Why do potentially traumatic events (PTEs) and substance use (SU) so commonly co-occur during adolescence? Causal hypotheses developed from the study of posttraumatic stress disorder (PTSD) and substance use disorder (SUD) among adults have not yet been subject to rigorous theoretical analysis or empirical tests among adolescents with the precursors to these disorders: PTEs and SU. Establishing causality demands accounting for various factors (e.g. genetics, parent education, race/ethnicity) that distinguish youth endorsing PTEs and SU from those who do not, a step often overlooked in previous research.

Methods

We leveraged nationwide data from a sociodemographically diverse sample of youth (N = 11,468) in the Adolescent Brain and Cognitive Development Study. PTEs and substance use prevalence were assessed annually. To account for the many pre-existing differences between youth with and without PTE/SU (i.e. confounding bias) and provide rigorous tests of causal hypotheses, we linked within-person changes in PTEs and SU (alcohol, cannabis, nicotine) across repeated measurements and adjusted for time-varying factors (e.g. age, internalizing symptoms, externalizing symptoms, and friends' use of substances).

Results

Before adjusting for confounding using within-person modeling, PTEs and SU exhibited significant concurrent associations (βs = .46–1.26, ps < .05) and PTEs prospectively predicted greater SU (βs = .55–1.43, ps < .05) but not vice versa. After adjustment for confounding, the PTEs exhibited significant concurrent associations for alcohol (βs = .14–.23, ps < .05) and nicotine (βs = .16, ps < .05) but not cannabis (βs = -.01, ps > .05) and PTEs prospectively predicted greater SU (βs = .28–.55, ps > .05) but not vice versa.

Conclusions

When tested rigorously in a nationwide sample of adolescents, we find support for a model in which PTEs are followed by SU but not for a model in which SU is followed by PTEs. Explanations for why PTSD and SUD co-occur in adults may need further theoretical analysis and adaptation before extension to adolescents.

Introduction

Posttraumatic stress disorder (PTSD) and substance use disorder (SUD) often co-occur. Three causal etiologic hypotheses currently exist to explain the co-occurrence: (a) the self-medication hypothesis, positing PTSD symptoms lead to SUD as individuals use substances to cope with their psychiatric symptoms (Brady, Dansky, Sonne, & Saladin, 1998; Chilcoat & Breslau, 1998a; Khantzian, 1997); (b) the susceptibility hypothesis, positing individuals with SUD are at greater risk of developing PTSD because they may engage in more risky behavior to obtain or use substances, and significant brain-related changes make them more susceptible to developing PTSD symptoms following exposure to traumatic events (Bonin, Norton, Asmundson, Dicurzio, & Pidlubney, 2000; Sharkansky, Brief, Peirce, Meehan, & Mannix, 1999; Stewart, Conrod, Samoluk, Pihl, & Dongier, 2000); and (c) the shared liability hypothesis, positing there are shared pre-existing factors (e.g. genetics) that lead to the development of both PTSD and SUD with neither disorder actually causing the other (Breslau, Davis, Peterson, & Schultz, 1997; Cottler, Nishith, & Compton, 2001; Krueger & Markon, 2006). These three hypotheses are not mutually exclusive – rather, at question is whether, to what extent, and under which conditions each explains the co-occurrence.

The precursors to PTSD and SUD during adolescence – potentially traumatic events (PTEs) and substance use (SU) – also co-occur frequently (Tunno, Pane Seifert, Cheek, & Goldston, 2022). Unknown is whether the causal etiologic hypotheses developed to explain comorbid PTSD + SUD among adults (i.e. self-medication, susceptibility, and shared liability) can also explain the co-occurrence of these precursors. One barrier to translation is that adolescents typically do not endorse clinical symptoms of disorders, especially among community and epidemiologic samples. Studying the applicability to precursors of PTSD and SUD is important because the events that lead to both disorders (index trauma, use of alcohol or drugs) typically first occur during adolescence (Felitti et al., 1998; Frewen, Zhu, & Lanius, 2019; Schulenberg & Maslowsky, 2009). Co-occurrence may be present earlier, among PTEs and SU rather than subsequent symptoms, as previously postulated. Thus, examining concurrent and prospective associations between PTEs and SU, not symptoms is critical to establishing the etiological connection between PTEs and SU and subsequently PTSD and SUD. An examination of comorbid PTSD + SUD among adults reveals a different etiological picture than co-occurrence of PTEs + SU in youth, which is expected. Adulthood and adolescence are distinct developmental stages and confer different experiences. Thus, it is important to recognize that etiologic hypotheses supported in adults may not apply to youth. To address why traumatic events and substance use coincide during adolescence, we must employ a developmental psychopathological framework. Developmental psychopathology posits that development is determined by dynamic, reciprocal interactions among affective, cognitive, social, and biological factors over time. Thus, outcomes are products of complex interactions of factors that tip an individual's trajectory toward risk or protection during the developmental process (Cicchetti, 1993; Kerig, 2017; McLaughlin, 2014; Rutter, 2013). In this case, psychopathology is best understood with a developmental lens because we cannot assume the same hypotheses explaining comorbid PTSD + SUD in adults explain PTEs + SU in youth (Berenz, McNett, & Paltell, 2019; Hinckley & Danielson, 2022). Employing this developmental psychopathological framework, we can explore the early stages of disorder development involving the co-occurrence of PTEs + SU.

Another consideration when translating causal hypotheses for PTSD + SUD among adults to PTEs + SU among adolescents is the timescale of the hypothesized effects. The mechanism for the self-medication hypothesis can be reduced to negative reinforcement pathology among adults. PTSD symptoms cause an individual to use more substances to cope (self-medication). The mechanism for the susceptibility hypothesis is vulnerability among adults. Chronic use of substances causes long-lasting changes in brain function and structure leading to PTSD onset (susceptibility). However, these same mechanisms do not make sense when applied to adolescents. As adolescent SU is mostly opportunistic, the pattern of use is episodic rather than regular, thus reinforcement is unlikely given the timing of use (Schulenberg, Maslowsky, Patrick, & Martz, 2019). Thus, reinforcement or vulnerability pathology is not the most plausible mechanism for adolescents. Given that adolescent substance use is episodic, a mechanism of victimization may be more plausible when applying the susceptibility hypothesis to adolescents. Here, exposure to PTEs would happen while an adolescent tries to obtain substances (e.g. in a dangerous situation while getting substances illicitly) or while intoxicated (e.g. getting into a fight). As such, the timescale of effects is different across developmental periods. It can be posited that susceptibility would operate on a shorter, more acute timescale during adolescence compared to adulthood. In contrast, the self-medication hypothesis would operate on a longer, more chronic timescale as the timing of use during adolescence is more sporadic. As such, the association between PTE exposure and SU would require an increased number of trials over a prolonged period to become associated with one another. Therefore, among adolescents, the susceptibility hypothesis may unfold more quickly, and the self-medication hypothesis may unfold more slowly.

