Research method
Cohort Study
17 studies
A cohort study follows a defined population over time, assigns an exposure (residence, a monitor surface, or a biospecimen), and then counts later outcomes. People are not randomised to polluted water or clean air. Some papers in this set are individual-level Cox or growth models; others are ecological standardised-mortality comparisons of whole municipalities; a few are pregnancy or surgery samples that measure chemicals without a later clinical endpoint.
Environmental epidemiologists reach for cohorts when they need years of follow-up for deaths, cancers, preterm birth or children's growth under real-world pollution. The design answers 'did people who lived here, or who carried this measured burden, have higher rates later?' Its main limitation is that an SMR for a 'Red area' is not an individual PFAS dose, a Cox HR per 10 μg/m³ is not a trial of cleaner air, and a single serum phthalate cannot establish that the chemical came first.
Studies
17
Findings
6
6 supporting · 0 challenging · 0 qualifying citations
Open tensions
2
Currently
What we know
- The Veneto 'Red area' (30 municipalities, ~153,525 people) is an ecological SMR study split around 1985 contamination. From 1985–2018 there were 51,621 deaths versus 47,731 expected (SMR 108; 3,890 excess) and no excess in 1980–1984. Later periods included kidney cancer (SMR 173 in one analysis) and testicular cancer in Red area A (SMR 256 for 1985–1999). Median serum PFOA was 14 ng/mL — a community marker, not a person-level dose.
- Individual-level Cox models in the Canadian Community Health Survey cohort assigned 3-year satellite/land-use PM₂.₅ and followed deaths up to 12 years. The fully adjusted non-accidental HR was 1.26 per 10 μg/m³; respiratory mortality was strongest (HR 1.52). Risk remained elevated below the then-WHO 10 μg/m³ guideline.
- In the Danish Diet, Cancer and Health cohort, time-weighted residential NOₓ from 1971 and traffic load within 200 m were screened against 20 non-lung cancers. After adjustment, cervical and brain cancers associated with residential NOₓ and/or nearby major streets; brain cancer was almost doubled near high-traffic streets. Median NOₓ was 21.9 μg/m³. Many smoking-related IRRs shrank after confounder control, and lung cancer was excluded because it was published separately.
- PROTECT followed 1,092 pregnancies with live singleton births in 21 Puerto Rican municipalities (2011–2018). About 9.1% of births were preterm; mean pregnancy PM₂.₅ was 6.8 μg/m³. An IQR increase raised preterm-birth risk only modestly (RR 1.014 individual-adjusted; RR 1.012 also municipal-adjusted). Phthalate metabolites did not modify the association. Monitor-based, municipality-at-birth assignment likely biases toward the null.
- A school cohort's multilevel BMI growth models (mean age 6.6 years; BMI rose ~2.6 units over 5 years) found non-freeway NOₓ still positively associated with BMI level and slope after SES, smoking, greenness, fast food and street-connectivity adjustment, while freeway NOₓ was not. Traffic density within 150 m was confounded, especially on the Spanish questionnaire.
Largest unresolved question
Papers sharing the cohort label are not one design. Veneto is an ecological SMR for municipalities; CCHS and PROTECT assign person- or pregnancy-level exposures and model individual outcomes; MIREC maps residues without a disease endpoint; the GraMo phthalate–adipose analysis is a cross-sectional surgery sample that cannot establish temporality (WQS β = −30.089 for GSH). 'Cohort evidence that pollution kills' collapses those differences.
Procedure
How it works
The main stages of the method. Specific protocols and analysis choices vary between studies.
Define the population and the start of follow-up
A group is enrolled or reconstructed from registries — residents of municipalities, survey respondents linked to death records, pregnancies, schoolchildren. Who is in the risk set, and when the clock starts, bound every later rate.
Assign an exposure without randomisation
Exposure is taken from residence, a monitor, a satellite or land-use surface, or a biospecimen. People are not randomised to polluted water or clean air. An ecological area label is not an individual dose; a municipality-at-birth assignment is not a personal monitor.
