Reader Comments
Post a new comment on this article
Post Your Discussion Comment
Please follow our guidelines for comments and review our competing interests policy. Comments that do not conform to our guidelines will be promptly removed and the user account disabled. The following must be avoided:
- Remarks that could be interpreted as allegations of misconduct
- Unsupported assertions or statements
- Inflammatory or insulting language
Thank You!
Thank you for taking the time to flag this posting; we review flagged postings on a regular basis.
closeConsidering the claims again
Posted by Colin479 on 13 Oct 2013 at 14:36 GMT
Reading the article by Olivier and Walter I feel it claims are not quite in keeping with the available evidence.
The title “Bicycle Helmet Wearing Is Not Associated with Close Motor Vehicle Passing: A Re-Analysis of Walker, 2007” gives the impression that there is no connection between vehicles passing closer and helmet use. The report states; “Our analysis confirms Walker’s results regarding helmets and passing distance with an adjusted estimate of an additional 5.8 cm when not wearing a helmet. However, the magnitude of this effect is a 32% decrease from the unadjusted estimate and was less than the estimated effects of vehicle size, city and kerb distance all having a larger impact on passing distance.”
Using data provided in Table 8 provides for calculating the average passing clearance, my calculations suggest 1.519m for helmeted and 1.607m for non- helmeted, difference approximately 8.7 cm. Details provided show for up to 1.5m passing clearance, helmeted 584 readings from 1149, 50.8% of events and for non-helmeted, 518 from 1206, 42.9% of events. For passing clearance up to 1.0m, helmeted 60 of 1149, 5.2% and for non-helmeted 49 of 1206, 4.1%. It appears that in both results a higher proportion of helmeted events had vehicles passing closer. It seems that other factors have a greater effect on passing clearance than bicycle helmets but helmet use also has an effect based on the research, so I assume the title is not appropriate.
In the Introduction it states; “Despite Walker’s hypothesis, there is no clear evidence helmet wearing leads to an increase in injury risk for the cyclist.” This assertion opens a complicated issue of helmet use and the accident rate. Erke and Elvik stated: "There is evidence of increased accident risk per cycling-km for cyclists wearing a helmet. In Australia and New Zealand, the increase is estimated to be around 14 per cent."
Clarke 2007 referred to several reports suggesting helmet use increases the accident rate and data from 2012 shows for New Zealand that a higher accident rate occurred. The assertion that there is no clear evidence provides a point of view that may be incorrect.
The Introduction also includes “Also, despite strong evidence in support of bicycle helmet efficacy [6,8–10], and the benefit of mandatory helmet legislation [11–14], laypersons have used Walker’s findings to justify the removal of mandatory helmet laws [15,16].” References 6,8,10,11-14 do not relate to the accident rate per km of travel for helmet wearers and non-wearers, so they all suffer from substantial weaknesses. None of the references provided evaluates the health implications of reduced cycling associated with helmet legislation, e.g survey data for New south Wales show a 44% drop in children cycling from 1991 to 1993 following mandatory helmet legislation (Walker 1991). Dr Mayer Hillman from the UK's Policy Studies Institute calculated the life years gained by cycling outweigh life years lost in accidents by a factor of 20 to 1. In contrast to claiming a benefit from mandatory helmet legislation Curnow 2008 concluded: “Compulsion to wear a bicycle helmet is detrimental to public health in Australia but, to maintain the status quo, authorities have obfuscated evidence that shows this” and "Cycling declined after the helmet laws by an estimated 40% for children, with loss of the benefits of the exercise for health. As serious casualties declined by less, the risks to cyclists, including death by head injury, increased".
In brief, some of the comments and the title seem inappropriate and the conclusions may be incorrect.
RE: Considering the claims again
jakeolivier replied to Colin479 on 17 Oct 2013 at 05:15 GMT
Colin Clarke,
Our title comes from the analyses summarized in Tables 4, 5 and 8. When passing distances are categorized by the one meter rule, the adjusted odds ratio for helmet wearing is not statistically significant (aOR=1.13,=0.54). This is also true when categorized using the cut-point 0.75m (p=0.993). There are significant differences in passing distance when wearing or not wearing a helmet only for passing distances greater than 2m as noted in Table 8. Over different intervals, the average passing distance is very similar when Walker either wore or did not wear a helmet. So, for passing distances within 2m, there is no significant difference when wearing or not wearing a helmet. This is clearly not a safety issue as for passing distances less than 2m, the cyclist will not be struck by the vehicle and the lateral forces are not substantial for a heavy vehicle traveling at 100km/h (citation 17). Can you provide evidence that, when passing distance is greater than 2m, there is a safety concern? Can you also provide evidence that distances greater than 2m can be categorized as ‘close’ as in the title of our paper?
There is no need to calculate summary statistics from Table 8 for passing distances when wearing or not wearing a helmet as the data is readily available (see link to Walker’s website). Given as mean (standard deviation), they are 1.52m (0.35) when wearing a helmet and 1.61m (0.41). Average passing distances are less when wearing a helmet, but they are also less variable. This is an important distinction when performing any statistical analysis. Comparing only average passing distance for the two conditions does not constitute a statistical analysis as the variability of these estimates should also be considered. Passing distance stratified by helmet wearing appears reasonably normally distributed by QQ plots (again, this can easily be checked using the available data). We can therefore use the means and standard deviations in conjunction with the 68-95-99.7 rule. This rule states that 99.7% of observations from a normal distribution should be within three standard deviations of the mean. These intervals are [0.46m, 2.58m] when wearing a helmet and [0.39m, 2.82m] when not wearing a helmet. So, although average passing distance is less when wearing a helmet, there is greater probability in the left tail of the distribution of passing distances (i.e., greater probability of a close overtaking maneuver) when not wearing a helmet than when a helmet is worn. This result further supports the wording of our title.
You state “Details provided show for up to 1.5m passing clearance, helmeted 584 readings from 1149, 50.8% of events and for non-helmeted, 518 from 1206, 42.9% of events. For passing clearance up to 1.0m, helmeted 60 of 1149, 5.2% and for non-helmeted 49 of 1206, 4.1%. It appears that in both results a higher proportion of helmeted events had vehicles passing closer. It seems that other factors have a greater effect on passing clearance than bicycle helmets but helmet use also has an effect based on the research, so I assume the title is not appropriate.” Summarizing Walker’s sample of n=2355 overtaking events gives us those percentages. If we took another sample of n=2355 overtaking events, would those proportions be the same? The answer is most definitely not. That is true from sample to sample. That is precisely why we must take variability of our estimates into account and that is exactly what our analysis does. As noted above, the adjusted odds ratio for safe/unsafe passing distance by the one meter rule for helmet wearing is small and not statistically significant (aOR=1.13,=0.54). Also, the inclusion or exclusion of helmet wearing from this model does not drastically influence the other estimates. From Table 8, average passing distances are incredibly similar at various intervals – 0.66m vs. 0.61, 0.90m vs. 0.90, 1.29m vs. 1.30 and 1.70m vs. 1.72 when wearing and not wearing a helmet respectively. A difference in average passing distances is only noticeable after two meters. Therefore, helmet use did not have an effect on close overtaking based on this research.
The Erke and Elvik technical report can be found here. Has this work been peer-reviewed? The vast majority of such reports are not.
https://www.toi.no/getfil...
This report does state cycling in Australia is 14% less safe after helmet legislation. However, this is not accompanied by any citation, computational results or analytic derivation. Perhaps you can shed some light on where this figure originated. It seems to just appear without any justification.
The remainder of your comments does not appear related to our current paper and are instead an attack on helmets generally.
You mention a 2007 report you authored. However, I could find no such record using either Google Scholar or Scopus. Where is this report published? Was it peer-reviewed?
You also mention your 2012 article in the New Zealand Medical Journal. In this paper, you assess the helmet law in New Zealand using, mostly, data found in a paper by Tin Tin and colleagues (2010). The original peer-reviewed paper can be found here.
http://www.biomedcentral....
You state “data from 2012 shows for New Zealand that a higher accident rate occurred”; however, you fail to note that all injuries per million hours cycled declined when comparing data from years 1988-91 with 1996-99. The New Zealand helmet law became effective between those dates on January 1, 1994 and helmet wearing increased accordingly. Although helmets are designed to protect the head, you did not consider head injuries alone and instead focused on all cycling injuries. Additionally, Tin Tin et al. (2010) found a statistically significant decline in serious traumatic brain injury (TBI) per million hours cycled during the same period when helmet wearing sharply rose. The rate of serious TBI remained steady as helmet wearing also remained steady. This is not mentioned in your paper. Although the data in your paper are estimates of injury rates, you do not include any measure of uncertainty (such as a confidence interval) nor do you perform any statistical tests. Would you care to comment why you do not provide any recognized statistical analysis in your paper or why you omit the serious TBI results from the Tin Tin et al. (2010) paper?
You state the articles we cite demonstrating a benefit to helmet legislation “do not relate to the accident rate per km of travel.” This is but one way to represent the number of injuries as a rate relative to a measure of exposure such as distance or time travelled by bicycle or per population. No data on time/distance travelled exists for New South Wales. That is precisely why we compared bicycle-related hospitalizations with a diagnosed head injury with diagnosed limb injury. The amount of cycling, whether it is distance or time, is captured in that comparison. That is unless a cyclist can remove either his/her limb or head while cycling, e.g., 10k cycled with your head is 10k cycled with your arms. This kind of comparison is called a dependent non-equivalent no-treatment control and is the most recommended type of comparison in the statistics literature for improved causal inference. This is because it is unethical to carry out a randomized controlled trial to determine the efficacy of helmets in a bicycle crash. It is unethical because we know that helmets protect the cyclist’s head in a collision.
You state “survey data for New south Wales show a 44% drop in children cycling from 1991 to 1993 following” MHL. This data comes from a series of four reports commissioned by the NSW Roads and Traffic Authority (RTA). They can be found online here.
http://www.bicycleinfo.ns...
Over a four year period (1990-1993), cyclists were observed to verify compliance with the helmet law. Surveys were taken in one month within each year. Therefore, out of a possible 48 months, only 4 months were surveyed, i.e., an 8.3% response rate. Further, as noted, these surveys were designed to estimate helmet wearing and not cycling exposure. For these reasons, caution should be exercised when interpreting this data.
However, if you insist on relying on the RTA helmet wearing reports as estimates of cycling exposure, then note the results are not unequivocal. There was a 22% increase in adult cycling counts in Sydney after their helmet law on January 1, 1991, and a 7% increase for NSW adults overall. The original survey counted adult cyclists at road intersections only and recreation areas were added to the following year’s survey. There was a decline in the counts of adult cyclists at road intersections comparing April 1991 to April 1993 (5478 to 4027); however, there was a larger increase in the counts of adult cyclists in recreation areas (1095 to 2638). There is a 1.4% increase in adult cycling comparing those years combining counts at road intersections and recreation areas (6573 to 6665).
Other researchers have noted the mixed results using this data. Nolén and Lindqvist (2003), cited within the same paragraph of the Erke and Elvik technical report you quote, state “Several studies, however, indicates that a cycle helmet law may result in a reduction in cycling by young people and in a certain, but temporary, reduction in cycling by younger children. On the other hand, cycling by adults is probably not influenced.” If helmet legislation is a deterrent to cycling, it is age dependent with permanent effects only detectable in young people. Other surveys have found no such decline for any age group. Further, young people eventually become adults and no research has studied whether these potential effects have been permanent. The large increase in Australians cycling, adults or otherwise, over the past decade or more suggests the effects have not been permanent.
If there was such a decline in cycling exposure due to the helmet law it would be noticeable in the hospitalization record. If the 44% decline in cycling (or the 40% you quote later) did, in fact, actually happen due to the helmet law, shouldn’t there be a large decline in all cycling injuries not just those to the head? Our 2011 analysis published in Accident Analysis and Prevention found no significant declines in arm or leg injury hospitalizations comparing the eighteen months pre- and post-helmet law (p=0.28 and p=0.62 respectively). However, head injury hospitalizations significantly declined by about 29% when helmet wearing increased from about 25% to 80%.
You directly quote a paper by Curnow (2008) – “Compulsion to wear a bicycle helmet is detrimental to public health in Australia but, to maintain the status quo, authorities have obfuscated evidence that shows this”. What exactly is your purpose here? Are you accusing us of unethical behavior? Are you implying helmet legislation is a government conspiracy? If so, that is a serious accusation which has no basis.