In addition, these hypotheses explaining comorbid PTSD + SUD are causal: they say not just that PTSD and SUD are correlated but what factors cause their co-occurrence. Yet the small literature that has tested both the self-medication and susceptibility hypotheses in adults has not used research designs that provide strong evidence of causality. Individuals who experience PTEs and engage in SU are influenced by various factors (e.g. race/ethnicity, sex, parent education, genetics) that complicate the direct causal link between PTEs and SU. To account for this, longitudinal data and analytic techniques must be utilized to address this complexity. Rather than the conventional approach of comparing differences between youth endorsing PTEs and SU to those who do not, a more robust causal model emerges by examining changes within each youth over time. This method establishes a stronger causal relationship since each youth serves as their own control, naturally adjusting for between-youth factors through within-youth analysis.

In summary, causal hypotheses developed to explain comorbid PTSD + SUD in adults may explain the co-occurrence of PTEs + SU among youth, but this possibility has not been subject to rigorous empirical tests. Thus, we leveraged data from a nationwide sociodemographically diverse sample of youth from the Adolescent Brain and Cognitive Development (ABCD)™ Study. Leveraging longitudinal data from late childhood to midadolescence, we have the opportunity to examine the causal etiologic hypotheses at early stages of disorder development, well before symptom onset. We evaluated the concurrent and prospective causal associations between PTEs and SU by longitudinally examining within-youth changes in PTEs and SU. By doing so, we aim to parse the relative nonexclusive contributions of each causal etiological hypothesis during the developmental transition between late childhood to mid adolescence (i.e. test how PTEs and SU co-occur). That is to say, each hypothesis is not exclusive to a developmental stage, but rather, at each developmental stage, what are the influences of each etiological hypothesis?

Methods

Participants

Between 2016 and 2018, 11,876 youth aged 9–10 years old were enrolled in the ABCD® Study. Youth were recruited primarily from schools at 21 sites across the United States. Entry criteria were minimal and the sample was intended to reflect national demographics (Garavan et al., 2018). At baseline, 48% of the sample was female. 52% identified as White, 15% as Black, 20% as non-white Hispanic, 2% as Asian, and 11% as Other.

Families completed a baseline assessment at ages 9–10 years old then a follow-up assessment each year for 3 years thereafter (yielding four total waves). At every annual visit, parents and youth reported on the occurrence of PTEs and use of alcohol and drugs. See Table 1 for prevalence of PTEs and substance use across time points.

Table 1. Sample characteristics, potentially traumatic event exposure rates, and drinking prevalence

Variable	Baseline	1-year follow-up	2-year follow-up	3-year follow-up	4-year follow-up
N	11,867	11,220	10,973	10,336	4,754
Age in years	9.9 (0.62)	10.9 (0.64)	12.0 (0.67)	12.9 (0.65)	14.1 (0.68)
% female (sex at birth)	47.8
Race/Ethnicity
White	61.6
Black	15
Hispanic	20.3
Asian	2.1
Other	10.5
Household income ($ USD)	97,330 (62,334)	102,171 (62,580)	106,006 (62,608)	109,676 (62,338)	114,510 (61,986)
Any PTE		57	55	53	53
# of PTEs		1.16 (1.51)	1.09 (1.44)	1.05 (1.47)	1.07 (1.45)
Any common substance use		1.3	2.0	3.4	7.8
Any alcohol		0.3	0.5	1.3	3.6
Any cannabis		0.5	0.8	1.0	3.6
Any nicotine		0.8	1.1	2.4	5.2
Internalizing symptoms	48.45 (10.64)	48.59 (10.62)	47.69 (10.54)	47.72 (10.6)	47.44 (10.83)
Externalizing symptoms	45.73 (10.34)	45.21 (10.12)	44.40 (9.83)	44.32 (9.55)	43.41 (9.28)
Friends' substance use	0 (0)	0.12 (0.79)	0.23 (1.09)	0.38 (1.41)	0.78 (2.06)

Values presented are prevalences (percentages) at each time point for sex (% female), race/ethnicity, Any PTEs (potentially traumatic events), any common substance use, any alcohol, any cannabis, and any nicotine. Means and standard deviations are presented for age, household income, # of ptes, internalizing symptoms, externalizing symptoms, and friends' substance use.

Measures

Potentially traumatic events

At each annual visit, youth reported lifetime experiences of 33 life events [e.g. ‘someone in the family died’ to ‘saw crime or accident’; PhenX Life Events (Tiet et al., 1998)]. If youth endorsed event occurrence, they were asked whether the event occurred in the past year, whether the event was good or bad (‘mostly good’, ‘mostly bad’, ‘not applicable’, or ‘don't know’), and how much the event impacted them on a Likert scale (0 = ‘not at all’; 1 = ‘a little’; 2 = ‘some’; 3 = ‘a lot’). We operationalize PTEs in this study as events categorized as occurring in the past year and are rated as ‘mostly bad’ allowing for a more inclusive range of PTEs beyond DSM-5-TR Criterion A events for PTSD. The count of all ‘bad’ events endorsed at each time point was used within analyses. See Table S1 for the breakdown of all events for each annual visit and whether they are considered criterion A events according to the Diagnostic Statistical Manual for Mental Disorders-5th edition (American Psychiatric Association, 2013). Table S2 provides zero-order correlations between all variables of interest within the current study.

Substance use

At each annual visit, youth were interviewed by a trained research assistant about their use of alcohol, cannabis, or nicotine since they last had contact with the research team (Lisdahl et al., 2018). Youth reported on their substance use within the past 6 months as youth also completed a mid-annual visit asking about their substance use within the past 6 months. Thus, we can assess substance use, across both visits, over the past year. Youth responded with a ‘yes’ or ‘no’ for three of the most commonly used substances (alcohol, cannabis, nicotine). For alcohol, youth responded to different consumption levels (e.g. a sip of alcohol, a full drink, and more than one full drink). The youth responded to different forms of use for cannabis (e.g. puffing, flower, edibles, concentrates, etc.) and nicotine (e.g. cig, e-cig, hookah, pipe, chew, etc.). We created dichotomous indicators from responses of whether the youth had used alcohol, cannabis, a nicotine product, or any thereof. An indicator for common substance use combining alcohol, cannabis, nicotine (i.e. ‘yes’ for any one substance and ‘no’ for all three) was created because during adolescence, substance use is generally not substance specific. Furthermore, we also checked each individual category of use as adult comorbid PTSD + SUD literature indicates there could be differences for use of only alcohol versus use of other substances (Bailey & Stewart, 2014).

Time-varying covariates

We also measured covariates with the expectation they would change over time (i.e. time-varying) and are associated with PTEs and SU; this would entail that they should be adjusted for during analysis. For example, given the longitudinal design, youth aged as the study progressed with more opportunities for PTE exposure and SU, thus age needed to be adjusted as a time-varying covariate within the model. As such, age, internalizing symptoms (assessed by Child Behavior Checklist from the Achenbach System of Empirically Based Assessment [ASEBA]; Achenbach & Rescorla, 2001, 2003), externalizing symptoms (assessed by the Child Behavior Checklist from the ASEBA; Achenbach & Rescorla, 2001, 2003), and friends' use of substances (assessed by Youth Substance Use Attitudes; Johnston, O'Malley, Miech, Bachman, & Schulenberg, 2017) were included as time-varying covariates in the concurrent and prospective models.

