Memorandum from GH—2004
DETECTING GROWTH
HORMONE ABUSE
Executive Summary
— GH-2004 is a research study funded
by the World Anti Doping Agency (WADA) and the United States Anti
Doping Agency (USADA). It "stands on the shoulders of"
GH-2000, a multi-national research project funded by the European
Union (EU) and the International Olympic Committee IOC), that
developed the "marker method" for detecting growth hormone
(GH) abuse. The "marker" method of detecting GH abuse
was developed using rigorous scientific methods.
— GH-2004 aims to validate this methodology
across a spectrum of ethnic groups and sporting scenarios. Further
validation has also been undertaken using data collected by other
unrelated but collaborating research groups:
— It has been shown that the GH-2000
discriminant functions work as well in detecting those taking
GH in an unrelated German GH administration study as it does on
the original UK data.
— It has been shown that the GH-2000
discriminant function did not generate any false positives when
applied to more that 1,000 blood samples from a variety of elite
athletes from many ethnic groups in Europe, Australia and Japan.
— Sporting injuries do not lead to
false positive results.
— The "marker" test will
detect GH abuse in a wide range of ethnic groups.
— Variation between commercial assays
for the markers is a significant problem.
Proposal
— UK Sport leads the world by initiating
a study of the operational viability of this test of GH abuse
in the run up to the London Olympic Games.
— UK Sport leads the world to accumulate
experience of testing and re-testing volunteer athletes to allow
them to demonstrate that their "passports are clean"
so far as growth hormone is concerned.
— We undertake further research to
develop more suitable assays for the markers IGF-I and P-III-P.
INTRODUCTION
1.1 Over the last two decades, there has
been accumulating evidence that growth hormone (GH) is widely
abused by athletes for its anabolic and lipolytic properties.
Its use was banned by the International Olympic Committee (IOC)
in 1989 and GH appears on the World Anti Doping Agency (WADA)
list of prohibited substances. The detection of exogenously administered
GH poses a formidable challenge, as it is almost identical to
that which is produced naturally in the body. A major multi-national
European research project—GH-2000, led by Professor Peter
Sonksen of St Thomas' Hospital, London and funded jointly by the
European Union (EU) and the IOC proposed a detection method based
on the measurement of two GH-dependent blood markers, insulin-like
growth factor—I (IGF-I) and type 3 pro-collagen (P-III-P),
both of which rise in response to growth hormone administration
in a dose dependent manner. The changes in these markers during
a placebo controlled double blind GH administration study were
used to construct gender specific formulae that gave good discrimination
between those taking GH and those taking placebo.
1.2 The results of this study were presented
to the IOC at a workshop in Rome in March 1999. The conclusion
of the workshop was that although the results presented were of
considerable interest and importance, further validation studies
needed to be undertaken before the test would be robust enough
to withstand the forensic examination that would be needed to
win a case in the Court of Arbitration in Sport (CAS). The particular
validation studies considered of most importance were:
1. Confirmation that there were no significant
racial or ethnic effects that might influence the interpretation
of results.
2. Confirmation that the effects of injury
would not influence the interpretation of results.
1.3 Funding for a further multi-national
research project necessary to answer these points—GH-2004—was
sought from the EU but although reaching the final stage it was
not financed on the basis of cost. Meanwhile, the IOC under its
new President Jacques Rogge was re-organising the structure of
its Doping Laboratories through the creation of "The World
Anti-Doping Agency—WADA". Under the new structure the
IOC no longer funded research and this was taken over by WADA.
The proposal for GH-2004 was submitted to WADA who responded that
they were interested in funding a "cut down" version
of the project based wholly in the UK (where they considered that
there was sufficient ethnic diversity to enable the research).
1.4 Meanwhile the United States Anti Doping
Agency (USADA) also had a call for proposals for research on methods
for the detection of GH abuse. Professor Sonksen had retired from
his post at St Thomas' Hospital (to live near Winchester, Hampshire)
and wished to continue (and hopefully complete) this research
and for this he sought the help of Dr Richard Holt (Senior Lecturer
in the Medical School at Southampton) to submit a joint application
for funding. Dr Holt had previously been a Lecturer with Professor
Sonksen in London.
