Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Potential health benefits of eliminating traffic emissions in urban areas

  • Shahram Heydari ,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Validation, Visualization, Writing – original draft, Writing – review & editing

    s.heydari@soton.ac.uk

    Affiliation Department of Civil, Maritime, and Environmental Engineering, University of Southampton, Southampton, United Kingdom

  • Masoud Asgharian,

    Roles Investigation, Methodology, Writing – review & editing

    Affiliation Department of Mathematics and Statistics, McGill University, Montreal, Canada

  • Frank J. Kelly,

    Roles Investigation, Methodology, Writing – review & editing

    Affiliation Humphrey Battcock Chair of Community Health & Policy, Environmental Research Group, School of Public Health, Faculty of Medicine, Imperial College London, London, United Kingdom

  • Rahul Goel

    Roles Conceptualization, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Transportation Research and Injury Prevention Centre, Indian Institute of Technology Delhi, New Delhi, India

Abstract

Traffic is one of the major contributors to PM2.5 in cities worldwide. Quantifying the role of traffic is an important step towards understanding the impact of transport policies on the possibilities to achieve cleaner air and accompanying health benefits. With the aim of estimating potential health benefits of eliminating traffic emissions, we carried out a meta-analysis using the World Health Organisation (WHO) database of source apportionment studies of PM2.5 concentrations. Specifically, we used a Bayesian meta-regression approach, modelling both overall and traffic-related (tailpipe and non-tailpipe) concentrations simultaneously. We obtained the distributions of expected PM2.5 concentrations (posterior densities) of different types for 117 cities worldwide. Using the non-linear Integrated Exposure Response (IER) function of PM2.5, we estimated percent reduction in different disease endpoints for a scenario with complete removal of traffic emissions. We found that eliminating traffic emissions results in achieving the WHO-recommended concentration of PM2.5 only for a handful of cities that already have low concentrations of pollution. The percentage reduction in premature mortality due to cardiovascular and respiratory diseases increases up to a point (30–40 ug/m3), and above this concentration, it flattens off. For diabetes-related mortality, the percentage reduction in mortality decreases with increasing concentrations—a trend that is opposite to other outcomes. For cities with high concentrations of pollution, the results highlight the need for multi-sectoral strategies to reduce pollution. The IER functions of PM2.5 result in diminishing returns of health benefits at high concentrations, and in case of diabetes, there are even negative returns. The results show the significant effect of the shape of IER functions on health benefits. Overall, despite the diminishing results, a significant burden of deaths can be prevented by policies that aim to reduce traffic emissions even at high concentrations of pollution.

Introduction and background

Motorised traffic is growing rapidly in many low- and middle-income countries (LMICs) resulting from increasing ownership of vehicles and rapid urbanisation. In these settings, emissions are still increasing compared to Europe and the United States where emissions have stabilised or are decreasing [1]. As expected, fine particulate matter (PM2.5) concentrations in East Asia, South Asia, and parts of Sub-Saharan Africa have increased markedly over the past decades and are currently the highest in the world [2, 3]. These concentrations of pollution result in disproportionate share of premature mortality due to cardiovascular and respiratory diseases in LMICs [3]. New evidence on the impact of PM2.5 pollution on infant mortality [4] and diabetes [5], now included in the Global Burden of Disease estimates [6], has added to the previously known health burden of PM2.5. Air pollution is one of the key pathways through which transport impacts public health in the cities [7]. Health impact studies that focus on reducing the use of motorised travel through a shift to active modes of travel have highlighted the health benefits from reduction in traffic emissions [8, 9].

Several previous studies investigated different aspects of traffic-related air pollution and its health impacts [1015]. For example, Pan et al. [10] estimated potential impacts of electric vehicles on air quality and health endpoints in Houston (USA) in 2040. Tong et al. [13] examined health effects of PM2.5 emissions from on-road vehicles during weekdays and weekends in Beijing, China. Teixeira et al. [14] estimated the impact of PM emissions from heavy-duty trucks on human health. A review by Health Effects Institute [16] found suggestive evidence of a causal relationship between exposure to traffic-related air pollution and onset of childhood asthma, non-asthma respiratory symptoms, impaired lung function, total and cardiovascular mortality, and cardiovascular morbidity. These studies have all pointed towards the potential of achieving health benefits from reduction in traffic-related air pollution in urban settings.

