Anabolic/Androgenic Steroid Use and Aggression I: A Review of the Evidence

by Jack Darkes, PhD
Assistant Professor,
Department of Psychology
Director of Interventions,
Alcohol and Substance Use Research Institute
University of South Florida

Introduction

The association between anabolic/androgenic steroid (AAS) use and aggression ("’roid rage") has been widely accepted in the culture in general, the mainstream media, and the resistance training subculture. This view has been bolstered by the use of AAS "induced" rage as a legal defense (Pope & Katz, 1990). And, although AAS use is not limited to those who perform resistance exercise, the evidence suggests that lifters using AAS are likely to use much higher doses than are those engaging in other athletic endeavors. Therefore, aggression has been both expected and reported to be more prevalent among weight trainers and this phenomenon has become part of the culture of bodybuilding, as well. More recently, naturally occurring androgen precursors have also entered the discussion (Ueki & Okano, 1999; Yesalis, 1999).

This series will examine the support for and potential strength of the causal link between AAS use and aggression and discuss putative processes associated with it. In this installment, representative research on the AAS use and aggression relationship in humans is briefly reviewed, including limited coverage of research on endogenous testosterone levels and aggressive behavior, in order to highlight prevalent themes in the literature. For a more in-depth analysis, recent reviews by Bahrke, Yesalis, & Wright (1996) and Sharp and Collins (1998) are suggested. Further installments will evaluate the evidence for a direct causal relationship between AAS use and aggressive behavior in humans, and a model in which aggression in AAS users is moderated by distal antecedent factors, and partially mediated though proximal psychological variables, will be proposed.

This series will not discuss the pharmacology of the potential AAS/aggression relationship or potential undesirable physical effects of AAS use. Such issues are addressed in many peer-reviewed and popular periodicals, including Mesomorphosis. This series is not intended to suggest a lack of potential psychiatric or medical risk involved in AAS use, nor to endorse or condemn AAS use.

In general, although there has been tacit acceptance of the direct relationship between AAS use and aggression in most quarters, a review of the literature finds that support for this relationship is equivocal. In fact, in studies that controlled for extraneous factors through rigorous inclusion criteria and random assignment, there is little evidence to suggest that moderate AAS use leads to aggressive behavior. However, experimental research addressing real-world patterns and levels of use is needed.

Testosterone, aggression, and dominance

The association of endogenous testosterone (T) with dominance, aggression, or aggressive behavior has a long history in the literature (see Bahrke, Yesalis, & Wright, 1990; 1996 for a full review). The role of T in dominant behavior among males is largely uncontested. However, the notion that dominance and aggression are the same phenomenon is not universally accepted (see Mazur & Booth, 1998). For instance, similar endogenous T levels have been found in both socially dominant but nonaggressive prisoners and their aggressive counterparts (Ehrenkranz, Bliss, & Sheard, 1974). In fact, most studies supporting an endogenous T and aggression link might also be interpreted as suggesting a T – dominance link (Mazur, 1976).

Studies unequivocally supporting a direct relationship between endogenous T and aggression have largely been accomplished with animals. This hypothesis is more rarely supported in humans. Some studies accomplished with "pathological" populations, such as prison inmates, have found that higher T relates to higher probabilities of committing violent crime, being viewed as dominant, and increased rule breaking while incarcerated (Dabbs, 1996). However, this could also reflect a link between T and dominance. Should studies support such a link, a major interpretive hurdle remains; incarcerated individuals are likely to differ from the general populace in many ways that might relate to aggressive behavior, T levels, or both. The generality of such findings is limited, providing little information about T and aggression in the general populace. Indeed, Dabbs (1996) noted that "Relatively few people out of the entire population engage in criminal behavior, regardless of their testosterone levels (p. 180)" suggesting crucial differences between incarcerated subjects and the general population that are not exclusively related to or a result of endogenous T. Such studies highlight the difficulty in generalizing from index cases (such as prisoners or individual "pathological" cases) to the general population.