RE: RE: Considering the claims again
Colin479 replied to jakeolivier on 18 Oct 2013 at 09:05 GMT
I gather that the report may contain important errors and these may explain some of the points I raise. My reply is based on the publish report.
Reply
Thank you Jake for your detailed reply. It seems to me that this topic requires more fundamental research rather than just a re-analysis of an original work.
Walker in Figure 2 provides information on the ratio of particularly close to particularly distant overtaking events as a function of the bicyclist’s riding position and helmet-wearing or not.
Walker reported
“These ratios are displayed in Fig. 2, showing that as riding
position shifts from the road edge towards the road centre, the
likelihood of experiencing near overtaking events rather than far
overtaking events increases, in line with the findings of the above
ANOVA. Also in line with the ANOVA, near overtaking events
were more likely to be experienced than far overtaking events
when a helmet was being worn, and this effect became more
pronounced as riding position moved away from the road edge,
reaching a peak for the 1.25m condition in which there were
4.33 times as many near events as far events when a helmet was
worn. This helmet effect was not seen in the 1m riding position,
in which overtaking was no more likely to be near than far when
a helmet was worn.”
Your title you say comes from Tables 4, 5 and 8, e.g. Table 4. Univariate and multiple logistic regression analysis for close passing (less than 1 m).
It appears that your approach is based on assuming that a 1 m passing clearance at 64km/hr is safe. Cyclists quite often catch another rider’s back wheel when they are riding in groups and this result in falls. A safe passing distance may be “Motorists to provide a minimum passing distance of one metre between the motor vehicle and a bicyclist for speeds up to 50km/hr. At higher speeds more than 1 metre should be provided and sufficient to prevent contact with the person operating the bicycle if the person were to fall into the driver’s lane of traffic.” This could be dependent on several factors and could be up to 2 m.
The data in Table 8 details for up to 1.5m passing clearance, helmeted 584 readings from 1149, 50.8% of events and for non-helmeted, 518 from 1206, 42.9% of events. For passing clearance up to 1.0m, helmeted 60 of 1149, 5.2% and for non-helmeted 49 of 1206, 4.1%. It appears that in both results a higher proportion of helmeted events had vehicles passing closer. I take your point about being similar at various intervals – 0.66m vs. 0.61, 0.90m vs. 0.90, 1.29m vs. 1.30 and 1.70m vs. 1.72 when wearing and not wearing a helmet respectively. Figure 1 of your report shows a wider range for passing distances for helmeted compared to no helmet. Walker reported being hit twice when wearing a helmet. Small differences in average passing distance I think could be expected but if a wider range occurs this may be important. Possibly your report could have considered this in more detail.
Figure 1 also shows approximate values of;
at 0.25m from road edge, one case of helmeted pass at 0.4m,
at 0.50m from road edge, one case of no helmet pass at 0.5m,
at 0.75m from road edge, one case of helmeted at 0.49m, no helmet at 0.52m.
Table 5, a note reports *There were only two passing events less than 0.5 m. In each case, Walker did not wear a helmet.
The information presented in Figure 1 and note in Table 5 do not seem to tie up.
The other reports you mention I will take a look at and comment later so as not to detract from considering the primary reports.
Colin Clarke
RE: RE: RE: Considering the claims again
jakeolivier replied to Colin479 on 20 Oct 2013 at 01:56 GMT
Colin Clarke,
When you state “this topic requires more fundamental research rather than just a re-analysis of an original work”, I get the sense you do not understand research. We used readily available data, kindly provided by Ian Walker on his website, which we re-analyzed using commonly used statistical methods – linear model and logistic regression analysis. Note that ANOVA, as used by Ian Walker in his original paper, is a type of linear model as is linear regression. Our paper is clearly a re-analysis of the data analyzed in another paper. We did not develop new methods of data collection nor did we develop new methods for analyzing data. Our paper is clearly not basic science research.
You state “Walker in Figure 2 provides information on the ratio of particularly close to particularly distant overtaking events as a function of the bicyclist’s riding position and helmet-wearing or not.” As we note in our paper, Walker categorized close and far overtaking events using the quartiles estimated from the data. We also cite articles that discuss problems when using data-driven cut points as used by Walker. Note our final model included kerb distance, so the estimated effect of helmet wearing has accounted for bicyclist’s riding position.
Instead, we categorized according to the most commonly used definition of ‘safe’ overtaking as evidenced by the many citations we give. Certainly safe passing distance is a function of vehicle size and speed, i.e., more distance is needed for larger vehicles going at greater speed. But, as we note below, vehicle speed was not measured in Walker’s data set.
You state “It appears that your approach is based on assuming that a 1 m passing clearance at 64km/hr is safe.” This is true for the analyses found in Tables 4 & 5, but not Table 8. It is clear from table 8 that overtaking distances when wearing a helmet or not differ only for passing distances greater than 2 meters. The average difference between helmet wearing or not does not exceed 2cm for passing distances under 2m.
Could you give a citation for this quote? “Motorists to provide a minimum passing distance of one metre between the motor vehicle and a bicyclist for speeds up to 50km/hr…” It appears in you comment without any justification. In addition to the citation pointing to one meter being a safe passing distance up to 64 kmh, The Amy Gillett Foundation recommends one meter for speeds up to 60 kmh.
http://www.amygillett.org...
With regards to Walker’s data, neither the recommended speed limit nor the actual speed of the vehicle while overtaking is included.
You state regarding safe passing, “This could be dependent on several factors and could be up to 2 m”. Our analysis demonstrates helmet wearing does not play a significant role in overtaking distances less than 2m. So, it would appear the title of our paper is justified by your quote.
You mention Ian Walker was struck by a vehicle twice, both times when wearing a helmet. These were unfortunate occurrences and my sympathy goes out to Ian Walker and any cyclist who has ever been hit by a motor vehicle. I would not wish that on anyone. However, you fail to mention both instances involved a large vehicle – a bus and a heavy goods vehicle. Our overall analysis points to vehicle size as a significant factor in close overtaking. But note that no statistical analysis is possible on just two observations. In fact, the estimated 95% confidence interval for the probability of being struck by a motor vehicle when wearing a helmet using this data is [0,1] (actually, it is [0,1.5] but probabilities cannot exceed 1).
Regarding the data contained in Table 8, you state “Small differences in average passing distance I think could be expected but if a wider range occurs this may be important. Possibly your report could have considered this in more detail.” The p-values given in that table actually consider the variability in passing distance when wearing a helmet or not. Yet, it is possible to give more information here. Below are the 95% confidence intervals for the differences in overtaking distance for each interval.
Interval | Difference | 95% CI
0-0.75 | -0.052 | -0.224, 0.121
0.75-1 | 0.003 | -0.061, 0.067
1-1.5 | 0.007 | -0.012, 0.027
1.5-2 | 0.017 | -0.003, 0.037
2-inf | 0.072 | 0.034, 0.109
With the exception of the first interval, when helmet wearing results in GREATER passing distance, the confidence intervals are quite narrow. This implies a good level of certainty in these estimates.
For the remainder of your comments, you seem to have overlooked my comment from two days ago.
http://www.plosone.org/an...
My comment states “It has been brought to my attention Figures 1 and 2 have been incorrectly labeled. The categories "Helmet" and "No Helmet" should be swapped. The journal has been notified and I have requested this be reflected in the proof. I have also supplied the journal with R script that performs the analyses and creates the figures using Walker's raw data.”
This confusion originated from a disparity between Table 1 in Walker’s original paper and Walker’s data posted to his website. We actually note this in our paper where we state “We obtained the same frequency of events for each category, but they were swapped for helmet versus no helmet.” Note that if Walker’s Table 1 were correct, Walker’s original analysis would estimate a benefit of helmet wearing on passing distance.
The R script file can also be found here.
https://www.researchgate....
RE: RE: RE: RE: Considering the claims again
Colin479 replied to jakeolivier on 21 Oct 2013 at 19:31 GMT
Thank you Jake for all the extra details provided and I will consider.
With the initial report having the helmet and no helmet data mislabeled in Fig 1 and 2 plus Table 8 suggesting a higher proportion of helmet wearers being passed closer, these gave one impression but with the corrections mentioned it lends more to the conclusions you have drawn.
The paper was based on 2355 passing events, the risk of accident from motor vehicles overtaking cyclist has been calculated at 1.5 per 10 million passes. The data base of 2335 appears to be small. (2355/6.66 million = 0.035% of an accident risk). This is why I suggested more research. Your paper considered there may be too much data when in fact there may be too little.
See Monash univ report of 1988, Drummond and Lee, Fig 23 for risk details. Risk data from Melbourne and passing data from UK, large potential for errors.
Jake you ask
“Could you give a citation for this quote? “Motorists to provide a minimum passing distance of one metre between the motor vehicle and a bicyclist for speeds up to 50km/hr…”
Hard to find but the following may help
------------------
The driver of a motor vehicle must maintain a lateral distance of at least 1.5 meters, to prevent accidents between their vehicle and cycles transiting in the same lane. - See more at:
http://www.ecf.com/news/v...
http://www.ecf.com/news/v...
http://www.pedalpower.org...
In France the rule is that motorists need to give cyclists a minimum of 1m in towns where the speed is lower
http://cyclistsinternatio...
“But even a 3-foot passing law is considered inadequate by cyclists and urban planning professionals who often ride bicycles on the road.”
“The standard 3-foot bicycle passing law found in most states is an inadequate minimum distance to pass a bicyclist at any speed,” said Andy Besold, an urban planner who studied at the Rutger’s Bloustein School and is now a planner in Boise, Idaho.
https://www.gov.uk/using-... see photo of what is considered safe passing.
Previous comments a reply;
Erke and Elvik 42 stated: "There is evidence of increased accident risk per cycling-km for cyclists wearing a helmet. In Australia and New Zealand, the increase is estimated to be around 14 per cent."
Jake says
“This report does state cycling in Australia is 14% less safe after helmet legislation. However, this is not accompanied by any citation, computational results or analytic derivation. Perhaps you can shed some light on where this figure originated. It seems to just appear without any justification.”
https://www.toi.no/getfil... page 28
My understanding is that the authors are respected researchers, e.g
https://www.toi.no/staff/...
The Robinson 1996 report shows data for children in NSW. The equivalent number of pre-law injuries increased from 1310 (384 head + 926 other injuries) in 1991 to 2083 (488 head + 1595 other injuries) in 1993. Children’s injury rate increased by 59%, from 1310 to 2083.
I think the 14% figure came from assessing data from a range of reports but I cannot provide precise details.
Jake says
“The remainder of your comments does not appear related to our current paper and are instead an attack on helmets generally.”
Your report included “Given the potential increase in harm that may arise from not
wearing a helmet,” It was therefore appropriate to point to information questioning the claimed benefits of helmets.
Jake asks
“You mention a 2007 report you authored. However, I could find no such record using either Google Scholar or Scopus. Where is this report published? Was it peer-reviewed?”
http://www.nationaler-rad...
Thursday: Best Practice
Poster Presentation: Mankind and Health
Colin Clarke (United Kingdom) - Cyclists' Touring Club: The Case against Bicycle Helmets and Legislation (Paper, 250 kb)
http://www.nationaler-rad...
Jake says
You state “data from 2012 shows for New Zealand that a higher accident rate occurred”; however, you fail to note that all injuries per million hours cycled declined when comparing data from years 1988-91 with 1996-99. The New Zealand helmet law became effective between those dates on January 1, 1994 and helmet wearing increased accordingly. Although helmets are designed to protect the head, you did not consider head injuries alone and instead focused on all cycling injuries. Additionally, Tin Tin et al. (2010) found a statistically significant decline in serious traumatic brain injury (TBI) per million hours cycled during the same period when helmet wearing sharply rose. The rate of serious TBI remained steady as helmet wearing also remained steady. This is not mentioned in your paper. Although the data in your paper are estimates of injury rates, you do not include any measure of uncertainty (such as a confidence interval) nor do you perform any statistical tests. Would you care to comment why you do not provide any recognized statistical analysis in your paper or why you omit the serious TBI results from the Tin Tin et al. (2010) paper?”
http://journal.nzma.org.n...
http://www.biomedcentral....
The two reports provide a range of information. Head injury data is provided and a section Head injuries provided. The 2012 was limited to 3000 words for publication criteria purposes. Adding statistical tests may have complicated the paper for some readers and may not have been reliable considering the limited data. Collins et al reported accident data for 1988, ‘Fifty-one percent of those hospitalised were aged 5–14, and males accounted for 70% of all admissions. Part of the problem becomes knowing the reduction in head injury due to less cycling, age related group and changes in road safety risk.
