Bicycling is an active mode of transportation that integrates physical activity into daily life. The bicycle is an attractive alternative to the automobile as an urban mode of transport. Cycling is associated with a range of individual and public health benefits such as improved physical and mental health, decreased obesity, and reduced risk of cardiovascular and other diseases 123456, as well as ancillary benefits such as reduced emissions of noise, air pollutants and greenhouse gases 78. There are significant economic costs of physical inactivity 9, and benefit-cost analyses suggest that the benefits of increased cycling are worth approximately four to five times the costs of investing in new cycling infrastructure 1011. These potential benefits suggest that it is important to increase the use of the bicycle as a mode of active transportation.

It is clear that the health benefits of cycling are significant, and at this point there is no reason to assume that health risks outweigh those benefits. However, a full public health understanding requires that attention be paid not only to long-term population health and environmental benefits of bicycling, but also to the factors that influence risk of injury and fatality. Bicyclists are vulnerable because they must frequently share the same infrastructure with motorized vehicles, and yet bicycles offer their users no physical protection in the event of a crash. In addition, the mass of a typical automobile is at least an order of magnitude greater than a bicycle plus its rider, and motorized vehicles have top speeds that are considerably faster than bicycles. As a result, bicycle riders who are involved in a crash are exposed to a much higher risk of injury compared to motor vehicle users (with the exception of motorcycle riders).

To date, most studies of cycling safety - especially in North America - have emphasized helmet design, regulation, and implementation to mitigate the severity of cycling injuries when a crash occurs 1213. This is particularly true for children 1415. In many North American jurisdictions children who cycle (and sometimes also adult cyclists) are required by law to use helmets, although this is not the case in most European countries. While helmets are effective in reducing the severity of head injuries, they do not address impacts to other parts of the body 1617. More importantly, they do not prevent incidents from occurring in the first place 18, and legislating their use may even discourage cycling 19.

The built environment has been implicated as an important determinant of bicycling rate 20212223, but these relationships are complex and a positive correlation has not always been found 24. It is equally important to understand how the built environment affects bicycling safety because there may be an opportunity to prevent injuries by modifying transportation infrastructure. Infrastructure improvement meets several important conditions for successful injury prevention measures: (a) it is population based, rather than requiring initiative on the part of the individual; (b) it is passive, rather than requiring active participation; and (c) it is accomplished with a single action, rather than requiring repeated reinforcement 18.

In this paper we review the evidence on how different types of transportation infrastructure affect bicyclists' safety. This paper is organized as follows: first we provide an overview of bicycling safety and ridership. Next we offer definitions of, and alternative terminology for, the transportation infrastructure used by cyclists that might be expected to influence their safety (Table 1). We describe our literature search methodology and the inclusion and exclusion criteria, and present the results of the search in two detailed tables. Table 2 describes studies that assess the safety of intersections for cyclists, and Table 3 describes studies related to straightaways (i.e. roads, lanes, paths). We conclude by discussing the findings of this review, critiquing the methodological approaches used, and offering recommendations for future research.

North Americans remain less likely than Europeans to choose bicycle transport for either short or long trips. In part, this may be due to differences in urban form between North American and European cities, particularly density and interconnectedness 25, but perceived and actual injury risks are also important.

Data on the share of trips made by bicycle are not often directly comparable between jurisdictions owing to differences in the survey methods employed (e.g. sampling scheme, definition of a trip, etc.), but comparisons are typically justified by the inability of these methodological disparities to explain the substantial difference observed (e.g. about 1% of trips are made by cycling in North America vs. an estimated 10% of trips in Switzerland, Germany, Austria, Sweden, Finland, and Belgium (Flanders), and more than 20% of trips in Denmark and the Netherlands) 26. Along with these lower cycling rates, there is also a higher risk of injury associated with cycling in North America: an analysis of traffic injuries indicated a two to three fold higher risk of death and an eight to 30 fold higher risk of injury while cycling in the United States vs. Holland and Germany, using either of the traditional transportation denominators: per trip or per kilometer traveled 27. While these comparisons underscore cycling injury risks, they also provide reason for optimism. If cycling is safer in European cities, it can be made safer in North America.