Count later outcomes
Deaths, cancers, preterm birth, growth, or a biomarker are recorded after the exposure window. Some papers stop at concentrations and never record a clinical endpoint.
Adjust for measured confounders
Cox, growth, or standardised-mortality models hold measured covariates fixed. Residual confounding, exposure error, and reverse causation remain; a single serum measurement cannot establish that the chemical came first. Adjustment is not randomisation.
Do not read the follow-up as a trial of cleaner air
A cohort answers whether people who lived here, or who carried this measured burden, had higher rates later. An SMR for a contaminated area or an HR per 10 μg/m³ is not evidence that cleaning the air would have prevented those events.
Common misconceptions
A pollution cohort is basically an RCT of living in dirty air.
No one was randomised. Veneto compares municipal death rates; CCHS and PROTECT adjust for measured covariates and still leave residual confounding, mobility and indoor sources on the table. An HR or SMR is not an assigned treatment effect.
Median serum PFOA of 14 ng/mL in the Red area means each extra death is tied to that person's PFAS dose.
The mortality analysis is ecological. Private wells misclassify some Red area B residents, food export may expose people outside the zone, and the SMR does not use individual PFAS measurements as the exposure.
If a pregnancy cohort measures chemicals in blood, it has shown those chemicals caused disease.
MIREC reports residue levels and predictors (parity, fish, foreign-born, sampling year), not clinical outcomes. GraMo's phthalate–redox associations are same-time surgery measurements; short-lived phthalates make one serum sample a noisy exposure metric.
Related
Claim ledger
What the evidence shows
Drawn from 17 studies in this library. Mix labels say which citation roles are present; they are not a strength score. Supports means evidence for a finding; Challenges means evidence against a stated position; Qualifies marks scope.
The Veneto 'Red area' (30 municipalities, ~153,525 people) is an ecological SMR study split around 1985 contamination. From 1985–2018 there were 51,621 deaths versus 47,731 expected (SMR 108; 3,890 excess) and no excess in 1980–1984. Later periods included kidney cancer (SMR 173 in one analysis) and testicular cancer in Red area A (SMR 256 for 1985–1999). Median serum PFOA was 14 ng/mL — a community marker, not a person-level dose.
Individual-level Cox models in the Canadian Community Health Survey cohort assigned 3-year satellite/land-use PM₂.₅ and followed deaths up to 12 years. The fully adjusted non-accidental HR was 1.26 per 10 μg/m³; respiratory mortality was strongest (HR 1.52). Risk remained elevated below the then-WHO 10 μg/m³ guideline.
In the Danish Diet, Cancer and Health cohort, time-weighted residential NOₓ from 1971 and traffic load within 200 m were screened against 20 non-lung cancers. After adjustment, cervical and brain cancers associated with residential NOₓ and/or nearby major streets; brain cancer was almost doubled near high-traffic streets. Median NOₓ was 21.9 μg/m³. Many smoking-related IRRs shrank after confounder control, and lung cancer was excluded because it was published separately.
PROTECT followed 1,092 pregnancies with live singleton births in 21 Puerto Rican municipalities (2011–2018). About 9.1% of births were preterm; mean pregnancy PM₂.₅ was 6.8 μg/m³. An IQR increase raised preterm-birth risk only modestly (RR 1.014 individual-adjusted; RR 1.012 also municipal-adjusted). Phthalate metabolites did not modify the association. Monitor-based, municipality-at-birth assignment likely biases toward the null.
A school cohort's multilevel BMI growth models (mean age 6.6 years; BMI rose ~2.6 units over 5 years) found non-freeway NOₓ still positively associated with BMI level and slope after SES, smoking, greenness, fast food and street-connectivity adjustment, while freeway NOₓ was not. Traffic density within 150 m was confounded, especially on the Spanish questionnaire.
