Farren et al (1) state that they, ”were surprised by the prevalence of PTSD symptoms in our study”, and that, ”Future research should be aimed at assessing the risk factors for PTSD”.
My research of 1993, into Post Traumatic Stress Disorder in women who have undergone Obstetric and/or Gynaecological Procedures (2), which was widely reported in the press, obstetric and midwifery literature, including the N.I.C.E Guidelines, identified several risk factors for the development of PTSD in the women affected.
The key areas in the procedures seem to be those concerning the level of control which the woman has, the attitude of the doctor, the degree to which the patient’s views are heeded, and whether or not consent was perceived to have been given for the procedure. The gender of the examiner did not seem to be relevant.
The women were asked in the questionnaire to describe their experiences, some of which were quite explicit and reminiscent of assault: “dehumanising and painful”; “degrading and distressing….no account was taken of my feelings”; “my opinions were dismissed as irrelevant although it was my body which was being invaded”; “I came away hurting and feeling violated”; “very brutal internal was excruciating”; “you begin to feel like a thing”; “I cried and shouted but was held down and told to stop making a noise”; “humiliation…as if it happened yesterday”; “I felt assaulted and then abandoned”; “I felt abused…like a piece of meat on a slab”; “the most terrifying I have ever experienced….I was a carcass to be dealt “; “I felt mutilated by the procedure”; “traumatic and brutal”; “I was offered no information whatever, hence my extreme fear”; “it felt undeniably like rape”.
There appear to be similarities between a traumatic obstetric/gynaecological experience and the experience of sexual assault, particularly where the woman perceived herself to be powerless. PTSD is a recognised sequela of sexual assault and factors influencing its onset seem to be related to experiences which are perceived as painful, humiliating, mutilating and occurring in an unsympathetic environment in which the woman feels powerless to resist.
Perhaps the authors could return to the patients in the study and ask them to identify the features of their experiences which may have triggered their PTSD?

Dr Janet Menage M.A. M.B. Ch.B., GP (retired); qualified psychological counsellor.
Competing interests: None

(1) Farren J, Jalmbrant M, Ameye L, et al. Post-traumatic stress, anxiety and depression following miscarriage or ectopic pregnancy: a prospective cohort study. BMJ Open 2016;6:e011864. doi:10.1136/bmjopen-2016011864
(2) Menage, J (1993) Post-traumatic stress disorder in women who have undergone obstetric and/or gynaecological procedures: A consecutive series of 30 cases of PTSD. Journal of Reproductive and Infant Psychology, Vol 11, pp221-228

In their recent paper Dodds et al. analysed cycling-related injuries recorded in the NHS England Trauma Audit and Research Network (TARN) Database for the period from 14 March 2012 to 30 September 2017(Dodds et al., 2019). They claim their methods show an association between cycle helmet use and reductions in, crude 30-day mortality, severe traumatic brain injury (TBI), intensive care unit requirement and neurosurgical intervention. Cycle helmets are light structures, generally weighing 250g to 600g, and are typically composed of a thin shell of stiff plastic outside a thicker shell of expanded polystyrene foam. The standard approval tests simulate simple falls, with no other vehicles involved (ROSPA, 2018). They are not rated for high-energy impacts involving moving motor vehicles. Eighty four per cent of fatality and serious cyclist accidents reported to the police involve motor vehicles (ROSPA, 2017). Cycle helmets are shaped to "cup" the top of the skull rather than enclose the head, which means they need a system of well-adjusted straps to stay attached in the event of a crash. When worn by non-enthusiast cyclists they are often poorly adjusted or the wrong size (Parkinson and Hike, 2003; Thai et al., 2015). When claims are made of "significant correlation between use of cycle helmets and reduction in adjusted mortality and morbidity associated with TBI and facial injury" (Dodds et al., 2019), then some observers will be immediately sceptical and seek closer testing of such claims. Scepticism will be intense when, as here, Dodds et al. make no apparent distinction between injuries from collisions with moving motor vehicles and those resulting from simple falls.