Data analysis

Analyses were performed in R v4.2.1 (Croissant & Millo, 2008; Lumley, 2020; R Core Team, 2022) and MPlus v8.2 (Muthén & Muthén, 2017) and were not preregistered. We test both concurrent and prospective associations between PTEs and SU. In each case, we start by fitting the nonadjusted between-person regression models (commonly used in literature) that do not account for known factors confounding the association between PTEs and SU. Next, we conduct more rigorous tests by fitting a within-person regression model that controls for the confounding bias introduced by all factors stable from year-to-year (e.g. genetics, race, parent education). We fit fixed effects regression models (Allison, 2009) to test concurrent associations, and we fit random-intercept cross-lagged panel models (RI-CLPMs; Hamaker, Kuiper, & Grasman, 2015) to test prospective associations with the cross-lagged effects constrained to equality over time. Fixed effects regression and RI-CLPMs each test the within-person association between intraindividual deviations along two variables (Hamaker et al., 2015; Mulder & Hamaker, 2021). The RI-CLPM is likely more familiar to psychologists but unsuitable for testing concurrent associations (see Figure S1 for depiction of exact model fitted for analyses). Because these models examine whether PTEs and SU are correlated within a given adolescent over time, the observed associations cannot possibly be explained by differences between adolescents on any potential confounder that is stable from observation to observation (e.g. genetics, race, parent education). To obtain further control over confounding and thus strengthen our test, we also adjusted for the four time-varying covariates described in the previous section (age, internalizing symptoms, externalizing symptoms, and friends' substance use). Here, the cross-lagged paths from PTEs to SU and vice versa provide tests of the self-medication and susceptibility hypotheses, respectively. In the case that an association between PTEs and substance use still exists despite no evidence emerging for self-medication or susceptibility, then this covariation must be explained by shared liability. As such, the test for the shared liability comes from a lack of support for the other two hypotheses. Additionally, comparing the random intercepts from RI-CLPMs provides a test of shared liability as the random intercept within those models represents the between-youth variance accounted for by shared time-invariant factors.

To study developmental impacts, we used likelihood ratio tests. We compared RI-CLPMs with full constraints (all cross-lagged PTEs-SU paths equal) to models with one hypothesis's paths freed (e.g. PTEs to subsequent SU for self-medication). Likelihood ratio tests assess model fits by parameter changes. Here, we make only one change: constraint on cross-lagged paths for one hypothesis. By comparing each hypothesis's unconstrained paths to the fully constrained model, we capture evolving associations as youth age. However, no significant differences in model fit emerged, indicated by robust, nonsignificant likelihood ratio tests (see Table 2). As a result, all cross-lagged paths remained constrained to equality in models.

Table 2. Coefficients and standard errors for each cross-lagged path in the unconstrained RI-CLPMs with time-varying covariates for the self-medication and susceptibility hypotheses

	Coef. (SE)	Coef. (SE)	Coef. (SE)	Coef. (SE)
	Common substance use	Alcohol	Cannabis	Nicotine
Self-medication
WX1 → WY2	0.30 (0.18)	0.18 (0.10)	0.23 (0.13)	0.22 (0.16)
WX2 → WY3	0.79 (0.22)***	0.43 (0.14)**	0.47 (0.15)**	0.62 (0.19)**
WX3 → WY4	1.32 (0.39)**	0.73 (0.29)*	1.02 (0.31)**	0.60 (0.31)
Likelihood ratio test results	0.09	0.30	0.18	0.39
Susceptibility
WY1 → WX2	−0.03 (0.25)	–	0.08 (0.32)	0.17 (0.39)
WY2 → WX3	−0.25 (0.14)	–	−0.22 (0.19)	−0.37 (0.23)
WY3 → WX4	0.17 (0.17)	–	−0.21 (0.31)	0.23 (0.22)
Likelihood ratio test results	0.41	–	0.92	0.37

WX = potentially traumatic event prevalence; WY = substance use outcome prevalence (i.e. Common Substance Use, Alcohol, Nicotine, Cannabis). Likelihood ratio test results are from comparing the adjusted model with covariates not constraining the cross-lagged paths testing each hypothesis to the fully constrained adjusted model with covariates presented in Table 3. Coefs, SEs, and likelihood ratio test results are not presented for the nonconstrained susceptibility model due to lack of model fit.
***p < 0.001; **p < 0.01; *p < 0.05.

Lastly, given 34% of the observations in the current study were collected during the COVID-19 pandemic as defined between the dates of March 19, 2020 to May 11, 2023,¹ we created a variable to indicate whether an observation was collected during the pandemic (score of 1) or outside of the pandemic (score of 0). After fitting concurrent and prospective fixed effects panel models, we found no statistically significant differences when including the pandemic time-varying covariate versus without, thus, we report results of models not including the time-varying covariate for within pandemic observations.

Results

Findings are organized into: (a) a section for each etiologic hypothesis (self-medication, susceptibility, shared liability) and then (b) within each section, by the timescale of association (concurrent, prospective). In each case, we first present the results of a nonadjusted regression model then the results of a within-person model accounting for confounding factors (see Table 3 for findings across all models).

Table 3. Nonadjusted and adjusted regression model results for the causal etiologic hypotheses

	Nonadjusted regression	Adjusted model	Adjusted model with covariates^b	Nonadjusted regression	Adjusted model	Adjusted model with covariates
	Concurrent^a			Prospective
	Coef. (SE)	Coef. (SE)	Coef. (SE)	Coef. (SE)	Coef. (SE)	Coef. (SE)
Self-medication hypothesis^c
PTEs^d → Common substance use^e^,^f	1.26 (0.11)***	0.38 (0.08)***	0.23 (0.08)**	1.43 (0.12)***	0.81 (0.15)***	0.55 (0.15)***
PTEs → Drinks^d	0.46 (0.07)***	0.22 (0.05)***	0.14 (0.05)**	0.55 (0.07)***	0.31 (0.08)***	0.28 (0.08)**
PTEs → Cannabis^d	0.58 (0.07)***	0.08 (0.05)	-0.01 (0.05)	0.77 (0.09)***	0.48 (0.11)***	0.38 (0.10)***
PTEs → Nicotine^d	0.92 (0.09)***	0.30 (0.06)***	0.16 (0.06)*	1.01 (0.10)***	0.56 (0.13)***	0.36 (0.13)**
Susceptibility hypothesis
Common Substance Use^b^,^d → PTEs	0.96 (0.08)***	0.23 (0.05)***	0.15 (0.05)**	0.06 (0.07)	0.20 (0.12)	−0.03 (0.11)
Drinks^d → PTEs	0.97 (0.13)***	0.33 (0.08)***	0.23 (0.08)**	−0.20 (0.10)	0.26 (0.18)	–^g
Cannabis^d → PTEs	1.14 (0.12)***	0.12 (0.08)	-0.01 (0.08)	−0.08 (0.09)	0.03 (0.18)	−0.14 (0.18)
Nicotine^d → PTEs	1.07 (0.09)***	0.28 (0.06)***	0.16 (0.06)*	0.16 (0.09)	0.34 (0.16)*	0.02 (0.16)
Shared liability hypothesis^h
PTEs & common substance use	–	–	–	–	0.23**	0.27***
PTEs & Drinks	–	–	–	–	0.05	0.06
PTEs & cannabis	–	–	–	–	–ⁱ	0.47
PTEs & nicotine	–	–	–	–	0.19**	0.22***