Table 4 details the changes, (change relative to 1988–91), the data for 96/99, 21.38 (-17). The 17% reduction is clearly shown in bold, as is the reduction for all other groups of other road users. The other 4 groups had on average a reduction of 18.7%. Highlighting the reduction for cyclists may have given a misleading impression that cyclists gained more than other road user groups.
Jake says
“Although helmets are designed to protect the head, you did not consider head injuries alone and instead focused on all cycling injuries.”
Tin Tin stated there was a substantial increase in injury risk from crashes not involving a motor vehicle. Figure 3 indicates head injuries reduced from about 12 to 6, other injuries increased from about 15 to more than 30. Much of the reduction in head injuries came from crashes not involving motor vehicles, these are usually less severe and often child related. The proportion of children cycling reduced appreciably. Making strong claims for helmets could have been misleading.
Jake says
“You state the articles we cite demonstrating a benefit to helmet legislation “do not relate to the accident rate per km of travel.” This is but one way to represent the number of injuries as a rate relative to a measure of exposure such as distance or time travelled by bicycle or per population. No data on time/distance travelled exists for New South Wales. That is precisely why we compared bicycle-related hospitalizations with a diagnosed head injury with diagnosed limb injury. The amount of cycling, whether it is distance or time, is captured in that comparison. That is unless a cyclist can remove either his/her limb or head while cycling, e.g., 10k cycled with your head is 10k cycled with your arms. This kind of comparison is called a dependent non-equivalent no-treatment control and is the most recommended type of comparison in the statistics literature for improved causal inference. This is because it is unethical to carry out a randomized controlled trial to determine the efficacy of helmets in a bicycle crash. It is unethical because we know that helmets protect the cyclist’s head in a collision.“
As detailed above
“The Robinson 1996 report shows data for children in NSW. The equivalent number of pre-law injuries increased from 1310 (384 head + 926 other injuries) in 1991 to 2083 (488 head + 1595 other injuries) in 1993. Children’s injury rate increased by 59%, from 1310 to 2083.”
The ratio head to arm may change do to several factors in addition to helmets.
Jake claims - “No data on time/distance travelled exists for New South Wales”
Day–to-Day Travel in Australia, CR 69, INSTAT, FORS 1988 http://www.infrastructure...
Census data
Travel to school information
4 years of surveys and other surveys
These combined provide indications of travel by bicycle.
Jake says;
“However, if you insist on relying on the RTA helmet wearing reports as estimates of cycling exposure, then note the results are not unequivocal. There was a 22% increase in adult cycling counts in Sydney after their helmet law on January 1, 1991, and a 7% increase for NSW adults overall. The original survey counted adult cyclists at road intersections only and recreation areas were added to the following year’s survey. There was a decline in the counts of adult cyclists at road intersections comparing April 1991 to April 1993 (5478 to 4027); however, there was a larger increase in the counts of adult cyclists in recreation areas (1095 to 2638). There is a 1.4% increase in adult cycling comparing those years combining counts at road intersections and recreation areas (6573 to 6665).”
The 1990 survey was conducted in spring, with poor Sydney weather conditions. Walker reported: “As it turns out, the first survey was conducted in overcast conditions in Sydney and, in some areas, was interrupted by rain whereas the second survey was conducted in sunny conditions.
The rural counts, 1990 -1993 of adults were, 2650, 2402, 1933, 1660, reduction 37%.
In April 1991 the law already applied to adults, so you are comparing periods 91to 93, with legislation in force. The data you provide for adult recreational, 1095 and 2638, the 1095 is described as ‘Miscellaneous’ and the criteria for including adults changed from 1991 to 1993. For children the figures are 1899 in 1991 to 1205 in 1993, down 37%. Road sites for children 1741 counted in 1991 to 902 in 1993, down 48%.
The statement “1.4% increase in adult cycling comparing those years” is probably incorrect, cycling reduced by 37% for rural adult road sites.
Jake refers to
“On the other hand, cycling by adults is probably not influenced”
Adult count in Melbourne down 29%, NSW adults rural road sites 37% etc.
Jake says
“The large increase in Australians cycling, adults or otherwise, over the past decade or more suggests the effects have not been permanent.”
Jake says
“No data on time/distance travelled exists for New South Wales.”
Jake says
“However, head injury hospitalizations significantly declined by about 29% when helmet wearing increased from about 25% to 80%” – it provides no evidence of a benefit because of the above data showing reduced cycling levels.
Jake says
“You directly quote a paper by Curnow (2008) – “Compulsion to wear a bicycle helmet is detrimental to public health in Australia but, to maintain the status quo, authorities have obfuscated evidence that shows this”. What exactly is your purpose here? Are you accusing us of unethical behavior? Are you implying helmet legislation is a government conspiracy? If so, that is a serious accusation which has no basis.”
I think Bill Curnow has a valid point. The Governments of Australia had a responsibility to provide full health, safety and other details of the helmet laws, they failed to provide full reports. Prior to the law in Victoria they had provided annual reports on cycling activity for city and rural areas. Vic Roads failed to publish similar reports in the 1990’s. The internal report IR90-15 was not published. Survey details from NSW after 1993 were not published, e.g. children counted in 1991-6788 and by 1993 3798, down 44% and a very serious outcome but the Government did not publish follow on surveys. Four in 10 children had probably been directly affected with health implications of reduced exercise and many adults especially in rural areas. The governments had a duty of care to monitor the effects of the helmet law and to fully disclose. They failed to do so.
RE: RE: RE: RE: RE: Considering the claims again
jakeolivier replied to Colin479 on 28 Oct 2013 at 00:14 GMT
Colin Clarke,
You are still confused about statistical inference on two events and, in particular, sample size. First, if you want to include the two times Ian Walker was hit by a vehicle, there are actually n=2357 overtaking events. Our comments about overly large sample sizes in the paper are in regards to measuring passing distance which is on a quantitative scale (specifically, in meters). This differs drastically from trying to get information from the specific events when Walker was hit by a vehicle (the sample size here is n=2). My comment was about estimating the probability of being hit while wearing a helmet using a confidence interval for the proportion of such events. This was [0,1] and therefore uninformative. The sample size, and therefore the inference, does not change when you multiply (or divide) by a constant to get a rate. If you would like to draw conclusions about why motor vehicles hit cyclists using Walker’s data, the sample size is indeed small (n=2) and helmet wearing is confounded with the size of the vehicle, as I mentioned.
There is plenty of data to assess overtaking distance. However, this data was collected by one cyclist exclusively in two UK cities. It is clear more research needs to done in this area and this should’ve been clear back in 2007 when Walker’s paper was published (as opposed to after we pointed to the lack of a helmet effect upon re-analysis).
You state “Risk data from Melbourne and passing data from UK, large potential for errors” and mention the Drummond and Lee MUARC report (found here).
http://www.monash.edu.au/...
As best I can tell, this has nothing to do with passing distance and just a measure of the riskiness in such maneuvers. So, you still haven’t answered my question. What evidence do you have that when overtaking distance is between [1.5m, 2m) that extra passing distance of 1.7cm (the size of my thumbnail) makes a difference in risk? Or, when passing distance is between [2m, inf) an extra 7.2cm (less than the length of my index finger) make a difference in risk? On the other hand, can you explain how an extra 5.2cm does not make a difference when passing is between [0,0.75) as is the case when wearing a helmet?
It is clear the recommended overtaking distance of one meter is not adequate for all situations and I would agree with 1.5m as a recommendation for higher speeds and larger vehicles. The link I provided last time to a report by The Amy Gillett Foundation even has this to say “will include greater separation for higher speed zones (1.5m in speed zones over 60km/h)” even though their general recommendation is for a one meter rule. However, this still doesn’t justify why a thumbnail difference (1.7cm) in overtaking distances greater than 1.5m is important for road safety? What are the additional lateral forces there?
You have not answered my question about the Erde and Elvik quote regarding a 14% increase in cycling risk in Australia and New Zealand. By stating they are “respected researchers” does not justify the quote or whether their report was peer-reviewed (I think it’s quite clear it is not). Based on what I can gather from comments you have made on the internet, you don’t seem to respect any researcher who demonstrates a benefit to helmets or helmet laws with regards to believing their research is true. Why is that?
You have tried justifying the Erde and Elvik quote by stating “The Robinson 1996 report shows data for children in NSW. The equivalent number of pre-law injuries increased from 1310 (384 head + 926 other injuries) in 1991 to 2083 (488 head + 1595 other injuries) in 1993. Children’s injury rate increased by 59%, from 1310 to 2083.”
In this same paper, Robinson states, regarding the Walker NSW helmet wearing studies, “Comparable figures were not available for adults.” Can you explain this statement? Each Walker report and the Smith and Milthorpe report have data for adult helmet wearing and, therefore, adult cycling count data. Why was this data not included in any of her papers? The adult cycling counts increased by 22% in Sydney from the Oct 90 to Apr 91 reports. She also ignores child cycling counts found in the Apr 90 report which, when combined with the Apr 91 report, shows an existing downward trend. Can you explain why Robinson did not factor the adult cycling counts or the existing downward trend in her calculated “Equivalent no of injuries for pre law numbers of cyclists”? You are listed in the Acknowledgements of Robinson’s 1996 paper, so I assume you would know. You seem to believe weather is the ultimate determining factor here. Why? Believing the Oct 90 survey is not comparable to the Apr 91 survey is not the same as stating this data does not exist.
You seem to have ignored my comment earlier regarding the Walker reports stating “Therefore, out of a possible 48 months, only 4 months were surveyed, i.e., an 8.3% response rate.” You seem to put full faith in these reports with lots of missing data that has been repurposed to draw conclusions about cycling rates. However, you are ready to ignore the complete census of all NSW hospitalization data around the same time. Why?
So, am I safe to assume your 2007 ‘report’ was not peer-reviewed?
You justify your general attack on helmets by stating “It was therefore appropriate to point to information questioning the claimed benefits of helmets” because we note in our paper “the potential increase in harm that may arise from not wearing a helmet.” There has now been a Cochrane review (Thompsons and Rivara), a meta-analysis (Attewell et al) and “several” re-analyses of the meta-analysis (Elvik) each demonstrating a benefit to helmets protecting the head in a crash. Each of these studies is in line with the biomechanical evidence. Further, Curnow’s many hypotheses regarding the negatives of helmet wearing are not supported by available evidence, specifically his diffuse axonal injury hypothesis. We therefore made that statement based on the existing peer-reviewed literature.
Regarding your own 2012 paper, you state “Adding statistical tests may have complicated the paper for some readers and may not have been reliable considering the limited data.” That is not true and you would never be allowed to publish in the vast majority of scientific publications without an appropriate statistical analytic plan. That is precisely why many top-tier medical journals have statistical reviewers and/or a statistical editorial board. You now state the data was limited, yet you contribute the NZ helmet law on 53 premature deaths per year. That is a very strong conclusion which should only be made after a rigorous statistical analysis using relevant data.
You have also not answered my question regarding not including serious traumatic brain injury or why you did not include any analyses only involving head injuries. Helmets are designed to protect the head and failure to assess the impact on that type of injury is a failure of your paper. It is acceptable to compare head injuries to other injuries (and may even help with drawing conclusions), but ignoring head injuries completely is a head scratcher.
Here’s another question regarding your 2012 paper. In Table 4, serious cycling injuries (abbreviated injury scale >=3) per million hours cycled decreased 53% (1996-99) and 39% (2003-07) compared to 1988-91. This would appear to be an important result considering the NZ helmet law became effective 1 January 1994, yet you never mention this in your paper. Why? Was that not a benefit attributable to the helmet law?
I noted the lack of cycling time/distance travel data for NSW and you point to a 1988 report. Since the helmet law for adults (1 January) and children (1 July) occurred in 1991, how does this data help? Since you claim helmets deter cycling, how does this report provide any evidence to support your conclusion?
If somehow weather has played a part in influencing the counts of cyclists for the NSW helmet use surveys, doesn’t this point to weather as a major deterrent to cycling? You seem to attribute the 22% increase in Sydney cycling counts to weather. Doesn’t that point to weather, not helmet laws, as the main cycling deterrent? You seem to dismiss the increase in recreational cycling counts. Why? The 141% increase in adult counts in recreational areas is a massive increase at a time when helmet legislation is supposedly driving people away. This also includes rural areas where counts were 545 in Oct 1992 and 1293 in Oct 1993 (a 237% increase).
