February 9, 2016

historic air pollution exposure and long-term mortality risks

In one of the longest-running studies of its kind, researchers from Imperial College London have published in the journal Thorax their investigation into the possible effects of long-term air pollution exposure over nearly 40 years on various health outcomes including mortality. This analysis accompanied a briefing.

Title, Date of Publication & Journal

‘Historic air pollution exposure and long-term mortality risks in England and Wales: prospective longitudinal cohort study’ by Anna Hansell et al published in Thorax Tuesday 9 February 2016.

Study’s main claims – and are they supported by the data

The paper does not prove the claim that “air pollution exposure has long-term effects on mortality”

This is an observational study and cannot prove cause and effect. The press release should make it clear that air pollution was associated with mortality risk from cardiovascular diseases, such as heart disease.

Most results in the paper do not adjust for smoking at all.

Strengths/Limitations

Strengths

This is a large study, making careful use of air pollution measures.  Air pollution modelling appears to be carefully done.  The study adjusts for smoking, social class, region age and sex.  There also some analyses which (partially) adjust for smoking.

Limitations

There is selective reporting in places.  “Largest effect sizes” – this will show exaggerated effects.

Despite the careful analysis, I am still not convinced that the effects are due to air pollution – and are perhaps due to a confounder.

• Some of the confounders are expected to have quite large effects.  The impacts of air pollution are relatively small.  It is thus difficult to be certain of the effect of air pollution.
• The largest associations are consistently with COPD – which is known to be associated with smoking, and also with working in dusty environments (e.g. coal mines).
• The levels of air pollution were much higher in 1971 (Table 1) than 1981 or 1991, but the odds ratios for impact of air pollution were appreciably lower in 1971 compared to 1981 or 1991. This is strange to me, if it is a genuine effect of air pollution. This is a pattern I would expect if air pollution is acting as a surrogate for other variables (e.g. smoking, residual effects of socio-economic status).

Historic levels of air pollution were much higher – meaning that the effects relating to historic air pollution might not be relevant to today’s levels of air pollution.

Additionally, the effects of historic air pollution could predate the specific levels of pollution in 1971, so measuring exposure from that point onwards might not be optimal.  This would include the smogs in the UK of the 1960’s

Some of the air pollution measures are modelled on a 1km square grid.  For factors such as traffic related pollution – there is likely to be considerable variation within the areas.

Glossary

“All cause mortality” in the paper presumably refers to “All causes excluding accidents”

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.