Harnessing the global expertise in drug use and drug prevention in physical activity settings: results from the Anabolic Steroid Prevention Survey

Abstract Recent decades have seen increased efforts internationally to prevent the use of anabolic androgenic steroids (AAS) and other image and performance-enhancing drugs (IPEDs) in gyms and fitness environments. Yet, very little is known about effective prevention strategies. This study aimed to identify key risk factors for AAS use and assess the relevance of these risk factors as targets of intervention. Seventy four (n = 74) IPED experts participated in the Anabolic Steroid Prevention Survey (response rate: 62.4%). A total of 18 psychosocial and two behavioral risk factors identified in a literature review were rated by participants along two dimensions: importance and preventability. The results show that most IPED experts (91%) believe that preventing AAS use in gyms is important to public health, and that AAS use can be prevented to a certain degree (91%), but not eliminated altogether (85%). Based on participants’ assessment, six risk factors were categorised as very promising (e.g. the descriptive norm and poor knowledge on AAS alternatives), 10 as promising (e.g. body dissatisfaction and drive for muscularity), and four as unpromising but worthy of consideration (e.g. AAS-using peers and perceived benefits of AAS use). To effectively prevent AAS use in gyms, interventions should attempt to reduce these risk factors.


Background
Each year substantial human and financial resources are directed at efforts to prevent and reduce the non-medical use of anabolic androgenic steroids (AAS) and other image and performance-enhancing drugs (IPEDs). Whereas such efforts have previously been associated exclusively with antidoping in elite sport, recent decades have seen a growing public, political, and scientific interest in doping prevention in non-competitive, recreational sport. Several countries, most notably in Scandinavia, have instigated national and local prevention campaigns to address recreational athletes' use of AAS and other doping substances in parallel to their anti-doping programme in elite sport Christiansen et al., 2020;EHFA, 2012;Johannisson et al., 2012;UK Anti-Doping, 2019). One setting in particular has attracted considerably more attention than others in relation to AAS, namely the strength training environment that can be found in private health clubs, fitness centers, and gymnasiums (henceforth: gyms). This is hardly surprising since ordinary gym members, predominantly young men who lift weights to enhance muscularity, outnumber elite athletes and most other user groups when it comes to AAS usage (Sagoe, Molde, et al., 2014).
Given the secrecy surrounding the practice, it is methodologically challenging to obtain accurate prevalence estimates of AAS use based on self-report data (de Hon et al., 2015). In addition, the study quality and characteristics vary considerably between surveys of gym members. This is reflected in the great variation in prevalence estimates, which oscillate between 0.4 and 13.5% in studies conducted in mainstream gym facilities, but with figures as high as 70% in studies of 'hardcore' bodybuilding clubs (Bergsgard et al., 1996;EHFA, 2012;Kartakoullis et al., 2008;Lenehan et al., 1996;Molero et al., 2017;Pedersen, 2010;Sagoe, Molde, et al., 2014;Stubbe et al., 2014;Vogels et al., 1996;Wiefferink et al., 2008). Despite the paucity of experimental studies, there is mounting evidence to suggest that AAS use is detrimental to the user's health and wellbeing. Most AAS users self-report one or more adverse effects that they attribute to their AAS usage such as hypertension, testicular atrohpy, and mood swings (Begley et al., 2017;Bonnecaze et al., 2020;Ip et al., 2011;Parkinson & Evans, 2006). This is consistent with findings from recent clinical and epidemiological studies which have documented the long-term impact of AAS use on cardiovascular health, reproductive function, and mental health (Baggish et al., 2017;Christou et al., 2017;Horwitz et al., 2019;Rasmussen et al., 2016Rasmussen et al., , 2018Thiblin et al., 2015). Adding to this are a range of potential AAS-related harms such as those associated with polydrug use, blood-borne viral infections such as HIV and hepatitis C, 1 and the putative AAS-induced increases in aggression, violence, and criminal behavior (Christoffersen et al., 2019;Dodge & Hoagland, 2011;Hauger et al., 2021;Hope et al., 2013Hope et al., , 2021Lundholm et al., 2015;Sagoe et al., 2015). Thus, an effective prevention strategy holds promise to limit the societal and public health impact of AAS use.