It is long reported that markers associated with family income or social status show an inverse relationship with predisposition to injury and death. For children, being poor is associated with increased risk of injury from a range of sources: traffic, drowning, poisoning, burns, falls (Laflamme, 2010). This pattern is well characterised for children, but it also exists for adults. In England and Wales a random sample of nearly 300,000 people aged between 16 and 65 were selected at the 1981 census and followed up for nearly nine years (Slogett and Joshi, 1994). It was found that adverse personal or household socioeconomic factors were associated with all-causes mortality; living in rented accommodation and not having access to a car were powerful predictors of mortality. For traffic injuries it is well established that child pedestrians in deprived areas have increased risk of being in collisions with motor vehicles. The same relationship is reported for child cyclists. In their 2010 review, Laflamme et al. cite 11 papers – from the UK, Sweden, Canada and Ireland – that reported increased risk of cycling injuries for children either of families in a lower socioeconomic position or living in the most deprived areas.

In 1997, for England and Wales, Roberts calculated child mortality rates over eight years using data for 1985–1992 and stratified these according to a five-level parental social-class score (Roberts, 1997). Child cyclists of social class I (the wealthiest) had a motor vehicle collision death rate of 2.4 per 100,000 per eight years, while the death rate for social class V children (the poorest) was 10.7 per 100,000 per eight years – a nearly 4.5-fold difference. Edwards et al. conducted a similar exercise for England and Wales, analysing rates of death from injury in children using the eight-class version of the National Statistics Socioeconomic Classification (NS-SEC) (Edwards et al., 2006a). They found that compared with children in NS-SEC 1 (higher managerial or professional occupations), children of parents classified as never having worked or as long-term unemployed (NS-SEC 8) had a death rate that was 27.5 (6.4 to 118.2) times higher.

Hippisley-Cox et al. conducted a cross-sectional survey of hospital admission data for injury for children aged 0–15 in the Trent region (UK) for 1992–1997 (Hippisley-Cox et al., 2002). Depending on electoral ward of home address, patients were categorised according to a five-level Townsend Score assigned to each ward. The score serves as a proxy for material deprivation, with variables for unemployment, overcrowding, lack of a car, and non-owner occupation. The authors found socioeconomic gradients for both number of injuries and injury severity. Children from the least deprived electoral wards had a pedal-cycle injury rate of 25.6 (0.0 to 48.0) per 10,000 children. The equivalent rate for the most deprived quintile was 46.0 (22.2 to 66.1).

Silversides et al. carried out a cross-sectional survey of 479 injuries in Belfast children aged 0–12 and compared rates between the most and least deprived areas, based on a locally developed Noble Economic Deprivation score (Silversides et al., 2005). They reported a significant correlation between economic deprivation and injury rate (r = 2.14, p <= 0.001). Children from the most deprived areas were particularly likely to be involved in road traffic accidents (relative risk RR = 3.25, p = 0.002). For cycling injuries, children from the least deprived areas had a mean injury rate of 0.95 per 1000 children. The rate for child cyclists from the most deprived areas was 2.70 per 1000 children (relative risk RR = 2.43, p = 0.22).

Edwards et al. (2006b) used a ten-level Index of Multiple Deprivation (IMD) based on 36 indicators across seven domains of deprivation. London was divided into 4,765 geographical zones in 33 boroughs, with each zone assigned an IMD score. Road traffic injury collision data was analysed with reference to each zone. The cycling injury rate for adults in the most deprived areas was 2.1 (1.5 to 2.8) times that for adults in the least deprived areas. There was also some suggestion of a relationship between injuries to child cyclists and deprivation, but numbers were too low to constitute good evidence. In the US research has tended to focus on disparities in cycling injury based on ethnic background rather than direct measures of socioeconomic status. Barajas reviewed 7,088 bicycle crashes over a three-year period in the San Francisco Bay Area (Barajas, 2018). Black cyclists were involved in the most crashes per person and per distance travelled; their crash was rate nearly eight times that of White cyclists. Hispanic cyclists experienced bicycle crashes at a rate 2.5 times greater than White cyclists per distance travelled.

