Following some early feedback from the Beta, a couple of new features have been added to the Navigator.
First, to make the Navigator quicker and easier to find, there’s a new big green button on the home page of this website:
(Depending upon your browser, you can probably see the picture of the buttons above as well as the actual buttons.) Old School access is still, of course, available using the web address navigator.crickles.org.
Second, when you’re using the Navigator there is now a small block of help text pertaining to each report. As in this example, this appears in the side panel and changes according to the tab that you’re looking at (in this case Fit-Fat):
General help on the use of the Navigator, such as how to reset the Date range quickly, is still available here.
You may have noticed recently that there is a new report on the Navigator that requires a password, and that a few of the posts on this website are now password protected. This reflects some significant changes to Crickles that are being released initially in a BETA programme. If you’d like to participate and check out the new features please read on…
To access the BETA functionality
Send an email requesting access to email@example.com.
I will need your Date of Birth – this will enable us to factor in age into the new analysis.
This aside, you are only asked to agree to:
Accepting the confidentiality terms;
Giving me feedback;
Acknowledging the limitations of the beta.
This is all described further in the Mechanics of the beta section below.
Please do try it out!
Here’s what’s in the beta:
More analysis of heart rate data
One of the main features of the BETA is a new Regularity report. This gives a new type of analysis on your heart rate data. It also gives some analysis on the “strap crap” that is filtered out by the Crickles data cleaning routines.
For those on the programme, a full description is available here.
Just your data
To date, all of the analysis on the Navigator has been available to everyone on an “all see all” basis. The new report is potentially more confidential in character and so each athlete can see only his or her own data. This is why we now require a password to access that report. In future, we may password-protect more information, such as FTP curves – subject to what people would prefer. However, peer comparisons are super-useful and we’ll keep these as a prominent feature.
To date, you can compare yourself to others using the Group dropdown. However, apart from the gender sub-selection that this offers, this isn’t useful for most of the Crickles population. On the beta, once you’re logged in you can now choose Strava Friends from the Group dropdown. This then enables you to compare yourself specifically against your Strava friends on the relevant reports (Relative CSS, Relative Profile, All-in and This week). This is much more meaningful – and you can still compare yourself to the overall Crickles cohort and your gender group.
Mechanics of the beta
Please email me as above to request a password. All passwords are encrypted in use and cannot be hacked in plain from a web server. However your password is not encrypted “in flight” when I send it to you. If you are concerned about security I can text you your password instead of emailing it, and if you’re very worried I can give it you by phone.
At this stage I haven’t built a mechanism for you to choose or re-set your own password. Also, the password setting process is still manual and it may take me a while to get round to sending out yours if a lot of people ask for them.
As far as I’m aware, this new analysis, and indeed some of the existing analysis, is unique and unavailable anywhere else. While I’m continuing to develop it, please treat it as commercially sensitive – for example, don’t email Strava describing it and asking for them to copy it!
To log in, choose the Regularity report, which will throw the log-in screen (unless you’re already logged in). The User ID you need is simply your Strava ID (e.g. mine is 301194). You can see this on the Athlete dropdown in brackets. Once you’ve logged in you’ll be returned to the main page (Relative CSS) and will need to choose Regularity again to see it.
While this functionality is still in beta, you can always escape back to the current way that the Navigator works by refreshing the page. This will log you out. On Safari at least, the browser caches your ID and password so you don’t need to re-enter them. The iPhone doesn’t do this – when we move beyond beta I intend to build this into Crickles so that you will rarely need to re-enter your credentials.
You will also get a general-purpose password for accessing the protected posts on crickles.org.
If you sign up to try out the beta, please could you give me feedback on the new features. It would also be super-helpful to know:
what you use Crickles for;
what you never use;
roughly how often you look at the Navigator;
whether you ever use Crickles Activity Charts;
what further improvements would make Crickles most helpful.