AAS interventions
The vast majority of AAS interventions are anti-doping and drug education programmes delivered to adolescents and young adults in high schools, universities, and sports clubs. This is paradoxical given that the largest population of AAS users are regular gym goers (Sagoe, Molde et al. 2014), and because the content and delivery methods of these interventions may not apply to gyms due to differences in, for example, population characteristics, the availability and qualifications of intervention providers, and the implementation context. Moreover, most published interventions typically address only a subset of risk factors for AAS use, predominantly at the individual level, such as athletes' knowledge, beliefs, attitudes, and skills. Consequently, critical social-contextual factors are left out of the equation such as the physical and cultural environments in which AAS use takes place as well as the expectations and behavior of significant others (Bates, Begley, et al., 2019;Codella et al., 2019;Gebert et al., 2017;Hurst et al., 2020;Kavussanu et al., 2020;Lucidi et al., 2017;Marcello et al., 1989;Medina et al., 2019;Molero et al., 2016;Murofushi et al., 2018;Nicholls et al., 2020;Ntoumanis et al., 2021;Wicki et al., 2018;Yager et al., 2019).
In 2014, an expert group conducted a study for the European Commission to map existing approaches to doping prevention in recreational sport across the EU member states. Another aim of the study was to develop an evidence base to inform these efforts . This work was followed up and expanded in 2019 to identify changes in the prevention landscape, and to report on good prevention practice. Interestingly, most of the participating national anti-doping agencies reported that they regard doping prevention in recreational sport as important, and almost half of them reported to have examples of good practice. However, very little is known about the impact of these efforts due to the lack of robust evaluations (Bates & Vinther, 2021;Christiansen et al., 2020). Furthermore, out of 29 doping projects funded under the Erasmusþ programme, which all had doping prevention as their primary aim and a shared budget totalling 7,944,402 EUR, none of the projects provided any specific recommendations to guide the development of prevention strategies. 2 In sum, there is remarkably little evidence to inform interventions to reduce the use of AAS and other IPEDs despite three decades of research and prevention efforts in this area.

Study aims and research questions
Behavior change interventions are more likely to be effective if they are based on a thorough understanding of the behavior and its underlying causes as well as realistic expectations of the extent to which the behavior is malleable. Epidemiologically speaking, this means targeting the full spectrum of relevant risk factors. That is, variables which (1) can explain why some individuals, but not others, adopt the behavior, and (2) can be modified (Bartholomew et al., 2006). Several studies of both quantitative and qualitative nature have summarized risk factors for AAS use and examined their explanatory power, but with little or no consideration of their modifiability Blank et al., 2016;Cafri et al., 2005;Nicholls et al., 2017;Ntoumanis et al., 2014;Tavares et al., 2019). The present study was designed to address this gap by (1) identifying key risk factors for AAS use in young men, and (2) assessing the importance of these risk factors and the extent to which they can be reduced in the context of gyms (defined here as preventability). Since AAS use is predominantly a male phenomenon, the study focused exclusively on men's use of AAS (Pope et al., 2014;). An additional aim of the study was to explore whether consensus exists in the international community of IPED experts regarding the importance, scope, and limits of AAS use prevention. The following research questions were developed to address the study aims: RQ1: What are the most important risk factors for young men's use of AAS? RQ2: Which of these risk factors are feasible as targetes of intervention in the context of gyms? RQ3: To what extent do experts agree on the importance, scope, and limits of AAS use prevention in gyms?

Participant recruitment
One hundred and twenty five (n ¼ 125) IPED experts were invited to complete The Anabolic Steroid Prevention Survey. Two types of experts were considered eligible for inclusion in the study: (1) researchers studying the use of IPEDs in elite, amateur, and recreational sport, and (2) practitioners responsible for coordinating or delivering IPED interventions with a particular focus on AAS. Participants were recruited strategically to obtain a sample with the greatest possible representation of IPED experts with specific knowledge on AAS use and prevention. Hence, an individual assessment was made in each case based on the person's publication record, work experience, personal qualifications, etc.
Participants were recruited in several ways. First, invitations were sent to a sample of IPED experts (n ¼ 46) that participated as interviewees in a qualitative study conducted as part of a larger research project. Second, additional IPED experts were identified through systematic searches in member lists of two international academic networks: The Human Enhancement Drugs Network (HEDN) and the International Network of Doping Research (INDR). Together, these networks represent a large proportion of the international community of IPED and doping researchers. Third, potential candidates were identified from author lists and bibliographies of key publications on IPED and AAS use. Fourth, the author's personal network was searched to identify further participants. Finally, the snowball sampling technique was applied by asking survey participants to identify and encourage IPED experts in their own network to take part in the study.

Data collection
An Internet-based survey was conducted between November 2019 and March 2020 3 using the online survey management tool SurveyXact# (Rambøll Management Consulting, Denmark). The survey was administered to invitees by e-mail in which they were given a brief description of the study aims. The e-mail also informed invitees about the estimated completion time and emphasised the anonymity of their responses. To maximise the response rate, reminders were sent to nonresponders twice during the study period (in January 2020 and February 2020). Invitees who accessed the survey link were redirected to an introduction page with more detailed information about the study, including a description of the working definition of prevention adopted in the survey and their rights as research participants. Participants were informed that proceeding with the survey would indicate their active, informed consent. After submitting their responses, participants were given the opportunity to contact the principal investigator for questions related to the study.