Linda Ward has provided more conflicting evidence regarding support for this hypothesis. When this evidence is taken as a whole, it doesn’t lend strong support for the ‘helmet law as a cycling deterrent’ hypothesis.
I still don’t understand the Curnow reference and quote. Our paper ultimately has nothing to do with helmet legislation or even Australian cycling. We performed a re-analysis on readily available data using fairly standard statistical methods (actually, not much different than Ian Walker’s original paper). We have even provided the R script for creating each figure and table using the original data. What exactly do you think we are hiding here?
I also don’t understand what you think the government is hiding here. The data you and others with your viewpoint come from government reports. They are also readily available online. By the way, a colleague of mine asked the NSW government to make all their cycling reports available to the public. We originally had a very difficult time finding electronic copies, although print versions could be viewed in person at the NSW Roads and Traffic Authority library. There was no pushback from the government to make their reports publicly available online (as noted, they were readily available to the public at their library).
http://www.bicycleinfo.ns...
You close by stating “Four in 10 children had probably been directly affected with health implications of reduced exercise and many adults especially in rural areas. The governments had a duty of care to monitor the effects of the helmet law and to fully disclose. They failed to do so.” Where is your evidence here? You are drawing very strong conclusions from very weak data.
RE: RE: RE: RE: RE: RE: Considering the claims again
Colin479 replied to jakeolivier on 30 Oct 2013 at 19:44 GMT
Jake,
The Highway Code advice may be improved by including the following;
Take extra care when passing cyclists, preferably leave 1.5m or more but at least a minimum passing distance of one metre between the motor vehicle and a bicyclist for speeds up to 30 mph. At higher speeds more space should be provided and sufficient to prevent contact with the person operating the bicycle if the person were to fall into the driver’s lane of traffic. Pass with care and not at excessive speed.
About the Erde and Elvik quote regarding a 14% increase in cycling risk in Australia and New Zealand. Seven reports were used in evaluating the 14 % figure.
Jake asks about
"Robinson states, regarding the Walker NSW helmet wearing studies, “Comparable figures were not available for adults.” Can you explain this statement? "
The comparisons were for April, at the same time of year, in 1991 the law already applied to adults but not to children. Recreational survey information was comparable for children but not for adults because they changed the counting criteria for adults. Children could be compared at the same time of year from before their law to after, adults could not.
Jake says
"So, am I safe to assume your 2007 ‘report’ was not peer-reviewed?"
Obviously the EFC review material submitted for including in their conferences and they have people well versed in the helmet topic having produced their own reports. On balance I would say it was ‘conference reviewed’ rather than peer reviewed. Much of the material in the 2007 paper was also presented in other papers that were reviewed. Some parts will have been reviewed 3 times.
Jake asks
"What evidence do you have that when overtaking distance is between [1.5m, 2m) that extra passing distance of 1.7cm (the size of my thumbnail) makes a difference in risk?"
I pointed you to data for the risk of accident for cars passing in the Monash report, HGV or buses have a higher risk. Walker’s Fig. 3. Mean overtaking proximities as a function of overtaking vehicle type. Cars about 1.34m, HGV /Bus about 1.10m, about 24 cm closer on average but the risk level increases, perhaps by about 5/30 fold, see http://www.ncbi.nlm.nih.g...
http://www.ncbi.nlm.nih.g...
Jake asks about
"the NZ helmet law on 53 premature deaths per year", the details and basis are provided in the paper.
Jake mentions
"You close by stating “Four in 10 children had probably been directly affected with health implications of reduced exercise and many adults especially in rural areas. The governments had a duty of care to monitor the effects of the helmet law and to fully disclose. They failed to do so.” Where is your evidence here?"
None of the state governments in Australia or any other country that has imposed helmet laws have provided health and safety assessments with reliable data.
The CTC information is worth viewing, http://www.ctc.org.uk/cam...
It seems clear that issues still remain about helmets but efforts to improve safety regardless of helmets should be the priority.
RE: RE: RE: RE: RE: RE: RE: Considering the claims again
jakeolivier replied to Colin479 on 04 Nov 2013 at 05:32 GMT
Colin Clarke,
You mention advice from The Highway Code (UK?, citation?) regarding ‘safe’ passing distance. You only seem to be repeating what I have mentioned previously, that this is somewhat a function of the size and speed of the vehicle. The Khan and Love articles (citations 17 and 18) are the only research articles I could find that attempt to justify any distance, one meter or otherwise. I would generally agree a conservative approach, e.g., lower overtaking speed, is good for cycling safety. However, this still doesn’t address how an extra thumbnail distance (1.7cm) for passing maneuvers over 1.5m is important. To put it another way, how much lateral variability does an average cyclist exhibit when cycling in a ‘straight’ line? My guess is the range is much larger than 1.7cm.
Regarding the Erke and Elvik quote, you state “Seven reports were used in evaluating the 14 % figure”. Your response still does not answer my question regarding how they arrived at a 14% increase. As fundamental as this quote is to your arguments, including your 2012 paper as well, shouldn’t you know how it was derived? You seem very quick to question any research demonstrating a non-negative effect of helmet wearing, yet you don’t seem to question anti-helmet arguments. Why?
Regarding Robinson’s omission of NSW cycling count data, you say “The comparisons were for April, at the same time of year, in 1991 the law already applied to adults but not to children. Recreational survey information was comparable for children but not for adults because they changed the counting criteria for adults. Children could be compared at the same time of year from before their law to after, adults could not.” The counting criteria did not change from year to year, instead more locations were included (namely recreation sites). The city of Albury was included in some but not all surveys. None of this changes the fact that comparable figures at road intersections do exist for NSW adults and Robinson (1996) did not include them in her paper. As I’ve noted before, the adult counts in Sydney increased 22% and 6% for NSW overall, and counts at recreation sites increased 141% from the 2nd to 4th surveys. None of that supports the ‘helmet as cycling deterrent’ hypothesis. Again, Robinson makes no argument for not including this data in her analysis and instead claims such data does not exist.
The authors of the fourth study, Smith and Milthorpe, wrote about their findings in the proceedings of a conference in 1994. In the abstract, they state “The 1993 survey of over 10,000 observations found no drop in adult ridership following legislation. There were fewer children riding, but numbers varied with the area – suggesting multiple reasons for the decrease.” Later on, the authors state “The unevenness in the change in ridership – up at some sites, down in others – makes it difficult to draw conclusions about trends.”
Smith, N.C., Milthorpe, F.W. (1994). Bicycle helmet wearing in New South Wales after legislative mandate. Proceedings of Pedestrian and Bicyclist Safety and Travel Workshop.
The European Cyclist’s Federation (EFC) is a cycling advocacy group and not a research organization. So, as I suspected, your paper was not peer reviewed by research experts in public health, epidemiology, road safety, statistics or any other academic discipline related to this topic. If it were my article, it would not count as research output for my university (and probably most if not all universities worldwide). My department head would probably question why I didn’t submit to a peer-reviewed journal instead.
You mention HGV/Buses pass on average 24cm closer than cars with an increase in risk (1.10m vs 1.34m). I don’t understand what this has to do with a 1.7cm difference in overtaking for distances already greater than 1.5m. I don’t think I can emphasize enough how small a distance that is.
Regarding my comments of your 2012 paper, you state “the details and basis are provided in the paper.” Since you do not discuss your decision to not include head injury data to assess the impact of helmet legislation in your paper, I am assuming you are refusing to answer the question. On that subject, I also found it strange that you don’t discuss the changes in helmet wearing in New Zealand around that time. Much of this data is readily available online or from previous research papers, including Robinson (2001).
You state “None of the state governments in Australia or any other country that has imposed helmet laws have provided health and safety assessments with reliable data” and you include a link to the Cyclist’s Touring Club, a cycling advocacy group to which you belong (Is this a conflict of interest?). Our 2011 paper found a 29% decline in bicycle related head injury hospitalizations relative to limb injuries in NSW following helmet legislation. Limb injuries did not change significantly at that time. Our paper from this year demonstrated this decline has been maintained over the following two decades. I’m confused here. Do you not consider a census (as in all) of NSW hospitalizations to not be ‘reliable’ data? I am not a government employee of Australia, but our work is an assessment of the impact the NSW helmet law had on cycling injuries. Is it because our conclusions do not match your strongly held beliefs?
You close by stating “It seems clear that issues still remain about helmets but efforts to improve safety regardless of helmets should be the priority.” It may be clear to you, a cycling advocate, that there are issues around helmets. My experience, as an academic statistician and road safety researcher, is the actual evidence against helmets relies heavily on the omission of relevant data, analyses that lack statistical rigor or the misrepresentation of data. It is a sideshow that unfortunately dominates any discussion about cycling safety. I do believe crash avoidance, like segregated cycleways, is an important part of any road safety policy. What is your recommended safety strategy when a crash does occur? Since you clearly don’t like helmets, what can a cyclist do to protect themselves from injury?
RE: RE: RE: RE: RE: RE: RE: RE: Considering the claims again
Colin479 replied to jakeolivier on 21 Nov 2013 at 12:30 GMT
Jake asked for a reply, details are provided and may require checking, the tables may not line up but hopefully they can be used.
Highway Code UK - below include suggestions for changes - so they not match the existing Highway Code.
Reply the current Highway Code advice is at
Overtaking (162 to 169)
https://www.gov.uk/using-... ... 162-to-169
163
Overtake only when it is safe and legal to do so. You should
(a list is provided in the Code)
• give motorcyclists at least as much room as you would when overtaking a car. Cyclists and horse riders require passing at lower speeds with plenty of room.
The caption on the picture requires re-phasing
Suggest
‘pass vulnerable road users with extra care and generous clearance '
213
Take extra care passing cyclists, preferably leave 1.5m or more but at least a minimum passing distance of one metre between the motor vehicle and a bicyclist for speeds up to 30 mph. At higher speeds more room should be provided and sufficient to prevent contact with the person operating the bicycle if the person were to fall into the driver’s lane of traffic. Pass with care and not at excessive speed.
The current Highway Code advises ‘as much room as you would when overtaking a car’ but this is almost meaningless and little to no legal standing for enforcement.
Erke and Elvik quote, you state “Seven reports were used in evaluating the 14 % figure”.
The 7 reports covering from 1988 to 2000 were used to evaluate the 14% figure. Helmet wearing rates before to after and injury rates. As far as I know the authors have not published details explaining their work in full. It would be better for the authors to explain their findings rather than me provide my view.
NSW comment you say
“The counting criteria did not change from year to year, instead more locations were included (namely recreation sites).” It appears that you are mistaken,
http://www.bicycleinfo.ns... page 15 explains the changes.
I think your error is processed into your recent report. I think your adult road count view is mistaken and in error regarding Robinson.
The 1993 survey of over 10,000 observations found no drop in adult ridership following legislation.
http://trid.trb.org/view....
Smith NC, Milnthorpe FW. An observational survey of law compliance and helmet wearing by bicyclists in NSW. Roads and Traffic Authority of NSW. 1993
http://www.bicycleinfo.ns...
Using data from the 1993 report for road sites, Table
1990 1991 1992 1993
Sydney 2730 3332 2796 2591
Rural 2650 2402 1933* 1660
Totals 5380 5734 4729 4251
• The 1992 survey did not detail the adult count for Albury (1990 – 262, 1991 - 256, 1993 – 224)
• See Table 12, 91 report, Table 9, 92 report, Table 3.1 93 report.
I would have to consider the claims made misleading based on the road survey data.
Considering census data, with estimates
1986 (30 June) 1.09% 0.69% 2.42%
1990 estimate** 1.2% 0.76% 2.66%
1991 (6 August) 0.96% (0.81*) 0.67% (0.88*) 1.85% (0.69*)
1996 (6 August) 0.87% (0.73*) 0.58% (0.76*) 1.58% (0.59*)
** 1990 estimate is based on accidents increased by approximately 40% between 1986 to 1990 for the 17+ age group 10% extra has been added for cycling to work, taking the 1.09 value to 1.2.
Jake
“My department head would probably question why I didn’t submit to a peer-reviewed journal instead.”
The paper was 10000 words and probably too long for most journals.
World Transport Policy & Practice Volume 12, No. 2, 2006 The case against bicycle helmets and legislation http://www.eco-logica.co.... did review and publish a shorter version.
Jake
“Regarding my comments of your 2012 paper” – refer http://theconversation.co... for more details.
