Can air pollution negate the health benefits of cycling and walking?

1Several health impact modelling (HIM) studies have estimated the health benefits and risks of active travel (cycling, walking) in different geographical areas. In most of these studies, the health benefits due to physical activity (PA) from increased active travel are significantly larger than the health risks caused by increases in exposure to air pollution.

2Most of the existing active travel HIM studies have been carried out in cities in high income countries with relatively low air pollution levels. This raises the question on the risk–benefit balance in highly polluted environments. Health risks of air pollution are usually thought to increase linearly with increased exposure for low to moderate levels of air pollution, whereas the benefits of PA increase curvy-linearly with increasing dose. Thus, at a certain level of background air pollution and of active travel, risks could outweigh benefits, which would directly imply that, from a public health perspective, active travel could not be always recommended.

3In this study we compare the health risks of air pollution with the PA-related health benefits from active travel across a wide range of possible air pollution concentrations and active travel levels. We use two thresholds to compare PA benefits and air pollution risks (Fig. 1): At the “tipping point” an incremental increase in active travel will no longer lead to an increase in health benefits (i.e. max. benefits have been reached). Increasing active travel even more could lead to the “break-even point”, where risk from air pollution starts outweighing the benefits of PA (i.e. there are no longer net benefits, compared to not engaging in active travel).

4Our approach followed a general active travel HIM method. Air pollution exposures due to active travel were quantified by estimating the differences in the inhaled dose of fine particulate matter (PM2.5) air pollution. We selected PM2.5 because it is a commonly used indicator of air pollution in active travel HIM studies, and because of the large health burden caused by PM2.5. For both air pollution and PA we used all-cause mortality as the health outcome because there is strong evidence for its association with both long-term exposure to PM2.5 and long-term PA behaviour.

5The reduction in all-cause mortality from active travel was estimated by converting the time spent cycling or walking to metabolically equivalent of task (MET) and calculating the risk reduction using dose–response functions (DRFs) adapted from Kelly et al.'s3 meta-analysis. From the different DRFs reported in Kelly et al. we chose the one with the “0.50 power transformation” as a compromise between linear and extremely non-linear DRFs. Non-linearity in a DRF means that the health benefits of increased active travel would level out sooner and a tipping point would be reached earlier than with more linear DRFs. See supplementary material for the sensitivity analysis with different DRFs. To convert cycling and walking time to PA we used the values of 4.0 METs for walking and 6.8 METs for cycling, based on the Compendium of Physical Activities. The walking and cycling levels used in this study are assumed to reflect long-term average behaviour.

6The health risks of PM2.5 were estimated by converting background PM2.5 concentrations to travel mode specific exposure concentrations, and by taking into account ventilation rate whilst being active. For background PM2.5 we used values between 5 and 200 μg/m3 with 5 μg/m3 intervals. We also estimated tipping points and break-even points for the average and most polluted cities in each region included in the World Health Organization (WHO) Ambient Air Pollution Database, which contains measured and estimated background PM2.5 concentrations for 1622 cities around the world.

7The mode specific exposure concentrations were estimated by multiplying background PM2.5 concentration by 2.0 for cycling or 1.1 for walking, based on a review of studies. The counterfactual scenario for the time spent cycling or walking was assumed to be staying at home (i.e. in background concentration of PM2.5). See supplementary file for the sensitivity analysis with counterfactual scenarios where cycling time would replace motorised transport time. The ventilation rates differences whilst at sleep, rest, cycling and walking were taken into account when converting exposure to inhaled dose. For sleep, rest, walking and cycling we used ventilation rates of 0.27, 0.61, 1.37 and 2.55, respectively. The sleep time was assumed to be 8 h in all scenarios and the resting time was 16 h minus the time for active travel.

8For the PM2.5 DRF we used a relative risk (RR) value of 1.07 per 10 μg/m3 change in exposure. We assumed that DRF is linear from zero to maximum inhaled dose. As a sensitivity analysis we used non-linear integrated risk function from Burnett et al. (see supplementary material for details).

9The model used for all calculations is provided in Lumina Decision Systems Analytica format in supplementary file 2 (readable with Analytica Free 101, http://www.lumina.com/products/free101/), and a simplified model containing the main results is provided in Microsoft Excel format in supplementary file 3.