There are clearly bicycling safety and popularity "gaps" between (and within) Europe and North America 28. In addition, there is an important safety gap between cyclists and other transport modes: estimates from both continents suggest that cyclists are seven to 70 times more likely to be injured, per trip or per kilometer traveled, than car occupants 2729. It is likely that public perception of a lack of safety acts as a deterrent to cyclists in North America: in surveys asking about factors that affect the choice of cycling as a mode of transportation, concern about safety is one of the most frequently cited deterrents [3031323334, and Winters M, Davidson G, Kao D, Teschke K: Motivators and deterrents of bicycling: factors influencing decisions to ride, submitted]. For example, in a survey of adults in the Vancouver metropolitan area, the following were among the top deterrents: the risk of injury from car-bike collisions; the risk from motorists who don't know how to drive safely near bicycles; motorized vehicles driving faster than 50 km/hr; and streets with a lot of car, bus, and truck traffic 33. The good news is that there is evidence that perceived safety improvements in bicycle transportation have an aggregate elasticity value greater than one (i.e. a 10% increase in perceived safety results in greater than 10% increase in the share of people commuting by bicycle) 32.

Increased ridership rates may result in improved safety for cyclists: injury rates have been shown to decrease with increased cycling rates. This principle of "safety in numbers" is supported by studies of injury and ridership patterns in California, Australia, and Europe, as well as between cities and within cities over time 35363738. There are a number of potential explanations. Motor vehicle drivers may not expect cyclists when there are few of them on the roads, and thus make so-called "looked-but-failed-to-see" errors that can result in collisions 39. When motorists and cyclists are unaccustomed to sharing the road, both parties may hold incorrect assumptions about what the other party will do 40. Increased cycling rates may mean that more motorists also use bicycles as a mode of transport, making motorists more attuned to cyclists and their movements, and encouraging them to drive in a way that accounts for potential interactions 36. Finally, a larger cycling population means stronger lobbying power for cycling resources.

Finally, it is worth considering long-term temporal trends in motor vehicle injuries. The injury rate from motor-vehicle crashes has steadily declined since the 1920s in many parts of the world, in part attributable to improvements in road-related infrastructure 41. This provides reason for optimism: the risk of injury or death from traffic crashes is modifiable, and this is likely to extend to the infrastructural determinants of cycling injuries.

In total, 23 papers were identified that met the inclusion criteria. Eight examined infrastructure related to intersections, and are abstracted in detail in Table 25354555657585960. Fifteen papers examined infrastructure related to "straightaways", i.e. roads, lanes, paths, etc., and are abstracted in Table 3162961626364656667686970717273. Studies are presented in the tables first by type of infrastructure, then by year for each type.

Ten of the 23 studies reviewed used injuries (or both injuries and crashes) as a metric of bicyclist safety, four examined injury severity, and nine examined crashes (i.e. collisions and/or falls). Most of the studies were published since 1994, except two US studies which were published in the mid-70s 6162 and one which was published in 1988 63. All the study designs were observational. Five of the intersection-related papers 5356575859, but only one of the road/lane/path-related papers 63, were "before-after" studies that quantified the change in cyclist safety before and after some infrastructure-related intervention took place. The remaining papers were classified as "non-intervention" observational studies. Most of the studies based their analyses and conclusions on at least 150 observations of injury or crash events, and seven studies based their analyses on more than one thousand observations. However one study of roundabouts examined only 67 crashes, 58 of which resulted in injuries 54, and two non-intersection studies examined 87 and 89 crashes on roads with and without marked bike lanes 63, and on sidewalks versus roads 71 respectively.

Thirteen of the studies were published in public health related journals (mainly Accident Analysis & Prevention and the Journal of Safety Research), and nine were published in transportation engineering journals (mostly Transportation Research Record). The remaining study (on the safety of different road/lane/path infrastructure types) was conducted as part of a Master of Science thesis 61.

All but one of the studies about intersection-related infrastructure (Table 2) were conducted in European countries. Five of the European intersection-related studies examined the safety of roundabouts and two examined marked bicycle crossings. The non-European study examined how intersection design in Japan influenced number of bicycle-motor vehicle collisions 60. Cyclists in Japan are required by law to travel on the sidewalk, so the results from this study may not be generalizable to countries with different traffic rules.

The findings of the roundabout studies show some consistency, with elevated risks for cyclists after installation of roundabouts with multiple traffic lanes or with marked bike lanes, whereas there were risk reductions or no apparent increase in risk at roundabouts with separated cycle tracks 545657. One study showed a decreased risk for cyclists and moped riders after installation of roundabouts in the Netherlands 53, but the authors did not disaggregate the results for these two road-user groups. The finding from this study - that roundabouts with separated cycle tracks had a greater safety effect than those with on-road marked bike lanes or no bicycle infrastructure - is consistent with other research. Another study on roundabout safety in Flanders found a similar effect for "vulnerable road users" 74, but we have not included this study in our table because the vulnerable road user population included pedestrians and motorized two-wheeler riders as well as cyclists. It is likely that the safety effect of roundabouts, as measured in such "before-after" studies, will depend on the "before" configuration of the intersections in question.

