With the release of erythropoietin (EPO) for medical application in the late 1980s, illegal blood doping practices have become a major concern in endurance sport. These practices aim to increase oxygen transport, which consequently increase oxygen uptake and endurance performance (6). Because total hemoglobin mass (Hbmass) is a major determinant of oxygen transport (32), the goal of any blood doping practice is to increase the total amount of hemoglobin in the blood (17).
Although there are methods available to detect illegal homologous blood transfusions (19), there is currently no test for autologous transfusions. It has recently been suggested that measures of Hbmass may have merit in detecting autologous blood transfusions (5,21,23). Effective screening of Hbmass for antidoping purposes requires a test that is accurate and reliable and sensitive enough to detect fluctuations in Hbmass of ∼7%–12% that could occur with abuse of EPO or transfusion of 1–2 U of red cells (20,21,23). The optimized carbon monoxide (CO) rebreathing method (30) has been shown to be an accurate and reliable method of measuring Hbmass, with reported typical error values for measures taken a few days apart ranging from 1.1% (9) to 2.2% (10). It has recently been shown that the optimized CO method has the sensitivity to detect changes in Hbmass induced by blood withdrawal and autologous reinfusion of 1–2 U of packed red cells (21).
Models such as the athlete’s biological passport (16,36) and the Abnormal Blood Profile Score (35) use indirect markers of altered erythropoiesis such as hemoglobin concentration and percentage of reticulocytes to detect abnormal changes arising from blood doping. These and other blood doping models (26) adopt a Bayesian approach using longitudinal blood profiles along with other factors such as ethnicity, age, and gender to produce a model with good sensitivity to detect blood doping (26,36). An athlete’s test result is compared to their individual baseline with probability-based deviations from baseline providing evidence of doping (36). These models rely on the assessment of both the within-subject and between-subject variations of the specific markers (36). As the number of measurements on an individual subject increases, there is a greater reliance on the within-subject variation and less on the between-subject variation. Furthermore, the natural effects of accelerated erythropoiesis at altitude may confound both within-subject and between-subject variations in Hbmass (25,32).
If measures of Hbmass are to be used for antidoping purposes, it is important that there is minimal biological variation in Hbmass. Eastwood et al. (5) reported a short-term coefficient of variation (CV) of 2.1% for Hbmass for measures taken a few days apart in recreationally active males undertaking consistent regular training for 100 d. Similarly, elite female cyclists showed a CV of 3.3% in Hbmass for measures taken ∼1 month apart during a 6-month competitive season (8). In a similar study, Prommer et al. (23) showed that the maximum individual oscillation in Hbmass during a 12-month period was 6.9% for 24 athletes. In contrast, however, a 14% reduction in Hbmass was observed in a female cyclist during a period of immobilization, which was subsequently rectified on return to training (33). Therefore, it is important to quantify the effect of a reduction in training load on Hbmass if this measure is to be used for antidoping purposes.
The primary aim of this study therefore was to quantify the within-subject variation of Hbmass in elite athletes during training for a period of approximately 1 yr. The secondary aims were to investigate the effects of altitude exposure and periods of reduced training on measures of Hbmass and to determine whether Hbmass might be incorporated into antidoping models to detect abnormalities in blood profiles resulting from illegal blood doping practices.
A total of 130 subjects participated in the study (43 females and 87 males). Subjects were elite (international standard) or semielite (national standard) athletes in the sports of rowing, swimming, running, cycling, kayaking, and Australian rules football. The characteristics (mean ± SD) of the subjects at the start of the study were age 25.5 ± 5.6 yr for males and 24.3 ± 5.2 yr for females and body mass 75.8 ± 13.1 kg for males and 58.1 ± 7.1 kg for females. All procedures were approved by the Australian Institute of Sport and Flinders University Ethics Committees, and subjects gave written informed consent before any testing.