The potential of gaining health benefits in a city through reduction in traffic emissions depends on the proportion of PM2.5 concentrations that is contributed by this sector as well as the total PM2.5 burden. Since the dose-response functions are non-linear, with a curve that is steep at low concentrations and flattens towards higher concentrations, there are diminishing returns of reduction in pollution levels at higher concentrations [2]. The two factors (proportion of traffic and overall PM2.5 concentrations) vary greatly across the world [1719]. For example, Heydari et al. [19] showed that traffic contribution estimates as well as uncertainties around these estimates vary largely across various cities and regions worldwide. Many LMIC cities have high concentrations of pollution because of contributions from multiple sectors, of which transport is only one of them [18, 20]. Many of the high-income countries have achieved cleaner air due to the reduction of emissions across multiple sectors. Based on the WHO source apportionment database [21], employing a population-weighted approach, Karagulian et al. [17] conducted a systematic review of local source contributions of PM in cities across the world. Using the same database and based on a Bayesian meta-regression approach, Heydari et al. [19] estimated the expected percentage contribution of traffic to PM2.5 and PM10, and their respective uncertainties, in various cities and regions worldwide.

The primary goal of this research is to estimate the health benefits that can be gained by reduction in traffic emissions. This is achieved by carrying out a rigorous meta-analysis exercise, with the aim of pooling strength over several previous studies on the concentrations of PM2.5 in multiple cities. To this end, in this work we introduce and discuss an analytical framework that can draw valuable inferences regarding the overall (due to all sources) and traffic-related (exhaust and non-exhaust emissions) concentrations of air pollutant concentrations in various locations worldwide from a collection of previous studies. Our specific objectives are summarised as follows:

  1. Develop a meta-regression model that simultaneously analyses overall and traffic-related PM2.5 concentrations in urban areas based on the previous studies collected in the WHO source apportionment database. Doing so, (i) we can explain variability in the reported concentrations by previous studies; (ii) estimate the magnitude of dependence between overall and traffic-related PM2.5 concentrations; and (iii) estimate expected concentrations of PM2.5 of different types (traffic-related, non-traffic-related, and overall due to all sources) with their associated uncertainties in multiple cities worldwide.
  2. Use the above estimates (specifically, the estimates obtained in 1.iii) to investigate the potential of achieving cleaner air and preventing premature mortality from multiple disease outcomes through reductions in traffic-related PM2.5.

Materials and methods

Review framework and study selection

We present a meta-analysis of traffic-related PM2.5 and overall (due to all sources) PM2.5 concentrations reported in the latest available World Health Organization (WHO) database (at the time of writing) on source apportionment studies [21]. The WHO database reports overall PM2.5 concentrations and the contributions of different source categories (i.e., traffic, industry, domestic fuel burning, natural sources, and unspecified sources of human origins) to particulate matter for various locations. Given the aim of our research, we excluded studies that did not report the share of traffic. Also, we only included studies that were reported from urban areas, excluding other site typologies such as industrial, rural, etc., to obtain a homogenous sample, reducing non-comparability between studies. This improves the quality of our meta-analysis.

For each study, we obtained traffic-related PM2.5 concentrations by multiplying overall PM2.5 concentrations by the reported percentage traffic contributions to PM2.5. To carry out our quantitative synthesis of previous research, we considered a series of explanatory variables available in the WHO database. These included publication year, study location (city, country, region, and continent), population, geographic coordinates, and estimation method. Another potentially relevant information was whether a study reported sea salt contribution to PM2.5. To better capture variability in the data, cities were assigned to 12 different regions mostly according to geographic proximity, and classifications reported by [17] and [19]. These are North America, Central Europe, East Asia, East/West Africa, Middle East, North-western Europe, Oceania/Japan, South/Central America, South-eastern Asia, Southern Asia, Southwestern Europe, and Western Europe. List of countries in each region are reported in Table B of the S1 File.