A couple of weeks ago Cardio Mark ran the Dublin marathon in a very creditable time of under four hours. I’m impressed. When I used to run regularly, I once or twice managed half marathon distance and I never finished wishing I’d done twice as much. Completing a marathon requires months of training and resolute determination.
A couple of days before Mark’s marathon I went for a bike ride. I had the day to myself so cycled over to Dunkery Beacon – the highest point in South West England and the region’s signature bike climb. It’s not the longest climb and not the absolute steepest but its combination of length and severity is nonetheless testing. The rest of my ride was not without interest, extending to 93 miles with 7.5k feet of ascent and taking me around 6:40 with a couple of stops. It was equivalent in difficulty to a reasonably stiff domestic sportive. Fit cyclists do not need to train for such sportives – they’re easy enough that you can just rock up and do them.
Out of curiosity, I compared Mark’s marathon with my bike ride using Crickles Activity Charts. Charting a histogram of our heart rates gives this, with Mark’s histogram in pink lying over mine in blue:
On the morning of my ride my resting heart rate was 46 bpm, which is marginally above Mark’s (44 bpm). Crickles shows our prevailing Lactate Threshold Heart Rates to be similar too (mine was 162 bpm and Mark’s was 157 bpm). The chart thus shows that not only did Mark record far fewer beats during his marathon than I did during my ride, he also managed his effort much more prudently with the entirety of his cardiac activity well under his LTHR in contrast to my ride in which I was well over my LTHR for a significant portion of the time. It is graphically evident that my ride placed a much greater load on my heart than Mark’s marathon placed on his. This is reflected in our respective Cardiac Stress Scores – 570 for me and 300 for Mark. (Reflecting on this, I realised that the Crickles CSS methodology does not fully capture the restorative effect of my two coffee stops. I calculated the effect of these and it decreases my CSS by 17 points – just under 3%, so not a significant amount.)
By contrast, the differences in our Suffer Scores on Strava was much narrower – Mark’s Suffer Score was 255 compared to my 299. This confirms my previous finding that the Strava Suffer Score is not a good measure of cardiac stress.
There is an important point here for cyclists. Mark’s marathon felt hard and was hard because of the intense amount of corporeal stress that running for four hours places on the body. Skeletal muscle and the bones, ligaments and tendons to which it is attached, can feel the pain. There is also that intangible of “the wall” as it is harder to refuel adequately when running. Cycling inflicts far less strain on skeletal muscle and doesn’t pound our body in the same way. Unlike our peripheral muscles, our exercising heart does not feel pain and we thus have no direct real-time index of the stress to which we subject it. So long as we ride within our strength limits and eat and drink well, the wall of bodily pain that marathon runners have to run through has no analogy in cycling and there is nothing to indicate how much cardiac stress we’re accruing. The only direct evidence that our sportive was equivalent in cardiac stress to two marathons might be a histogram or a CSS number. An appreciation of this might spur us to think more seriously about the amount of recovery we need to build into our exercise schedule.
I have updated this post now. Previously I just attached the word file – time constraints. But today I have some time, so here goes. As far as I can tell, this is the most complete analysis of the data that exists to date.
A search using the term “Marathon” and “Death” or “Mortality” on Pubmed yielded 167 references (most recently performed 20/10/17). The abstracts of these were reviewed. The reference lists of the final articles included in this paper were scanned for further relevant literature and review articles were obtained and read and their reference lists reviewed. On-line web searches and direct approaches to authors were made to provide additional data.
The End-Note library can be found here. It does not include the texts of some of the papers which were behind pay-walls.
A total of 8 papers were included in the final analysis. The papers can be divided into two categories. The first category comprises a series of papers which looked broadly at hundreds of marathons run across the USA over the past 40 years. The second category comprises a number of papers that have focused on two marathons in the USA – The Marine Corps Marathon and the Twin Cities Marathon – and also the London Marathon.