Subjects were measured for Hbmass at either the Australian Institute of Sport (AIS) in Canberra (n = 82) or the South Australian Sports Institute (SASI) in Adelaide (n = 48). To minimize between-operator error, all measures on an individual subject were conducted by the same researcher, using the same equipment. Measures were taken during a 3-yr period, the average number of measurements taken on a subject was 6.3 (range = 3–20) and the average period of the measures was 333 d (range = 29–1186 d). The time interval between measures was not standardized.
Hbmass was assessed using a modified version of the CO rebreathing procedure first described by Schmidt and Prommer (30). This modified version has been described in detail by Prommer and Schmidt (22). Briefly, this procedure comprised inhalation of a bolus of 99.5% chemically pure CO (BOC Gases, Sydney, Australia) in a dose of 1 mL·kg−1 of body mass and was rebreathed for 2 min. Arterialized capillary blood samples (200 μL) were taken from a prewarmed fingertip and analyzed in quintuplicate for percent carboxyhemoglobin (%HbCO) using a diode array spectrophotometer (OSM3 Hemoximeter; Radiometer, Copenhagen, Denmark) before as well as 6 and 8 min after commencing the rebreathing. After the 2-min rebreathing period, the volume of CO not taken up by the body was calculated as the product of the rebreathing bag volume and the concentration of CO in it, which was measured using a handheld analyzer with parts-per-million sensitivity (Fluke model 220; Mississauga, Canada). Seven minutes after commencing rebreathing, end-tidal CO concentration was measured with the same CO analyzer and multiplied by the estimated alveolar ventilation (5.25 L·min−1) (39) to account for the CO exhaled after disconnecting from the spirometer. The total Hbmass was calculated according to the methods described by Schmidt and Prommer (30).
On each visit to the laboratory, athletes were required to complete a questionnaire regarding their training since the last measurement of Hbmass. The questionnaire requested training hours, any reduction in training, illness and injury, and altitude exposure including time and type (live high–train high (classic) or live high–train low (LHTL)). Classic altitude exposure ranged from 1350 to 2850 m, whereas all LHTL exposures were at 3000 m.
Using the statistical package R1 (Version 2.10.0; R Foundation for Statistical Consulting) (24), linear mixed models were fitted with log(Hbmass) as the response variable, sport (six levels), sex, institute (AIS vs SASI), body mass and altitude as fixed effects, and athlete as a random effect. Altitude was considered as a factor with different levels based on the type (classic vs LHTL) and time spent at altitude. Sea level was the reference level of altitude (0) with six other levels: classic altitude for ≤7 d; classic altitude for 14–21 d; 7 d after classic altitude; LHTL for ≤7 d; LHTL for 10–42 d; and after LHTL for 7–14 d. To simplify interpretation, estimates of the fixed effects, all of which were obtained on the log scale, were converted to a percentage change in Hbmass.
In addition, the fitted models included an allowance for within-subject autocorrelation; that is, the expectation that the magnitude of between-pair differences in Hbmass changes with the number of days between measurements. To allow for the potentially separate effects of analytical and biological variation of Hbmass, we used the exponential (auto) correlation structure (37):
where d is the time (d) between two measurements of Hbmass, nugget reflects the magnitude of time-independent analytical variation, and the range determines the rate at which autocorrelation associated with within-subject biological variation decreases (to 0) as the time increases between measurements of Hbmass: the larger the range, the slower the rate of decline.
If SDwithin denotes the total within-subject SD of log(Hbmass), including both analytical and within-subject biological components for measurements taken a long time apart (>∼4 months), then SDwithin√ (nugget) can be interpreted as the analytical SD of log(Hbmass) (SDanalytical). SDanalytical is analogous to the typical error where measurements are taken a few days apart and the biological error is minimal. The long-term biological SD, after allowing for analytical variation, is given by √ (SDwithin2 – SDanalytical2), and the SD of within-subject changes over d days, which includes analytical variation, is given by
This SD can be used for determining whether a change between two readings is unusually large.