1.5 A research proposal for GH-2004 was
then submitted to both WADA and USADA from the Medical School
at Southampton with Professor Sonksen and Dr Holt as joint "Principal
Investigators". Approval for a grant to cover most of the
project was approved by USADA on 24 December 2002, three days
before a skiing accident that left Professor Sonksen severely
disabled with a spinal cord injury. Dr Holt took de facto control
of the project although Professor Sonksen was able to re-join
the project after leaving hospital in the summer of 2003. WADA
later agreed to contribute a further trenche to complete the necessary
finance.
1.6 The details of this project are given
at: http://www.gh2004.soton.ac.uk/. The project base at Southampton
University was recently reviewed by Richard Caborn MP (Minister
of Sport).
1.7 The project is now nearing completion,
all the studies have been finished and the thousands of blood
samples collected have been analysed by Professor David Cowan's
team at The UK IOC/WADA-Accredited Laboratory at King's College,
London. The results are currently being analysed by the team statisticians
and the results will shortly be submitted to leading scientific
journals for peer-reviewed publication. Some preliminary results
have been presented in a recent closed USAD/WADA workshop in the
USA.
1.8 In summary, these results strongly supported
the conclusions previously reported by the GH-2000 group and showed
no evidence of a significant "ethnic" effect that might
interfere with the global use of the test. The preliminary results
also showed no important "injury" effect that might
interfere with the test.
1.9 Meanwhile a number of papers from independent
research groups have been published that contribute usefully to
the validation of this approach (now commonly referred to as the
"marker method"). The most important of these came from
the IOC/WADA accredited laboratory at Kreischa in Germany.
1.10 Investigators at the Kreischa laboratory
undertook a further placebo controlled double blind GH administration
study and developed an alternative formula based on IGF-I and
P-III-P but also including IGF-I Binding Protein 3—IGFBP3.
1.11 The statistical procedure used to generate
the discriminant functions by both research groups (GH-2000 and
Kreischa), involved splitting the data collected into two; a "training"
set of data to calculate the discriminant function and a "confirmatory"
set for validating the sensitivity and specificity of the discriminant
function and determines whether the function is successful in
discriminating between the treatment groups. The confirmatory
set is required in order to ensure the model is applicable to
the general population and not just the "training" set.
There is a valid criticism of this approach that sees it as a
potentially "self-fulfilling" method. Ideally, further
validation is needed using completely independent data sets to
evaluate whether the GH discriminant function formulae perform
reliably on completely different sets of data.
1.12 We have investigated the validity of
the GH-dependent marker approach by assessing whether the GH-2000
discriminant function could be used reliably to detect those receiving
GH when applied to the Kreischa dataset.
2. RESULTS
2.1 Kreischa Study
The blood samples had been analysed for the
GH-sensitive markers using different laboratory methods to those
used in GH-2000 and GH-2004 and the results were not directly
comparable without manipulation. We developed a method for adjustment
based on "normalising" the data using the results from
the untreated study populations. This then allowed direct comparison
of results.
2.2 Performance of GH-2000 Discriminant
Function on "their own" and "the others" laboratory
results: Success rates and estimated average sensitivity were
calculated using a specificity ("false positive" rate)
of approximately of 1 in 10,000 with a cut-off point of 3.7s.d.
Sensitivity was calculated as number of observations placed over
the 3.7 cut-off value out of the total number of observations
for the particular day from volunteers receiving GH treatment.
Results in bold are from days during GH administration.
| Discriminant Function
| Data | Day 3
| Day 5 | Day 7
| Day 9 | Day 11
| Day 13 | Day 16
| Day 19 |
| GH-2000 | Kreischa
| 1/10 (10%) | 2/10 (20%)
| 6/10 (60%) | 7/10 (70%)
| 6/10 (60%) | 9/10 (90%)
| 5/10 (50%) | 5/10 (50%)
|
| Discriminant Function
| Data | Day 21
| Day 28 | Day 30
| Day 33 | Day 42
| Day 84 |
| GH-2000 | GH-2000
| 23/28 (82%) | 24/28 (86%)
| 13/24 (54%) | 7/27 (26%)
| 2/28 (7%) | 0/27 (0%)
|
2.3 Thus it can be seen that the GH-2000 Discriminant
Function behaves equally well on the Kreischa (90% detected on
day 13) as on their own data (>80% on days 21 and 28 of GH
administration). The dose of GH administered in the German study
was lower than the lowest in the GH-2000 trial and much lower
than that believed to be used by athletes. Because of this the
actual sensitivity of the test in practice may well be better
than that in the trials.