Characteristics of the final data

Our final dataset includes 182 observations of source contributions of PM2.5 concentration in urban areas. These observations are measurements reported by 118 studies (Table A of the S1 File), corresponding to 117 cities worldwide, from 1987 to 2014. A summary of the final sample used in our study is reported in Tables 1 and 2. Around 36% of the reported measurements were estimated based on studies conducted in North America or Oceania. Around 10% were from North Western or Western Europe while 23% of the observations were from studies conducted in the rest of Europe. Two-thirds (74.2%) of the measurements were reported after year 2005, and 50% of the observations in the data reported percentage contribution of sea salt to PM2.5. In our final data, the reported overall PM2.5 concentrations varied largely across cities: from around 12 ug/m3 to 97 ug/m3, with a mean (and standard deviation) of 35.11 ug/m3 (36.97 ug/m3) at a global level. Similarly, traffic-related PM2.5 varied from 1.10 ug/m3 to 64.02 ug/m3, with a mean (and standard deviation) of 9.13 ug/m3 (12.41 ug/m3).

Meta-regression

We adopted a joint meta-regression approach to identify factors that can explain variations in reported overall and traffic-related PM2.5 concentrations in the WHO database. This allowed us to model both outcomes simultaneously through a system-equation approach rather than modelling each outcome separately, improving the reliability of our statistical inferences. We assumed the log-transformed concentrations follow a multivariate normal density. Let yki denote the vector of log-transformed concentrations of k different types reported by previous studies i (i = 1, 2, …, N). Here k = 2; therefore, y = (y1i, y2i), where y1i, y2i denote PM2.5 due to traffic and overall PM2.5, respectively. Let Xk = (Xk1, Xk2, …, Xkm) be the vector of m explanatory variables (e.g., population) associated with the outcomes of interest (y1i, y2i) with their respective regression coefficients γ = (γk1, γk2,…, γkm). Let η = (η1, η2) denote the vector of intercepts corresponding to y1i and y2i, respectively. Let R and K denote the scale matrix and the degrees of freedom, respectively, in a Wishart distribution. We can then write (1)

As it can be seen in the above model, the dependency across outcomes is captured through the covariance matrix Σ. This specification allowed us to investigate the magnitude of correlation between overall PM2.5 and traffic-related PM2.5 concentrations.

Prior specification and model computation

Normally distributed non-informative priors, normal (0,100), were used for the regression coefficients. In the joint modelling of correlated outcomes, it is a common practice to specify a Wishart distribution for the inverse of covariance matrix Σ-1 [22], with K = 2 (for two correlated outcomes) and a 2x2 scale matrix R (R[1,1] = R[2,2] = 0.01 and R[1,2] = R[2,1] = 0), which leads to a non-informative prior specification. For model computation, we employed WinBUGS [23] to draw posterior densities for our Markov chain Monte Carlo simulations running two chains each containing 15,000 iterations. The posterior densities are based on 20,000 samples as the first 5,000 iterations were discarded for convergence requirements. Based on the Gelman-Rubin statistic [24], history plots, and Monte Carlo errors, this number of iterations was sufficient.

Computing probabilities of exceeding the WHO-recommended concentration of PM2.5

We estimated the potential of a city to attain the WHO-recommended concentration of PM2.5 concentrations (5 ug/m3) [18] if all traffic emissions were removed. For this, we estimated the posterior densities of non-traffic-related PM2.5 concentrations for each city based on our meta-regression approach. We then estimated the probability that the latter posterior mean exceeds 5 ug/m3. The higher this probability, the lower the ability of a city to reach the WHO-recommended concentration of PM2.5 even from the complete removal of traffic emissions. To compute these probabilities, we created an Mx1 matrix of indicator variables [Ic] for each city c, where M is the total number of cities under investigation. At each iteration of our MCMC simulations and for each city, we compared the expected non-traffic-related PM2.5 concentration with the WHO-recommended value of 5 ug/m3 as shown in (2). Finally, we averaged the indicator variable value over all iterations to obtain the probabilities of exceeding for each city.

