Predictors of doping intentions, susceptibility, and behaviour of elite athletes: a meta-analytic review

TLDR: This critical review aims to summarize studies that identified predictors of doping intentions, susceptibility, and behaviour in elite athletes and to analyse in how far previous research included aspects beyond athlete-centred approaches, such as context and sporting culture.

Abstract: Research in doping has focused on potential intervention strategies, increasingly targeting predicting factors. Yet, findings are inconsistent, mostly athlete-centred and explain only limited variances in behaviour. This critical review aims to (a) summarize studies that identified predictors of doping intentions, susceptibility, and behaviour in elite athletes and to (b) analyse in how far previous research included aspects beyond athlete-centred approaches, such as context and sporting culture. We reviewed 14 studies that focused on elite athletes. Situational temptation, attitudes, and subjective norms seem to be strong predicting variables of doping intentions (r ≥ 0.50), but intention was no predictor for behaviour. Attitudes were a significant predictor for both, doping susceptibility (r = 0.47) and behaviour (r = 0.30). Most of the predictors are athlete-centred and ignore macro-level factors that might help to explain how certain individual traits impact on the decision making process. The findings from this review call for a critical discussion of whether current doping-prevention research needs to take new directions. We propose future research to bridge findings of psychologists and sociologists, as it appears that doping behaviour cannot be explained by ignoring the one or the other. Impacts of sporting culture that have been identified in qualitative approaches need to be integrated in future quantitative approaches to test for its external validity. Inclusion of both, micro- and macro level factors may enable an integrative prevention program that creates a sporting culture without doping.

Full text

Blank
et al. SpringerPlus (2016) 5:1333
DOI 10.1186/s40064-016-3000-0
REVIEW
Predictors ofdoping intentions,
susceptibility, andbehaviour ofelite athletes:
a meta-analytic review
Cornelia Blank
1*
, Martin Kopp
2
, Martin Niedermeier
2
, Martin Schnitzer
2
and Wolfgang Schobersberger
1,3
Abstract
Research in doping has focused on potential intervention strategies, increasingly targeting predicting factors. Yet,
findings are inconsistent, mostly athlete-centred and explain only limited variances in behaviour. This critical review
aims to (a) summarize studies that identified predictors of doping intentions, susceptibility, and behaviour in elite
athletes and to (b) analyse in how far previous research included aspects beyond athlete-centred approaches, such as
context and sporting culture. We reviewed 14 studies that focused on elite athletes. Situational temptation, attitudes,
and subjective norms seem to be strong predicting variables of doping intentions (r ≥ 0.50), but intention was no
predictor for behaviour. Attitudes were a significant predictor for both, doping susceptibility (r = 0.47) and behaviour
(r = 0.30). Most of the predictors are athlete-centred and ignore macro-level factors that might help to explain how
certain individual traits impact on the decision making process. The findings from this review call for a critical discus-
sion of whether current doping-prevention research needs to take new directions. We propose future research to
bridge findings of psychologists and sociologists, as it appears that doping behaviour cannot be explained by ignor-
ing the one or the other. Impacts of sporting culture that have been identified in qualitative approaches need to be
integrated in future quantitative approaches to test for its external validity. Inclusion of both, micro- and macro level
factors may enable an integrative prevention program that creates a sporting culture without doping.
Keywords: Doping prevention, Micro-level, Macro-level, Sporting culture
© 2016 The Author(s). This article is distributed under the terms of the Creative Commons Attribution 4.0 International License
(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium,
provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license,
and indicate if changes were made.
Background
Doping is generally considered to be unsportsmanlike
and believed to create unfair advantages while destroying
the values of sport. In view of the importance to protect
the athletes’ health and integrity, results of a previous
survey indicated that preventing doping in elite sport is
considered as the highest priority from 35 international
sporting federations (Mountjoy and Junge 2013). Nev
-
ertheless, Wagner and Pedersen (2014) have previously
reported that the general population does not trust the
doping management of international sporting federa
-
tions due to ongoing doping scandals despite the imple-
mentation of the World Anti-Doping Agency (WADA)
in 1999. Significant amounts of money have been spent
on (a) identifying new policies and measures to prevent
doping and (b) implementing these measures. However,
if the general population perceives this money to be
spent on inefficient prevention strategies, then their trust
may decrease even further. erefore, the identification
of processes that lead to doping behaviour should be of
interest to both science researchers and sport govern
-
ing bodies entrusted with doping prevention. Prevention
strategies that are grounded on transparently accumu
-
lated scientific evidence might help to reduce the above
mentioned lack of trust not only of the general popula
-
tion but also of the athletes themselves (Overbye 2016).
