted cities differ in many ways. When a labor dispute closed the steel mill in Utah Valley for thirteen months in the 1980s, particulate levels fell sharply and children's hospital admissions for respiratory illness fell by roughly half, then rose again when the mill reopened. When Dublin banned coal sales in 1990, respiratory and cardiovascular death rates in the city dropped measurably in the following years. The exposure changed for reasons unrelated to health, and health changed with it.

Mechanism. Controlled human exposure studies and animal work supply the biological pathway connecting the two.

The point: Fine particle mortality rests on cohort associations, natural experiments where exposure changed for unrelated reasons, and a demonstrated biological mechanism, which is about as strong as observational environmental evidence gets.

Ozone, which nobody emits

Ground-level ozone is worth understanding separately because it behaves unlike the others. Nothing emits it. It forms when nitrogen oxides and volatile organic compounds react in sunlight, which is why ozone peaks on hot sunny afternoons, in summer, and often downwind of a city rather than in it.

That chemistry produces a genuinely counterintuitive result. Near heavy traffic, fresh nitric oxide destroys ozone through a reaction called titration, so the most congested corridors can show lower ozone than the suburbs downwind. Reduce traffic on a weekend and ozone in the city center can go up. Control strategies therefore have to be regional, and they have to consider which precursor is limiting: in some airsheds cutting nitrogen oxides helps, and in others it must be paired with cutting volatile organics or it does little.

Ozone itself is a powerful oxidant. It inflames airways, reduces lung function measurably in healthy exercising adults within hours, worsens asthma, and is associated with respiratory mortality. Outdoor workers and children who play outside on summer afternoons receive the highest doses, because dose depends on ventilation rate.

Indoor air, where people actually are

Americans spend roughly ninety percent of their time indoors, so an outdoor monitor measures a place most people are not. Indoor concentrations of some pollutants routinely exceed outdoor ones.

The largest global indoor problem is combustion for cooking and heating with wood, dung, charcoal, or coal, which Atlas's Global Health course covers as a leading cause of death worldwide. In wealthier settings the list runs differently: secondhand tobacco smoke, radon seeping from soil, volatile organic compounds from furnishings and cleaning products, carbon monoxide from malfunctioning appliances, mold following water damage, and nitrogen dioxide from unvented gas cooking.

Gas stoves deserve an honest paragraph because the topic became politically charged. It is well established that unvented gas cooking raises indoor nitrogen dioxide, often above outdoor standards, and that nitrogen dioxide exposure is associated with childhood asthma. A 2022 analysis estimated that a substantial share of childhood asthma in the United States, on the order of one case in eight, could be attributed to gas stove use. That estimate depends on assumptions about exposure and on transferring risk estimates from other settings, and it has been contested by other researchers. The defensible summary is that the exposure is real and measurable, the direction of the association is consistent, and the size of the population effect is uncertain. Ventilation, specifically a range hood vented outdoors, addresses it directly.

What has improved and what has not

American air is dramatically cleaner than it was. EPA's aggregate emissions of the six criteria pollutants fell by roughly three quarters between 1970 and 2020 while the economy and vehicle miles traveled grew substantially, which is one of the strongest available rebuttals to the claim that environmental regulation and economic growth are incompatible.

Three things complicate the good news. Wildfire smoke has begun reversing particulate gains across the western United States, and smoke is not a small contributor: in bad years it dominates the annual average in affected regions. Exposure remains unequally distributed within metropolitan areas, which is the subject of a later lesson. And globally the picture is far worse: the World Health Organization estimates that nearly the entire world population breathes air exceeding its guidelines, and attributes millions of premature deaths each year to ambient and household air pollution combined.

Common misconceptions

• Air pollution mainly causes lung disease. The largest share of attributable deaths is cardiovascular, through systemic inflammation and effects on the vasculature.
• Ozone is emitted by cars. Cars emit its precursors; ozone forms photochemically and often peaks downwind, so heavy traffic corridors can show lower ozone than suburbs.
• Indoor air is cleaner than outdoor air. For several pollutants the reverse is true, and people spend about ninety percent of their time indoors.
• The WHO guideline and the US standard disagree about the science. Guidelines are health-based recommendations; the US primary standard is an enforceable requirement set under a specific statute.
• Air quality regulation stopped economic growth. Criteria pollutant emissions fell by roughly three quarters from 1970 to 2020 while output and vehicle travel rose.

What to carry forward

• Donora in 1948 and London in 1952 established that air pollution kills, and made acute episodes the model for a chronic problem.
• Six criteria pollutants carry primary health-based standards and secondary welfare standards, reviewed every five years.
• Particle size determines where a particle deposits, and fine particles reach the alveoli and drive systemic cardiovascular effects.
• Cohort studies, natural experiments, and mechanism converge on fine particulate mortality.
• Ozone is a secondary pollutant whose control requires regional strategy and attention to which precursor is limiting.
• Indoor air matters because that is where people are, and gas cooking, radon, and combustion sources dominate it.

Sources

1. U.S. Environmental Protection Agency. (n.d.). Criteria air pollutants. epa.gov
2. U.S. Environmental Protection Agency. (n.d.). NAAQS table. epa.gov
3. World Health Organization. (n.d.). Air pollution. who.int
4. Dockery, D. W., Pope, C. A., Xu, X., Spengler, J. D., Ware, J. H., Fay, M. E., Ferris, B. G., and Speizer, F. E. (1993). An association between air pollution and mortality in six U.S. cities. New England Journal of Medicine, 329(24), 1753-1759.
5. Pope, C. A., Burnett, R. T., Thun, M. J., Calle, E. E., Krewski, D., Ito, K., and Thurston, G. D. (2002). Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. JAMA, 287(9), 1132-1141.

Four hundred thousand people in one city

In late March 1993, pharmacies across Milwaukee ran out of antidiarrheal medication. Absenteeism spiked in schools and at hospitals. The city's own laboratory found that turbidity at one of its two water treatment plants, a measure of cloudiness that serves as a proxy for whether filtration is working, had been unusually high for days.

The cause was Cryptosporidium, a protozoan parasite whose thick-walled oocysts pass through chlorine essentially unharmed. Filtration is the barrier that stops it, and the filtration at the Howard Avenue plant had degraded. An estimated 403,000 people became ill, the largest documented waterborne disease outbreak in United States history, and dozens died, most of them people with advanced HIV disease whose immune systems could not clear the infection.

Milwaukee was not a poor city with no treatment plant. It had a modern system, in compliance with the standards of the day, and the standards of the day did not require what would have caught this. The outbreak drove a national tightening of filtration and turbidity rules.

Why this matters: Drinking water safety is a chain of physical barriers, and the pathogen that gets through is the one your barriers were not designed for.

The treatment train

Conventional surface water treatment is a sequence, and each step handles what the previous one leaves.

• Coagulation and flocculation. A coagulant such as aluminum or iron salt is added, neutralizing the electrical charges that keep fine particles suspended, so they clump into larger flocs.
• Sedimentation. The flocs, now heavy enough, settle out in a quiet basin.
• Filtration. Water passes through sand, anthracite, or membranes, physically removing remaining particles including protozoan cysts and oocysts. This is the barrier that failed in Milwaukee.
• Disinfection. Chlorine, chloramine, ozone, or ultraviolet light inactivates bacteria and viruses. A residual disinfectant is maintained in the distribution system so that contamination entering through a pipe break does not travel unchecked.

Groundwater from a protected aquifer is often naturally filtered by the soil column and may need only disinfection, which is why the treatment requirements differ by source. Turbidity is the workhorse operational measurement: it is cheap, continuous, and correlates with whether filtration is doing its job, which is why regulators use it as a surrogate for pathogen removal rather than testing for parasites directly.

One barrier deserves special note. Chlorine is inexpensive, provides a lasting residual, and inactivates bacteria and viruses effectively. It is nearly useless against Cryptosporidium, which is why filtration and, increasingly, ultraviolet disinfection carry that load. Ultraviolet light damages the organism's DNA and works well on protozoa while providing no residual, so systems using it still add a chemical disinfectant afterward.

Two numbers for every contaminant

Under the Safe Drinking Water Act, EPA sets two values for each regulated contaminant, and the difference between them is one of the most instructive things in environmental regulation.

The maximum contaminant level goal is the concentration at which no known or anticipated adverse health effect occurs, with an adequate margin of safety. It is purely health-based, it considers no cost or feasibility, and it is not enforceable. For contaminants treated as having no safe threshold, including genotoxic carcinogens and lead, the goal is zero.

The maximum contaminant level is the enforceable standard, set as close to the goal as is feasible using the best available technology, taking cost into account. Where measuring a contaminant reliably is impractical, EPA instead specifies a treatment technique, a required process rather than a number.

So when you read that the goal for a contaminant is zero and the enforceable limit is five parts per billion, you are not looking at an inconsistency. You are looking at the exact place where health science hands the problem to engineering and economics, in public, with both numbers published.

Remember: The goal states what health science would prefer, the enforceable level states what is technically and economically achievable, and the gap between them is a policy choice made in the open.

The disinfection tradeoff

Chlorine reacts with natural organic matter in source water to produce disinfection byproducts, principally trihalomethanes and haloacetic acids, some of which are associated with bladder cancer in epidemiologic studies and are regulated at low levels.

This creates a genuine tension, and it is worth being precise about how it resolves. The risks are not comparable in size. Waterborne infectious disease killed enormous numbers of people in American cities before disinfection and still kills more than a million people a year worldwide. The estimated cancer risk from byproducts at regulated concentrations is small. Utilities manage the tradeoff by removing organic matter before disinfecting, adjusting the disinfectant, and monitoring byproducts, not by disinfecting less.

The lesson generalizes beyond water. Any intervention with a large benefit and a small, more easily measured harm will attract disproportionate attention to the harm, because the benefit is invisible: it consists of illnesses that did not happen.

What the standards do not cover

Federal drinking water regulation applies to public water systems. Several important populations fall outside it.

Private wells serve roughly one American in eight, and they are not regulated by the Safe Drinking Water Act. Nobody tests them unless the owner pays for it. Well water can carry naturally occurring arsenic, uranium, or radon, nitrate from fertilizer and septic systems, and bacteria from surface infiltration. Nitrate above the standard causes methemoglobinemia in infants, the condition known as blue baby syndrome, and it is a specific hazard for households mixing formula with well water in agricultural areas.

Premise plumbing, meaning the pipes inside a building, is where the water quality a person actually drinks is determined, and it is largely outside the regulatory system. Legionella grows in warm stagnant water in building systems: cooling towers, hot water tanks held at low temperatures, decorative fountains, and rarely used fixtures. Outbreaks of Legionnaires' disease are building problems more than treatment plant problems, and they have increased substantially in reported incidence in the United States over the past two decades.

Households without adequate sanitation. This one surprises students. In Lowndes County, Alabama, a 2017 study tested residents in an area where many homes rely on failing septic systems or straight-pipe discharge because the soil does not percolate and municipal sewer service does not reach them. About a third of participants showed evidence of hookworm infection, a parasite associated with inadequate sanitation and generally assumed to have been eliminated from the United States decades ago. Similar wastewater gaps exist in colonias along the southern border, on tribal lands, and in parts of Appalachia.

Sanitation, and the global picture in one paragraph

Wastewater treatment mirrors drinking water treatment in reverse. Primary treatment settles solids. Secondary treatment uses microorganisms to consume dissolved organic matter, dramatically lowering biochemical oxygen demand so the effluent does not strip oxygen from receiving waters. Tertiary treatment, where required, removes nutrients such as nitrogen and phosphorus that would otherwise drive algal blooms, and disinfects.

Older cities carry a structural problem called a combined sewer, in which stormwater and sewage share one pipe. In dry weather everything goes to the plant. In heavy rain the combined flow exceeds capacity and the system is designed to overflow, discharging diluted sewage into rivers and harbors. Hundreds of American communities still have combined systems, and separating them costs billions.

Globally, the World Health Organization and UNICEF estimate that around two billion people lack safely managed drinking water and around three and a half billion lack safely managed sanitation. Atlas's Global Health course treats that burden, including the diarrheal disease it causes and the interventions that reduce it, in detail. What the American cases above show is that the distinction between rich-country and poor-country water problems is a generalization with real exceptions inside wealthy countries.

Common misconceptions

• Chlorine handles everything. Cryptosporidium is highly resistant to chlorine; filtration and ultraviolet disinfection are the barriers that stop it.
• The health goal and the legal limit should be the same number. The goal is purely health-based, and the enforceable limit reflects treatment feasibility and cost, which the statute directs EPA to consider.
• Disinfection byproducts make chlorination a bad tradeoff. The infectious disease risk prevented is orders of magnitude larger; utilities manage byproducts by removing organics first, not by disinfecting less.
• Well water is naturally pure. Private wells are unregulated and can carry arsenic, nitrate, uranium, radon, and bacteria, and nobody tests them unless the owner does.
• Waterborne parasitic disease from poor sanitation does not occur in wealthy countries. Testing in rural Alabama found hookworm in about a third of participants in an area with failing septic systems.

Pulling it together

• Milwaukee in 1993 sickened an estimated 403,000 people because a filtration barrier failed against a chlorine-resistant parasite.
• Coagulation, sedimentation, filtration, and disinfection each remove what the previous step leaves, with turbidity as the operational proxy.
• Maximum contaminant level goals are health-based and unenforceable; maximum contaminant levels are feasible, enforceable, and often higher.
• Disinfection byproducts are a real but far smaller risk than the waterborne disease disinfection prevents.
• Private wells, premise plumbing, and unsewered communities fall largely outside the regulatory system.
• Combined sewers discharge during heavy rain by design, and globally billions still lack safely managed water and sanitation.

Sources

1. U.S. Environmental Protection Agency. (n.d.). National primary drinking water regulations. epa.gov
2. U.S. Environmental Protection Agency. (n.d.). Safe Drinking Water Act (SDWA). epa.gov
3. WHO and UNICEF Joint Monitoring Programme for Water Supply, Sanitation and Hygiene. (n.d.). Household data. washdata.org
4. Mac Kenzie, W. R., Hoxie, N. J., Proctor, M. E., Gradus, M. S., Blair, K. A., Peterson, D. E., et al. (1994). A massive outbreak in Milwaukee of Cryptosporidium infection transmitted through the public water supply. New England Journal of Medicine, 331(3), 161-167.
5. McKenna, M. L., McAtee, S., Bryan, P. E., Jeun, R., Ward, T., Kraus, J., et al. (2017). Human intestinal parasite burden and poor sanitation in rural Alabama. American Journal of Tropical Medicine and Hygiene, 97(5), 1623-1628.

The building they called loony gas

In October 1924, at Standard Oil's Bayway refinery in New Jersey, workers producing tetraethyl lead began hallucinating. Five of them died within days. Dozens more were hospitalized with tremors and psychosis. Similar deaths had already occurred at DuPont's plant in Deepwater. Newspapers called the facility the loony gas building.

Leaded gasoline had gone on sale in Dayton, Ohio the previous year, added to fuel because it suppressed engine knock cheaply. After the deaths, several cities banned it and the Surgeon General convened a conference in 1925. The industry argued that leaded gasoline was essential to progress, that the refinery deaths reflected handling of the concentrated additive rather than the product, and that no evidence showed harm from ordinary use. A committee found no proof of danger at expected exposures. Sales resumed. The framework that emerged, associated with the industry-funded toxicologist Robert Kehoe, placed the burden of proof squarely on anyone alleging harm: absent positive evidence of injury, use should continue.