2.4 This is a most important validation of the GH-2000
discriminant function showing it to have the ability to detect
up to 90% of men taking GH and up to 50% of men who stopped taking
GH as long as five days ago (Kreischa Day 19) with a false positive
rate set at 1:10,000.
2.5 The GH-2004 Study
Only preliminary results are available now but to illustrate
the comparability of results of the measurement of marker levels
in the blood of volunteer elite athletes from different ethnic
groups, Figure 1 shows the (assay adjusted) IGF-I levels against
age for volunteers from the different ethnic groups. The blue
line shows the 99% confidence intervals (CI) for the Caucasian
volunteers. It is clear firstly that the results depend on the
age of the athlete and secondly, that the results different ethnic
groups scatter randomly across the results from the Caucasians.
2.6 Figure 1
2.7 Figure 2
2.8 It can be seen from Figure 2 that the results of
(assay adjusted) P-III-P values from elite volunteers again show
a strong age-dependence but do not differ between ethnic groups.
The three volunteers with values slightly above the 99% CI of
Caucasians are not the same as those above the 99% CI for IGF-I
shown in Figure 1 (With >1,000 results up to 5 values would
be expected to exceed the upper 99% CI by chance alone).
2.9 The discriminant function developed by GH-2000 uses
both IGF-I and P-III-P and has to have a value above 3.7 to suggest
GH abuse. It was not above this value in any of the volunteer
athletes. Of course, we do not know that all these athletes were
"clean" but the results do suggest that this was the
case.
2.10 Collaboration with Professor Ken Ho in Sydney, Australia
allowed us a further validation using results he obtained from
more than 800 blood samples from a variety of ethnic groups.
— After correction for assay, virtually all of the
Australian samples lie within the 99% prediction intervals for
the white subjects from the GH-2000 study.
— Application of the male and female discriminant
functions to the Australian subjects showed that no individual
would have been falsely accused of doping.
— The age correction applied in the GH-2000 study
results in a small over-correction in the Australian study. The
reason for this is unclear but probably reflects differences in
assays.
2.11 Preliminary analysis of the double blind placebo
controlled studies of GH administration to volunteers from different
ethnic groups shows a similar response in all groups. This indicates
that this "marker" approach should be able to detect
GH abuse in all ethnic groups.
2.12 Preliminary analysis of the "Injury" study
showed no person in whom the discriminant function exceeded the
3.7 cut-off value that indicates GH abuse.
CONCLUSIONS
— The "marker" method of detecting GH
abuse has been developed using rigorous scientific methods.
— Validation of this method has been undertaken
using data collected by other unrelated research groups.
— It has been shown that the GH-2000 discriminant
functions work as well in detecting those taking GH in an unrelated
German GH administration study as it does on the original UK data.
— It has been shown that the GH-2000 discriminant
function did not generate any false positives when applied to
more that 1,000 blood samples from a variety of elite athletes
from many ethnic groups.
— Sporting injuries do not lead to false positive
results.
— The "marker" test will detect GH abuse
in a wide range of ethnic groups.
— Variation between commercial assays for the markers
is a significant problem.
PROPOSAL
— UK Sport leads the world by initiating a study
of the operational viability of this test of GH abuse in the run
up to the London Olympic Games.
— UK Sport leads the world to accumulate experience
of testing and re-testing volunteer athletes to allow them to
demonstrate that their "passports are clean" so far
as growth hormone is concerned.
— We undertake further research to develop more
suitable assays for the markers IGF-I and P-III-P.
May 2006
| 