The two studies of the safety effect of marked bicycle crossings at intersections looked at different design aspects (one on physically elevated crossings, one on colored crossings) and did not provide clear conclusions. Although the study on elevated crossings showed a small increase in the number of crashes after the crossing was installed, the bicycle traffic volume grew by 50% on the streets after the intervention, as compared to unchanged streets in the area, and this was not adjusted for in the analysis 58. The second study showed a reduction in injury or crash risk when there was one colored bicycle crossing at an intersection, but an increase in injury or crash risk when there were two or more colored crossings 59.

Of studies examining infrastructure related to straightaways on roads, lanes, and paths (Table 3), all but one were conducted in Canada or the US. The only European study in this category is very different in its focus: the safety effect of rural street lighting in the Netherlands 73. Perhaps unsurprisingly, that study found that the presence of street lighting on rural roads reduced the rate of cyclists' injuries by half. The effect was corroborated by an injury severity study that found that crashes resulting in more severe injuries were significantly associated with unlit roads at night 69.

Most of the remaining studies in this category compared cyclist injury or crash rates on different types of road- or path-related infrastructure that cyclists commonly travel, namely major and minor roads without specific cycling facilities, roads with wide curb lanes or marked bike lanes, on-road bike routes, off-road bike-specific or multi-use paths, and sidewalks. A difficulty with this literature was that several facilities (between two and seven in number) were grouped into categories, such that facilities with potentially different risks were classified within a single category. In addition, the categorizations differed from study to study, and the terminology used was sometimes not clearly defined or consistently used. Despite these limitations, there are still some consistent messages.

On-road marked bike lanes were found to have a positive safety effect in five studies, consistently reducing injury rate, collision frequency or crash rates by about 50% compared to unmodified roadways 6162656667. Three of those studies 616667 found a similar effect for bike routes. One study 63 found that there was an increase in crash rates in the year following installation of marked bike lanes on a major road, especially for a section counter to on-road traffic flow, but this effect was not sustained over the long term.

There is less consistent evidence about off-road riding, possibly because this category encompassed a wide variety of facility types. There may have been confounding factors such as whether the surface was paved or unpaved, or for bicycles only or multiple user groups. Two studies examined off-road bike paths and found reduced risks, ranging from 0.11 to 0.67 times the risk of cycling on minor roads 6467. Two studies that grouped paved and unpaved, bicycle only and multi-use urban trails in their off-road path category found elevated risks, 1.6 to 3.5 times higher than riding on-road 296668. Studies that examined unpaved off-road trails as a separate category found risks of injury 2.5 to 7.2 times higher than on-road cycling 616566 and 8 to 12 times higher than bike routes, lanes, or paths 6566.

Most studies that considered sidewalk-riding suggested that it is particularly hazardous for cyclists, with estimates of 1.8 to 16 times the risk of cycling on-road 2966676871. However one study found that the risk of traveling on the sidewalk was the same or lower than riding on residential streets 64. Another considered the direction of travel and found that the elevated risk when sidewalk cyclists entered intersections was almost exclusively related to cycling against the flow of adjacent on-road traffic 71.

Four studies examined the association between various infrastructural characteristics and injury severity 16697072. More severe injuries were significantly associated with motor vehicle involvement, unlit roads at night, wider roads, perceptible road grades, and one-way streets. Injury severity does not reflect risk of an incident, but rather the outcome of the incident once it occurs. In comparison, the studies that examined injury or crash rates, as opposed to those that concentrated on injury severity, were our primary focus since we are most interested in shaping transportation infrastructure for injury prevention.

