select search filters
roundups & rapid reactions
factsheets & briefing notes
before the headlines
Fiona fox's blog

air pollution and death

Research in The Lancet suggested exposure to tiny particles of soot or dust may be more deadly below current European Union (EU) air quality limits than previously thought. This before the headlines analysis accompanied roundup comments.



Title, Date of Publication & Journal

Effects of long-term exposure to air pollution on natural-cause mortality: an analysis of 22 European cohorts within the multicentre ESCAPE project, Dec 9th 2013, Lancet


Claim supported by evidence?

The paper provides some evidence that exposure to PM2.5 is associated with a rise in natural-cause mortality, but provides no evidence of a causal relationship between the two (because observational studies cannot demonstrate causality).



  • A well conducted observational study examining a possible link between air pollution and natural-cause mortality.
  • Objective: to determine whether any forms of air pollution exposure is associated with increased natural-cause mortality.
  • This is a meta-analysis of results from 22 cohort studies in Europe.  Overall, the results appear to add some weight to the association between fine particulate air pollution (PM2.5) exposure and natural cause mortality, even at PM2.5 levels below the European guideline.
  • The association is observed in men, but not in women


Study Conclusions

The paper’s main conclusion is “long-term exposure to fine particulate air pollution is associated with natural-cause mortality, even at concentration ranges well below the present European annual mean limit value.”

It appears as though there may be a statistically significant association between increased PM2.5 exposure and increased mortality due to natural causes in men.

The results seem to suggest that even at levels below the European guideline (25 μg/m3), the higher the exposure to PM2.5 the greater the increased mortality rate due to natural causes (i.e. a dose-response).

However, the paper only looks at pollution levels over a short time, so cannot establish whether the association depends on prolonged exposure. The authors acknowledge that “The PM2.5 threshold analysis focused on the concentration distributions [of PM2.5] in 2008-11 and not on cumulative exposures during the follow-up period”.

The authors acknowledge “we did find that pooled HRs* for PM2.5 were raised in men but not in women. No explanation for such sex differences has yet been proposed”.

*HR: Hazard ratio, a measurement of risk of death resulting from being exposed to the exposure(PM2.5)

The authors also studied the effect of five other air pollutants and two measures of traffics (Table 4), and did not find they are associated with increased natural cause mortality.



Uses information from many big cohorts covering many countries in Europe. Each cohort seems to fairly represent the general population at the geographical area they cover.

Uses standard protocol for analysing the data for the different cohorts. The results from the cohorts were then properly combined using meta-analyses.

Many confounders, especially those related to smoking, were accounted for. Although the authors acknowledge that “residual confounding by, for example, smoking, can never be excluded completely”.

The authors used the concentration distributions in 2008-11, and no historical PM2.5 records are available prior to this time.  This is a limitation as it means the pollution exposure of individuals can only be measured over a short period of time.  Because many deaths from the cohorts would have occurred before 2008, the association may therefore be non-causal. In addition, because levels of PM2.5 change over time (decrease in general in Europe), it is not certain whether the study population with exposure level classified as below the European guideline had not had higher exposure levels in the past. This introduces doubts to the claim that below European guideline, increased exposure to PM2.5 was associated with increase natural cause mortality.

The authors reported the main results for the whole population, but there are significant differences between men and women. The association is observed in men, but not in women. The authors did not provide any aetiological explanation for this difference.


‘Before the headlines’ is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry(PSI) and experienced statisticians in academia and research.  A list of contributors, including affiliations, is available here

in this section

filter Headlines by year

search by tag