The 95% and 99.9% prediction limits for a change in Hbmass were calculated assuming a normal distribution for the within-subject changes in log (Hbmass). The variability that could be observed by either 5% or 0.1% of readings, for two readings taken a long time apart (>∼4 months), was calculated as:
with different values of SDwithin for males and females. The upper and lower values of the prediction limits were calculated as exp(z-score × √2SDwithin) and 1/exp(z-score × √2SDwithin) (37).
A second analysis was conducted to assess the suitability of the assumed autocorrelation structure used for part 1. After allowing for the effects of reduced training and altitude found in part 1, we calculated all possible pairwise differences of log(Hbmass) for each athlete, partitioned them into 14 groups according to the number of days between measurements (Table 1), and then compared the observed SD of the differences in these groups with those predicted by equation 2.
Hbmass of males was on average one-third higher than that of females even after the highly significant effect of body mass was taken into account, and the Hbmass values of athletes measured at SASI were ∼13% lower than those measured at the AIS (Table 2). On average, reduced training (35 observations) resulted in a significant decrease in Hbmass of ∼3%. Also in Table 2, we see that the effects of altitude on Hbmass were similar for classic (42 observations) and LHTL (152 observations) altitude training. There was little evidence of any change in Hbmass for exposure up to 7 d “at altitude” and a significant increase of ∼2% after 14–21 d of classic altitude as well as for 10–42 d of LHTL. However, the percentage increase in Hbmass less than 1 wk after return from classic altitude training (2.9%) was nearly twice that approximately 2 wk after completion of LHTL (1.5%).
The parameter estimates for the raw random effects are shown in Table 3 and expressed as percentage CV in Table 4 for the within-subject components of variability. The within-subject CV for males was smaller than the corresponding estimates for females: 3.4% for males versus 4.0% for females (P = 0.02; Table 4). The analytical CV was estimated to be ∼2.0% for both males and females (Table 4), whereas the long-term biological CV, after allowing for analytical variation, was estimated to be 2.8% (√(3.42–1.92)) for males and 3.5% (√(4.02–2.02)) for females. The SDchange(d→∞) (Table 3) represents the “long-term” SD of changes in log(Hbmass); that is, when the number of days between measures of Hbmass approaches infinity. From the estimates of SDchange(d→∞) (Table 3), 95% prediction limits for long-term changes in Hbmass were calculated to be +9.9% to −9.0% (Hbmass ×/÷ 1.098) for males and +11.7% to −10.5% (×/÷ 1.117) for females; the corresponding 99.9% prediction limits were +17.1% to −14.6% (×/÷ 1.171) and +20.5% to −17.0% (×/÷ 1.205), respectively. The assumption of normality, on which the prediction intervals are based, was assessed by applying the Shapiro–Wilk test to the normalized residuals from the linear mixed model and found to be reasonable, P = 0.20. From equation 1, and using the parameter estimates given in Table 3, the autocorrelation is estimated to decrease to the negligible level of 0.1 after 113 d for males and 89 d for females.
The comparison of observed SD of within-subject changes in log(Hbmass) and those values obtained from linear mixed modeling are shown in Figure 1. Subjectively, the agreement between the observed and predicted SD was close for up to 150 d, but not as satisfactory thereafter, particularly for females. The observed SD reached a peak at around 150 d and then decreased for both males and females.
This study aimed to establish estimates of within-subject variation for Hbmass, which might be incorporated into antidoping models to detect abnormal changes in an individual athlete’s Hbmass arising from illegal blood doping practices. The finding of substantial autocorrelation for measures of Hbmass up to ∼100 d apart for both males and females has important implications. Either the timing of measurements of Hbmass would need to be incorporated into the antidoping models, which would make the models quite complex, or, alternatively, a conservative single value of within-subject variability would be needed for antidoping purposes. The obvious conservative, single-value estimates of within-subject variation of Hbmass derived from the linear mixed modeling of this study are the CV values of 3.4% and 4.0% for males and females, respectively.
