Introduction

The heightened co-occurrence of antisocial behavior and substance use during adolescence can predict subsequent psychopathology (e.g., antisocial personality disorder [APD] and substance use disorder [SUD] [1]) and their co-occurrence is associated with a poorer prognosis [[2], [3], [4]]. For example, recent meta-analyses have revealed that the comorbidity between alcohol use disorder (AUD) and APD is as high as 76.6% [5] and that such comorbidity predicts treatment dropout [6]. Differences in impulsivity have long been acknowledged as important predictors of adolescent antisocial behavior and alcohol use [7,8] and subsequent disorders. As per the behavioral disinhibition model (BDM), impulsivity, evident by early adolescence, underlies the co-emergence of antisocial behavior and substance use in early and mid-adolescence [1]. By late adolescence, these problem behaviors predict the subsequent development of APD and SUDs, such as AUD [1,9].

The BDM hypothesizes that the cascading effects of impulsivity on early to mid-adolescent antisocial behavior and substance use [10] contribute to these problem behaviors becoming more direct predictors of psychopathology in late adolescence rather than impulsivity itself [1,9]. In addition, it has long been acknowledged that antisocial behavior (such as stealing and fighting) and substance use co-occur, with antisocial behavior typically emerging first [11]. As such, early impulsivity may continue to predict later alcohol use and AUD, through the cascading chains of mid to late adolescent antisocial behavior, but not directly [1,9]. Indeed, most studies suggest that adolescent impulsivity becomes less predictive of problem behaviors after mid-adolescence or late-adolescence [[12], [13], [14], [15]].

While there is evidence to support the BDM [16], this research has focused on individual differences in impulsivity as predictors of subsequent individual differences in psychopathology. However, such tests of the theory do not directly inform us whether changes in impulsivity during adolescence only predict subsequent outcomes such as AUD and APD via behavioral tendencies such as alcohol use and antisocial behavior that emerge from impulsivity. That is, prior tests of the BDM have not reliably demonstrated whether impulsivity only continues to predict AUD and APD as mediated by the behaviors (i.e., alcohol use and antisocial behavior) that lead to those conditions. The present study was designed to test this prediction. If this prediction of the BDM were to hold, it would suggest that intervening on impulsivity will only provide benefit if done before its sequelae have emerged.

A relevant study by Elkins et al. [16] that showed support of BDM, used a twin-sample and 3-wave longitudinal design and reported that conduct disorder symptoms at age 11 predicted substance use at age 14 and SUD at 18, controlling for attention deficit hyperactivity disorder symptoms. However, the odds ratios for conduct disorder as a predictor of SUD were at least 1.5 times larger than the odds ratios of attention deficit hyperactivity disorder symptoms. These findings provide support for BDM but also suggest that besides impulsivity, conduct problems can begin to play a bigger role in predicting substance use especially later in adolescence. One explanation is that although impulsivity may decline in high-risk adolescents after its initial increase [17], the youth in that high-risk trajectory still show increased AUD later on [17]. Considering the findings of Elkins et al. [16], the abovementioned finding [17] emerged perhaps due to the sequelae of impulsivity via conduct problems leading to late AUD and APD. But this explanation is yet to be tested.

The present study uses repeated measurements of impulsivity, antisocial behavior, and alcohol use from early to late adolescence to test this cascading hypothesis. Namely, we simultaneously investigated (a) whether within-person changes in repeatedly measured impulsivity predict subsequent within-person changes in adolescent alcohol use and antisocial behavior (cf. [18]) and (b) whether those problem behaviors predict subsequent psychopathology (AUD and APD) in late adolescence rather than impulsivity itself (Figure 1). The longitudinal design is an advantage of the present study, because despite the support for the BDM, in many studies impulsivity is assumed to be stable and it is either measured in childhood or early adulthood [19]. In fact, in a recent meta-analysis [20] on self-control and problem behaviors, none of the included longitudinal adolescent studies had multiple assessments of impulsivity across the entire adolescent period. This leaves the potential role of impulsivity as a predictor of subsequent problem behavior across the adolescent period unexamined.

Figure 1.

All examined paths of the RI-CLPM. Within-wave associations and control variables (gender, SES, and ethnicity) are not depicted. In the final RI-CLPM, all paths from T4 impulsivity to T4 Alcohol Use Disorder symptoms and from T4 impulsivity to T4 Antisocial Personality Disorder symptoms could be constrained to 0.

Figure 2.

Significant standardized paths of the final RI-CLPM. Within-wave associations and control variables (gender, SES, and ethnicity) are not depicted.

Participants

The community sample of adolescents in the present study took part in the Philadelphia Trajectory Study [21], which was approved by the Institutional Review Board of Children's Hospital of Philadelphia. In this 6-wave study, participants between ages 10–12 years at baseline were tested annually from 2004 to 2010, with a final 2-year follow-up in 2012. We used audio computer-assisted self-interviewing [22] for all self-reports except for information on conduct problems which was collected via paper and pencil forms. Most of the samples (70%) were students at 7 public and private schools in the Philadelphia area. The remaining 30% of participants were recruited via flyers distributed at other local venues (e.g., libraries) [21]. We received assent from the adolescents and informed consent from their parents.

The present study only uses waves 3 to 6 due to low levels of alcohol use in prior waves. These 4 waves will be referred to as T1 (baseline), T2, T3, and T4. Attrition was 5.3%, 5.7%, 13.4%, and 24.8% from waves 3 to 6, respectively. Valid data were available for 364 adolescents (at baseline: 51.8% female; M_age = 13.51 years). At T4, participants were between ages 18–21 (M_age = 18.78 [0.72]). We define T1 as early adolescence, T2 and T3 as mid adolescence, and T4 as late adolescence (or “emerging adulthood”). The participants were diverse in ethnicity: 55.07% non-Hispanic White, 26.85% non-Hispanic Black, 9.04% Hispanic, and 9.04% identified as other [10,21]. Most participants had a low-middle (socio-economic status [SES]; Hollingshead Two-Factor Index of Social Status: M = 47.38 ± 15.43; reversed scored) [10].