Jake says
“You state “None of the state governments in Australia or any other country that has imposed helmet laws have provided health and safety assessments with reliable data” and you include a link to the Cyclist’s Touring Club, a cycling advocacy group to which you belong (Is this a conflict of interest?). Our 2011 paper found a 29% decline in bicycle related head injury hospitalizations relative to limb injuries in NSW following helmet legislation. Limb injuries did not change significantly at that time. Our paper from this year demonstrated this decline has been maintained over the following two decades. I’m confused here. Do you not consider a census (as in all) of NSW hospitalizations to not be ‘reliable’ data? I am not a government employee of Australia, but our work is an assessment of the impact the NSW helmet law had on cycling injuries. Is it because our conclusions do not match your strongly held beliefs?”
No conflict of interest, Your 2011 paper compares injuries – see result from Tin Tin fig 3, non-TBI increased by 129% - your paper becomes unreliable due to its methods. A new paper for NSW is required. e.g
Fatality data comparing pedestrians and cyclists
1886-90 91-95 96-2000 01-05 06-10
Cyclists –C 114 58 56 64 54
Pedestrians -P 924 616 564 457 312
C/P% 12.3% 9.4% 9.9% 14.0% 17.3%
Jake asks
Since you clearly don’t like helmets, what can a cyclist do to protect themselves from injury?
The short answer is ride with care, avoid icy roads, heavy traffic and high speeds, avoid the inside space for HGVs and buses, check the bike, wear light coloured clothing etc, avoid night time riding if possible. Ride steady down hills. Helmets will increase the accident risk so avoid them. After age 60, watch out for less control, slower reactions, poorer vision that come with age and ride accordingly. Without going into reams of details the above approach is probably a summary that comes to mind.
RE: RE: RE: RE: RE: RE: RE: RE: RE: Considering the claims again
jakeolivier replied to Colin479 on 25 Nov 2013 at 00:02 GMT
Colin Clarke,
When I asked for your reply, I actually meant for you to reply to my questions. I will try to make myself clearer this time. You just seem to be reiterating points that are not on topic. None of your responses are relevant to our paper.
The one metre rule is recommended in the UK, and in Australia, for slower speeds (whether 30mph or 50km/h in UK and Australia respectively). At those speeds, there is no significant association with helmet wearing a close passing. For greater speeds, the 1.5m recommendation is given and there is only a 1.7cm difference for helmet wearing for passing distances in the interval [1.5m, 2m). The differences are not statistically significant for passing less than 1.5m, although there is 5.2cm of MORE space when wearing a helmet at passing distances less than 0.75m. So, how is a thumbnail difference (1.7cm) important for safety when cars pass at distances greater than 1.5m? And, if you want to make that argument, how is an extra 5.2cm when passing at distances in the interval [0m, 0.75m) not important?
To reiterate my previous point, you seem to be satisfied with the Erke and Elvik quote of 14% without any justification given anywhere in their work. Yet, you are quick to criticize any study demonstrating any research demonstrating helmets are beneficial. You are clearly an anti-helmet advocate and, based on your seemingly poor knowledge of research fundamentals, are clearly not a researcher.
It is you who is mistaken with regards to our peer-reviewed paper “Statistical Errors in Anti-helmet Arguments,” you state “I think your adult road count view is mistaken and in error regarding Robinson.” This paper can be found here.
http://acrs.org.au/events...
These figures were taken directly from Tables 3.1 (p. 16) and 3.10 (p. 28) of Smith and Milthorpe. We did not include the Albury data in our tables because data was not consistently collected at road intersections or recreation areas. However, the adult counts at Albury road intersections were 262 and 256 in Oct 90 and Apr 91 respectively which is also inconsistent with the ‘helmet law deters cycling’ hypothesis.
I don’t understand how this could ever be in “error regarding Robinson” since she never used the adult counts and instead has never recognized this data exists in any of her publications. Could you explain? An argument data is not comparable is not the same as denying data exists.
You further introduce the Ride to Work data which is a census taken on one day every 5 years in Australia. You forget to mention that the 1986 cycling mode share before any helmet legislation was 1.14% in Australian capital cities, where most Aussies live, and was 1.13% in 1991 when most Australians of all ages were subjected to such laws. You quote 1.09% and 0.96% in 1986 and 1991 respectively (Is this for Sydney? NSW?). How is a difference of 0.13% important? Note that taking ratios of such small percentages over-accentuates any ‘real’ change that may have occurred. As you’ve demonstrated many times, you have very little understanding of statistics or even the fundamentals of research.
This difference is also incredibly small considering the historical estimates of cycling mode share in Australia. The peak of around 8-9% occurred around World War II and slowly declined thereafter. The cycling mode share change profoundly 45 years before any helmet legislation in Australia (see Figure 2.1, p. 27). How does helmet legislation have anything to do with this decline?
http://www.infrastructure...
I have no idea where this ‘estimate’ comes from, “1990 estimate is based on accidents increased by approximately 40% between 1986 to 1990 for the 17+ age group 10% extra has been added for cycling to work, taking the 1.09 value to 1.2.” Could you explain? Where does this data come from? What do columns 3 and 4 represent? Are they 95% confidence intervals?
Since this is an ‘estimate’ (and I’m giving you the benefit of the doubt here), it will inherently have some associated amount of uncertainty attached. With that in mind, how is a change from 1.09% to 1.2% to 0.96% important? Further, note that each mode share estimate is contained in the 95% confidence interval of the previous survey. So, mode share estimates from one survey to the next are contained in the range of ‘typical’ values indicating no trend. This is further evidence unsupportive of the ‘helmet law deters cycling’ hypothesis.
You still have never addressed my questions regarding your 2012 NZMJ paper, here or anywhere else. Instead you point to a commentary I wrote that is highly critical of your paper and the comments contained therein; you also did not address my criticisms there either. Some of them are
1. How can you draw such strong conclusions when you've performed no statistical analyses?
2. Does your study meet any of the Bradford-Hill criteria for causal inference?
3. How do you assess helmet legislation properly (an intervention directed at lowering head injury) by not analyzing head injury data?
4. Why didn't you include estimates of helmet wearing in your study?
Your response seems to be you either (1) had no space or (2) statistics would just confuse the issue. The flaws in your paper are FATAL which your excuses do not actually address. You also rely on data that is not near the helmet law date. This is a clear limitation to your study, yet you do not discuss your study’s limitations (and there are plenty).
You state “No conflict of interest.” I completely disagree. You are clearly an anti-helmet advocate with no interest presenting a balanced view of helmets. This is demonstrated by your criticism of any research supportive of helmets and you seem to blindly accept research critical of helmets or helmet laws. I definitely do not think helmets are a panacea as they are only designed to protect the head and do not help with crash avoidance (the evidence also suggests they do not increase crash risk). There are also crashes where the forces are so great or directed to other body parts that a helmet will not be helpful. I also believe more research is needed in this area (and less advocacy) so that we better understand the mechanics by which cycling safety interventions (includes helmets, but also cycling infrastructure) can mitigate cycling injury.
You state "Your 2011 paper compares injuries – see result from Tin Tin fig 3, non-TBI increased by 129%” and conclude “your paper becomes unreliable due to its methods. A new paper for NSW is required.” What exactly is your point here? The Tin-Tin paper shows serious TBI decreased by a large margin before and after helmet legislation in NZ and this decline was maintained. The increase in other cycling injuries is consistent with increases in recreational cycling (e.g., mountain biking, road) which is known to have occurred over the past 20 years or so. In a later paper, we also found cycling head and arm injuries diverged after the NSW law which is consistent with the NZ experience. So, why are head injuries not increasing while other cycling injuries are? Perhaps an intervention directed at mitigating head but not other cycling injuries?
What is the point of the table you’ve presented? The NZ helmet law was Jan 1, 1994 and you’ve aggregated data by 1986-90, 1991-95, etc. You’ve not split your intervals up appropriately to assess changes relative to the NZ helmet law. However, note that cycling fatalities dropped 49% (114 to 58) and pedestrian fatalities dropped 33% (924 to 616) comparing 1968-90 with 1991-95. What could be causing cycling fatalities to decline at a faster rate than pedestrian fatalities? Perhaps an intervention directed at one and not the other?
As discussed in this paper and in the one linked below, there is no strong evidence helmet wearing increases the likelihood of a crash. Clearly, crash avoidance is important; however, so is the mitigation of injury when a crash occurs. So, again, you clearly do not like helmets, but how can a cyclist protect themselves from injury in a crash? Crashes do occur and your response is completely silent on that issue.
http://acrs.org.au/wp-con...
RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: Considering the claims again
Colin479 replied to jakeolivier on 25 Nov 2013 at 22:15 GMT
The one metre rule is recommended in the UK, ........the interval [0m, 0.75m) not important?
From the above info it does not seem significant. However I did notice drivers seem to leave much more space at far distances, this would be quite welcome to many cyclists. Cycling is more than safety issues of course.
To reiterate my previous point, you seem to be satisfied with the Erke and Elvik quote of 14%,................. are clearly not a researcher
I am anti-helmet legislation and pro choice.
The 1996 paper, page 464, http://www.cycle-helmets.... 10 lines down RHS. Robinson does mention adult cycling.
The basic figures we agree on (what a change) but you have it incorrect from what I can see.
Road Intersections (without Albany)
Sydney .2730 3332 2796 2591
Rural ….2388 2146 1933 1436
Subtotal 5118 5478 4729 4027
The 1990 survey was conducted in spring, with poor Sydney weather conditions. Walker reported: “As it turns out, the first survey was conducted in overcast conditions in Sydney and, in some areas, was interrupted by rain whereas the second survey was conducted in sunny conditions”. Adult rural road cycling reduced by 40%, according to a simple estimate (1436/2388 = 0.60). (37% if including Albany perhaps)
From 1991 to 1993, a drop of 22% occurred in Sydney. In 1991, the helmet law already applied to adults, effectively disregarding the low Sydney count for 1990 due to poor weather. Rural comparison 1991 to 1993 – down 33%.
From 1991 to 1993 both followed a similar pattern
Pop changes etc
Location 1986 1991 1996
NSW ……….. 5.532 5.899 6.205
Sydney……... 3.466 3.673 3.881
Rest of State.. 2.066 2.226 2.324
Inner Sydney pop
Location…………………………… 2001 pop. 2011 pop. % increase
Homebush Bay – Silverwater….. 4600 12000 159%
Waterloo – Beaconsfield……….. 11200 22700 103%
Sydney - Haymarket - The Rocks 14500 25000 73%
Above combined areas………….. 30300 59700 97%
Recreational adults etc.
They changes the counting criteria, 1990 no details, 1991 16-19 year for adults, over 20 not counted but some were, 1992 all adults. Plus it seems the number of sites may have been changed plus they estimated when it rained for one site, boosting numbers.
Year**(date)………… NSW ……………. Sydney…………. Rest of State
1986 (30 June)-……. 1.09% …………… 0.69%.................. 2.42%...........
1990 estimate……… 1.2%.................... 0.76%.................. 2.66%..........
1991 (6 August)…… 0.96% (0.81*)…… 0.67% (0.88*)…… 1.85% (0.69*)
1996 (6 August)…… 0.87% (0.73*)…… 0.58% (0.76*)…… 1.58% (0.59*)
Figures like (0.81*) would be the percentage reduction from the estimate for 1990.
"times, you have very little understanding of statistics or even the fundamentals of research."
Obviously you don’t know me. But I have corrected quite a few mistakes in research papers over the years, even in the last few weeks examples come to mind. Your papers has some and Tin Tin had one at least. I have corrected DfT papers on the helmet topic. These are facts.
"I have no idea where this ‘estimate’ comes from, “1990 estimate is based on accidents increased by approximately 40% between 1986 to 1990 for the 17+ age group 10% extra has been added for cycling to work, taking the 1.09 value to 1.2.” Could you explain? Where does this data come from? What do columns 3 and 4 represent? Are they 95% confidence intervals?"
The accident data comes from the NSW official stats, eg 0-16 age group had 733 reported, 10 killed I think in 1990, by 1999 – 340 injured, even less now. The 40% is from 1986 – 1990. I think you can probably understand it now with a touch of luck. Cycling was increasing in Au in the late 1980s due to several factors and the number and proportion cycling to work by 1990 would have been higher than in 1986. I estimated a 10% gain but adult accident stats went up by 40% so my estimate may have been low. 10% gain 1.09 to 1.2 etc."