The within-subject CV in Hbmass of ∼3%–4% for both male and female elite athletes shown in this study support claims that Hbmass is relatively stable during training (8,23) compared with changes of up to 12%, which may be seen with blood doping (20,21) suggesting it may have merit for antidoping purposes. However, assuming a normal distribution of within-subject changes in log(Hbmass), based on a within-subject CV of 3.4% for males and 4.0% for females, 5% of pairs of readings on males would show random increases of at least 9.9% and random decreases of at least 9.0% and 5% of pairs of readings on females would show changes of at least +11.7% or at least −10.5% by chance alone for measures taken months apart. A study by Sharpe et al. (34) proposed two thresholds that could be used to target or exclude athletes based on the variation in certain blood markers. A 1-in-100 cutoff was recommended as suspicious, which would require follow-up testing or investigation; the second cutoff, 1-in-1000, was recommended as a no-start. Based on the results of the current study, 1-in-1000 pairs of readings on males would show changes greater than ∼17% and although those for females would be greater than ∼20% by chance alone for measures taken months apart. Therefore, authorities choosing to use Hbmass to detect blood doping would need to enforce conservative limits of ∼17% and 20% for males and females, respectively, based on 99.9% prediction limits to ensure minimal likelihood of false-positive results. These limits would be too large for Hbmass alone to be of much use for antidoping purposes considering that autologous reinfusion of two units of red cells or abuse of EPO has been shown to increase Hbmass by ∼7%–12% (20,21).
In addition to the changes in Hbmass that arise due to chance alone, the influence of factors such as reduced training and altitude exposure on Hbmass must also be considered. Periods of reduced training were found to decrease Hbmass by ∼3% and exposure to altitude caused increases in Hbmass ranging from 1.5% to 2.9% depending on the type of altitude (classic vs LHTL) and the length of exposure. If statistical methods, such as the athlete biological passport (36), are to be used to monitor longitudinal changes in Hbmass in elite athletes as an indicator of blood doping, the confounding effects of detraining and altitude must also be incorporated into the model.
Males versus females
In the present study, the Hbmass of males was ∼33% higher than that of females when the effect of body mass was taken into account. Allowing for uncertainty of measurements, this result is not too dissimilar from to that of Schmidt and Prommer (31), who reported that erythrocyte volume, estimated from Hbmass, was 20% higher in males than in females when normalized for body mass. The effect of testosterone on erythrocyte production (13) is likely responsible for the differences between males and females when body mass is taken into account. However, independent of sex, our results also suggest that total body mass, which, in lean athletes, is strongly related to fat-free mass, is also closely associated with Hbmass; for instance, a 10% difference in body mass would be associated with a ∼4.9% difference in Hbmass.
The 4.0% within-subject CV for female athletes in this study support previous findings of Garvican et al. (8), who reported a CV of 3.3% in Hbmass in elite female cyclists during a competitive season of ∼6 months. Garvican et al. (8), however, made no allowance for autocorrelation, and therefore, their CV values could be expected to be smaller than those reported in this study. The within-subject CV for males was lower than for females: 3.4% versus 4.0%, respectively. It may be suggested that the greater variation in the Hbmass of females could be due to the loss of Hbmass during the menstrual cycle. It has been shown that normal menstrual blood loss is around 35 mL and ranges from 10 to 80 mL (7). A heavy flow (80 mL) would result in a loss of 12 g of Hbmass, assuming a [Hb] of 15 g·dL−1. Assuming a Hbmass of 680 g (the average for a female in our study), 12 g would represent 1.7% of their Hbmass. However, it is also likely that the difference in the CV of Hbmass of males and females is merely due to sampling variation considering the overlapping confidence limits (Table 4).
Our results show a 2.8% decrease in Hbmass with reduced training supporting previous claims that Hbmass is sensitive to changes in training load (8). However, the period of reduced training was not measured in this study, and therefore, a fully quantitative effect of a set amount of reduced training on Hbmass remains unclear from the present results. Despite this, the fact that a period of reduced training has the potential to decrease Hbmass means it must be considered in antidoping models.