Table displays the unstandardized coefficients and standard errors for the relevant paths from each model fitted for the Self-Medication and Susceptibility Hypotheses. Correlations are presented for the Shared Liability Hypothesis. Coefficients in the prospective models refer to the average effect of PTEs on sSU or vice versa across four-time assessments. The nonadjusted regression model is a standard generalized linear model. The concurrent adjusted model is a fixed effects panel model and prospective adjusted model is a random-intercept cross-lagged panel model (RI-CLPM), both of which control for variance accounted by time-invariant confounds to provide a more robust estimate of how PTEs and SU are associated. The concurrent adjusted model with covariates is a fixed effects panel model with time-varying covariates added and the prospective adjusted model with covariates is a RI-CLPM with time-varying covariates included. These models control for the variance accounted for by both time-invariant and time-varying covariates to provide even a more robust estimate of the relationship between PTEs and substance use.
^a The concurrent models are identical for self-medication and susceptibility hypotheses, except the predictor and outcome are reversed in the regression. We report coefficients estimated both ways such that there is an interpretable coefficient to quantify the association in either unit.
^b Time-varying covariates are child age, internalizing symptoms, externalizing symptoms, and friends' substance use.
^c SU indicators were multiplied by 100 so coefficients are in units of percentage points for all self-medication hypothesis models. For example, if a participant endorsed drinking, they would have a score of 1. This score was multiplied by 100 to create a prevalence of drinking, so that individual would now have a score of 100.
^d PTEs are a sum score of all items endorsed within the past year and rated as ‘mostly bad’ on the PhenX Life Events Checklist [range: 0–33].
^e Common substance use is defined by prevelance of any of the following: full drink, cannabis use, and nicotine use.
^f All substance variables had the midannual visit values added to the annual follow-up visit data to capture endorsements of substance use across multiple assessments within the same year.
^g Values not presented due to lack of model fit (see results section for more detail).
^h Correlations are only presented for the prospective adjusted model and adjusted model with covariates as correlation value is extracted from the association between the latent variable random intercepts within the RI-CLPM fitted to test prospective associations and is not available in standard linear regressions and fixed effects panel models.
ⁱ Value reported in model was implausible and uninterpretable and is not reported.
***p < 0.001; **p < 0.01; *p < 0.05.

Self-medication hypothesis: PTEs causes SU

Concurrent tests of the self-medication hypothesis

When fitting a nonadjusted regression (i.e. not adjusting for between-youth factors), there was a significant positive concurrent association between PTEs and common substance use (β = 1.26, SE = 0.11, p < .0001) with the pattern holding for the three substances within the broader category: alcohol (β = 0.46, SE = 0.07, p < .0001), cannabis (β = 0.58, SE = 0.07, p < .0001), and nicotine (β = 0.92, SE = 0.09, p < .0001). However, within our fixed effects panel model (i.e. a more rigorous test adjusting for between-youth factors), we saw a significant positive concurrent association between PTEs and common substance use (β = 0.38, SE = 0.08, p < .0001), alcohol (β = 0.22, SE = 0.05, p < .0001), and nicotine (β = 0.30, SE = 0.06, p < .0001) but not cannabis (β = 0.08, SE = 0.05, p = .12). For full results, see Table 3.

Prospective tests of the self-medication hypothesis

When fitting a nonadjusted regression (i.e. not adjusting for between-youth time-invariant and variant factors), prior PTE exposure positively and significantly predicted subsequent common substance use (β = 1.43, SE = 0.12, p < .0001) with the pattern holding for the individual substances: alcohol (β = 0.55, SE = 0.07, p < .0001), cannabis (β = 0.77, SE = 0.09, p < .0001), nicotine (β = 1.01, SE = 0.10, p < .0001). Within a random-intercept cross-lagged panel model (RI-CLPM), which adjusted for between-youth time-invariant factors to parse out the within-youth variance over time, we saw a significant prediction of common substance from prior PTE exposure (β = 0.81, SE = .15, p < .0001). This prospective pattern remained significant for alcohol (β = 0.31, SE = .08, p < .0001), cannabis (β = 0.48, SE = .11, p < .0001), and nicotine (β = 0.56, SE = .13, p < .0001). Furthermore, when adjusting for time-varying covariates in addition to the time-invariant between-youth factors in the RI-CLPM, the prospective association remained significant across the waves where PTEs lead to common substance use (β = 0.55, SE = .15, p < .0001). This pattern of significance held for each individual substance: alcohol (β = 0.27, SE = 0.08, p = .001), cannabis (β = 0.38, SE = 0.10, p < .0001), and nicotine (β = 0.36, SE = .13, p = .005). For full results, see Table 3.

Susceptibility hypothesis: SU causes PTEs

The susceptibility hypothesis posits that SUD leads to PTSD. In our operationalization, SU leads to PTEs. As such, all results in this section are framed with SU as the predictor and PTEs as the outcome.

Concurrent tests of the susceptibility hypothesis

When fitting a nonadjusted regression (i.e. not adjusting for between-youth factors), there was a significant positive concurrent association between common substance use and PTEs (β = 0.96, SE = 0.08, p < .0001) with the pattern holding for the three substances within the broader category: alcohol (β = 0.97, SE = 0.13, p < .0001), cannabis (β = 1.14, SE = 0.12, p < .0001), and nicotine (β = 1.07, SE = 0.09, p < .0001). However, when fitting a fixed effects panel model (i.e. a more rigorous test adjusting for between-youth factors), we saw a significant positive concurrent association between common substance use and PTEs (β = 0.23, SE = 0.05, p < .0001) with the pattern holding across the individual substances for alcohol (β = 0.33, SE = 0.08, p < .0001) and nicotine (β = 0.28, SE = 0.06, p < .0001) but not cannabis (β = 0.12, SE = 0.08, p = .12). For full results, see Table 3.