(2)

Estimating health benefits of reducing traffic emissions

We estimated the percentage reduction in health burden resulting from a complete removal of traffic emissions. To estimate changes in health burden we used Integrated Exposure Response (IER) functions for six disease endpoints that were used in Global Burden of Disease (GBD) 2017 [6]. These are ischemic heart disease (IHD), stroke, chronic obstructive pulmonary disease (COPD), lung cancer, lower respiratory infections (LRI), and type-II diabetes (diabetes). For IHD and Stroke, IER is age-specific and, for these, we present calculations for 55–60 years age group for illustration of the method. The relative risk using IER function is calculated as: (3) where ϑ, ω, δ are parameters specific to each disease end-point and x is a counterfactual value below which the assumption is that there are no increased mortality. The value of x is obtained from a uniform distribution, representing its uncertainty, with lower and upper bounds of 2.4 ug/m3 and 5.9 ug/m3, respectively [25]. The dose-response function for the six disease end-points are presented in Fig 1. This graph presents the average value of relative risks at each concentration value, calculated for 1000 iterations of the four IER parameters ϑ, ω, δ, x reported by Burnett [26].

thumbnail
Fig 1. Integrated exposure response function (age group 55–60 years for IHD and stroke).

https://doi.org/10.1371/journal.pone.0264803.g001

We used the comparative risk assessment approach to estimate the population attributable fraction (PAF) for each city and expressed it as percentage (see Eq 4). We defined counterfactual PM2.5cf as the concentrations achieved after the complete removal of traffic emissions. (4) where RRcf and RRb are, respectively, the relative risk values (see Eq 3) for the concentrations of PM2.5cf and PM2.5b; the latter stands for overall PM2.5 concentrations at the baseline. For each city, we calculated PM2.5b as the mean value of the posterior densities of overall PM2.5 concentrations, obtained from our meta-regression analysis. We calculated PM2.5cf as the difference between the mean values of the posterior densities of overall and traffic-related PM2.5. Lastly, using the pair of these concentrations, for each city, we calculated PAF for each of the 1000 iterations of the parameters in the IER functions, and we present the mean PAF of those iterations.

Results

The results relating to the estimated posterior densities of PM2.5 for the cities included in our study are reported in the S1 File (Tables C-E in S1 File). Note that to investigate robustness of the multivariate normal density assumption in (1), we used a scale mixing approach that can address skewness in the data; this confirmed the suitability of our assumption.

Posterior estimates of the meta-regression coefficients

As mentioned previously, we considered variables available in the WHO database to develop our models. Through an exploratory data analysis, we noticed a relatively significant difference between measurements reported by studies published before 2005 and those published after 2005 with respect to traffic-related PM2.5 concentrations. This is in accordance with [19]. We therefore created a categorical variable for publication year with the aim of capturing the above difference in the model. Also, we created several categorical variables to include the location of studies in the models, considering different possibilities and combinations of regions according to geographic proximity, income, and classifications reported by [17] and [19]. The final location variable had four categories: (1) North America, Oceania or Japan; (2) North Western or Western Europe; (3) the rest of Europe; and (4) the rest of the world. Japan being a high-income country and having PM2.5 concentrations similar to North America and Oceania was included in the first category mentioned above. Table 3 reports the posterior summary of the estimated regression coefficients.

Correlation between overall PM2.5 and traffic-related PM2.5

Our modelling approach allowed us to estimate the magnitude of the correlation between overall PM2.5 and traffic-related PM2.5 across a sample of 117 cities worldwide. The mean (standard deviation) of the correlation is 0.63 (0.05), with a 95% credible interval varying from 0.54 to 0.72. This indicates that the correlation between overall and traffic-related PM2.5 is statistically important and that traffic-related PM2.5 and overall PM2.5 are highly correlated.

Probability of non-traffic emissions exceeding the WHO threshold

Fig 2 presents the probabilities of PM2.5 concentrations exceeding the WHO-recommended concentration of PM2.5 (i.e., 5 ug/m3) for the cities under investigation if all traffic emissions are removed. This probability accounts for PM2.5 concentrations as well as their associated uncertainties, both of which are represented by posterior distribution of concentrations, estimated based on our analysis. Fig 2 shows that, when the overall concentration is beyond 15 ug/m3, the probability of exceeding the WHO-recommended concentration of PM2.5 remains greater than 80 percent.

thumbnail
Fig 2. Probability of exceeding WHO-recommended PM2.5 if traffic emissions are removed.

