Strengths and weaknesses
The strength of our assessment is especially the quantitative comparison of benefits and risks, in a common scenario for the three stressors evaluated. It could be argued that the Copenhagen (Andersen et al. 2000
) and Chinese studies (Matthews et al. 2007
) of the effects of bicycling on mortality have already demonstrated the net effect of physical activity on all-cause mortality, including the negative effects of fatal traffic accidents and air pollution. However, the size of the potential negative health effects was not quantified separately in those studies. Therefore it is difficult to transfer the net effect of these studies to other locations, where traffic accident rates and air pollution may be different. Because in our assessment the separate risks have been disentangled, it is possible to make assessments for different settings, by using other input data (e.g., traffic mortality rates).
We performed our calculations for the Netherlands, where an extensive cycling infrastructure exists and priority is given to cyclists over other traffic—factors that contribute to regular cycling. Restrictions to car use through traffic calming in residential areas and car-free zones influence cycling behavior as well (Pucher and Dijkstra 2003
). Apart from the highest average distance cycled per person, the Netherlands is also one of the safest countries in terms of fatal traffic accidents. In such countries as the United Kingdom, Spain, and France, the risk of a fatal traffic accident for cyclists is substantially higher, probably also relative to car driving [Supplemental Material, Figure 2 (doi:10.1289/ehp.0901747)]. When we repeated the traffic accident calculations for the United Kingdom, where the risk of dying per 100 million km for a cyclist is about 2.5 times higher [Supplemental Material, Figure 2 (doi:10.1289/ehp.0901747)] and assuming the same fatality risk for car drivers as in the Netherlands, resulting life expectancy losses were approximately 14 days/person, based on 2005 population data from the United Kingdom and Wales. Overall, benefits of cycling are still 7 times larger than the risks.
Calculations on mortality impacts were performed for people 18–64 years of age, because people in that age range are more likely to make the modal shift. Age-specific analysis showed that the relative benefits of cycling are highest in the older age categories. This may have been even more pronounced if we had taken into account that the relative risks of physical activity may be larger for the elderly (Vogel et al. 2009
). The empirical evidence for higher relative risks in elderly related to long-term exposure to air pollution is weak; for example, in the large American Cancer Society study there were no differences in relative risk for PM2.5
(Pope et al. 2002
). We did not include children in our assessment because they are unable to drive a car, so a modal shift is not possible. Because of our focus on mortality effects (being extremely rare in children), we could not quantitatively compare risks for children as car passengers or as cyclists for physical activity and air pollution. The benefits of physical activity in children are considered important, however, both for current and for future health.
Overall relative risks may largely reflect the response from sensitive subgroups. For all stressors, the elderly are likely more susceptible, and we documented in an additional analysis that the ratio of benefits and risks was highest for ≥ 65-year-olds. For air pollution, subjects with preexisting cardiorespiratory disease may be more susceptible, and for physical activity, sedentary people may be more susceptible; these are subgroups that may partly overlap. Hence, both the risks and benefits may be higher than in the population average analysis.
In summary, it is unlikely that the conclusion of substantially larger benefits from cycling than risks is strongly affected by the assumptions made in the scenario, including the use of data from the Netherlands. Because concentration–response functions are mostly based on studies in Europe and North America, they may not apply in developing countries. For air pollution, there are no studies on long-term mortality effects in developing countries. The generally higher ambient air pollution concentrations could lead to higher losses in life-years comparing cycling and car driving. Traffic accident statistics for the Netherlands are probably not transferable to developing countries. For physical activity, there is evidence from a Chinese study (Matthews et al. 2007
), with very similar benefits. Hence, very large differences in concentration–response functions for air pollution and traffic accidents from the functions we used would be necessary to tip the balance between benefits and risks.
For air pollution, there is considerable evidence that long-term and short-term exposures are related to increased cardiopulmonary mortality (Brunekreef and Holgate 2002
). There are no studies of mortality effects specifically related to in-traffic exposures. We estimated the effect of shifting mode using two major long-term mortality cohort studies (Beelen et al. 2008
; Pope et al. 2002
), making assumptions about the contribution of traffic participation to the total inhaled dose of PM2.5
and (diesel) soot. Relative risks comparing cycling and car driving were small for both approaches, with the lower estimates based upon BS probably most realistic, because this component is more specific for traffic emissions.
The actual risk may be smaller because cyclists could more easily choose a low-traffic route. The substantial influence of route has been documented in various monitoring and modeling studies (Adams et al. 2001
; Hertel et al. 2008
; Kingham et al. 1998
; Strak et al. 2010
). A study in Utrecht found 59% higher UFP exposure for cyclists along a high-traffic route compared with a low-traffic route (Strak et al. 2010
). Walking close to the curb in London greatly increased personal exposures (Kaur et al. 2005
). For cyclists, position on the road is likely important as well, because it determines distance to motorized traffic emissions. Urban planning may also contribute by separating cycle lanes from heavily trafficked roads (Thai et al. 2008
For society, reduced overall air pollution levels may result in lower mortality from long-term exposure of city dwellers. The potential benefits we estimated based on NO2 reductions were in the same order of magnitude as the potential risks for the individuals shifting.