Deterrence andeducation
Increasing doping controls as part of a deterrence
approach represents one possibility to prevent dop
-
ing. Nevertheless, results from current research offer a
Open Access
*Correspondence: cornelia.blank@umit.at
1
Institute for Sports Medicine, Alpine Medicine & Health Tourism, UMIT,
Eduard-Wallnöfer-Zentrum 1, 6060 Hall in Tyrol, Austria
Full list of author information is available at the end of the article

Page 2 of 14
Blank
et al. SpringerPlus (2016) 5:1333
number of explanations as to why this approach does not
seem to be successful on its own. For example, as already
mentioned above, there is a lack of trust among athletes
in effective doping controls (Overbye 2016). Trust of the
athletes is crucial for bans to work as a deterrent, as it is
not the act of doping but the discovery of doping that is
punished (Petróczi and Haugen 2012). Highlighting the
negative health effects of doping has also been unsuccess
-
ful (Engelberg etal. 2015; Huybers and Mazanov 2012;
Probert and Leberman 2009; Schnell et al. 2014). e
only deterrents that appear to be partially effective are
social sanctions and humiliation (Huybers and Mazanov
2012; Overbye etal. 2014), which were both included in
the Australian Anti-Doping Agency’s prevention pro
-
gram called “You can never win your reputation back”
(Huybers and Mazanov 2012). Based on this evidence,
current doping prevention measures focus on more than
mere deterrence strategies. Additional education-based
prevention approaches (i.e., Goldberg’s ATLAS program)
(Goldberg etal. 2000) have been increasingly applied to
prevent negative behaviour before it occurs, especially
with respect to athletes’ health, integrity, and fairness,
as well as values in sports. Nonetheless, education in
the sense of transferring information does not appear
to be successful either. Previous studies have identified
only weak, if any, association between knowledge about
doping and its side effects, and doping intentions and/
or behaviour (Blank etal. 2014; Ntoumanis etal. 2014).
It appears that the effectiveness of the health message in
trying to prevent doping is questionable (Engelberg etal.
2015). As a result of the acknowledged complexity of the
doping phenomenon scientists suggest that only a firm
understanding of factors involved in doping as well as
their relationships will potentially result valid pro-social
interventions for doping (Johnson 2012). erefore, sub
-
sequent research that has focused on identifying reasons
for doping behaviour, has been inspired mainly, but not
exclusively, by research in the field of health- and social
psychology (Lazuras etal. 2015) and focusing on the indi
-
vidual athlete.
Psychological factors
Doping is considered to be a complex behaviour. A
recent meta-analysis by Ntoumanis et al. (2014) sum
-
marized and compared literally all known psychological
predictors of doping behaviour in all physical settings
and at all performance levels. Findings included vari
-
ables from the theory of planned behaviour (TPB) (Ajzen
and Madden 1986), additional attitude-behaviour rela
-
tions (Bentler and Speckart 1979), deterrence (Pater-
noster 1987) and self-determination theory (Ryan and
Deci 2000), as well as combinations of these variables
(Donovan etal. 2002; Strelan and Boeckmann 2003). In
addition, implicit measures were also included (Brand
etal. 2014; Petroczi etal. 2008). Even though this meta-
analysis provides extensive information for the scientific
and practical community, including all competition levels
might dampen the significance of the result for the elite
athletes. A number of studies included within this meta-
analysis involve recreational athletes who are not part of
sporting organizations that signed the WADA anti-dop
-
ing code and are therefore not directly confronted with
the offense of doping (Arandjelovic 2015). Additionally,
motivations for doping in these sports may be differ
-
ent from motivations of elite athletes who are training
to compete in major sporting events, such as the Olym
-
pic Games or World Championships (Bilard etal. 2011;
Elliot and Goldberg 1996; Wiefferink et al. 2008). e
Olympic Games are considered to be the most impor
-
tant event in an athlete’s life, and winning a medal at
the Olympics is the highest goal to which an athlete can
strive. Chester and Wojek (2015) have recently criticized
existing research in recreational athletes as not necessar
-
ily being representative for elite athletes. It is expected
that elite athletes face different situational pressures
within their daily training routine, and a previous study
has shown that situational factors mediate several predic
-
tors of doping behaviour (Barkoukis etal. 2013). In line,
only a few of the findings from Ntoumanis etal. (2014)
were observed in a recent qualitative study by Engelberg
et al. (2015), who analysed interviews with doped ath
-
letes. Some correlations observed in the meta-analysis of
Ntoumanis etal. (2014) were in the opposite direction of
correlations observed in Engelberg etal. (2015). Explana
-
tions for these diverse findings might be (a) the different
target populations, (b) difficulties of evaluating doping
behaviour (i.e. in most research athletes are asked to self-
report about their doping behaviour) and (c) different
methodological approaches that are hardly comparable.
