AbstractSports authorities exclude athletes with abnormal levels of blood parameters. Here, the consideration of longitudinal blood profiles together with heterogeneous factors such as ethnicity and age produces a model with enhanced sensitivity to detect blood doping. Sports disciplines with heterogeneous populations now have a general method to introduce the no-start rule.
In sports, rHuEPO, the recombinant form of erythropoietin, is illicitly used to improve physical performance. Some sports federations with a fairly homogeneous population of athletes have discouraged rHuEPO doping by introducing hematocrit (Hct) and hemoglobin (Hb) limits. Athletes tested above these limits are declared unfit for competition: the so-called no-start rule. Unfortunately, since Hct and Hb present elevated between-subject variations1 and can easily be manipulated,2 the efficiency of both variables remains limited. Models using multiple hematological variables, such as Abnormal Blood Profile Score (ABPS),3 were proposed to improve sensitivity/specificity. Along with this, the idea of a hematological passport was also suggested.4 Sportsmen with significant differences between new test results and an individual historical baseline could be excluded from competition.1
Hematological variables depend on various factors such as gender, ethnicity and age. Even though the effects of these factors on hematological parameters have been well described,5 they have not been taken into consideration in the formulation of a blood model. This leads to an unpredictable number of false-positive findings for heterogeneous populations. Unsurprisingly, sports federations with highly heterogeneous populations refrained from introducing a no-start rule.
We, therefore, propose a blood test that combines a multiparametric approach for increased specificity, a hematological passport for individual longitudinal monitoring, and the formal integration of various factors for heterogeneous populations. The blood test is based on a global Bayesian inference approach for the detection of abnormal values over time.6 Hb and ABPS3 (with Hct, Hb, RBC count, reticulocytes percentage, mean corpuscular volume, mean corpuscular Hb, mean corpuscular Hb concentration variables) markers are used. Effects of gender [male, female], ethnicity [Caucasian, Asian, African, Oceanian], age [<19 years, 19–24 years, >24 years], altitude [<610, 610–1730, >1730], sport [endurance, non-endurance] on the mean of each parameter were taken from published data.5 Except for gender, the variance of blood parameters is considered independent of the factors and modeled as a log-normal distribution with parameters estimated from data collected on control subjects. Firstly, 135 blood profiles were collected from 22 top-level elite endurance athletes, 11 males and 11 females, participating in a study commissioned by the Swiss Federal Office for Sport to promote drug-free sports. Regular anti-doping tests were conducted on each athlete over a period of 2 years (6 blood and 11 urine samples in average) returning only negative test results. Secondly, 572 blood profiles were collected from 47 male amateur athletes over a period of two months (347 with an Advia 120, Bayer Diagnostics, Zurich, Switzerland, 225 with a XT-2000i analyzer, Sysmex, Norderstedt, Germany). Sensitivity was estimated from data collected in a two month rHuEPO clinical trial described elsewhere. 3,7 To summarized, 32 healthy volunteer males participated in this study. The aim was to reproduce rHuEPO doping habits practised in some sports. Eight subjects received subcutaneous injections of 40 IU/kg of EPREX three times a week (group R40), 8 subjects 80 IU/kg doses (group R80), 8 subjects received placebo (group P), and the last 8 subjects (group NT) had no treatment. rHuEPO administration in R40 and R80 groups was nevertheless individualized. A full dosage was administered for Hct below 45%, half doses for Hct between 45 and 50%, and substitution by isotonic saline when Hct exceeded 50%. Five hundred and ten blood samples (178 from doped subjects) were analyzed on a Cell-Dyn 4000 instrument (Abbott Diagnostics Division, Baar, Switzerland). In total, 1,039 samples collected in the three studies were used to estimate the specificity. For didactic purposes, we applied the Bayesian model to two subjects (Figure 1). The first subject is a top-level female Caucasian endurance athlete, 28 years old, living at low altitude. The second subject is a top-level male African endurance athlete, 36 years old, living and training occasionally at high altitude. Consideration of a new individual Hb value induces a new distribution of possible values of Hb, and, therefore, new individual reference range. The significant decrease between the first (a population-based threshold) and the last cut-off limit (an individual threshold) can be associated to a larger between-than within-subject variance of the hematological variable.1,6,8
All three aspects of the model, multiparametric approach, longitudinal analysis and consideration of heterogeneous factors, lead to a decrease of the overall variance of hematological data. As shown in Table 1, a heightened sensitivity to discontinuous rHuEPO treatment was observed. A population-based strategy on Hb gives only 3 true positives for a specificity of 0.999, whereas the Bayesian model returned as much as 11 times more true positives. With the ABPS marker, sensitivity was even further improved.
special recognition is due to the subjects who volunteered for this study
- Funding: this study was funded and supported by UCI (International Cycling Union), IAAF (International Association of Athletics Federations) and FOSPO (Swiss Federal Officefor Sport).
- Sharpe K, Ashenden MJ, Schumacher YO. A third generation approach to detect erythropoietin abuse in athletes. Haematologica. 2006; 91:356-63. PubMedGoogle Scholar
- Kuipers H, Brouwer T, Dubravcic-Simunjak S, Moran J, Mitchel D, Shobe J. Hemoglobin and hematocrit values after saline infusion and tourniquet. Int J Sports Med. 2005; 26:405-8. PubMedhttps://doi.org/10.1055/s-2004-821156Google Scholar
- Sottas PE, Robinson N, Giraud S, Taroni F, Kamber M, Mangin P. Statistical classification of abnormal blood profiles in athletes. International J Biostat. 2006; 2:1-21. Google Scholar
- Malcovati L, Pascutto C, Cazzola M. Hematologic passport for athletes competing in endurance sports: a feasibility study. Haematologica. 2003; 88:570-81. PubMedGoogle Scholar
- Sharpe K, Hopkins W, Emslie KR, Howe C, Trout GJ, Kazlauskas R. Development of reference ranges in elite athletes for markers of altered erythropoiesis. Haematologica. 2002; 87:1248-57. PubMedGoogle Scholar
- Sottas PE, Baume N, Saudan C, Schweizer C, Kamber M, Saugy M. Bayesian detection of abnormal values in longitudinal biomarkers with an application to T/E ratio. Biostatistics. 2007; 8:285-96. PubMedhttps://doi.org/10.1093/biostatistics/kxl009Google Scholar
- Robinson N, Saugy M, Buclin T, Gremion G, Mangin P. The interpretation of secondary blood markers can get hazardous in case of a discontinuous rhEPO treatment. Haematologica. 2002; 87:ELT28. PubMedGoogle Scholar
- Harris EK. Effects of intra- and interindividual variation on the appropriate use of normal ranges. Clin Chem. 1974; 20:1535-42. PubMedGoogle Scholar
- Schmidt W, Prommer N. The optimised CO-rebreathing method: a new tool to determine total haemoglobin mass routinely. Eur J Appl Physiol. 2005; 95:486-95. PubMedhttps://doi.org/10.1007/s00421-005-0050-3Google Scholar