https://doi.org/10.1371/journal.pone.0264803.g002

Health benefit potential of reducing traffic emissions

Fig 3 presents percent reduction in the premature mortality due to four disease end-points (COPD, IHD, lung cancer and diabetes) for a counterfactual scenario of complete removal of traffic-related PM2.5 concentrations. Note that stroke and LRI have similar shaped curves as IHD and Lung Cancer, respectively, and their results are not shown here. According to these graphs, the benefits of prevented mortality increase up to a point (30–40 ug/m3) for COPD, IHD and lung cancer, at which point there are large variations, and then flatten off. The flattening is most prominent for IHD and less so for COPD and lung cancer. The diabetes-related mortality reduction show that the largest benefits are limited to concentrations below 25 ug/m3, at which point there is large variation across the cities. Above this concentration, the benefits in terms of reduced mortality become progressively smaller.

thumbnail
Fig 3. Percentage reduction in mortality resulting from total removal of traffic emissions (for COPD and IHD, age group is 50–55 years).

https://doi.org/10.1371/journal.pone.0264803.g003

Discussion

Statement of principal findings

We used the WHO database of source apportionment studies to estimate overall, traffic-related, and non-traffic-related PM2.5 concentrations using a Bayesian meta-regression approach. For the posterior distributions of the expected concentration of non-traffic-related PM2.5, we estimated the probability that concentrations remained higher than the WHO-recommended annual concentration of PM2.5 if traffic-related emissions were removed completely. We found that this probability rises steeply up to 20 ug/m3, and then remains greater than 90 percent. In other words, for the cities with overall concentrations greater than 20 ug/m3, the removal of traffic-related emissions are highly unlikely to reduce concentration levels down to the WHO-recommended PM2.5 annual guideline.

We estimated the percentage reduction in premature mortality due to four disease end-points (COPD, IHD, lung cancer and diabetes) if all traffic-related emissions are removed. We used non-linear IER functions along with the comparative risk assessment approach to estimate population attributable fraction corresponding to this reduction in concentrations. We found that for COPD, IHD and lung cancer, the percent reduction in mortality increases up to 30–40 ug/m3, and at higher concentrations, it flattens off, showing diminishing returns. The flattening is far more prominent in IHD than in COPD and lung cancer. This is expected from the shape of their IER functions (Fig 1). IER for COPD and lung cancer have steeper functions of relative risk than IHD. In case of Diabetes, the pattern is opposite to that of the other three outcomes. With increasing concentrations, there are negative returns in the reduction of premature mortality. This is expected given that diabetes is the only disease outcome that has a prominently flat IER function after a steep jump up to 20 ug/m3. Mathematically, a flat function implies that the ratio of relative risks in Eq 3 approaches unity at higher concentrations.

Strengths and weaknesses of the study

We employed a multivariate meta-regression approach to estimate the expected concentrations of PM2.5 of different types (in terms of source) for 117 cities in various regions worldwide. Note that compared to relaying on information from one study only, conclusions from a meta-analytic approach are more reliable for evidence-based policy making. Further, we modelled traffic-related and overall PM2.5 concentrations jointly; therefore, our estimates have superior statistical properties. This is due to the fact that data points borrow strength from other related data points [27]. Also, a joint analysis allowed us to investigate the magnitude of correlation between traffic related PM2.5 and overall PM2.5 based on a systematic approach, providing further insight into the relationship between the two types of concentrations. Understanding this correlation is quite interesting as it allows estimating the range of one type of concentration from the other one when both pieces of information are not available for a given location. Note that our meta-regression, being developed under the Bayesian paradigm, accounted for uncertainties in both regression parameter estimates and predicted values fully.

One of the limitations of our study is that we have only estimated reduction in mortality due to different diseases as percentage of baseline health burden. Depending on the magnitude of baseline health burden, a given percentage reduction will translate to highly divergent values across the cities. This is because, in certain countries that have much greater proportion of older adults, baseline incidence rates of mortality from the cardiovascular, respiratory, and metabolic diseases are much greater than countries with much younger population. Also, LMICs have much greater incidence rates of mortality than HICs. Secondly, many city estimates used in our analysis are more than a decade old or even older. As a result, our study does not represent the latest situation of most of the cities. However, our study provides an analytical approach that can be readily updated as new data (other studies) become available.