Sociological approaches
Given the constant number of positive doping samples,
one could either speculate that despite the growing body
of research that helps understanding the underlying psy
-
chological processes of doping behaviour, the preven-
tive strategies seem to lack success or that the analytical
detection methods have improved. Considering the
first hypothesis, many researchers opened the debate to
shift from an athlete-centred approach to a much wider
approach. Especially in sociology this debate already has
an extensive history. Stewart and Smith (2008) have pro
-
posed a macro model that also includes sporting con-
text and—culture. ey further acknowledged that an
athlete’s decision to dope might not always be rational
and be influenced by a range of impacts. erefore, even
though beneficial, socio-psychological theories to explain

Page 3 of 14
Blank
et al. SpringerPlus (2016) 5:1333
doping behaviour might not always fit as they are mostly
based on rational and intentional decision making. It has
to be acknowledged that athletes are embedded in com
-
munity cultures and practices (Wagner 2010). Copeland
and Potwarka (2016) claimed that doping can be under
-
stood on a cognitive level of the individual, but further-
more that the contextual-organizational level must not
be ignored. Parts of this contextual-organizational level
are for example impacts of commercialization and glo
-
balization as well as sporting culture and the perception
of its own identity (Stewart and Smith 2008).
Summarizing, there already exists a rapidly growing
body of research aimed at explaining doping behaviour.
Nevertheless, findings are very heterogeneous and appear
to have a mainly athlete-centred focus which might be
due to several reasons outlined above. is heterogene
-
ous landscape of research findings renders the formula-
tion of clear prevention strategy difficult and there is the
need to summarize these findings to possibly identify
generalizable common predictors.
Aims andobjectives
Building on the comprehensive work of Ntoumanis etal.
(2014), this meta-analytic review has two major objec
-
tives. First, this review aggregates and interprets research
efforts towards the identification of predictors of (a)
doping behaviour, (b) intentions that are most proximal
to doping behaviour (Armitage and Conner 2001; Bilic
2005; Elliot et al. 2003; Godin and Kok 1996), and (c)
susceptibility to doping behaviour, which is commonly
used as a substitute for doping behaviour itself. Given
the expected differences between amateur and elite ath
-
letes that have been outlined above, only studies that
included applied multivariate analyses of elite athletes
competing at the national level or higher are included in
this review. Findings based on different empirical mod
-
els are reviewed and compared as appropriate. With this
approach, we aim to determine whether it is possible to
identify common predictors of doping intentions, suscep
-
tibility, and behaviour of elite athletes. More importantly,
we aim to critically discuss whether it is reasonable to
quantify overall effect sizes of these predictors due to dif
-
ferent operationalization methods of equal psychological
constructs.
e second major aim is to analyse whether or not pre
-
vious research, that allows meta-analytic calculations,
included aspects beyond the athlete-centred approach.
Findings and possible non-findings will be critically dis
-
cussed to provide both, a scientific data compilation that
might allow for the definition of preventive strategies
as well as starting points for future research to close a
potential gap between micro- and macro level oriented
research approaches.
