Environment Counts | Changes in methodology doubles WHO estimates of deaths linked to environmental risk factors
Author: Geoff Zeiss – Published At: 2016-03-22 18:13 – (712 Reads)
|
The World Health Organization (WHO) has recently released a report with an infographic that headlines as “fact†that 23% of all global deaths are linked to the environment, approximately 12.6 million deaths a year. Since these estimates are widely used to support policy decisions, this article aims to elucidate some of the key sources of uncertainty in their calculation. In particular these estimates include limited and unreliable statistics on cause of death, a reliance on generalizing from epidemiological studies in the U.S. and Western Europe to the rapidly developing cities of Asia, and the limited data about the health effects of the different chemical composition of air pollution in different localities. Our conclusion is that while the latest WHO report is based on the best evidence available, there are still important gaps and concerns with the data underlying the estimates.Preventing Disease through Healthy Environments Second Edition 2016
|
Introduction
The recent World Health Organization (WHO) report Preventing Disease through Healthy Environments Second Edition 2016 headlines that 12.6 million deaths, or 23% of worldwide deaths, were attributable to environmental risk factors in 2012. The report gives as 95% confidence intervals between 13–34% of all global deaths.
This article is intended as a backdrop for a separate article 7 million people die every year from air pollution – Double previous estimates on this site which presents the latest WHO results for mortality linked to air pollution, both indoor and outdoor.
Since the cause of death reported on death certificates specifies a disease or diseases, not environmental risk factors such as air pollution, statistical and other mathematical techniques are necessary to produce these estimates. In this article we investigate how the WHO estimates are calculated and the sources of uncertainty in these estimates.
Calculating environmental burden of disease
The WHO estimates are calculated using data and methods from the Global Burden of Disease (GBD) framework, a large worldwide initiative which began in 1990. As part of the GBD study mathematical techniques were developed to calculate the proportion of the incidences of diseases attributable to environmental risk factors.
In simplified form the estimates of number of deaths due to a disease such as lung cancer that are attributable to an environmental risk factor such as air pollution is calculated as
- The population at risk from air pollution is the urban or rural population of a country or region and is generally well known.
- The fraction of the population that is expected to contract lung cancer can be estimated using simple accounting from cause of death on death certificates when they are available. But complete death registration covers only 1/3 of the world’s population. Some information on another third is available through death registration systems maintained by some cities and national sample registration systems in India and China. For the remaining third of the world population limited information is available from “verbal autopsies”, for example. Another source of error is that even physician assigned causes of death are not always reliable. A study has found that 1/3 of physician assigned causes of death are incorrect.
- The population attributable fraction (PAF) is the proportional reduction in death or disease that would occur if exposure to a risk is reduced to a “counterfactual exposure distribution”. This is the minimum exposure distribution currently achieved in certain population groups, for example households not using solid fuels, or that which could be achieved by changes in the environment, for example, populations with low atmospheric pollution.
In the best case, the method used for estimating the population attributable fraction due to environmental risk factors is based on rigorous epidemiological studies, for example, cross-sectional or longitudinal studies that compare health outcomes in populations with high atmospheric air pollution with populations that have lower air pollution risk.
Integrated exposure-response function (IER)
When reliable epidemiological studies are available, a function called an integrated exposure-response function (IER) is used. For example, for air pollution, an IER is used to estimate the dependence of disease incidence on the concentration of particulate matter, typically measured as PM2.5 or PM10. For the range of particulate concentrations found in Europe, typically 5-30 micrograms/cubic meter (μg/m3), for example, a study of over 300,000 people found that for every increase of 10 μg/m3 in PM10 the lung cancer rate rose 22%. The study also found a 36% increase in lung cancer per 10 μg/m3 increase in PM2.5.
For the 2010 GBD study, the IER used to estimate the population attributable fraction from exposure to ambient PM2.5 is a non-linear function fitted to the observed relationships from epidemiologic studies of long-term exposure to particulate matter not only from outdoor air pollution (AAP), but also second hand smoke (SHS), direct smoking (ATS), and studies of household air pollution from solid cookfuel (HAP). A source of uncertainty in the IER is extrapolating from lower PM2.5 concentrations typical of North American cities where there are epidemiological studies to the higher PM2.5 concentrations typical of rapidly developing cities in Asia.
These models were used to estimate the percentage of population attributable fraction associated with exposure to ambient PM2.5 using average annual PM2.5 for each of the 187 countries included in the GBD 2010 project.
A source of uncertainty is the use of an IER for the long-term effects of PM10, which are currently mainly available from studies in the United States. The validity of the use of such risk measures in burden of disease studies in other countries is questionable, since air pollution chemical mixtures (for which PM10 is an indicator) and average population susceptibility may vary between countries.
For many environmental risk factors there may be very limited data and a lack of reliable methods for estimating the population attributable fraction for certain diseases. A study of environmental risk factors concluded that only for three environmental risk factors, water sanitation and hygiene, solid fuel use, and outdoor air pollution, were the necessary methodology and enough exposure data available to make sensible global estimates at a country level.
