In defence of Paula Radcliffe and her biological passport

Most posts on here will be about evolutionary genetics, but here’s a diversion into athletics and Sports Science. Paula Radcliffe, the marathon world record holder and UK running legend, has been in the news here (and I guess elsewhere) following a Parliamentary Culture, Media and Sport Committee Hearing on doping in sport. Although she wasn’t named, a comment was made that could only really point the finger at her and nobody else. She released a press statement later that morning denying that she has ever cheated, and explaining both her anger, and the reasons for her unusual tests – more details below.

The current process for trying to detect drug cheats in sport is a process where athletes are regularly blood tested, in order to build up a picture of their ‘biological passport’. Various blood parameters are measured, and the idea is that a ‘clean’ athlete will show pretty consistent readings for these parameters while a cheat will show much more variation between samples. The data are formally analysed using a Bayesian framework. Testers can then identify suspected cheats and test them more rigorously. The Sunday Times, one of the UK broadsheet (i.e. not tabloid) papers has been running a series of investigations about doping in sport, especially athletics. They obtained access to biological passport data from 2001 and 2002, and without naming any athletes, suggested that unusual biological passports were very common, and that doping was rife in endurance sports such as long-distance running. They argued that the IAAF, the body which governs world athletics was ignoring much of these data, and that many Olympic and World Championship medallists showed these patterns, among them (shock, horror), athletes from the UK, including one very well known one. There has been some speculation on social media etc who this unnamed athlete was, and what the results showed. Well now we know who it was, and we know a fair bit about the samples.

What happened yesterday:
During the Parliamentary committee meeting yesterday (available here at about 12:18 – the chair essentially revealed the identity of that well known athlete as Paula Radcliffe. He implied that the list of ‘suspected’ athletes included London marathon winners, including a British one. Paula Radcliffe is the only British winner of the London marathon since (well before) 2001. Because the meeting took place in parliament, he is protected from legal action by ‘Parliamentary Privilege’. I didn’t actually notice any reporting of his comments or of people putting two and two together, until Paula Radcliffe released a statement condemning his ‘implication’ that she was cheating. The statement gives quite a bit of background to her samples, including the revelation she has produced three ‘abnormal’ readings that were seen by the Sunday Times.

Athletes releasing their Passport data:
In the months following The Sunday Times article and a BBC Panorama documentary on doping in athletics (where it was alleged that Mo Farah’s coach was involved in helping athletes to cheat), there has been quite a bit of discussion about whether athletes should make their passport data available to the public to ‘prove’ that they are clean. In fact, Mo Farah and 9 other top UK athletes released their data. Each of them showed very little fluctuation over the time series of their readings and their parameters were well within the normal boundries. Paula Radcliffe didn’t release hers (she had no obligation to), but she did point out (correctly) that a biological passport could not prove that somebody was clean, and perhaps more worryingly, clean athletes might still produce abnormal readings which, although explainable, would look suspect if taken out of context. It’s pretty clear now, that she falls into this latter category.

What do Paula Radcliffe’s samples show, and what does the science say?
We can only really glean what was ‘abnormal’ about Paula Radcliffe’s samples from her statement yesterday – it’s in full here, – because the data are not available, and anyway they are probably impossible to analyse in isolation. What is clear though, is that the three ‘abnormal’ samples were taken following periods of altitude training, and that one of them was taken right after a race held in high temperatures. The questions I had, when I read this defence, were ‘Do these things matter?; Do they affect the readings?’ A quick delve into the scientific literature and it’s pretty clear that they do matter…… a lot. As she writes in her statement, experts have analysed her data and concluded there is no case to answer.