With respect to potential biases in carrying out a meta-analysis [28], firstly, we ensured comparability of the studies considered in our research by focusing only on those conducted in urban settings, removing other site typologies (e.g., industrial sites) from the data. Secondly, we avoided strict inclusion/exclusion criteria in our study selection to prevent limiting generalisability. For example, we did not select urban areas of a specific type or from a specific location. Thirdly, our large sample size helps reduce bias in our estimates. Since we relied on a valid collection of previous studies made available by WHO [21], publication bias should be negligible in our meta-analysis for the period of study as this bias mostly arises when several important studies are not included in a meta-analysis. Overall, we believe such biases would not result in any important issue in our inferences. Lastly, as we adopted a Bayesian meta-regression approach, one source of bias could relate to the choice of prior distributions for the model parameters. Since we used non-informative priors (see the section of prior specification and model computation), we are confident about the reliability of our estimates.

Meaning of the study: Possible mechanisms and implications for policymakers

We found that only for a handful of cities could complete reduction in traffic emissions result in achieving the WHO-recommended guideline for annual PM2.5 concentrations. We used the WHO guideline for the purpose of illustration which, for many settings across the world, is a highly ambitious scenario. The method presented here can be applied for more realistic targets of pollution concentrations and could include reductions across multiple sectors. Using a stochastic approach, we can make a probabilistic judgement of the impact that policies will have. Our approach, being developed under the Bayesian framework, can be used to estimate updated probabilities as more information is added or updated information is added for the same cities. The health benefits that we presented are also for a highly ambitious scenario, in which all of traffic emissions are removed. However, due to non-linearity in IER functions, we found that health benefits are not proportionally as large. While traffic as a polluting sector gains a lot of attention due to visibility of its sources (i.e., vehicles), our results imply that cities with high concentrations of pollution need a multi-sectoral framework to reduce anthropogenic emissions. This will not only help clean air much faster, it will also make investments more cost-effective as concentrations near the steeper part of the curve.

PM2.5 emissions from vehicles are largely proportional to sulphur content in the fuel, and cleaner fuel with lower sulphur content can significantly reduce traffic emissions [29, 30]. While Europe and North America implemented use of low- and ultra-low-sulphur diesel (less than 50 ppm and less than 15 ppm sulphur content, respectively) in late 2000s [31], the progress in many of the low- and middle-income countries has been much slower [30]. A timeline of sulphur concentrations in diesel for 2013–2020 period across the world [32] shows that many countries are gradually progressing towards the use of low-sulphur fuel. However, more than half of the world’s countries are still using high-sulphur fuels. These are mainly low-and middle-income countries spread across Latin America, the Caribbean, Africa, the Middle East, and Asia-Pacific [30]. It is only with low-sulphur fuels that vehicles with stricter emission standards can be effective. Thus, accelerating the desulphurisation of fuel and adoption of cleaner vehicle standards in large parts across the world has large potential to prevent health burden attributed to traffic.

Rapid adoption of electric passenger cars will add to these efforts to reduce on-road emissions. However, electric vehicles, being heavier than internal combustion engine vehicles, have greater contribution to PM2.5 due to non-exhaust emissions such as tyre wear, break wear, road surface wear, and resuspension of road dust. As a result, they have only slightly lower PM emissions than an internal combustion engine vehicle [33]. Therefore, electric vehicles may have the greatest benefits in settings where exhaust emissions continue to be high due to lagging emission standards.

The policies to reduce transport emissions should not be restricted to cleaner fuels or vehicle technology. Instead, they should be developed within a broader framework of transport and its impact on health. Use of an electric car may result in lower emissions than conventional vehicles but does not contribute to physical activity or reduce danger on the road. Alternatively, a mode shift from passenger cars to active travel or public transport will not only reduce emissions, but also improve population health through the pathways of physical activity and reduced road injuries. This mode shift can be achieved through a transport system that prioritises widespread and safe walking, cycling, and public transport infrastructure [34].

Unanswered questions and future research

As highlighted in the limitations, the underlying dataset of cities [21] reporting source apportionment has studies that were done a decade or longer ago. Therefore, an updated review of source-apportionment studies could greatly improve our understanding of different sources of pollution in cities. The results on health impacts presented here are based on IER functions that were used in GBD 2017. The different types of risk functions for PM2.5 have been shown to have significant impact on their respective estimates of disease burden [35]. As evidence improves on different health outcomes as well as from settings with high concentrations of pollution, we could expect changes in these functions and their respective health impact estimates.