Measures

Impulsivity (T1–T4) was measured using 13 yes/no items (e.g., ‘Do you usually do and say things without stopping to think?’) from the Junior Eysenck Impulsivity Scale [23,24], which has been validated in previous studies (e.g., [25]). We focused on the acting without thinking about dimension of impulsivity, which pertains to behavior undertaken without adequate consideration of its consequences [26] because (1) it has been consistently linked to both substance use and antisocial behavior [[18], [19], [20]]; and (2) it is theoretically related to the behavioral disinhibition concept of BDM [14]. We summed the scores on these items, and thus this resulting sum-score had a continuous scale, ranging from 0 to 13. Higher scores indicated more impulsive behavior. Cronbach alpha's across waves 1–4: 0.78, 0.82, 0.77, and 0.80.

Alcohol use (T1–T4) was measured with two items that measured ever (yes/no) and past 30-day use. These two items were combined and recoded into one scale, with the following response categories: never drank (= 0), used but not in past 30 days (= 1), drank 1–9 days in past 30 days (=2), and drank 10–30 days (= 3).

AUD criteria (AUDc) at T4 were measured with a questionnaire that tapped abuse and dependence as defined in the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) [27,28]. We used these items to measure AUD as per DSM-5 criteria which do not distinguish between abuse and dependence, but our measure did not assess craving. The scale consisted of 12 items and the following categories were used: 0 = 0 criteria met, 1 = 1 criterion met, 2 = 2 criteria met, and 3 = 3 criteria met. The final 2 categories correspond to a mild AUD diagnosis as per DSM-5 [24]. A total of 13.5% of participants met criteria for a mild AUD. Nevertheless, we used the criterion scores (range: 0–3) to measure severity of AUDc.

Antisocial behavior was assessed using 15 items of the Conduct Problems scale of the Youth Self Report [29,30] at T1–T3 that assess symptoms of conduct disorder as defined in DSM-IV. This scale measures antisocial behaviors such as “stealing” and “physically attacking people”, within the last 6 months, on a scale of: not true (=0), somewhat or sometimes true (=1), and very true or often true (=2). There were no items pertaining to substance use. At T4, participants were 18 years or older; thus the Antisocial Personality Problems scale on the Adult Self-Report [30] was used to assess antisocial behaviors. This scale consists of 20 items that tap the symptoms of APD as defined in DSM-IV, with no items pertaining to substance use. Reliabilities for these scales [31] are high (Cronbach alpha's >0.70) and they have been validated in other studies [31]. Mean scores were used for both scales.

APDc were measured at T4 using DSM-IV criteria [30,32]. This DSM-oriented scale comprises 20 items that were rated as being very consistent with APD [32]. The following categories were used: 1 = normal; 2 = borderline; and 3 = clinical. A total of 5.2% adolescents met the borderline range and 2.4% met the clinical range.

Strategy of analyses

The RI-CLPM was specified in Mplus 7.3 [33]. This model included stability paths for antisocial behavior, alcohol use, impulsivity, and cross-lagged paths between these variables. Concurrent associations between these variables of interest were also estimated in this model. Furthermore, we controlled for gender, race, ethnicity, and SES. Next, the variance between persons (i.e., stable time invariant traits) was parceled out from the variance of the observed scores [34], by controlling the random intercept in the scores of antisocial behavior, substance use, and impulsivity. As a result, the cross-lagged linkages signify whether a within-person change in alcohol use, for example, can be predicted by deviation from one's own mean score on impulsivity and vice versa [34]. We constrained the cross-lagged paths to be equal [34] across T1–T3 as the time-lag between these time points were equal and since this did not worsen the model fit, X²(6) = 8.90, p = .179. To account for developmental changes in the means, we did not constrain the means to be equal over time [35]. Finally, to test for indirect effects between initial impulsivity and T4 AUD and APD, we included cross-sectional paths between T4 alcohol use and antisocial behavior with those outcomes.

To test the significance of the mediational paths, we conducted mediation analyses using bootstrapped standard errors [36]. Given no apparent bias in participant dropout [17,37], we used Full Information Maximum Likelihood for missing data [33]. The magnitude of these cross-lagged paths were interpreted as 0.03 (small effect), 0.07 (medium effect), and 0.12 (large effect) [38].

Results

The means of alcohol use increased over the waves but antisocial behavior and impulsivity peaked at T3 and T2, respectively, and declined thereafter (Table 1). All variables were concurrently correlated in the expected directions at each wave (Table 2).

Random intercept cross-lagged panel model

The results of the final model are reported in Figure 2 and Tables 3 and 4. While the cross-sectional path between T4 impulsivity and T4 APDc was not significant (β = .043; p = .506), the cross-sectional path between T4 impulsivity and T4 AUDcwas significant (β = .235; p = .004). But we found that (1) there was no stability in impulsivity from T3 to T4; (2) no direct cross-lagged paths from T3 impulsivity to T4 alcohol use or from T3 impulsivity to T4 antisocial behavior; and (3) no significant indirect mediational effects via T4 impulsivity to T4 AUDs or from T4 impulsivity to T4 APDc. Therefore, we constrained the paths between T4 impulsivity and T4 AUD and APD to be zero. The fit of the model (Chi-squared (57) = 113.603 p < .001) including these constraints was good (Comparitive Fit Index = 0.98; Root Mean Square Error of Approximation = 0.05; Standardized Root Mean Squared Residual = 0.04). The Bayesian Information Criterion also confirmed that the more parsimonious model without those paths provided a good fit (Bayesian Information Criterion = 9255.241 vs. 9255.414). These constraints are consistent with BDM, which posits that early adolescent impulsivity (T1) predicts APD and AUD (T4) via mid-adolescent (T2-T3) antisocial behavior and alcohol use (and not via impulsivity). Nevertheless, as a robustness check, we also tested whether impulsivity at T3 might predict subsequent T4 AUD and APD. However, this link and any indirect paths via this link were nonsignificant.

∗p < .05; ∗∗p < .01

SE = standard error; AUDc = Alcohol Use Disorder criteria; APDc = Antisocial Personality Disorder criteria.

Table 4. Significant specific mediational links to alcohol use disorder criteria (AUDc) and antisocial personality disorder criteria (APDc)

AB = Antisocial Behavior; APDc = Antisocial Personality Disorder criteria; AU = alcohol use; AUDc = Alcohol use Disorder criteria; IMP = Impulsivity.

