And Then You're Dead

The Autopsy / Case File · Environmental & Climate

The Nine Causes

The air pollution rule everyone's already meeting doesn't mean the air is safe. A landmark study of 4.5 million veterans found it kills through nine distinct diseases, and nearly all of the deaths happen below the legal limit.

Fine particulate matter — PM2.5, particles small enough to lodge deep in the lungs and cross into the bloodstream — has been linked to heart and lung disease for decades. What a 2021 study in JAMA Network Open added was a longer, more specific list: nine separate causes of death, three of them never previously connected to air pollution at all, built from a decade of health records on nearly every U.S. veteran enrolled in VA care.

And Then You're Dead · September 2026 · 9 min read

Join the discussion →
Industrial smoke creates a haze lingering over the Pittsburgh skyline along the Monongahela River
Steel-mill smoke settles over Pittsburgh in this 1973 EPA photograph. Modern PM2.5 exposure is invisible by comparison — and, per the underlying research, still deadly well below what current law defines as safe.

Air pollution research has spent half a century building the case that PM2.5 — particulate matter smaller than 2.5 micrometers, roughly a thirtieth the width of a human hair — is dangerous. What's changed more recently isn't the basic premise; it's the resolution. A 2021 study out of Harvard, drawing on the health records of 4.5 million U.S. military veterans followed from 2006 to 2016, didn't just confirm that PM2.5 kills. It identified, cause by cause, exactly how — and found that three of the ways it kills had never been documented before.

Six Known Causes, Three New Ones

Before this study, the epidemiological consensus already linked PM2.5 exposure to six specific causes of death: cardiovascular disease, cerebrovascular disease (stroke), chronic obstructive pulmonary disease, type 2 diabetes, lung cancer, and pneumonia. The veteran cohort, tracked with a median follow-up of ten years, confirmed all six and added three more: chronic kidney disease, hypertension, and dementia. None of those three had been established as PM2.5-attributable causes of death in prior research at this scale — they emerged specifically because the study's size (4.5 million people, one of the largest cohorts ever assembled for this kind of analysis) gave it the statistical power to detect associations too subtle for smaller studies to isolate.

0
total U.S. deaths from all nonaccidental causes estimated to be attributable to PM2.5 exposure, per the study's national extrapolation
0
newly identified PM2.5-linked causes of death — chronic kidney disease, hypertension, and dementia — none previously established at this scale

The breakdown by cause shows how unevenly the burden falls across those nine diseases. Cardiovascular disease alone accounted for an estimated 56,070 of the attributable deaths — the single largest share — followed by cerebrovascular disease at roughly 40,466. But the three newly identified causes weren't a rounding error: hypertension-related deaths totaled an estimated 30,697, larger than either lung cancer (17,545) or pneumonia (8,855), the two most historically well-established PM2.5-linked causes. Dementia added an estimated 19,852 deaths and chronic kidney disease roughly 7,175 — both categories that, before this study, wouldn't have appeared on an air-pollution mortality accounting at all.

The 99% Finding

The study's single most consequential result wasn't the length of the list — it was where on the exposure curve the deaths were happening. The EPA's current annual PM2.5 standard sets 12 micrograms per cubic meter as the legal ceiling for what regulators consider an acceptable, health-protective average exposure level. The veteran study found that 99% of the attributable nonaccidental deaths in its dataset occurred among people exposed to PM2.5 levels below that legal limit.

Ninety-nine percent of the deaths this study attributes to fine particulate pollution happened at exposure levels the EPA's own standard already classifies as legally acceptable air quality.

That finding reframes what a "legal" pollution limit actually protects against. A regulatory ceiling implies a threshold below which exposure is safe enough to tolerate; this research found essentially no such floor for PM2.5 mortality risk. The dose-response relationship the study documented was close to linear across the range of exposures in the data, meaning incremental reductions in pollution, even from levels already well under the legal cap, continued to correspond with measurably lower mortality risk. There was no clean cutoff where the risk curve flattened out.

Who Absorbs the Burden

The study didn't stop at identifying causes and thresholds — it also broke the attributable mortality down by race and community-level socioeconomic status, and the disparity was stark and consistent with decades of separate environmental-justice research.