One candidate identified for socioeconomic disparities in traffic injuries is simply that the road conditions faced by the less wealthy are more hostile. Compared with wealthy areas, deprived areas may tend to have higher traffic levels, more roads of an arterial nature, and junction geometries that are more hostile for vulnerable road users (Morency et al., 2012). The less wealthy also have restricted choices about when and how they can travel. What is sometimes termed “Transport Poverty” is a field of research in itself (Titheridge et al., 2014). Medical doctors, stockbrokers or IT consultants have choices about where and when they might choose to cycle. A recent immigrant who works by cycling as a food delivery courier is in a very different situation.

The particular relevance of the 1997 Roberts paper is that it covers a period before cycle helmets were in widespread use (1985–1992). In 1994, the Transport Research Laboratory published a literature review on cycle helmets (Royles, 1994). Based on limited sources, they reported an adult cycle-helmet-wearing rate of 3% for Southampton in 1986. For 1991, the review suggests potential junior school wearing rates of 4%–13% and potential secondary school wearing rates of 2%–7% in 1991 (the lower value is for “always wore a helmet”; the higher value is the rate of helmet ownership). This establishes an argument that the socioeconomic gradient in cycling deaths and injuries pre-dates helmet use and was present before various observational studies that claimed helmet wearing was associated with reduced rates of death and injury.

Observational studies such as the paper under discussion cannot be used to show causation and are vulnerable to bias by confounding factors. Lee Johnson writes that "All studies have weaknesses; observational studies have the scientific weakness that they can be used only to find associations between risk factors and responses, but alone they cannot establish causation" (Lee Johnson, 2018). Associations obtained using observational studies are often portrayed to the public and politicians as having much greater weight than the methods actually allow. What is an interesting statistical correlation in an academic paper can become the direct basis of policy and legislative changes and may inform findings of fault in court cases concerning traffic collisions. It is in the public interest that methods used to show such correlations be valid and avoid bias. Selection bias is one form of bias and occurs “when individuals or groups in a study differ systematically from the population of interest leading to a systematic error in an association or outcome” (Nunan et al., 2017).

The healthy-user bias or healthy-user effect is a type of selection bias where a group receiving an intervention inherently have better health and health outcomes than a comparison group. This has resulted in observational studies that produced exaggerated claims of benefit that were later overturned when assessed by other methods (Shrank et al., 2009). In their own discussion, Dodds et al. acknowledge that there are numerous confounders. But they do not discuss the potential presence of selection bias or a healthy-user effect. Instead they state that "the primary find of this paper can be considered robust to selection bias", and that they assumed the populations of helmeted and unhelmeted cyclists to be similar.

It has been shown above that for persons using bicycles the exposure to risk is not uniform but follows gradients associated with socioeconomic status and related markers. In this context, higher socioeconomic status can be treated as a source of healthy-user effect that may act as confounder of observational studies of cycling injuries and deaths. The ownership and wearing of cycle helmets is associated with markers of socioeconomic status and can of itself be treated as such a marker. A cross-sectional study of schools in Nottingham found that children from more deprived areas had lower rates of helmet ownership than children in less deprived areas (Kendrick and Royal, 2003). Children in more deprived areas who owned bicycles were more likely to use them than their peers in less deprived areas.

Lang investigated factors associated with helmet wearing among British children using survey responses (Lang, 2007). This work divided the heads of household into six social classes ranging from "Professional" to "Unskilled". Respondents reporting that they always wore a helmet showed a socioeconomic gradient. Children of the "Professional" social class reported an "always wears" rate of 28.5%. The "always wears" proportion decreased for each subsequent class cohort, with respondents in the "Unskilled" class recording a rate of 8.8%. Similar associations between helmet wearing and household income are reported in the US (Jewett et al., 2012) and Canada (Irvine et al., 2002). In the US case, adults having household income greater than $85,000 a year were more likely than those earning less than $40,000 to wear helmets (adjusted prevalence ratio 1.13, 95% CI: 1.05, 1.20) (Jewett et al., 2012). For adults in Ontario, Irvine et al. found adjusted odds ratios associating helmet wearing with high-income households (OR = 1.43, 95% CI 1.06, 1.93 p = 0.02) and holders of university or college degrees (OR = 1.68, 95% CI 1.30, 2.18 p = 0.00).