The results of the present study show classic altitude exposure had more effect on Hbmass than on LHTL. This is consistent with the concept that the extent to which erythropoiesis is accelerated in response to hypoxia is dose dependent (15), and furthermore, it is easier to accumulate more hours in natural altitude than in a synthetic altitude environment located near sea level. Measures taken 7 d after classic altitude were ∼3% higher compared to 1.5% after LHTL. The results of the present study show that either type of altitude needs to be in excess of 7 d for a noticeable effect on Hbmass, although previous reports suggest that 3 wk of moderate altitude exposure >2000 m is required for a response (27). Clark et al. (4) suggest that an athlete may increase Hbmass by ∼1% per hundred hours of hypoxia (LHTL), so that after 7 d (168 h) of classic altitude, the extent of increase is difficult to discern because of the time taken to produce new red blood cells (14) and the imprecision of the method used to assess total hemoglobin (10). It has been shown that a combination of classic altitude exposure and LHTL for 21 d increased Hbmass by 7%; however, there were also substantial differences between individuals (29). LHTL at 2500 m for 24 d was found to increase Hbmass by 5.3% in elite orienteers (38); similarly, Robertson et al. (25) showed that Hbmass increased by ∼3% in elite runners after 3 wk of LHTL at 3000 m for 14 h·d−1, with individual athletes experiencing larger changes. The results of the present study show that, on average, Hbmass was increased by 1.5% after LHTL for 10–42 d, which is considerably lower than that on previous reports (25,29,38). It is possible that any LHTL exposures less than 21 d in the present study resulted in a relatively small Hbmass response, as it has been shown that 12 d of LHTL at 2650 m was not sufficient to cause an increase in Hbmass in elite cyclists (2). Therefore, the 1.5% change reported in this study may underestimate the magnitude of the Hbmass response to LHTL altitude, particularly in some individual athletes (25). The results of the present study also show that the effect of altitude on Hbmass persists for at least 7 d and possibly 14 d after altitude, although data beyond that time were not available to include in our analysis. The fact that altitude has a significant and substantial effect on Hbmass implies that it has the potential to increase within-subject variation in Hbmass and therefore must be considered in antidoping models.
Differences between institutes
The Hbmass scores of the athletes measured at the AIS were, on average, ∼13% higher than those measured at SASI. This is likely due to a greater proportion of elite endurance athletes being tested at the AIS, whereas several team sport athletes of lower aerobic fitness comprised the SASI cohort. The present study also showed that there was a significant difference in the Hbmass scores between sports. The AIS athletes were from the sports of running, cycling, and swimming compared to rowing, kayaking, and Australian rules football at SASI. It has been shown that endurance-trained athletes have a greater Hbmass than strength- or power-trained athletes (12), which may account for some of the differences between sports and institutes in the present study. In addition, although each athlete was tested by the same researcher using the same equipment, different OSM3 analyzers were used in the different institutes, which may also account for some of the difference in the Hbmass scores between institutes (11).
Application for antidoping
The results of the present study indicated that the SD between athletes was about double the within-athlete SD for both males and females. Our findings support previous claims that reference ranges for Hbmass based on population estimates of between-athlete SD are of limited use for antidoping purposes (36) and highlight the importance of using within-subject SD such as those established by the present study as a reference in antidoping models (36).