Leaded gasoline was finally banned for on-road vehicles in the United States in 1996, seventy-three years after that first sale. Every step of the intervening argument is documented, which is why lead is the best case study environmental health has for how a society revises what it believes is safe, and how slowly.

Key idea: The lead record is not a story about a chemical that turned out to be worse than expected; it is a story about who carries the burden of proof while evidence accumulates.

The scientist who could not get a clean measurement

In the early 1950s a geochemist named Clair Patterson was trying to determine the age of the Earth by measuring lead isotopes in meteorites. His results kept coming back contaminated. Chasing the contamination, he built what amounted to the first ultra-clean laboratory, and in the process discovered something more important than the number he was after: lead was everywhere, at concentrations far above anything natural.

He and colleagues then measured lead in Greenland ice cores, reading deposition year by year through layers of snow. The record showed a rise during Roman-era smelting, a larger rise with industrialization, and then, after the 1920s, a nearly vertical climb. In 1965 Patterson published an argument that average Americans carried body burdens on the order of a hundred times natural levels. He spent the following decades in conflict with an industry that funded much of the field's research, and he was for a time excluded from advisory panels on lead.

The important methodological point is that Patterson did not begin by studying health. He began by establishing what the baseline was. Without a natural baseline, an elevated level looks normal, because everyone has it.

The number kept falling

Blood lead concentration, measured in micrograms per deciliter, became the standard exposure metric. The official level at which public health action was warranted fell repeatedly, and each fall followed better studies rather than any change in lead.

Year	Level	What changed
1960s	60	Roughly the threshold for overt clinical poisoning
1971	40	Recognition of subclinical effects on heme synthesis
1975	30	Accumulating pediatric evidence
1985	25	Neurodevelopmental studies at lower exposures
1991	10	Cohort studies linking low exposures to cognitive deficits
2012	5 as a reference value	Abandonment of the idea of a threshold; a population-based value replaced the level of concern
2021	3.5 as a reference value	Update reflecting falling population exposure, not a new safety finding

The 2012 change deserves attention, because it is a different kind of move. Before it, the number was described as a level of concern, implying that below it there was no concern. CDC replaced that with a blood lead reference value set at the 97.5th percentile of measured levels in young children, meaning it identifies children with unusually high exposure relative to their peers. It is explicitly not a safety threshold, and the 2021 reduction to 3.5 happened because the population distribution shifted downward, not because 5 became newly dangerous.

Two pieces of work drove that reconceptualization. Herbert Needleman, in a 1979 study, collected shed baby teeth from schoolchildren in Chelsea and Somerville, Massachusetts, measured lead in dentin, and found that children with higher tooth lead had lower IQ scores and worse teacher-rated classroom behavior, at exposures then considered harmless. He was subsequently the target of misconduct allegations, and was cleared. In 2005 a pooled analysis of seven international cohorts by Bruce Lanphear and colleagues examined the dose-response relationship directly and found measurable IQ decrements below 10 micrograms per deciliter, with the curve steeper at low concentrations than at high ones. That shape is important: the first few micrograms do more damage per unit than the next ten, which means there is no safe level and the biggest gains come from protecting the least exposed children.

The upshot: Every downward revision of the lead standard followed better measurement, and the final revision abandoned the idea of a threshold entirely.

What lead does, and one contested claim

In children, lead impairs neurodevelopment: measured effects on IQ, attention, executive function, and behavior, plus hearing and growth. At high levels it causes anemia, encephalopathy, and death. In adults it raises blood pressure, damages kidneys, and is associated with cardiovascular mortality; a large analysis of national survey data linked historical lead exposure to a substantial number of annual cardiovascular deaths. Lead crosses the placenta freely and is remobilized from maternal bone during pregnancy, so a woman's childhood exposure can reach her fetus decades later.

One further claim deserves an honest hearing rather than either endorsement or dismissal. The removal of lead from gasoline in the 1970s and 1980s was followed, roughly two decades later, by a large decline in violent crime, and researchers including Rick Nevin and Jessica Reyes documented that the timing held across countries and across American states with different phasedown schedules. The proposed mechanism, impaired impulse control and executive function from early exposure, is biologically plausible and consistent with the neurodevelopmental literature. The objections are equally serious: these are ecological correlations, many other things changed over the same decades, and individual-level cohort studies have produced mixed results, with some finding associations with antisocial behavior and arrests and others finding little after adjustment. The defensible position is that lead exposure plausibly contributed something to the crime decline, that its share is not well estimated, and that the case for removing lead never depended on it.

Where lead is now

The gasoline phasedown, beginning in the mid-1970s and completed for on-road vehicles in 1996, produced one of the clearest public health graphs ever drawn: national survey data show mean blood lead levels falling by roughly three quarters between the late 1970s and the early 1990s, tracking the removal of lead from fuel almost exactly. Economic analyses of the phasedown consistently found benefits exceeding costs by a wide margin, which is why it is a standard example in cost-benefit teaching.

What remains:

• Paint. Lead paint was banned for residential use in 1978. Housing built before then still contains it, and the hazard is not the intact paint but the dust generated as it deteriorates or is disturbed by renovation.
• Soil. Decades of vehicle exhaust and paint weathering deposited lead near roads and around old buildings, where it persists indefinitely and reaches children through hand-to-mouth contact and tracked-in dust.
• Water. Lead service lines, leaded solder banned in 1986, and brass fixtures. The 1986 definition of lead free still permitted brass containing up to eight percent lead; that was tightened in 2011 to a weighted average of a quarter percent.
• Consumer products. Imported spices, traditional remedies, cosmetics, and glazed ceramics have repeatedly been found with high lead content.
• Aviation fuel. Piston-engine aircraft still use leaded gasoline, and EPA issued a finding in 2023 that emissions from it endanger public health.
• Occupational take-home. Workers in battery plants, smelters, radiator shops, and shooting ranges carry lead home on clothing and in vehicles.

Flint, step by step

On 25 April 2014, under a state-appointed emergency manager and facing a budget crisis, Flint, Michigan switched its drinking water source from Detroit's treated Lake Huron supply to the Flint River, to be treated at the city's own plant.

The chemistry that followed is worth understanding precisely, because it explains why the crisis was a water treatment failure rather than a contamination event. Lead service lines do not normally deliver much lead, because a protective scale of lead-containing mineral deposits builds up on the pipe's interior over decades. That scale is stable only under the right water chemistry. Utilities maintain it deliberately through corrosion control, typically by adding orthophosphate, which forms a low-solubility coating.

Flint River water was more corrosive than Lake Huron water, with higher chloride from road salt. The city did not add a corrosion control inhibitor. Over months, the protective scale dissolved, exposing bare lead to the water.

Residents complained immediately about color, odor, and taste. In October 2014 General Motors stopped using the city's water at its engine plant because it was corroding parts. In early 2015 sampling at one resident's home, LeeAnne Walters, returned lead concentrations far above any regulatory threshold. In August 2015 a team from Virginia Tech led by Marc Edwards conducted independent citywide sampling and reported widespread severe exceedances. In September 2015 Dr. Mona Hanna-Attisha, a pediatrician at Hurley Medical Center, analyzed blood lead results for Flint children before and after the switch and found that the proportion with elevated levels had roughly doubled citywide, with larger increases in the neighborhoods where water lead was highest. The city returned to Detroit water in October 2015. A Legionnaires' disease outbreak during the same period killed at least a dozen people.

Two features of Flint generalize. First, this had happened before: Washington, D.C. experienced a serious lead release between 2001 and 2004 after a change in disinfectant altered water chemistry, and the lessons were available. Second, residents identified the problem correctly and early, and were told the water met standards, which was in a narrow sense true under a rule that tests a small number of taps and reports a ninetieth percentile.

The federal Lead and Copper Rule sets an action level of 15 parts per billion at the ninetieth percentile of sampled taps, which triggers corrosion control, public education, and service line replacement. It is a treatment technique trigger, not a health standard, and the health-based goal for lead in drinking water is zero. A 2024 revision requires most systems to replace lead service lines generally within ten years and lowers the action level to 10 parts per billion.

Common misconceptions

• Lead was banned once scientists showed it was harmful. Harm was documented in 1924; the on-road gasoline ban came in 1996, and the intervening argument was about burden of proof.
• Blood lead below the reference value is safe. The reference value is a population percentile, not a threshold, and effects have been measured well below it.
• Intact lead paint is the hazard. The hazard is dust from deteriorating or disturbed paint, which is why renovation of older housing is a specific risk.
• Flint's water was contaminated by an industrial spill. The lead came from the city's own service lines after the absence of corrosion control dissolved their protective scale.
• The 15 parts per billion action level is a health standard. It is a trigger for treatment actions at the ninetieth percentile of sampled taps; the health-based goal is zero.

Looking back

• Workers died producing tetraethyl lead in 1924, and the resulting framework put the burden of proof on those alleging harm.
• Clair Patterson established a natural baseline first, which is what made modern burdens visible as abnormal.
• The official level of concern fell from 60 to 10 and was then replaced by a reference value, because no threshold could be defended.
• The dose-response curve is steeper at low exposures, so the first micrograms matter most.
• Gasoline removal cut population blood lead by roughly three quarters; paint dust, soil, service lines, and imports remain.
• Flint was a corrosion control failure that dissolved protective scale inside existing lead lines, and it had a documented precedent.

Sources

1. U.S. Environmental Protection Agency. (n.d.). Lead. epa.gov
2. Agency for Toxic Substances and Disease Registry. (n.d.). Lead toxicological profile and health effects. Centers for Disease Control and Prevention. atsdr.cdc.gov
3. Wikipedia contributors. (n.d.). Flint water crisis. en.wikipedia.org
4. Needleman, H. L., Gunnoe, C., Leviton, A., Reed, R., Peresie, H., Maher, C., and Barrett, P. (1979). Deficits in psychologic and classroom performance of children with elevated dentine lead levels. New England Journal of Medicine, 300(13), 689-695.
5. Lanphear, B. P., Hornung, R., Khoury, J., Yolton, K., Baghurst, P., Bellinger, D. C., et al. (2005). Low-level environmental lead exposure and children's intellectual function: An international pooled analysis. Environmental Health Perspectives, 113(7), 894-899.
6. Hanna-Attisha, M., LaChance, J., Sadler, R. C., and Champney Schnepp, A. (2016). Elevated blood lead levels in children associated with the Flint drinking water crisis. American Journal of Public Health, 106(2), 283-290.

A successful intervention that poisoned tens of millions

In the 1970s, children in Bangladesh were dying in enormous numbers from diarrheal disease contracted from surface water. The response was sensible, cheap, and effective: sink tube wells into the shallow groundwater, which is naturally free of the pathogens that contaminate ponds and rivers. Aid agencies and the government drilled millions of them. Child mortality from waterborne infection fell substantially.

The alluvial sediments beneath the Ganges delta contain naturally occurring arsenic, mobilized into groundwater under the reducing chemical conditions in those aquifers. Nobody tested for it, because nobody was looking for a geological contaminant in a program aimed at bacteria. Arsenicosis takes years to appear, so the first skin lesions surfaced long after the wells were in use. Arsenic in groundwater was identified in neighboring West Bengal in the 1980s and confirmed in Bangladesh in 1993.

In 2000 a paper in the Bulletin of the World Health Organization estimated that between 35 and 77 million people were at risk and described the situation as the largest mass poisoning of a population in history.

Nothing about this was malice or negligence in the ordinary sense. It was a chain reasoned correctly at every link except one that nobody thought to check.

Key idea: The most consequential environmental exposures are frequently created by interventions that solved a different problem competently.

Arsenic

Arsenic is a metalloid, ubiquitous in the earth's crust, and its toxicity depends sharply on form. Inorganic arsenic, in its trivalent and pentavalent states, is the toxic form and the one that dissolves into groundwater. Organic arsenic compounds found in seafood, chiefly arsenobetaine, are largely excreted unchanged and are not considered a significant hazard, which matters because total arsenic measurements in food can look alarming without meaning anything.

Chronic exposure produces a distinctive progression. First come skin changes: hyperpigmentation, and hyperkeratosis on palms and soles. Later come cancers of the skin, bladder, and lung, along with elevated risks of cardiovascular disease, diabetes, and adverse pregnancy outcomes. Prenatal and early childhood exposure is associated with reduced cognitive performance. Latency runs from years to decades, which is precisely why the Bangladesh wells operated for two decades before the problem became visible.

The United States lowered its drinking water standard from 50 to 10 micrograms per liter in a rule finalized in 2001, with compliance required by 2006. That change was contested at the time on cost grounds, and it illustrates the risk assessment lesson from Module 1: at 50 micrograms per liter, the estimated lifetime cancer risk was far above the range regulators normally accept. American exposure is concentrated in private wells in parts of the Southwest, New England, and the upper Midwest, which are outside the regulatory system entirely.

Food is the other route. Rice takes up arsenic from paddy soil and water more efficiently than most crops, and rice-based infant cereal became a specific concern because infants eat a large amount of a single food relative to body weight. The Food and Drug Administration set an action level for inorganic arsenic in infant rice cereal, and varying grains and rinsing rice reduce intake.

Remediation in Bangladesh has been genuinely hard. Wells were tested and painted green or red. Deeper wells often draw from aquifers with less arsenic but risk drawing contamination downward over time. Filtration works and requires maintenance. And moving households back to surface water reintroduces the pathogen problem the wells were dug to solve. This is a real tradeoff, not a rhetorical one.

Mercury, in three chemical forms

Mercury behaves so differently in its three forms that treating them as one substance guarantees confusion.

Form	Main exposure	Principal effect
Elemental mercury vapor	Artisanal gold mining, industrial use, spills, dental amalgam	Absorbed through the lungs; tremor, cognitive and behavioral effects
Inorganic mercury salts	Some skin-lightening products and traditional preparations	Corrosive to gut; kidney damage
Methylmercury	Fish and marine mammals	Crosses placenta and blood-brain barrier; developmental neurotoxicity

Methylmercury is the public health problem, and the pathway that creates it is worth following. Coal combustion and artisanal gold mining release mercury to the atmosphere, where it travels globally. It deposits into water bodies. In anoxic sediments, sulfate-reducing bacteria methylate it. Methylmercury then biomagnifies through the aquatic food web, so concentrations rise at every step, and a long-lived predatory fish such as swordfish or shark carries orders of magnitude more than the water around it. Artisanal and small-scale gold mining, where mercury is used to amalgamate gold from ore and then burned off, is the largest anthropogenic source of mercury emissions to air worldwide.

The catastrophe that named the disease occurred at Minamata Bay in Japan, where the Chisso Corporation discharged methylmercury-bearing effluent from an acetaldehyde plant for decades. The disease was officially recognized in 1956. Its cruellest feature was congenital: children were born with severe neurological damage to mothers who had few or no symptoms themselves, because the placenta concentrates methylmercury into the fetus. The international treaty adopted in 2013 to control mercury supply, use, and emissions is named the Minamata Convention.

Two cohorts, two answers

How much prenatal methylmercury matters at ordinary fish-eating exposures has been studied in two long-running birth cohorts that reached different conclusions, and the disagreement is genuinely informative.

The Faroe Islands cohort, followed by Philippe Grandjean and colleagues, studied a population whose methylmercury came substantially from pilot whale meat. It found associations between prenatal exposure and deficits in attention, language, and memory in childhood, persisting at follow-up.