"Some of them are 1-4"
Sorry see other place for answers perhaps. I may simply not work in your style and consider issues from other directions. You may be asking why I don't work like you.
I am pro choice, but this entails answering questions about helmets, so I provide evidence that I can support, pro helmet people see this as anti helmets but the opposition is to helmet laws due to the social damage plus they increase the accident rate, even if you dispute this.
"You state "Your 2011 paper compares injuries – see result from Tin Tin fig 3, non-TBI increased by 129%” and conclude “your after the NSW law which is consistent not increasing while other cycling injuries are? Perhaps an intervention directed at mitigating head but not other cycling injuries?
You simply do not get it, injuries per million hours increased.
The ratio cyclist/pedestrian deaths in blocks of 5 years for NZ, first 2 figures mainly pre law, 12.3 and 9.4, other 3, post law, 9.9, 14, 17.3.
C/P% 12.3% 9.4% 9.9% 14.0% 17.3%
They almost suggest that helmets cost lives.
Jake
You simply do not understand the research and have arrived at a pro helmet law position. If Sydney Uni pays my air ticket (or if they go half) I will come and explain it to you and others in some detail.
RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: Considering the claims again
jakeolivier replied to Colin479 on 03 Dec 2013 at 05:04 GMT
Colin Clarke,
None of your latest comments had anything to do with our current paper. You seem to have drifted off into other areas which perhaps should be discussed elsewhere.
You state “I am anti-helmet legislation and pro choice” and later state I “have arrived at a pro helmet law position”. However, our paper is completely silent on this issue as we state “The evidence from this study does not justify recommendations around helmet wearing, but rather highlights the more important factors of kerb distance, road characteristics and traffic type which may inform more effective cycling safety improvements.”
I am neither anti- or pro-helmet/helmet legislation, but I am pro-science. The available evidence indicates they are beneficial at mitigating head injury in a crash. The hypothesized negative impact of helmets or helmet legislation (deterrent effects, rotational injuries, risk compensation, safety in numbers, etc) is not back backed by available evidence which we discuss in our ACRS paper. Even the Elvik (2011) paper, which is often paraded around as ‘proof’ helmets don’t work, estimated a substantial benefit of helmets mitigating head injury. Note the results of the Elvik paper have changed drastically over the past two years due to errors as we note in our paper. Can you cite solid evidence helmets are not beneficial? I would re-evaluate my position given new evidence, as would any scientist.
http://acrs.org.au/wp-con...
I stated that Robinson (1996) does not mention adult cycling in NSW in her papers and you counter that “Robinson does mention adult cycling” and give a specific location in her paper. The direct quote is
“Comparable figures were not available for adults.”
How exactly is this Robinson mentioning the adult counts? She also doesn’t mention the 1990 counts for children. The best way to discuss this issue is to present all available data and discuss any limitations that may exist and what impact they have on the analytic results. Robinson, on the other hand, never admits this data exists.
My colleagues and I have discussed the problems with using the RTA helmet use surveys (includes Walker, 1990, 1991, 1992; Smith & Milthorpe, 1993). We state
http://www.sciencedirect....
“The use of limb injuries as a comparison group to head injuries, which we employed as a means to overcome the lack of cyclist exposure data, also takes into account changes in cyclist numbers.
A 30–40% reduction in cycling participation in NSW after the commencement of MHL is an often misquoted or selectively quoted figure (Robinson, 1996; Curnow, 2008; Rissel and Wen, 2011) from the surveys by Smith and Milthorpe (1993). In fact, the survey results show a drop in school-aged riders, while adult cycling participation remained largely unchanged in the Sydney metropolitan region where the majority of the population lives, with a modest reduction in rural regions. However, the authors of the survey series explicitly warn that their study was not designed to assess participation, and thus should not be relied on as definitive evidence of changes in participation one way or the other. On the other hand, hospitalisation records as used in our analysis are a census of injured cyclists admitted to hospital and are thus not susceptible to weaknesses common to survey data. In any case, changes in the cycling exposure denominator that occurred during the analysis period obviously applied equally to both head and arm injury rate numerators.”
In our peer-reviewed ACRS conference paper, we state
“Caution should be taken when interpreting statistical results using this survey data whether supportive or opposed to helmet legislation. Smith and Milthorpe (1993) note the surveys were designed to estimate helmet wearing in NSW and not to estimate cycling exposure. Also, over a forty-eight month period, data was only collected over four months (akin to an 8.3% response rate). However, the use of these surveys for that purpose only supports Robinson’s conclusions when the adult data are ignored since those counts clearly increased from pre- to post-MHL. Additionally, Robinson’s conclusions are tempered with the inclusion of cycling counts for children in 1990.”
I clearly do not believe this data is adequate in estimating cycling exposure or changes in cycling. But, if you insist on their use, you need take this data in full and not cherry-pick results. As I’ve pointed out before, the authors of the 1993 report, Smith and Milthorpe, state “The 1993 survey of over 10,000 observations found no drop in adult ridership following legislation. There were fewer children riding, but numbers varied with the area – suggesting multiple reasons for the decrease.” Later on, the authors state “The unevenness in the change in ridership – up at some sites, down in others – makes it difficult to draw conclusions about trends.”
Smith, N.C., Milthorpe, F.W. (1994). Bicycle helmet wearing in New South Wales after legislative mandate. Proceedings of Pedestrian and Bicyclist Safety and Travel Workshop.
You again state the counting criteria had changed, which is not true. More sites were added, including recreation areas, but the method of counting had not. If the surveys changed their counting method from year to year, the data would not be comparable at all and it would nullify any research based on this data (e.g., Robinson, 1996). If you continue to argue the 1990 data was affected by weather, as in less cycling than to be expected, then that also affects the child cycling counts. This would mean the apparent decline in child cycling from 1991 to 1992 would be part of a longer term trend in less cycling that began long before the helmet law.
You again cite the ride to work data. Your 1990 ‘estimate’ is not part of this data and comparability with the other years is, at best, questionable. A point about this data I made last time is that future surveys are within the margin of error of previous surveys and, therefore, do not suggest a trend.
You state “Obviously you don’t know me. But I have corrected quite a few mistakes in research papers over the years, even in the last few weeks examples come to mind.” This would not get you a job as a researcher. It might only get a job as a fact checker at an anti-helmet advocacy group. You have no research record and, from what I can discern, no research training.
You state “Your papers has some and Tin Tin had one at least. I have corrected DfT papers on the helmet topic. These are facts.” What errors have you found? There was a mistake with a few of our figures, but I stated there was an error on PLOS ONE before you ever commented. I even pointed that out to you. Do you not read my responses to you?
Regarding your 1990 estimate, you state “I think you can probably understand it now with a touch of luck.” You’ve made no attempt to justify your estimate with regards to comparability with the other estimates. Estimates are essentially guesses and have an inherent amount of uncertainty. This can usually be represented by estimates of standard error and/or a confidence interval. You’ve supplied neither. I reiterate, your estimates and the ride to work estimates do not differ by much and, I suspect, are all within the normal range if there is no trend. That is why a proper statistical analysis is important and you can’t get away with drawing strong conclusions from weak, or in your case non-existent, analyses.
You state “Cycling was increasing in Au in the late 1980s due to several factors and the number and proportion cycling to work by 1990 would have been higher than in 1986.” Where is your citation? As I pointed out last time, cycling in Australia peaked at 8-9% mode share during World War II then declined steadily thereafter.
http://www.infrastructure...
I included these four criticisms of your work.
1. How can you draw such strong conclusions when you've performed no statistical analyses?
2. Does your study meet any of the Bradford-Hill criteria for causal inference?
3. How do you assess helmet legislation properly (an intervention directed at lowering head injury) by not analyzing head injury data?
4. Why didn't you include estimates of helmet wearing in your study?
Your response is that you don’t work like me. What is that supposed to mean? These are simple questions to answer if you’ve done the research properly. Since you’ve never given valid responses here or elsewhere, I will continue to assume the conclusions of your paper are invalid.
You again cite NZ fatality data yet ignore the comparison most relevant to helmet legislation. As I stated previously, “cycling fatalities dropped 49% (114 to 58) and pedestrian fatalities dropped 33% (924 to 616) comparing 1986-90 with 1991-95. What could be causing cycling fatalities to decline at a faster rate than pedestrian fatalities? Perhaps an intervention directed at one and not the other?” Since the helmet law was 1 January 1994, how is the data from 1996 onwards more relevant than these two time periods? But, if you want to compare the pre-law period (1986-90) with the next post-law period (1996-2000), cycling fatalities declined by 50% and pedestrian fatalities by 38%.
The simple fact that you don’t seem to grasp is that data nearest the helmet law date is more relevant than data farther away in time.
You conclude with “You simply do not understand the research and have arrived at a pro helmet law position. If Sydney Uni pays my air ticket (or if they go half) I will come and explain it to you and others in some detail.” I’ve demonstrated over and over that you don’t understand the fundamentals of research. This also includes a lack of understanding of statistics which is essential to rigorous research.
Note that I am a researcher at The University of New South Wales, as clearly indicated here.
http://www.plosone.org/ar...
Given my comments, perhaps you could explain why a research organization would pay for your traveling expenses.
RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: Considering the claims again
Colin479 replied to jakeolivier on 06 Dec 2013 at 10:44 GMT
Jake, a short reply to some of the important points. you say
"You again state the counting criteria had changed, which is not true. More sites were added, including recreation areas, but the method of counting had not."
Page 15, 1992 http://www.bicycleinfo.ns...
clearly the criteria changed for adults in recreational surveys but your report miss this important point.
Page 3, 1992, Summary of aims, No7 to investigate any decrease in the number of cyclists, again your reports miss this important main objective.
By 1993 with different authors they changed the outlook and provided more weight to covering up the reduction in cycling. Equally the population increase and their locations were not reflected.
Your reports reflect not only on the authors but also on the University of NSW.
If we can discuss these two issues and agree, then some progress may be possible.
RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: RE: Considering the claims again
Colin479 replied to Colin479 on 06 Dec 2013 at 10:50 GMT
Page 14 of the 1992, see Table 9, Note 1. 'Number of adults at recreational sites not comparable, under different instructions etc.
RE: Considering the claims again
Linda_Ward replied to Colin479 on 27 Oct 2013 at 12:04 GMT
Colin479 (Clarke) cites Curnow (2008): 'Compulsion to wear a bicycle helmet is detrimental to public health in Australia but, to maintain the status quo, authorities have obfuscated evidence that shows this'.
A 2009 Cochrane review (by Thompson) contains this description of Curnow's contribution to the helmet debate:
'His commentary contains factual errors and misinterpretations of the data. In contrast to Curnow’s claims, the Thompson 1996 study found that all types of bicycle helmets (hard shell, soft shell and foam) provided substantial protection against head, brain and severe brain injuries for bicyclists involved in motor vehicle crashes and crashes due to other causes (Thompson 1996: Tables 3 and 4). In Curnow’s Table 1 (Curnow 2005) he compares brain-injured cases to head injured cases without brain injury. He interprets the 1.06 odds ratio from this exercise as showing that helmets don’t protect against brain injury. The correct interpretation is that the protective effect of helmets is similar for both head and brain injury. Cummings 2006 explains that many of Curnow’s criticisms stem from misconceptions about the studies that have been done and about case-control studies in general. . . Hagel 2006 rebuts Curnow’s arguments and points out the advantages that well conducted case-control studies have over ecologic study designs. In reply, Curnow 2006 continues the discussion and repeats arguments which have been addressed both in this review and the comments which follow at the end of the review.'
With respect to Curnow's 2005 article, Hagel (2006) commented that "Ironically, and most damingly, Curnow’s paper would not comply with the rigorous criteria required of a Cochrane systematic review because he fails to present all relevant evidence for the effect of bike helmet use and legislation in a balanced way".
An aspect of Curnow's 2005 article that was not mentioned by Hagel was Curnow's use of Australian Road Transport Safety Bureau (ATSB) fatality data for 1994 and 1998 to support his claim that 'Despite a decrease in cycling, deaths to cyclists, even those by head injury, declined by less than other road users. No benefit from the helmet laws is evident.' ATSB data shows that during 1992-94 (helmet laws in all Australian jurisdictions), compared to 1987-89 (no helmet laws), cyclists fatalities fell by considerably more (45%) than pedestrian fatalities (28%).