Also of interest is the fact that the relationship between dominance and endogenous T is not uni-directional. Endogenous T levels not only predict dominant behavior, but are also predicted by it. Winning (the act of dominating) has been associated with an increase in T from pre to post-competition (see Elias, 1981; Gladue, Boehler, & McCaul, 1989; Mazur & Booth, 1998). Hence, increased levels of T in dominant samples might be a result rather than a cause, although this finding has not been universally supported (see Suay et al., 1999, for instance). In addition, some researchers have reported pre-contest rises in T, suggesting an anticipation of future need. This anticipatory rise in endogenous T suggests a system whereby a classically conditioned expectation exerts its influence, a system with implications for psychological theories of the AAS/aggression relationship.

In summary, the relationship between endogenous T and aggression is complex. As with most relationships between physiology and complex behavior, it reflects a "biopsychosocial" process, involving an interaction between the biological substrate of hormonal action, the psychology of the individual, and the social environment in which behavior occurs. Additionally, inconsistent definitions and operationalizations (e.g., discriminating dominance from aggression), the bi-directional effects of T and dominance/aggression, and the lack of longitudinal studies of the T/aggression link in large representative samples, are a few of the factors that complicate the examination of this relationship.

AAS and aggression in humans

Even a cursory search of the psychological and psychiatric literature finds it replete with empirical reports and case studies suggesting that AAS users score more highly than the norm on personality scales measuring hostility. Regardless of this seeming consensus, it has recently been acknowledged that, although AAS use and aggression are correlated, the full extent and nature of the relationship remains unexplained and a clear inference of causality cannot be drawn (Beel, Maycock, & McLean, 1998). For instance, Riem and Hursey (1995) presaged Dabbs’ (1996) sentiments regarding T and aggression, but in relation to AAS use, commenting that "In sum, not all AAS users exhibit aggressive behavior, even though all experience increases in sex steroids (p. 250)." Although AAS use is reportedly widespread (see Brower, 1992), relatively few AAS users exhibit overtly aggressive behavior (rage). Factors that might underlie this variability will be discussed later in this series.

The literature on endogenous T and aggression/hostility provides little assistance in clarifying the potential AAS/aggression relationship in humans for a number of reasons. First, in contrast to endogenous T, AAS use is a behavioral choice. Hence, it is not randomly distributed within the population and AAS users are likely to differ from nonusers. Secondly, AAS ingestion and injection are not simply physical or chemical events, but also behavioral events, part of a sub-culture and a ritual.

The literature on AAS use and aggression encompasses a range of research methods. As with most drug use literature, it is heavily laden with descriptive statistics. For example, lifetime prevalence of AAS use has been reported as 9.1% for males in Great Britain (Korkia & Stimson, 1997). Between 4% and 11% of males in the U.S. have tried AAS (Brower, 1992). And 6.3% of high school football players in Indiana are current or former AAS users (Stilger & Yesalis, (1999). [For a full review of the epidemiology of AAS use see Yesalis, Kennedy, Kopstein, & Bahrke (1993).] An abundance of anecdotal "personal stories" appear in the popular bodybuilding press (e.g., Lefavi, 1998) and case studies are also frequent in the scientific literature (e.g., Corrigan, 1996; Pope & Katz, 1990; Schulte, Hall, & Boyer, 1993; Wilson-Fearon & Parrott, 1999). These data represent naturalistic evidence of this relationship. Evidence from such reports, while rich in individual detail, contributes little to an understanding of the relationship between AAS use and aggression in the larger population. They are biased in that any number of characteristics might differentiate such individuals from the general population besides their use of AAS, again highlighting the difficulty in attempting to speculate about "normal" processes, pharmacological or psychological, in "abnormal" cases. Nonetheless, such cases constitute the majority of the evidence to which the populace is exposed.

More rigorous studies involve the observation of the concurrent correlation between variables within large groups (empirical research) or comparisons between existing groups on concurrent measures (cross-sectional research). Changes in relationships may be evaluated over time, either within or between existing groups (longitudinal or prospective studies). Lastly, treatments (i.e., the administration of AAS/placebo) may be applied to either pre-existing groups (quasi-experimental designs) or to groups of randomly assigned subjects (true experimental designs) who are then evaluated over time.

Empirical and Case Studies.