In this review we have described two categories of infrastructure: the first related to intersections; and the second related to straightaways on roads, lanes, and paths. It is of interest to note that studies of the former type of infrastructure were conducted almost entirely in Europe, while studies of the latter were conducted almost entirely in North America. The reason for this may be the substantial differences in urban form, existing cycling infrastructure, cycling rates, and even the culture of cycling between Europe and North America. Pucher and colleagues have discussed this issue extensively 2675. There is also significant variety in infrastructure design from one country to another, and even within a given city. Despite this, our review has revealed that relatively few types of infrastructure have been studied. For example, some common types of infrastructure in North American cities have not been assessed: traffic circles; bike boxes; sharrows; speed bumps/humps; and traffic diverters (Table 1). In addition, except for studies of roundabouts, we did not find any injury or crash studies that investigated cycle "tracks", a bicycle-specific design that is frequently available in high modal share European cities. One of the limitations of this review is that we have only included studies in the English scientific literature, although we are aware that there may be studies reported only in other (particularly European) languages.

The principal trend that emerges from the papers reviewed here is that clearly-marked, bike-specific facilities (i.e. cycle tracks at roundabouts, bike routes, bike lanes, and bike paths) were consistently shown to provide improved safety for cyclists compared to on-road cycling with traffic or off-road with pedestrians and other users. Marked bike lanes and bike routes were found to reduce injury or crash rates by about half compared to unmodified roadways. The finding that bicycle-specific design is important applies also to intersections with roundabouts, where it was found that cycle tracks routing cyclists around an intersection separately from motor vehicles were much safer than bike lanes or cycling with traffic. It has been suggested that the reason for high rates of bicycle-motor vehicle collisions at intersections is that motor vehicle drivers may be making "looked-but-failed-to-see" errors, whereby they search for oncoming motor vehicles but do not recognize that a cyclist is approaching because they are not looking for them 3940.

Although roundabouts at intersections are not common in North America, they are relatively popular in many European countries. It is possible that they may see more widespread use in North America in the future because of evidence that conversion of intersections to roundabouts reduces crash risk for motor vehicle road users by 30-50% 76, especially when they replace intersections that were not previously signal-controlled. However, because the cyclist-specific safety effect of roundabouts appears to be highly dependent on their design, transportation infrastructure planners should carefully consider interactions between cyclists and other traffic modes. A literature review on the safety effect of roundabouts, prepared for the 18th Workshop of the International Co-operation on Theories and Concepts in Traffic Safety 77, came to similar conclusions. It may be prudent to avoid installing roundabouts in areas where there is a high proportional volume of bicycle traffic, for example along designated bicycle routes on residential roads. In some North American cities there is retrofitting of "traffic circles" at intersections in residential areas. Since these are quite different from the larger-diameter roundabouts found in Europe, their effect on cyclist safety should be investigated before more widespread use is advocated.

The reviewed literature also confirms some things that may already be "common-sense" for transportation planners and safety experts: that streets used by cyclists at night should have good street-lighting, road surfaces should be paved and well-maintained, and bike routes should avoid excessive grades wherever possible.

An issue with the literature to date, especially that related to roads, lanes, and paths, is that some investigators did not define the terminology used. For example, the meaning of bike "path" was not defined in the paper by Tinsworth et al. 64. Other investigators clearly defined their infrastructure terms, but grouped facilities that may have different injury risks. For example, the studies of Aultman-Hall et al. 2968 defined paths as "an off-road (usually multi-use) paved or unpaved path or trail," grouping paths for bikes only, which were found by others to have lower risks than cycling on roads 6467, with unpaved trails, which were found by others to have higher risks 616566. Definitions of terminology are especially important in questionnaire-based studies to ensure that study participants are all answering with the same infrastructure in mind; photos can be helpful in this regard 33.

Clear and specific categorization is also vital to transportation planners and engineers, so they can distinguish sometimes subtle differences between successful and problematic design characteristics. One of the difficulties of the studies in the English-language literature to date is that the range of infrastructure studied is small compared to the range of configurations used between and within jurisdictions. Some examples are described above, but there are many other features that merit investigation: stop signs; numbers of roads intersecting; junctions such as driveways and lanes; cyclist lane of travel in relation to parked cars; surface features such as cobble stones or street-car (tram) tracks; traffic calming measures such as diverters or road humps; and road/lane/path curvature.