The Seychelles Child Development Study, led by Gary Myers, Philip Davidson and colleagues, studied a population eating ocean fish at high frequency. It found no consistent adverse neurodevelopmental associations, and in some analyses positive associations with fish consumption.

Several explanations for the difference are on the table, and they are not mutually exclusive. The exposure vehicles differ: whale meat carries polychlorinated biphenyls alongside mercury, and fish carries long-chain omega-3 fatty acids that support brain development. The exposure patterns differ, with whale meals delivering large intermittent doses versus daily fish delivering steady lower ones. The outcome measures and ages differ. What both cohorts agree on is that the vehicle matters as much as the contaminant, which is why public advice now takes the form of naming which fish to eat rather than telling pregnant women to avoid fish. Advising avoidance would remove a documented benefit to reduce a documented harm, and the net effect could go either way.

What matters here: When two good cohorts disagree, the productive question is what differed between them, and here the answer was the whole diet rather than the single contaminant.

Cadmium

In the Jinzu River basin of Toyama Prefecture, Japan, water carrying cadmium from an upstream mine was used to irrigate rice paddies for decades. Beginning in the early twentieth century, local residents, overwhelmingly older women who had borne several children, developed a condition of such severe bone pain that it was named itai-itai, meaning it hurts. Their bones fractured under ordinary movement. In 1968 the Japanese government officially recognized it as caused by cadmium pollution.

Cadmium accumulates in the kidney, where its biological half-life is measured in decades, and it damages the renal tubules. The bone disease follows partly from impaired vitamin D metabolism and calcium handling and is worst in people whose calcium reserves are already depleted by pregnancy and poor nutrition, which explains who fell ill in Toyama.

Contemporary exposure is less dramatic and widespread. For smokers, tobacco is typically the dominant source, because the tobacco plant takes up cadmium efficiently and inhalation absorbs it well. For nonsmokers, food dominates: leafy vegetables, grains, shellfish, and organ meats. Occupational exposure occurs in battery manufacture, smelting, and pigment production. Because the half-life is so long, cadmium in the kidney reflects a lifetime of intake, which is the body burden concept from Module 1 in its purest form.

One principle these three share

All three are natural. Arsenic in a delta aquifer, mercury in coal, cadmium in an ore body: none was invented. What human activity does is mobilize them, concentrating them, moving them into water and air and food chains, and delivering them to people at rates that geology alone never would.

That should permanently settle the idea that natural means safe. It should also settle its mirror image, that synthetic means dangerous. The question for any agent is the same one from the first lesson: what dose, by what route, in whom, over how long.

Common misconceptions

• Natural contaminants are less dangerous than industrial ones. Arsenic, mercury, and cadmium are geological in origin and among the most damaging exposures in this course.
• All arsenic in food is equally hazardous. Organic arsenic compounds in seafood are largely excreted unchanged; inorganic arsenic is the toxic form regulators target.
• Pregnant women should avoid fish because of mercury. Current advice names which fish to eat, because removing fish removes documented developmental benefits along with the contaminant.
• Mercury in fish comes from local water pollution. Atmospheric emissions travel globally, deposit, are methylated by sediment bacteria, and biomagnify, so remote oceans are affected.
• A current blood or urine test captures cadmium exposure. Cadmium accumulates in the kidney with a half-life of decades, so the burden reflects a lifetime rather than recent intake.

What to remember

• Tube wells solved a pathogen problem in Bangladesh and created an arsenic exposure affecting tens of millions.
• Inorganic arsenic causes skin lesions and cancers of skin, bladder and lung after long latency; the US standard fell from 50 to 10 micrograms per liter.
• Mercury's three forms differ in route and effect, and methylmercury is the developmental neurotoxicant.
• Atmospheric mercury deposits, is methylated by sediment bacteria, and biomagnifies to high concentrations in predatory fish.
• The Faroe Islands and Seychelles cohorts disagreed because the whole diet differed, not because one study was wrong.
• Cadmium accumulates in the kidney over decades; tobacco dominates smokers' intake and food dominates everyone else's.

Sources

1. World Health Organization. (n.d.). Arsenic. who.int
2. U.S. Environmental Protection Agency. (n.d.). Mercury. epa.gov
3. Agency for Toxic Substances and Disease Registry. (n.d.). Toxicological profiles for arsenic, mercury, and cadmium. Centers for Disease Control and Prevention. atsdr.cdc.gov
4. Smith, A. H., Lingas, E. O., and Rahman, M. (2000). Contamination of drinking-water by arsenic in Bangladesh: A public health emergency. Bulletin of the World Health Organization, 78(9), 1093-1103.
5. Grandjean, P., and Landrigan, P. J. (2014). Neurobehavioural effects of developmental toxicity. The Lancet Neurology, 13(3), 330-338.

Sixty-nine thousand blood samples

In 2005, as part of a class action settlement with DuPont over releases from its Washington Works plant near Parkersburg, West Virginia, roughly 69,000 residents of six affected water districts gave blood and completed health questionnaires. It became one of the largest community exposure studies ever conducted. The settlement also created something unusual: an independent panel of three epidemiologists, agreed on by both sides in advance, empowered to determine whether the chemical at issue was linked to disease, with the company bound by the finding.

The chemical was perfluorooctanoic acid, known in the plant as C8. Between 2011 and 2012 the panel reported probable links to six conditions: kidney cancer, testicular cancer, ulcerative colitis, thyroid disease, pregnancy-induced hypertension, and elevated cholesterol.

The mechanism of the exposure was ordinary. The plant released the compound to air and water. It entered groundwater. People drank it. What was not ordinary was the compound's behavior once inside them: it did not break down, and it did not leave quickly.

Key idea: Persistence changes everything about an exposure, because a compound that neither degrades in the environment nor clears from the body converts a past release into a permanent one.

Why forever chemicals are forever

PFAS, per- and polyfluoroalkyl substances, are a family of thousands of compounds built around chains of carbon atoms with fluorine attached. The carbon-fluorine bond is among the strongest single bonds in organic chemistry, which is exactly why these compounds were useful: they resist heat, oil, water, and chemical attack. Non-stick cookware, stain-resistant textiles, grease-resistant food packaging, and the aqueous film-forming foams used to fight fuel fires all depend on that stability.

The same property means they do not meaningfully degrade in the environment. In humans, elimination half-lives for the legacy long-chain compounds are measured in years rather than hours or days, so blood concentrations reflect years of accumulated intake.

National biomonitoring detects PFAS in nearly all Americans tested. The pattern over time is informative: serum concentrations of the two most studied compounds, PFOA and PFOS, fell substantially after United States manufacturers phased them out in the early 2000s and under a later stewardship agreement. That decline is evidence both that the exposure was largely industrial in origin and that removing a source works, even for a compound that persists.

It also produced the field's standard cautionary pattern. Long-chain compounds were replaced by shorter-chain and structurally modified alternatives, marketed partly on the basis that they clear from the body faster. Some of those replacements are now themselves regulated, having turned out to be mobile in groundwater and detectable downstream. Chemists call this regrettable substitution: replacing a known problem with a less studied member of the same family.

In April 2024 EPA set the first national drinking water standards for PFAS, with enforceable limits of 4 parts per trillion for PFOA and PFOS, 10 parts per trillion for three other named compounds, and a hazard index approach for mixtures. The health-based goals for PFOA and PFOS were set at zero. Parts per trillion is worth pausing on: this is a limit roughly a thousand times lower than typical limits for older regulated contaminants, and meeting it requires granular activated carbon, ion exchange, or reverse osmosis.

One honest qualification. For most people, drinking water is not the largest PFAS exposure route; diet and consumer products contribute substantially. For people living near a fluorochemical plant, a military airfield where firefighting foam was used, or a landfill receiving PFAS-bearing waste, drinking water dominates completely. Population averages and community realities diverge here more than almost anywhere else in this course.

Pesticides: designed to be toxic

Pesticides are the only chemicals in this course intended to kill something. That makes the central question narrower and sharper: how selective is the toxicity, and how much reaches people?

Organochlorines came first at scale, and DDT is the emblem. It was extraordinarily effective against insect vectors and agricultural pests, cheap, and persistent. That persistence, plus its fat solubility, produced biomagnification through food chains. In birds of prey, the metabolite DDE interfered with calcium deposition in eggshells, so shells thinned and broke under the weight of incubating parents, and populations of bald eagles, peregrine falcons, and brown pelicans collapsed. Rachel Carson's 1962 book made the mechanism public, and the United States banned agricultural use in 1972.

The honest complication is that DDT is still used, legitimately, for indoor residual spraying against malaria vectors under an exemption in the international treaty governing persistent organic pollutants. Atlas's Global Health course covers that tradeoff. It is a clean example of a substance whose risk-benefit balance genuinely differs between settings.

Organophosphates largely replaced organochlorines because they break down faster in the environment. Their mechanism is precise: they inhibit acetylcholinesterase, the enzyme that clears acetylcholine from synapses. Acetylcholine accumulates, and the result is continuous stimulation of cholinergic receptors, producing salivation, tearing, urination, diarrhea, vomiting, bronchial secretions, slowed heart rate, muscle twitching, and, at high dose, seizures and respiratory failure. Treatment is atropine to block the receptors and pralidoxime to reactivate the enzyme if given early enough.

The regulatory arguments concern chronic low-level exposure rather than acute poisoning. Three birth cohorts in New York City and California's Salinas Valley followed prenatal organophosphate exposure, measured through maternal urinary metabolites or cord blood, and reported associations with lower cognitive scores and altered neurodevelopment in childhood. EPA cancelled residential uses of chlorpyrifos in 2000 and moved in 2021 to revoke its food tolerances entirely, a decision a federal appeals court set aside in 2023, leaving the compound's status contested.

Farmworkers carry the highest exposures by far, through application, drift, and residue on foliage, and they are the population least protected by the standard assumption that dietary residues are the exposure of concern.

Glyphosate: the same evidence, two answers

In 2015 the International Agency for Research on Cancer classified glyphosate as probably carcinogenic to humans. EPA and the European Food Safety Authority have concluded that it is unlikely to pose a carcinogenic risk at expected exposures. Both statements are defensible, and the reason they can coexist is the distinction from Module 1.

The cancer agency performs hazard identification: can this agent cause cancer under any circumstance? It weighs mechanistic and animal evidence heavily and considers only publicly available data. The regulatory agencies perform risk assessment: does it cause cancer at the doses people actually receive? They consider exposure and include unpublished registrant studies.

So the disagreement is partly definitional and partly real. It is definitional in that the two bodies are answering different questions. It is real in that they weighted the epidemiology differently, and reasonable scientists have argued about whether the registrant studies should carry the weight regulators gave them. What is not defensible is quoting one conclusion while omitting that the other exists, which both advocacy sides routinely do.

The point: A hazard classification and a risk assessment can reach opposite-sounding conclusions about the same chemical without either being wrong, because they are not answering the same question.

Endocrine disruption, and a dispute that is not resolved

Hormones work at concentrations far below those of ordinary toxicants, in parts per trillion, and they work through receptors whose response is tuned rather than proportional. An endocrine disrupting chemical is one that interferes with that system, by mimicking a hormone, blocking a receptor, or altering hormone synthesis, transport, metabolism, or elimination.

The proof of concept is pharmaceutical and unambiguous. Diethylstilbestrol, a synthetic estrogen, was prescribed to pregnant women from around 1940 into the early 1970s to prevent miscarriage, which it did not do. In 1971 physicians reported a rare vaginal cancer, clear cell adenocarcinoma, in young women whose mothers had taken it during pregnancy, alongside structural reproductive abnormalities. The exposure was prenatal, the outcome appeared two decades later, and the mothers were unaffected. Every feature that makes endocrine disruption hard to study was present at once.

The scientific dispute concerns whether environmental exposures to compounds such as bisphenol A, certain phthalates, and some flame retardants produce comparable effects at ordinary doses. Two positions, stated at their strongest.

Many academic endocrinologists argue that standard regulatory toxicology is structurally unable to detect these effects. Guideline studies test high doses and extrapolate downward, on the assumption that the dose-response curve is monotonic. Hormone systems do not necessarily behave that way: receptor saturation and feedback can produce curves that are U-shaped or inverted-U, so an effect at a low dose can vanish at a high one. Guideline studies also use endpoints such as organ weight and gross pathology, which can miss changes in mammary gland development, prostate morphology, or behavior, and they often dose at the wrong life stage. On this view, hundreds of published low-dose findings are being discounted for failing to conform to a testing paradigm built for a different kind of chemical.

Many regulatory toxicologists respond that guideline studies exist because reproducibility is the point. Low-dose findings have often failed to replicate; some come from studies with small groups, uncontrolled dietary phytoestrogens, or caging materials that themselves leach test compounds; and a non-monotonic curve reported without a mechanism is difficult to distinguish from noise across many measured endpoints. On this view, adopting every positive academic finding would make regulation unstable and unfalsifiable.

A joint program called CLARITY-BPA was designed to settle part of this, running a guideline-compliant core study and academic sub-studies on animals from the same source and dosing stream. Published in 2018, the core study found few consistent low-dose effects, while several academic arms reported effects on specific tissues. Both camps read the result as supporting their position, which is itself a finding about how underdetermined the evidence is.

The clearest illustration of where this leaves regulation came in 2023, when the European Food Safety Authority reduced its tolerable daily intake for bisphenol A by roughly four orders of magnitude, based principally on immune system effects. The United States Food and Drug Administration reviewed the same literature and stated that its position, that bisphenol A is safe at current exposure levels, had not changed. Two competent agencies, overlapping evidence, opposite conclusions. If you want a single image of how much interpretation sits between data and a regulatory number, that is it.

Common misconceptions

• PFAS exposure comes mainly from drinking water. For most people diet and consumer products dominate; for communities near plants, airfields, or landfills, water dominates completely.
• Replacing a banned chemical with a new one solves the problem. Regrettable substitution is common, and several PFAS replacements are now themselves regulated.
• IARC and EPA contradict each other about glyphosate because one is wrong. One performs hazard identification and the other risk assessment, and they also weighted the epidemiology differently.
• Testing a chemical at high doses is always protective. That holds only for monotonic dose-response curves, which is precisely what the endocrine dispute contests.
• The endocrine disruption question has been settled. Two major regulatory bodies reached opposite conclusions about bisphenol A in 2023 on overlapping evidence.

Summing up

• The C8 settlement produced an independent panel that found probable links between PFOA and six conditions among 69,000 exposed residents.
• The carbon-fluorine bond makes PFAS persistent in the environment and slow to clear from people, so past releases remain present exposures.
• EPA set drinking water limits of 4 parts per trillion for PFOA and PFOS in 2024, with health-based goals of zero.
• DDT's eggshell effect and organophosphates' acetylcholinesterase inhibition are the two clearest pesticide mechanisms, and farmworkers carry the highest exposures.
• Hazard classification and risk assessment answer different questions, which is why glyphosate conclusions differ without either being fabricated.
• The low-dose endocrine dispute turns on whether standard testing can detect non-monotonic effects, and it is unresolved.

Sources

1. U.S. Environmental Protection Agency. (n.d.). PFAS. epa.gov
2. U.S. Environmental Protection Agency. (n.d.). Pesticides. epa.gov
3. National Institute of Environmental Health Sciences. (n.d.). Endocrine disruptors. National Institutes of Health. niehs.nih.gov
4. Carson, R. (1962). Silent Spring. Houghton Mifflin.
5. Vandenberg, L. N., Colborn, T., Hayes, T. B., Heindel, J. J., Jacobs, D. R., Lee, D. H., et al. (2012). Hormones and endocrine-disrupting chemicals: Low-dose effects and nonmonotonic dose responses. Endocrine Reviews, 33(3), 378-455.