MIREC measured POPs in first-trimester and cord plasma from 1,983 pregnant women in ten Canadian cities (2008–2011). PFOS had the highest maternal PFAS geometric mean (4.56 μg/L); maternal PFOA was four times cord PFOA (1.65 vs 0.35 μg/L); median detections were 14 POPs in maternal plasma versus 2 in cord plasma. The paper reports concentrations and predictors, not clinical health effects.
Debates
Tensions and limits
Some items are genuine disagreements on the same question. Others mark different assays, populations, or outcomes.
Papers sharing the cohort label are not one design. Veneto is an ecological SMR for municipalities; CCHS and PROTECT assign person- or pregnancy-level exposures and model individual outcomes; MIREC maps residues without a disease endpoint; the GraMo phthalate–adipose analysis is a cross-sectional surgery sample that cannot establish temporality (WQS β = −30.089 for GSH). 'Cohort evidence that pollution kills' collapses those differences.
Papers sharing the cohort label are not one design. Veneto is an ecological SMR for municipalities; CCHS and PROTECT assign person- or pregnancy-level exposures and model individual outcomes; MIREC maps residues without a disease endpoint; the GraMo phthalate–adipose analysis is a cross-sectional surgery sample that cannot establish temporality (WQS β = −30.089 for GSH). 'Cohort evidence that pollution kills' collapses those differences.
- PFAS-contaminated water linked to excess Veneto deaths
- Low PM2.5 still tracks Canadian mortality
- Prenatal PM2.5 linked to preterm birth in Puerto Rico
- Canadian pregnancy cohort maps POP levels in blood
- Serum phthalates track adipose redox markers
Study Role Design N Population Outcome PFAS-contaminated water linked to excess Veneto deaths Supports CohortSMR comparison of Veneto PFAS-contaminated Red area mortality vs surrounding provinces N=153525 · Red area population ~153,525 (2020); analyses of 51,621 deaths 1985–2018 Residents of 30 Veneto municipalities with PFAS drinking-water contamination All-cause, CVD, and cancer mortality SMRs after water contamination Low PM2.5 still tracks Canadian mortality Supports CohortCCHS cohort Cox models of mortality vs low ambient PM2.5 N=299500 · 299,500 Canadian Community Health Survey respondents after exclusions Canadian adults in the CCHS linked to ambient PM2.5 Non-accidental and cause-specific mortality per 10 μg/m³ PM2.5 Prenatal PM2.5 linked to preterm birth in Puerto Rico Supports CohortPROTECT pregnancy cohort with pregnancy-average PM2.5 and preterm-birth risk ratios N=1092 · 1,092 PROTECT participants with live singleton births in 21 municipalities Pregnant women in the PROTECT cohort in Puerto Rico Preterm birth associated with prenatal PM2.5 (with phthalate-interaction tests) Canadian pregnancy cohort maps POP levels in blood Supports CohortMIREC pregnancy cohort measuring POPs/PFAS in maternal and cord plasma N=1983 · 1,983 pregnant women in 10 Canadian cities (2008–2011) Canadian pregnant women and newborns in MIREC Maternal and cord POP/PFAS concentrations and demographic correlates Serum phthalates track adipose redox markers Supports Cross-sectionalGraMo surgery patients with serum phthalates and adipose oxidative-stress biomarkers N=308 · 308 with adipose redox markers; 143 with overlapping exposure and markers; 230 with serum phthalates Spanish adults undergoing surgery in the GraMo study Adipose oxidative-stress biomarkers associated with serum phthalates and mixtures
PaperFren reads this as a limit on how far one study travels — different assays, populations, or outcomes — not a forced fight between papers.
Traffic signals split by source and by how hard confounding is controlled. Children's BMI tracked non-freeway NOₓ but not freeway NOₓ, and near-road traffic density was confounded. In the Danish cancer screen, cervical and brain associations remained after adjustment while many smoking-related IRRs shrank, and the analysis was a 20-cancer screen without HPV data for cervix.