Survey instrument
A 50-item, self-report questionnaire was developed for the purpose of this study (see Supplementary file 1). The questionnaire is divided into four parts: (1) assessment of demographic and personal characteristics, (2) assessment of attitudes and beliefs regarding the prevention of AAS use in gyms, (3) a risk factor rating task, and (4) a validity assessment. In part one, respondents were requested to provide their age, gender, and level of expertise (i.e. years of professional experience with AAS use and/or prevention). In part two, respondents were asked to indicate their level of agreement with four different statements concerning AAS use prevention in gyms (example item: 'Preventing steroid use in gyms is important to public health'). Each statement was rated using a 6-point Likert scale with response options ranging from Strongly disagree (1) to Strongly agree (6).
In part three, respondents were instructed to imagine a situation in which they were being commissioned to develop a public health intervention to prevent the initiation of AAS use by young men training in gyms. This was followed by an explanation of the concept 'risk factor' and the role risk factors play in the design of public health interventions. Respondents were presented with a list of risk factors for AAS use. They were then instructed to rate each risk factor according to the following criteria: (1) importance and (2) preventability. In brief, respondents rated all risk factors using a 4-point Likert-type scale with response options ranging from Not at all important (1) to Very important (4). They were guided by the question: 'How important is it to reduce the following risk factors in order to prevent steroid use? Imagine a young, male gym member who … ', followed by a description of each risk factor in a non-technical language. For example, the risk factor 'descriptive norm' was described as: ' … believes that steroid use is more prevalent than it actually is'. In a second round, respondents rated the same risk factors using a different 4-point, Likert-type scale with response options ranging from Very unlikely (1) to Very likely (4). Here, respondents were asked: 'How likely is it that the following risk factors, from a prevention perspective, can be reduced? Imagine a young, male gym member who … [description of risk factor]'. Thus, each risk factor was rated twice; once for importance and once for preventability.
In the final part, respondents were asked to indicate their level of agreement with three statements concerning how well they understood the instructions for the rating task, the confidence they had in their own judgement, and whether they believed the 20 risk factors cover the most important aspects of AAS use (example item: 'I fully understood the instructions for the rating task'). These statements were scored on the same scale as in part one. Respondents were also given the opportunity to suggest additional risk factors and to provide feedback and comments.

Scale properties
Scales with no midpoint (e.g. 'neither/nor') were employed in the questionnaire to elicit forced responses because research has shown that inclusion of midpoints in attitudinal and similar scales may undermine the validity of the scale. First, if not clearly defined, verbal midpoint anchors such as 'neither/nor' may be interpreted in ways not intended by the investigator. Second, midpoints tend to be overused by individuals who select this option to avoid negative responses in the belief that doing so will present them in a more favourable light, thereby increasing social desirability bias. Third, some individuals may choose the midpoint category as a 'quick' and 'easy' option instead of making a decision (Nadler et al., 2015).
Likewise, undecided choice options (e.g. 'no opinion' or 'don't know') were omitted from the scales because they may introduce bias by incorrectly dragging respondents from a definite to an undecided position. Such response options may be useful when survey participants have limited knowledge on the subject matter as it allows them to skip items to which they feel incapable of responding (Nadler et al., 2015). However, because participants in the present study were experts in the research topic, the benefits of obtaining a more complete dataset outweighed the potential bias associated with forced responses and the inconvenience that some respondents may have experienced.
The decision to employ scales with varying sensitivity across the examined domains (6-point scales for agreement and 4-point scales for importance and preventability) was based on the assumption that the items in part two and four lend themselves to more detailed responses than the items in part three. Importantly, both scales fall within the range of response options (4-7) that are associated with high reliability and validity (Boateng et al., 2018;Nadler et al., 2015).