Ambient (outdoor) air pollution
For 2008, the number of premature deaths attributable to urban outdoor air pollution (AAP) was estimated as 1.34 million worldwide. Globally, the most recent study estimated that 3.7 million deaths were attributable to ambient air pollution in 2012.
| Year | Premature deaths attributed to AAP (millions) |
| 2004 | 1.15 |
| 2008 | 1.34 |
| 2010 | 3.1 |
| 2012 | 3.7 |
The WHO cited primarily the following methodological reasons for the doubling of the AAP attributable burden compared with the previous estimate:
- better data on the relationship between exposure and health outcomes and the use of integrated exposure-response functions. The 2010 GBD was the first to use integrated exposure response models for estimating the burden of disease attributable to risk factors. For mortality due to ambient air pollution as measured by ambient PM2.5 concentrations IER functions were developed over the entire global exposure range for particulate pollution for causes of mortality in adults for most important diseases were based on available risk information from studies of ambient air pollution (AAP), second hand tobacco smoke, household solid cooking fuel, and active smoking. The counterfactual concentration was selected to be between 5.8 and 8.8 µg/m3.
- the inclusion of the rural population, whereas the previous estimate only covered the urban population, and
- the use of a lower counterfactual, i.e. the baseline exposure against which the effect of air pollution is measured.
The only non-methodological factor cited as leading to the increased estimate was an increase in non-communicable diseases as a result of urbanization.
Household air pollution
A comparative risk assessment of household air pollution (HAP) done as part of the 2010 GBD study concluded that in 2010 HAP was responsible for 3.9 million premature deaths, nearly double the 2004 estimate of 2 million. More recently, it was estimated that 4.3 million deaths were attributable to household air pollution in 2012.
| Year | Deaths attributed to household air pollution (millions) |
| 2004 | 2 |
| 2010 | 3.9 |
| 2012 | 4.3 |
The same integrated exposure-response functions (IER) developed for AAP were used for most important diseases.
The large increase found in 2010 and 2012 compared with the previous estimate of 2 million deaths from household air pollution in 2004 were ascribed to primarily two methodological factors:
- additional outcomes such as cerebrovascular diseases and ischaemic heart disease were included in the analysis
- new data on the relationship between exposure and health outcomes and the use of integrated exposure response functions
As in the case of AAP, the one non-methodological factor cited for an increase in the most recent estimates was a rise in non-communicable diseases.
A source of uncertainty cited is the integrated exposure-response functions (IER) developed for the GBD 2010 study. These exposure-response relationships derive from modeling across a broad range of mean exposures (5 to over 100 µg/m**3) with different patterns of exposure using epidemiologic evidence from very different populations (ambient air pollution, secondary tobacco smoke, direct tobacco smoking, and household air pollution). The authors qualify their results by saying that although sufficiently compelling to be used for the AAP and HAP studies, they need to be confirmed with more direct evidence in HAP settings for the major disease outcomes.
These AAP and HAP analyses assume stable conditions of exposure. For example, it is assumed that the attributable burden of each disease today is the result of exposures to HAP over long periods in which HAP levels did not change precipitously.
Conclusion
The estimated number of deaths attributed to AAP and HAP in the most recent WHO study is based on a number of important assumptions which contribute to uncertainty in accepting the findings as fact. These assumptions include limited and unreliable causes of death statistics, generalizing from epidemiological studies in the U.S. and Western Europe to the more polluted cities of Asia, and the limited data about the health effects of the different chemical composition of air pollution in different locations around the globe. The WHO report tries to capture this level of uncertainty in the estimated 95% confidence intervals of 13–34% of all global deaths. The study concluded that for three environmental risk factors, water sanitation and hygiene, solid fuel use, and outdoor air pollution, is there the necessary methodology and enough exposure data available to make sensible global estimates at a country level. This then raises questions about the reliability of the estimates for other environmental risk factors (such as second-hand smoke, crowding, ionizing radiation) and therefore the conclusion that 12.6 million deaths are attributable to environmental risk factors. Perhaps,the best way to position the most recent WHO report is that it is based on the best evidence available, but that there are important gaps and concerns with the data underlying the estimates.
Supplementary Material
Supplementary Linking Diseases to Environmental Risk Factors.pdf
Sources
Preventing Disease through Healthy Environments Second Edition 2016
Preventing Disease through Healthy Environments 2006
Lim SS, Vos T, Flaxman AD, Danaei G, Shibuya K, Adair-Rohani H, et al. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010, The Lancet 2012;380(9859):2224-60. doi: 10.1016/S01406736(12)61766-8.
Lancet. 2013 Feb 23;381(9867):628. AlMazroa, Mohammad A (added); Memish, Ziad A (added).
Lancet. 2013 Apr 13;381(9874):1276.
