when you look at NHANES II (1976 to 1980) blood-lead by birth year, the curve runs nearly flat at 14 to 15 micrograms per deciliter from the 1910 cohort all the way through the 1960 cohort. that flatness is the question. it cannot be the case that a child born in 1925 was getting the same daily lead dose as a child born in 1968. so what is the flat line actually summing? the answer turns out to be four separate lead eras layered on top of one peak-exposure moment in time, and the cohort structure only resolves when you stop looking at one survey round and start looking at all ten.
this is a page for the person who has been told the United States solved its lead problem and cannot reconcile that with what they see in the data. specifically: the NHANES record across older adults today is high relative to their younger relatives, and the older the adult, the higher it tends to be. the popular story (leaded gasoline peaked in the 1970s, banned in 1996, problem mostly retired) does not predict that pattern. a 90-year-old today was already 45 when the phasedown started. they should be fine by any model that treats 1980 as the watershed.
they are not fine. and the reason is that the NHANES blood-lead measurement is not a record of what one person was exposed to in one decade. it is a snapshot of a body that has been collecting lead from every available source since birth, sequestering most of it in bone, and then leaking it back out across the second half of life. for a person born before 1965 the sources were many and they overlapped. the cross-sectional table hides all of that.
what follows walks through the actual mechanics. every claim is anchored to a paper you can read.
start with the empirical observation. in the unified NHANES record across ten survey rounds from 1976 through 2023 (n = 99,961 blood-lead measurements), the NHANES II round shows mean adult blood lead at about 14.6 µg/dL. when you stratify that round by birth year, the curve is almost flat across every cohort born from roughly 1910 to 1960. that is the plateau.
the reason the plateau is flat in 1976 to 1980 is not that the cohorts had identical lifetime exposures. the reason is that every adult tested in 1976 to 1980 was being measured at the contemporaneous peak of leaded-gasoline atmospheric lead, peak lead-soldered plumbing, peak lead-painted housing stock, and peak occupational lead in the working-age cohorts. the contemporaneous signal was so loud it drowned out the cohort-of-origin differences. that signal had a half-life in blood of about 30 days (Rabinowitz, 1991), so what NHANES II measured in 1978 was mostly what those people had been breathing and ingesting that same month.
once the gasoline phasedown took effect, the contemporaneous signal collapsed. in the 2021 to 2023 round the mean adult blood lead is about 1.0 µg/dL, a 93 percent reduction. and now, with the loud signal gone, the cohort structure becomes visible. the same NHANES curve plotted by birth year shows older Americans (born before 1965) with measurably higher blood lead than middle-aged Americans born in the 1980s or later. they are not currently being exposed. their skeletons are releasing what they stored sixty years ago.
the plateau in NHANES II is not flat because the cohorts had the same childhoods. it is flat because in 1978 everyone was being measured under the same fire hose of contemporaneous gasoline air. when the fire hose turned off, the cohort signal resolved underneath it.
a person born in 1925 did not live through one lead era. they lived through four, stacked on top of each other, each with its own pathway and its own peak.
a person born in 1925 ate three of these eras (Victorian residue, paint peak, leaded gasoline) before they hit middle age, and a fourth (occupational) on top of any working life in industry. a person born in 1970 mostly got hit by one of these (leaded gasoline childhood) hard, plus residual paint and dust. different shape of exposure, different total dose, different distribution across the body. but the 1976 NHANES II snapshot reads them all out at about the same blood-lead concentration because in 1978 they were all breathing the same air.
the same cohort logic from this article, as a working widget. change the outcome series and watch the integrated childhood-lead curve track it. this is the cohort signal the NHANES II plateau hides.
children cooked by lead (house paint, dominant for births roughly 1910 to 1955, plus leaded gasoline peaking around 1972) became the adults who acted out decades later, each in the calendar year they reached the age that behavior happens. this is an ecological association across birth cohorts, not a claim about any individual.
childhood lead was a high flat plateau through 1976 and then a cliff, so the only defensible claim here is the post-1976 synchronized decline as cleaner cohorts age in (the canonical Nevin / Reyes result), never a mid-century rise. series that do not fit are kept and sequestered with a stated reason, not deleted. that honesty is part of the argument.