I have not included results relevant to half marathons or ultra-marathons or other events, such as triathlons, except where indicated in the text. Furthermore, I have not included mortality from marathons run as part of an event, such as a triathlon.
There have been a number of efforts to collect data from a larger number of marathons that have been staged across America. It is probable that the data from these papers overlaps. Furthermore, it is more likely that deaths will be missed compared with the studies which have focused on specific marathons, and therefore these papers have the potential to underestimate the true fatality rate.
Redelmeier et al. 20071
Redelmeier and Greenwald published an analysis of 26 marathons and their related deaths, staged in the USA, which they followed for up to 30 years. The principal focus of the paper was to compare the number of marathon deaths with the expected number of motor vehicle fatalities to determine whether running a marathon was safer than not running one.
The authors screened the 328 American marathons listed in Runner’s World on 01/01/2005. They excluded those with less than 20 years of data, fewer than 1000 participants annually, or those that were located primarily on off-road trails or that were part of triathlons or other combined endurance events. They selected 26 marathons randomly from the remainder. Note that the Marine Corps data is included in this analysis.
Data on “sudden cardiac deaths” was obtained from local newspapers on the days after each marathon. Race directors were contacted.
A total of 3,292,268 participants were included in the analysis. There were 26 “sudden cardiac deaths”. 15 marathons had no death, 6 had one death, and 5 had more than one (Boston, New York, Chicago, Honolulu and the Marine Corps Marathon). New York had 2 deaths in 1994.
The average age of the fatalities was 41 years. 81% (21) of those who died were men. 5 deaths occurred in people who had completed a marathon before.
There were autopsy results on 24; 21 had coronary artery disease. Coronary anomalies were noted in 2. Electrolyte abnormalities were thought to be significant in 4 and heat stroke in 1. Most died within a mile of the finish line (Figures 1a and 1b).
The overall risk of a fatality, as estimated by this paper was 0.79/100,000 or 1 death for every 126,626 finishers.
Figure 1a. Location of Fatalities, Redelmeier et al. 2007
Figure 1b. Location of Fatalities, Redelmeier et al. 2007
Mathews et al. 20122
Using publically available racing (MarathonGuide.com, Athlinks, and The Association of Road Racing Statisticians) and news (Google, LexisNexis) databases and by directly contacting race organisers, the authors collected statistics on marathon finishers and deaths between 2000 and 2009. Data was cross-referenced with MarathonGuide.com.
The denominator – the numbers completing marathons during the 10-year period – was 3,718,336. There were 2,255,060 men who completed a marathon and 1,463,276 women.
They identified 28 people who died during the race or within 24 hours of finishing. 6 women and 22 men died. The male death rate was 0.98/100,000 and the female death rate was 0.41/100,000. The overall death rate was 0.75/100,000.
The median age of death was 41.5 years (IQR 25.5, range 22-68).
14 deaths occurred in people over the age of 45. 13 of those deaths were caused by atherosclerosis (Table 1). In younger racers, none of the deaths were caused by cardiovascular disease.
The median distance travelled before dying was 22.5 miles (IQR 10.6). 7 completed the marathon before dying. 18 deaths occurred after mile 20. Cardiac and cardiovascular aetiologies accounted for 24/28 deaths.
People were more likely to die in October marathons (n=11); 27% of marathon participants raced in October. This was of no consolation to myself when I ran the Dublin Marathon, in October, aged 47.
Table 1. Cause of Death, Mathews et al. 2012
Kim et al. 20123
In 2012, a study by Kim et al was published in the New England Journal of Medicine. They looked at cardiac arrests that occurred whilst running or within an hour of running a marathon or half marathon. I have not reported the ½ marathon data.
The database of cardiac arrests was compiled prospectively from Jan 1st 2000 to 31st May 2010 – a slightly different time period to Mathews et al. The arrests were cross referenced using LexisNexis and Google. Further searches were performed directed at particular race events and their local newspapers. Contact was also made with race officials. Cases of cardiac arrest were retained for the final analysis if they were identified in 3 separate data sources or confirmed with race staff. The next of kin of those who died were written to asking for further data about exercise history and family history and asking for consent to access medical data.