References

  1. 1. Crippa M, Guizzardi D, Muntean M, et al. Gridded emissions of air pollutants for the period 1970–2012 within EDGAR v4.3.2. Earth Syst Sci Data. 2018;10(4):1987–2013.
  2. 2. Apte JS, Brauer M, Cohen AJ, et al. Ambient PM2.5 Reduces Global and Regional Life Expectancy. Environ Sci Technol Lett. 2018;5(9):546–51.
  3. 3. Burnett R, Cohen A. Relative risk functions for estimating excess mortality attributable to outdoor PM2.5 air pollution: Evolution and state-of-the-art. Atmosphere (Basel). 2020;11(6):1–13.
  4. 4. Heft-Neal S, Burney J, Bendavid E, et al. Robust relationship between air quality and infant mortality in Africa. Nature [Internet]. 2018;559(7713):254–8. Available from: http://dx.doi.org/10.1038/s41586-018-0263-3 pmid:29950722
  5. 5. Bowe B, Xie Y, Li T, et al. The 2016 global and national burden of diabetes mellitus attributable to PM 2·5 air pollution. Lancet Planet Heal [Internet]. 2018;2(7):e301–12. Available from: http://dx.doi.org/10.1016/S2542-5196(18)30140-2
  6. 6. Stanaway JD, Afshin A, Gakidou E, et al. Global, regional, and national comparative risk assessment of 84 behavioural, environmental and occupational, and metabolic risks or clusters of risks for 195 countries and territories, 1990–2017: A systematic analysis for the Global Burden of Disease Stu. Lancet. 2018;392(10159):1923–94. pmid:30496105
  7. 7. Khreis H, May AD, Nieuwenhuijsen MJ. Health impacts of urban transport policy measures: A guidance note for practice. J Transp Heal [Internet]. 2017;6:209–27. Available from: https://www.sciencedirect.com/science/article/pii/S2214140516304145
  8. 8. Mueller N, Rojas-Rueda D, Salmon M, et al. Health impact assessment of cycling network expansions in European cities. Prev Med (Baltim). 2018;109:62–70. pmid:29330030
  9. 9. Tainio M. Burden of disease caused by local transport in Warsaw, Poland. J Transp Heal [Internet]. 2015;2(3):423–33. Available from: http://dx.doi.org/10.1016/j.jth.2015.06.005 pmid:26516622
  10. 10. Pan S, Roy A, Choi Y, et al. Potential impacts of electric vehicles on air quality and health endpoints in the Greater Houston Area in 2040. Atmos Environ [Internet]. 2019;207:38–51. Available from: https://www.sciencedirect.com/science/article/pii/S1352231019301840
  11. 11. Liang X, Zhang S, Wu X, et al. Air quality and health impacts from using ethanol blended gasoline fuels in China. Atmos Environ [Internet]. 2020;228:117396. Available from: https://www.sciencedirect.com/science/article/pii/S1352231020301357
  12. 12. Ramacher MOP, Matthias V, Aulinger A, et al. Contributions of traffic and shipping emissions to city-scale NOx and PM2.5 exposure in Hamburg. Atmos Environ [Internet]. 2020;237:117674. Available from: https://www.sciencedirect.com/science/article/pii/S1352231020304064
  13. 13. Tong R, Liu J, Wang W, et al. Health effects of PM2.5 emissions from on-road vehicles during weekdays and weekends in Beijing, China. Atmos Environ [Internet]. 2020;223:117258. Available from: https://www.sciencedirect.com/science/article/pii/S1352231019308969
  14. 14. Rodrigues Teixeira AC, Borges RR, Machado PG, et al. PM emissions from heavy-duty trucks and their impacts on human health. Atmos Environ [Internet]. 2020;241:117814. Available from: https://www.sciencedirect.com/science/article/pii/S1352231020305483
  15. 15. Chambliss SE, Silva R, West JJ, et al. Estimating source-attributable health impacts of ambient fine particulate matter exposure: Global premature mortality from surface transportation emissions in 2005. Environ Res Lett. 2014;9(10).
  16. 16. Helth Effects Institute. Traffic-Related Air Pollution: A Critical Review of the Literature on Emissions, Exposure, and Health Effects. HEI Special Report 17. Health effects Institute. Boston, MA; 2010.
