Our nosology defines substance use disorders by specific behaviors and their consequences. Yet current research, as exemplified in the article by Ersche et al. in this issue (1), is now based on the implicit assumption that addictive disorders depend on characteristics that preceded—and can be dissociated from—the behavior that defines them. Identifying these characteristics is part of a general strategy in psychiatric research to investigate underlying features (perhaps heritable) that predispose some individuals to more overt signs of a disorder. In fact, substance use may well be an epiphenomenon, albeit a potentially dangerous and toxic one, of a more pervasive neurobehavioral problem. For example, substance use disorders are highly prevalent in affective, anxiety, psychotic, and impulse control disorders (2). Therefore, susceptibility might be related to neural functions like the regulation of reward sensitivity, motivation, or arousal or to behavioral constructs like impulsivity or impaired decision making, which would underlie these conditions as well as substance use (3, 4). The Ersche et al. study used judiciously chosen markers for such characteristics, many of which were characterized in terms of underlying neural mechanisms (1).

Studies of such underlying traits or endophenotypes typically require at least three subject groups: probands with the condition in question, their nonaffected family members, and comparison subjects without the condition (or a family history of it) who are otherwise comparable to the probands (5). Endophenotype studies of substance use disorders are challenging. An individual diagnosed with a substance use disorder has already been exposed to the substance, resulting in potentially lasting behavioral, cognitive, or physiological changes. Furthermore, individuals with substance use disorders are likely to be currently using the substance, complicating the process of dissociating the underlying disorder from the drug. Requiring abstinence would bias recruitment to individuals who may not be as ill, and it raises further questions about the period of abstinence that should be required and the confounding withdrawal effects (6).

During the Ersche et al. study (1), 45 of the 50 stimulant-dependent probands were actively using stimulants (interestingly, 76% were enrolled in a drug treatment program). Despite the problems discussed above, this sample provided a representative population without the confounding effects of withdrawal. Arguably, since the probands were generally taking stimulants, the authors were not measuring endophenotypes in a strict sense, but they were definitely investigating characteristics very likely to mark genetic risk for the clinical condition. The authors measured social, environmental, cognitive, and behavioral characteristics, and the analyses focused on similar features shared by drug-dependent individuals and their unaffected siblings that were not shared by healthy comparison subjects.

Among the findings that appear salient in terms of mechanisms and consequences of substance use disorders, impulsive behavior—particularly impaired response inhibition—was strongly related to stimulant dependence. Impulsivity is multifaceted, and one of the strengths of the Ersche et al. study was that the authors used laboratory-, behavioral-, and personality-based measures of traits related to impulsivity. The results showed greater impulsive and anxious personality traits in the probands (on measures of these traits, probands > siblings > comparison subjects) and positive correlations between measures of impulsivity and anxiety. The potential association between impulsivity and anxiety, possibly related to poor regulation of arousal (7), may appear counterintuitive but has important consequences. Negative urgency (8) is a form of impulsivity where impulsive, poorly considered behavioral responses are associated with negative affect or stressful conditions. This contributes to risk for suicidal behavior, a prominent factor in substance use disorders whether independent or co-occurring with other psychiatric disorders (9).

Another notable finding had to do with the role of early environmental stressors and disruptions of family life. Stimulants and stressful events interact to produce behavioral sensitization, in which the activating effects and the self-administration of stimulants are increased (10). The development of sensitization may be part of the process whereby drug use progresses from sporadic to addictive (11). Probands and their siblings were generally exposed to the same early environments, but the siblings did not develop pathological substance use. This raises two possibilities: First, susceptibility to a substance use disorder may lie not in behavioral sensitization, but in susceptibility to sensitization, which may be greater in probands than in their siblings and is elevated in impulsive individuals (12). Second, characteristics in the siblings may confer resilience to substance use disorder, despite the presence of shared environmental and behavioral-cognitive characteristics with the probands. In these respects, it was advantageous that the subject groups were in their thirties, because those without a stimulant use disorder were relatively unlikely to develop one.

Substance use disorders appear to be part of a general class of addictive disorders with similar pathophysiology. The question of specificity across substances as well as between substance use disorders and other behavioral addictions remains. In the Ersche et al. study of stimulant users, opiates were the most commonly used second drug. Individuals using stimulants and opiates did not differ across measures from those using stimulants alone, suggesting that the candidate endophenotypes might be valid across a range of addictive disorders. Alcohol and cigarettes were also used more by probands than by their siblings. The study was too small, however, to definitively address questions of specificity.

There are caveats. Probands and their siblings were exposed to stressful environments that could have had appreciable epigenetic consequences. The genome reflected by endophenotypes is not the genome that the individual was born with but one that has been modified epigenetically, including by stress conditions and, for probands, stimulant exposure. The differences between probands and siblings in characteristics like anxiety, trait impulsivity, or response inhibition may reflect previously inherited characteristics associated with susceptibility to substance use or may be the later epigenetic consequences of stimulant exposure. Finally, as the authors note, their method does not address the question of resilience in nonaffected siblings. Our efforts have been aimed at illness more than resilience, but a better understanding of resilience mechanisms could provide new strategies for prevention and treatment.