Running USA compiled statistics for participation (not finishers) rates in marathons or half-marathons in the USA. It was estimated that 3,949,000 people participated in a marathon, and 6,922,000 participated in a half marathon.
59 cardiac arrests were identified, 40 in marathons and 19 in half marathons. The incidence of cardiac arrest was 1.01/100,000 in marathons. 51 of the 59 people who arrested were male. 34 of the 40 arrests in marathons were in men. Male marathon participants had a rate of cardiac arrest of 1.41/100,000. The mean age of those who arrested was 42±13 years.
42 of the 59 died. The mean age of those who died was 39±9 years. The death rate was 0.63/100,000 during marathons (n=25) and 0.25/100,000 (n=17) during half-marathons. Overall men were more likely to die than women (0.62/100,000 vs. 0.14/100,000) during marathons and half-marathons.
Most arrests occurred in the latter quartile of the race (Figure 2). The data was not separated by race type or by mortality.
Hypertrophic cardiomyopathy was the most common underlying diagnosis overall. In those who survived, coronary disease was more common.
It was recognised that the method of ascertainment of cardiac arrests may have missed some cases, and therefore, the true rates of cardiac arrest and death may be higher. Complete clinical data was missing on approximately 50% of the cases. It was also recognised that the participants may have run multiple marathons, and therefore the total number of unique participants, would have been lower.
Figure 2. Location of Cardiac Arrests, Kim et al. 2012
Webner et al. 20124
In 2009 this group sent 33 item surveys to 400 race directors of US Marathons to ask about the number of marathon participants and associated deaths. 88 (22%) returned the surveys. The marathons were run between 1976 and 2009.
There were a total of 1,710,052 runners. Races had between 30 and 30,000 participants. 30 arrests and 10 deaths were reported. The risk of death was therefore 1 in 171,005, or 0.58/100,000. The cause of death was coronary artery disease in 7. One person had an anomalous coronary artery. The cause of death was not established in 2 cases. 28 of the 30 runners who arrested were male. The mean age of those who arrested was 49.7 years. The mean age and sex distribution of those who died was not specified in the paper.
The location of the arrests was again skewed towards the final miles of the marathon (Figure 3).
There were more participants in this paper, and one less death, than reported when the abstract was presented in abstract form in 2011. Discrepancies between abstracts and final papers are common as the peer review process often picks up errors.
Figure 3. Location of Cardiac Arrests, Webner et al. 2012
These papers can be brought together. It is clear that the typical person who arrests is a middle-aged male around 40 years old. The cause of death is often considered to be coronary artery disease. Typically, cardiac arrest and death occurs in the latter quartile of the race.
The risk of death estimated by these studies was approximately 0.70 per 100,000 finishers, that is 1 death per 142,000 participants (Table 2).
Table 2. Risk of Death During Marathons
Of the two studies that reported marathon deaths by sex, that is Redelmeier et al. and Mathews et al, there were 43 males (80%) who died and 11 females (20%). Mathews et al. reported the breakdown of participants by sex in addition. In that study there were 22 men who died and 6 women. The rate of male deaths was 0.98/100,000 (1 per 102,503) vs. 0.41/100,000 (1 per 243,879) in females (Table 3).
Table 3. Risk of Death by Sex
If further data was available from the papers, more accurate estimates could be made.
The Marine Corps Marathon and the Twin Cities Marathons
A number of papers have been published by the medical team associated with two marathons in the USA: The Marine Corps Marathon (MCM), held in Washington DC and the Twin Cities Marathon (TCM) in Minneapolis. Three papers have been published over the years, looking at data from these two marathons. All are freely available online.
Maron et al. 19965
This paper focused on the MCM from 1976-1994 and the TCM from 1982-1994. All deaths were included. The data is reproduced in table 4.