Taken together, results showed that changes in impulsivity only predicted changes in alcohol use and antisocial behavior in the first two waves of the series. In other words, impulsivity was no longer a significant predictor from T3 to T4. Instead, T3 antisocial behavior predicted both T4 alcohol use and T4 impulsivity. As for indirect effects, significant mediational cascading links from T1 impulsivity (via antisocial behavior) to T4 AUD (Total indirect: B = 0.014, 95% CI = 0.003, 0.026, p = .017; β = 0.043, 95% CI = 0.008, 0.082, p = .023) and to T4 APD (Total indirect: B = 0.009, 95% CI = 0.001, 0.016, p = .019; β = 0.062, 95% CI = 0.009, 0.104, p = .015) emerged. The link from T1 alcohol use to T2 alcohol use was nonsignificant, and thus we tested the mediational path beginning at T2 alcohol use instead of T1. Not surprisingly, we found significant mediational paths from T2 alcohol use to T4 AUDc (Total indirect: B = 0.050; 95% CI = 0.007, 0.106, p = .046; β = 0.039, 95% CI = 0.005, 0.084; p = .055) and significant mediational paths from T1 antisocial behavior to T4 APD (Total indirect: B = 0.173; 95% CI = 0.057, 0.345, p =. 017; β = 0.091; 95% CI = 0.034, .162, p = .015). For the specific indirect effects, please refer to Table 4. The significant cross-lagged paths were large in magnitude [38], while the significant indirect paths were small in magnitude. To verify our decision to only focus on one form of impulsivity, we also tested a model using delay discounting as the indicator of impulsivity but did not observe any indirect effects of this form of impulsivity to late adolescent psychopathology via mid-adolescent problem behaviors.

Discussion

Guided by the BDM [1], we investigated whether cascading mediating links exist from early adolescent impulsivity to late adolescent/emerging adulthood AUD and/or APD via mid-adolescent alcohol use and/or antisocial behavior (problem behaviors). Consistent with the BDM, the results showed that from early to mid-adolescence changes in impulsivity predicted changes in antisocial behavior and alcohol use. However, changes in impulsivity did not predict changes in antisocial behavior and alcohol use from mid-adolescence to late-adolescence. Similarly, other studies suggest that adolescent impulsivity is less predictive of problem behaviors after mid-adolescence or late-adolescence [[12], [13], [14], [15]]. Unlike those studies, the present study included repeated assessments of impulsivity (and problem behaviors), spanning early to late adolescence, and thus more reliably determines whether the predictive power of impulsivity for problem behaviors is the same during the different stages of adolescence.

Our findings appear to be inconsistent with a previous study [39] that found that declines in impulsivity were correlated with declines in problematic alcohol use in adulthood. Of note, unlike this study, the growth models used in that study [39] did not disentangle a possible role of antisocial behavior, which limits the comparisons between the results of that study with the present study. Nevertheless, similar to that study [39], we also found correlations between impulsivity and alcohol use in our adolescent sample, although impulsivity did not consistently predict the changes in those outcomes as adolescents aged.

Of note, we found that whereas impulsivity predicted problem behavior (and not vice versa) from early to mid-adolescence, a reverse link emerged in mid-adolescence to late-adolescence with increases in antisocial behavior predicting increases in impulsivity. The BDM does not hypothesize such a reversed link. Nevertheless, at least four studies [[40], [41], [42], [43]] that investigated this reversed link have found evidence for it. Perhaps, individuals who engage in antisocial behavior may be labeled as “rule breakers” [18,[41], [42], [43]], which could lead them to describe themselves as more impulsive as they age. This interpretation, extrapolated from labeling theory within the criminology–sociology literature, suggests that labeling someone as criminal or antisocial causes the individual to assume similar attributes [44]. Adolescents might be even more susceptible to such labeling effects due to their ongoing identity development. It would be of an added value for future research to explore this hypothesis.

As for mediational links, we further found support for the BDM, as mediational analyses showed that early adolescent impulsivity predicted late-adolescent APD via mid-adolescent antisocial behavior. However, we did not find that early adolescent impulsivity predicted late-adolescent AUD via mid-adolescent substance use. Instead, we found that early adolescent impulsivity predicted late-adolescent AUD via mid-adolescent antisocial behavior. These mediational results showed small mediation effects but they are nevertheless consistent with the conclusion that impulsivity is a factor in early development of adolescent problem behavior but that this influence wanes as youth develop such that its effects are only evident in the consequences of early impulsivity on alcohol use and antisocial behavior.

To the best of our knowledge, this is the first study to investigate cascading mediating links between impulsivity, alcohol use, and antisocial behavior across adolescence. Nevertheless, at least one study similarly showed that both early adolescent hyperactivity/impulsivity and antisocial behavior predicted alcohol use by age 14 [16]. In addition, early hyperactivity/impulsivity predicted AUD [16]. Our results are consistent with those results [16] and further show that the link between early impulsivity and late AUD is mediated by intervening changes in other problem behavior, beyond within-wave associations and controls for ethnicity, gender, and SES.

In sum, in a large, ethnically, and socioeconomically diverse community sample, we found evidence of antisocial behavior not only as a consequence of impulsivity but also as a predictor of alcohol use from mid-adolescence to late-adolescence. In other words, mid-adolescent antisocial behavior (T3) continued to predict substance use 2 years later in late adolescence/emerging adulthood (T4), whereas mid-adolescent impulsivity at T3 did not. Zooming in on the sequencing of antisocial behavior and alcohol use, we only found evidence of a unidirectional link from antisocial behavior to alcohol use [45]. Finally, as mentioned previously, antisocial behavior additionally predicted AUD, whereas impulsivity did not. Taken together, these results imply that antisocial behavior is more relevant than impulsivity in predicting alcohol use, AUD, and APD in late adolescence. However, intervening early is critical to further avoid the consequences of impulsivity which are more difficult to reverse once psychopathology has developed. Prevention of heightened impulsivity in earlier years also needs attention, in particular focusing on the childhood precursors of impulsivity during early adolescence. Research suggests that individual factors such as childhood executive functioning and social factors such as parenting styles during childhood are important to consider [46]. In the present study, SES effects were accounted for and we found that SES was a significant predictor of impulsivity at each wave. Future research could further investigate the mechanisms by which early exposure to socioeconomic disadvantage influences heightened impulsivity during adolescence, including impacts on child executive functioning and parenting behaviors [47,48]. As for prevention and intervention programs, individual-level interventions such as mindfulness training have shown promising effects for reducing both impulsivity [49] and antisocial behavior [50] in youth. Family-based interventions have also shown promising results [51].

