Air Pollution Impairs Marathon Performance: A Cross-Sectional Analysis of 2.7 Million Finishers Across Six World Marathon Majors
Sports medicine - open · 2026
Higher race-day NO2 is linked to slower marathon times, especially in recreational runners, but this is an ecological correlation across only 75 race-days, not proof of a direct causal hit.
The paper
Cross-sectional observational analysis, n=2,756,553 finishing times (1,815,188 with age data) from six World Marathon Majors, 2010-2024.
What they found
Each 1 µg/m³ rise in race-day NO₂ was associated with finishing times 1.43 min slower in men (95% CI 0.66-2.21) and 1.56 min slower in women (95% CI 0.63-2.48); PM2.5 showed no significant association (men -0.13, 95% CI -0.67 to 0.40; women -0.14, 95% CI -0.76 to 0.49). Recreational runners' times were roughly six times more sensitive to NO₂ than elites (2.12 vs 0.34 min/µg/m³).
Study design
Retrospective cross-sectional observational study of 2.7 million individual marathon finishers across 75 marathon-years (six World Marathon Majors, roughly 15 years), using multilevel regression with a random intercept for event to relate finishing time to race-day ambient air pollution. This sits low on the hierarchy: it is registry-based observational data, not a trial, so it can show association only.
Methodology
Air pollution exposure comes from CAMS modelled reanalysis assigned once per event, meaning the entire NO2/PM2.5 predictor really only varies across 75 city-days despite the huge runner-level N, so this is closer to a 75-unit ecological comparison than a true individual-exposure study, and personal exposure (route, pace, time on course) is not measured. The random intercept for event cannot fully rule out city-day confounders that travel with pollution, such as course profile, historical field quality, or humidity, which the authors did not adjust for, so some of the NO2 effect may reflect which cities or conditions tend to produce slower races generally rather than a direct physiological hit. Finishing time itself is objectively measured by chip timing, which limits outcome misclassification, and the elite-versus-recreational gradient was assessable directly from reported subgroup coefficients (2.12 vs 0.34 min/µg/m3), giving a useful, low-cost sanity check that a clinician can look for in any future replication. No randomisation, blinding, or control over confounders was possible given the design, and with two pollutants and multiple subgroups tested, there is some risk that the NO2 result was emphasised over the null PM2.5 finding.
The appraisal
The NO2 association is well estimated statistically (CIs clear of zero) and shows a genuinely useful internal consistency check: recreational runners, who spend far longer exposed on course, show roughly six times the effect of elites, and this gradient survives even when expressed as a percentage of finishing time. A diffuse city-day confounder (heat, humidity, field quality) would be expected to slow everyone by a similar proportion, not hit recreational runners six-fold harder, so this pattern does argue for a real exposure-duration effect rather than pure coincidence. That said, this remains association, not causation, and the size of the effect needs stress-testing: NO2 was measured once per event, so the exposure variable itself only varies across 75 marathon-years, however large the runner-level N. Across six cities and 15 years, real-world NO2 differences are easily 20-30 µg/m3, and the fitted slope implies pollution alone could add 30+ minutes to recreational finishing times between the cleanest and dirtiest race days, a large effect for NO2's known airway physiology and one that leaves room for city-level confounders (course profile, qualifying standards, typical field make-up) that a random intercept for event doesn't fully strip out.
The gap
CAMS is a modelled reanalysis product, not measured personal exposure: it gives ambient background pollution for the city on race day, not what an individual runner actually breathed on their specific route and at their specific pace. Because NO2 only varies between events, not within them, the model is essentially comparing 75 city-days, so it can't fully separate a direct pollution effect from other things that travel with high-NO2 days in a given city, such as course profile, historical field composition, or humidity, which wasn't adjusted for. Individual-level exposure data and adjustment for humidity and 15 years of field-quality drift would sharpen the causal claim considerably; as it stands, some of the NO2 signal may be a marker of which cities/conditions produce generally slower races rather than a direct physiological hit.
Landmark context
This extends earlier work linking air pollution (ozone, PM10) to marathon performance in smaller elite-runner samples, such as analyses correlating ambient pollutant levels with top finishing times across major marathons; this study is a substantial scale-up using mass-participation data rather than elite times alone, and is one of the first to show the recreational-versus-elite exposure gradient explicitly.
What to do Monday
Not something that changes clinical management on Monday, but useful talking points for endurance athletes and coaches: on high-NO₂ race days, recreational runners in particular should expect and plan for slower times, and event organisers/clinicians advising mass-participation runners might factor race-day air quality into pacing or health advice for at-risk (respiratory/cardiac) participants.
In practice
This doesn't change assessment or dosage for anyone, it's a race-day environmental-load conversation. In clinic, when advising a patient with known respiratory or cardiac disease who's training for a mass-participation marathon, add race-day air quality forecast to the usual heat/humidity check-in, and remind them that a slower-than-goal time on a high-NO2 day isn't a fitness failure, it's a known population-level effect, largest for recreational-pace runners who're out on course for hours longer. On the S&C/performance floor, build a simple contingency into pacing plans for endurance athletes racing in polluted cities (this cohort's data comes from Berlin, Boston, Chicago, London, NYC, Tokyo specifically): coach acceptance of a slower target split rather than chasing a fixed goal time when air quality is poor, since fighting the clock against a physiological headwind late in a marathon adds unnecessary risk. Caveat: this is city-level modelled pollution, not what any individual actually breathed, so treat it as a reason to check the forecast and set expectations, not as a precise dose to plan around.
The five papers that matter — every week, free.
We read the whole sports-medicine, rehab & performance literature and appraise the papers worth your time.
Get the free weekly issueRead in your language: English · Français · Deutsch · Español · Italiano · Nederlands · Dansk · Svenska