Identification of AAS risk factors
Key risk factors for AAS use were identified through five different sources (a publication record for each source can be found in Supplementary file 2). First, reviews and meta-analyses that examine correlates and predictors of IPED use in various populations and settings were searched for variables deemed relevant for understanding AAS use in gyms. This also included studies that integrate findings from empirical research into theoretical frameworks to explain the interrelationship between such variables. Second, cross-sectional studies of gym members and similar populations (e.g. bodybuilders), including studies that were not analyzed in any of the reviews, were searched for additional variables; For instance, the authors of a recent meta-analysis excluded 41 (potentially useful) articles because they were unable to access the required study information (Ntoumanis et al., 2014). These studies were retrieved mainly through searches in scientific databases and relevant grey literature publications. Third, experimental studies of interventions to prevent the use of AAS and other IPEDs were searched for variables that were not reported in the epidemiological and social science literature. These studies were extracted from a recent systematic review  and identified through a separate literature search undertaken by the author in May 2019 (for search strategy, see Supplementary file 3). Fourth, qualitative research exploring the aetiology and trajectory of AAS use initiation based on in-depth interviews and/or ethnographic field work were examined for the same reasons as intervention studies. Finally, a preliminary list of risk factors was presented to a convenience sample of IPED experts, and additional risk factors were added based on their suggestions.
Risk factors were selected according to the following criteria: 1. Explanatory power (i.e. the variable helps explain in a plausible way why some individuals, but not others, decide to start using AAS) 2. Measurability (i.e. the variable can be measuredeither through existing or new instruments) 3. Feasibility (i.e. whether it is practically possible, desirable, or ethically defensible to attempt to modify the variable) Only risk factors that met all of the above critera were included in the study. Explanatory power was determined in quantitative studies through a subjective assessment of each variable and the strength of its association with AAS use or a similar behavioral outcome (e.g. doping or use of performance enhancing drugs). In qualitative studies, this assessment was not based on any objective measures but on subjective judgement alone.
To create a meaningful overview of the findings, and to aid future intervention development, the risk factors were grouped according to the constructs of the COM-B model and the associated domains of the Theoretical Domains Framework (TDF). This enables intervention designers to select appropriate behavior change techniques for each variable. COM-B (Capability, Opportunity, Motivation, and Behavior) is a generic model of human behavior that can be applied to any situation in order to explain, predict, and provide a theoretical point of departure for changing that behavior. This model posits that reflective and automatic motivational processes is essentially what drives behavior. However, motivation is not a sufficient condition. An individual must have the physical and psychological capability to enact the behavior, and there must be an opportunity afforded by the physical and social environment for the behavior to occur (Michie et al., 2011). TDF is an integrated and validated framework of behavioral influences developed from a synthesis of 128 theoretical constructs from 33 theories of behavior (Cane et al., 2012). When used in conjunction with the COM-B model, the TDF permits a more detailed behavioral analysis which in turn provides the basis for developing a more robust intervention strategy (Atkins et al., 2017;Michie et al., 2018). The COM-B model will also be used to structure the discussion.

Pilot testing
To enhance the quality and validity of the survey, the questionnaire was tested in a two-stage piloting phase in August 2019. This involved (1) examining the target groups' understanding of the instrument, including its content, definitions, and instructions, (2) assessing the suitability of items (form and wording) and response options (scale sensitivity and verbal anchors), and (3) identifying potential sources of bias and strategies for minimising bias. In stage one, a convenience sample with individuals from the target group was recruited to gather preliminary feedback on the questionnaire before performing a formal evaluation.
In stage two, a final questionnaire draft was administered to conference delegates who participated in a workshop on AAS use prevention organized by the author in connection with the 8 th International Network of Doping Research (INDR) conference. On day one, workshop participants completed the questionnaire and took part in a subsequent focus group to discuss the findings from the survey. On day two, the author approached each workshop participant in person and asked them to explain in their own words the purpose of the rating task. All workshop participants provided explanations that indicated they had fully understood the task.

Data analysis
Data from the survey were analyzed descriptively using Microsoft Excel V R : Numbers and percentages were calculated to summarize the frequency and proportion of participants' responses to the items in part one, two, and four, whereas the items in part three are presented as means and standard deviations.
Responses to the open-ended question will be analyzed separately in a forthcoming qualitative study.

Results
In total, 78 out of 125 invited IPED experts responded to the survey during the study period (response rate: 62.4%). Of these, 66 completed the survey and 12 returned a partially completed questionnaire (completion rate: 84.6%). Data obtained from partially completed questionnaires were included only if they contained responses to all items in part one and two, or more. Four participants were excluded; two due to incomplete data sets and two upon their own request. This left a final sample of 74 IPED experts, representing 59.2% of invitees, whose responses were included in the data analysis. Inspection of the full dataset confirmed that participants' responses were in all probability credible and free of deliberate attempts to undermine the survey. A subgroup analysis further confirmed that the inclusion/exclusion of data from partially completed questionnaires did not affect the findings. Table 1 presents the demographic and personal characteristics of the study sample, which is composed predominantly of highly expertised, middle-aged, and mostly male, IPED experts from Western countries. More than half of participants (59%) have eight or more years of experience working with AAS use/prevention as a researcher and/or practitioner, highlighting the substantial amount of accumulated expertise in the sample. It should be noted, however, that a significant proportion of the sample deviate from this general pattern with respect to gender (27% are female), age (23% are between 18 and 34 years), and level of expertise (14% have less than three years of experience working with AAS use/prevention). Moreover, despite being limited to Western countries, the sample is geographically diverse in that 16 different countries are represented in the study (Australia, Austria, Belgium, Denmark, Finland, France, Germany, Greece, Iceland, Italy, Netherlands, Norway, Sweden, Switzerland, the United Kingdom, and the United States).