This is the first study to monitor longitudinal changes in Hbmass in a large group of elite and semielite athletes. Previous longitudinal studies on Hbmass in elite athletes were conducted with smaller numbers of athletes (23), which is a limitation for antidoping research. In the present study, analytical variation was found to be ∼2%. This estimate of measurement error is consistent with previous studies, which have reported typical error values ranging from 1.1% (9) to 2.2% (10). It is important to note that the typical error estimate of 2.2% is that derived for measures taken about 1 d apart when biological variation should be minimal; therefore, the typical error reported in these studies is analogous to the analytical CV reported in the present study. Both the typical error data and our analytical error support previous claims that the optimized CO rebreathing method provides an accurate and reliable measure of Hbmass (10,30). Biological variation for ∼4 months, independent of analytical variation, was estimated to be 2.9% for males and 3.5% for females in the current study. Collectively, these findings support previous claims that measures of Hbmass are relatively stable in elite athletes undertaking intensive training (23).
However, the within-subject CV of ∼4% found in this study for ∼4 months suggests that measures of Hbmass alone are of little use for antidoping purposes. For antidoping purposes, it is the change between two measures of Hbmass taken possibly months apart that is relevant. Both the first and second measures of Hbmass will have error, so that the appropriate CV values are 4.8% (3.4√2) for men and 5.7% (4.0√2) for women. That is, in an antidoping context, it is appropriate to be conservative to minimize the likelihood of false-positive results and to consider the maximum CV of change in Hbmass and the associated prediction limits. The findings of this study suggest that, for measures of Hbmass taken several months apart, the change limits of ∼17% for males and ∼20% for females would need to be adopted by antidoping authorities to reduce the likelihood of false-positive readings to 1-in-1000. The maximum variability of change in Hbmass was found to occur around 4 months in the present study; at this time, the observed values of differences in Hbmass were closely approximated by the asymptotic value given by the linear mixed model (0.048 for males and 0.057 for females; Fig. 1). For males, the agreement between observed and modeled differences was good, but for females, the observed SD of pairwise differences were more erratic with some relatively extreme values; however, these extreme values were obtained from quite small samples (n ≤ 15; Table 1).
Modifications to the CO method could improve its accuracy and minimize the analytical error associated with measurements resulting in a lower within-subject variation, which could potentially be incorporated into antidoping models. For example, the typical error of the CO method may be reduced by increasing the dose of CO (3) and ensuring at least five measurements of HbCO at each time point (1). Nevertheless, even if the analytical error were reduced to just 1% the contribution of biological variation independent of analytical variation would remain at ∼2.8% for males and 3.5% for females. Assuming these analytic and biological errors to be independent, the long-term within-subject CV would be 3.0% and 3.6% for males and females, respectively, and the associated 1-in-1000 change limits would be +15.0% or −13.0% and +18.2% or −15.4%. With an analytical error of just 1%, and based on a 12% increase in Hbmass, which is possible after infusion of 2 U of blood or abuse of EPO (20,21), the probability of exceeding the 15% cutoff limit for males is estimated to be ∼0.27. If, based on the findings of the current study, the analytical error was indeed 2%, the probability of exceeding the associated 17% cutoff limit for males is estimated to be ∼0.18, which suggests that Hbmass has limited value to detect 1–2 U of blood doping or abuse of EPO. Thus, it seems that Hbmass is not likely to be used as a stand-alone measure to detect blood doping. Instead, its greatest value may be when used in combination with other measures such as percentage of reticulocytes (18). However, the efficacy of such a model has also been challenged (28).
A major limitation of the present study is that there were no objective measures of the athletes training load. Therefore, although we show an average 2.8% decrease in Hbmass with reduced training consistent with findings of others (8), it is not possible to fully quantify the effects of a period of reduced training on Hbmass. Moreover, although it was assumed that the athletes were not doping, not all athletes were drug tested during the study, and there were no records of which athletes were tested and when; therefore, the possibility of doping cannot be fully excluded. If some athletes were doping during the study, the estimates of within-subject variation reported in this study would potentially be larger than if no athletes were doping.
In elite athletes undertaking regular training, the (maximal) within-subject CV for Hbmass is ∼4%. In addition, factors such as altitude training and periods of reduced training have the potential to increase or decrease an individual’s Hbmass score by up to 3%. At a 1-in-1000 probability, an athlete may demonstrate within-subject changes of Hbmass of up to ∼20% by chance alone for measures taken a few months apart, suggesting that stand-alone measures of Hbmass may be of limited use for antidoping purposes. Possibly, the most likely application of Hbmass in antidoping is in combination with other makers of blood manipulation.
