Long-term residential exposure to PM2.5, PM10, black carbon, NO2, and ozone and mortality in a Danish cohort

This was an observational cohort study, not an experiment. The researchers tested whether long-term residential exposure to five air pollutants — PM2.5 (fine particles <2.5 micrometres), PM10 (coarse particles <10 micrometres), black carbon (soot from combustion), nitrogen dioxide (NO2, traffic-related gas), and ozone (O3) — was associated with mortality from all causes, cardiovascular disease (CVD), and ischaemic heart disease (IHD).

The primary comparison was between people living in areas with higher versus lower concentrations of each pollutant. The researchers also adjusted for road traffic noise to see whether air pollution effects were independent of noise — a common confound because both come from traffic.

The outcome measures were:

All-cause mortality (death from any cause)

Cardiovascular disease mortality (death from heart attack, stroke, etc.)

Ischaemic heart disease mortality (death from narrowed heart arteries)

The study used the Danish Diet, Cancer and Health cohort, which recruited 57,053 participants aged 50–64 years at baseline (1993–1997) from the greater Copenhagen and Aarhus areas. Participants were cancer-free at enrolment. After excluding those with missing address history or pollutant data, the final analytic sample was 49,564 individuals.

Key characteristics:

47% male, 53% female

Age range at baseline: 50–64 years (mean ~56 years)

All were born in Denmark (homogeneous genetic and cultural background)

Follow-up lasted from baseline until 2015 (up to 22 years)

During follow-up, 9,329 deaths occurred from all causes, 3,563 from CVD, and 1,207 from IHD

This is a large, well-characterised cohort with very long follow-up, but it is not representative of younger adults, non-Danish populations, or people with pre-existing disease at baseline.

Air pollution exposure was estimated using a sophisticated modelling approach, not personal monitors:

**PM2.5, PM10, black carbon, NO2, and O3** were modelled using the Danish AirGIS system, which combines:

- A dispersion model (predicts pollution based on traffic, meteorology, and background levels)

- Land use data (e.g., proximity to roads, green space)

- Address history for each participant (updated every year from 1993–2015)

Modelled concentrations were assigned to each participant's residential address for each year of follow-up

Exposure was averaged over the entire follow-up period (time-weighted average)

Mortality was identified through the Danish Civil Registration System and the Danish Cause of Death Registry, which are complete and reliable.

Road traffic noise was measured using the Nordic prediction method, modelled as Lden (day-evening-night level) at each address.

**Why this matters:** Modelled exposure is less precise than personal monitoring (e.g., wearing a PM2.5 sensor), but it captures long-term average exposure across decades, which is what matters for chronic disease. Personal monitors would be impractical for 22 years in 50,000 people.

**Study design:** Prospective observational cohort study.

**Key design features:**

Participants were recruited at baseline (1993–1997) and followed forward in time until death, emigration, or end of follow-up (2015)

Exposure was assigned based on residential address history, not randomised

Statistical models used Cox proportional hazards regression, which estimates the hazard ratio (HR) — the relative risk of death per unit increase in pollutant concentration

Models were adjusted for: age (as time scale), sex, smoking status (never, former, current), smoking intensity (grams/day), smoking duration (years), alcohol intake (g/day), physical activity (yes/no), education level, occupation, fruit/vegetable intake, and road traffic noise (Lden)

Two main models: Model 1 adjusted for age, sex, and lifestyle factors; Model 2 additionally adjusted for road traffic noise

**What this design can prove:**

Association between long-term air pollution exposure and mortality

Dose-response relationships (higher exposure = higher risk)

Independence from noise (if the association persists after noise adjustment)

**What this design cannot prove:**

Causation. This is observational, not randomised. People are not randomly assigned to live in high- vs low-pollution areas. Those living in polluted areas may differ in many ways (socioeconomic status, diet, healthcare access) that also affect mortality.

Direction of causality. While exposure precedes death (temporality is established), residual confounding by unmeasured factors (e.g., stress, diet quality, healthcare access) could explain the findings.

Mechanism. The study cannot tell us *how* air pollution causes death (e.g., inflammation, oxidative stress, plaque formation).