Participant characteristics
The participating researchers represent a diverse range of scientific disciplines and research areas, including sociology, criminology, anthropology, social work, psychology, history, sport management, sports science, public health, sports medicine, psychiatry, forensic medicine, and endocrinology. Likewise, the practitioner subgroup is comprised of various professionals who represent different national and local organizations, institutions, agencies, and services (both governmental and non-governmental). These include: National anti-doping organizations, public health agencies, drug monitoring and education organizations, primary healthcare providers, needle and syringe programs, specialized IPED harm reduction clinics, AAS outpatient clinics, and addiction treatment and rehabilitation services. Participants' publication records and job descriptions indicate that AAS is the class of IPEDs in which most of them hold their primary expertise.

AAS risk factors
A total of 18 psychosocial 4 and two behavioral risk factors met all the inclusion criteria and were included in the study (RQ1). Table 2 provides an overview of these risk factors including how they were operationalized in the survey. Over half of them represent the reflective motivation (7) and social opportunity (5) constructs of the COM-B model, whereas the remaining risk factors represent psychological capability (1), physical opportunity (1), automatic motivation (2), and actual behavior (2). No physical capability variables were included. In addition, the risk factors represent both purely empirical variables (e.g. drive for muscularity and body dissatisfaction) as well as constructs from a diverse range of theories, models, and frameworks of behavior: the Theory of Planned Behavior, the Health Belief Model, Social Cognitive Theory, Social Norms Theory, the Sport Drug Control Model, and two socio-ecological frameworks of performance enhancing drug use (Ajzen, 1991;Bandura, 1986;Cialdini & Goldstein, 2004;Donovan et al., 2002;Janz & Becker, 1984;Stewart & Smith, 2008

Importance and preventability
A table with mean importance (n ¼ 70) and preventability (n ¼ 69) scores for each risk factor based on the IPED experts' assessment (RQ2) can be found in Supplementary file 4. The mean importance scores range from 2.0 to 3.5 (SD: 0.7-1.1). A total of 13 risk factors obtained mean scores above 3, indicating moderate to high importance. Mean scores for the remaining seven risk factors were between 2 and 3, indicating low to moderate importance. Thus, on average, no risk factors were rated as less than 'slightly important'. Mean preventability scores range from 2.2 to 3.2 (SD: 0.7-1.1). Only two risk factors obtained mean scores above 3, indicating moderate to high preventability, whereas the mean scores for the remaining 18 risk factors lie between 2 and 3, indicating low to moderate preventability. Again, on average, no risk factors were rated as less than 'unlikely'.

Preventability clusters
To enable a more meaningful interpretation of the findings, and to support decision-making in relation to intervention Exposure to sale of sports supplements in the gym 'Is training in a gym that sells legal sport supplements (e.g. protein powders or creatine)' Social opportunity Social influences Descriptive norm 'Believes that steroid use is more prevalent than it actually is' Subjective norm 'Feels pressure from friends or peers to use steroids' AAS-using peers 'Has steroid-using friends or peers' Exposure to AAS-using role models 'Is exposed to steroid-using role models or idols (e.g. on social media or in the gym)' Training in a 'hardcore'  development, the risk factors were grouped according to four pre-specified 'preventability clusters'. Cluster 1 (very promising risk factors) contains risk factors that a large majority of participants (more than 70%) rated as either 'very likely' or 'likely' on the preventability scale. Cluster 2 (promising risk factors) consists of those risk factors that a majority (50-70%) rated as either 'very likely' or 'likely'. Similarly, cluster 3 (unpromising risk factors, but worthy of consideration) and cluster 4 (unpromising risk factors) include risk factors that were rated as either 'unlikely' or 'very unlikely' by 50-70% and by more than 70% of participants, respectively. The cutoff points were established on the basis of the assumption that, all other things being equal, larger groups of experts make better judgements and more accurate predictions than smaller groups of experts in relation to their area of expertise. Accordingly, risk factors in cluster 1 are presumed to be more modifiable than risk factors in cluster 2. The same is true for the other clusters. Table 3 presents an overview of the clusters and the risk factors within them. No risk factors were assigned to cluster 4. Mean importance scores range from 2.9 to 3.5 (SD: 0.7-0.9) and from 2.0 to 3.5 (SD: 0.7-1.0) across the six 'very promising' and eight 'promising' risk factors, respectively. For the six risk factors categorised as 'unpromising, but worthy of consideration', mean importance scores range from 2.1 to 3.5 (SD: 0.8-1.1).