Underreporting of some events is an issue that is common to all studies of bicycle injuries and crashes. Many of the studies reviewed here relied on administrative data sources including hospital records 166264, police reported accidents 54555657585960616970717273, and national or city-maintained registries 5363, all of which are likely to miss less severe events. For example, one of the large surveys 67 found that 9.8% of the respondents had had a crash in the last year, but only two in five crashes (38.2%) had been reported to police. Over half (56.6%) required medical attention, but only one in twenty crashes (5.5%) required admission to a hospital. This underreporting may create bias in infrastructure-specific risk calculations, since collisions involving motor vehicles may be more likely to be reported to police for insurance reasons and to hospitals because they are more severe, as compared to collisions that happen with non-motorized users (which may happen more frequently on off-street paths). Results of studies using these data sources should be interpreted as reflecting risk of severe events. Other studies in this review used data from cyclist surveys that may capture a wider range of crash types, including those that are less severe 296165666768. However, survey data will not capture events that resulted in fatalities (though these are extremely rare) or catastrophic incapacitating brain, spinal cord or other injuries and, depending on the method of survey administration, may not capture individuals who no longer cycle following a crash 2968. No single study design can overcome these reporting problems, thus the importance of looking for consistency of results across different designs.

A great challenge in studying cycling injuries is ensuring that comparisons control for the number of cyclists at risk (also called "exposure to risk"). The before-after studies reviewed here aimed to do this by comparing numbers of injuries on the same intersection or roadway prior to and post introduction of an infrastructure intervention, with the assumptions that underlying traffic levels, injury rates, and types of cyclists stay the same. These assumptions may not hold 58, so some of these studies also adjusted for temporal trends in traffic volumes 585963 or injury rates in the area 53, or made additional comparisons to unchanged intersections 56575859. The non-intervention studies needed to include methods to derive bicycling trip volumes on the infrastructure types being compared. Sometimes these came from administrative data collected by transportation authorities 5455607173, and sometimes from study participants describing the route of an injury trip or their typical cycling location 29616465666768. Injury severity studies made comparisons within the injured populations, so did not require trip volume denominators 16697072, but this meant that they examined differences in severity of the outcome once in an injury event, not the original risk of the event itself.

Though the most basic requirement for studies examining risk of crashes or injuries is to account for exposure to risk, there are many other factors that may confound comparisons and that ideally would be controlled in study design or adjusted for in analyses. For example, men and women or people in different age groups may choose to cycle on different facility types, and might have different skill levels or risk-taking behavior, thus creating the potential for confounding associations between infrastructure and injury. While it is difficult to control for all potential confounders, many of the non-intervention studies reviewed here did adjust for personal factors such as age 162964657071, sex 29646571, cycling experience 2968, bicycle type 65, and environmental factors such as time of day 6469707273 and weather 65697072. Most injury severity studies adjusted for helmet use 166972. A style of observational study that can control for most potential confounders is the case-crossover design 78. Such a study is underway in the Canadian cities of Toronto and Vancouver. It will compare infrastructure at the injury site to that of randomly selected control sites on the same trip, thus within-trip factors (including age, sex, cycling experience, propensity for risk taking, alcohol or drug use, bicycle type and condition, visibility via clothing or bicycle lights, weather, time of day, etc.) are controlled in the design.

Although the effect of infrastructure design on cyclist safety was first studied more than three decades ago, the literature on the topic remains remarkably sparse. This review highlights opportunities for more detailed and controlled studies of infrastructure and cycling injuries.

The evidence to date suggests that purpose-built bicycle-only facilities (e.g. bike routes, bike lanes, bike paths, cycle tracks at roundabouts) reduce the risk of crashes and injuries compared to cycling on-road with traffic or off-road with pedestrians. Street lighting, paved surfaces, and low-angled grades are additional factors that appear to improve cyclist safety. The major advantage of infrastructure modifications, compared to helmet use, is that they provide population-wide prevention of injury events without requiring action by the users or repeated reinforcement. Given the influence of safety on individuals' decisions to cycle, the importance of cycling modal share to safety, and the ancillary benefits of this active and sustainable mode of transportation, infrastructure enhancements have the opportunity to promote an array of improvements to public health.

OECD: Organisation for Economic Cooperation and Development; km: kilometer; ICD: International Classification of Diseases; AIS: Abbreviated Injury Scale.

The authors are part of a research team that is studying the association between bicyclists' injuries and the cycling environment in Vancouver and Toronto. The Heart and Stroke Foundation of Canada and the Canadian Institutes of Health Research have funded this three-year study. See http://www.cher.ubc.ca/cyclingincities/injury.html for more information.

KT, CR, AH and PC conceived of the study and developed the literature search strategy. CR and AH conducted the literature search. CR, AH, MW, and KT reviewed the included papers, and abstracted them for the detailed tables prepared for this paper. CR wrote the initial draft of the manuscript, and the other authors all contributed to its development, particularly the discussion. All authors reviewed and approved the final manuscript.