Direct surveys of UK adult helmet wearing and socioeconomic status are more difficult to find, but it is possible to pull together inferences from several sources. Compared with the general population, the helmet wearing rate among enthusiast sports cyclists (“club cyclists” or “road cyclists”) is very high, typically 100% for organised events. In recent years this type of cycling has become associated with higher-income groups: An LSE report on the British Cycling Economy calculates the cost of entry for a new participant in the cycling enthusiast cohort at £2,495 to cover the bicycle and accessories (Grous, 2011). Membership of British Cycling, the main sports cycling body, grew from 75,000 to 150,000 between 2013 and 2019 (British Cycling, 2013, 2019).

A 2008 Transport Research Laboratory report on surveys of helmet-wearing rates notes that the rate among cyclists in London (69.5%) was significantly higher than the rate outside London (29.9%) (Sharratt et al., 2009). Reports compiled by Transport for London remark that the cycling demographic in London is disproportionately dominated by people who are typically white, under 40, male, with medium to high household income (Transport for London, 2011). Most recently it was reported that, on some routes, the proportion of cyclists from households with incomes above £75k/year is significant (often exceeding 20 or 30 per cent) (Transport for London, 2019). The 2008 TRL helmet survey report also noted that the helmet-wearing rate on recreational routes was higher. Weekend surveys of two recreational routes, the Ridgeway cycle track in Oxfordshire and the Bath to Bristol cycle route, recorded a helmet wearing rate of 52.1% versus a weekday wearing rate of 34.2% at other sites (Sharratt et al., 2009). English National Travel Survey (NTS) data and Department for Transport statistics indicate that members of high-income quintiles cycle more than those in low-income quintiles, and a greater proportion of that cycling is recreational.

Based on data in NTS Table NTS0705, in the period 2012 to 2017 for cycling as the main mode, the average trips per person per year was 15 for the lowest real-income quintile but 19 for the highest real-income quintile (Department for Transport, 2019). For distance travelled, the disparity was large: the annual average distance per year was 40 miles for the lowest real-income cohort and more than double that, 82 miles, for the highest real-income cohort. Data in Table CW0203 gives a breakdown of utility to recreational cycling for five NS-SEC Occupation classifications (Department for Transport, 2016). The ratio of utility to recreational cycling was 1:1.5 for the “Never worked and long-term unemployed” class but 1:1.9 for the “Higher managerial, administrative and professional” class. A 2015 review of pedestrian and cycling casualties in Scotland noted a recent increase in adult cyclist casualties, with the greatest increase occurring in the least deprived quintile; the authors speculate that this may be due to increased prevalence of cycling (Whyte and Waugh, 2015).

Researchers who separate patients into groups according to whether or not they wear cycle helmets are selecting their groups based on a marker associated with socioeconomic status. Higher socioeconomic status is associated with reduced risk of injury and death. Any unadjusted method based on such selection should be assumed to be inherently biased towards showing a reduction in deaths and serious injuries. Various observational studies claim to have shown an association between helmet wearing and reductions in fatality and serious injury (Cyclehelmets.org, 2020). A better explanation is that these are simply detecting a side-effect of other factors also associated with socioeconomic status. In this author’s view, nothing meaningful about the effects of cycle helmets, whether positive or negative, can be concluded from studies using the methodology employed by Dodds et al.

In order to draw meaningful conclusions about the likely influence of cycle helmets, some form of matched analysis would be needed. The matching would need to involve attributes of the patients such as age, sex, alcohol consumption or being a holder of a British Cycling licence. Matching should also include the type of cycling: recreational, utility, off-road, etc. Matching would also be needed on attributes of the crash, such as involvement of other vehicles, type of vehicle (e.g., HGV or car), urban vs rural location, class of road, proximity to junctions, time of day, weather conditions, and so on. Most of these factors have a bearing on likely crash severity (ROSPA, 2017).