Prospective tests of the susceptibility hypothesis

When fitting a nonadjusted regression (i.e. not adjusting for between-youth time-invariant and variant factors), prior common substance use did not significantly predict subsequent PTE exposure (β = 0.06, SE = 0.07, p = .34) with nonsignificance holding for alcohol (β = −0.20, SE = 0.10, p = .052), cannabis (β = −0.08, SE = 0.09, p = .38), and nicotine (β = 0.16, SE = 0.09, p = .06). Fitting a random-intercept cross-lagged panel model (RI-CLPM), which adjusted for between-youth time-invariant factors to parse out the within-youth variance over time, we saw no significant prediction of PTE exposure from prior common substance use across waves (β = 0.20, SE = .12, p = .09). This nonsignificant prospective pattern held for alcohol (β = 0.26, SE = .18, p = .16) and cannabis (β = 0.03, SE = .18, p = .85), but was significant for nicotine (β = 0.34, SE = .16, p = .04). Furthermore, when adjusting for time-varying covariates in addition to the time-invariant between-youth factors in the RI-CLPM, prior common substance use did not lead to PTEs (β = −0.03, SE = .11, p = .81). This pattern of nonsignificance held for each individual substance: alcohol,² cannabis (β = −0.14, SE = .18, p = .42), and nicotine (β = 0.02, SE = .16, p = .89). For full results, see Table 3.

Tests of shared liability hypothesis: shared factors cause PTEs and SU

The shared liability posits that shared underlying factors lead to PTSD and SUD. In our operationalization, time-invariant factors that differ between youth lead to PTEs and SU. As such, in the current section, we presented correlations between the random intercepts created in the prospective adjusted model with covariates to report the relative contribution of the shared liability hypothesis within our results. There was significant contribution of the shared liability hypothesis between PTEs and common substance use (r = .27, SE = .05, p < .0001) and PTEs and nicotine (r = .22, SE = .05, p < .0001) but not PTEs and alcohol (r = .06, SE = .04, p = .13) or PTEs and cannabis (r = .47, SE = 0.28, p = .10). Despite the higher coefficient for cannabis use, the relationship was still nonsignificant due to the higher standard error, given greater variance in cannabis use among youth in our sample.

Discussion

This study probed whether causal hypotheses developed to explain the co-occurrence of PTSD and SUD in adults extended to explain co-occurrence of PTEs and SU in adolescence. When tested rigorously in a nationwide sample of adolescents, adjusting for between-youth time-invariant and time-varying confounds, we find results consistent with the self-medication and partially consistent with the shared liability hypotheses, but not the susceptibility hypothesis for youth from late childhood to midadolescence. Few studies have examined the causal etiologic associations between trauma and substance use generally. Prior literature presents mixed support for the causal etiologic hypotheses among adults. For self-medication, PTSD symptoms predict SU (Bountress et al., 2019; Hicks et al., 2022) and SUD symptoms (Berenz et al., 2017; Chilcoat & Breslau, 1998a, 1998b; Haller & Chassin, 2014; Rappaport, Cusack, Sheerin, & Amstadter, 2021). For susceptibility, SUD symptoms predict PTEs (Bountress et al., 2019; Haller & Chassin, 2014) and SUD symptoms predict PTSD symptoms (Berenz et al., 2017; mediated by PTE exposure: Hicks et al., 2022). Overall, slightly more support for the self-medication hypothesis exists within the literature, which our study adds further to these findings. However, given the incorporation of symptoms and differences in analytic techniques (e.g. mediation), it is difficult to directly compare current results with prior literature. Furthermore, we found partial support for the shared liability hypothesis among our analyses.

There are multiple reasons why our results may differ from those of prior studies. One, we examined precursor event co-occurrences among adolescents, not adults. Two, we examined precursor event co-occurrence exclusively with no incorporation of symptoms. Three, our analytic techniques adjusted for pre-existing between-youth differences and within-youth time-varying confounds. Overall, it is highly likely that there are developmental differences in the comorbid presentation of precursor event (co-occurrences between adults and youth.

Comparing the coefficients in models not adjusting versus adjusting for confounding (time invariant: genetics, race/ethnicity, parent education, etc. and time-varying: age, internalizing, externalizing, friends' substance use), the effect size shrunk across both hypotheses for all substance categories concurrently (self-medication range: 70%–102%; susceptibility range: 76%–101%) and prospectively (self-medication range: 49%–64%; susceptibility range: 75%–150%). Despite the decrease in effect size with the adjustment for causal factors, the results imply the hypotheses still hold but explain much less of the co-occurrence than at first glance. Attenuation in the causal effects (as observed by decreases in effect sizes) can be attributed to some of common underlying factors, especially for certain substances (common substance use and nicotine), as posited by the shared liability hypothesis. As such, the current hypotheses need to be elaborated on with additional explanations of the casual mechanisms of self-medication, susceptibility, and shared liability beyond a simple direct relationship as they are currently posited.

The conclusion that PTEs significantly predict subsequent SU under nonadjusted and adjusted designs aligns with the self-medication hypothesis via negative reinforcement, a delayed cyclical process. Psychological distress from PTE exposure drives SU for coping, which becomes negatively reinforced with PTEs, leading to co-occurrence over time. Our annual measurements captured the negative reinforcement consistent with the self-medication hypothesis over time, but not the susceptibility hypothesis. SU did not predict subsequent PTE exposure but they were significantly associated concurrently, perhaps implying that SU leads to PTEs acutely among adolescents. This would be consistent with victimization, rather than the vulnerability mechanism as posited within the susceptibility hypothesis. Our assessments might miss this due to their annual interval (missing the initial signal), which is evidenced by the nonsignificance among the prospective nonadjusted models.

Strengths, limitations, and future directions

Within the current study, we sought to explain how etiology may differ across the life span by testing prevalent adult-based hypotheses among youth from late childhood to midadolescence, the theory for which is currently lacking. Using youth data on PTEs and SU, we explored how adult-based causal etiologic hypotheses apply differently to youth and to the early stage of disorder development characterized by the co-occurrence of PTEs and SU, rather than PTSD and SUD symptoms. We also addressed the timescale of the hypotheses, or lack thereof, within the theory by testing associations with an interval of 1 year between PTEs and SU. Lastly, we treat the etiological hypotheses as causal, as they are implied according to the theory, and developed analytical models that reflect that causality.

This study had six limitations. The first four limitations are related to participant data whereas the last two are related to data analysis. First, low prevalence rates of SU, particularly alcohol, could explain why we did not see evidence for the susceptibility hypothesis as youth would have fewer occasions of possible victimization while intoxicated. Trauma may impact alcohol use more specifically than other SU given that alcohol is the easiest substance to access while underage (Kirby & Barry, 2012; Wagenaar et al., 1993), should youth have a coping motive for use. Second, our sample spanned from 9.0 to 15.8 years old, but the processes explaining the co-occurrence of PTEs and SU differ in late adolescence. SU is more common in later adolescence compared to early/midadolescence. This provides more chances for the victimization mechanism in the susceptibility hypothesis. Additionally, Many adolescents have already processed past PTEs by this stage. Consequently, negative reinforcement might be sustained rather than emerging. Thus, the influence of both susceptibility and self-medication could be seen during late adolescence though we hypothesize susceptibility would have a greater role to play given the developmental stage of late adolescence and the trajectory of disorder development. Future research should assess for PTEs and SU weekly along with key time-varying factors, such as acute physiological and emotional distress following PTEs that may predispose some youth to seek out substances as a coping method to elucidate whether the mechanism of negative reinforcement can be observed on an acute timeline. Third, the sample was biased toward families of higher socioeconomic status (SES). This limits the generalizability of the findings. SES is predictive of more trauma exposure during childhood, however, is protective against early initiation of substance use, especially alcohol. So while the ABCD Study® has a large sample of lower SES families than most studies due to sheer sample size, future studies need to examine the moderating impact of SES on these causal hypotheses. Fourth, five new PTEs were added to the study protocol at the 3-year follow-up (see Table S1). As such, the total number of PTEs a youth could endorse increased from 28 to 33. This may have impacted the results.

