News & Comment / In Brief

In this issue:
Tests for EPO Abuse

Has your Patient/Client/Subject Changed?
Limits to Performance

These aritcles reviewed by John A Hawley, RMIT University, Melbourne, Australia.



Will G Hopkins PhD, Physiology and Physical Education, University of Otago, Dunedin 9001, New Zealand. Email: Sportscience 4(2),, 2000 (1051 words)

    EPO (erythropoietin) is a hormone produced by the kidneys. It travels via the circulation to the bone marrow, where it keeps the supply of red cells ticking over. More EPO means more red cells, which boost endurance performance by transporting more oxygen to the muscles. A sojourn at real or simulated altitude is a safe and legal way for an athlete to get more EPO. Injections of EPO are even more effective: athletes can expect enhancements in endurance performance a massive 5% or more (Sawka et al., 1996; Birkeland et al., 2000). But injections of EPO can make the blood so thick with red cells that it clots throughout the circulation and kills the athlete. Amongst cyclists alone EPO abuse is thought to have caused 20 sudden deaths in recent years. Not surprisingly, EPO is on the IOC's list of banned substances.
    Unfortunately there has been no dependable and fair test for EPO abuse. The International Cycling Union (UCI) now tests for the thicker blood by measuring the proportion of red cells (the hematocrit or packed-cell volume) in a blood sample. By itself, this test is not a good indicator of EPO abuse, because a few athletes have a naturally high hematocrit, while others can get a high proportion from altitude training. A cyclist exceeding the upper limit is therefore not banned for EPO abuse, but is simply not permitted to compete because of the health risk. In any case, cyclists can cheat the test. When told they are to be tested, apparently they have 10 minutes to report to the medical team. Why 10 minutes? A cynical informant claims that's long enough for an athlete to run 500 ml of saline into a vein. By diluting the blood, the saline immediately brings the hematocrit down by a few percent. The normal hematocrit for "clean" elite cyclists is around 44% (Saris et al., 1998; Schumacher et al., 2000). So it's possible for a cyclist to take enough EPO to increase the hematocrit to around 52%, then infuse saline just before the test to bring the hematocrit back below the limit of 50% (or 51%, to allow for error of measurement). As a bonus, the saline infusion itself almost certainly enhances performance in long hot events like the Tour de France.
    You can't really blame the athletes for cheating, or the UCI for allowing the cheating to continue. Selfish behavior by athletes and their sporting bodies is driven by genes that evolution can't eliminate from the gene pool of social animals. Those of us who inherit the genes are driven to cheating when we figure the rewards outweigh the risks. Sure, but one of the rules of public competitive sport is that we must catch and punish the cheats. What to do in the case of EPO abuse?
    A better test would help. Right now there are two on offer: a urine test and a blood test. The urine test, which has just been published in Nature (Lasne and de Ceaurriz, 2000), is based on the technique of immunoblotting to directly detect the artificial "recombinant" EPO that the drug companies produce for treatment of patients. The authors of the test first showed that it worked on urine from patients taking recombinant EPO, then they turned their attention to frozen samples of urine from 102 cyclists in the 1998 Tour de France. A routine test for EPO revealed 28 positives. When the authors immunoblotted the 14 samples with the highest concentration, all were positive for recombinant EPO. If we assume at least half of the remaining 14 were positive, at least 20% of competitors abused EPO during or immediately before the Tour.
    Scientists at the Australian Institute of Sport have been working on the blood test for the last couple of years, and they're hoping the International Olympic Committee will adopt it for the games in Sydney. The test is based on detecting the effects of EPO on red cells, rather than EPO itself. The AIS team have found that a sensitive and specific indicator of EPO injections is an increase in the number of immature red cells (reticulocytes) in the blood. By analyzing the properties of these immature cells with state-of-the-art equipment, they can distinguish between athletes who have injected EPO and those who have had a natural increase in red cells from real or simulated exposure to altitude.
    How good are these tests? My guess--and it is a guess, because no-one at the AIS will comment--is that the urine will test positive only if the last injection of EPO was within a few days of the test. Any earlier and the EPO will have disappeared from the circulation and therefore from the urine. The blood test might detect an injection within the last couple of weeks, because that's about how long it takes the new red cells to mature. If the IOC decides to use either of these tests at the Sydney Olympics, athletes will simply stop injecting a week or so before arriving at the Games village. The ergogenic effect of a course of EPO injections lasts several months, because that's the lifetime of red cells in the circulation of athletes training hard. So the cheats will win again, but hopefully for the last time. The real value of these tests will be apparent when they are used for random testing between the Olympics. Athletes at Sydney will seem to be clean. Athletes at Athens really will be clean.