A substantial amount of empirical research supports the AAS/aggression relationship. For instance, AAS users report higher levels of anger-arousal and hostile outlook than a group that never used AAS (Lefavi, Reeve, & Newland, 1990). Interestingly, data collected from former AAS users was not reported, so it is uncertain if they differed reliably from either group. AAS users exhibit increased instances of mood disorder (Pope & Katz, 1994), higher scores on aggression scales on personality measures (Galligani, Renck, & Hansen, 1996; Yates, Perry, & Murray, 1992) and measures of mood (Bond, Choi, & Pope, 1995). Nonetheless, as with the T/aggression relationship, findings of reliable differences in psychometrically assessed psychological characteristics between AAS users and non-users are not universal (e.g., Malone, Dimeff, Lombardo, & Sample, 1995; Swanson, 1989).

Several case studies (e.g., Pope & Katz, 1990) and retrospective evaluations of forensic records (e.g., Thilbin, Kristiansson, & Rajs, 1997) have also reported associations between AAS and aggression or other psychopathology. However, as noted previously, generalizing from case study data or criminal index cases to the larger population is, at best, a tenuous proposition.

The majority of the empirical and case studies suffer from methodological flaws, such as inconsistent operationalizations of aggression and differing psychometric measures (Bahrke, Yesalis, & Wright, 1996), making comparisons across studies difficult. Most rely exclusively on self-report measures of aggression, a method susceptible to several sources of bias. And, as mentioned earlier, inferring causation using such data is problematic in that AAS use is not randomly distributed in the population. The choice to use AAS, potentially at high doses, is likely to be confounded with a number of predisposing individual differences. For example, current or past AAS users might value aggression and consider aggressive responding a desirable outcome.

Ultimately, the data are largely inconsistent and inconclusive (Uzych, 1992) and a causal relationship between AAS use and aggression has not been established (Isacsson & Bergman, 1993).

Choi, Parrott, & Cowan (1990) followed current AAS users and a non-using control group over a period of several months in a prospective and to some extent quasi-experimental design. The AAS group was evaluated both when using and not using AAS (an ABBA design) and non-users where evaluated at the same times, but never used AAS. A significant group (user/non-user) by drug phase (on/off) interaction for aggression, assessed by the Buss-Durkee Hostility Inventory (BDHI) resulted. Subsequent tests found no reliable effect for drug phase or user status. On the other hand, although there was no significant interaction for hostility (BDHI), there was a reliable effect for group: AAS users were more hostile than non-users, regardless of drug phase. This longitudinal (prospective) quasi-experimental (self-selected and administered treatments - used or did not use) study suggests that those who chose to use AAS were more hostile over time, whether using or not. The assessment of hostility prior to first ever drug use (difficult to accomplish given the low base rate of AAS use) would be more illuminating.

This study was quasi-experimental; there was no random assignment to conditions. Users self-selected drug use and had a prior history of use, and the controls chose not to use AAS and were lifetime nonusers. AAS users and nonusers have, in other empirical studies, differed in their mean scores on a variety of self-report and psychometric measures of personality and aggression (e.g., Galligani, Renck, & Hansen, 1996; Moss, Panzak, & Tarter, 1992). Therefore, any between group effects (as compared to "cycling on or off" differences) merely replicate the cross-sectional findings and might represent dispositional factors related to self-selection, rather than AAS use.

In a within subject, double-blind, prospective design, Su et al., (1993) examined four within subject drug phases: placebo baseline, low dose (40 mg/day) and high dose (240 mg/day) Methyltestosterone and placebo withdrawal. Each phase lasted 3 days. Significant increases in positive mood, negative mood, and cognitive impairment during high dose administration resulted. One out of twenty-nine (approximately 3.4%) participants exhibited a hypomanic episode (an atypical, but non-severe elevation of mood). Although changes in hostility across time showed a dose response relationship, the only reliable differences were between placebo and high dose time periods. These authors note that "The increased symptoms we noted during anabolic steroid administration, while significant, were subtle, reflecting several factors. First, the response to anabolic steroids across members of the subject group was highly variable, ranking from negligible to dramatic (p. 2763)." They acknowledged that marked increases in a small number of subjects were sufficient to create significant differences across time periods and, perhaps most interestingly, noted that "Symptomatic differences did not, however, reflect differences in plasma anabolic steroid levels (p. 2763)." It must be noted that this dosing pattern, a single AAS used at relatively low doses for a very short period of time, does not generalize to typical use in a naturalistic setting. In fact, as the quote above suggests, any behavioral or psychological response in this sample had less to do with blood levels of AAS than with other apparently unmeasured variables.