**Major methodological strengths:**

Very large sample (nearly 50,000)

Very long follow-up (22 years)

Complete mortality ascertainment (no loss to follow-up)

Detailed adjustment for smoking (pack-years, duration, intensity) — crucial because smokers also tend to live in more polluted areas

Adjustment for road traffic noise (a major confound)

Address history updated annually (reduces exposure misclassification)

**Major methodological weaknesses:**

Exposure is modelled, not measured personally (measurement error, likely non-differential, which biases results toward the null — meaning true effects may be larger)

No adjustment for indoor air pollution (e.g., cooking, smoking indoors, wood stoves) — this could confound results if indoor sources correlate with outdoor levels

No adjustment for time spent outdoors or commuting (personal exposure depends on where you actually are)

Single-country, homogeneous population (limits generalisability)

No data on pre-existing disease progression or medication use during follow-up

All results are hazard ratios (HR) with 95% confidence intervals (CI) per specified increment in pollutant concentration. Model 2 (adjusted for noise) is the primary analysis.

**All-cause mortality (primary outcome):**

PM2.5 (per 5 µg/m³ increase): HR = 1.15 (95% CI: 1.08–1.23) — 15% increase in risk

PM10 (per 10 µg/m³ increase): HR = 1.10 (95% CI: 1.04–1.16) — 10% increase

Black carbon (per 1 µg/m³ increase): HR = 1.08 (95% CI: 1.03–1.14) — 8% increase

NO2 (per 10 µg/m³ increase): HR = 1.23 (95% CI: 1.15–1.31) — 23% increase

O3 (per 10 µg/m³ increase): HR = 0.97 (95% CI: 0.93–1.01) — not statistically significant (p > 0.05)

**Cardiovascular disease mortality (secondary outcome):**

PM2.5 (per 5 µg/m³): HR = 1.19 (95% CI: 1.08–1.31) — 19% increase

PM10 (per 10 µg/m³): HR = 1.13 (95% CI: 1.03–1.24) — 13% increase

Black carbon (per 1 µg/m³): HR = 1.10 (95% CI: 1.01–1.19) — 10% increase

NO2 (per 10 µg/m³): HR = 1.30 (95% CI: 1.17–1.44) — 30% increase

O3 (per 10 µg/m³): HR = 0.95 (95% CI: 0.89–1.01) — not significant

**Ischaemic heart disease mortality (secondary outcome):**

PM2.5 (per 5 µg/m³): HR = 1.21 (95% CI: 1.03–1.43) — 21% increase

PM10 (per 10 µg/m³): HR = 1.15 (95% CI: 1.00–1.33) — 15% increase (borderline significant)

Black carbon (per 1 µg/m³): HR = 1.13 (95% CI: 0.99–1.29) — not significant

NO2 (per 10 µg/m³): HR = 1.31 (95% CI: 1.11–1.55) — 31% increase

O3 (per 10 µg/m³): HR = 0.93 (95% CI: 0.84–1.03) — not significant

**Key nuance:** After adjusting for road traffic noise (Model 2 vs Model 1), the associations for PM2.5, PM10, and black carbon were slightly attenuated but remained statistically significant. For NO2, the association was essentially unchanged. This suggests that the effects of PM2.5 and black carbon are partly confounded by noise, but NO2 effects are independent.

**Dose-response:** The authors tested for non-linearity and found no evidence of a threshold — risk increased linearly across the range of exposures. This means even small reductions in pollution (e.g., 1–2 µg/m³ PM2.5) would be expected to reduce mortality risk.

To put these numbers in perspective:

A 15% increase in all-cause mortality per 5 µg/m³ PM2.5 means that if you live in an area with PM2.5 of 15 µg/m³ (typical for a European city) versus 10 µg/m³ (cleaner suburban area), your risk of dying over 22 years is about 15% higher. In absolute terms, if the baseline risk of death over 22 years is ~19% (as in this cohort), a 15% relative increase means an absolute risk of ~22% — so about 3 extra deaths per 100 people over 22 years.

The NO2 effect was stronger: a 23% increase in all-cause mortality per 10 µg/m³. For context, living next to a major road versus a quiet side street might easily differ by 10–20 µg/m³ NO2. That would translate to a 23–46% increase in mortality risk.

These effect sizes are comparable to the mortality risk of being a former smoker versus never-smoker (roughly 20–30% increase), though much smaller than current smoking (200–300% increase).

The effects were larger for cardiovascular and ischaemic heart disease mortality than for all-cause mortality, consistent with the known cardiovascular toxicity of air pollution.