Methods
Search methods foridentication ofarticles
A systematic literature research was performed to docu-
ment the findings of previous studies with respect to
predicting factors on doping, intentions, susceptibil
-
ity, and behaviour. Our search included original studies
published in scientific peer-reviewed journals between
1999 (the founding year of WADA) and January 2016
and indexed in the MEDLINE and/or EBSCO (including
SocIndex, Academic search elite, Business source pre
-
mier, Cinahl, Pre-Cinahl, Hospitality and Tourism index,
Inspec, PsychArticles, PsychInfo, SportsDiscus, Lista)
databases using the search terms “doping”, “performance
enhancing drugs”, “drugs AND sport” as well as by com
-
bining each of these terms with “determinants”, “corre-
lates”, “risk factors”, “precipitating factors”, and “model”.
is search term strategy was previously used by Back
-
house etal. (2007) for a final report to the WADA. Addi-
tionally, we manually searched the reference lists of every
primary study for additional publications.
Assessed outcomes
e main outcomes of interest were predicting factors
for doping susceptibility and doping behaviour; however,
as intentions have been said to be an important proxi
-
mal factor to behaviour (Armitage and Conner 2001;
Bilic 2005; Elliot etal. 2003; Godin and Kok 1996), we
included predicting factors for intentions as the third
outcome of interest. Outcomes were exclusively recorded
by self-reporting questionnaires and were displayed in
prevalence percentages and/or computed scores as the
results of regression analyses and/or structural equation
modelling.
Data extraction
Two researchers (CB, WS) independently performed
the literature research, quality assessment, and data
extraction. Any disagreements about inclusion of trials
were resolved by discussion with the three remaining
researchers (MK, MN, MS). According to a standardized
form, the investigators extracted data that was method
-
ologically and scientifically sound and collected the fol-
lowing variables: author and year of publication, journal
title, characteristics of the target population of the study
(including sample size and age), the dependent variable(s)
of the study (intentions, susceptibility, behaviour), the
included psychological concepts, and the outcome of the
study (tested model and/or individual predicting factors).
Inclusion andexclusion criteria
Due to the meta-analytic approach, the review was lim-
ited to studies that evaluated predicting factors with
respect to doping intentions, susceptibility, and/or

Page 4 of 14
Blank
et al. SpringerPlus (2016) 5:1333
behaviour of elite athletes competing on national level
and above. We excluded manuscripts in which the sam
-
ple was described with words such as “non-competitive”,
“amateur” and/or “competing at club level”. Studies with
focus on no matter which age groups, type of sports, and
country were included. Studies that were not aimed at
evaluating predictors with respect to intentions, suscep
-
tibility, and/or behaviour but rather included findings of
this kind as ancillary results were excluded. Studies that
focused on body building, gym-users and high-school/
college sports were excluded, especially given the evi
-
dence of different reasoning (i.e., body image) for taking
prohibited substances within these sports (Laure et al.
2004; Leifman etal. 2011). Studies that focused on ado
-
lescents participating in amateur level (i.e. high-school
sport) were excluded. Publications that reported results
from a mix of elite and non-elite athletes as per definition
of the authors were excluded if such results were undis
-
tinguishable. Studies not reporting Pearson’s correlation,
odds ratio or the mean standardized difference were not
included in the meta-analytical statistics but included in
the study. Finally, reports that described solely theoreti
-
cally developed models without empirical testing were
excluded. Figure 1 summarizes the study search and
inclusion process.
Excluded articles
e study search yielded a total of 1107 results, and 793
of these studies were rejected based on the title. A total
of 324 abstracts were screened, and the full text of 71 of
these studies including the bibliography was analysed.
In sum, 57 articles were rejected due to the target group
(31), no doping relation (7), no predictors (6), qualita
-
tive research (3), no report on Pearson’s correlation, odds
ratio or the mean standardized difference (4), or descrip
-
tion of theoretical models (6).
Included articles
In total, 14 studies were included in the analyses. Four
studies focused on doping intentions (Barkoukis et al.
2013, 2015; Lazuras et al. 2010, 2015), four studies
focused on doping susceptibility (Barkoukis etal. 2014;
Gucciardi etal. 2011; Hodge etal. 2013; Whitaker etal.
2014), six studies focused on doping behaviour (Dona
-
hue etal. 2006; Dunn and omas 2012; Jalleh etal. 2014;
Mazanov etal. 2008; Petroczi 2007; Uvacsek etal. 2011).