Additional risk factors
A number of additional risk factors for AAS use were suggested by participants (for a detailed overview, see Supplementary file 5). Thematically, they fell into six categories: Individual factors (e.g. gender and a predisposition to take risks), family and social network factors (e.g. vulnerable upbringing and parent perceptions), socio-cultural factors (e.g. the medicalisation of social problems and body Improved self-esteem as training purpose 63 3.2 (0.7) Lack of concern for long-term health 60 3.2 (0.9) Body dissatisfaction 58 3.4 (0.7) Desire for increased sex drive 57 2.4 (0.9) Training in a 'hardcore' gym 55 3.4 (0.8) Illicit drug use 55 2.9 (0.9) Exposure to sale of sports supplements in the gym 54 2.0 (0.9) Subjective norm 52 3.5 (0.7) Drive for muscularity 52 3.4 (0.8) Personal norm/morality 51 2.6 (1.0) Cluster 3 (Unpromising risk factors, but worthy of consideration c ) AAS-using peers 68 3.5 (0.8) Exposure to AAS-using role models 58 3.3 (0.8) Perceived benefits of AAS 55 2.9 (1.1) Sports supplement use 51 2.1 (0.8) a Includes risk factors that more than 70% of IPED experts rated as either 'likely' or 'very likely' on the preventability scale. b Includes risk factors that 50-70% of IPED experts rated as either 'likely' or 'very likely' on the preventability scale. c Includes risk factors that 50-70% of IPED experts rated as either 'unlikely' or 'very unlikely' on the preventability scale.   ideals), regulatory and drug market factors (e.g. weak law enforcement and availability of AAS), criminality and polysubstance use factors (e.g. having a criminal record and previous injecting drug use), and employment and sportsrelated factors (e.g. seeking a career where size is an asset and sports injuries). Only a few of these risk factors met all of the inclusion criteria, and those that did were covered by at least one of the risk factors already included in the survey.

