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

aircraft noise and cardiovascular disease (before the headlines)

A study of residents living close to Heathrow airport, published in the BMJ, found an association between exposure to high levels of aircraft noise and an increased risk of cardiovascular disease. Results were presented at an SMC briefing, with accompanying roundup comments.

 

Title, Date of Publication & Journal

‘Aircraft noise and cardiovascular disease near Heathrow airport in London: small area study’, by Anna Hansell et al, published in the BMJ on Tuesday 8 October 2013.

 

Claim supported by evidence?

  • The evidence in the paper suggests, but does not prove, that there is an association between the highest levels of aircraft noise and increased risk of hospital admissions for stroke, CHD and CVD.
  • There is no evidence for any increased risk in areas with mid-levels of aircraft noise.
  • The evidence for increased numbers of deaths from these causes in relation to aircraft noise levels is weaker than that for increased hospital admissions.

 

Summary

The authors took aircraft noise data from the Civil Aviation Authority records for several areas in London near Heathrow.  They looked at hospital admissions and numbers of deaths from stroke, coronary heart disease and cardiovascular disease (data from Office for National Statistics and the Department of Health) for people living in certain areas (they defined small areas using national census geographic units).  They found that those areas experiencing the very highest levels of aircraft noise (>60 dB) to have greater numbers of hospital admissions for stroke, CHD and CVD.  However, these findings apply to areas rather than individual people so confounding factors (e.g. smoking, ethnicity, depravation) are harder to control for so we can’t be sure from this that the aircraft noise was itself responsible for increased hospital admissions.  The evidence for increased numbers of deaths in these areas is weaker and we cannot be sure any deaths were due to aircraft noise.

 

Study Conclusions

There is suggestion of a link between very high levels of aircraft noise and hospital admissions.

Causality cannot be concluded as this is an observational study.

The relative risks given in the press release only apply to areas with the highest level of aircraft noise compared to areas with the lowest levels.  The highest levels were only regularly experienced by a small proportion of the study population (less than 2% of sample, or about 60,000 people out of ~3,000,000; supplemental table 1a).

Most areas in the study experienced low levels of noise.

There is no evidence of detrimental health effects of low or mid-level noise.

 

Strengths/Limitations

Strengths

The authors have acknowledged the limitations of the data and clearly recognise this in their paper.  These are echoed in the expert quotes in the press release.

This is a well-conducted study and the analysis methodology is appropriate.

In their discussion the authors reference other studies that did not show similar health effects because noise levels were lower – this is supportive of the lack of health effects for those living in areas with low noise levels in this study.

 

Limitations

Lack of individual-level information on confounders e.g. smoking means we cannot be sure aircraft noise has an effect at the level of an individual person.

Page 8 of the paper states that “results at the area-level may not be applicable to individuals (ecological fallacy)”; in other words, not everyone in a designated area will be a smoker of South Asian ethnicity, but the paper cannot allow for that.

The authors wanted to compare data for different factors to see if they were linked, but to do this they had to compare a dataset of one size with a dataset of another, e.g. the area size for ethnicity data (COA) was different from the area size of deaths data (SOA), so analysing them together means the variability within each of the different areas is unknown and we cannot tell how applicable the results are to individuals within the areas.

Some categories are rather broad (e.g. South Asian Ethnicity: <4%, 4%-8%, 8%-50% and >50%) so adjustment for ethnicity may be poor.

The authors have attempted to model data at the ‘area’ level.  Two different ‘areas’ have been used – COA was used for hospital admission data whereas SOA was used for deaths (because there were relatively few deaths so larger area sizes were required for the statistical model used); see supplementary tables 1a and 1b.  The larger you make an area, the more approximate the value used for each confounder for an area becomes (so evidence for deaths is weaker).

Data analysis of day-time noise assumes population stays continuously within the COA/SOA; this might not be the case so we don’t know how relevant the results are for people who move around between areas.

The number of data for those at high aircraft noise volumes is limited (most groups in the study lived in areas with lower noise volumes) so the claimed association between high aircraft noise and cardiovascular health problems may only be applicable to ~6,000 of the ~3,000,000 people in the study population.

Presentation of relative risks and not absolute risks – it is difficult to calculate absolute risks because they have not looked at individual people.

 

Glossary

COA and SOA are standard geographical areas created for census data.

COA = Census Output Area = Average of 397 inhabitants, 0.13 square km;

0.13 square km would be a square 360m by 360m.  Distances within an average COA could be longer if the area was not square / round.

COA used for hospital admission data.

SOA = Super Output Area = 5 COAs = Average of 1,510 inhabitants, 0.65 square km;

0.65 square km would be a square 800m by 800m.

SOA used for deaths data.

Carstairs index = measure of deprivation used in paper.

CHD = coronary heart disease.

CVD = cardiovascular disease.

dB = decibel.

 

‘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