Dodds et al. do not appear to have classified their patients according to whether the injuries resulted from a collision with a moving motor vehicle or from simple falls. This by itself will invalidate their findings for some. Time of day is important because although cycling crashes at night are less common, they are much more likely to be severe, with a high rate of fatality. The same arises with urban vs rural locations, the latter having a higher fatality rate. Attributes associated with socioeconomic status, such as occupation, salary or education level, may serve to stratify patients by a marker of likely exposure to risk. It may be possible to use a deprivation index for the district of the crash (e.g., based on census data zone or postcode) as an indirect indicator of likely road conditions, but it is not clear that the TARN database captures this information accurately.

Like the use of cycle helmets, membership of the medical profession is a marker of high socioeconomic status. When doctors conduct crude injury comparisons of helmet wearers vs non-wearers, they are using a marker of their own social status to make what will inevitably become public assertions about the “reasons” for death and injury among the rest of the population. Outside observers could feel that statistics are being used in a manner that amounts, however unwittingly, to criticising poor people for being poor.

Biographical note
Shane Foran was educated as a biochemist and worked in medical diagnostics before moving to IT. He has been a volunteer advocate for active travel for most of his adult life. He was a member of the editorial board of the Bicycle Helmet Research Foundation, which set up the Cyclehelmets.org website that was created in part to track concerns about the quality of research being published regarding cycle helmets. He is a "Right to Ride Representative" with Cycling UK (formerly the Cyclists' Touring Club) and contributes to the helmets working group of the European Cyclists' Federation. He is a member of the National Panel of Sustainability Experts for the Irish Department of Transport, Tourism and Sport. He is a qualified UK National Standard Cycling Instructor (Instructor registration number CTSB503841C).

References

Barajas 2018
J.M. Barajas
Not all crashes are created equal: Associations between the built environment and disparities in bicycle collisions.
J. Trans. Land Use, 11 (2018), pp. 865-882
https://www.jstor.org/stable/26622434 (Accessed January 19 2020)

British Cycling 2013
British Cycling reaches 75,000 member milestone
British Cycling, June 2013
https://www.britishcycling.org.uk/about/article/20130617-British-Cycling... (Accessed 15 January 2020)

British Cycling 2019
British Cycling reaches 150,000 member milestone for the first time
British Cycling, 2019
https://www.britishcycling.org.uk/campaigning/article/20190502-campaigni... (Accessed 15 January 2020)

Cyclehelmets.org 2020
Published evidence supportive of helmet effectiveness or promotion
Cyclehelmets.org 2020
https://www.cyclehelmets.org/1147.html (Accessed 19 January 2020)

Department for Transport 2016
[Dataset] Table CW0203 Walking and cycling levels demographic breakdown
Walking and cycling demographic tables in England, Department for Transport, published 12 July 2016
https://www.gov.uk/government/statistical-data-sets/walking-and-cycling-... (Accessed 09 January 2020)

Department for Transport 2019
[Dataset] Table NTS0705: Travel by household income quintile and main mode or mode: England
Statistical data set Mode of travel Data about people travelling by mode of transport, produced by Department for Transport, 2019
https://www.gov.uk/government/statistical-data-sets/nts03-modal-comparisons (Accessed 09 January 2020)

Edwards et al. 2006a
P Edwards, I Roberts, J Green, S Lutchmun
Deaths from injury in children and employment status in family: analysis of trends in class specific death rates
Br. Med. J., 333 (2006), pp. 119-121
DOI: https://doi.org/10.1136/bmj.38875.757488.4F
PMID: https://www.ncbi.nlm.nih.gov/pubmed/16829537
PMCID: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1502180/