Discussion
This study represents the first attempt to systematically map key risk factors for young men's use of AAS and determine their relevance as targets of intervention in the context of recreational strength training in gyms. Consistent with the recommendation made by  to quantify and examine the relative strength of factors influencing the decision to use AAS, the present study identified 18 psychosocial and two behavioral risk factors and subsequently assessed their importance and preventability drawing upon a sample of leading international experts in the field. The use of expert knowledge to assess these characteristicsa crucial piece of information for designing effective interventionsis what makes the present research novel and unique compared with previous research that have examined correlates and predictors of IPED use Blank et al., 2016;Cafri et al., 2005;Nicholls et al., 2017;Ntoumanis et al., 2014;Tavares et al., 2019). Notably, all risk factors were rated by the IPED experts as at least 'slightly important', suggesting that each risk factor has some explanatory power. This finding is corroborated by the validity assessment in which nine out of 10 experts reported that they belive the risk factors cover the most important aspects of AAS use. Goldberg et al. (1991) made an observation over three decades ago that is by no means banal: The use of AAS represents goal-directed behavior that is learned and reinforced in a particular social context through the influence of significant others. Thus, a necessary first step towards understanding the aetiology of AAS use initiation is to consider what goals AAS are being used to achieve. Although research has identified various personal reasons for why men pursue physical transformation through lifting weights in the gym, whether it is for body image, occupational, or other reasons, the desire for a more muscular physique is fundamentally what drives some individuals to chemically enhance their training with AAS. The desired level of muscularity, body fat, and strength may vary between individuals, but increased muscle mass nevertheless stands out as the overarching goal of this behavior. It is therefore essential that interventions in this area seek to address motivational processes related to the muscular ideal and the use of AAS as a means to this end (Andreasson & Johansson, 2020;Christiansen, 2020;Sagoe, Molde, et al., 2014;. In this study, 11 motivational risk factors for AAS use were identified, representing various beliefs, goals, and emotions. Individually, these risk factors can only partly explain the motivational forces that influence a person's decision to use AAS, and they should not be understood as deterministic. Not all men who use AAS have body image issues even though body dissatisfaction is a risk factor for AAS use in this population (Goldfield & Woodside, 2009;Hilkens et al., 2021;Kanayama et al., 2003;Pope et al., 2012). However, when considered together, they form the countours of an ideal-typical motivational profile of AAS-using young men. This motivational profile can be described as follows: A high drive for muscularity combined with body dissatisfaction and low selfesteem provide the impetus to seek extraordinary means to achieve the desired body size and shape. Attitudes and personal values in favour of AAS, reinforced by little concern with long-term health and a perception that the benefits of AAS outweigh the potential harms, facilitate the process of gradually recognising AAS as a legitimate way to enhance muscularity. This process is further catalyzed by anticipations of an increased sex drive and by feelings of being unable to build the desired physique without using AAS and to resist the temptation when offered, encouraged, or inspired to take AAS.
Another important behavioral influence is capability (Michie et al., 2011). Only one risk factor was identified within this domain: Poor knowledge of AAS alternatives. Several lines of research suggest that an effective prevention strategy should promote safe and effective alternatives to AAS. For instance, in a recent cross-European survey of recreational athletes who were mainly involved in fitness, bodybuilding, and amateur weightlifting, the second most frequently reported reason for not using AAS and other IPEDs was: 'I want to see what I can do naturally' (Lazuras et al., 2017, p. 6). Additionally, a number of AAS interventions have been successful in improving this risk factor in school settings by teaching students effective training techniques and proper sports nutrition (Goldberg et al., 1996;MacKinnon et al., 2001;Sagoe et al., 2016). Thus, knowing how gains in muscle size can be achieved by 'natural' means, and the advantages of doing so, can serve as a competing influence when the motivation to experiment with AAS is strong. This may in turn protect against future AAS use. There are of course situations where such knowledge is not protective, for example when the desired level of muscularity or the preferred pace of bodily change exceed the person's physiological limitations. It is reasonable to assume that it is almost inevitable that individuals who are strongly determined to build a muscular physique very rapidly and/or beyond the 'genetic maximum' will regard AAS as the only viable option to reach their goal. The fact that there are currently no licit or illicit drugs or methods available that can match the efficacy of AAS in terms of increasing lean body mass attests to this (Hoffman et al., 2009).
Like other behaviors, the decision to use AAS does not take place in a vacuum; It is greatly influenced by the physical, social, and cultural context in which the behavior occurs . This study identified six risk factors for AAS use within the opportunity domain. Two types of norms are particularly important in this regard: (1) the descriptive norm, which refers to an individual's belief about the prevalence of AAS use in a specific reference group, and (2) the subjective norm, which refers to an individual's belief about others' approval or disapproval of using AAS and his motivation to comply with this. Research has shown that exaggerated perceptions of how widespread AAS use is (i.e. 'everyone is doing it!') and perceived social pressure to use AAS is associated with an increased likelihood of using AAS Ntoumanis et al., 2014;Vogels et al., 1996;Wiefferink et al., 2008). Thus, preventive interventions should aim to: (1) correct erroneous normative beliefs about AAS use, and (2) promote effective strategies to tackle social pressure to use AAS. This could also include social pressure to conform with certain body ideals and notions of masculinity. As also indicated by the findings presented here and elsewhere, interventions should target the sources through which these norms and ideals are transmitted, internalized, and translated into behavior. These include, but are not limited to, the peer group, the physical environment and the social climate within the gym, and hypermuscular role models displaying their bodies publicly in, for example, movies, tv shows, magazines, and on social media platforms (Bergsgard et al., 1996;Leifman et al., 2011;Lenehan et al., 1996;Ntoumanis et al., 2014;Wright et al., 2000).
Past behavior has been shown to be a strong predictor of future behavior (Ajzen, 2011). The present research also identified two behavioral risk factors that interventions should address in relation to the prevention of AAS use: Illicit drug use and sports supplements use. These behaviors have consistently been found to be positively associated with AAS use, but it remains poorly understood whether there is a causal relationship between these behaviors and AAS use, and through which pathways users of illicit drugs and sports supplements may progress to AAS use ( Ntoumanis et al., 2014;Solheim et al., 2017;Yager & O'Dea, 2014).
It has recently been suggested that intervention effectiveness tends to increase with the number of targeted risk factors and behavior change techniques (Bates, Begley, et al., 2019). However, some risk factors may be more resistant to change than others, regardless of their importance. Hence, efforts to reduce them may be futile. Therefore, the preventability clusters were created to serve as a priority mechanism to aid intervention designers in selecting those risk factors that are most promising as targets of intervention. As a minimum, it is recommended that interventions target risk factors in cluster 1 and 2 as these are (presumably) the most malleable. Risk factors in cluster 3 could also be considered, but only when appropriate behavior change techniques have been identified to address those in cluster 1 and 2. Moreover, since behavioral influences operate on various levels ranging from the individual and social network levels to the more distant institutional, community, and societal levels, it is highly unlikely that all the risk factors presented here can be addressed in a single intervention . Individual behavior change interventions delivered in gyms will have to form part of a broader and more comprehensive programme that attempts to interfere with the entire 'system' of behavioral influences leading to AAS use (Bates & Vinther, 2021). Arguably, this includes a wide range of supply-side and demand-side interventions with different functions (e.g. education, training, and environmental restructuring), delivered in different settings (e.g. schools, sports clubs, and gyms), and targeted at different populations (e.g. parents, coaches, and gym staff).
A second aim of the study was to establish whether consensus exists in the international community of IPED experts on the importance, scope, and limits of AAS use prevention in gyms. As regards the importance, there is accumulating evidence to suggest that the non-medical use of AAS is associated with increased morbidity and mortality, and leading medical experts in the field have described the phenomenon as a 'hidden epidemic' and a 'looming public health concern' (Goldman et al., 2019;Horwitz et al., 2019;Kanayama et al., 2008;Lindqvist et al., 2014;P€ arssinen et al., 2000;Thiblin et al., 2015). However, a UK study that used an expert consensus approach to assess the individual and societal harms of various licit and illicit drugs found that, compared with other drugs, the overall harm of AAS is low and similar in magnitude to drugs such as khat, ecstacy, and LSD (Nutt et al., 2010). Since this assessment was made by traditional drug experts and on the basis of the available evidence at the time, the harms of AAS may very well have been underestimated in this work given what we know today about the health risks of AAS. Still, however, AAS use in the general population is rare in most countries including in the UK, especially in comparison to other health compromising behaviors with a more well established impact on the burden of disease such as the use of tobacco, alcohol, and illicit drugs (ACMD, 2010;Rehm et al., 2006;Sagoe, Molde, et al., 2014). For instance, in a nationwide study of Danish youth, 3.4% of male vocational school students reported to have used AAS in the past year. In contrast, 37% were daily smokers, 21% had engaged in binge drinking four times or more in the past month, and 40% had used cannabis in the past year (Bendtsen et al., 2014). This raises the question whether AAS use, while certainly being an individual health issue, is also a public health issue? Defining the threshold in a population above which a behavior is widespread and harmful enough to pose a threat to public health is difficult. Such undertaking inevitably involves an element of subjective judgement. Notably, the present study provides evidence in support of an affirmative answer in that nine out of 10 participating IPED experts believe that preventing AAS use is important to public health.
Another question concerns the possibility of reducing AAS use in gyms, and the extent to which AAS use can be reduced in these environments. Many drug strategies and policy documents contain bold statements that express a strong belief in the prospects of reducing and even eliminating the use of AAS and other doping substances despite the paucity of evidence to support such ambitions (Council of Europe, 1989;Marriott-Lloyd, 2010;Ministry of Health & Social Affairs, 2016;WADA, 2020). In the present study, only about one in 10 IPED experts reported that they believe AAS use in gyms can be eliminated altogether. However, nine out of 10 IPED experts believe that AAS use in gyms can be prevented to a certain degree. This finding adds credibility to claims that prevention is a viable strategy to tackle AAS use in gyms. Finally, there was widespread agreement among IPED experts that a public health response to AAS use in gyms should also consist of harm reduction interventions targeted at current users of AAS. This resonates with previous arguments that prevention and harm reduction should not be viewed as conflicting ideologies, but rather as complementary behavior change interventions that each serve a unique purpose in terms of reducing the societal and public health impact of AAS use (Bates & Vinther, 2021).