Limitations and future directions

Despite the novel aspects of this study, some caveats should be considered when interpreting the results. Inspired by the BDM, the present study focused on impulsivity as an individual-level predictor. Although social risk factors (e.g., deviant friends) are also acknowledged by this model, impulsivity is hypothesized to be the common risk factor that leads to such social risk factors. Nevertheless, social influences should also be considered in future tests of the model. Of note, a prior study [17] using the current sample showed via latent growth modeling that even the subgroup trajectory wherein impulsivity increased in early adolescence and declined thereafter, still predicted late SUD in late adolescence. The present study further suggests that in that trajectory group, increased levels of antisocial behaviors (as a result of increased impulsivity in early adolescence) predicted the SUD that was observed in that study. Hence, future studies are encouraged to consider antisocial behavior (in addition to impulsivity) when investigating SUD. As for measurement limitations, of note is that our impulsivity measure used binary (yes/no) response options, which may not have adequately captured the variability in impulsivity, as compared to a continuous measure. Nevertheless, research suggests that binary response formats versus ordinal multicategory response formats are equally reliable [52]. Finally, our results for AUD and APD were based on questionnaire responses that have been validated in past research but may not be as robust as diagnoses determined by trained interviewers.

Conclusion

Current results show that targeting impulsivity early in adolescence could halt the cascading chain of events leading to late adolescent AUD and APD, by slowing growth in antisocial behavior during mid-adolescence. However, it appears that during mid to late adolescence, intervening on the consequences of impulsivity (i.e., antisocial behavior) might be more useful to treat alcohol use, AUD, and APD rather than intervening on impulsivity itself. Thus, the findings suggest the importance of antisocial behavior as a source of risk for alcohol use, AUD, and APD in late adolescence. The prevalence rates of AUD and APD in the current ethnically and socioeconomically diverse sample were comparable to nationally representative samples [53,54], and hence the current results may apply to community-based samples of youth. To conclude, clinicians should be aware that our results suggest that adolescents with increasing levels of antisocial behavior are at risk of both AUD and APD.

Antisocial Behavior as a Mediator of the Link Between Impulsivity and Psychopathology in Adolescence

Introduction

The frequent co-occurrence of antisocial behavior and substance use during adolescence is a significant predictor of later psychopathology, including antisocial personality disorder (APD) and substance use disorder (SUD) [1]. This comorbidity is associated with poorer prognoses [[2], [3], [4]], with a meta-analysis indicating a comorbidity rate as high as 76.6% between alcohol use disorder (AUD) and APD [5]. This co-occurrence also predicts negative outcomes like treatment dropout [6]. Impulsivity is recognized as a key factor influencing adolescent antisocial behavior and alcohol use [7,8], and consequently, the development of related disorders.

The behavioral disinhibition model (BDM) posits that impulsivity, observable by early adolescence, underlies the co-emergence of antisocial behavior and substance use in early to mid-adolescence [1]. These problem behaviors then become predictors of APD and SUDs, such as AUD, by late adolescence [1,9]. The BDM further suggests that the influence of impulsivity on psychopathology is mediated by its cascading effects on early to mid-adolescent antisocial behavior and substance use [10]. These behaviors, rather than impulsivity itself, become more direct predictors of psychopathology in late adolescence [1,9]. This is consistent with observations that antisocial behavior, like stealing and fighting, often precedes and co-occurs with substance use [11]. Consequently, early impulsivity might indirectly predict later alcohol use and AUD through a chain of mid to late adolescent antisocial behavior [1,9]. This notion is supported by studies indicating that the predictive power of adolescent impulsivity for problem behaviors diminishes after mid or late adolescence [[12], [13], [14], [15]].

While existing research provides support for the BDM [16], it has predominantly focused on individual differences in impulsivity as predictors of individual differences in psychopathology. However, these studies don't definitively demonstrate whether changes in impulsivity during adolescence predict outcomes like AUD and APD solely through behavioral pathways (i.e., alcohol use and antisocial behavior). This study addresses this gap by examining whether impulsivity's predictive power for AUD and APD is mediated by alcohol use and antisocial behavior. Such a finding would suggest that interventions targeting impulsivity are only effective before these sequelae emerge.

Elkins et al. [16] conducted a relevant twin-sample, three-wave longitudinal study that supports the BDM. They found that conduct disorder symptoms at age 11 predicted substance use at age 14 and SUD at 18, even after controlling for attention-deficit/hyperactivity disorder symptoms. Importantly, the odds ratios for conduct disorder predicting SUD were considerably higher than those for attention-deficit/hyperactivity disorder symptoms. These findings, while supporting the BDM, suggest that conduct problems, in addition to impulsivity, play an increasingly important role in predicting substance use, particularly in later adolescence. One possible explanation is that while impulsivity may decline in high-risk adolescents after an initial increase [17], these individuals still exhibit higher rates of AUD later on [17]. Considering Elkins et al.'s findings [16], this persistent risk for AUD despite declining impulsivity could be attributed to the enduring consequences of early impulsivity, manifesting as conduct problems that ultimately lead to AUD and APD. However, this explanation requires further investigation.

This study utilizes repeated measurements of impulsivity, antisocial behavior, and alcohol use from early to late adolescence to examine this cascading hypothesis. We investigate whether within-person changes in impulsivity predict subsequent within-person changes in adolescent alcohol use and antisocial behavior (cf. [18]). We also examine whether these problem behaviors, rather than impulsivity itself, predict psychopathology (AUD and APD) in late adolescence (Figure 1). This longitudinal design is a significant strength, as it allows for a more nuanced understanding of impulsivity's role compared to studies that rely on a single measurement of impulsivity in childhood or early adulthood [19], despite the dynamic nature of impulsivity across adolescence. Indeed, a recent meta-analysis on self-control and problem behaviors [20] revealed that none of the included longitudinal adolescent studies incorporated multiple assessments of impulsivity across the entire adolescent period, highlighting the need for the present investigation.

[Figure 1. Omitted for brevity. Please refer to the original article.]