Four children

In January 1993, physicians in Seattle began seeing children with bloody diarrhea and kidney failure. The organism was Escherichia coli O157:H7, and the source was undercooked hamburger patties served at Jack in the Box restaurants across four western states. More than seven hundred people were sickened. Four children died.

The failure was a temperature. Washington state had recently raised its required internal cooking temperature for ground beef, and the chain's grills were still set for the old standard. A difference of a few degrees, applied to patties containing an organism with an infectious dose of well under a hundred cells, killed four children.

The consequences were structural. In 1994 the Department of Agriculture declared O157:H7 an adulterant in raw ground beef, meaning its presence alone made the product illegal to sell, which had never before applied to a bacterium in raw meat. In 1996 the department required meat and poultry plants to operate under a systematic preventive framework rather than relying on inspectors examining carcasses by sight and touch.

Key idea: Food safety regulation in the United States is largely a record of outbreaks, each one converting a specific failure into a general rule.

The scale of the problem

The Centers for Disease Control and Prevention estimate that foodborne illness causes about 48 million illnesses, 128,000 hospitalizations, and 3,000 deaths in the United States each year. That is roughly one American in six getting sick from food annually, most of them never diagnosed, never reported, and never linked to a source.

The ranking depends on which outcome you count, and the differences are instructive.

Pathogen	Notable for	Typical vehicles
Norovirus	The most illnesses by a wide margin	Infected food handlers, shellfish, ready-to-eat foods
Nontyphoidal Salmonella	The most hospitalizations, and among the most deaths	Poultry, eggs, produce, reptiles and their environments
Campylobacter	Very common; associated with Guillain-Barre syndrome	Raw and undercooked poultry, unpasteurized milk
Listeria monocytogenes	A very high case fatality rate	Deli meats, soft cheeses, melons, refrigerated ready-to-eat foods
E. coli O157:H7	Hemolytic uremic syndrome, especially in children	Ground beef, leafy greens, unpasteurized cider
Clostridium perfringens	Large institutional outbreaks	Meats and gravies held at improper temperatures

Two of these deserve extra attention because they violate intuitions.

Listeria grows at refrigeration temperature. Cold slows most foodborne bacteria to a halt; this one keeps multiplying in the deli case. It also crosses the placenta, and pregnancy raises susceptibility substantially, which is why the advice to pregnant women about soft cheeses and deli meat is not fussiness. Its case fatality among diagnosed invasive cases is on the order of one in five, higher than any other common foodborne pathogen.

E. coli O157:H7 produces Shiga toxin, which damages the endothelium of small blood vessels, and in a minority of cases, concentrated among young children and older adults, produces hemolytic uremic syndrome: destruction of red cells, low platelets, and acute kidney failure. The infectious dose is remarkably low, which is why cross-contamination in a kitchen matters as much as cooking temperature, and why antibiotics are generally avoided in suspected cases because of concern that killing the organisms releases more toxin.

From inspection to prevention

For most of the twentieth century, American food safety relied on inspection: a person looking at product. That approach detects gross problems and is nearly useless against an invisible pathogen distributed unevenly through a lot.

HACCP, hazard analysis and critical control points, replaced it with a systems approach originally developed to keep astronauts from getting sick in orbit. The logic is to identify where in a process a hazard can be controlled, set a measurable limit at that point, monitor it continuously, and document everything. Its seven principles are: analyze hazards; identify critical control points; establish critical limits; establish monitoring; establish corrective actions; establish verification; and establish record-keeping.

A cooking step with a required internal temperature is a critical control point. So is a pasteurization hold time, a metal detector, or a pH adjustment. The strength of the approach is that it moves control from after-the-fact detection to during-the-process assurance, and it produces records that let a traceback investigation reconstruct what happened.

The Food Safety Modernization Act of 2011 extended the same philosophy to the roughly four fifths of the American food supply that the Food and Drug Administration regulates, requiring preventive controls in facilities, setting the first federal safety standards for produce growing and harvesting, and requiring importers to verify their foreign suppliers. The remaining fifth, meat, poultry, and processed egg products, is regulated by the Department of Agriculture, a split that dates from the early twentieth century and produces the peculiar result that a cheese pizza and a pepperoni pizza fall under different agencies.

Why this matters: Prevention systems work because they control a process continuously; inspection works only if the defect is visible, and pathogens are not.

How a multistate outbreak gets solved

Modern outbreak detection is a molecular problem. Twenty people getting sick in twenty states over three weeks looks like nothing at all in local surveillance data. What makes them visible is subtyping.

The national laboratory network established in 1996 originally fingerprinted bacterial isolates by cutting their DNA with enzymes and comparing the resulting band patterns. Since then it has moved to whole genome sequencing, which reads the entire genome of each isolate. Two patients whose isolates differ by a handful of nucleotides almost certainly acquired the organism from a common source, even if they live two thousand miles apart and ate a month apart.

Once a cluster is identified, investigators interview cases about what they ate, compare against expected consumption patterns, and generate a hypothesis. Then traceback works backward through distribution records to find a common lot, farm, or processor. Sequencing has made this dramatically more powerful: clusters that would have been invisible are now detected, which is one reason reported outbreak counts have risen even as food safety has improved. Better detection produces more outbreaks on paper.

What a person can actually do

Four practices carry nearly all the household benefit, and each targets a specific mechanism.

• Clean. Handwashing and surface cleaning interrupt transmission from raw ingredients and from infected handlers, which is how most norovirus reaches food.
• Separate. Cross-contamination from raw meat to ready-to-eat food defeats cooking entirely, because the salad is never heated. Separate boards and utensils address a low infectious dose directly.
• Cook. Internal temperature, verified with a thermometer rather than by color, is the actual control. Ground beef requires 160 degrees Fahrenheit, poultry 165, and whole cuts of beef and pork 145 with a rest period. Ground meat needs a higher temperature than a steak because grinding distributes surface organisms throughout.
• Chill. Between about 40 and 140 degrees Fahrenheit, bacterial growth is rapid. Prompt refrigeration and shallow containers for cooling large volumes address the mechanism behind most institutional outbreaks.

The chemical side of food

Pathogens dominate acute risk, and food also carries chemical hazards worth knowing.

Mycotoxins are produced by molds growing on crops. Aflatoxin, produced by Aspergillus species on maize and groundnuts stored warm and damp, is among the most potent liver carcinogens known, and its effect multiplies with chronic hepatitis B infection, which is why it matters most where both are common. Atlas's Global Health course covers that interaction.

Heavy metals reach food from soil and water, as covered in the previous lesson. Testing of commercial baby foods has repeatedly found detectable arsenic, lead, and cadmium, prompting federal attention to action levels for products aimed at infants.

Process contaminants form during cooking. Acrylamide forms when starchy foods are fried, roasted, or baked at high temperature, and it is classified as a probable human carcinogen on the basis of animal data, with human epidemiology inconsistent.

Natural toxins include ciguatera from reef fish, scombroid poisoning from histamine in improperly chilled fish, and shellfish toxins from harmful algal blooms, which are becoming more frequent as coastal waters warm.

Common misconceptions

• Refrigeration stops bacterial growth. Listeria grows at refrigeration temperature, which is exactly why it dominates ready-to-eat food risk.
• Brown in the middle means a burger is done. Color is an unreliable indicator; only a thermometer verifies the internal temperature that actually kills the organism.
• More reported outbreaks means food is getting less safe. Whole genome sequencing detects clusters that were previously invisible, so better surveillance raises the count.
• One agency regulates food. The Department of Agriculture handles meat, poultry, and processed egg products; the Food and Drug Administration handles roughly the rest.
• Antibiotics are the right treatment for E. coli O157:H7. They are generally avoided because of concern that killing organisms increases Shiga toxin release and hemolytic uremic syndrome risk.

The takeaway

• The 1993 Jack in the Box outbreak converted a cooking temperature failure into an adulterant declaration and a national preventive framework.
• An estimated 48 million illnesses, 128,000 hospitalizations, and 3,000 deaths occur annually in the United States from foodborne agents.
• Norovirus causes the most illness, Salmonella the most hospitalizations, and Listeria has the highest case fatality.
• HACCP controls hazards during a process rather than detecting them afterward, and the 2011 modernization act extended prevention to FDA-regulated foods.
• Whole genome sequencing links geographically scattered cases into solvable outbreaks and raises reported outbreak counts.
• Clean, separate, cook, and chill each target a distinct transmission mechanism, and chemical hazards including mycotoxins and metals sit alongside pathogens.

Sources

1. U.S. Department of Health and Human Services. (n.d.). FoodSafety.gov. foodsafety.gov
2. U.S. Food and Drug Administration. (n.d.). Food. fda.gov
3. U.S. Department of Agriculture Food Safety and Inspection Service. (n.d.). Food safety. fsis.usda.gov
4. Scallan, E., Hoekstra, R. M., Angulo, F. J., Tauxe, R. V., Widdowson, M. A., Roy, S. L., Jones, J. L., and Griffin, P. M. (2011). Foodborne illness acquired in the United States: Major pathogens. Emerging Infectious Diseases, 17(1), 7-15.

Two neighborhoods across a street from each other

During the week of 13 July 1995, Chicago recorded 739 excess deaths. Air temperatures reached 106 degrees Fahrenheit, and the heat index went higher. The dead were overwhelmingly elderly, overwhelmingly poor, and disproportionately people living alone. Refrigerated trucks were brought in because the medical examiner's office ran out of room.

The sociologist Eric Klinenberg later examined the mortality data neighborhood by neighborhood and found something the weather could not explain. North Lawndale and South Lawndale sit beside each other on Chicago's west side. Both were poor. Both were predominantly nonwhite. Their death rates during the heat wave differed by a factor of several.

What differed was the physical and social fabric. South Lawndale, also called Little Village, had busy commercial streets, dense occupied housing, and shops elderly residents walked to daily, which meant people saw each other and noticed absences. North Lawndale had lost most of its commerce and much of its population, leaving vacant lots, abandoned buildings, and streets that older residents were afraid to walk. Isolated people in hot apartments died with their windows closed.

Key idea: The built environment is not scenery around a health problem; the arrangement of buildings, streets, and shops determined who survived a week of hot weather.

Inside the house

Americans spend roughly ninety percent of their time indoors, and most of that at home, which makes housing the single most concentrated exposure setting most people have. Federal healthy homes guidance organizes the hazards into principles that are easy to remember and each map onto a mechanism: keep a home dry, clean, pest-free, ventilated, safe, contaminant-free, maintained, and thermally controlled.

Hazard	Mechanism	Principal outcome
Dampness and mold	Fungal growth and microbial products in indoor air	Asthma exacerbation, cough, wheeze, upper respiratory symptoms
Cockroach and mouse allergen	Sensitization plus continuing exposure in bedding and dust	Asthma morbidity, especially in children
Lead paint dust	Deteriorating or disturbed pre-1978 paint, hand-to-mouth transfer	Neurodevelopmental damage
Radon	Soil gas entering through foundation cracks and sumps	Lung cancer
Carbon monoxide	Incomplete combustion from heaters, stoves, and generators	Acute poisoning and death
Fall and injury hazards	Stairs, lighting, tub surfaces, unguarded windows	Fractures and head injury, especially in older adults and toddlers
Extreme indoor temperature	Absent cooling or heating, poor insulation, energy insecurity	Heat and cold mortality

The asthma pathway deserves the closest look, because it is where housing intervention has the best evidence behind it. A landmark study of children with asthma in nine American cities found that the combination of being sensitized to cockroach allergen and living with high cockroach allergen exposure predicted hospitalization and symptom days better than sensitization or exposure alone. That is an important structure: neither the biological susceptibility nor the environmental exposure alone did the damage; the interaction did.

A follow-on trial then tested whether changing the environment changed the outcome. Families received an individualized intervention, addressing whichever allergens the child was actually sensitized to, using pest management, mattress and pillow covers, and high-efficiency vacuuming. Children in the intervention group had significantly fewer symptom days, and the benefit persisted after the intervention period. That is unusually strong evidence for an environmental health intervention, because it was randomized.

Remember: Housing interventions for asthma work best when they are individualized to the allergens a specific child is sensitized to, rather than applied as a generic cleaning package.

Outside the front door

Neighborhood design shapes health through several distinct channels.

Physical activity. Walkable environments, with mixed land use, connected street networks, and destinations within walking distance, are associated with more walking and lower body mass index. The honest qualification is that people who like to walk move to walkable places, so cross-sectional associations overstate the causal effect. The stronger evidence comes from studies following people who move and from before-and-after studies of new transit and trail infrastructure.

Injury. Traffic kills tens of thousands of Americans annually, and pedestrian deaths have risen sharply since around 2009 even as overall vehicle occupant safety improved. Three design factors dominate. Speed: the probability that a struck pedestrian dies rises steeply with impact speed, so a person hit at 20 miles per hour usually survives and one hit at 40 usually does not. Vehicle geometry: taller, blunter front ends strike an adult's torso rather than legs and are more likely to push a pedestrian under the vehicle. And road design: wide lanes, long blocks, and infrequent crossings both increase speeds and lengthen exposure. The Vision Zero approach treats these as engineering problems with engineering solutions rather than as individual behavior failures.

Noise. Chronic environmental noise from roads, rail, and aircraft is associated with sleep disturbance, hypertension, and ischemic heart disease, through pathways involving sleep fragmentation and stress hormone response. Studies of children near airports have found associations between aircraft noise exposure and reading comprehension. Noise is among the most underregulated common environmental exposures in the United States.

Green space. Access to parks and vegetation is associated with better mental health, and here too the design question is whether the association is causal. One of the strongest tests came from Philadelphia, where researchers randomly assigned vacant lots to be cleaned and greened, cleaned only, or left alone. Residents living near greened lots reported significantly lower rates of feeling depressed, and a companion analysis found reductions in nearby gun violence. Randomizing the intervention is what makes this evidence rather than correlation.

Heat. Cities are hotter than surrounding areas because dark impervious surfaces absorb solar radiation, buildings trap it, vegetation that would cool by evapotranspiration is absent, and vehicles and air conditioners add waste heat. The effect is not uniform within a city. Research mapping present-day surface temperatures against the 1930s federal mortgage security maps found that formerly redlined neighborhoods are consistently hotter today, by several degrees Celsius in many cities, because they have less tree canopy and more paved surface.

The problem that makes this research hard

People choose where they live, and the choice is shaped by income, employment, family, discrimination, and preference. So a comparison of health between neighborhoods confounds the effect of place with the characteristics of the people who ended up there. This is the central methodological problem in the entire literature, and Atlas's Epidemiology course treats the general form of it.

The most informative response has been an actual experiment. Beginning in 1994, the Moving to Opportunity demonstration randomly assigned about 4,600 families in public housing in five cities to one of three groups: a voucher usable only in a low-poverty neighborhood with counseling, an unrestricted voucher, or no voucher. Because assignment was random, differences that emerged could be attributed to the move.

The results were mixed in an instructive way. Adults who moved showed improvements in mental health, and reductions in obesity and diabetes, but no gains in employment or earnings. Effects on children initially looked disappointing. Then a later analysis using tax records found that the age at which a child moved mattered enormously: children who moved to lower-poverty neighborhoods before about age thirteen had substantially higher earnings as adults, while those who moved as adolescents did not benefit and in some measures did worse.