Sept 8: The blood and urine tests will both be used at the Sydney Olympics. Any athlete who tests positive in both tests will be disqualified. Some action may be taken against athletes who pass the urine test but fail the blood test.

Birkeland KI, Stray-Gundersen J, Hemmersbach P, Hallen J, Haug E, Bahr R (2000). Effect of rhEPO administration on serum levels of sTfR and cycling performance. Medicine and Science in Sports and Exercise 32, 1238-1243

Lasne F, de Ceaurriz J (2000). Recombinant erythropoietin in urine. Nature 405, 635


Saris W, Senden JMG, Brouns F (1998). What is a normal red-blood cell mass for professional cyclists? Lancet 352, 1758


Sawka MN, Joyner MJ, Miles DS, Robertson RJ, Spriet LL, Young AJ (1996). The use of blood doping as an ergogenic aid. Medicine and Science in Sports and Exercise 28(6), R1-R8


Schumaker YO, Grathwohl D, Barturen JM, Wollenweber M, Heinrich L, Schmid A, Huber G, Keul J (2000). Haemoglobin, haematocrit and red blood cell indices in elite cyclists. Are the control values for blood testing valid? International Journal of Sports Medicine 21, 380-385


See also previous articles on this topic by Stephen Seiler and Dave Martin under Blood Tests on our Sports Medicine index page.


Will G Hopkins PhD, Physiology and Physical Education, University of Otago, Dunedin 9001, New Zealand. Email: Sportscience 4(2),, 2000 (716 words)

    Any time you test or measure someone, the value you get has "noise" (random error) in it. If the noise is small, the test will reliably track any changes in the subject. But if the noise is too big, any change you see could be due to the noise rather than to any real change. So here's a really important question: how can you tell whether the change you see is real or noise?
    Some researchers have had an honest shot at this question, using a reference value for changes called 95% limits of agreement. According to these researchers, you can trust an observed change only if it's greater than the limits of agreement (Atkinson and Nevill, 1998). Sounds really cool, but there's a problem: the limits of agreement are so big that clinically important changes often fall within them. So if you use limits of agreement, you may have to ignore an important change in your subject. I made this point and others in a recent review of reliability in the journal Sports Medicine (Hopkins, 2000). I also wrote about assessing an individual on my stats pages and provided a spreadsheet to use when making decisions about change. Since then Atkinson and Nevill have written a letter to the editor, and the editor has invited me to respond. Their letter and my response may appear in the October issue of the journal.
    In writing the response, I've found that the best reference value for making decisions about change is the magnitude of the noise itself. I've previously called this noise the typical error. It's also known as the standard error of measurement, the technical error of measurement, and the within-subject standard deviation. Limits of agreement are about 3 times as big as the typical error, which explains why they're too big to use in clinical decision-making. I've also found that you have to keep one eye on the smallest "signal"--the smallest clinically important change in your subject. This element has been missing from previous publications on the topic of measurement error. I've put the noise and the signal together in the following advice for clinicians and other practitioners:


Use of Typical Error When Monitoring an Individual

  • Find the noise in your measure--the value of the typical error from a short-term reliability study of individuals similar to the one you are monitoring.
  • Decide on the smallest signal--the smallest clinically or practically worthwhile change in the measure for the individual.
  • If the noise is less than the smallest signal, you can trust observed changes of any magnitude between a single test and retest.
  • If the noise is greater than the smallest signal, you can trust expected changes greater than the noise, but you will need to do multiple tests before you can trust any changes less than the noise.