[Figure 2. Omitted for brevity. Please refer to the original article.]

Participants

This study utilizes data from the Philadelphia Trajectory Study [21], a six-wave longitudinal study approved by the Institutional Review Board of Children's Hospital of Philadelphia. Participants (N = 364 at baseline; 51.8% female; Mage = 13.51 years) aged 10–12 years at baseline were assessed annually from 2004 to 2010, followed by a final assessment in 2012. Audio computer-assisted self-interviewing [22] was used for all self-report measures except for conduct problems, which were assessed via paper-and-pencil forms. The sample consisted of adolescents recruited from public and private schools (70%) and through flyers distributed at local venues (30%) [21]. Assent was obtained from adolescents, and informed consent was obtained from their parents.

Only data from waves 3 to 6 were used in this study due to low alcohol use prevalence in earlier waves. These four waves are referred to as T1 (baseline), T2, T3, and T4, representing early adolescence, mid-adolescence, and late adolescence/emerging adulthood, respectively. Attrition rates were 5.3%, 5.7%, 13.4%, and 24.8% from waves 3 to 6, respectively. At T4, participants were aged 18–21 (Mage = 18.78 [0.72]). The sample was ethnically diverse: 55.07% non-Hispanic White, 26.85% non-Hispanic Black, 9.04% Hispanic, and 9.04% other [10,21]. Most participants came from low-middle socioeconomic backgrounds (SES; Hollingshead Two-Factor Index of Social Status: M = 47.38 ± 15.43; reversed scored) [10].

Measures

• Impulsivity (T1–T4) was measured using 13 yes/no items from the Junior Eysenck Impulsivity Scale [23,24], a measure previously validated in similar research (e.g., [25]). This study specifically focused on the "acting without thinking" dimension of impulsivity [26], which pertains to actions undertaken without considering potential consequences. This dimension was chosen due to its established link to substance use and antisocial behavior [[18], [19], [20]] and its theoretical relevance to the BDM's concept of behavioral disinhibition [14]. A sum score (range: 0–13) was calculated, with higher scores indicating greater impulsivity. Cronbach's alpha for the scale across waves 1–4 were: 0.78, 0.82, 0.77, and 0.80.

• Alcohol use (T1–T4) was measured using two items assessing lifetime (yes/no) and past 30-day alcohol use. These items were combined into a single scale with four categories: never drank (= 0), used but not in the past 30 days (= 1), drank 1–9 days in the past 30 days (= 2), and drank 10–30 days (= 3).

• AUD criteria (AUDc) at T4 were assessed using a questionnaire based on DSM-IV criteria for abuse and dependence [27,28]. While DSM-5 combines abuse and dependence, our measure did not include craving. The scale consisted of 12 items, with response categories ranging from 0 (no criteria met) to 3 (3 criteria met). The final two categories correspond to a mild AUD diagnosis as per DSM-5 [24]. While 13.5% of participants met the criteria for mild AUD, the criterion scores (range: 0–3) were used to represent AUDc severity.

• Antisocial behavior was assessed using the Conduct Problems scale of the Youth Self-Report [29,30] at T1–T3. This scale measures past 6-month conduct disorder symptoms as defined by DSM-IV, encompassing behaviors like stealing and physical aggression, using a 3-point scale: not true (= 0), somewhat or sometimes true (= 1), and very true or often true (= 2). The scale does not include items related to substance use. At T4, given participants were 18 years or older, the Antisocial Personality Problems scale from the Adult Self-Report [30] was used. This scale assesses DSM-IV symptoms of APD using 20 items and does not include substance use-related items. Both scales demonstrate high reliability (Cronbach's alpha > 0.70) and have been validated in previous studies [31]. Mean scores were used for both scales.

• APDc at T4 were measured using a DSM-IV based scale [30, 32] consisting of 20 items rated for consistency with APD [32]. Response categories were: 1 = normal; 2 = borderline; and 3 = clinical. In total, 5.2% of adolescents met the criteria for the borderline range and 2.4% for the clinical range.

Strategy of Analyses

A random intercept cross-lagged panel model (RI-CLPM) was specified in Mplus 7.3 [33] to analyze the data. This model included stability paths for antisocial behavior, alcohol use, and impulsivity, as well as cross-lagged paths between these variables. Concurrent associations between the variables were also estimated. The model controlled for gender, race, ethnicity, and SES. To separate between-person variance (stable traits) from within-person variance, the random intercept of each individual's scores on antisocial behavior, substance use, and impulsivity was controlled [34]. This approach ensures that the cross-lagged effects represent within-person changes, indicating whether deviations from an individual's mean score on one variable predict subsequent deviations from their mean score on another variable [34]. Cross-lagged paths were constrained to be equal across T1–T3 [34] as the time lag between these points was equal and this constraint did not negatively impact model fit, χ2(6) = 8.90, p = .179. To account for potential developmental changes, the means were not constrained to be equal over time [35]. To test for indirect effects of initial impulsivity on T4 AUD and APD, cross-sectional paths were included between T4 alcohol use and antisocial behavior with those outcomes.

Mediation analyses employing bootstrapped standard errors [36] were conducted to test the significance of the mediational paths. Given the lack of apparent bias in participant dropout [17, 37], Full Information Maximum Likelihood was used to handle missing data [33]. The magnitude of cross-lagged paths was interpreted using the following guidelines: 0.03 (small effect), 0.07 (medium effect), and 0.12 (large effect) [38].

Results

Alcohol use showed an increasing trend across waves, while antisocial behavior and impulsivity peaked at T3 and T2, respectively, before declining (Table 1). Concurrent correlations between all variables at each wave were in the expected directions (Table 2).

[Table 1. Omitted for brevity. Please refer to the original article.]

[Table 2. Omitted for brevity. Please refer to the original article.]