Quality analyses
Based on the nature of research in doping preven-
tion, most studies are either quantitative self-reporting
or qualitative interviews. Until now, barely any rand
-
omized controlled trials and/or experiments have been
performed to evaluate predictors of doping attitudes,
intentions, susceptibility, and/or behaviour. erefore,
applying methods such as the scale developed by Jadad
et al. (1996) for data quality analysis was not feasible;
however, we evaluated data in terms of their quality
based on sample size, response rate, reliability (Cron
-
bach’s alpha), and comparability of questionnaires.
Data synthesis
To structure the outcomes with respect to different psy-
chological concepts, we aligned the organization to psy-
chological constructs from the literature. erefore,
attitudes, subjective norms, perceived behavioural con
-
trol (PBC), and intentions were subsumed under the con-
cept of TPB (Ajzen 1991). (Non-) user favourability and
(non-) user similarity were subsumed under the construct
of prototype modelling used by Whitaker etal. (2014),
and additionally, we added descriptive norms to this con
-
struct (operationalized in all of the included studies as
the estimated prevalence of doping in others) (Barkoukis
etal. 2013, 2014, 2015; Lazuras etal. 2010, 2015; Uvac
-
sek etal. 2011; Whitaker etal. 2014). As suggested by the
theoretical drugs in sports deterrence model (DSDM)
(Strelan and Boeckmann 2003), which is based on the
deterrence theory, we subsumed situational temptation
(as mostly pressure from the outside), personal morality,
affordability, availability, legitimacy, threat, and benefit
appraisal under the construct of deterrence theory (DT).
Variables describing autonomous/intrinsic, controlled/
extrinsic, and amotivation as well as coach-controlled,
controlling teammate, autonomy-supportive coach, and
teammate climate were combined under the concept of
self-determination theory (SDT) (Ryan and Deci 2000).
Variables of sport motivation (win orientation, competi
-
tiveness, goal orientation, mastery avoidance/effort, per-
formance avoidance/ability, and performance approach/
external reasons) were aggregated under the construct of
achievement goal orientation/sport orientation (AGO)
(Gill and Deeter 1988). Sportspersonship was the term
used for the sportspersonship orientation scale, which
was developed by Vallerand et al. (1997) and includes
items such as respect for rules, opponents, and officials.
Any kind of moral operationalization was also subsumed
under sportspersonship. Experience was the term used
for the combination of knowledge, past and current
behaviour.
As effect sizes, Pearson’s correlation coefficients were
given for the examination of the relation of two continu
-
ous variables and odds ratios (OR) for the examination
of the relation of two dichotomous variables, respec
-
tively. Whenever different operationalization methods
were applied for the same concept in different studies,
i.e. both, dichotomous and continuous variables were
used for the same construct in different studies, the OR

Page 5 of 14
Blank
et al. SpringerPlus (2016) 5:1333
or the mean standardized difference was converted to
Pearson’s correlation to allow comparison of the studies.
is procedure was previously proposed by Borenstein
etal. (2011). According to Schmidt and Hunter (2015),
random-effect models in meta-analysis of correlations
are superior to fixed-effect models in terms of the accu
-
racy of the confidence intervals. erefore, meta-analytic
methods were applied according to the random-effect
method of Hedges and Olkin (1985), described in detail
by Field (2005).
e mean effect size for every predictor was calculated
using weighted Fisher’s r-to-Z and Z-to-r transformation
for Pearson’s correlations (Fisher 1921). Due to the small
number of included studies, this method is less biased
than other methods (e.g. Hunter & Schmidt method)
(Field 2005). Additionally, 95% confidence intervals were
calculated for the mean effect sizes, when possible. Pear
-
son’s correlations between 0 and 0.1 were considered
as small, 0.1 and 0.3 as medium and 0.3 and 0.5 as large
(Cohen 2013). e classification of OR was defined as
small (1.68–3.47), medium (3.47–6.71) and large (>6.71)
(Chen etal. 2010).
Results
Methodological quality ofincluded studies
Available Cronbach’s alpha ranged between 0.51 and
0.98. After discussion among the authors, no study was
excluded due to quality issues, even though Cronbachs-α
Fig. 1 Flow chart of search strategy