Between Februari 2000 and February 2009 the International Skating Union (ISU) detected elevated levels of reticulosytes in 14 out of 95 blood samples taken from Claudia Pechstein. Elevated levels indicate possible blood doping by EPO. Based on these results, the ISU suspended Pechstein on the 1st of July 2009 for a period of two years. Pechstein denies all allegations and accuses the ISU of gross negligence. She says “Samples were mislabeled and mixed up. Moreover, results of the same sample measured in two laboratories show big differences. A laboratory in Lausanne reported 1.32% Ret whereas a German laboratory in Kreischa reported 2.4%”.
Is Pechstein a victim of a false positive?
To answer this question we have to consider whether the cut-off value is right considering the precision of the method. Doping is concluded when a value is found above that cut-off value. Cut-off values are related to the normal range. This range is determined by two factors. The first one is the biological variation (if any) of the %Ret in blood within a person and between persons. The second factor is the variation of test results obtained in one and the same laboratory and in different laboratories. The variation between laboratories is caused by the usage of different methods, apparatus and qualified personnel. The Lausanne laboratory applies a cut-off value of 2.4%. However, by a comparison of the results obtained by two independent ISO 17025 accredited laboratories, they concluded that too many fluctuations should be avoided by carrying out the test under standardized pre-analytical conditions and with identical analyzers of the same manufacturer1. Translated into normal words, this means that this cut-off value is probably only applicable under restricted conditions of a single laboratory. Bessman2, who is not involved with doping tests but with regular clinical testing, reports that the reticulocyte percentage in duplicate samples varied up to 80% among five samples over two months. “Because some laboratories are more liberal than others about designating cells as reticulocytes,..... our range of normal values is 0.4 to 3%” says Bessman. This large variation is also mentioned in an interlaboratory study3 (310 samples, 8 laboratories, 11 instruments, 4-month period) reporting a possible difference between two duplicate samples (for the expert, I note that this is the value of the Reproducibility limit) up to 67% relative at a %Ret level equal to 2.5 and 193% at a 0.5% level (calculated from the CV in3).
What does this all mean for the Pechstein case?
If the German and Swiss laboratories that analyzed the same Pechstein blood sample used different methods and instruments, a relative difference of about 100% is possible between their results. Thus the seemly big difference between the Lausanne and Kreischa results is perfectly explainable just by the measurement error! Harm Kuipers, member of the Council for the Fight against doping of the ISU says ”Yes, small differences between laboratories are possible”. A real understatement!
From the data reported in ref(3) we derive a standard deviation (s) equal to 0.34 at the 0.5% level and 0.6 at the 2.5% level. The (upper) cut-off value is found at [true value +3s] (we choose the value 3s to minimize the probability of false positives – see my June post). Assuming that the true but unknown % Ret value equals the theoretical value for a healthy person, which is between 0.7 and 0.9%2 then a cut-off value for %Ret is obtained between [0.7 + 3*0.34] = 1.7 and [0.9 + 3*0.34] = 1.9 which is well below the cut-off value applied by the ISU. However, the overall mean %Ret obtained by measuring 1510 endurance sportsmen in 2001 was 1.3% (Robinson4), which is well above the theoretical value and can be considered as the true value. Assuming a test error s = 0.6 at that level, a cut-off value is obtained equal to [1.3 + 3*0.6] = 3.1.
So, Pechstein may have a difficult, but not impossible case to fight her suspension, although the %Ret = 3.4 found in one of the samples is definitely too high. This sample should be re-evaluated by a second laboratory in order to make sure that this value was not an outlier. Or, other parameters should be considered as well.
Finally, we note that the Council for the Fight against Doping of the International Cycling Union Cycliste Internationale (UCI) and the Lausanne laboratory co-authored several papers.
Disclaimer: Considering that my findings are based on a very preliminary study, the reported cut-off value should be considered a as rather crude estimation of the lower bound for the true but unknown cut-off value
1N. Robinson, L. Schattenberg, M. Zorzoli, M.Saugy, Haematological Analysis at the Departure of the Tour de France 2001, Int. J. Sports Med 2005 (26), 200-207
2 J. David Bessman in Clinical Methods, The History, Physical and Laboratory Examinations, 3rd edition, H. Kenneth Walker, W. Dallas Hall, J. Willis Hurst, Butterworths
3 B.H. Davis, N.C. Bigelow, J.A. Koepke, M.J. Borowitz, B. Houwen, J.W. Jacobberger, R.V. Pierre, L. Corash, K.A. Ault, J.D. Batjer, Flow Cytometric Reticulocyte Analysis – MultiInsttutional Interlaboratory Correlation Study, Am. J. of Clin. Pathology, 102 (4), 48-477 (1994)
4 N. Robinson, Y. Pfister, P. Mangin and M. Saugi, One year's experiment with the blood screening test to fight against rh-EPO doping, poster