215,413 runners successfully completed the marathons during this time. 4 sudden deaths occurred during (n=3) or shortly after (15 minutes, n=1) completion of the marathon. 3 were male and 1 was female. The 3 men had coronary disease, the 1 woman an anomalous coronary artery. The one woman died in 1990 in the MCM, aged 19. One man died in the TCM in 1989, aged 40. Two men died in 1986 and 1993 in the MCM, aged 32 and 58 respectively.
Table 4. Maron et al. 1996
Roberts et al. 20006
This paper looked at the 81,277 entrants in the TCM from 1982 to 1994. It therefore contains no new information over and above that of Maron et al. 1996 with regards to deaths.
Its focus was describing all medical issues experienced by marathon runners. For interest medical encounter rates were 25.3/1,000.
Of note, it was stated that 60,757 finished the course – slightly different to the 60,379 from the 1996 paper. There were 48,330 male finishers and 12,427 females. There appears to be a further typographical error in the text, and this number is derived from the table reproduced below (Table 5).
Table 5. Roberts et al. 2000
Updating the data from 1996 yields the following (Table 6):
Table 6. Combination of Maron et al. 1996 and Roberts et al. 2000
Roberts et al. 20137
This paper used data from the two marathons between 1982 and 2009. 1982 was chosen, as data was reported by sex from that year onwards.
The paper reported that there were 540,892 finishers during the study period. 379,863 were male and 168,227 were female.
In total 14 runners collapsed suddenly. 7 were successfully resuscitated. There were 7 sudden cardiac deaths. There was still only 1 female fatality (age 19, anomalous coronary artery, 1990, MCM). The rest who died were male. The underlying diagnosis for all men was coronary artery disease. The mean age of the men was 48 years. The location of the collapse was on average at 16 miles.
The rate of death overall was 1/78,299 finishers or 1.28/100,000. In men the rate was 1/63,311 finishers (1.58/100,000) whereas in women it was lower at 1/68,227 (0.59/100,000).
Details of the deaths are as follows (Table 7).
Table 7. Details of Deaths. Roberts et al. 2013
The Marine Corps Marathon have now published data on their website of the number of finishers since the race began in 1976. It can be found here, and is reproduced in table 8.
Table 8. Marine Corps Marathon Finishers
This allows for a combination of all of the papers, leading to a final estimate of risk:
Between 1982 and 2009 there were 540,892 finishers. From 1976 to 1981 a further 26,387 completed the MCM. The total finishers of the two races from their inception (MCM 1976, TCM 1982) to 2009 was therefore 567,279. There were 7 deaths in that time. The rate of death, therefore, was 1.23/100,000 finishers, or 1 death for every 81,039 finishers.
The London Marathon
Dan Tunstall-Pedoe was the London Marathon Medical Director between 1981 and 2006. He presented data on the first 26 London Marathons in 2007.7
In the paper, data was obtained from St John’s ambulance, receiving hospitals and the coroner. The paper stated that there were a total of 650,000 finishers. A total of 8 deaths were recorded in this time.
The first death was in 1990, and ascribed to hypertrophic cardiomyopathy (HCM). In 2001 and 2005 there were two further deaths from HCM. Deaths in 1993, 1995, 1996, 1997 and 2003 were ascribed to ischaemic heart disease.
This yielded a crude rate of death of 1/81,250 (1.23/100,000).
There is also corroborative data online, written by him and published at Peak Performance with data up to and including 2003. The data there is slightly different. HCM deaths were reported as occurring in 1990 and 2001 – as above. 5 deaths from ischaemic heart disease were reported as occurring in 1991, 1994, 1995, 1997 and 2003. This was the same number of deaths as in the published paper over the same timeframe. Data was published on the number of finishers up until 2003. This yielded a death rate of 1/67,414 (1.48/100,000).