Strengths and limitations
This study has several strengths including its use of multiple sources to identify risk factors, the large and geographically diverse sample of IPED experts, the high response rate, and the high content validity of the measures applied. However, there are a number of limitations that should be taken into account when interpreting the findings. First, risk factors were predominantly identified in studies with a cross-sectional design. Thus, they may be consequences rather than causes of AAS use, or both. However, longitudinal studies have shown that variables identical or conceptually similar to the risk factors included in the present study actually predict drug use initiation (Holm et al., 2016;Litt & Dodge, 2008;Van Den Berg et al., 2007). Second, the preventability assessment was based exclusively on subjective judgedment. However, even experts may be wrong, and critics could rightly argue that experimental studies are needed to determine whether they are in fact modifiable. However, this approach represents a cost-effective way to assess the modifiability of risk factors already during intervention development. Third, there may be overlap between some of the risk factors because they tap into the same psychological and social mechanisms. A study of Dutch gym users found that the subjective norm and the personal norm were highly correlated (r ¼ 0.51), indicating a great deal of shared variance between these variables. However, this is mainly of theoretical interest as it would simply mean that by addressing both variables, interventions would slightly overaccount for the underlying mechanism (Wiefferink et al., 2008). Finally, non-responders always pose a threat to the external validity of survey research because answers from this group could have altered the distribution of responses. However, inspection of the completion-by-email statistic indicated that the majority of participants with specific expertise in AAS use prevention completed the survey, indicating alignment between the sample characteristics and the study aims. Notes 1. It is currently unclear exactly how these viruses are transmitted among AAS users but potential routes of transmission include the reuse and sharing of injecting equipment and unprotected sexual intercourse. 2. Based on a review of 84 Erasmusþ projects which include the term 'doping' in their project description (Undertaken by the author in May 2020: https://ec.europa.eu/programmes/erasmus-plus/ projects_en) 3. Three additional participants were included in August 2020. 4. One risk factor (exposure to sale of sports supplements in the gym) relates to the physical and not the social environment.