Gradually increasing doses of testosterone cypionate (150, 300 and 600 mg/week) or placebo were injected, in blocks of two weeks, into eight normal male volunteers, including both prior AAS users and nonusers (Kouri, Lukas, Pope, & Oliva, 1995). Aggression was operationalized as the number of button pushes chosen in order to subtract points from a fictitious opponent. The fictitious opponents’ subtraction of points from participants represented provocation. Two participants failed to believe the sham opponent deception and were dropped, leaving six participants for subsequent within subject comparisons. Increased "aggressive responding" in response to provocation, as compared to both placebo administration and baseline measures, followed testosterone administration. Higher scores were also reported on the Aggression Questionnaire at post testosterone as compared to baseline, largely due to increases in the Physical Aggression score. Whether the participants included (five lifters and 3 non-lifters: 3 with a prior history of AAS use) and the measure of aggression used provide much insight into the AAS/aggression relationship is uncertain. It was not clear which participants were excluded or, in light of the exclusions, how to interpret the statement "Since many of the subjects could not discriminate the testosterone treatment from the placebo treatment… (pp. 77-78)" in view of the small number of participants included in the analyses.

Swanson (1989) examined concurrent differences between current AAS users, non-AAS using athletes, and non-using non-athletes on aggressive behavior. Group membership was verified by urinalysis. A sham reaction time competition was used and the participants’ choice of a noise level to which their "opponent" was exposed if the opponent were slower on the task constituted the measure of aggression. Participants also completed the BDHI. No between group differences were found in behavioral or self-report indices. This study is subject to the previous caveats regarding self-selection when using pre-existing groups, as well as issues related to the operationalization of aggression. Even so, while certain correlations were significant within the AAS using group, there were no differences reported between AAS users, non-using athletes, and non-using non-athletes.

Several true experimental studies, incorporating random assignment of non-using participants to AAS or placebo treatments, have recently appeared. Although the ability of such studies to generalize to self-initiated and self-maintained AAS use can be limited, they address a number of the problems associated with the cross-sectional, prospective, and quasi-experimental designs reviewed above. They constitute a true test of the AAS/aggression relationship while controlling for biases associated with self-selection and the existence of predisposing characteristics.

Bjorkqvist, Nygren, Bjorklund, and Bjorkqvist (1994) randomly assigned twenty-seven male participants to receive no-treatment control, placebo, or 40 mg/day orally administered testosterone (Panteston) over a seven day period. Both self-reported and observer-rated mood showed no effect of drug treatment. In fact, the only reliable differences reported, for self-reported anger, irritation, frustration, and impulsivity and for observer ratings of frustration, indicated that the placebo group scored higher than the no-treatment or testosterone treated groups. While, as in earlier studies, the low dose level certainly impacts the applicability of these results to real world AAS use, it is clear that anticipation and expectation played a part in participants’ observer rated behavior and self-report. However, as these authors point out "What is surprising and calls for an explanation, is the absence of a placebo effect in the group receiving testosterone (p. 24)."

Tricker et al. (1996), reported on mood and behavioral changes in a sample in which physical performance changes were reported separately by Bhasin et al., (1996). Testosterone administration (600 mg/week testosterone enanthate in 3 ml. sesame oil or a placebo of 3 ml. of sesame oil, IM) and exercise (strength training v. no exercise) were completely crossed to create four treatment cells. Forty-three males were randomly assigned to the four conditions and evaluated over a 30-week period in the following order – 4-week control period, 10-week treatment period, and 16-week recovery period. Forty participants completed the study. Attrition was unrelated to adverse drug effects. No between group difference in mood or behavior assessed via psychometric instrument, self-report, or observer (significant other) ratings were reported. As before, both dose and the use of a single drug may not accurately reflect naturalistic practice. Nonetheless, the administration of a supraphysiological dose of AAS over a 10-week period to randomly assigned participants found no reliable differences in aggression between those receiving AAS and those receiving placebo.