An online search to determine the number of finishers yielded data from additional sites. There are race reports, hosted on the London Marathon website, which mention the number of finishers in the very early races. Wikipedia provides data on the first London Marathon. Marathonguide.com has data on the number of London Marathon finishers. I also searched the London Evening Standard and BBC News Websites. Combining the data from these sources yields the following table (Table 9). All the data is from marathonguide.com unless otherwise specified.
I also emailed Professor Sanjay Sharma, the current director of the London Marathon, who was kind enough to respond directly. He stated that there have been 14 deaths.
As of 2017 over 1,000,000 people had completed the London Marathon (1,038,733). This yields a death rate of 1/74,105 or 1.35/100,000. I look forward to seeing a more detailed paper from Professor Sharma with confirmed finishing statistics and further medical details, including the age and sex, of those who died.
Table 9 summarises the data as far as is possible.
4 Tunstall-Pedoe D. Marathon Cardiac Deaths. Sports Med 2007;37:448-50.
HCM Death from hypertrophic cardiomyopathy
IHD Death from ischaemic heart disease
These two sets of data can be drawn together, yielding a death rate of 1/76,477 finishers or 1.31 deaths/100,000 finishers. The two separate marathons have strikingly similar results (Table 10). The rates of death are higher than those ascertained by more general surveys, as would be predicted as the authors were directly linked with providing medical support to the marathons over many years.
Table 10. Mortality During Specific Marathons
Men around the age of 40 are the most likely to die participating in a marathon. Death typically occurs in the latter stages of the marathon, and is often caused by previously undiagnosed ischaemic heart disease, although hypertrophic cardiomyopathy and increasingly heat stroke are also concerns.
Different methodologies have yielded conflicting estimates of the death rates during marathons. Large surveys of multiple marathons have suggested that the death rate is 0.70/100,000 finishers. Detailed studies of a small number of events has suggested the rate is higher at 1.31/100,000 finishers.
As always, the truth is somewhere in between. But what is clear is that the chances of dying during a marathon are very low.
The data are frustrating. The authors of the papers almost certainly have some further data, not included in the publications that could facilitate a more formal meta-analysis.
What is needed, however, is a prospective registry of entrants / finishers and medical events from race directors. It would help understand what the true extent of the problems are. It would help concentrate medical expertise and help train event professionals to deal with the most common scenarios.
How to screen for these problems in advance? An ECG can be helpful and is often advised, but does not reliably pick up either ischaemic heart disease or hypertrophic cardiomyopathy. An echocardiogram can diagnose hypertrophic cardiomyopathy reliably. A cardiac CT can pick up asymptomatic coronary disease. Judging from the lack of symptoms and training, an exercise tolerance test or stress-echocardiogram is likely to be falsely reassuring. But that is a lot of testing to prevent a rare event. It’s not cost effective, and can probably not be supported by the NHS in the UK, at least.
The Crickles Navigator has a new Summary tab giving headline figures for the athlete and a comparison of each with peers. Currently, there are six figures shown in the Value column:
Period CSS shows the total Cardiac Stress Score for the period defined by the Date range. This period defaults to the last six weeks but can easily be changed in the side panel.
Period XSS extends this measure to cover also activities on Strava for which there is no heart rate data. The estimate of cardiac stress is less good than CSS, which requires a heart rate monitor, but better than assuming that it’s zero. If you always wear a heart rate monitor CSS and XSS will be equal (though not for the peer group). If you never wear a heart rate monitor, CSS will always be zero but XSS will usually be positive.
Current LTHR is the latest Crickles estimate of your Lactate Threshold Heart Rate. (Occasionally this estimate may lag one activity behind your last effort.) This only appears if you use a heart rate monitor.
Current Fitness, Fatigue and Form are as described here. Since they are current estimate, neither these values nor Current LTHR change as you change the Date Range.