I don’t claim this to be an exhaustive review, but it is a summary of a fair bit of the most recent, most relevant and most cited literature on biological passports. For a couple of decent reviews of the Athletes Biological Passport (ABP), check out [1, 2]. They explain how the ABP works and what parameters are fitted in the model (in case you are wondering, they are gender, ethnicity, age, altitude – both at the time of the sample and beforehand, type of sport and the make/model of the analysis equipment). The model estimates the expected range of each blood parameter, based on both previous readings and the parameters above, and then reports whether scores exceed the e.g. 99%, 99.9% sensitivity threshold. In PR’s statement she says one of readings was above the 1 in 100 false positive test threshold (99% sensitivity) but below the 1 in 1000 threshold (99.9% sensitivity). Although the ABP model takes into account altitude and type of sport, it does this in a fairly crude way. For example, it fits into the model whether it was an endurance sport or not, rather than distinguishing between, say, a 5000m race and a marathon. With altitude, it simply fits whether the training or test were conducted above or below a certain elevation e.g. 1000m. It doesn’t distinguish between different types of endurance event, or between having just trained at altitude and having returned some time ago from training at altitude. The extent to which a sample will be an outlier will also depend on how many previous tests were carried out in similar (or different) conditions. There is strong evidence that readings are sensitive to the time since competition or intense training [3]. In fact, it is recommended that samples are taken at least two hours post a race [note Paula Radcliffe’s comment that hers was taken right after a half marathon race].

A couple of studies have actually looked at how often ‘abnormal’ results are returned when elite athletes are sampled while competing or training at altitude. The sample sizes are small and the studies were done on swimmers and cyclists rather than runners (although both are endurance sports, like marathon running). The athletes were all thought to be clean, based on long-term monitoring of their samples. The findings are remarkable though. The study of swimmers showed that after altitude training, 6 out of 10 swimmers showed ABP scores above the 99% threshold [4]. Among cyclists competing in a multiple day race at altitude, 5 out of 25 exceeded the 99% threshold (1 in 100) and 2 of them were above the 99.9% threshold [5]. The combination of being at altitude and competing intensely gave results that, taken in isolation, would be regarded as abnormal or even highly abnormal. In light of these findings, Paula Radcliffe’s results look far from abnormal; they actually look pretty typical.

I don’t believe for one second that Paula Radcliffe is a drugs cheat. Perhaps more than any other athlete she has taken a strong and clear stance against cheating in her sport. Most famously perhaps, when she and a team-mate held up a sign reading ‘EPO Cheats Out’ in the 2001 World Championships, after an athlete failed a drugs test and was reinstated to the 5000m. Bringing that attention on herself is not the action of a cheat. It looks to me as though the Sunday Times dataset has been analysed in a less than robust way, and therefore the number of ‘abnormal results’ far exceeds the number of genuinely illegal samples. I don’t doubt that cheating is pretty common in athletics, and there is evidence that systematic doping is happening in some countries; see [6] for example. The science behind the ABP looks pretty robust, and the leading experts are well aware that the underlying models need constant refining e.g. “In particular, modifications of haematological parameters during and after exposure to different altitudes/hypoxic protocols need to be properly included within detection models.”[7]. In fact, I was impressed with the sophistication of the analyses in some of the papers I read – I was prejudiced about the rigour of sports science. The ABP is a vital tool in catching cheats, but misusing it can have devastating consequences for clean athletes. In the case of Paula Radcliffe, even if subsequent expert opinions clear her of wrong-doing (as I’m sure they will), some mud is sure to stick. People will remember the controversy now, far more than any subsequent lower profile ‘actually the samples look ok’ story. My hope is that her reputation is not tarnished, and that ABP testing continues to be refined, because there is no doubt that when used properly, it is a powerful weapon to catch cheats.