0
age-standardized PM2.5-attributable deaths per 100,000 among Black veterans, versus 51.0 per 100,000 among all other veterans in the cohort
0
deaths per 100,000 in the most socioeconomically disadvantaged communities, versus 46.1 per 100,000 in the least disadvantaged — a gap wider than the racial disparity alone

The socioeconomic gap was, by the study's own numbers, larger than the racial gap — a reminder that PM2.5 exposure tracks not just with race but with the specific, mappable geography of industrial zoning, highway proximity, and land-use decisions that concentrate pollution sources in lower-income areas regardless of who lives there. The two disparities compound rather than substitute for each other: Black Americans are also disproportionately likely to live in the socioeconomically disadvantaged communities carrying the highest exposure, meaning the mortality burden isn't spread evenly even within the group already carrying more of it.

Why a Veteran Cohort, and What It Can't Tell You

The study's power came from a tradeoff worth naming directly. VA health records offered something almost no other dataset could: a genuinely massive population, with consistent long-term follow-up, geocoded home addresses that could be matched against detailed PM2.5 monitoring and modeling data, and death records with reliably coded causes of death. That's an unusual combination of scale and data quality in environmental epidemiology, where studies are more often limited to tens of thousands of participants in a single city or region.

The tradeoff is a cohort that was 93.8% male with a median age over 64 at baseline — military veterans skew older, male, and, relative to the general population, carry a somewhat different baseline health profile shaped by military service itself. The researchers accounted for this statistically, but it means the precise attributable-death counts are best read as a rigorous estimate calibrated to a veteran population, then extrapolated nationally, rather than a direct census of the general public. What doesn't depend on that extrapolation is the more basic finding: the specific disease list itself, the near-linear dose-response curve, and the 99%-below-legal-limit result have since been broadly consistent with findings from separate Medicare-beneficiary cohort studies examining PM2.5's cause-specific mortality effects using an entirely different population.

A Standard Written Around an Incomplete List

The EPA's PM2.5 standard, like most air-quality regulation, is calibrated against the epidemiological evidence available at the time it was last revised — evidence that, until this study, didn't include chronic kidney disease, hypertension, or dementia as PM2.5-attributable causes of death at all. A regulatory limit built around six known causes of death is answering a different, narrower question than one built around nine. The gap between what the standard was designed to prevent and what the fuller disease list now shows it doesn't prevent is not a hypothetical distinction — it's the space in which 99% of this study's attributable deaths took place.

The EPA did eventually move the number. In February 2024, the agency finalized a tightened annual PM2.5 standard, lowering the legal ceiling from 12.0 micrograms per cubic meter to 9.0 — the first change to the annual limit in over a decade, and one the agency's own modeling projected would prevent an estimated 4,500 premature deaths a year by 2032, alongside hundreds of thousands of avoided asthma cases and lost workdays. That is a real, measurable tightening, and it moves millions of Americans in the highest-exposure counties into a bracket the EPA will now formally classify as out of compliance, which triggers mandatory state cleanup plans.

But a near-linear dose-response curve doesn't stop mattering just because the ceiling moved. If 99% of attributable deaths in the veteran study occurred below 12.0 micrograms, a meaningful share of that same mortality risk persists below 9.0 as well — the newer standard did not identify, and was not designed around, an exposure floor below which the risk disappears, because the underlying research suggests no such floor exists in the range these studies can measure. A tighter legal limit reduces the population living in the very highest-risk exposure bands; it does not resolve the more unsettling part of the finding, which is that "meeting the standard" and "the air not killing you" are not the same claim.

And Then You're Dead

Nobody's death certificate says "particulate matter." It says hypertension, or stage 5 chronic kidney disease, or a stroke that came without warning — the same downstream diagnoses that would have appeared on that certificate regardless of what was in the air outside. The particulate matter is the thing that made an ordinary disease arrive years earlier than it otherwise would have, in air that, by law, was already fine.

That's it. That's the whole thing.

Comments

Loading…

Comments are anonymous and moderated.

Sources

Photo: "Smoke from Industry Creates a Haze Which Lingers over the City of Pittsburgh, Pennsylvania," by John L. Alexandrowicz, EPA DOCUMERICA project, 1973, courtesy of the National Archives and Records Administration (NARA ID 557233), via Wikimedia Commons (public domain, U.S. government work).