Random Intercept Cross-Lagged Panel Model

The results of the final model are presented in Figure 2 and Tables 3 and 4. While the cross-sectional path between T4 impulsivity and T4 APDc was not significant (β = .043; p = .506), the path between T4 impulsivity and T4 AUDc was significant (β = .235; p = .004). However, the analysis revealed: (1) no stability in impulsivity from T3 to T4; (2) no direct cross-lagged paths from T3 impulsivity to T4 alcohol use or T4 antisocial behavior; and (3) no significant indirect effects of T4 impulsivity on T4 AUDs or T4 APDc. Consequently, the paths between T4 impulsivity and T4 AUD and APD were constrained to zero. The model fit with these constraints was good (Chi-squared (57) = 113.603 p < .001; Comparative Fit Index = 0.98; Root Mean Square Error of Approximation = 0.05; Standardized Root Mean Squared Residual = 0.04). The Bayesian Information Criterion also supported the more parsimonious model without these paths (Bayesian Information Criterion = 9255.241 vs. 9255.414). These constraints align with the BDM, which posits that early adolescent impulsivity (T1) predicts APD and AUD (T4) through mid-adolescent (T2-T3) antisocial behavior and alcohol use, rather than through impulsivity itself. To ensure robustness, a model with a path from T3 impulsivity to T4 AUD and APD was tested, but this path and any indirect paths through it were nonsignificant.

[Table 3. Omitted for brevity. Please refer to the original article.]

[Table 4. Omitted for brevity. Please refer to the original article.]

The results indicate that changes in impulsivity predicted changes in alcohol use and antisocial behavior only in the first two waves, suggesting that impulsivity's predictive power diminishes over time. Conversely, T3 antisocial behavior predicted both T4 alcohol use and T4 impulsivity, a finding not explicitly hypothesized by the BDM but supported by other studies [[40], [41], [42], [43]]. This reverse effect might be explained by labeling theory, which suggests that individuals engaging in antisocial behavior might internalize the label of "rule breaker" [18, [41], [42], [43]], leading them to perceive themselves as more impulsive as they age [44]. This is particularly relevant for adolescents who are navigating identity development. Further research is needed to explore this hypothesis.

Significant mediational pathways were found from T1 impulsivity to T4 AUD (through antisocial behavior; total indirect effect: B = 0.014, 95% CI = 0.003, 0.026, p = .017; β = 0.043, 95% CI = 0.008, 0.082, p = .023) and to T4 APD (total indirect effect: B = 0.009, 95% CI = 0.001, 0.016, p = .019; β = 0.062, 95% CI = 0.009, 0.104, p = .015). The path from T1 alcohol use to T2 alcohol use was nonsignificant, leading to the testing of the mediational path from T2 alcohol use. This path from T2 alcohol use to T4 AUDc was significant (total indirect effect: B = 0.050; 95% CI = 0.007, 0.106, p = .046; β = 0.039, 95% CI = 0.005, 0.084; p = .055), as was the path from T1 antisocial behavior to T4 APD (total indirect effect: B = 0.173; 95% CI = 0.057, 0.345, p = .017; β = 0.091; 95% CI = 0.034, .162, p = .015). For specific indirect effects, refer to Table 4. While significant cross-lagged paths showed large magnitudes [38], the significant indirect paths had small magnitudes. A model using delay discounting as an indicator of impulsivity was also tested but did not yield any indirect effects of this form of impulsivity on late adolescent psychopathology via mid-adolescent problem behaviors.

Discussion

In line with the BDM [1], this study investigated the presence of cascading mediating pathways from early adolescent impulsivity to late adolescent/emerging adulthood AUD and/or APD, mediated by mid-adolescent alcohol use and/or antisocial behavior. Consistent with the BDM, the findings indicate that changes in impulsivity from early to mid-adolescence predicted changes in both antisocial behavior and alcohol use. However, this predictive relationship was not observed from mid to late adolescence, aligning with previous research suggesting that impulsivity's influence on problem behaviors diminishes in later adolescence [[12], [13], [14], [15]]. Unlike these previous studies, the present study's longitudinal design and repeated assessments of impulsivity and problem behaviors offer a more robust examination of this relationship across different developmental stages.

Our findings appear to contradict a previous study [39] that found a correlation between declines in impulsivity and problematic alcohol use in adulthood. However, it is important to note that the growth models used in that study [39] did not account for the potential mediating role of antisocial behavior, limiting the comparability of the results. Nevertheless, consistent with that study [39], the present study also observed correlations between impulsivity and alcohol use in the adolescent sample, although impulsivity did not consistently predict changes in these outcomes as participants aged.

Interestingly, the study found a shift in the relationship between impulsivity and problem behavior between early to mid-adolescence and mid to late adolescence. While impulsivity predicted problem behavior (but not vice versa) in the earlier stages, a reverse relationship emerged later, with increases in antisocial behavior predicting increases in impulsivity. This finding, although not directly hypothesized by the BDM, has been documented in other studies [[40], [41], [42], [43]]. As previously mentioned, this might be attributable to the internalization of negative labels associated with antisocial behavior, leading individuals to perceive and consequently report higher levels of impulsivity.

Regarding mediational links, the study provides further support for the BDM, demonstrating that early adolescent impulsivity predicted late-adolescent APD through mid-adolescent antisocial behavior. Interestingly, the study did not find support for early adolescent impulsivity predicting late-adolescent AUD through mid-adolescent substance use. Instead, the pathway from early adolescent impulsivity to late-adolescent AUD was mediated by mid-adolescent antisocial behavior. These findings, though demonstrating small mediation effects, underscore the diminishing influence of impulsivity as individuals transition through adolescence and highlight the enduring consequences of early impulsivity on subsequent alcohol use and antisocial behavior.

This study is the first to examine cascading mediational relationships between impulsivity, alcohol use, and antisocial behavior across adolescence. The findings are consistent with a previous study [16] that found both early adolescent hyperactivity/impulsivity and antisocial behavior to be predictors of alcohol use by age 14, with early hyperactivity/impulsivity also predicting AUD [16]. The present study builds upon these findings by demonstrating that the link between early impulsivity and late AUD is mediated by changes in other problem behaviors, beyond within-wave associations and after controlling for ethnicity, gender, and SES.