    Researchers, don't feel left out! I've come up with similar advice for you. Once again, the typical error and smallest clinically important change are the crucial elements:


Use of Typical Error in Research Design

  • Find the noise in your measure--the value of the typical error from a reliability study with individuals and a time frame similar to those of your intended study.
  • Decide on the smallest signal--the smallest clinically or practically worthwhile change in the measure for your study group.
  • If the noise is less than the smallest signal, you can use the measure to make precise estimates of any experimental effects with a single test and retest and a sample of modest size (<10 in a crossover; ><36 in a fully controlled trial). >
  • If the noise is greater than the smallest signal, the measure will provide acceptable precision for effects smaller than the noise only with more testing (more subjects, or more pre and post tests).


    I give the rationale for this advice in my response in Sports Medicine. I am also emphatic that limits of agreement should be abandoned as a clinical tool and marginalized as a measure of reliability.



Atkinson G, Nevill AM (1998). Statistical methods in assessing measurement error (reliability) in variables relevant to sports medicine. Sports Medicine 26, 217-238


Hopkins WG (2000). Measures of reliability in sports medicine and science. Sports Medicine 30, 1-15


Will G Hopkins PhD, Physiology and Physical Education, University of Otago, Dunedin 9001, New Zealand. Email: Sportscience 4(2),, 2000 (423 words)

    A writer (Kurt Kleiner) preparing an article for an Olympics issue of New Scientist contacted me recently about limits to world records. I flicked his inquiry and my comments to the Sportscience mailing list for further comment. The full text of the inquiry and the responses are linked here. Below is a summary for the contributors, with a link to their responses. See also the article by Kurt at the New Scientist website. In summary: some people say the running records are leveling off; others aren't so sure.
    Will Hopkins thought that the differences between training and nutrition programs accounted for only a few percent. He thought that more people competing would lead to better records for purely statistical reasons. He also suggested eugenics arising from breeding between top athletes will lead to genetically superior athletes. He thought (wrongly?) performances were plateauing now. There haven't been big gains associated with drug use, so he suggested that top athletes tend not to use the drugs (too much to lose), or that top athletes might get less benefit anyway.
    Stephen Seiler had data on progression of records. In the last 50 years, improvements per decade have been approximately: sprinting, 1%; distance running, 1.5%; jumping, 2-3%; pole vaulting, 5%; swimming, 5%; skiing, 10%. According to Stephen, the increases show no sign of leveling off, even for sprinting. [But female times have pretty-much leveled off: see Stephen's article on the gender gap.] Introduction of drug testing appears to have made no difference [for the males, but it may have stopped females getting faster]. Reasons for the increase: technology, especially for transportation sports; more extreme outliers with increased participation; interbreeding of athletes; individualized training; drugs? For endurance events there is headroom in theory for ~10% improvement, if one individual had current best values of maximum oxygen consumption, anaerobic threshold, and exercise efficiency.
    Michael Green wondered if we would need to measure performances to an extra decimal place, as performances plateau.
    Frank Katch called our attention to a recent article in Scientific American, in which the authors suggested that manipulation of gene expression in muscles could produce superfast sprinters.
    Dan Wagman expressed regret that gene doping might stop us from ever knowing what our natural physiological limits are. He wondered whether sport psychologists will need to specialize in sport performance dehancement strategies, to help stop gene-doped athletes giving the game away.
    Will Hopkins commented on the paper in Scientific American. He noted that blood volume was more important than proportion of muscle-fiber type as a determinant of endurance performance. He was critical of the parallel the authors drew between detraining in previous sedentaries and tapering in elite sprinters. He was doubtful that turning on the expression of superfast Type IIb myosin in human muscles would lead to enhanced sprint performance, especially if top sprinters already have an optimum mix of Type IIa and IIx myosin.

Published Sept 2000