The numbers alongside these in the Crickles Percentile column show how you compare to the Group chosen in the side panel (by default, the entire Crickles population). If the Crickles Percentile is 100 you have the highest value. If the Crickles Percentile is 50 you’re on the median.
By design, this tab works particularly well on the iPhone.
Over time, we may add more information to this tab.
NB: There is a known bug that the Summary tab will be totally blank (i.e. no data) if you have included the character “(” in your Strava name. If this affects you and you’re interested in seeing your Summary data before I otherwise get round to fixing it, please let me know!
The Timeline (previously called CSS (Cardiac Stress Score) Map) on the Crickles Navigator has been overhauled to display data differently. Now, the x-axis is the timeline, which still reflects the Date Range on the side panel, and the y-axis shows the CSS for each activity. This makes it easy to pick out which activities have the highest CSS (those that lie highest on the chart) and when you did them (more recent to the right, least recent to the left). The composition of CSS into its elements is now encoded through size, which represents moving time – larger dots are longer activities – and colour – the least intense activities are green and the most intense are red.
Here’s an illustration – note that CSS Map now appears as Timeline as the tab name:
In this example, the activity with the highest CSS occurs half way along the chart at the top. The size (large) and colour (medium orange) indicates that the activity was long in duration but only moderately intense.
You will also see that there are some grey dots. This is because the Timeline now also includes activities for which there is no heart rate data and so Cardiac Stress has to be estimated from moving time alone.
Group selection is now under Athlete selection so that you only have to select from the groups to which you belong. For the majority of athletes, this reduces to a choice between “All” athletes (the default) and Male or Female.
LTHR and FTP
Lactate Threshold Heart Rate (LTHR) and Functional Threshold Power (FTP) are now charted on separate tabs. This makes changes in LTHR in particular easier to discern. Obviously if you don’t cycle with a power meter you won’t get an FTP chart.
Last loaded date/time
The notice of when Navigator data was last updated was previously given in the arcane UTC timezone beloved of computers. This is now changed to London time.
Please let us know if you have issues with any of the above.
There is no doubt there is a big difference between myself and Ian. He will be going off to Italy to cycle up a hill in October on a very nice bike. It will be sunny and there will be wine and olives. Later in the month, I am off to run a marathon in Dublin. I won’t bother checking the weather. It will be cold and wet. And there will be no wine at all, although hopefully some Guinness, so not all bad. There are times when I wish I had put a bit more effort into maths at school.
I do worry about running “long” distances, and what strain it puts upon the heart. But what are the risks of dying during a marathon? You can’t rely on newspaper headlines. They only report the bad events. They don’t write headlines along the lines of “Marathon run today, nobody died”. It’s therefore important to look at studies where it has been planned to study the outcomes, or where outcomes are tracked over a long period. But finding raw data is harder than you might think.
I thought I would start with the Berlin marathon. It’s on soon – typically taking place at the end of September. And it’s fast – a place to set world records. More importantly it’s run by a lot of people – 46,950 people in 2016. When there are big numbers there can be good data. Unfortunately, after emailing them, they don’t keep statistics. So that’s no good.
What about London? In 2001, I “ran” the marathon. It turns out that not running at all in the previous month because of a knee injury and drinking wine each night doesn’t make for a good marathon. Being overtaken by a tree is pretty galling.
The first London marathon was staged in 1981, and since then over 1,000,000 people have completed the race. There have (probably) been 14 deaths since then, although it depends on how things are counted, and there are discrepancies between reports.
From 2007, race statistics and details of deaths are pretty secure. David Rogers, aged 22, died that year from hyponatraemia (water intoxication). In 2011 Claire Squires, aged 30, died of heart failure in front of Buckingham palace. It was felt that DMAA, a now banned amphetamine stimulant, contributed (it wasn’t banned at the time). In 2014 Robert Berry, aged 42, died from heat stroke. In 2016 David Seath died from heat stroke (probably) aged 30.