To summarize, the study provides evidence for antisocial behavior not only as a consequence of impulsivity but also as a predictor of alcohol use from mid to late adolescence. Mid-adolescent antisocial behavior (T3) significantly predicted substance use two years later in late adolescence/emerging adulthood (T4), while mid-adolescent impulsivity at T3 did not. Furthermore, the study found evidence for a unidirectional relationship between antisocial behavior and alcohol use, with antisocial behavior predicting alcohol use but not the reverse [45]. Finally, antisocial behavior, rather than impulsivity, emerged as a significant predictor of AUD. These findings collectively suggest that antisocial behavior might play a more critical role than impulsivity in predicting alcohol use, AUD, and APD in late adolescence. Early intervention is crucial to mitigating the long-term consequences of impulsivity, which become more challenging to address once psychopathology develops. Future research should focus on identifying and addressing childhood precursors of impulsivity, such as executive functioning and parenting styles [46]. The study found SES to be a significant predictor of impulsivity at each wave, highlighting the need to investigate the mechanisms through which early socioeconomic disadvantage influences impulsivity during adolescence, including its impact on executive function and parenting behaviors [47, 48]. In terms of prevention and intervention, mindfulness training, both at the individual level [49] and within a family-based framework [51], has shown promise in reducing impulsivity and antisocial behavior in youth.

Limitations and Future Directions

Despite its strengths, the study has some limitations that need to be acknowledged. Firstly, although the BDM acknowledges social risk factors (e.g., deviant peer influence), the current study focused on impulsivity as an individual-level predictor, as the model suggests that impulsivity contributes to these social risks. Future research should incorporate social influences to provide a more comprehensive understanding of the pathways to psychopathology. Secondly, the study's reliance on a binary (yes/no) response format for the impulsivity measure might not have fully captured the variability in impulsivity compared to a continuous measure. However, research suggests comparable reliability between binary and ordinal response formats [52]. Lastly, AUD and APD outcomes were based on self-report questionnaires, which, although validated in previous research, might not be as robust as diagnoses derived from structured clinical interviews.

Conclusion

The findings of this study emphasize the potential of early interventions targeting impulsivity in adolescence to disrupt the cascading chain of events leading to AUD and APD by mitigating the development of antisocial behavior during mid-adolescence. However, during mid to late adolescence, interventions focusing on the consequences of impulsivity, namely antisocial behavior, might be more effective in addressing alcohol use, AUD, and APD than interventions solely targeting impulsivity. Clinicians should be aware of the heightened risk for AUD and APD among adolescents exhibiting increasing levels of antisocial behavior. Given the comparable prevalence rates of AUD and APD in the study's diverse sample to nationally representative samples [53, 54], the findings likely hold relevance for community-based youth populations.

Overall, this rewritten response aims to maintain the structure and content of the original article while enhancing clarity and readability for an academic audience.

Antisocial Behavior's Role in the Link Between Impulsivity and Later Psychopathology in Adolescents

Introduction

We know that teenagers who show both antisocial behavior (like breaking rules or acting aggressively) and substance use are at a higher risk for developing mental health problems later in life. These problems include things like antisocial personality disorder (ASPD) and substance use disorders (SUD). When these issues occur together, it's harder to treat them effectively. For instance, a large overlap exists between alcohol use disorder (AUD) and ASPD, and having both makes it more likely someone will drop out of treatment.

One model that tries to explain this is the Behavioral Disinhibition Model (BDM). It suggests that impulsivity, which is already present in early adolescence, is the underlying reason why teenagers might start showing antisocial behaviors and using substances. As time goes on and they reach late adolescence, these problematic behaviors can develop into serious conditions like ASPD and AUD.

The BDM proposes that impulsivity doesn't directly cause these disorders. Instead, it sets off a chain reaction: impulsivity first leads to behaviors like substance use and antisocial behavior, and these behaviors then increase the risk of developing ASPD and AUD later on. This idea is backed up by studies showing that impulsivity becomes less important in predicting problem behaviors as teenagers get older.

While the BDM has support, most studies haven't looked at whether impulsivity's effect on disorders like AUD and ASPD only happens through those problem behaviors. Our study aimed to test this idea. If true, it would mean that addressing impulsivity is only effective before these problematic behaviors appear.

A previous study supported the BDM and found that conduct disorder symptoms at age 11 predicted substance use at age 14 and SUD at age 18. Importantly, conduct problems were a much stronger predictor of SUD than just general attention problems. This suggests that while impulsivity is important, conduct problems play a larger role as adolescents get older. One explanation is that while impulsivity might decrease after initially increasing in at-risk adolescents, these individuals might still develop AUD due to the lasting effects of their earlier impulsivity, which manifested as conduct problems.

Our study examined this by repeatedly measuring impulsivity, antisocial behavior, and alcohol use in adolescents from early to late adolescence. We wanted to see if:

(a) Changes in impulsivity over time predicted changes in alcohol use and antisocial behavior. (b) These problem behaviors, rather than impulsivity itself, predicted whether someone would develop AUD and ASPD later on.

Participants

We used data from the Philadelphia Trajectory Study, a study that followed adolescents over time. Our study included 364 adolescents who were between 10-12 years old at the beginning of the study and 18-21 years old by the end.

Measures

We measured several things at different points in time:

• Impulsivity: We used a questionnaire to measure how often adolescents acted without thinking about the consequences.

• Alcohol Use: We asked adolescents how often they drank alcohol in the past 30 days.

• AUD Criteria (AUDc): We used a questionnaire to assess how many symptoms of AUD adolescents met.

• Antisocial Behavior: We used questionnaires to measure how often adolescents engaged in behaviors like stealing, fighting, or breaking rules.

• ASPD Criteria (APDc): We used a questionnaire to assess how many symptoms of ASPD adolescents met.

Strategy of Analyses

We used a statistical model called the Random Intercept Cross-Lagged Panel Model (RI-CLPM) to analyze the data. This model allowed us to see how changes in one variable, like impulsivity, predicted changes in another variable, like alcohol use, over time. We also controlled for factors like gender, race, ethnicity, and socioeconomic status.

Results

• Alcohol use increased over time, while antisocial behavior and impulsivity peaked in middle adolescence and then declined.

• Changes in impulsivity only predicted changes in alcohol use and antisocial behavior in the first two years of the study. After that, impulsivity was no longer a significant predictor.

• Interestingly, in later adolescence, increases in antisocial behavior predicted increases in impulsivity.

• Early adolescent impulsivity predicted later ASPD and AUD, but only indirectly through its effects on antisocial behavior. This means that early impulsivity led to increased antisocial behavior, which then increased the risk for developing these disorders.