Curnow's 2006 article reports data showing reductions in participation after the helmet legislation, and fails to report data showing that
- there was no decrease in (overall) cycling in South Australia (p<0.05, Marshall)
- adult cycling in New South Wales (NSW) increased (NSW RTA)
- cycling to work in the Australian Capital Territory increased (Australian Bureau of Statistics census data)
- a drop in cycling to work in Perth (Western Australia) co-incided with a transport 'megaproject' (Muhammad) that resulted in the 'revitalisation' (Mees) of Perth's rail system, and bus travel to work dropped by almost twice as much as cycling to work
(Curnow's 'theory' that helmets increase DAI has been discredited by the Walter and McIntosh studies.)
(Colin479) Clarke has also overlooked much evidence that is contrary to the helmets/laws-are-bad argument . . .
In his 2007 article, Clarke cites data from the NSW RTA reports as evidence of a drop in cycling, but fails to cite data (from the same series of reports) showing that the number of adult cyclist counted increased after the helmet law.
Clarke's 2007 article fails to note that Marshall found that there was no reduction in (overall) cycling (18 months after the helmet law, compared to 18 months before the law, p<0.05). The Marshall study also found that
- prior to the law, cycling to school comprised 20% of cycling in that age group
- after the law, a (40%) drop in cycling to school was accompanied by an increase of equivalent size in cycling to/around other venues
Colin479 claims that 'survey data for NSW show a 44% drop in children cycling', with no mention of the evidence to the contrary (eg. the participation data from the Marshall study, and cyclist non-head injury data from a 1995 study by Williams). He also fails to note that the NSW surveys showed that the number of child cyclists counted at rural NSW road intersections dropped by about 20% months before the child helmet law came into effect (the surveys were in Sep/Oct 1990 and April 1991, the child helmet law came into effect in July 1991, the injury data from the Carr study suggests that cycling levels are about 25% higher in April than Sep/Oct).
Clarke's 2007 article attributes a decrease in cycling to work to the helmet laws, but fails to report that the census data also shows that that other modes of travel to work decreased by at least as much as cycling
- between 1986 and 1996, cycling to work dropped by about the same amount as bus and train (and less than ferry/tram) travel to work
- at the 1991 census, jurisdictions with helmets laws comprised about 70% of the Australian population, between 1986 and 1991 cycle travel to work dropped by 7%, bus travel to work dropped by 14%.
- between 1991 and 1996, when the remaining 30% of the population was also subject to helmet laws, cycle travel to work dropped by 38%, walking to work fell by 36%.
His 2007 article cites New Zealand (NZ) Land and Transport Safety Authority data as showing that cycling in NZ declined between 1989 and 1997. Clarke fails to report that Scuffham et al. (1997) noted that there was a gradual (20%) decrease in the cyclist numbers 2 years BEFORE the helmet law.
The 2007 article also fails to note that a later (2000) study by Scuffham et al. reported that 'Results indicated that there was a positive effect of helmet wearing upon head injury and this effect was relatively consistent across age groups and head injury (diagnosis) types. We conclude that the helmet law has been an effective road safety intervention that has lead to a 19% (90% CI: 14, 23%) reduction in head injury to cyclists over its first 3 years.'
Clarke's 2007 article cites injury data from 1992 and 1994 Monash reports by Cameron et al. as showing no benefit, but fails to cite a 1995 Monash report by Carr, Cameron et al.. The injury data in the 1995 Monash (Carr) report shows that (after correcting the hospital admission data to remove the effects of changes in funding arrangements for public hospitals in Victoria), cyclist non-head injuries fell by about the same amount (20%) as pedestrian head injuries (indicating that there was no long-term reduction in amount of cycling, consistent with the Marshal results, and the NSW adult cyclist counts); and that serious/severe (AIS3/4) cyclist head injuries dropped by 60%.
Clarke (2007) also fails to note a West Australian study by Hendrie et al., which found there was a significant reduction in cyclist head injuries attributable to the helmet law.
RE: RE: Considering the claims again
Colin479 replied to Linda_Ward on 30 Oct 2013 at 21:45 GMT
Linda Ward refers to
’Compulsion to wear a bicycle helmet is detrimental to public health in Australia but, to maintain the status quo, authorities have obfuscated evidence that shows this’ and refers to a 2009 Cochrane review (by Thompson), plus articles in 2005 and 2006 by Curnow.
http://www.cyclehelmets.o... Table 1 provides data. The information shows that children were discouraged from cycling. Governments consider the health benefits of cycling outweigh the risks but no government has provided a health and safety assessment. They failed to address the health aspect of discouraging cycling.Thomson, Cummings or Hagel do not provide health and safety assessments.
Curnow’s 2005 paper referred to 1988 and 1994 and explains the information came from a ATSB report where data is available for alternative years. Ward provides a different comparison using 87-80 compared to 92-94 period. Clarke provides another comparison based on 4 year periods, 86-89 compared to 93-96 see
http://www.cycle-helmets....
Clarke report “Cycling was discouraged by approximately 30%+ due to the helmet law. Allowing for the reduction in cycling gives only a 13% reduction for cyclists, the smallest reduction of all road users.”
Linda says[
“In his 2007 article, Clarke cites data from the NSW RTA reports as evidence of a drop in cycling, but fails to cite data (from the same series of reports) showing that the number of adult cyclist counted increased after the helmet law.”
Adult cyclists counted at rural road sites reduced from 2650 in 1990 to 1660 in 1993, down by 37%. Adults counted in the Sydney area were 2730 in 1990 (The 1990 survey was conducted in spring, with poor Sydney weather conditions. Walker reported: “As it turns out, the first survey was conducted in overcast conditions in Sydney and, in some areas, was interrupted by rain whereas the second survey was conducted in sunny conditions”). In 1991 the count was 3332, 1992 – 2796, 1993 – 2591, reduction of 22% from 1991 to 1993.
In total adult road sites counted 5380 in 1990 and 4251 in 1993. Adult recreational counts changed their criteria for counting and therefore they did not provide reliable comparisons.
Various claims based on Marshall are made, these need to be investigated to see if reliable.
Linds says
“His 2007 article cites New Zealand (NZ) Land and Transport Safety Authority data as showing that cycling in NZ declined between 1989 and 1997. Clarke fails to report that Scuffham et al. (1997) noted that there was a gradual (20%) decrease in the cyclist numbers 2 years BEFORE the helmet law.”
http://www.ta.org.br/site...
The 2007 paper states “In New Zealand cycling declined by 34% from 1989 to 1997, 12% with helmet promotion prior to legislation and 22% after legislation. Head injuries and other injuries reduced with the reduced cycling. Scuffham and Langley reported on pre law, "results revealed that the increased helmet wearing percentages has had little association with serious head injuries to cyclists as a percentage of all serious injuries to cyclists".
Possibly different sources of data have resulted in variations.
A number of other points have been raised with regards to the 2007 report, these have little bearing on vehicles passing cyclists. It seems that some points are not valid, eg Marshall was reporting on South Australia and has no real bearing on children’s cycling levels in NSW.
References
Walker M.B. Law compliance and helmet use among cyclists in New South Wales, Dec 1990
http://www.bicycleinfo.ns...
Walker M.B. Law compliance and helmet use among cyclists in New South Wales. April 1991
http://www.bicycleinfo.ns...
Walker M.B. Law compliance among cyclists in New South Wales, April 1992
http://www.bicycleinfo.ns...
Smith NC, Milnthorpe FW. An observational survey of law compliance and helmet wearing by bicyclists in NSW. Roads and Traffic Authority of NSW. 1993
http://www.bicycleinfo.ns...
Colin Clarke (Colin479)
RE: RE: RE: Considering the claims again
Linda_Ward replied to Colin479 on 03 Nov 2013 at 09:41 GMT
You claim that the data in Curnow's table 1 'shows that cycling were discouraged from cycling'. In fact, what the data in table 1 shows is that children were discouraged from cycling TO SCHOOL. The SA data in from the Marshall study, Curnow neglects to report the results from the same study which showed that prior to the helmet law, cycling to school comprised only 20% of cycling exposure in that age group (tables 6a and 6b, on p19). He also fails to report that the same study found that (18 months after the law, compared to 18 months before the law), there was no decrease in overall child cycling, because the decrease in cycling to school was accompanied by an increase of equivalent size in cycling to/around other venues (p19). A 40% drop in cycling to school equates to an 8% drop in overall child cycling, which is miniscule compared to the commonly (mis-) quoted 40% drop. Curnow also fails to cite the results of the same study showing that (18 months after the law, compared to 18 months before the law) there was no reduction in the proportion of adults who cycled at least once a week (p<0.05, p2-3).
In response to Jake Olivier's question as to why you did not cite the TBI results from the Tin Tin et al. (2010) paper, you replied that 'part of the problem becomes knowing whether the reduction in head injury is due to less cycling'. The 2nd sentence of the Tin Tin abstract states that exposure-based rates were used, the 4th sentence of the abstract states that 'time spent cycling was used as the measure of exposure. Your (2012) table 1 is reproduced from Tin Tin table 4, with the addition of % change in the number of injuries per million hours spent travelling. The TBI rates in the same Tin Tin paper are clearly labeled as 'Number of injuries per million hours spent cycling'. Yet you state that there is a problem knowing whether the reduction in 'head injury' RATES were due to less cycling.
You state that Tin Tin figure 3 "indicates head injuries reduced from about 12 to 6", however there is no 'head injury' category in figure 3, and no TBI results showing a drop from 12 to 6. You also state that "Much of the reduction in head injuries came from crashes not involving motor vehicles, these are usually less severe", however figure 3 shows that for serious (AIS>2) injuries, the TBI rate dropped by more than 60% (from just over 3 to about 1).
Your statement that 'the proportion of children cycling reduced appreciably' is presumably based the decrease between 1989/90 and 2005/08 reported in the Tin Tin study. Scuffham et al. (1997) noted that there was a (gradual) 20% decrease in the number of cyclists counted in the 2 years BEFORE the helmet law, Povey et al. (1995) noted that there was no evidence of any reduction in cycling as a result of the helmet law in NZ. Your suggestion that a drop in child cycling could have been responsible for the drop in injury rates is 100% incorrect, the rates in figure 3 are 'per million hours spent cycling'.
Carr (1995), Povey (1999), Walter (2011) and Jake Olivier (above) have noted that declines in cycling would be reflected by declines in cyclist non-head injuries.
If the Tin Tin figure 3 rates were, for example, per year, a drop in cycling would also be expected to cause decreases in the number of non-head injuries, not increases as shown.)
You claim that 'The ratio head to arm may change do to several factors in addition to helmets', but do not provide any supporting evidence. The Povey and Walter studies, both published in peer-reviewed journals, were based on the assumptions of a (reasonably) constant head to non-head injury ratio. Robinson's 2001 article noted that the Povey study was based on this assumption, and made no suggestion that the assumption was not valid.
Moving on to your claims/s of a 40% reduction in child cycling in NSW . . .
In the NSW data from the Walter study, the correlation coefficient for cyclist head vs non-head injuries was 0.68 before the law, and 0.63 after the law. So, as noted by Jake Olivier, a 40% reduction in (child) cycling would be noticeable in the non-head injuries.
As has been brought to your attention multiple times (on Wikipedia and TheConversation), the injury data in the (1995) Williams study shows that
- between 1990 and 1993 serious casualties for all NSW road users dropped by 15%
- between 1998/90 and 1991/92 NSW child cyclist non-head injury hospital admissions dropped by 13%
This indicates, consistent with the Marshall results, that any reduction in child cycling would have been small, and short-lived.
You state that Robinson's 1996 article shows that the children's injury rate increased (by 59%) between 1991 and 1993. This increase was (presumably) calculated by 'applying' a 40% reduction in children's cycling. The data in Robinson's table 1 indicates that NSW child cyclist non-head injuries dropped by 15% between 1990 and 1993, which is the same as serious road casualties for all NSW road users. So Robinson's non-head injury data indicates (consistent with the Marshall results for SA) that between 1990 and 1993, NSW (overall) child cycling reduced by 0%. The table 1 data also indicates that in 1992/93, compared to 1989/90, NSW child cyclist non-head injury hospital admissions reduced by 13%, and head injuries reduced by 37%. This indicates that there was a 37%-13%=24% reduction in NSW child cyclist head injuries attributable to the helmet law, which is 'amazingly' close to the Walter (NSW) results.
Hagel (2006) noted that the correlation coefficients corresponding to one of Robinson's graphs show that there is a strong (negative) correlation between helmet use and head injuries.