258,911 men have completed the race since 2007 (including 2007 and 2017) and 140,271 women. So, the death rates are 1.2/100,000 for men and 0.71/100,000 for women.
If you accept that 14 have died over the course of the race, and that 1,039,225 people have completed it (a combination of race reports from the London Marathon website, marathonguide.com, and a page on peakendurance sport.com) then the overall risk rises to 1.5/100,000 for both sexes.
Boston is another famous city with a famous marathon, unfortunately for sad reasons at present. I’ll be looking out for Stronger when it is released in the UK. There is a professor at Harvard who has collated some data. When it arrives in the post, I will update the blog with the key points.
The New York Marathon is the world’s largest. It started in 1970. They have a great analytics page (http://www.tcsnycmarathon.org/analytics). Tata Steel aren’t popular in the UK, but Tata Consultancy Services are my new friend. 1,070,784 people have participated in the NY Marathon since 1970 – 764,609 men and 306,175 women. The average man aged 40-49 completes the distance in 04:15:57. Finding the deaths is a little harder. 3 died in 2008. They don’t keep those stats on the website, and more data trawling is required. I have emailed them, but won’t hold my breath.
What have I learned? Finding out accurate data is hard. I still have more stones to look under. And marathon organisers don’t want to hear about deaths or advertise them (odd that…). If you have access to any data please let me know. But one thing is certain: deaths running a marathon are mercifully rare. It’s about the same risk as spending an hour on a plane (http://www.besthealthdegrees.com/health-risks/).
I’ll book a return ticket back from Dublin. The flight is quite short.
CSS is a great measure of Cardiac Stress and, quite naturally, it relies on the athlete wearing a heart rate monitor (HRM) during exercise. However, sometimes we exercise without a HRM and sometimes we want this to “count” towards our overall accrued CSS. This is recognised in the Fit-Fat tab, which makes an estimate of CSS even for activities in which a HRM is not used.
This same estimation methodology has been extended to give an estimate of CSS its components; this is shown on a new All-in tab. Here, you can see four charts:
First, an estimate of CSS that takes into account all activities recorded on Strava, even those where an HRM was not used:
Here, the purple mass (technically, a density plot) shows the distribution of estimates amongst all Crickles athletes over the chosen period. We can see in this example that the most common value is at about 2,500, there are quite a few athletes (around 1/5th of the number at 2,500) at 5,000 and even a blip indicating one athlete over 20,000.
The vertical black line picks out the value for the chosen athlete (me in this example), and you can see that in this case the athlete (me) is bang on the most common value.
If an athlete wears an HRM all the time the estimate of CSS will be as good as we can get. Conversely, to the extent that an HRM is not used the quality of the estimate will decline. The second chart shows this quality for the given athlete over the chosen period:
Again, the purple mass is a density plot showing where Crickles athletes overall lie on the quality chart. Happily, the most common value is at over 90%. About 1/6th of the number at this high value are at zero, indicating that they didn’t use an HRM at all over this period. Again, the vertical black line represents the chosen athlete and again it’s me. I’m at about 87.5% usage. To verify consistency, you can multiply this percentage (87.5%) by the CSS estimate (2,500) we just saw and that will show the CSS that you see on the Relative CSS tab (about 2,188).
Very thoughtful readers may correctly spot that when you’re on a bike with a power meter Crickles is able to give a high quality estimate of CSS even in the absence of a HRM. This is indeed true but this is not reflected on this quality chart.
The third and fourth charts show the components of the full CSS estimate. These are Intensity:
and average weekly exercise hours:
The Intensity estimate is subject to the same estimate quality as the full CSS estimate (more HRM usage -> a better estimate) whereas the weekly exercise hours are a simple sum of “moving times” and are unaffected by HRM (non-)usage.
These charts work the same way as the others: for example, the most common weekly exercise time amongst Crickles users over this period is shown by the last chart to be five hours (the peak of the purple mass) and I did fractionally more than that (the vertical black line).