Discussion

Our findings support the BDM's idea that impulsivity has a cascading effect: it first leads to problem behaviors like antisocial behavior and alcohol use, and these behaviors then increase the risk of later disorders like ASPD and AUD.

We also found that antisocial behavior, once it emerges, can then predict later increases in impulsivity. This suggests that engaging in antisocial behavior might reinforce impulsive tendencies in some adolescents.

This is one of the first studies to investigate this chain of events across adolescence. It highlights the importance of addressing impulsivity early on, before problem behaviors emerge. It also suggests that interventions targeting antisocial behavior in middle adolescence could be crucial in preventing the development of AUD and ASPD.

Limitations and Future Directions

Our study has some limitations. We only focused on impulsivity as an individual factor, but social factors like peer influence are also important. Additionally, we used questionnaires to measure AUD and ASPD, which might not be as accurate as clinical diagnoses.

Future studies should investigate the role of social factors and use more comprehensive measures of AUD and ASPD.

Conclusion

Our findings show that early adolescent impulsivity can set off a chain reaction, leading to antisocial behavior and eventually increasing the risk for AUD and ASPD. Targeting impulsivity early in adolescence and addressing antisocial behavior in middle adolescence could be crucial for preventing these serious mental health problems.

Does Impulsivity in Early Adolescence Lead to Alcohol and Antisocial Problems Later On?

Introduction

You might have heard that teenagers who act impulsively and get in trouble a lot are more likely to have problems with alcohol and continue their antisocial behavior as they get older. This idea is based on something called the Behavioral Disinhibition Model (BDM), which says that being impulsive early on can set off a chain reaction that leads to bigger problems later in life.

The BDM suggests that teenagers who act without thinking might start getting into trouble with things like stealing or fighting during early adolescence. Then, as they reach mid-adolescence, they may also start using alcohol more often. By the time they're young adults, these same individuals might be diagnosed with serious conditions like Antisocial Personality Disorder (APD) or Alcohol Use Disorder (AUD). Importantly, the BDM says that these later problems are more likely to be caused by the earlier drinking and antisocial behavior than by impulsivity itself.

This study wanted to test these ideas by looking at a group of adolescents over several years. The researchers were interested in seeing if: (a) changes in impulsivity early on would predict changes in alcohol use and antisocial behavior later, and (b) whether alcohol use and antisocial behavior in mid-adolescence were better predictors of AUD and APD in young adulthood than impulsivity itself.

Who Was in the Study?

This study used information from the Philadelphia Trajectory Study, which followed a large group of young people over time. The researchers focused on a group of 364 adolescents (who were about 13 years old at the beginning of the study) and looked at information collected when they were 13, 14, 15, and 18 years old.

What Did They Measure?

The researchers were interested in four main things:

• Impulsivity: They measured how impulsive the teenagers were using a questionnaire with questions like "Do you usually do and say things without stopping to think?".

• Alcohol Use: They asked the teenagers how often they drank alcohol and how much they drank.

• Antisocial Behavior: They measured how often the teenagers engaged in behaviors like stealing, fighting, or lying using a questionnaire.

• Symptoms of AUD and APD: When the participants were 18, the researchers used questionnaires to measure symptoms of AUD and APD.

What Did They Find?

Here are the key things the researchers learned:

• Impulsivity predicted problems in mid-adolescence, but not later. Changes in impulsivity between ages 13 and 15 predicted changes in alcohol use and antisocial behavior during that same period. However, impulsivity at age 15 did not predict these problems at age 18.

• Antisocial behavior became a stronger predictor over time. While impulsivity was important early on, antisocial behavior in mid-adolescence (age 15) was a better predictor of both alcohol use and impulsivity at age 18.

• Early impulsivity was linked to later problems through a chain reaction. The study found that impulsive behavior at age 13 predicted symptoms of AUD and APD at age 18 indirectly. In other words, being impulsive at age 13 led to more antisocial behavior, which in turn led to greater risk for AUD and APD.

What Does It All Mean?

This study provides evidence for the idea that impulsivity can set off a chain of events that leads to later problems with alcohol and antisocial behavior. However, it also suggests that impulsivity itself becomes less important over time. Instead, the things that happen because of impulsivity, like engaging in antisocial behavior, become stronger predictors of future problems.

Why is This Important?

These findings highlight the importance of addressing impulsivity early on. If we can help young people learn to control their impulses, we might be able to prevent them from developing more serious problems down the road. However, this study also tells us that it's crucial to address antisocial behavior directly, as this seems to be a key factor that keeps the cycle going.

Why Do Some Teenagers Have Problems with Drinking and Breaking Rules?

Introduction

Have you ever noticed that some teenagers who break rules also seem to have problems with drinking? This can be a big problem because it can lead to even bigger problems later in life, like becoming an adult who always breaks rules or can't stop drinking. This is like a chain reaction where one problem leads to another!

Scientists have a theory called the "behavioral disinhibition model" (let's call it the BDM) that tries to explain this. The BDM says that teenagers who act without thinking (being impulsive) might start breaking rules and drinking more because they don't think about the consequences. This can then turn into a bigger problem when they become adults.

Scientists have studied the BDM, but they haven't figured out if impulsivity only causes problems later in life because of the drinking and rule-breaking in teenage years. This study tried to find that out! If they are right, it means that it's important to help impulsive teenagers early on, before the drinking and rule-breaking starts!

Who was in the study?

The study looked at a large group of teenagers from different backgrounds in Philadelphia. The teenagers answered questions every year from ages 13 to 18 about their impulsivity, if they drank alcohol, and if they broke rules.

What did they find?

• As teenagers got older, they drank alcohol more often.

• Teenagers were most impulsive and likely to break rules around age 15, and then it decreased.

• As the BDM predicted, teenagers who were impulsive at the beginning of the study were more likely to drink and break rules a few years later.

• However, being impulsive at age 16 did not predict drinking or rule-breaking problems at age 18.

• Instead, teenagers who broke rules at age 16 were more likely to have problems with drinking at age 18! They were also more impulsive at age 18.

What does this mean?

This study tells us that it's super important to help teenagers who act impulsively early on, maybe around age 13. If we can help them then, they might be less likely to have problems with drinking and breaking rules when they're older. However, once teenagers start breaking rules, we might need to focus on helping them stop those behaviors, rather than only focusing on impulsivity. This could help prevent even more serious problems later in life.