Among the limitations related to data analysis, first, we collapsed across PTEs, potentially resulting in nonsignificance for PTEs but there could be significant effects for some types of PTEs, especially those that are criterion A events (e.g. childhood sexual abuse) compared to others (e.g. emotional neglect). Second, we did not adjust for all possible confounds within our analyses – while our models were a significant improvement over prior work, the observed associations between PTEs and SU still may not be entirely causal. It is possible that there are time-varying confounds (e.g. acute physiological distress, coping motives, deliquency, etc.) that influence the relationship between PTEs and SU that we did not measure or could not include within our analyses. Moreover, considering the shared liability hypothesis, since the RI-CLPM addresses time-invariant confounding factors by introducing random intercepts, it remains uncertain which specific time-invariant factors are encompassed by these random intercepts. Subsequent research should meticulously assess both time-invariant and time-varying confounding variables, enabling intentional adjustments in analyses. Specifically exploring personal and environmental time-invariant factors (e.g. sex, race/ethnicity, parent education, etc.) will help fully characterize the causal etiologic hypotheses and differences in their presentation based on these factors. Time-varying factors such as those included in the current study should be examined as mediators along the cross-lagged paths between PTEs and SU to see whether these factors shed light on the mechanisms of action for the causal hypotheses attempting to explain the comorbidity between PTEs and SU among youth. Furthermore, future research should also examine the interactions between time-invariant and time-varying confounds such as sex-internalizing/externalizing symptoms, SES-internalizing/externalizing symptoms to establish a more clear narrative of how these factors influence the causal etiology of co-occurring PTEs and SU. This will enhance the establishment of causal connections between PTEs and SU, aligned with shared underlying vulnerabilities.

Conclusion

We examined the etiology of comorbid PTSD + SUD at an early stage of disorder development, well before symptom onset, during the co-occurrences of PTEs + SU. We showed that causal hypotheses developed to explain the comorbidity of PTSD and SUD in adults extended only partially to explain the co-occurrence of PTEs and SU in adolescents. In prospective models, we found evidence consistent with the self-medication and shared liability hypotheses. Thus, explanations for why PTSD and SUD co-occur in adults may need further theoretical analysis and adaptation before extension to prediagnostic co-occurrence among adolescents or that adolescents may need their own unique models unique from adults. Adaptation and/or development of models will require a developmental psychopathological framework to explain co-occurring PTEs and SU across an individuals' developmental stages but also disorder development.

Acknowledgements

Data used in the preparation of this article were obtained from the Adolescent Brain Cognitive DevelopmentSM (ABCD) Study (https://abcdstudy.org), held in the NIMH Data Archive (NDA). This is a multisite, longitudinal study designed to recruit more than 10,000 children ages 9–10 years old and follow them over 10 years into early adulthood. The ABCD Study® is supported by the National Institutes of Health and additional federal partners under award numbers U01DA041048, U01DA050989, U01DA051016, U01DA041022, U01DA051018, U01DA051037, U01DA050987, U01DA041174, U01DA041106, U01DA041117, U01DA041028, U01DA041134, U01DA050988, U01DA051039, U01DA041156, U01DA041025, U01DA041120, U01DA051038, U01DA041148, U01DA041093, U01DA041089, U24DA041123, and U24DA041147. A full list of supporters is available at https://abcdstudy.org/federal-partners.html. Additional support for this work was made possible from supplements to U24DA041123 and U24DA041147, the National Science Foundation (NSF 2028680), and Children and Screens: Institute of Digital Media and Child Development Inc. A listing of participating sites and a complete listing of the study investigators can be found at https://abcdstudy.org/Consortium_Members.pdf. ABCD consortium investigators designed and implemented the study and/or provided data but did not necessarily participate in analysis or writing of this report. This manuscript reflects the views of the authors and may not reflect the opinions or views of the NIH or ABCD consortium investigators. The ABCD data used in this report came from the ABCD 5.1 data release (DOI: 10.15154/z563-zd24). DOIs can be found at https://nda.nih.gov/study.html?id=2313. For research presented within the current article, the first author was supported by the Banting Postdoctoral Fellowship and the California Department of Cannabis Control. The last author was supported by the National Institute on Health (DA055935, DA058314, AA030197) and a Brain and Behavior Research Foundation (BBRF) Young Investigator Award. The authors have declared that they have no competing or potential conflicts of interest.

Endnotes

1 These dates were chosen to reflect the start and end of the pandemic for analytic purposes as March 19, 2020 is when California (the first state to do so) initiated the first lockdown in response to the COVID-19 pandemic within the United States. May 11, 2023 represents the date when the United States federal order considering COVID-19 a public health emergency expired.

2 Statistical results are not reported for alcohol for fitting a RI-CLPM with covariates due to lack of model fit.

Key points

Current causal etiologic hypotheses for comorbid posttraumatic stress disorder and substance use disorders were developed to explain symptoms among adults.
Unknown is whether these causal etiologic hypotheses can also explain the co-occurrence of precursors to posttraumatic stress disorder (PTSD) and SUD among youth: potentially traumatic events (PTES) and substance use (SU).
We test these causal etiologic hypotheses among a nationwide sociodemographically diverse sample of youth using rigorous models adjusting for stable between-youth and within-youth time-varying confounds.
We find prospective evidence supporting the self-medication hypothesis but not the susceptibility hypothesis.
Thus, explanations for why PTSD and SUD co-occur in adults may need further theoretical analysis and adaptation before extension to adolescents.

Open Research

Data availability statement

Data used in the preparation of this article were obtained from the Adolescent Brain Cognitive Development (ABCD) Study^TM (https://abcdstudy.org), held in the NIMH Data Archive (NDA). The ABCD data used in this report came from the ABCD 5.1 data release (DOI: 10.15154/z563-zd24). DOIs can be found at https://nda.nih.gov/study.html?id=2313.