1. Sanchis-Gomar, F., et al., Current limitations of the Athlete’s Biological Passport use in sports. Clin Chem Lab Med, 2011. 49(9): p. 1413-5.
2. Sottas, P.-E., et al., The Athlete Biological Passport. Clinical Chemistry, 2011. 57(7): p. 969-976.
3. Schumacher, Y.O. and G. d’Onofrio, Scientific Expertise and the Athlete Biological Passport: 3 Years of Experience. Clinical Chemistry, 2012. 58(6): p. 979-985.
4. Bonne, T.C., et al., Altitude training causes haematological fluctuations with relevance for the Athlete Biological Passport. Drug Testing and Analysis, 2015. 7(8): p. 655-662.
5. Schumacher, Y.O., et al., High altitude, prolonged exercise, and the athlete biological passport. Drug Testing and Analysis, 2015. 7(1): p. 48-55.
6. Sottas, P.-E., et al., Prevalence of Blood Doping in Samples Collected from Elite Track and Field Athletes. Clinical Chemistry, 2011. 57(5): p. 762-769.
7. Sanchis-Gomar, F., et al., Altitude exposure in sports: the Athlete Biological Passport standpoint. Drug Testing and Analysis, 2014. 6(3): p. 190-193.

Posted in Uncategorized | Comments Off on In defence of Paula Radcliffe and her biological passport

Wild Animal Genomics Meeting – Trondheim

The plan is to get into a (fairly) regular habit of blogging about evolutionary genetics research (and perhaps a bit of running and biking stuff). Anyway, 1st post is about a meeting I recently attended in Trondheim.

The first Wild Animal Genomics meeting was run by Josephine Pemberton and I, using funds from our ERC grants. That took place in Scotland three years ago, and was aimed at bringing together researchers that were using, or planned to use, genomics tools to study quantitative genetics, selection and microevolution in wild populations. In June this year Henrik Jensen and Arild Husby put together a follow-up meeting, which was held just outside Trondheim. I really like these kinds of meetings – fairly small (~30-35 people) and everybody working on similar topics. The meetings are modelled on the WAMBAM meetings which have been running for around 10 years and have proven hugely successful in bringing together a community of researchers using ‘animal models’ to study quantitative genetics in pedigreed populations.

One of the most striking things at the Trondheim meeting was how quickly research in the field has moved on. Three years ago, we were only just starting GWAS analyses, virtually nobody had full genome assemblies, and a state-of-the-art SNP chip had 10,000 markers. There was a recognition that QTL mapping by linkage analysis was hugely under-powered, but a (in hindsight very naive) hope that association scans would get around this, and that many of the quantitative traits we were studying would harbour some loci of moderate-large effect that might explain measurable variation in fitness. At the recent meeting, there were examples where reseaerchers had sequenced whole genomes of >1000 individuals, and a number of presentations showing GWAS results with high density SNP chips or genotyping-by-sequencing methods. In general (and there were some exceptions) the attempts to find loci underpinning quantitative variation hadn’t found much. Lessons from studies of complex traits in humans suggest this isn’t suprising, although there are clearly some wild populations where patterns of high LD and recent admixture mean that some loci will be found.

So does this all mean that the genomic approach to understanding genetic variation in wild populations has failed? I don’t think so. If one regards ‘gene hunting’ as the main reason for doing this work, then sure, there will be some disappointed researchers. However, the most exciting thing about the recent advances in molecular quantitative genetics has been some of the cool research that has gone beyond a GWAS and given us genuine insight into the genetic architecture of quantitative traits. It isn’t necessary to find the loci that explain significant variation in order to perform analyses such as chromosome partitioning, genomic prediction, regional heritability mapping, and inbreeding coefficient estimation. Not only that, but these analyses do not require pedigrees and are robust to different kinds of genetic architecture. With that in mind, I still think these are exciting times for evolutionary genomics of wild populations. In the next few years we should expect to see research that describes the architecture of traits, identifies genetic changes in response to selection, gives us new insight into the mechanisms of inbreeding depression, helps us understand sexually dimorphic traits and no doubt addresses a whole load of other questions. Not only that, but we can start to do so in any population where phenotypes and DNA can be collected; the constraint of needing decades worth of pedigree data is no longer there, and so we should get broader taxonomic representation in this field. The biggest challenge will probably be in collecting good phenotypic data in large enough numbers to perform robust analyses. In other words, the most important researchers in the genomics era will still be the dedicated natural historians and ecologists that have underpinned so much of what we do today.

Posted in Research | Tagged | Comments Off on Wild Animal Genomics Meeting – Trondheim