Pollution and pneumonia

Behind the Headlines

Tuesday April 15 2008

“Air pollution ‘kills as many as the smogs of the 1950s’”, reads the headline in the Daily Mail today. It goes on to say that scientists have looked at levels of emissions and causes of death in 352 local authority areas in England. They found that after making adjustments for social factors, deaths from pneumonia were strongly linked to emissions.

The lead researcher, George Knox is reported as saying that many of the deaths from pneumonia were probably caused by “direct chemical injury” and that “Total annual losses as a result of air pollution, through pneumonia, probably approach those of the 1952 London smog, which killed 4,000 people”.

This study looked at associations between emission levels and deaths from various causes across England. It approached this question from a population level, which means that it did not assess the exposure for each individual. Instead, it estimated the emissions for each area and looked for an association with deaths from various causes in the same region.

This relatively complex study does indicate that there is an association between certain emissions and deaths from pneumonia. However, because it does not directly look at an individual’s exposures and outcomes, it cannot by itself prove a link. A larger body of information about this association would need to be considered before drawing this type of conclusion.

Where did the story come from?

Professor George Knox carried out the research. No sources of funding were reported for the study. The study was published in the peer-reviewed: Journal of Epidemiology and Community Health.

What kind of scientific study was this?

In this ecological, cross-sectional study, data on deaths from different causes in areas of England were compared to air pollution levels for those areas.

Using data compiled and published by the Oxford Cancer Intelligence Unit, the researcher obtained standardised mortality ratios (SMRs) for 352 local authorities in England between 1996 and 2004.

Standardised mortality ratios are used to compare the proportions of deaths from different causes between different years, or between different populations, for example, the populations residing in different areas. The expected number of deaths from a specific cause over a certain period in an area are calculated based on data for the entire population, and adjusted (standardised) for any differences in age and gender between the area in question and in the population as a whole. The SMR is the ratio of actual (observed) deaths to the expected deaths.

From this data, the researcher obtained SMRs for 45 specific diseases that had sufficiently complete data and which according to the researcher, were “suitable for analysis”.

Annual estimates of particulate and gaseous emissions for each square kilometre of the UK National Grid were obtained from the National Atmospheric Emissions Inventory (NAEI). This data is grouped by the source of the emissions (e.g. road transport, power production, industrial) and by local authority.

The researcher also used maps displaying accumulated data for major emissions to estimate the amount of material diffusing from more distant major sources. Because each local authority varies in size, density and pattern of inhabitants, a “population centre point” was identified for each. This centre point was defined as the area with the highest carbon dioxide emissions from commercial institutional and residential combustion (that is, space heating of schools, homes, and businesses) within the local authority area.

The areas identified on the National Grid were then linked to data on potential social confounders within these areas such as poverty, poor education, hazardous employments and lifestyles. This was obtained from multiple sources such as the Index of Multiple Deprivation (IMD, 2004) and government data about each local authority based on social surveys.

The researcher used statistical methods to look for links (correlation) between the SMRs and the emission data. These analyses were adjusted for five main social factors: IMD, smoking, binge drinking, and distance east and north of the population centre.

What were the results of the study?

The author measured the variation, which is the degree to which the values found for each item varied across the areas. Between the local authority areas, there was a lot of variation found in emission scores and in social variables (e.g. levels of binge drinking). There was also a lot of variation in the SMRs for some diseases, such as lung cancer, pancreatic cancer, asthma, and chronic obstructive pulmonary disease. There was less variation in the SMRs for other cancers, such as oesophageal, breast, and prostate cancer.

The researcher then looked at how closely these factors were linked to each other. Substantial correlation was found between different emissions and between different social variables, and between these two types of variables. SMRs for some cancers were positively correlated with each other, for example, in areas where the SMR for lung cancer was high the SMR for stomach cancer were also high. In other cases, there was a negative correlation, for example, in areas where SMR for melanoma was high, the SMR for stomach cancer was low, and vice versa.

There were associations between emission levels and standardised mortality ratios (SMRs) for certain diseases, however, once these were adjusted for the five main social variables, most of the associations were no longer significant. The positive associations between emissions and SMRs for cancer of the lung and stomach, rheumatic heart disease, chronic obstructive pulmonary disease, peptic ulcer, and pneumonia remained significant.

The strongest associations seen were for the SMR for pneumonia. The emissions that showed these associations were mostly those that arose from oil combustion and road transport.

What interpretations did the researchers draw from these results?

The researcher concluded that there was “a strong correlation between deaths from pneumonia and engine exhaust emissions, together with other transport-related substance”.

What does the NHS Knowledge Service make of this study?

This was a relatively complex study, which does indicate that there is an association between certain emissions and deaths from pneumonia. There are some points to consider when interpreting this study:

  • This study was carried out at the population level. This means that it did not examine the exposures for individual people who died. Because of this, it cannot confirm that the deaths were a direct result of these exposures. The study cannot in itself be taken as proof that these emissions cause pneumonia. A larger body of information about this association would need to be considered before drawing this type of conclusion.
  • The values for emissions were based on figures from 2004. These figures may not be representative of past exposures of the residents of these regions.
  • As the study reports, there was a strong association between emission level and various social factors. For example, people who live in industrial areas with high emission levels may be more likely to have other social risk factors, such as lower socioeconomic status and more unhealthy lifestyle. Although the results were adjusted for some of these social factors, the effect of these and other factors (such as diet) on mortality may still be present. Therefore, it is not clear now much of this association could be due to these other factors.

Analysis by Bazian

Edited by NHS Choices

Links to the headlines

Pneumonia 'linked' to pollution. BBC News, April 15 2008

Pneumonia from traffic fumes 'kills thousands'. The Daily Telegraph, April 15 2008

Traffic fumes are killing thousands every year. The Scotsman, April 15 2008

Pneumonia deaths' 'pollution link'. Channel 4 News, April 15 2008

Air pollution kills as many as the smogs of the 1950s, say scientists. Daily Mail, April 15 2008

Links to the science

Knox EG. Atmospheric pollutants and mortalities in English local authority areas. J Epidemiol Community Health 2008; 62:442–447


How helpful is this page?

Average rating

Based on 1 ratings

All ratings

1  ratings
0  ratings
0  ratings
0  ratings
0  ratings

Add your rating