10The tipping point and break-even point for different average cycling times and background PM2.5 concentrations are shown in Fig. 2. For half an hour of cycling every day, the background PM2.5 concentration would need to be 95 μg/m3 to reach the tipping point. In the WHO Ambient Air Pollution Database less than 1% of cities have PM2.5 annual concentrations above that level. The break-even point for half an hour of cycling every day was at 160 μg/m3 (Fig. 2). For half an hour of walking the tipping point and break-even point appear at a background concentration level above 200 μg/m3 (Fig. S3, supplementary file). For the average urban background PM2.5 concentration (22 μg/m3) in the WHO database, the tipping point would only be reached after 7 h of cycling and 16 h of walking per day.

11Tables S2 and S3 (supplementary file) show the tipping point for cycling and walking, respectively, in different regions of the world. In the most polluted city in the database (Delhi, India, background concentration of 153 μg/m3), the tipping and break-even points were 30 and 45 min of cycling per day, respectively (Table S2, supplementary file). In most global regions the tipping points for the most polluted cities (44 μg/m3 to 153 μg/m3) varied between 30 and 120 min per day for cycling, and 90 min to 6 h 15 min per day for walking (Table S3, supplementary material).

12When we assumed that time spend cycling would replace time driving a car, benefits always exceeded the risks in the background air pollution concentrations below 80 µg/m3, a concentration exceeded in only 2% of cities (World Health Organization (WHO), 2014). Other sensitivity analyses showed that the results are sensitive to the shape of the DRF functions. With the linear DRF for active travel the break-even point would be reached with background PM2.5 concentrations of 170 μg/m3 regardless of the active travel time (Fig. S4, supplementary material); a level not currently found in any of the cities in the WHO air pollution database. With the most curved DRF (0.25 power) the PM2.5 concentration where harms exceed benefits for 1 h of cycling per day would drop from 150 μg/m3 to 130 μg/m3 (Fig. S4, supplementary material), a level currently found only in 9 cities. With a non-linear DRF for PM2.5 the break-even point was not reached in any background PM2.5 concentration when using “power 0.50” DRF for cycling and walking. Other input value modifications had small or insignificant impact to the results.

13This study indicates that, practically, air pollution risks will not negate the health benefits of active travel in urban areas in the vast majority of settings worldwide. Even in areas with high background PM2.5 concentrations, such as 100 μg/m3, up to 1 h 15 min of cycling and 10 h 30 min of walking per day will lead to net reduction in all-cause mortality (Fig. S5, supplementary material). This result is supported by epidemiological studies that have found the statistically significant protective effects of PA even in high air pollution environments. However, a small minority engaging in unusually high levels of active travel (i.e. bike messengers) in extremely polluted environments may be exposed to air pollution such that it negates the benefits of PA.

14Some considerations of the limitations and the strengths of our study need to be applied when generalising these findings.

15In this analysis we took into account only the long-term health consequences of regular PA and chronic exposure to PM2.5. Impacts of short-term air pollution episodes, where concentrations significantly exceed the average air pollution levels for a few days, may induce additional short term health effects. We have also only worked with all-cause mortality and have, thus, not taken into account the morbidity impact.

16For the health risks of air pollution we only estimated the increased risk during cycling and walking, not the overall health risk from everyday air pollution. Air pollution causes a large burden of diseases all over the world and reducing air pollution levels would provide additional health benefits. Since transport is an important source of air pollution in urban areas, mode shifts from motorised transport to active travel would not only improve health in active travellers, but also help to reduce air pollution exposures for the whole population.

17The results are sensitive to assumptions of the linearity of dose–response relationships between active travel-related PA and health benefits, and between PM2.5 and adverse health effects. With linear DRFs for PA the benefits always exceeded the risks at all levels of PM2.5 concentrations. Evidence for a linear DRF for high PM2.5 concentrations is small and, for example, the Global Burden of Disease study applied non-linear, disease specific DRFs for PM2.5. If the risks of PM2.5 level out after PM2.5 concentrations over 100 μg/m3, the health benefits of PA would always exceed the risks of PM2.5.

18It should also be taken into account that the results are based on generally representative values without detailed information on local conditions, or from the background PA and disease history of individuals. For individuals highly active in non-transport domains the benefits from active travel will be smaller, and vice versa.

19The benefits from active travel generally outweigh health risks from air pollution and therefore should be further encouraged. When weighing long-term health benefits from PA against possible risks from increased exposure to air pollution, our calculations show that promoting cycling and walking is justified in the vast majority of settings, and only in a small number of cities with the highest PM2.5 concentration in the world cycling could lead to increase in risk.

20MT made the calculations and drafted the first version of the manuscript. AJN, TG, MJN, SK, THS, DRR, PK and JW participated in designing the scope of the study. AJN and TG helped to clarify the message of the study. All authors contributed to the writing of this paper. All authors approved the final version to be submitted for consideration of publication.

21The authors declare that there are no conflicts of interests.