That is a more useful finding than a simple positive or negative result. It says that place matters, that it matters cumulatively through childhood, and that short follow-up periods can make a real effect look absent.

What actually gets fixed

Interventions at this scale are unglamorous and specific: housing code enforcement with real inspection capacity, weatherization programs that improve insulation and reduce both energy cost and temperature extremes, integrated pest management instead of pesticide spraying, lead hazard control grants for pre-1978 housing, radon-resistant construction requirements in new building codes, street redesign that lowers speeds physically rather than by posting a sign, tree planting targeted at the hottest and least canopied blocks, and zoning that allows the mixed uses that make a neighborhood walkable.

None of these looks like health care. All of them have measurable health effects, and several have been tested in randomized or quasi-experimental designs, which is more than can be said for a good deal of what does look like health care.

Common misconceptions

• Heat deaths are caused by temperature alone. Chicago's 1995 mortality varied several-fold between adjacent equally poor neighborhoods, tracking social and physical infrastructure.
• Mold is dangerous because of toxins in the air. The well-supported effects are respiratory and allergic, driven by dampness and microbial growth; claims about widespread toxic mold illness go well beyond the evidence.
• Walkable neighborhoods cause weight loss. Much of the association reflects who chooses to live there; the credible evidence comes from movers and from infrastructure natural experiments.
• Pedestrian deaths reflect careless walking. Impact speed, vehicle front-end geometry, and road design explain most of the variation, which is why engineering approaches outperform education campaigns.
• Neighborhood effects cannot be studied rigorously. Moving to Opportunity randomized housing vouchers, and the greening trials randomized lots, producing genuine causal evidence.

What you now know

• Chicago in 1995 showed that neighborhood fabric, not temperature alone, determined who died in a heat wave.
• Housing hazards run from dampness and pest allergens to lead dust, radon, carbon monoxide, falls, and temperature extremes.
• Randomized asthma trials show individualized allergen reduction lowers symptom days, with benefits persisting after the intervention.
• Neighborhood design affects health through activity, traffic injury, noise, green space, and urban heat.
• Residential self-selection confounds neighborhood research, which is why Moving to Opportunity and randomized greening trials matter.
• The effective interventions are code enforcement, weatherization, pest management, street redesign, and canopy, not clinical care.

Sources

1. U.S. Environmental Protection Agency. (n.d.). Indoor air quality. epa.gov
2. U.S. Department of Housing and Urban Development. (n.d.). Healthy homes and lead hazard control. hud.gov
3. National Highway Traffic Safety Administration. (n.d.). Pedestrian safety. U.S. Department of Transportation. nhtsa.gov
4. Klinenberg, E. (2002). Heat Wave: A Social Autopsy of Disaster in Chicago. University of Chicago Press.
5. Chetty, R., Hendren, N., and Katz, L. F. (2016). The effects of exposure to better neighborhoods on children: New evidence from the Moving to Opportunity experiment. American Economic Review, 106(4), 855-902.

The stone that was supposed to be a better countertop

Engineered stone is made by crushing quartz, binding the grains with polymer resin, and pressing the mixture into slabs. It is harder than granite and cheaper than marble, and by weight it can be more than ninety percent crystalline silica, against roughly a quarter to a half for granite. Cutting it dry with a handheld grinder, in a shop with a box fan in the window, fills the air with respirable silica particles small enough to reach the deep lung.

In 2023 the California Department of Public Health and the Centers for Disease Control and Prevention described what that had produced in one state: a cluster of more than fifty silicosis cases among countertop fabrication workers, all of them men, nearly all Latino immigrants, with a median age in the forties. Ten had died. Several were diagnosed only after being treated for tuberculosis or sarcoidosis first, because the textbook picture of silicosis is an old miner, not a fabricator in his thirties.

Nothing about the hazard was new. Agricola described the lung disease of miners in 1556. Between 1930 and 1931 workers drilling the Hawks Nest Tunnel at Gauley Bridge, West Virginia, cut through rock so rich in silica that the company began mining it, and several hundred of them, mostly Black migrant laborers, died of acute silicosis within a few years. The control measures are not technically difficult either: cut the stone wet, capture the dust at the blade, ventilate the room. What was new was that the exposure had migrated into hundreds of small shops that no inspector had ever visited, working a material whose silica content nobody had thought to ask about.

Bottom line: Occupational disease rarely arrives because a hazard was unknown. It arrives because a known hazard moved into a workplace where nobody was looking.

Why the workplace is the sharpest exposure setting

Return to the dose arithmetic from Module 1 and change three inputs. Concentration in an uncontrolled workplace can run hundreds or thousands of times ambient levels. Duration is eight hours a day, five days a week, for a working lifetime of forty years. Activity level is elevated, so breathing rate is higher. The result is that a worker in a bad job can receive more of a given agent in a month than the surrounding community receives in a lifetime.

That is why the workplace is where we learned most of what we know. Percivall Pott connected scrotal cancer to soot in chimney sweeps in 1775. Ludwig Rehn reported bladder cancer in aniline dye workers in 1895. Christopher Wagner and colleagues linked mesothelioma to crocidolite asbestos in the Cape Province of South Africa in 1960. In 1974 a physician at a B. F. Goodrich plant in Louisville reported three cases of angiosarcoma of the liver, a tumor so rare that three in one workforce was itself the finding, and vinyl chloride was identified as a human carcinogen within months. The great majority of agents recognized as human carcinogens were identified first in occupational cohorts, because those cohorts have high doses, defined membership, and payroll records that let you count person-years.

One complication runs through all of it. Employed people are healthier than the general population: they were well enough to be hired and well enough to stay. Comparing the mortality of a group of workers to the mortality of the whole country therefore makes the job look protective. This is the healthy worker effect, and it biases occupational studies toward finding nothing. When a study comparing workers to the general population still finds excess disease, the true excess is usually larger than the number reported.

Exposure does not always stop at the plant gate. Asbestos fibers carried home on work clothes have caused mesothelioma in wives who laundered them. Lead dust rides home on boots, in cars, and on hands, and children of workers in battery plants and firing ranges have shown elevated blood lead from that route alone. Take-home exposure is the reason changing rooms, on-site laundering, and separate street clothes are not fussiness.

The core of it: Work delivers the largest routine chemical doses in society to a population you can name, follow, and count, which is exactly why occupational epidemiology has done so much of the field's discovery.

The hierarchy of controls, read from the top down

Ask a room of people how to protect a worker from dust and most will say masks. That answer is last on the list for a reason, and the list itself is the single most useful thing in occupational health practice.

Level	What it does	Silica in a countertop shop	Why it ranks there
Elimination	Removes the hazard entirely	Do not use a high-silica material	Nothing left to fail
Substitution	Replaces it with something less harmful	Specify a low-silica or resin-free slab	Protection does not depend on behavior
Engineering controls	Separates people from the hazard	Wet cutting, blade-mounted water feed, local exhaust ventilation, enclosed cutting booths	Works whether or not anyone remembers; needs maintenance, not vigilance
Administrative controls	Changes how people work	Task rotation, wet housekeeping instead of dry sweeping, exposure monitoring, training	Depends on supervision and continuous compliance
Personal protective equipment	Protects at the individual	Fit-tested respirator with correct cartridges	Fails silently, one person at a time

The ranking is not a preference for expensive solutions. It is a statement about failure modes. A ventilation hood that stops working is noticeable and affects everyone at once, so it gets fixed. A respirator that leaks affects one person, produces no symptom, and is invisible to everyone including the wearer. A tight-fitting respirator requires medical clearance, an annual fit test, a clean shave at the sealing surface, correct cartridge change schedules, and consistent wear during every minute of exposure, including the five minutes of cleanup when people take them off. Each of those is a place to fail, and the protection factor printed on the box assumes none of them do.

Notice also which levels survive a change of staff. Elimination and engineering keep working after the trained supervisor quits. Administrative controls and protective equipment do not.

Why this matters: When you evaluate any workplace, the diagnostic question is not whether people are wearing protection. It is how far up the hierarchy the employer went before falling back on it.

A permissible exposure limit is not a safe level

The Occupational Safety and Health Act was signed in December 1970 and took effect in April 1971, creating two bodies with deliberately different jobs. The Occupational Safety and Health Administration, in the Department of Labor, writes and enforces standards. The National Institute for Occupational Safety and Health, in what became the CDC, does research and recommends. NIOSH cannot make a rule and OSHA is not obliged to adopt its recommendations, which explains a great deal of what follows.

To have something enforceable on day one, OSHA adopted several hundred existing consensus limits wholesale, most of them the 1968 threshold limit values of a private organization, the American Conference of Governmental Industrial Hygienists. Those became permissible exposure limits with the force of law. In 1989 OSHA tried to update more than four hundred of them at once. In 1992 the Eleventh Circuit vacated the update, holding that the agency had not made the required showings for each substance individually. The old numbers snapped back. The result is that most permissible exposure limits enforced today were set from a 1968 list, older than the majority of the workers they cover.

Number	Who sets it	Legal force	Example
Permissible exposure limit (PEL)	OSHA	Enforceable; citations and penalties	Noise at 90 A-weighted decibels as an 8-hour average
Recommended exposure limit (REL)	NIOSH	None; advisory	Noise at 85 A-weighted decibels as an 8-hour average
Threshold limit value (TLV)	ACGIH, a private body	None, unless adopted by a state or referenced in a contract	Updated annually by committee

Two further features of these numbers matter in practice. First, nearly all are time-weighted averages over eight hours, so a worker can spend twenty minutes in an enormous concentration and remain in compliance if the rest of the shift is clean; that is why some substances also carry short-term limits or ceilings that may never be exceeded. Second, a permissible limit is not the level at which harm begins. It is the level the agency concluded it could defend as both risk-reducing and economically and technologically feasible, which is a different question and the subject of the course's final lesson.

The silica rule shows the machinery working slowly but working. OSHA proposed a new standard in 2013 and finalized it in 2016, cutting the general industry limit to 50 micrograms per cubic meter as an eight-hour average, setting an action level of 25 that triggers monitoring, and giving construction employers a table of specified control methods they can follow instead of measuring. It took roughly four decades from the point at which the evidence was uncontested.

Four hazards, four different failures

Coal mine dust. Black lung was supposed to be finished. Then, beginning around 2014, NIOSH surveillance and Appalachian clinics began finding progressive massive fibrosis, the advanced destructive form, in miners too young to have worked under the old rules; clinicians in southwestern Virginia identified hundreds of cases in a few years, the largest cluster ever reported. The leading explanation is that the thick seams are gone, so machines now cut through more sandstone to reach thin coal, and the dust the miners breathe contains far more silica than the dust the standard was written for. The Mine Safety and Health Administration lowered the respirable dust limit to 1.5 milligrams per cubic meter in a 2014 rule.

Asbestos. Mesothelioma appears twenty to fifty years after exposure, which means today's cases record decisions made in the 1970s and 1980s and today's decisions will not be visible until the 2050s. Latency of that length defeats ordinary feedback: nobody in the plant sees the consequence of the year they are living through. The United States never enacted a general ban; EPA's 1989 rule was largely overturned in 1991, and only in 2024 did the agency finalize a prohibition on the remaining ongoing uses of chrysotile asbestos.

Noise. Noise-induced hearing loss is the most common occupational illness in the country and it is permanent, painless, and untreatable. It also shows the gap between the two kinds of numbers with unusual clarity. OSHA permits 90 decibels averaged over eight hours and uses a 5-decibel exchange rate, meaning allowed exposure time halves for each 5-decibel increase. NIOSH recommends 85 decibels and a 3-decibel exchange rate, which is what the physics of sound energy actually implies. The difference between those two rules is a large amount of hearing across a working population.

Heat. Exertional heat stroke kills agricultural, construction, and warehouse workers every summer, and the chronic story may be larger: an epidemic of chronic kidney disease of nonspecific origin among young sugarcane cutters in Central America, in men with none of the usual risk factors, is best explained by repeated heat strain and dehydration across seasons. California adopted an outdoor heat illness standard in 2005 and several states followed; there is still no federal heat standard, though OSHA published a proposed rule in 2024. Until one is final, federal heat enforcement runs through the General Duty Clause, section 5(a)(1) of the Act, which requires employers to furnish a workplace free from recognized hazards likely to cause death or serious harm. It is a real tool and a weak one, because the agency must prove the hazard was recognized and that a feasible abatement existed.

Who gets counted

The Bureau of Labor Statistics reports roughly five thousand fatal traumatic work injuries a year in the United States. That number is reasonably solid: a death on a job site is hard to miss. Occupational disease is a different matter, and the undercount is structural rather than accidental.

A lung cancer diagnosed at 67 carries no label naming the foundry where the person worked at 27. Death certificates record disease, not cause of exposure. Employer injury logs capture cut fingers well and capture asthma and hearing loss poorly. Workers with insecure status do not report, and some incentive schemes have effectively paid crews not to. The workers' compensation system, the no-fault bargain that began with state laws in the 1910s, was designed for acute injury: in exchange for guaranteed payment without proving fault, employees generally gave up the right to sue their employer. Applied to a disease with a thirty-year latency, multiple employers, and a smoking history to argue about, that system denies far more than it pays.

Protection is also distributed unevenly. Temporary and staffing-agency workers are frequently assigned the most hazardous tasks and are covered by a divided responsibility that lets each employer point at the other. Small shops, like the countertop fabricators this lesson opened with, fall below the radar of an inspectorate numbering a few thousand federal and state officers for millions of workplaces; the realistic probability that a given small business is inspected in any year is very close to zero. Immigrant workers, day laborers, and independent contractors sit further outside the system still.

One tool is worth knowing by name. Any group of employees can request a NIOSH Health Hazard Evaluation, in which federal investigators come to the workplace, measure exposures, and publish findings. It is free, it can be requested confidentially, and it does not require an employer to agree.

Common misconceptions

• If a workplace complies with the legal limit, workers are safe. Most permissible exposure limits descend from a 1968 list, and even current ones are set at a level judged feasible, not at a level shown to be harmless.
• Protective equipment is the core of workplace safety. It is the last resort in the hierarchy of controls precisely because it fails invisibly, one worker at a time, and depends on continuous correct use.
• Silicosis and black lung are historical diseases. Both are appearing in young workers now, in engineered stone shops and in mines cutting thin seams through silica-rich rock.
• A study finding no excess mortality in a worker cohort shows the job is safe. The healthy worker effect biases such comparisons toward the null, so a flat result is weaker evidence than it looks.
• Occupational exposures affect only the worker. Asbestos and lead carried home on clothing have caused disease in spouses and children.

Recap

• A cluster of silicosis in engineered stone fabricators shows an old hazard reappearing in a workplace nobody was inspecting.
• Work delivers society's highest routine doses to an enumerable population, which is why most human carcinogens were identified in occupational cohorts.
• The healthy worker effect biases occupational comparisons toward finding no effect.
• The hierarchy of controls ranks by failure mode: elimination and engineering keep working, administrative controls and equipment depend on people.
• Most permissible exposure limits date from a 1968 consensus list; NIOSH recommendations are more protective and carry no legal force.
• Coal dust, asbestos, noise, and heat each defeat a different part of the system: changing geology, long latency, a weak exchange rate, and no standard at all.
• Occupational disease is undercounted by design, and the least protected workers are temporary, small-shop, and immigrant employees.