Traffic signals split by source and by how hard confounding is controlled. Children's BMI tracked non-freeway NOₓ but not freeway NOₓ, and near-road traffic density was confounded. In the Danish cancer screen, cervical and brain associations remained after adjustment while many smoking-related IRRs shrank, and the analysis was a 20-cancer screen without HPV data for cervix.
Study Role Design N Population Outcome Non-freeway NOx tracks children's BMI growth Supports CohortLongitudinal multilevel BMI growth models vs traffic density and CALINE-like NOx N=4550 · 4,550 children in 13 Southern California communities Schoolchildren in Southern California communities BMI level and 4-year growth related to non-freeway vs freeway NOx and near-road traffic Traffic NOx linked to cervical and brain cancer risk Supports CohortDiet, Cancer and Health cohort with modeled residential NOx and nearby traffic load N=54304 · 54,304 cohort members followed ~9.6 years (from 57,053 recruited) Danish adults aged 50–64 in Copenhagen and Aarhus areas Incidence of 20 non-lung cancers related to residential traffic pollution
PaperFren reads this as a limit on how far one study travels — different assays, populations, or outcomes — not a forced fight between papers.
Timeline
How understanding moved
Study years are when the paper was published. Evidence edits are dated changes to this page's claims. Explanations are when PaperFren added a Discovery — not a claim that the science happened that day.
2026
Concept page published
Cohort Study
Change log
What changed
Dated edits to this page's evidence: studies added or removed from a claim, claims added or withdrawn, and new explanations tagged here. Coverage edits are distinct from reviewed research changes and corrections. Dates record PaperFren updates; study publication dates appear in the timeline.
- Coverage updated · Concept page published
Papers
17 studies in this library bear on Cohort Study, ordered by citations.
- Non-freeway NOx tracks children's BMI growth
In 4,550 Southern California children, non-freeway NOx associated with BMI at age 10 and 4-year growth; freeway NOx and near-road traffic density were weaker after confounding control.
- Rome NO2 maps 12 years apart still predict deaths
Land-use regression NO2 surfaces for Rome in 1995/96 and 2007 were highly correlated, and both predicted higher natural-cause mortality in a 2001–2006 adult cohort.
- Roads and air pollution track neurologic disease
In Metro Vancouver, living near roads raised incidence of non-Alzheimer dementia, Parkinson’s, Alzheimer’s, and MS; air pollution linked to NAD and PD, noise was null.
- Low PM2.5 still tracks Canadian mortality
In 299,500 Canadians, a 10 μg/m³ rise in PM2.5 was linked to HR 1.26 for non-accidental death even at mean 6.3 μg/m³.
- Traffic NOx linked to cervical and brain cancer risk
In 54,304 Danish adults, long-term traffic-related NOx at home was associated with cervical and brain cancer, with a weaker liver-cancer traffic signal.
- Traffic PM and nearby roads raised diabetes incidence
In the Heinz Nixdorf Recall cohort, living within 100 m of a busy road and higher total PM10 were associated with more new type 2 diabetes over ~5 years.
- Canadian pregnancy cohort maps POP levels in blood
MIREC found many POPs in maternal plasma but few in cord blood; parity, age, fish, and birthplace predicted levels.
- Prenatal PFAS tracked 5-year child overweight risk
In 412 Scandinavian mother–child pairs, higher maternal PFAS associated with higher child BMI and triceps skinfold z-scores and with overweight/obesity at age 5, with geographic differences.
Show 9 more studiesShow fewer studies
- Pregnancy PFAS and lead track lower birth weight
In Cincinnati’s HOME Study, PFAS, lead, and organophosphate pesticides associated with lower birth weight; phthalates and BPA did not.
- Serum PFAS mixes reveal seafood vs product exposure
Long-chain PFCAs mark seafood/whale diets, while PFHxS and N-EtFOSAA mark carpets and consumer products in children.
- PFAS in Veneto youth tracked with higher blood pressure
In 16,224 Veneto adults aged 20–39 exposed via contaminated drinking water, higher serum PFOA, PFOS, and PFHxS associated with hypertension, with the signal concentrated in men.