shows that the modal shift will lead to an increase in traffic accident deaths. The relative risk may be lower than we used because of the “safety-in-numbers” effect [Supplemental Material, Figure 2 (doi:10.1289/ehp.0901747)]. Car drivers may take more account of cyclists, resulting in fewer accidents per car-kilometer, when cyclists form a bigger part of the traffic (Jacobsen 2003
). Traffic fatality and injury rates in Germany and the Netherlands (with relatively high levels of cycling and walking) were relatively low compared with those of the United States (Pucher and Dijkstra 2003
). However, whether this reduction is attributable to a safety-in-numbers effect or a result of more biking lanes cannot easily be disentangled. The WHO concluded that if promotion of active commuting is accompanied by suitable transport planning and safety measures, active commuters are likely to benefit from the safety-in-numbers effect (WHO 2007b
). The relative risks could also be higher because the less experienced cyclists making the shift could be more vulnerable to accidents. We cannot quantify this effect.
Even when origin and destination are the same, cars and bicycles often take different routes (Witlox 2007
). The same short trip for a car may be 20–50% longer than for a bicycle (our calculations are based on comparisons per kilometer). If we could make a trip-based comparison, a lower relative risk for fatal accidents for cyclists compared with car drivers would be found. Furthermore, we did not take into account the concept of constant travel time budgets (van Wee et al. 2006
): A change of travel time will be compensated by a change of destination. When taking the bicycle, the shop next door is preferred over the shop with greater choices farther away. These factors would lead to lower relative risks than we used.
Relative risks for different physical activity definitions (total physical activity, meeting the physical activity guideline, active commuting) were quite consistent. An important issue is whether the comparison between subjects with lower and higher physical activity can be used to assess the health effects of a change in physical activity related to a shift toward active commuting. Bauman (2004)
showed that persons who were already in the highest quartile of fitness at baseline had a significantly lower mortality when they became even more active. In another study, people who went from unfit to fit over a 5-year period had 44% relative risk reduction compared with people who remained unfit (Blair et al. 1995
). The largest improvements in health status are seen in inactive persons who change their lifestyle and become physically active (Warburton et al. 2006
). A review by Erikssen et al. (1998)
suggested similar health benefits from an increase in physical activity for active and sedentary persons. Already active persons could have lower benefits of the extra physical activity, leading to relative risks up to 0.90. If only active persons shift mode of transport, lower overall benefits of cycling compared with car driving will be found (ratio of life-years gained vs. lost, 4 instead of 9).
An increase in cycling does not necessarily lead to an increase in total physical activity, if it is associated with reduced activity in another domain (Forsyth et al. 2008
; Thomson et al. 2008
). The empirical evidence for substitution is weak, and increased fitness could also lead to more physical activity in leisure time. If we assume that for 25% of the population no health gains occur because of substitution, the ratio of benefits to risks (central estimates from ) would be reduced from 8.9 to 6.7. Only if for 89% of the population no increase in total physical activity occurs because of substitution would benefits and risk become equal.
We have not considered the negative effects of physical activity on health—namely, musculoskeletal injury and fatal and nonfatal cardiac events (Institute of Medicine 2007
). Cycling can be considered a moderate type and duration of sport and has lower injury risk than do more vigorous types (running, scholastic athletics) and longer durations of physical activity (Hootman et al. 2001
; Parkkari et al. 2004
). Exercise has acute cardiac risks as well, but the absolute risk of a cardiac event during exercise seems to be low (Institute of Medicine 2007
). Regular physical activity also reduces the acute risk of a cardiovascular event (Tofler and Muller 2006
Restriction to mortality
We limited the quantitative assessment to mortality. It is difficult to evaluate the comparison between cycling and car driving if morbidity is included because of the lack of solid concentration–response relationships for air pollution and physical activity for morbidity outcomes. A meta-analysis reported a consistent positive association between physical activity and health-related quality of life (Bize et al. 2007
). The largest cross-sectional study showed that people meeting the recommended levels of physical activity had an adjusted odds ratio of “having 14 or more unhealthy days during the previous months” of 0.4 (95% CI, 0.36–0.45) over the inactive subjects (Bize et al. 2007
). Quality of life may even further improve apart from the increases in life-years. Concentration–response functions for air pollutants and morbidity outcomes such as hospital admissions are lower than for mortality: in the range of 1% compared with 6% per 10-μg/m3
increase in PM2.5
). Traffic injuries may differ even more between cyclists and car drivers than fatal accidents [Supplemental Material, Table 4 (doi:10.1289/ehp.0901747)], if underreporting of especially cyclist accidents is accounted for. This would reduce the ratio between benefits and risks.
We did not include all stressors in the quantitative evaluation. Cycling contributes to other benefits, including reduced emissions of carbon dioxide relevant for reducing climate change, reduced use of physical space (e.g., related to parking), and reduced traffic noise for city dwellers, which may result in less annoyance. We are not aware of exposure studies or health effects studies that have compared traffic noise during transport for cyclists and car drivers.