Sources

1. National Institute for Occupational Safety and Health. (n.d.). Workplace safety and health topics, including the hierarchy of controls and recommended exposure limits. Centers for Disease Control and Prevention. cdc.gov/niosh
2. Occupational Safety and Health Administration. (n.d.). Law and regulations, permissible exposure limits, and the respirable crystalline silica standard. U.S. Department of Labor. osha.gov
3. U.S. Bureau of Labor Statistics. (n.d.). Injuries, illnesses, and fatalities program. bls.gov
4. Hamilton, A. (1943). Exploring the Dangerous Trades: The Autobiography of Alice Hamilton, M.D. Little, Brown and Company.
5. Cherniack, M. (1986). The Hawk's Nest Incident: America's Worst Industrial Disaster. Yale University Press.

The blue powder in Goiania

In September 1987 two men scavenging an abandoned radiotherapy clinic in Goiania, Brazil, removed a heavy metal cylinder and sold it to a scrapyard. Inside was a source of cesium-137, about 93 grams of cesium chloride salt. The scrapyard owner broke it open, found that the powder glowed blue in the dark, and thought it was beautiful. He shared it. People rubbed it on their skin. A six-year-old girl ate a sandwich with it on her hands.

By the time the source was identified two weeks later, roughly 112,000 people had to be screened with monitors set up in a football stadium. About 250 were contaminated. Twenty needed hospital treatment for acute radiation effects. Four died, the girl among them. Topsoil, houses, and thousands of cubic meters of material were removed as waste.

The accident is instructive because every element of radiation health physics appears in it. A sealed source is harmless in a shielded container and lethal opened. The powder was dangerous mostly once it was on and in people. Nobody felt anything at the time, because ionizing radiation stimulates no sensory nerve. And the response consumed vastly more effort in screening the worried than in treating the exposed, which is a pattern that recurs at every scale.

The point: Radiation is the most precisely measurable hazard in this course and one of the most poorly reasoned about, and those two facts are related.

What ionizing means, and why the type decides the danger

Ionizing radiation carries enough energy per particle or photon to strip an electron from an atom. That matters biologically because ionization in or near DNA produces strand breaks, directly or through reactive species formed from water. Below that energy threshold, radiation can heat tissue or excite molecules but cannot ionize, which is why ultraviolet light, radio waves and visible light belong in a separate discussion at the end of this lesson.

Type	What it is	Stopped by	Where it is dangerous
Alpha	Helium nucleus, heavy and highly charged	A sheet of paper or the dead outer layer of skin	Almost entirely internal: inhaled or ingested emitters deposit enormous energy in a small volume of tissue
Beta	Electron or positron	A few millimeters of plastic or aluminum	Skin burns externally; internally when the emitter concentrates in an organ, as iodine-131 does in thyroid
Gamma and X-ray	High-energy photons	Centimeters of lead or meters of concrete	External whole-body exposure; this is what a shielded source delivers through a wall
Neutron	Uncharged nuclear particle	Hydrogen-rich material such as water or polyethylene	Reactors and some accelerators; not an environmental exposure for the public

Two internal emitters explain most of what follows. Radon-222 is an alpha emitter, harmless to unbroken skin and a lung carcinogen once inhaled. Iodine-131 is a beta emitter with an eight-day half-life that the thyroid actively concentrates, which is why the exposure route that mattered most after Chernobyl was children drinking milk from cows that had eaten contaminated grass, and why potassium iodide works: flooding the gland with stable iodine blocks uptake of the radioactive form, provided it is taken before or very soon after exposure.

The practical rules for external exposure are time, distance, and shielding. Distance is the strongest of the three, because intensity falls with the square of it: standing twice as far away cuts the dose rate to a quarter.

Three units, and a scale you can hold in your head

Confusion about radiation news usually starts with units. There are three, and they answer different questions. The becquerel measures activity: one decay per second. It says how radioactive a thing is, not what it does to you. The gray measures absorbed dose: one joule of energy deposited per kilogram of tissue. The sievert measures effective dose, taking the gray and weighting it for how damaging the radiation type is and how sensitive the exposed tissues are. Health risk statements use sieverts. Older sources use rem, where 1 sievert equals 100 rem, and rad, where 1 gray equals 100 rad.

Exposure	Approximate effective dose
Dental or chest X-ray	0.005 to 0.1 millisieverts
Flight from New York to London	About 0.04 millisieverts
Average annual dose from natural sources in the United States	About 3.1 millisieverts, of which roughly 2.3 is radon
Average annual dose from medical imaging in the United States	About 3.0 millisieverts, giving a total near 6.2
Abdominal and pelvic computed tomography	Roughly 10 millisieverts for the pair
Annual limit for occupationally exposed workers	50 millisieverts
Onset of acute radiation syndrome, whole body over minutes	About 1,000 millisieverts
Roughly half of untreated people die within sixty days	About 4,000 millisieverts

Read the table twice. The gap between a chest X-ray and acute radiation sickness is a factor of about ten thousand, and the largest routine dose most Americans receive is not from any reactor or weapon but from a gas seeping out of the ground and from medicine.

Worth holding on to: Becquerels tell you a source is present, sieverts tell you whether to care, and the difference between the two accounts for a large share of bad radiation reporting.

Radon, which is the one that matters

Uranium is present in ordinary rock and soil everywhere. Its decay chain passes through radium and then radon-222, a chemically inert gas with a 3.8-day half-life that percolates through pore spaces, enters buildings through foundation cracks, sump openings, and slab penetrations, and accumulates where ventilation is low. The gas itself is largely exhaled. Its short-lived decay products are solid, attach to dust, lodge in the airway lining, and deliver alpha particles directly to bronchial epithelium.

The evidence came in two stages. Underground uranium miners in the Colorado Plateau, Czechoslovakia, and elsewhere showed steep excess lung cancer, but skeptics could argue that miners breathed a great deal else. Then residential case-control studies were pooled, in Europe and in North America, and found a consistent increase in lung cancer risk of roughly 8 to 16 percent per 100 becquerels per cubic meter of long-term average exposure, with no evidence of a threshold. EPA estimates that radon causes about 21,000 lung cancer deaths a year in the United States, second only to smoking.

The interaction with smoking is the part people miss. Radon and tobacco together produce far more disease than either alone, so the great majority of radon-attributable deaths occur in smokers and former smokers. That does not make it a smokers' problem: it makes radon testing an unusually cheap intervention in any household containing a smoker.

EPA recommends action at 4 picocuries per liter, and mitigation is a fan and a pipe that vents sub-slab air above the roofline, typically for a cost in the low thousands of dollars. Short-term test kits cost less than a restaurant meal. This is the rarest thing in environmental health: a large, proven risk with a cheap, individually available fix, ignored mostly because the gas has no smell.

The argument about small doses

The backbone of radiation risk estimation is the Life Span Study, a cohort of roughly 120,000 survivors of Hiroshima and Nagasaki followed since 1950, with individual dose estimates reconstructed from location and shielding. It gives clear, quantified excess cancer risk at moderate and high doses. The problem is that most regulatory questions concern doses below about 100 millisieverts, where any excess is far too small to detect against a background lifetime cancer risk of roughly forty percent.

Regulators therefore extrapolate. The linear no-threshold model assumes risk falls proportionally to zero dose with no safe floor, and the 2006 BEIR VII review of the National Research Council endorsed it as the most reasonable description for solid cancers. Critics make two distinct arguments that are often blurred together. The first is scientific: cells repair damage, and at very low dose rates repair may be efficient enough to produce a threshold, or even a small protective response, a claim called hormesis. The second is practical: multiplying a tiny individual dose by a very large population produces impressive numbers of theoretical deaths that no study could ever confirm, and those numbers drive expensive decisions, including evacuations that carry their own mortality.

Both sides have a real point, and the dispute is unusually resistant to resolution because the study that would settle it cannot be run. To detect a risk that small you would need millions of people with individually reconstructed doses and decades of follow-up. Large studies of nuclear workers and of patients receiving computed tomography in childhood have been broadly consistent with a linear extrapolation, but not decisively so.

The regulatory answer is to keep the model for protection purposes while refusing to treat its arithmetic as a body count. That is the meaning of ALARA, the requirement to keep exposure as low as reasonably achievable rather than merely below a limit.

Chernobyl and Fukushima, counted as they happened

On 26 April 1986 reactor 4 at Chernobyl exploded during a badly designed test, and burned in the open for days. Two workers died at once. Of about 134 emergency workers diagnosed with acute radiation syndrome, 28 died within months. Iodine-131 fell on pasture, and because Soviet authorities did not stop milk distribution or issue potassium iodide, several thousand children and adolescents later developed thyroid cancer, a cancer that is normally rare in childhood and, treated, rarely fatal; the United Nations Scientific Committee counted roughly fifteen deaths from it in the following two decades. Beyond thyroid cancer and the emergency workers, UNSCEAR has not found a consistent radiation-attributable increase in disease in the general population, while describing the mental health consequences of relocation, stigma, and fear as the accident's largest public health effect.

On 11 March 2011 an earthquake and tsunami cut power and cooling at Fukushima Daiichi, and three reactor cores melted. The radiological outcome and the human outcome diverge sharply here. No deaths have been attributed to radiation exposure, and UNSCEAR concluded that no discernible increase in cancer rates is expected. The thyroid screening program found many small cancers in children, which specialists attribute largely to the screening effect: ultrasound of any large population finds indolent tumors that would never have surfaced. Meanwhile more than two thousand deaths in Fukushima Prefecture have been classified as disaster-related, many of them among elderly patients moved from hospitals and nursing homes during the evacuation.

Holding both accidents in view yields a lesson neither side of the nuclear argument likes. Chernobyl shows that a reactor accident can inflict serious, measurable, preventable radiation harm, most of it through an exposure pathway, contaminated milk, that a competent response would have closed in a day. Fukushima shows that the response itself can kill more people than the radiation, and that protective action decisions are medical decisions about frail people, not only radiological ones.

The non-ionizing side

Ultraviolet radiation is the one non-ionizing exposure with an unambiguous cancer verdict. It damages DNA directly, causes keratinocyte carcinomas and melanoma, and the International Agency for Research on Cancer classified ultraviolet-emitting tanning devices as carcinogenic to humans in 2009. Shade, timing, clothing, and sunscreen are all effective, and the UV Index exists to make the exposure legible.

Radiofrequency fields from mobile phones and towers are a different case. IARC classified them as possibly carcinogenic, Group 2B, in 2011, on the strength of some case-control studies of heavy users. Against that sits the absence of the rise in glioma incidence that would be expected from a strong effect, across countries and decades in which mobile phone use went from nothing to universal, and large cohort studies that have not reproduced the association. A 2018 rodent study found tumors in male rats at whole-body exposures well above what a phone delivers. The honest summary is that a large effect is effectively excluded and a small one cannot be, which is not the same as proven safe and not remotely the same as established harm.

Extremely low frequency magnetic fields from power lines and household wiring sit in the same category, also 2B, and for a stranger reason: a fairly consistent epidemiologic association between residential fields above roughly 0.3 to 0.4 microtesla and childhood leukemia that has survived repeated analysis, without any laboratory mechanism or animal evidence to support it. The association is weak, the exposure is rare, and after fifty years no one has explained it.

Common misconceptions

• Radiation is radiation. An alpha emitter is stopped by skin and lethal when inhaled; a gamma emitter reaches you through a wall. Type and route decide the hazard.
• Becquerels measure danger. They measure decays per second. Effective dose in sieverts is the quantity that maps onto risk.
• Nuclear power plants are the public's main radiation exposure. For the average American, natural background and medical imaging dominate, with radon the largest single source.
• Chernobyl killed hundreds of thousands. Documented radiation deaths number in the dozens, plus roughly fifteen thyroid cancer deaths; the enormous projected totals come from multiplying tiny doses across large populations, an arithmetic no study can confirm.
• Fukushima's evacuation was pure benefit. No radiation deaths have been attributed to the accident, while over two thousand deaths were classified as disaster-related, concentrated among the frail elderly who were moved.
• Cell phones are proven to cause brain tumors. Population incidence has not risen as it should if the effect were large, though small effects cannot be excluded.

Where this leaves us

• Goiania shows the whole field in miniature: a source harmless when contained, invisible when not, and a response dominated by screening the worried.
• Alpha, beta, and gamma emitters differ in penetration, so the same nuclide is trivial outside the body and serious inside it, or the reverse.
• Becquerels measure activity, grays measure energy deposited, sieverts weight it for biological effect; risk statements use sieverts.
• Radon is the largest natural exposure and the second leading cause of lung cancer, with a cheap test and a mitigation system that works.
• The linear no-threshold model governs regulation because the doses that matter are too small to study directly, and its critics are right that its arithmetic should not be read as a death toll.
• Chernobyl's harm was real, largely preventable, and concentrated in thyroid cancer and emergency workers; Fukushima's harm came mostly from the evacuation.
• Among non-ionizing exposures, ultraviolet is a proven carcinogen while radiofrequency and power-frequency fields remain classified as possible on limited evidence.

Sources

1. U.S. Environmental Protection Agency. (n.d.). Radiation protection and radon. epa.gov/radon
2. United Nations Scientific Committee on the Effects of Atomic Radiation. (n.d.). Reports on the Chernobyl and Fukushima Daiichi accidents. unscear.org
3. Centers for Disease Control and Prevention. (n.d.). Radiation and your health. cdc.gov
4. National Research Council. (2006). Health Risks from Exposure to Low Levels of Ionizing Radiation: BEIR VII Phase 2. National Academies Press.
5. International Atomic Energy Agency. (1988). The Radiological Accident in Goiania. IAEA.

Forty-nine point six degrees

On 29 June 2021 the village of Lytton, British Columbia, recorded 49.6 degrees Celsius, about 121 degrees Fahrenheit. It was the highest temperature ever measured in Canada, beating the old national record by nearly five degrees, in a place five hundred kilometers north of Seattle. The next afternoon a fire moved through the village and destroyed most of it.

The heat dome that produced that reading sat over the Pacific Northwest for roughly a week. The British Columbia coroner attributed more than six hundred deaths to it, with hundreds more in Oregon and Washington. Most of the dead were older, most were indoors, and most were in dwellings with no air conditioning, in a region where air conditioning had never been necessary. An attribution study published within days concluded that an event of that magnitude would have been effectively impossible in the climate of the previous century.

That last sentence is the reason climate belongs in this course rather than in a course on atmospheric science. Climate change is not an exposure in the sense the rest of this course has used the word. It is a change in the conditions under which every other exposure operates: it moves the temperature distribution, the pathogens, the pollen, the smoke, the water supply, and the places people can live.

The core of it: The hazard is not a new agent. It is a shift in the distribution of old ones, which is why the health effects show up in categories that already existed.

What heat does to a body, and to which bodies

Humans hold core temperature near 37 degrees Celsius. When ambient temperature approaches or exceeds skin temperature, the only remaining route for shedding heat is evaporation of sweat, and evaporation depends on how much water the air can still absorb. This is why humidity matters so much: 40 degrees in dry air is survivable for a healthy adult who can drink and rest, while a far lower dry-bulb temperature can be lethal at high humidity.

The measure that captures both is the wet-bulb temperature, the reading of a thermometer wrapped in a wet cloth. A theoretical survivability limit of 35 degrees wet-bulb has been widely quoted, on the reasoning that above it the body cannot shed metabolic heat at all. Recent laboratory work on young healthy adults suggests the practical limit is lower, closer to 30 or 31 degrees under some conditions, which means the dangerous threshold arrives sooner than the theoretical figure implies.