Supporting Information

References

Achenbach, T.M., & Rescorla, L.A. (2001). Manual for the ASEBA School-age Forms & Profiles. Burlington, VT: University of Vermont, Research Center for Children, Youth, & Families.
Google Scholar
Achenbach, T.M., & Rescorla, L.A. (2003). Manual for the ASEBA Adult Forms & Profiles. Burlington, VT: University of Vermont, Research Center for Children, Youth, & Families.
Google Scholar
Allison, P. (2009). Fixed effects regression models. Thousand Oaks, CA: SAGE Retrieved July 31, 2023, from https://methods.sagepub.com/book/fixed-effects-regression-models
10.4135/9781412993869
Google Scholar
American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders: DSM-5 ( 5th edn). Washington, D.C.: American Psychiatric Association.
10.1176/appi.books.9780890425596
Google Scholar
Bailey, K.M., & Stewart, S.H. (2014). Relations among trauma, PTSD, and substance misuse: The scope of the problem. In P. Ouimette & J.P. Read (Eds.), Trauma and substance abuse, causes, consequences, and treatment of comorbid disorders ( 2nd edn, pp. 11–34). Washington, D.C.: American Psychological Association.
10.1037/14273-002
Google Scholar
Berenz, E.C., McNett, S., & Paltell, K. (2019). Development of comorbid PTSD and substance use disorders. In Posttraumatic stress and substance use disorders: A comprehensive clinical handbook ( 1st edn, pp. 11–27). New York, NY: Routledge.
10.4324/9781315442648-2
Google Scholar
Berenz, E.C., Roberson-Nay, R., Latendresse, S.J., Mezuk, B., Gardner, C.O., Amstadter, A.B., & York, T.P. (2017). Posttraumatic stress disorder and alcohol dependence: Epidemiology and order of onset. Psychological Trauma Theory Research Practice and Policy, 9, 485–492.
10.1037/tra0000185
PubMedWeb of Science®Google Scholar
Bonin, M.F., Norton, G.R., Asmundson, G.J.G., Dicurzio, S., & Pidlubney, S. (2000). Drinking away the hurt: The nature and prevalence of PTSD in substance abuse patients attending a community-based treatment program. Journal of Behavior Therapy and Experimental Psychiatry, 31, 55–66.
10.1016/S0005-7916(00)00008-2
CASPubMedWeb of Science®Google Scholar
Bountress, K.E., Cusack, S.E., Sheerin, C.M., Hawn, S., Dick, D.M., Kendler, K.S., & Amstadter, A.B. (2019). Alcohol consumption, interpersonal trauma, and drinking to cope with trauma-related distress: An auto-regressive, cross-lagged model. Psychology of Addictive Behaviors, 33, 221–231.
10.1037/adb0000457
PubMedWeb of Science®Google Scholar
Brady, K.T., Dansky, B.S., Sonne, S.C., & Saladin, M.E. (1998). Posttraumatic stress disorder and cocaine dependence: Order of onset. The American Journal on Addictions, 7, 128–135.
10.1111/j.1521-0391.1998.tb00327.x
CASPubMedWeb of Science®Google Scholar
Breslau, N., Davis, G.C., Peterson, E.L., & Schultz, L. (1997). Psychiatric sequelae of posttraumatic stress disorder in women. Archives of General Psychiatry, 54, 81–87.
10.1001/archpsyc.1997.01830130087016
CASPubMedWeb of Science®Google Scholar
Chilcoat, H.D., & Breslau, N. (1998a). Investigations of causal pathways between PTSD and drug use disorders. Addictive Behaviors, 23, 827–840.
10.1016/S0306-4603(98)00069-0
CASPubMedWeb of Science®Google Scholar
Chilcoat, H.D., & Breslau, N. (1998b). Posttraumatic stress disorder and drug disorders: Testing causal pathways. Archives of General Psychiatry, 55, 913–917.
10.1001/archpsyc.55.10.913
CASPubMedWeb of Science®Google Scholar
Cicchetti, D. (1993). Developmental psychopathology: Reactions, reflections, projections. Developmental Review, 13, 471–502.
10.1006/drev.1993.1021
Web of Science®Google Scholar
Cottler, L.B., Nishith, P., & Compton, W.M. (2001). Gender differences in risk factors for trauma exposure and post-traumatic stress disorder among inner-city drug abusers in and out of treatment. Comprehensive Psychiatry, 42, 111–117.
10.1053/comp.2001.21219
CASPubMedWeb of Science®Google Scholar
Croissant, Y., & Millo, G. (2008). Panel data econometrics in R: The plm package. Journal of Statistical Software, 27, 1–43.
10.18637/jss.v027.i02
Web of Science®Google Scholar
Felitti, V.J., Anda, R.F., Nordenberg, D., Williamson, D.F., Spitz, A.M., Edwards, V., … Marks, J.S. (1998). Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults: The adverse childhood experiences (ACE) study. American Journal of Preventive Medicine, 14, 245–258.
10.1016/S0749-3797(98)00017-8
CASPubMedWeb of Science®Google Scholar
Frewen, P., Zhu, J., & Lanius, R. (2019). Lifetime traumatic stressors and adverse childhood experiences uniquely predict concurrent PTSD, complex PTSD, and dissociative subtype of PTSD symptoms whereas recent adult non-traumatic stressors do not: Results from an online survey study. European Journal of Psychotraumatology, 10, 1606625.
10.1080/20008198.2019.1606625
PubMedWeb of Science®Google Scholar
Garavan, H., Bartsch, H., Conway, K., Decastro, A., Goldstein, R.Z., Heeringa, S., … Zahs, D. (2018). Recruiting the ABCD sample: Design considerations and procedures. Developmental Cognitive Neuroscience, 32, 16–22.
10.1016/j.dcn.2018.04.004
CASPubMedWeb of Science®Google Scholar
Haller, M., & Chassin, L. (2014). Risk pathways among traumatic stress, posttraumatic stress disorder symptoms, and alcohol and drug problems: A test of four hypotheses. Psychology of Addictive Behaviors, 28, 841–851.
10.1037/a0035878
PubMedWeb of Science®Google Scholar
Hamaker, E.L., Kuiper, R.M., & Grasman, R.P.P.P. (2015). A critique of the cross-lagged panel model. Psychological Methods, 20, 102–116.
10.1037/a0038889
PubMedWeb of Science®Google Scholar
Hicks, T.A., Bountress, K.E., Adkins, A.E., Svikis, D.S., Gillespie, N.A., Dick, D.M., & Amstadter, A.B. (2022). A longitudinal mediational investigation of risk pathways among cannabis use, interpersonal trauma exposure, and trauma-related distress. Psychological Trauma: Theory Research Practice and Policy, 15(6), 969–978.
10.1037/tra0001207
PubMedWeb of Science®Google Scholar
Hinckley, J.D., & Danielson, C.K. (2022). Elucidating the neurobiologic etiology of comorbid PTSD and substance use disorders. Brain Sciences, 12(9), 1166.
10.3390/brainsci12091166
CASPubMedWeb of Science®Google Scholar
Johnston, L.D., O'Malley, P.M., Miech, R.A., Bachman, J.G., & Schulenberg, J.E. (2017). Monitoring the future national survey results on drug use, 1975–2016: Overview, key findings on adolescent drug use. Ann Arbor, MI: Institute for Social Research, The University of Michigan.
Google Scholar
Kerig, P.K. (2017). Posttraumatic stress disorder in childhood and adolescence: A developmental psychopathology perspective. New York, NY: Momentum Press.
Google Scholar
Khantzian, E. (1997). The self-medication hypothesis of substance use disorders: A reconsideration and recent applications. Harvard Review of Psychiatry, 4, 231–244.
10.3109/10673229709030550
CASPubMedWeb of Science®Google Scholar
Kirby, T., & Barry, A.E. (2012). Alcohol as a gateway drug: A study of US 12th graders. Journal of School Health, 82, 371–379.
10.1111/j.1746-1561.2012.00712.x
CASPubMedWeb of Science®Google Scholar
Krueger, R.F., & Markon, K.E. (2006). Reinterpreting comorbidity: A model-based approach to understanding and classifying psychopathology. Annual Review of Clinical Psychology, 2, 111–133.
10.1146/annurev.clinpsy.2.022305.095213
PubMedWeb of Science®Google Scholar
Lisdahl, K.M., Sher, K.J., Conway, K.P., Gonzalez, R., Ewing, S.W.F., Nixon, S.J., … Heitzeg, M. (2018). Adolescent brain cognitive development (ABCD) study: Overview of substance use assessment methods. Developmental Cognitive Neuroscience, 32, 80–96.
10.1016/j.dcn.2018.02.007
PubMedWeb of Science®Google Scholar
Lumley, T. (2020). Survey: Analysis of complex survey samples.
Google Scholar
McLaughlin, K.A. (2014). Developmental Epidemiology. In M. Lewis & K.D. Rudolph (Eds.), Handbook of developmental psychopathology (pp. 87–107). Boston, MA: Springer US.
10.1007/978-1-4614-9608-3_5
Google Scholar
Mulder, J.D., & Hamaker, E.L. (2021). Three extensions of the random intercept cross-lagged panel model. Structural Equation Modeling: A Multidisciplinary Journal, 28, 638–648.
10.1080/10705511.2020.1784738
Web of Science®Google Scholar
Muthén, L.K., & Muthén, B.O. (2017). Mplus user's guide ( 8th edn, p. 950). Los Angeles, CA: Muthén & Muthén.
Google Scholar
R Core Team. (2022). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing.
Google Scholar
Rappaport, L.M., Cusack, S.E., Sheerin, C.M., & Amstadter, A.B. (2021). Intraindividual association of PTSD symptoms with binge drinking among trauma-exposed students. Journal of Counseling Psychology, 68, 571–581.
10.1037/cou0000543
PubMedWeb of Science®Google Scholar
Rutter, M. (2013). Developmental psychopathology: A paradigm shift or just a relabeling? Development and Psychopathology, 25, 1201–1213.
10.1017/S0954579413000564
PubMedWeb of Science®Google Scholar
Schulenberg, J.E., & Maslowsky, J. (2009). Taking substance use and development seriously: Developmentally distal and proximal influences on adolescent drug use. Monographs of the Society for Research in Child Development, 74, 121–130.
10.1111/j.1540-5834.2009.00544.x
PubMedWeb of Science®Google Scholar
Schulenberg, J.E., Maslowsky, J., Patrick, M.E., & Martz, M. (2019). Substance use in the context of adolescent development. In R.A. Zucker & S.A. Brown (Eds.), The Oxford handbook of adolescent substance use (pp. 19–36). New York, NY: Oxford University Press.
Google Scholar
Sharkansky, E.J., Brief, D.J., Peirce, J.M., Meehan, J.C., & Mannix, L.M. (1999). Substance abuse patients with posttraumatic stress disorder (PTSD): Identifying specific triggers of substance use and their associations with PTSD symptoms. Psychology of Addictive Behaviors, 13, 89–97.
10.1037/0893-164X.13.2.89
Web of Science®Google Scholar
Stewart, S.H., Conrod, P.J., Samoluk, S.B., Pihl, R.O., & Dongier, M. (2000). Posttraumatic stress disorder symptoms and situation-specific drinking in women substance abusers. Alcoholism Treatment Quarterly, 18, 31–47.
10.1300/J020v18n03_04
Web of Science®Google Scholar
Tiet, Q.Q., Bird, H.R., Davies, M., Hoven, C., Cohen, P., Jensen, P.S., & Goodman, S. (1998). Adverse life events and resilience. Journal of the American Academy of Child & Adolescent Psychiatry, 37, 1191–1200.
10.1097/00004583-199811000-00020
CASPubMedWeb of Science®Google Scholar
Tunno, A.M., Pane Seifert, H.T., Cheek, S., & Goldston, D.B. (2022). The intersection of trauma and substance use in adolescent populations. In R. Geffner, J.W. White, L.K. Hamberger, A. Rosenbaum, V. Vaughan-Eden, & V.I. Vieth (Eds.), Handbook of interpersonal violence and abuse across the lifespan: A project of the National Partnership to end interpersonal violence across the lifespan (NPEIV) (pp. 1533–1554). Cham: Springer International.
10.1007/978-3-319-89999-2_34
Google Scholar
Wagenaar, A.C., Finnegan, J.R., Wolfson, M., Anstine, P.S., Williams, C.L., & Perry, C.L. (1993). Where and how adolescents obtain alcoholic beverages. Public Health Reports, 108, 459–464.
CASPubMedWeb of Science®Google Scholar