  17. 17. Karagulian F, Belis CA, Dora CFC, et al. Contributions to cities’ ambient particulate matter (PM): A systematic review of local source contributions at global level. Atmos Environ [Internet]. 2015;120:475–83. Available from: http://dx.doi.org/10.1016/j.atmosenv.2015.08.087
  18. 18. World Health Organization. Ambient (outdoor) air pollution [Internet]. 2018. https://www.who.int/news-room/fact-sheets/detail/ambient-(outdoor)-air-quality-and-health#:~:text=%22WHO%20air%20quality%20guidelines%22%20estimate,related%20deaths%20by%20around%2015%25
  19. 19. Heydari S, Tainio M, Woodcock J, et al. Estimating traffic contribution to particulate matter concentration in urban areas using a multilevel Bayesian meta-regression approach. Environ Int [Internet]. 2020;141(May):105800. Available from: pmid:32474298
  20. 20. Guttikunda SK, Nishadh KA, Jawahar P. Air pollution knowledge assessments (APnA) for 20 Indian cities. Urban Clim [Internet]. 2019;27:124–41. Available from: http://www.sciencedirect.com/science/article/pii/S2212095518302402
  21. 21. World Health Organization. WHO’s Source Apportionment Database for PM10 and PM2.5 Updated to August 2014 [Internet]. 2015. http://www.who.int/quantifying_ehimpacts/global/source_apport/
  22. 22. Tunaru R. Hierarchical Bayesian models for multiple count data. Austrian J Stat. 2002;31(2&3):221–9.
  23. 23. Lunn DJ, Thomas A, Best N, et al. WinBUGS-a Bayesian modelling framework: concepts, structure, and extensibility. Stat Comput. 2000;10(4):325–37.
  24. 24. Gelman A, Rubin DB. Inference from iterative simulation using multiple sequences. Stat Sci. 1992;7(4):457–72.
  25. 25. Burnett RT, Arden Pope C, Ezzati M, et al. An integrated risk function for estimating the global burden of disease attributable to ambient fine particulate matter exposure. Environ Health Perspect. 2014;122(4):397–403. pmid:24518036
  26. 26. Burnett R. Integrated exposure response functions. Pers Commun. 2021.
  27. 27. Jackson D, Riley R, White IR. Multivariate meta‐analysis: potential and promise. Stat Med. 2011;30(20):2481–98. pmid:21268052
  28. 28. Joseph, L. A brief introduction to meta-analysis. Division of Clinical Epidemiology, Department of Medicine, Montreal General Hospital, Montreal, Quebec, Canada. 2000. http://www.medicine.mcgill.ca/epidemiology/Joseph/courses/EPIB-669/joseph2000.pdf.
  29. 29. Goel R, Guttikunda SK. Evolution of on-road vehicle exhaust emissions in Delhi. Atmos Environ. 2015;105:78–90. Available from: http://dx.doi.org/10.1016/j.atmosenv.2015.01.045
  30. 30. Climate and Clean Air Coalition. Cleaning Up the Global On-Road Diesel Fleet: A Global Strategy to Introduce Low-sulfur Fuels and Cleaner Diesel Vehicles [Internet]. 2016. https://www.unep.org/resources/report/cleaning-global-road-diesel-fleet-global-strategy-introduce-low-sulfur-fuels-and
  31. 31. EEA. Explaining road transport emissions: A non-technical guide. Copenhagen; 2016.
  32. 32. UNEP/CCAC. Global Low Sulfur Fuels, Cleaner Vehicles Progress Tracker [Internet]. United Nations Environment Programme (UNEP), Climate and Clean Air Coalition (CCAC). 2018. http://www.airqualityandmobility.org/gfeitoolkit/sulphur.html
  33. 33. Timmers VRJH, Achten PAJ. Non-exhaust PM emissions from electric vehicles. Atmos Environ. 2016 Jun;134:10–7.
  34. 34. Salvo D, Garcia L, Reis RS, et al. Physical Activity Promotion and the United Nations Sustainable Development Goals: Building Synergies to Maximize Impact. 2021;1–18.
  35. 35. Evangelopoulos D, Perez-Velasco R, Walton H, et al. The role of burden of disease assessment in tracking progress towards achieving WHO global air quality guidelines. Int J Public Health. 2020;65(8):1455–65. pmid:33057794