Edwards et al. 2006b
P Edwards, J Green, I Roberts, C Grundy, K Lachowycz
Deprivation and Road Safety in London: A report to the London Road Safety Unit
London School of Hygiene and Tropical Medicine, London, 2006
http://content.tfl.gov.uk/deprivation-and-road-safety.pdf (Accessed 11 January 2020)

Grous 2011
A Grous
The British Cycling Economy ‘Gross Cycling Product’ Report
London School of Economics, 2011
http://eprints.lse.ac.uk/38063/1/BritishCyclingEconomy.pdf (Accessed 12 January 2020)

Hippisley-Cox et al. 2002
J Hippisley-Cox, L Groom, D Kendrick, C Coupland, E Webber E, B Savelyich
Cross sectional survey of socioeconomic variations in severity and mechanism of childhood injuries in Trent 1992-7
Br. Med. J., 324 (2002), pp. 1132–1132
DOI: https://doi.org/10.1136/bmj.324.7346.1132
PMID: https://www.ncbi.nlm.nih.gov/pubmed/12003886
PMCID: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC107914/

Irvine et al. 2002
A Irvine, BH Rowe, V Sahai
Bicycle Helmet-wearing Variation and Associated Factors in Ontario Teenagers and Adults
Can J Public Health 93 (2002) pp. 368–373
DOI: https://doi.org/10.1007/BF03404572
PMID: https://www.ncbi.nlm.nih.gov/pubmed/12353460

Jewett et al. 2012
A. Jewett, L.F. Beck, C. Taylor, G. Baldwin
Bicycle helmet use among persons 5years and older in the United States, 2012
J Safety Res, 59 (2016), pp. 1-7
DOI: https://doi.org/10.1016/j.jsr.2016.09.001/
PMID: https://www.ncbi.nlm.nih.gov/pubmed/27846992
PMCID: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5189688/

Kendrick and Royal 2003
D Kendrick, S Royal
Inequalities in cycle helmet use: cross sectional survey in schools in deprived areas of Nottingham
Arch Dis Child 88 (2003) pp. 876–880
DOI: http://dx.doi.org/10.1136/adc.88.10.876

Laflamme et al. 2010
L Laflamme, M Hasselberg, and S Burrows
Review Article 20 Years of Research on Socioeconomic Inequality and Children’s—Unintentional Injuries Understanding the Cause-Specific Evidence at Hand
Int J Pediatr. Volume 2010, Article ID 819687
DOI: https://doi.org/10.1155/2010/819687
PMID: https://www.ncbi.nlm.nih.gov/pubmed/20706660
PMCI: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2913857/

Lang 2007
I Lang
Demographic, socioeconomic, and attitudinal associations with children's cycle-helmet use in the absence of legislation.
Inj. Prev., 13 (2007), pp. 355-358
DOI: https://doi.org/10.1136/ip.2007.015941
PMID: https://www.ncbi.nlm.nih.gov/pubmed/17916896
PMCID: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2610622/

Lee Johnson 2018
L Lee Johnson 2018
Chapter 17 - Design of Observational Studies
In; J.I. Gallin, F.P. Ognibene, L. Lee Johnson (Eds) Principles and Practice of Clinical Research (Fourth Edition). Academic Press 2018 Pages 231-248
DOI: https://doi.org/10.1016/B978-0-12-849905-4.00017-4

Morency et al. 2012
P Morency, L Gauvin, C Plante, M Fournier, C Morency
Neighborhood Social Inequalities in Road Traffic Injuries: The Influence of Traffic Volume and Road Design
Am. J. Public Health. 102 (2012), pp. 1112–1119
DOI: https://doi.org/10.2105/AJPH.2011.300528
PMID: https://www.ncbi.nlm.nih.gov/pubmed/22515869
PMCID: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3483951/

Nunan et al. 2017
D Nunan, C Bankhead, JK Aronson.
Selection bias. Catalogue of Bias Collaboration Catalogue Of Bias 2017
http://www.catalogofbias.org/biases/selection-bias/ (Accessed 13 January 2020)