This study was funded by the World Anti-Doping Agency (research grant no. 05A5FS).
The authors have no conflicts of interest.
The results of the present study do not constitute endorsement by the American College of Sports Medicine.
1. Alexander AC, Garvican LA, Burge CM, Clark SA, Plowman JS, Gore CJ. Standardising analysis of carbon monoxide rebreathing for application in anti-doping. J Sci Med Sport. 2011; 14: 100–5.
2. Ashenden MJ, Gore CJ, Martin DT, Dobson GP, Hahn AG. Effects of a 12-day “live high, train low” camp on reticulocyte production and haemoglobin mass in elite female road cyclists. Eur J Appl Physiol. 1999; 80: 472–8.
3. Burge CM, Skinner SL. Determination of hemoglobin mass and blood volume with CO: evaluation and application of a method. J Appl Physiol. 1995; 79: 623–31.
4. Clark SA, Quod MJ, Clark MA, Martin DT, Saunders PU, Gore CJ. Time course of haemoglobin mass during 21 days live high:train low simulated altitude. Eur J Appl Physiol. 2009; 106: 399–406.
5. Eastwood A, Hopkins WG, Bourdon PC, Withers RT, Gore CJ. Stability of hemoglobin mass over 100 days in active men. J Appl Physiol. 2008; 104: 982–5.
6. Ekblom B, Goldbarg AN, Gullbring B. Response to exercise after blood loss and reinfusion. J Appl Physiol. 1972; 33: 175–80.
7. Fraser IS, Warner P, Marantos PA. Estimating menstrual blood loss in women with normal and excessive menstrual fluid volume. Obstet Gynecol. 2001; 98: 806–14.
8. Garvican LA, Martin DT, McDonald W, Gore CJ. Seasonal variation of haemoglobin mass in internationally competitive female road cyclists. Eur J Appl Physiol. 2010; 109: 221–31.
9. Gore CJ, Bourdon PC, Woolford SM, Ostler SM, Eastwood A, Scroop GC. Time and sample site dependency of the optimized CO-rebreathing method. Med Sci Sports Exerc. 2006; 38 (6): 1187–93.
10. Gore CJ, Hopkins WG, Burge CM. Errors of measurement for blood volume parameters: a meta-analysis. J Appl Physiol. 2005; 99: 1745–58.
11. Gough CE, Sharpe K, Ashenden MJ, et al.. Quality control technique to reduce variability of longitudinal measurement of haemoglobin mass. Scand J Med Sci Sports. 2011; 21 (6): e365–71.
12. Heinicke K, Wolfarth B, Winchenbach P, et al.. Blood volume and hemoglobin mass in elite athletes of different disciplines. Int J Sports Med. 2001; 22: 504–12.
13. Hero M, Wickman S, Hanhijarvi R, Siimes MA, Dunkel L. Pubertal upregulation of erythropoiesis in boys is determined primarily by androgen. J Pediatr. 2005; 146: 245–52.
14. Jelkmann W. Erythropoietin: structure, control of production, and function. Physiol Rev. 1992; 72: 449–89.
15. Levine BD, Stray-Gundersen J. Dose–response of altitude training: how much altitude is enough? Adv Exp Med Biol. 2006; 588: 233–47.
16. Malcovati L, Pascutto C, Cazzola M. Hematologic passport for athletes competing in endurance sports: a feasibility study. Haematologica. 2003; 88: 570–81.
17. Morkeberg J, Belhage B, Ashenden M, et al.. Screening for autologous blood transfusions. Int J Sports Med. 2009; 30: 285–92.