The results of the Walter study (and the results obtained above, from correctly analyzing Robinson's NSW data) are consistent with the head injury results in the Thomson/Cochrane, Attewell, and Elvik meta-analyses. The AIS3/4 injury reductions reported in the Carr study are consistent with the brain injury results in the Thompson/Cochrane, Attewell, and Elvik meta-analyses. Given the consistency between the case-control and (correctly analysed) longitudinal results, any increase in risk as claimed by proponents of the 'safety in number' and 'risk compensation' theories would have been, at most, very small.
There is only one part of the 'equation' that 'does not fit' - a drop in cycling of anywhere near the size claimed by those opposed to helmets/laws.
You claim that a number of the issues identified with respect to your 2007 article have 'little bearing on vehicles passing cyclists'. Firstly, it was you who stated that 'In contrast to claiming a benefit from mandatory helmet legislation Curnow 2008 concluded: “Compulsion to wear a bicycle helmet is detrimental to public health in Australia but, to maintain the status quo, authorities have obfuscated evidence that shows this”'.
Secondly, the purpose of documenting the deficiencies in your 2007 article is to provide information to others reading your comments to enable them to decide how much weight to attach to your comments (ie. whether you are a reliable source).
As previously noted, Thompson (Cochrane) has observed that Curnow's commentary "contains factual errors and misinterpretations of the data" and that "many of Curnow’s criticisms stem from misconceptions about the studies that have been done". Some examples of YOUR lack of understanding of key epidemiological/statistical concepts . . .
Jake Olivier has explained (above) how you are 'still confused about statistical inference on two events and, in particular, sample size'.
In response to Olivier's question as to the absence of any recognised statistical analyses in your 2012 article, you stated that statistical tests 'may have not have been reliable considering the limited data'. This response indicates a dire ignorance of basic statistical concepts, the objective of statistical tests is PRECISELY to determine whether the 'data' is 'reliable', and if the test/s indicate that the results are not 'reliable', then the results should NOT be cited as 'evidence'.
Further examples of your lack of understanding of (basic) statistical methods, specifically with respect to your lack of understanding of the results of Olivier's re-analysis of the Walker data, can be found at http://btawa.org.au/2013/....
Your claim that 'Marshall was reporting on South Australia and has no real bearing on children’s cycling levels in NSW' demonstrates an ignorance of concepts that are taught in introductory epidemiology courses. If after the helmet laws were introduced, (child) cycling dropped in one jurisdiction but not another, it would indicate that the drop in cycling was not due to the helmet laws, but something else (eg. the weather).
(For those who are interested, further examples of your lack of understanding of basic epidemiological concepts are described at http://theconversation.co....)
And you have (repeatedly) failed to grasp the fact that the Marshall participation results ARE consistent with both the NSW school cyclist counts and NSW child non-head injury counts.
You state that 'Various claims based on Marshall are made, these need to be investigated to see if reliable.' I have not made any 'claims', I have cited the results presented in the report. And I have cited the same results, many times, over many months, in various Wikipedia and TheConversation forums in which you have been participating. An electronic version of the report is freely available, nobody has challenged the results that I have cited. (I have provided the relevant table and page numbers at the top of this post.)
Moving on to Hagel's (2006) observation that Curnow "fails to present all relevant evidence" . . .
Those who are not familiar with this topic may be interested in the numerous examples, from your 2012 article, of your failure to report evidence that is contrary to your helmet/laws-are-bad argument, that are detailed at http://theconversation.co... (search for 'Povey (1999)').
In summary, in addition to failing to cite evidence from the Scuffham (1997) and Povey (1999) studies that there was no evidence that the helmet law reduced cycling in New Zealand, and failing to cite the TBI results from the study by Tin Tin et al., you
- cited a 2005 National Children's Bureau (NCB) document as evidence that the NCB does not support helmet use/promotion
- failed to note that
- the British Medical Association DOES support helmet legislation
- a 2009 NCB document ('Active Transitions') indicates that the NCB DOES support helmet use/promotion: "Schools can undertake various measures to improve the safety of children on the school journey . . . Providing secure storage for pupils’ cycle helmets, to help alter the view that helmets are inconvenient. This will help encourage children cycling to school to wear helmets. . . Raffles for children who travel to school by foot, bike, etc. with tickets handed out with prizes like cycle helmets"
- a 2010 NCB document ('Research Summary 4') that is also supportive of helmet use: "The use of bicycle helmets reduces cycling-related injuries . . . Studies from outside the UK reported that legislation mandating the use of helmets . . . reduced injuries for children"
RE: RE: RE: RE: Considering the claims again
Colin479 replied to Linda_Ward on 03 Nov 2013 at 14:58 GMT
Just replying to some of the main points.
1
Marshall – a reference or link should have been provided, so that we or others know what precisely is being discussed.
I assume this 1994 report
https://www.bicyclenetwor...
2
NZ related
http://www.ncbi.nlm.nih.g...
Trends in cycle injury in New Zealand under voluntary helmet use.
Scuffham PA, Langley JD. 1997
‘Results revealed that the increased helmet wearing percentages has had little association with serious head injuries to cyclists as a percentage of all serious injuries to cyclists for all three groups, with no apparent difference between bicycle only and all cycle crashes.’
http://webcache.googleuse...
‘Head injuries to bicyclists and the New Zealand bicycle helmet law’
Paul Scuffham, Jonathan Alsop, Colin Cryer, John D. Langley 2000
(Data 1988–1996, inclusive)
Method
“We chose to model the data as event rates, where the numerator was the number of head injuries in cyclists. We also required a denominator, ideally this would be
a person–time exposure, which in this case is the number of hours spent by cyclists on public roads. However, given the absence of cycling exposure data we chose to use the number of non-head injuries as a proxy for this.”
Abstract;
“Results indicated that there was a positive effect of helmet wearing upon head injury and this effect was relatively consistent across age groups and head injury (diagnosis) types. We conclude that the helmet law has been an effective road safety intervention that has lead to a 19% (90% CI: 14, 23%) reduction in head injury to cyclists over its first 3 years.”
Section 2.4
“The proportion of head injury admissions for non-cyclists has changed over time. We assumed that these patterns were a result of unaccounted changes in practices by hospital admitting doctors in New Zealand or other social and economic factors not accounted for.”
Tin Tin 2010
http://www.ncbi.nlm.nih.g...
Discussion;
Included
“Our analysis showed the declining trend in rates of traumatic brain injuries from 1988-91 to 1996-99. However, it is unclear whether this reflects the effectiveness of the mandatory all-age cycle helmet law implemented in January 1994 or simply reflects a general decline in all road injuries during that period.”
3
One issue is exposure, Scuffham et al “absence of cycling exposure data we chose to use the number of non-head injuries as a proxy for this”.
Other reports have also used non-head injuries as a proxy for exposure. However, Erde and Elvik quote a 14% increase in cycling risk due to helmets, in Australia and New Zealand. Seven reports were used in evaluating the 14 % figure. Clarke 2007 reported on a number of reports indicating a higher accident rate may occur with helmet use. http://www.nationaler-rad...
Using data from Robinson 1996 and the count of cyclists as the proxy for exposure, then the outcome was reduced safety for children, see Table 2 Robinson DL; Head injuries and bicycle helmet laws; Accid Anal Prev, 28, 4: p 463-475, 1996
http://www.cycle-helmets....
Considering data from Fig 3 Tin Tin, The figure is lacking information on ‘Other’ lower extremities. The full data shows for
88/91, Overall – TBI -11.87, non-TBI – 18.58, total 30.45
96/99, overall – TBI – 7.29, non-TBI – 25.72, total 33.01
03/07. Overall – TBI – 6.23, non-TBI – 42.63, total 48.86
These rates were per million hours spent cycling.
The non –TBI rates increased from 18.58 to 42.63, 129% increase.
Robinson’s 1996 for ‘other’ equivalent injuries increased, 926 to 1595 based on injuries and counts of child cyclists.
For Canada indication from my report also suggest an increase, see
http://www.cycle-helmets....
The NZ Tin Tin data shows it would be unreliable to use non-head injury data as a proxy for exposure and based on counts in Australia it also shows it could be a mistake.
RE: RE: RE: RE: RE: Considering the claims again
Linda_Ward replied to Colin479 on 17 Nov 2013 at 11:14 GMT
Re 'Marshall – a reference or link should have been provided, so that we or others know what precisely is being discussed. I assume this 1994 report', given the amount of attention that this report received in various Wikipedia forums in which you were a very active participant, I am stunned that you should make such a statement. Especially given that you posted some census data (attempting to discredit the contents of the report), as the sentence immediately following the link for the report (http://en.wikipedia.org/w... '<ref name=Marshall-and-White-1994>{{cite book|title=Evaluation of the compulsory helmet wearing legislation for bicyclists in South Australia. Office of Road Safety Report Series 8/94|year=1994|publisher=Office of Road Safety, South Australian Department of Transport|isbn=0 7308 0075 X|author=Marshall J|coauthors=White M|month=November|url=http://www.bicyclenetwork...}}</ref> Census data for Adelaide for 1986, 1991 and 1996 have cycling counts of 8061, 7186 and 4494 respectively and represent a reduction from 1986 to 1996 of 44%.<ref>Travel to work in Australian capital cities, 1976-2006: an analysis of census data</ref>'.
And it is more than a little hypocritical that you should make such a complaint, given that you repeatedly cite the Erde and Elvik 14%, which is not supported by so much a skerrick of evidence, and you have repeatedly failed to provide the citations for the 'seven reports' that you claim were 'used in evaluating the 14% figure'. Like the material at www.cyclehelmets.org (that you are so fond of quoting), your report at http://www.nationaler-rad... and your (cyclehelmets.org) 'Canadian' report, the Erde and Elvik article has not been peer-reviewed by experts in the relevant areas (if it had been, the evidence for the 14% claim would have needed to be cited). PLOSONE is an academic journal, you should refrain from citing non-peer reviewed material in this forum. (The fact that your Canadian report has not appeared in the CMAJ suggests that it was not considered to be of a quality worthy of publication.)
You cite the results of the (1997) Scuffham study, which showed that 'increased helmet wearing percentages has had little association with serious head injuries as a percentage of all serious injuries to cyclists'. Given that the 'increased' helmet wearing rates were only 62% for teenagers, and 39% for adults, it is hardly surprising that there was 'little association'.
You claim that 'The NZ Tin Tin data shows it would be unreliable to use non-head injury data as a proxy for exposure', but the Tin Tin data you have cited appears to be unrelated to this claim. You state that 'full' data you have cited corresponds to Fig. 3, yet it seems to be unrelated. For example, you state '88/91, Overall - TBI 11.87', yet there is no TBI rate of 11.87 in fig. 3 (the TBI rates in the figure look to be about 4, 8, 3 and 5).
Re your claim that Robinson's 1996 shows that helmet law reduced cycling/safety for children . . .
In Sydney, (3332-2442)/2442=36% MORE adult road cyclists were counted in the 1st post-law survey (1991) than in the pre-law survey (1990). There was also an increase in the number rural adult road cyclists counted in 1991. Yet in rural NSW, (668-850)/850=21% FEWER child road cyclists were counted in 1991 than in 1990, the 1991 counts were BEFORE the helmet law. The Sydney child road cyclists counts were about the same in 1991 as in 1990. So between 1990 and 1991, the number of adult road cyclists increased despite the helmet law for adults, yet the number of child road cyclists decreased, despite there being no helmet law for kids.
Jake Olivier has noted above that the use of a dependent non-equivalent no-treatment control is the most recommended type of comparison in the statistics literature. The correlation coefficient corresponding to the data in Robinson's Table 2 is 0.82, the use of non-head injuries as the comparator is clearly going to produce a MUCH more accurate result than Robinson's 'proxy measure'.
As previously noted, the data in the Williams study shows that between 1990 and 1993 serious casualties for all NSW road users dropped by 15%. Based on the data in Robinson's Table 2, between 1990 and 1993, child non-head injuries fell by exactly the same amount. So, as pointed out to you in my previous reply, even the data in Robinson's (/your) article indicates that there was no (significant) drop in (overall) child cycling in NSW as a result of the helmet law.
Also as previously noted, the table 2 data shows that in 1992/93, compared to 1989/90, NSW child cyclist non-head injury hospital admissions reduced by 13%, and head injuries reduced by 37%. This means that there was a 37%-13%=24% reduction in NSW child cyclist head injuries attributable to the helmet law (ie. INCREASED safety).