- PFAS-contaminated water linked to excess Veneto deaths
Residents of 30 Italian municipalities on a PFAS-contaminated aqueduct had about 3,890 extra deaths from 1985–2018, with later kidney- and testicular-cancer signals.
- Canopy nitrogen and tree carbon uptake
Canopy-applied 15N is recovered aboveground about three times more than soil-applied N.
- Prenatal PM2.5 linked to preterm birth in Puerto Rico
In the PROTECT birth cohort, an IQR rise in pregnancy-average PM2.5 was associated with higher preterm-birth risk among relatively healthy Puerto Rican women.
- Childhood phthalates linked to teen liver enzymes
In the Ewha Birth and Growth Cohort, higher childhood urinary phthalates—especially later LMW exposure—tracked with higher AST, ALT, and γ-GTP at ages 10–15, with BMI interactions in girls.
- Phthalates track weaker grip in Korean elders
Higher urinary DEHP and MnBP metabolites were inversely associated with handgrip strength in community-dwelling older adults, especially with a high omega-6/omega-3 diet.
- Serum phthalates track adipose redox markers
In 308 Spanish adults, low-molecular-weight phthalates associated with higher GPx and TBARS in fat, while a phthalate mixture index associated with lower GSH and GSSG.
Compare studies
Select 2–10 studies. Design and N are labels, not a ranking.
Nothing selected yet.
Questions
What is still open
Papers sharing the cohort label are not one design. Veneto is an ecological SMR for municipalities; CCHS and PROTECT assign person- or pregnancy-level exposures and model individual outcomes; MIREC maps residues without a disease endpoint; the GraMo phthalate–adipose analysis is a cross-sectional surgery sample that cannot establish temporality (WQS β = −30.089 for GSH). 'Cohort evidence that pollution kills' collapses those differences.
Traffic signals split by source and by how hard confounding is controlled. Children's BMI tracked non-freeway NOₓ but not freeway NOₓ, and near-road traffic density was confounded. In the Danish cancer screen, cervical and brain associations remained after adjustment while many smoking-related IRRs shrank, and the analysis was a 20-cancer screen without HPV data for cervix.
Ask PaperFren about Cohort Study
Study this conceptflashcards and short-answer questions
Veneto reports SMR 108 (51,621 observed vs 47,731 expected) after 1985 and no excess in 1980–1984. What does the calendar split buy you, and what does it still not prove?
The split is a before/after check around documented contamination: excess deaths appear after 1985, not in the short pre-period. It still does not assign individual PFAS doses, and well misclassification plus food export can move exposure across the Red-area boundary.
CCHS finds a fully adjusted non-accidental HR of 1.26 per 10 μg/m³ PM₂.₅, remaining elevated below the then-WHO 10 μg/m³ guideline. Why is that not the same claim as 'PM₂.₅ below 10 is proven to kill'?
It is an observational Cox contrast with wide uncertainty on the low-concentration spline, incomplete mobility capture, and possible residual confounding from smoking intensity or indoor sources. Risk remaining elevated below a guideline is a concentration–response finding, not a randomised demonstration of a safe threshold.
PROTECT's preterm RR is 1.014 per IQR at mean pregnancy PM₂.₅ of 6.8 μg/m³, with about 9.1% preterm births. Why might that underestimate a true effect?
Assignment is monitor-based and municipality-at-birth, so residential moves and fine-scale PM₂.₅ are missed; that exposure error typically biases toward the null. The cohort also excluded major comorbidities and had few preterm births for interaction tests.
A classmate treats MIREC (maternal PFOA 1.65 vs cord 0.35 μg/L; 14 vs 2 median POP detections) as evidence that PFAS caused infant disease. What did the study actually report?
Concentrations and demographic/diet predictors in a convenience sample of relatively educated, mostly Canadian-born women — not clinical endpoints. Cord:maternal ratios are only for pairs both above the LOD.