|Home | About | Journals | Submit | Contact Us | Français|
A large body of epidemiological research indicates that anxiety and mood disorders are highly comorbid with substance use disorders (SUDs). However, longitudinal research regarding their temporal relations is limited. The goal of this study was to assess whether emotional disorders (i.e., anxiety and mood disorders) predict the onset of SUDs, whether SUDs predict the onset of emotional disorders, or both.
The current study used data from baseline assessment (N = 627) and four years of follow-up assessments from the NU/UCLA Youth Emotion Project to examine this question.
In line with the self-medication hypothesis of emotional disorder/SUDs comorbidity, anxiety and unipolar mood disorders at baseline assessment were associated with later onsets of SUDs. In particular, social anxiety disorder (SAD) at baseline predicted onset of alcohol use disorders and PTSD predicted the onset of all SUDs. SUDs did not predict any anxiety or unipolar mood disorders with the exception that alcohol use disorders predicted the onset of obsessive compulsive disorder (OCD).
These findings, as well as the clinical implications and future directions for research, are discussed.
Epidemiological, cross-sectional studies have documented elevated prevalence of substance use disorders (SUDs) in individuals with emotional disorders (i.e., mood and anxiety disorders), and vice versa (e.g., Conway et al., 2006; Kessler et al., 2005a; Grant et al., 2004). This comorbidity is present across various emotional disorders and with different primary substances (e.g., Conway et al., 2006; Chilcoat & Breslau, 1998; Compton et al., 2007; Schneier et al., 2010).
Several theories of comorbidity patterns among emotional disorders and SUDs have been proposed. The “self-medication” theory suggests that individuals with emotional disorders use substances to alleviate distress (Kushner et al., 1990; Smith & Book, 2010). Indeed, drinking motives among those with mood or anxiety symptoms include escape from negative emotions and reduction of tension (Goldstein & Flett, 2009; Kuntsche et al., 2006). The theory is also supported by some of the age of onset research (e.g., Schneier et al., 2010; Kuo et al., 2006; Buckner et al., 2008) including treatment-seeking samples where anxiety disorders preceded alcohol use disorders (AUD) in the large majority of those with generalized anxiety disorder/AUD and social anxiety disorder/AUD comorbidity (Smith & Book, 2010; Terra et al., 2006).
In contrast, the substance-induced enhancement theory suggests that multiple intoxication and withdrawal experiences engender anxiety, thus contributing to anxiety disorder onset (Kushner et al., 1990; Zvolensky & Schmidt, 2003). Indeed, cocaine and marijuana use increase risk for panic attacks, and cocaine use is associated with panic disorder onset (Anthony et al., 1989). Also, some studies show SUDs confer risk for MDD and not vice versa (e.g., Fergusson et al., 2009; Rao et al., 2000). Further, neurobiological perspectives suggest that repetitive episodes of alcohol intoxication and withdrawal alter key inhibitory (GABAnergic) and excitatory (glutamatergic) neuronal circuits that modulate fear circuits, thereby inducing anxiety (Breese et al., 2005). Studies supporting either of these two theories have mostly used designs involving retrospective recall of ages of onset, leaving open the possibility of recall biases.
The current prospective study examined comorbidity of unipolar mood disorders (UMDs) and anxiety disorders (ADs) with SUDs using data from the NU/UCLA Youth Emotion Project, a prospective study of risk for emotional disorders. We tested the two models described above: (1) the self-medication theory, which predicts that emotional disorders confer risk for SUDs; and (2) the substance-induced enhancement theory, which predicts that SUDs confer risk for emotional disorders. A third possibility is that the relations of UMDs and ADs with SUDs are bi-directional.
High school juniors recruited from a school in suburban Chicago and a school in suburban Los Angeles were screened with the Neuroticism subscale from the revised Eysenck Personality Questionnaire (EPQ-R-N; Eysenck & Eysenck, 1975). Recruitment oversampled participants high in neuroticism to increase the likelihood of new onsets of emotional disorders (Clark et al., 1994; Krueger et al., 1996). There were 627 participants (68.9% female) who completed the baseline assessment. Participants were on average 16.9 years old (SD = 0.39) at baseline. The sample was 48.2% Caucasian, 15.3% Hispanic, 13.1% African-American, 4.3% Asian-American, 0.6% Native American/Pacific Islander/Alaskan Native, 13.1% multiracial, and 5.4% other racial/ethnic group. See Zinbarg et al., 2010, for more details.