Heat illness runs a gradient: cramps, then exhaustion with heavy sweating and weakness, then heat stroke, in which thermoregulation fails, core temperature exceeds about 40 degrees, and mental status changes. Heat stroke is a medical emergency with high mortality if cooling is delayed.

But most heat wave deaths are not diagnosed as heat stroke. They are cardiovascular and respiratory deaths in people whose systems could not tolerate the added strain, which is why heat mortality is usually measured as excess deaths during and just after the event rather than by counting death certificates that say heat. The risk factors are consistent across events: age over 65, living alone, chronic cardiovascular or renal disease, psychiatric illness, medications that impair sweating or thirst including anticholinergics and diuretics, homelessness, top-floor apartments, and no access to cooling. Night matters as much as day, because a night that stays above roughly 25 degrees denies the body its recovery period.

Outdoor workers are a separate population with a separate mechanism, and the chronic kidney disease among Central American cane cutters described in the previous lesson is the clearest current example of heat producing disease rather than acute death.

Three pathways, not one

The useful way to organize an enormous literature is by how far the causal chain runs between the climate signal and the health outcome.

Pathway	Mechanism	Examples	What makes attribution easier or harder
Direct	Weather acts on the body	Heat illness, drowning and injury in floods, storm deaths, cold snaps	Easiest: the event has a date and a mortality curve
Ecosystem-mediated	Climate changes the behavior of organisms and chemistry	Vector range shifts, waterborne pathogens, harmful algal blooms, aeroallergens, ozone formation, wildfire smoke	Moderate: trends are measurable but many drivers act at once
Socially mediated	Climate changes economies, food supply, and where people can live	Undernutrition, displacement, occupational loss, mental health effects, conflict over water	Hardest: mediated through institutions, so effects are large but hard to isolate

The gradient in that last column explains a persistent asymmetry in the evidence: the effects we quantify best are probably not the largest ones. Heat mortality is countable. Undernutrition from repeated crop failure is not, and the World Health Organization's estimate of about 250,000 additional deaths a year between 2030 and 2050 covers only four causes, undernutrition, malaria, diarrheal disease, and heat exposure, precisely because those were the ones that could be modeled.

When the map moves, the vectors move

Mosquitoes and ticks are ectotherms. Temperature governs their development rate, their survival through winter, their biting frequency, and the speed at which a pathogen replicates inside them, which is why small warming shifts can produce large changes in transmission potential at the edge of a range.

The blacklegged tick, Ixodes scapularis, has moved north into eastern Canada, where established populations now exist in areas that had none a generation ago. Reported Lyme disease in Canada rose more than tenfold in the decade after 2009. In the United States the tick has expanded into the upper Midwest and northern New England, and reported Lyme case counts have risen accordingly, though better recognition and testing contribute to that count as well.

Aedes aegypti and Aedes albopictus, the mosquitoes that carry dengue, chikungunya, and Zika, are expanding poleward and upward in elevation. The Americas recorded their worst dengue years on record in 2023 and 2024, and local transmission has occurred in Florida, Texas, Arizona, and California. Dengue's dependence on temperature is not simple, because very high temperatures shorten mosquito survival, so warming shifts the map rather than uniformly expanding it.

Vibrio vulnificus, a bacterium that causes severe wound infections and sepsis, requires warm brackish water. As coastal waters have warmed, infections have appeared further north along the United States Atlantic coast and in the Baltic Sea, in places with no prior history of them.

None of this means the tropics are simply moving north. Transmission depends on housing, screens, water storage, land use, and public health capacity as much as on climate, which is why dengue is endemic in places far cooler than parts of the United States that have almost none.

The air gets worse on hot days

Module 2 covered the criteria pollutants. Climate interacts with two of them in specific ways.

Ground-level ozone forms faster in heat and sunlight, so the same emissions produce higher concentrations on hotter days, an effect atmospheric chemists call the climate penalty. Meeting a fixed ozone standard therefore gets harder as summers warm, even with unchanged precursor emissions.

Wildfire smoke has become the dominant particulate exposure event in North America. On 7 June 2023, smoke from Canadian fires pushed New York City to the worst measured air quality in its modern record, turning midday orange and sending emergency department visits for asthma sharply upward across the Northeast. Wildfire smoke is mostly fine particulate matter, and there is evidence it is more harmful per microgram than urban particulate matter of the same mass. In much of the western United States, decades of gains in particulate air quality have been partly reversed by smoke.

Pollen has moved too. Analyses of North American pollen monitoring stations found that seasons have lengthened by roughly twenty days and pollen concentrations have risen by about a fifth since 1990, with the strongest changes at higher latitudes. Longer, denser pollen seasons mean more allergic rhinitis and more asthma exacerbation, and unlike most items in this lesson, that one is already measurable in clinic visits.

In short: Climate does not usually create a new disease. It raises the frequency of the conditions that trigger diseases we already track.

Counting deaths that no death certificate names

Attribution has become quantitative in two distinct ways, and confusing them is a common error.

Event attribution asks how much more likely or intense a particular event was made by warming, by comparing simulations of the actual climate with simulations of a counterfactual one. The Pacific Northwest study is an example, and such analyses now appear within days of an event.

Health attribution asks how many deaths in an event or a period would not have occurred otherwise, which requires an epidemiologic exposure-response function relating temperature to mortality, applied to the attributed change in temperature. Both steps carry uncertainty, and honest studies report wide intervals.

What is not in dispute is the direction and the shape of the exposure-response curve: it is U-shaped, with mortality rising at both temperature extremes, and the location of its minimum differs by city because populations adapt to their own climate. A 28-degree day is unremarkable in Phoenix and dangerous in Vancouver. That is one reason absolute thresholds are poor warning triggers and percentile-based ones work better.

Adaptation that has been measured

After the 2003 European heat wave, which produced on the order of 70,000 excess deaths across the continent, France built a national heat plan: a warning system tied to forecast thresholds, municipal registries of isolated vulnerable residents, defined actions for nursing homes and hospitals, and public communication. Mortality in subsequent heat waves came in below what the 2003 relationship would have predicted.

Ahmedabad, in Gujarat, adopted South Asia's first city heat action plan in 2013 after a lethal 2010 heat wave: color-coded forecast alerts, hospital preparation, water distribution, adjusted work and school hours, and cooling spaces. Evaluations have credited it with avoiding on the order of a thousand deaths a year, and dozens of Indian cities have since adopted versions of it.

The components that recur in effective plans are unglamorous: a forecast trigger, a named person responsible for pulling it, a list of who to check on, somewhere cool that is actually open in the evening, and instructions to the institutions holding the frailest people. Tree canopy, cool roofs, and building standards work on a slower timescale and address the underlying exposure rather than the emergency.

Adaptation can also misfire. Air conditioning saves lives and is the single strongest protective factor in heat waves, and it also raises peak electricity demand, exhausts waste heat into the street, and until recently ran on refrigerants that are potent greenhouse gases; the 2016 Kigali Amendment addresses the last of these by phasing down hydrofluorocarbons. A strategy that protects the people who can afford it while raising the outdoor temperature for everyone else is a partial answer.

The mitigation side carries a health argument of its own. Burning less fossil fuel reduces particulate and ozone exposure immediately and locally, decades before the climate benefit arrives. These health co-benefits are large enough that several analyses find they alone can offset a substantial share of mitigation costs, which changes the shape of the policy question: the air quality benefit is not a bonus, it is a near-term return.

Common misconceptions

• Heat deaths are mostly heat stroke. Most are cardiovascular and respiratory deaths in vulnerable people, which is why excess mortality is the right measure and death certificates undercount badly.
• Warmer winters will offset heat deaths. Cold-related mortality is real, but it tracks winter mortality patterns poorly, and the studies that examine both generally find heat increases outweigh cold decreases in warm and temperate regions.
• Climate change will bring tropical diseases straight north. Transmission depends on housing, water storage, screens, and public health capacity, not only on temperature, which is why dengue is endemic in some cooler places and nearly absent in some hot ones.
• The health effects are far in the future. Heat mortality, wildfire smoke exposure, and lengthened pollen seasons are already measurable in mortality and clinic data.
• Air conditioning solves heat. It is the strongest individual protection and simultaneously raises peak demand and outdoor heat, and it is distributed by income.

The short version

• Lytton reached 49.6 degrees Celsius in June 2021 and burned the next day; the heat dome killed more than six hundred people in British Columbia alone.
• Heat kills through cardiovascular and respiratory strain in the old, ill, isolated, and unairconditioned, and humidity matters as much as temperature.
• The health effects sort into direct, ecosystem-mediated, and socially mediated pathways, and the ones we measure best are probably not the largest.
• Ticks and mosquitoes are expanding their ranges, and warm-water pathogens are appearing further north.
• Heat raises ozone, wildfire smoke has reversed particulate gains in parts of North America, and pollen seasons are about twenty days longer than in 1990.
• Event attribution and health attribution are different calculations, and mortality follows a U-shaped curve whose minimum differs by city.
• Heat action plans in France and Ahmedabad have measurably reduced deaths, and cutting fossil fuel combustion delivers air quality benefits immediately.

Sources

1. World Health Organization. (n.d.). Climate change and health. who.int
2. U.S. Environmental Protection Agency. (n.d.). Climate change indicators in the United States. epa.gov
3. U.S. Global Change Research Program. (n.d.). National climate assessment and the climate and health assessment. globalchange.gov
4. Centers for Disease Control and Prevention. (n.d.). Heat and health, and tickborne disease surveillance. cdc.gov
5. Robine, J. M., Cheung, S. L. K., Le Roy, S., Van Oyen, H., Griffiths, C., Michel, J. P., and Herrmann, F. R. (2008). Death toll exceeded 70,000 in Europe during the summer of 2003. Comptes Rendus Biologies, 331(2), 171-178.

Five hundred arrests on a road in Afton

In 1978 a trucking company hired to dispose of oil contaminated with polychlorinated biphenyls solved the problem by opening a valve and driving. Over several nights it sprayed roughly 30,000 gallons along about 240 miles of North Carolina roadside shoulders. The state then had to decide where to bury the contaminated soil.

It chose Warren County, the county with the highest proportion of Black residents in the state and one of its poorest. In September 1982, when the trucks began arriving at the landfill site near Afton, residents lay down in the road in front of them. Over six weeks more than five hundred people were arrested, including clergy, civil rights organizers, and members of Congress. They did not stop the landfill.

What the protest did do was force a question into a form that could be studied. Was Warren County a coincidence? Within a year the General Accounting Office examined four hazardous waste landfills in the southeastern states and found that three of the four sat in communities that were majority Black. In 1987 the United Church of Christ Commission for Racial Justice published a national study reporting that race predicted the location of commercial hazardous waste facilities better than income did. Sociologist Robert Bullard, working from a 1979 Houston lawsuit, documented that all five city-owned landfills and six of eight municipal incinerators had been placed in Black neighborhoods in a city roughly a quarter Black.

By 1991 a national summit of community organizations had adopted a set of principles and a name for the field. In February 1994 President Clinton signed Executive Order 12898, directing every federal agency to identify and address disproportionately high and adverse health or environmental effects of its programs on minority and low-income populations.

Worth holding on to: The field began not with a hypothesis but with a road, a landfill, and a set of arrests, and the research followed the protest rather than preceding it.

The objection that had to be answered

Those early studies shared a structure. They looked at where facilities are today and at who lives there today, and found the two correlated. In 1994 Vicki Been, then a law professor at New York University, pointed out that this design cannot distinguish between two very different stories, both consistent with the same data.

In the first, a siting authority chooses a Black or poor neighborhood because its residents have less political capacity to object. That is discrimination in siting, and the remedy is to change how permits are granted.

In the second, a facility is built somewhere, property values near it fall, and over the following decades the households that move in are the households with the fewest housing options, who in a segregated housing market are disproportionately Black, Hispanic, and poor. Nobody in that story chose a neighborhood by race. The pattern is produced by the market and by the housing discrimination that shaped it, and the remedy is different: it lies in housing policy and in the compensation of residents rather than in permit review.

Been was not defending siting practice; she was arguing that a claim of disproportionate siting requires demographic data from the time of siting, not from the present. She then did that work. Re-examining the Houston facilities, she found that the neighborhoods had in fact been disproportionately Black when the facilities were placed, which supports the siting story in that city. A later national longitudinal study she conducted with Francis Gupta found that tracts receiving facilities had been disproportionately poor and disproportionately Hispanic at the time of siting, found weaker national evidence for disproportionate siting by Black population share, and found little sign that host neighborhoods changed dramatically afterward.

Mixed results, honestly reported, from someone testing her own hypothesis. That is what the argument looked like at its best.

How the question became answerable

Two methodological moves changed the picture. The first concerned distance. Most early studies used what Paul Mohai and Robin Saha called unit-hazard coincidence: take the zip code or census tract containing a facility and compare its demographics to tracts without one. The flaw is geometric. Tracts are large and irregular, a facility can sit at the very edge of one, and the people closest to it may live across the boundary in a tract counted as unexposed. When Mohai and Saha instead drew circles of one and three kilometers around facilities and weighted the population inside them, the measured racial disparities grew substantially. The disparities had been diluted by the units, not absent.

The second move was to make the studies longitudinal, comparing neighborhood composition before siting, at siting, and decades after. Reviewing that literature and adding their own national analysis, Mohai and Saha reported that host neighborhoods were disproportionately minority at the time of siting, and that post-siting demographic change was generally too small to account for present-day disparities. Other researchers using different samples and periods have found more evidence of subsequent change, particularly in Hispanic population growth around facilities.

Where that leaves an honest reader: disproportionate siting is supported by the strongest available evidence in most samples, market dynamics are real and contribute in some places, and the two are not mutually exclusive. A neighborhood can be chosen because it is poor and then become poorer because it was chosen. Note also what neither story does: whichever mechanism produced the arrangement, the residents breathing the air today are breathing it today.

What matters here: The causal question determines the remedy, not whether the disparity exists. Siting discrimination points at permits; market dynamics points at housing policy; a disparity produced by both requires both.

What current exposure data show

The siting debate is now only part of the evidence, because exposure itself can be measured directly rather than inferred from facility locations.

Fine particulate exposure has fallen steeply for everyone in the United States since 1980, and the disparities have persisted through the decline. A national analysis published in 2020 found that the places that were most polluted in 1981 were largely the same places that were most polluted in 2016, and that they remained disproportionately Black. A separate 2019 study took a different approach, tracing particulate exposure to the consumption that generates it, and found that Black and Hispanic Americans are exposed to more particulate pollution than their own consumption of goods and services produces, while white Americans are exposed to less. The disparity holds at low absolute concentrations, which matters because the exposure-response relationship for particulates has no established threshold.

The pattern repeats across this course's other chapters. Blood lead levels are higher in Black children and in low-income children, a legacy of housing age and of where lead-emitting industry sat. Serious asthma is concentrated in the same neighborhoods that carry old housing stock, truck routes, and highway proximity. On the Navajo Nation, several hundred abandoned uranium mines from Cold War procurement remain, with contaminated water sources and homes built of mine waste rock. Diesel corridors, port neighborhoods, and the industrial stretch of the lower Mississippi carry their own concentrated burdens.

Two other dimensions deserve naming, because reducing environmental justice to exposure numbers misses them. Procedural justice concerns who gets a hearing: whether notice is published in a language residents read, whether meetings are held when working people can attend, whether comments demonstrably change outcomes. And cumulative impact concerns the fact that permits are issued one at a time while residents live with the sum, so a community can host a dozen individually compliant facilities and be exposed to a mixture no permit ever evaluated.