Ozone showed no association (and a non-significant protective trend), which is unusual. The authors note this may be due to negative correlation between O3 and other pollutants (O3 is lower in traffic-heavy areas) or because O3 was measured at background stations rather than at residences.

**Acknowledged by authors:**

Exposure misclassification: modelled rather than personal measurements

No adjustment for indoor air pollution sources

No data on time spent at home versus away

Residual confounding by socioeconomic status (though education and occupation were adjusted)

Single-country, homogeneous population limits generalisability

Ozone exposure was from background monitoring stations, not modelled at residence

**Critical reader notes:**

The cohort was 50–64 at baseline, so results may not apply to younger adults or children

Participants were all cancer-free at baseline, but the study did not exclude people with CVD at baseline — this could bias results if people with CVD moved to cleaner areas (healthy survivor bias)

No adjustment for changes in smoking or diet during follow-up (people may quit smoking after diagnosis, which could confound)

The effect estimates for PM2.5 (HR = 1.15) are lower than some other European studies (which found HRs of 1.20–1.30), possibly because Denmark has relatively clean air (mean PM2.5 ~12 µg/m³) — effects may be steeper at lower concentrations

The study did not examine cause-specific mortality beyond CVD and IHD (e.g., respiratory mortality, lung cancer)

No data on medication use (e.g., statins, antihypertensives) that could modify the effect of air pollution

The statistical models assume proportional hazards (constant relative risk over time), which may not hold if pollution effects accumulate or wane with age

For someone running their own n=1 experiment to reduce personal air pollution exposure:

**What to test:**

Use a portable PM2.5 monitor (e.g., PurpleAir, AirVisual, or Dylos) to measure your personal exposure at home, at work, and during commuting

Test the effect of one intervention at a time:

- HEPA air purifier in bedroom (run 24/7, close windows)

- Changing commute route to avoid high-traffic roads

- Moving desk away from window facing a busy road

- Using an N95 mask during high-pollution days (PM2.5 > 35 µg/m³)

- Relocating to a lower-pollution neighbourhood (extreme, but testable)

**Minimum meaningful duration:**

For acute effects (e.g., heart rate variability, blood pressure, lung function): 2–4 weeks per condition

For chronic effects (e.g., inflammation markers, oxidative stress): 3–6 months per condition

The study found effects over decades, but measurable physiological changes (e.g., 2–5 mmHg blood pressure reduction) can occur within weeks of reducing PM2.5 by 10–20 µg/m³

**What to measure (specific metrics):**

**Primary:** Resting heart rate (beats per minute, measured same time each morning before getting up), blood pressure (systolic/diastolic, same time/conditions)

**Secondary:** Heart rate variability (HRV, using a chest strap or smartwatch — SDNN or RMSSD), peak expiratory flow (PEF, using a peak flow meter), C-reactive protein (CRP, via finger-prick blood test, monthly)

**Environmental:** Indoor PM2.5 (µg/m³, hourly average), outdoor PM2.5 at nearest monitoring station, temperature, humidity

**Subjective:** Daily symptom log (cough, phlegm, shortness of breath, headache, fatigue on 0–10 scale)

**Key confounds to control for:**

**Season:** Air pollution varies by season (winter inversions, summer ozone). Run each condition for at least 4 weeks in the same season, or use a crossover design (e.g., 4 weeks with purifier, 4 weeks without, repeat)

**Weather:** Temperature and humidity affect both pollution and physiology. Log daily weather and adjust statistically if possible

**Indoor sources:** Cooking (especially frying), candles, incense, wood stoves, smoking — standardise these across conditions (e.g., no candles during study)

**Activity level:** Exercise increases breathing rate and pollution dose. Keep exercise routine constant across conditions

**Time spent outdoors:** Log hours outdoors each day. If you spend more time outdoors during one condition, you'll get higher exposure regardless of indoor air

**Sleep quality:** Poor sleep affects HRV and blood pressure. Track sleep duration and quality (e.g., sleep diary or wearable)

**Diet and alcohol:** Both affect inflammation and blood pressure. Keep diet consistent or log and adjust

**Stress:** Major life events (job loss, relationship stress) can swamp any pollution effect. Note any major stressors

**What a positive result would look like:**

After 4 weeks with an air purifier in your