A recent study (Yates, Perry, MacIndoe, Holman, & Ellingrod, 1999) reported similar results. Of 42 participants randomly assigned to receive either 100, 250, or 500 mg/week of testosterone cypionate, 31 completed the study. The design included a 2-week period of placebo injections for all participants, followed by 14 weeks of injections at their assigned dose. Attrition was largely related to failure to attend weekly visits, although two 100 mg. dose dropouts were excluded due to psychological exclusions (personality disorder and high BDHI prior to treatment). One 250 mg. participant dropped out due to gynecomastia and one was lost to follow-up. One 500 mg. subject dropped out due to worsening acne and another withdrew due to adverse psychological effects (increased irritability, sleep-onset insomnia, and concentration problems – but no aggressive behavior). Analyses indicated no significant differences in attrition across the groups and no effect of non-completion on the results found with those who completed the study.

No reliable effects of any dose were found for measures of aggression, whether self-report or collateral ratings. Several quotes from these authors are noteworthy. First, they noted "…testosterone cypionate at doses of up to 500 mg/week is associated with minimal psychological effects for the majority of subjects in the study (p. 258)." However, "…the entry criteria were extremely rigorous. More than half of the potential subjects were excluded because of evidence of Axis I or II disorders or elevated psychometric measures of aggression (p. 259)." Again, the use of a single AAS and the range of doses administered do not reflect real world use, but neither do the exclusion criteria. Nonetheless, through the use of random assignment and rigorous exclusionary criteria, most potentially confounding variables (self-selection and pre-existing psychological factors) were controlled for in this study. The results suggest that, when such factors are controlled for, there is relatively little evidence to link AAS use with aggression at the doses used.

Summary

This brief review of the literature finds no clear, consistent, and unequivocal support for the hypothesis that AAS use causes aggression. Does this refute the anecdotal reports and case studies that depict heavy steroid users as aggressive? No. Not only can such idiographic results not be generalized to the larger population, but also the normative data cannot account for all individual cases. In addition, ethical concerns regarding the use of higher dose levels and multiple AAS in experimental studies, confounds the pattern of use with the method of data collection (naturalistic, empirical, or experimental). It certainly does not refute the existing evidence for the modulation of neurotransmitter systems associated with aggression by androgens (e.g., Cologer-Clifford, Simon, Richer, Smoluk, & Lu, 1999). Does the inconsistency call into question the reflexive and widespread assumption that the use of AAS inevitably leads to aggressive behavior in humans or that such behavior is a result of purely pharmacological events? It would seem so, at least to some extent, and within the limits set by issues of dose and simultaneous use of multiple AAS. Certainly the null hypothesis, that AAS use is not necessarily causally related to aggression, cannot be rejected.

In short, as Beel et al. (1996) suggested, the literature reveals a rather complex relationship between AAS use and aggressive behavior. Perhaps this complexity has been over-simplified for mass distribution, an occurrence that is common in such instances. If so, there may be several reasons for it. The complexities of the relationship may be distilled down to imprecise bits of information for dissemination to a populace that deals best and most comfortably with short, easily digestible answers. People often desire easy to grasp dichotomies, preferring simple and clear-cut conclusions even when faced with decidedly complicated and uncertain realities. Perhaps this simplification reflects the desire to curtail the potential abuse of AAS. Such statements, that a certain drug causes undesirable behavior, often become an integral part of "scare tactic" approaches, presenting extreme or worse case scenarios to enhance negative expectations. Unfortunately, such messages mean little to ongoing users, whose experiences might disconfirm the assertion. And, conversely, such statements may heighten the drug’s appeal, should the outcome (e.g., increased aggression) be desirable to the individual contemplating use or lead to negative outcomes, when the taking of the drug facilitates the expected outcome.

Part II of this series will examine the literature reviewed above and discuss its strengths and weaknesses as evidence for the inference of causation in the AAS use and aggression relationship.

References