panel 1
one cohort, marched through its life by age
the 1971 to 1976 birth cohort (peak childhood lead). each dot is the calendar year that cohort reached the modal age for that outcome. read left to right: school-leaving and young-offender age around 1990, fraud and financial-failure age in the 2000s, midlife now. green = direction matched the lead hypothesis, amber = it did not (kept, not hidden).
panel 2
cohort overlay explorer
panel 3
correlation: CORE (headline) vs SEQUESTERED (kept, reasoned)
the headline rests on the strong CORE fits, sorted top to bottom by strength: robbery (r=+0.93), fraud-perpetration arrests (+0.92), cigarettes per capita (+0.89), high-school dropout (+0.85), motor-vehicle death rate (+0.82). bar length is the Pearson r between the outcome and the integrated childhood blood lead of the birth cohort that produces it at its modal age.
panel 4
the driver: US childhood blood lead, 1900 to 2023
integrated paint plus gasoline childhood blood-lead profile (Annest 1983 / NHANES II 1976-80 / Pirkle 1994 / CDC NHANES). a high flat plateau near 16 ug/dL from 1930 to 1976, then the steep post-phaseout cliff. because the input is flat then falls, the cohort signal can only speak to the decline.
defensibility. these are ecological (population) associations between an outcome and the integrated childhood blood lead of the birth cohort that produces it at its modal age. association is not proof. lead is treated as a contributing, necessary-not-sufficient factor, not a sole cause. because childhood lead was high and flat before 1976, the defensible claim is the post-phaseout decline, not any mid-century rise. sequestered series are retained with a stated mechanism, never deleted. sources: recovered Reuters lead data; EPA EJScreen; CDC NCHS; FBI UCR and BJS; US Courts; Federal Reserve; NHTSA; NCES; Census; NIAAA; Fraser, Murderland (2025); driver Annest 1983 / NHANES / Pirkle 1994 / CDC.
this is the part that resolves the apparent paradox, and it is the reason Eric's question is the right question to ask. the blood-lead measurement is a snapshot of the circulating pool, which equilibrates with current exposure on a timescale of about thirty days (Rabinowitz, 1991). the bone is the integrating compartment. cortical bone has a half-life on the order of decades. trabecular bone is faster, but still measured in years.
roughly 90 to 95 percent of an adult's total body lead burden is in the skeleton. that is where the cohort-of-origin record actually lives. the synthesis paper is Hu et al., 1998, which made the case that bone lead measured by K-XRF (a noninvasive nuclear technique calibrated by Aro et al., 1994) is the right exposure biomarker for chronic toxicity.
when bone lead is measured directly, the cohort gradient that the blood-lead plateau hides shows up clearly. the Normative Aging Study in Boston (mostly men born between 1910 and 1940) was the cohort that built this literature. three findings from that work speak directly to Eric's question:
first, Kim et al., 1997 documented a clear age-and-secular trend in tibia bone lead in middle-aged and elderly men. older cohorts carried higher bone lead than younger ones. the cohort effect was visible in bone where it had been invisible in blood.
second, Weisskopf et al., 2009 showed that bone lead, not blood lead, predicted all-cause and cardiovascular mortality in those men over more than a decade of follow-up. blood lead at one measurement underestimated the risk. bone lead, which integrated decades of stored exposure, did not.
third, Silbergeld et al., 1988 established that postmenopausal bone resorption releases stored skeletal lead back into the blood at a rate that measurably elevates blood lead and contributes to hypertension risk. that is why elderly women in particular show a late-life rise in blood lead that has nothing to do with current external exposure. it is the bone giving back what it took in 1955.
the lead in your 80-year-old parent's blood today is not coming from this year. most of it is coming from 1955.