Why civil rights law turned out to be a weak instrument

The intuitive legal theory is discrimination. It has mostly failed, for reasons worth understanding precisely.

Constitutional equal protection claims require proof of discriminatory intent, not merely of disparate outcome. Siting decisions are made through zoning, land cost, transportation access, and existing industrial use, so the record almost never contains evidence of intent, and courts have consistently rejected these claims.

Title VI of the Civil Rights Act of 1964 looked more promising, because agency regulations under it reach practices with discriminatory effects regardless of intent. But in 2001 the Supreme Court held in Alexander v. Sandoval that private parties cannot sue to enforce those disparate-impact regulations. That left the administrative route: file a complaint with EPA and wait. Those complaints have historically taken years and rarely produced a permit denial.

Executive Order 12898 has its own limit written into it. It directs agencies to consider these effects, and it expressly creates no right enforceable in court. An executive order can also be revised or rescinded by a later president, which is a structural weakness rather than a comment about any administration.

What has actually changed practice

Progress has come from screening, from state statutes, and from research design.

Mapping tools made the disparities legible and comparable. California's OEHHA built CalEnviroScreen, which combines pollution burden indicators with population vulnerability indicators into a score for every census tract, and the state ties certain funds to it. EPA built EJScreen on similar logic for national use. These tools do not prove causation and are not risk assessments, but they force the cumulative question into a form a permit writer can see.

New Jersey went further in 2020, enacting a law requiring the state to deny a permit for a new facility in an overburdened community when it would contribute to disproportionate cumulative stressors, the first statute in the country to make cumulative impact a basis for denial rather than a factor to weigh. Whether it changes outcomes at scale is still being determined, which is the honest status of a rule that recent.

Research practice changed too. Community-based participatory research, in which residents help define the question and hold rights over the data, produced the bucket brigades that documented refinery emissions and the community air monitoring networks that have since been adopted by regulators.

And Warren County, eventually, got its remedy. After two decades of pressure, the state detoxified the landfill, treating the contaminated soil rather than leaving it. It took twenty years and it was not automatic.

Common misconceptions

• Environmental justice is only about hazardous waste siting. Siting was the origin; the evidence now covers particulate exposure, lead, asthma, heat, drinking water, and occupational hazards.
• If minority residents moved in after a facility was built, there is no injustice. A housing market shaped by segregation and by the facility itself is not a neutral sorting mechanism, and the exposure is identical regardless of arrival order.
• Disparities are entirely explained by income. The 1987 national study found race predicted facility location better than income, and racial disparities in particulate exposure persist within income strata.
• Air quality improvement has eliminated the disparities. Particulate concentrations fell steeply for everyone while the ranking of places stayed remarkably stable.
• Executive Order 12898 gives communities a legal right to sue. It directs agencies to consider these effects and expressly creates no enforceable right.

What to carry forward

• The field began with the 1982 Warren County protests, followed by GAO in 1983, the United Church of Christ study in 1987, and Executive Order 12898 in 1994.
• Been's market dynamics objection was a real methodological challenge: present-day demographics cannot distinguish siting discrimination from later demographic change.
• Longitudinal and distance-based studies mostly support disproportionate siting, while confirming that post-siting change contributes in some places.
• Measurement geometry matters: comparing host tracts to non-host tracts dilutes disparities that circle-based methods reveal.
• Particulate exposure fell for everyone while the most polluted places stayed the most polluted, and they remain disproportionately Black.
• Equal protection requires intent, and Sandoval closed private disparate-impact suits under Title VI, leaving slow administrative complaints.
• Screening tools, cumulative impact statutes such as New Jersey's, and community-based research have moved practice further than litigation has.

Sources

1. U.S. Environmental Protection Agency. (n.d.). Environmental justice. epa.gov
2. California Office of Environmental Health Hazard Assessment. (n.d.). CalEnviroScreen. oehha.ca.gov
3. U.S. Government Accountability Office. (1983). Siting of Hazardous Waste Landfills and Their Correlation with Racial and Economic Status of Surrounding Communities. GAO/RCED-83-168. gao.gov
4. Been, V. (1994). Locally undesirable land uses in minority neighborhoods: Disproportionate siting or market dynamics? Yale Law Journal, 103(6), 1383-1422.
5. Mohai, P., and Saha, R. (2015). Which came first, people or pollution? A review of theory and evidence from longitudinal environmental justice studies. Environmental Research Letters, 10(12), 125011.

The same number, chosen twice

On 22 January 2001 EPA published a rule cutting the drinking water limit for arsenic from 50 to 10 micrograms per liter, the first change to that standard since 1975. Two months later the incoming administration delayed the effective date and asked for further review, on the argument that the science and the cost estimates deserved another look. The National Research Council was asked to update its assessment. In October 2001 the agency announced that the standard would stand at 10.

Nothing about arsenic changed in those nine months. The toxicology was the same in October as in January; the NRC update in fact concluded that the risks at 10 micrograms per liter were still higher than regulators normally accept. What was actually being contested was a decision rule: how much cost a water system should bear for a given reduction in lifetime cancer risk, and who decides.

That episode is the whole subject of this lesson. Ask why one contaminant is regulated at a level near zero and a chemically similar one is barely regulated at all, and the answer is usually not the toxicology. It is which statute applies, and what question that statute instructs the agency to ask.

The upshot: A standard is a number produced by a decision rule applied to scientific evidence. Change the rule and the number moves, even when the evidence does not.

Where the science stops and the choice begins

Module 1 walked through the four steps of risk assessment: hazard identification, dose-response assessment, exposure assessment, and risk characterization. That process ends by handing the decision maker something like a reference dose, a slope factor, and an estimate of how much exposure the population actually receives.

None of that tells anyone what to do. A lifetime cancer risk of two in ten thousand is a fact about a population; whether it is acceptable is not. The step that follows, risk management, is where the statute takes over, and the statutes disagree with each other profoundly.

Five statutes, five different questions

Statute	What it regulates	The question the agency must answer	Where cost enters
Clean Air Act, criteria pollutants	Ambient concentrations of six pollutants	What level is requisite to protect public health with an adequate margin of safety, including sensitive groups?	Not at all in setting the standard; only in implementation planning
Safe Drinking Water Act	Contaminants in public water systems	What is the health goal, and how close to it can systems feasibly get with best available technology?	Directly: the enforceable limit is set as close to the goal as is feasible, with cost-benefit analysis added in 1996
Occupational Safety and Health Act	Workplace exposures	Is there a significant risk, and if so what level eliminates it to the extent feasible?	Only as feasibility, which asks whether the industry can survive the rule, not whether benefits exceed costs
Toxic Substances Control Act, as amended in 2016	Industrial chemicals	Does the chemical present unreasonable risk under its conditions of use?	Excluded from the risk determination, permitted when choosing the management measure
Federal Insecticide, Fungicide, and Rodenticide Act	Pesticide registration	Do the benefits of the use outweigh its adverse effects?	Built into the standard itself as an explicit risk-benefit balance

Read the third column as a ranking of stringency and you will predict most regulatory outcomes correctly. A pesticide is judged by whether it is worth it. An industrial chemical is judged by whether its risk is unreasonable. An air pollutant is judged only by what protects health.

That last point was tested and settled. Challenging the 1997 ozone and particulate standards, industry argued that EPA had to weigh implementation costs. In Whitman v. American Trucking Associations, decided in 2001, the Supreme Court held unanimously that the Clean Air Act does not permit EPA to consider cost in setting a national ambient air quality standard. Costs are real and get considered, but at the implementation stage, in the state plans that decide how to reach the number.

Two cases that fixed the rules for workplaces

The Occupational Safety and Health Act instructs the agency to set standards that most adequately assure, to the extent feasible, that no worker suffers material impairment of health. Two Supreme Court decisions turned that sentence into a procedure.

In 1980, reviewing OSHA's benzene standard, the Court held in Industrial Union Department v. American Petroleum Institute that the agency must first make a threshold finding that a significant risk exists at current exposures before it may regulate. The plurality offered a rough calibration that has been quoted ever since: a lifetime risk of one in a thousand might well be significant, while one in a billion plainly is not. That decision is why every OSHA health standard now opens with a quantitative risk assessment.

The following year, reviewing the cotton dust standard in American Textile Manufacturers Institute v. Donovan, the Court held that once significant risk is established, the Act requires the standard to be set at the level that eliminates it to the extent feasible, and that formal cost-benefit analysis is not required and not permitted. Feasibility asks whether the industry can absorb the rule and continue to exist, which is a different and generally more protective question than whether monetized benefits exceed monetized costs.

Bottom line: Under the Act that governs the highest exposures in the country, the agency must prove significant risk before acting and may not balance lives against dollars once it does.

Why one statute did almost nothing for forty years

The Toxic Substances Control Act of 1976 was supposed to be the general chemical law. Two design decisions crippled it. First, roughly sixty thousand chemicals already in commerce were grandfathered onto the inventory without review, so the burden fell on the agency to demonstrate a problem rather than on the manufacturer to demonstrate safety. Second, before restricting a chemical, EPA had to show that its rule was the least burdensome alternative that would adequately protect.

The consequences arrived in 1991, when the Fifth Circuit vacated most of EPA's asbestos ban in Corrosion Proof Fittings v. EPA. The agency had spent about a decade and compiled a record tens of thousands of pages long, on a substance whose carcinogenicity nobody disputed, and the court held that it had not adequately shown that a ban was less burdensome than the alternatives it had considered. After that decision the agency largely stopped trying: in four decades under the original statute, only a handful of existing chemicals were restricted.

The 2016 Lautenberg amendments rewrote the core of it. EPA must now evaluate chemicals on a schedule, must make risk determinations without consideration of cost, and may weigh cost only when selecting the management measure. The least burdensome language is gone. Whether the new machinery moves faster is being tested now, chemical by chemical.

Zero risk in food, until the chemistry improved

The 1958 food additives amendment contained a sentence with no equivalent anywhere else in American regulation. The Delaney Clause provided that no additive found to induce cancer in humans or animals could be deemed safe, at any level. It set the acceptable risk at zero and left the agency no discretion.

It was workable while laboratories could detect parts per million. As detection limits fell toward parts per trillion, the clause began to prohibit residues at concentrations with no plausible health significance, and to do so inconsistently: a pesticide residue concentrated in processed food fell under the zero standard while the identical residue on the raw commodity did not. When EPA tried to apply a de minimis exception, the Ninth Circuit held in 1992 that the statute did not allow one.

Congress resolved it in the Food Quality Protection Act of 1996, replacing the Delaney standard for pesticide residues with a single safety finding, a reasonable certainty of no harm, applied consistently to raw and processed food, and adding an additional tenfold safety factor to protect infants and children unless data show it is unnecessary. That extra factor is one of the few places in American law where children's greater susceptibility is written directly into the arithmetic.

How a rule actually moves, and where it stops

The pipeline from evidence to enforceable number has more steps than most people expect, and each one is a place where a rule can sit for years.

1. An assessment establishes hazard and dose-response. At EPA this is often an IRIS assessment; some have taken more than a decade.
2. A science advisory body reviews it. For air standards this is the Clean Air Scientific Advisory Committee, and the criteria pollutants are on a five-year review cycle.
3. The agency drafts a proposed rule and, if it is economically significant, sends it to the Office of Information and Regulatory Affairs in the White House for review under Executive Order 12866. Rules can sit there.
4. The proposal is published in the Federal Register for public comment. The agency must respond to significant comments, and failure to do so is itself a ground for reversal.
5. The final rule is published with a preamble that answers the comments and builds the record for the litigation that follows.
6. Someone sues. Courts review under the arbitrary and capricious standard, asking whether the agency examined the relevant data and articulated a rational connection between the facts and its choice.
7. Congress may disapprove a recent rule under the Congressional Review Act, and a future administration may begin a new rulemaking to revise it, which takes as long as the first one did.

The silica standard from Module 4 illustrates the timescale: a hazard described in the sixteenth century, uncontested by the 1970s, proposed in 2013, final in 2016. The arsenic rule took roughly a decade of assessment, then survived a nine-month reversal fight, and gave systems until 2006 to comply.

The other model, and what it asks instead

The European Union built its chemicals system on a different premise. Under REACH, in force since 2007, a company that manufactures or imports a substance above one tonne a year must register it with a dossier of data proportionate to volume. No dossier, no market. Substances of very high concern, including known carcinogens, mutagens, reproductive toxicants, and persistent bioaccumulative substances, require specific authorization for continued use, granted only on a showing of adequate control or overriding socioeconomic benefit with no suitable alternative.

The difference from the American default is the location of the burden of proof. The United States has generally required the government to demonstrate harm; REACH requires the manufacturer to demonstrate sufficient information for safe use. The Lautenberg amendments moved American law partway toward that position for new chemicals.

Both systems face the same underlying difficulty, which no procedural design removes: the number of chemicals in commerce vastly exceeds the number that have been thoroughly evaluated, and evaluation is slow, expensive, and contested. Neither model has solved that. The choice between them is a choice about who bears the cost of uncertainty while it lasts.

Key idea: Ask of any standard: who had to prove what, to whom, before anything could happen? That single question explains more about the resulting number than the toxicology does.

Common misconceptions

• Regulatory limits are set where harm begins. They are set where a particular statute's decision rule lands, given uncertainty factors, feasibility, and sometimes an explicit benefit calculation.
• Agencies always weigh costs against benefits. EPA may not consider cost in setting a national ambient air quality standard, and OSHA may not use formal cost-benefit analysis once significant risk is established.
• If a chemical is on the market, someone has evaluated it. Roughly sixty thousand chemicals were grandfathered onto the TSCA inventory in 1976 without review.
• Stricter is always better regulatory practice. The Delaney Clause's zero-risk rule became unworkable as detection limits fell, and Congress replaced it with a health-based safety finding plus a children's factor.
• Rules take a long time because the science is unsettled. Assessment, advisory review, White House review, comment response, and litigation each add years to standards whose science was settled long before.

Putting it together

• The arsenic standard was set at 10 micrograms per liter, delayed, and reinstated at the same value, because the dispute was about the decision rule and not the toxicology.
• Risk assessment ends by producing numbers; the statute decides what to do with them.
• The Clean Air Act forbids cost in setting the standard, the Safe Drinking Water Act builds in feasibility, and FIFRA balances benefits against harms.
• The benzene decision requires a finding of significant risk, and the cotton dust decision requires feasibility rather than cost-benefit balancing.
• TSCA grandfathered sixty thousand chemicals and demanded the least burdensome alternative, which the asbestos case showed to be nearly impossible; the 2016 amendments removed both barriers.
• The Delaney Clause set risk at zero until analytical chemistry made it unworkable, and the 1996 statute replaced it with a reasonable certainty of no harm plus a tenfold children's factor.
• REACH shifts the burden to the manufacturer, which is a different answer to the same unsolved problem of too many chemicals and too little data.

Sources

1. U.S. Environmental Protection Agency. (n.d.). Laws and regulations, and the Integrated Risk Information System. epa.gov
2. Occupational Safety and Health Administration. (n.d.). Law and regulations. U.S. Department of Labor. osha.gov
3. Office of the Federal Register. (n.d.). Proposed rules, final rules, and public comment. federalregister.gov
4. European Chemicals Agency. (n.d.). Understanding REACH. echa.europa.eu
5. National Research Council. (1983). Risk Assessment in the Federal Government: Managing the Process. National Academy Press.