Parkinson and Hike 2003
G.W. Parkinson, K.E. Hike
Bicycle Helmet Assessment During Well Visits Reveals Severe Shortcomings in Condition and Fit
Pediatrics, 112 (2003), pp. 320-323
DOI: https://doi.org/10.1542/peds.112.2.320
PMID: https://www.ncbi.nlm.nih.gov/pubmed/12897281

Roberts, 1997
I Roberts
Cause specific social class mortality differentials for child injury and poisoning in England and Wales
J. Epidemiol. Community Health, 51 (1997), pp. 334-335
DOI: https://doi.org/10.1136/jech.51.3.334
PMID: https://www.ncbi.nlm.nih.gov/pubmed/9229067
PMCID: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1060483/

ROSPA 2017
Road Safety Factsheet: Cycling Accidents
The Royal Society for the Prevention of Accidents, November 2017
https://www.rospa.com/rospaweb/docs/advice-services/road-safety/cyclists... (Accessed 13 January 2020)

ROSPA 2018
Road Safety Factsheet: Cycle Helmets
The Royal Society for the Prevention of Accidents, March 2018
https://www.rospa.com/rospaweb/docs/advice-services/road-safety/cyclists... (Accessed 18 January 2020)

Royles 1994
MR Royles
International literature review cycle helmets. TRL Project Report 76.
Transport Research Laboratory, Crowthome 1994 (Accessed 08 January 2020)
https://trl.co.uk/sites/default/files/PR076.pdf (Accessed 08 January 2020)

Sharratt et al. 2009
C Sharratt, O Anjum, L Walter
Cycle Helmet Wearing in 2008, Report PPR420
Transport Research Laboratory Limited 2009
https://trl.co.uk/reports/PPR420

Shrank et al. 2011
WH Shrank, AR Patrick, A Brookhart
Healthy User and Related Biases in Observational Studies of Preventive Interventions: A Primer for Physicians
J. Gen. Intern. Med., 26 (2011), pp. 546–550
DOI: https://doi.org/10.1007/s11606-010-1609-1
PMID: https://www.ncbi.nlm.nih.gov/pubmed/21203857
PMCI: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3077477/

Sloggett and Joshi 1994
A. Sloggett and H. Joshi
Higher mortality in deprived areas: community or personal disadvantage?
Br. Med. J., 309 (1994), pp. 1470–1474
DOI: https://doi.org/10.1136/bmj.309.6967.1470
PMID: https://www.ncbi.nlm.nih.gov/pubmed/7804047
PMCID: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2541648/

Thai et al. 2015
K.T. Thai, A.S. McIntosh, T.Y. Pang
Bicycle Helmet Size, Adjustment, and Stability
Traffic Inj. Prev., 16(2015), pp. 268-275
DOI: https://doi.org/10.1080/15389588.2014.931948
PMID: https://www.ncbi.nlm.nih.gov/pubmed/249495311

Titheridge et al. 2014
H Titheridge, N Christie, R Mackett, Dl Oviedo Hernández, R Ye
Transport and Poverty A review of the evidence
University College London, 1 July 2014
https://pdfs.semanticscholar.org/a19d/f0e8b6bb5f402cae550568968f5db07200... (Accessed 12 January 2020)

Transport for London 2011
What are the barriers to cycling amongst ethnic minority groups and people from deprived backgrounds?
Policy Analysis Research Summary, Transport for London, November 2011
http://content.tfl.gov.uk/barriers-to-cycling-for-ethnic-minorities-and-... (Accessed 13 January 2020)

Transport for London 2019
Cycling Trends Update, Travel in London report 11
Transport for London, July 2019
http://content.tfl.gov.uk/cycling-trends-update.pdf (Accessed 13 January 2020)

Whyte and Waugh 2015
B Whyte, C Waugh
Trends in pedestrian and cyclist road casualties in Scotland
Glasgow Centre for Population Health, August 2015
https://www.gcph.co.uk/assets/0000/5206/Pedestrian_and_cyclist_casualtie... (Accessed 10 January 2020)