"bone lead, integrated over decades of past exposure, is a better predictor of lead-related health endpoints in adulthood than concurrent blood lead, which reflects recent exposure plus a fraction of skeletal remobilization."
this is where Eric's intuition needs a small adjustment. the 1920 to 1940 cohort lived through more eras (Victorian, paint, gasoline, occupational), but the per-year intensity of childhood lead exposure peaked later, in the 1965 to 1980 window. childhood lead absorption is severalfold more efficient than adult, and the developing brain is the most vulnerable target.
Eric's NHANES paper places the maximum per-person childhood IQ loss in the 1973 birth cohort, at about 6.4 points lost per child (Lanphear et al., 2005 dose-response applied to era-specific NHANES blood-lead distributions). that cohort was 0 to 7 years old at the absolute peak of atmospheric lead deposition. they took the worst single-era dose of any twentieth-century US cohort.
so the answer is two-handed. the 1925 cohort took longer total exposure across more pathways. the 1965 to 1975 cohort took the most concentrated childhood dose during the most vulnerable developmental window. both end up at high bone lead by age 70, but for different reasons. the NHANES II plateau averages them together visually. the cohort-stratified analysis pulls them apart.
this is the difference that runs through all of our work. the question that matters is dose across a life, not parts-per-billion theater on one product label. find the source, count the µg per day, do not perform a number.
the empirical observation that Eric's NHANES paper puts most weight on is that the dementia wave is still ascending. cardiovascular mortality attributable to lead peaked between roughly 1970 and 1985 and is now declining as the highest-exposure cohorts pass through that age window. but dementia manifests with a 50-year latency from childhood exposure, and the 1965 to 1975 cohort (with the heaviest skeletal depot) is just now entering the 60-to-75 age window where dementia incidence accelerates.
the bone-lead-to-dementia connection was directly measured for the first time by Wang et al., 2026, who reported in a prospective NHANES cohort follow-up that bone lead (estimated patella concentration) was associated with all-cause dementia incidence at hazard ratios up to 2.15 comparing the top quartile to the bottom. that same study found no statistically significant association between current blood lead and incident dementia. the etiology is in the cumulative bone, not in the snapshot blood. Eric's paper applies those hazard ratios to the projected bone-lead distribution of the United States population and projects the annual lead-attributable dementia death rate to peak between 2025 and 2035 as the baby boom cohort ages into the 75-to-85 window.
which is the operational point of all of this. the blood-lead snapshot of an 80-year-old today is not actionable as a measure of what they got. it is the bone, and the bone is now leaking. the consequences of the lead they got in 1955 are scheduled for 2030.
none of this is meant to leave a reader with no next step. there is a stack of free tools on this site for exactly the question of where in this story you sit.
if you only do one thing this week, run the zip screener for the house you grew up in, and again for the house you live in now. the contrast between the two is usually the whole story.
if you want the full underlying NHANES pack and the next pieces of cohort math as they go up, join the Lead Safe University list. one email, no spam, the data goes to you first.
the bone lead in a 90-year-old is not a story about the past. the painted walls of the house they raised their kids in are still painted. the dust that comes off when a window opens has the same lead-pigment fingerprint it had in 1955. the FluoroSpec kit uses perovskite chemistry to flag lead on a specific surface in your house today, the kind of answer no historical chart can give you about the actual glass of water or the actual windowsill in front of you. one customer, Daniella, went from a child at 3.4 µg/dL to undetectable in four months once she tested the actual dishes and pulled the right one. that is what finding the source does.
see the kitone quiet detail. my mom, Karen, was born in 1947. she spent her childhood in pre-war housing stock with original paint, breathed peak gasoline air through her twenties and thirties, raised three kids in a town with mostly pre-1960 plumbing. on the cohort-stratified NHANES math, her bone-lead burden today is in the upper quartile of her birth cohort. that is the cohort that is now in the dementia-risk window. she is the reason the projection through 2035 is not abstract to me. it is also the reason this company exists. I built FluoroSpec to give parents the best data, the best tools, the best chemistry, everything aimed at letting families actually see lead in their own homes instead of inheriting it quietly. the kit and the data tools are for parents who would rather know than not, and the science is finally good enough that knowing is possible.
if you came in from the home quiz or the zip screener, the code LAUNCH10 still works at checkout.