Citing Literature

Volume65, Issue10

October 2024

Pages 1388-1397

Do traumatic events and substance use co-occur during adolescence? Testing three causal etiologic hypotheses

Abstract

Background

Methods

Results

Conclusions

Introduction

Methods

Participants

Measures

Potentially traumatic events

Substance use

Time-varying covariates

Data analysis

Results

Self-medication hypothesis: PTEs causes SU

Concurrent tests of the self-medication hypothesis

Prospective tests of the self-medication hypothesis

Susceptibility hypothesis: SU causes PTEs

Concurrent tests of the susceptibility hypothesis

Prospective tests of the susceptibility hypothesis

Tests of shared liability hypothesis: shared factors cause PTEs and SU

Discussion

Strengths, limitations, and future directions

Conclusion

Acknowledgements

Endnotes

Key points

Open Research

Data availability statement

Supporting Information

References

Citing Literature

References

Information

RESOURCES

About Wiley Online Library

Help & Support

Opportunities

Connect with Wiley

Do traumatic events and substance use co-occur during adolescence? Testing three causal etiologic hypotheses

Abstract

Background

Methods

Results

Conclusions

Introduction

Methods

Participants

Measures

Potentially traumatic events

Substance use

Time-varying covariates

Data analysis

Results

Self-medication hypothesis: PTEs causes SU

Concurrent tests of the self-medication hypothesis

Prospective tests of the self-medication hypothesis

Susceptibility hypothesis: SU causes PTEs

Concurrent tests of the susceptibility hypothesis

Prospective tests of the susceptibility hypothesis

Tests of shared liability hypothesis: shared factors cause PTEs and SU

Discussion

Strengths, limitations, and future directions

Conclusion

Acknowledgements

Endnotes

Key points

Open Research

Data availability statement

Supporting Information

References

Citing Literature

References

Related

Information

RESOURCES