18. Morkeberg J, Sharpe K, Belhage B, et al.. Detecting autologous blood transfusions: a comparison of three passport approaches and four blood markers. Scand J Med Sci Sports. 2011; 21: 235–43.
19. Nelson M, Popp H, Sharpe K, Ashenden M. Proof of homologous blood transfusion through quantification of blood group antigens. Haematologica. 2003; 88: 1284–95.
20. Parisotto R, Gore CJ, Emslie KR, et al.. A novel method utilising markers of altered erythropoiesis for the detection of recombinant human erythropoietin abuse in athletes. Haematologica. 2000; 85: 564–72.
21. Pottgiesser T, Umhau M, Ahlgrim C, Ruthardt S, Roecker K, Schumacher YO. Hb mass measurement suitable to screen for illicit autologous blood transfusions. Med Sci Sports Exerc. 2007; 39 (10): 1748–56.
22. Prommer N, Schmidt W. Loss of CO from the intravascular bed and its impact on the optimised CO-rebreathing method. Eur J Appl Physiol. 2007; 100: 383–91.
23. Prommer N, Sottas PE, Schoch C, Schumacher YO, Schmidt W. Total hemoglobin mass—a new parameter to detect blood doping? Med Sci Sports Exerc. 2008; 40 (12): 2112–8.
24. R Development Core Team. A Language and Environment for Statistical Computing
. Vienna (Austria): R Foundation for Statistical Computing; 2010 [cited 2009 November 9]. Available from: http://www.R-project.org/
25. Robertson EY, Saunders PU, Pyne DB, Gore CJ, Anson JM. Effectiveness of intermittent training in hypoxia combined with live high/train low. Eur J Appl Physiol. 2010; 110: 379–87.
26. Robinson N, Sottas PE, Mangin P, Saugy M. Bayesian detection of abnormal hematological values to introduce a no-start rule for heterogeneous populations of athletes. Haematologica. 2007; 92: 1143–4.
27. Rusko HK, Tikkanen HO, Peltonen JE. Altitude and endurance training. J Sports Sci. 2004; 22: 928–44.
28. Sanchis-Gomar F, Martinez-Bello VE, Gomez-Cabrera MC, Vina J. The hybrid algorithm (Hbmr) to fight against blood doping in sports. Scand J Med Sci Sports. 2010; 20: 789–90.
29. Saunders PU, Pyne DB, Gore CJ. Endurance training at altitude. High Alt Med Biol. 2009; 10: 135–48.
30. 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.
31. Schmidt W, Prommer N. Effects of various training modalities on blood volume. Scand J Med Sci Sports. 2008; 18: 57–69.
32. Schmidt W, Prommer N. Impact of alterations in total hemoglobin mass on V˙O2max
. Exerc Sport Sci Rev. 2010; 38 (2): 68–75.
33. Schumacher YO, Ahlgrim C, Ruthardt S, Pottgiesser T. Hemoglobin mass in an elite endurance athlete before, during, and after injury-related immobility. Clin J Sport Med. 2008; 18: 172–3.
34. Sharpe K, Ashenden MJ, Schumacher YO. A third generation approach to detect erythropoietin abuse in athletes. Haematologica. 2006; 91: 356–63.
35. 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.
36. Sottas PE, Robinson N, Saugy M. The athlete’s biological passport and indirect markers of blood doping. In: Thieme D, Hemmersbach P, editors. Doping in Sports, Handbook of Experimental Pharmacology. Berlin (Germany): Springer-Verlag; 2010: 305–26.
37. Venebles WN, Ripley BD. Modern Applied Statistics with S-PLUS. New York (NY): Springer Verlag; 1999. p. 441.
38. Wehrlin JP, Zuest P, Hallen J, Marti B. Live high–train low for 24 days increases hemoglobin mass and red cell volume in elite endurance athletes. J Appl Physiol. 2006; 100: 1938–45.
39. West JB. Respiratory Physiology—The Essentials. Baltimore (MD): Lippincott Williams and Wilkins; 1995. p. 15.