Structured Clinical Interview for DSM-IV (SCID-I/NP; First et al., 2002). The SCID was used to assess for DSM-IV ADs, UMDs, and SUDs. Interviews were conducted at baseline (T1) and then annually over the next 4 years. Inter-rater reliability was high: agreement across raters was 94% for ADs, 88% for UMDs, 96% for AUDs, and 98% for non-alcohol substance use disorders (non-alcohol SUDs).
After the baseline in-person SCID, participants completed an annual SCID on the phone or in person. Among the participants who completed a baseline assessment 79.1%, 67.0%, 67.3%, and 67.3% completed the first, second, third, and fourth follow-up assessments, respectively.
Logistic regressions were used in which the DV was the presence/absence of the diagnosis being predicted at any time point during the follow-up period. SUD variables included SUDs overall, AUDs, and non-alcohol SUDs. Cases in which the participant had a T1 diagnosis for the diagnosis being predicted were excluded. See Table 1 for disorder frequencies. Models included gender as a covariate, given that gender is associated with ADs, UMDs, and SUDs (Kessler et al., 2005b).
ADs at T1 predicted SUD onset over the follow-up period, OR = 1.98, p < .05, 95% CI: 1.07–3.65. This effect was due to AUD onset (OR = 2.71, p < .01, 95% CI: 1.39–5.29) whereas ADs did not predict non-alcohol SUDs (OR = 1.66, p > .19, 95% CI: 0.77–3.56). PTSD at T1 was the only specific anxiety disorder to predict the onset of SUDs in general (in a positive direction), OR = 7.29, p < .05, 95% CI: 1.18–45.25 (all other ps > .13). Social anxiety disorder at T1 was the only specific anxiety disorder to predict AUDs (in a positive direction) (OR = 2.52, p < .05, 95% CI: 1.10–5.80) (all other ps > .27). No ADs predicted non-alcohol SUD onset (all ps > .14).
Baseline UMDs predicted SUDs onset over the follow-up, OR = 2.35, p < .01, 95% CI: 1.34–4.13. This effect was driven by prediction of AUD onset [OR = 2.25, p < .05, 95% CI: 1.18–4.28] whereas UMDs did not predict non-alcohol SUD onset (OR = 1.48, p > .29, 95% CIs: 0.72–3.05).
SUDs at T1 did not predict the onset of any ADs, either as a general class of disorders or when considering each specific anxiety disorder (ps > .17). Similarly, non-alcohol SUDs at T1 did not predict anxiety disorder onset (all ps > .12). However, AUDs at T1 were associated with increased odds of developing obsessive compulsive disorder during follow-up, OR = 9.59, p < .05, 95% CI: 1.04–88.64 (all other ps > .09).
None of the SUD predictors at baseline predicted the onset of a subsequent UMD (all ps > .17).
In line with the self-medication hypothesis, emotional disorders robustly predicted the onset of SUDs, whereas SUDs predicted the onset of emotional disorders in only one case. This is the first prospective study to our knowledge examining the role of both ADs and UMDs in the subsequent onset of SUDs. Emotional disorders predicted the onset of alcohol use disorders but not other SUDs. The heterogeneous group of drugs represented in non-alcohol SUDs may have obscured effects that would be observed for certain drugs but not others. As expected in a non-clinical adolescent sample, there were too few onsets of non-alcohol SUDs other than cannabis use disorder to examine whether emotional disorders predicted non-alcohol SUDs for specific drugs.
In contrast, the presence of SUDs generally did not confer risk for emotional disorders. Thus, these findings show stronger support for the self-medication hypothesis than the substance-induced enhancement theory. This finding conflicts with some previous research in MDD/SUD comorbidity (e.g., Fergusson et al., 2009). There may be something unique about those who develop SUDs by late adolescence, as this is earlier than the typical age of onset for SUDs. Findings may have differed with an older sample.
In contrast to other studies (e.g., Falk et al., 2008), SUDs did not predict the onset of panic disorder, even though cannabis use disorders, which have previously been shown to be associated with panic disorder onset (Kushner et al., 1990), were well-represented in this sample. However, other drugs (e.g., stimulants) that may confer risk for panic disorder were not well-represented in this sample. Also, the low incidence of panic disorder over the follow-up period may have resulted in insufficient power to detect this effect. Future research should use longer follow-up periods to assess individuals into the mid-20s when panic disorder is more likely to onset (Kessler, Berglund et al., 2005).
Additionally, PTSD was associated with increased risk of SUDs. Research investigating the directionality of PTSD-SUDs comorbidity has been mixed. Our finding adds to the literature demonstrating that PTSD precedes SUDs (Cottler et al., 1992; Breslau et al., 2003) and suggests that those with PTSD may misuse substances to reduce symptoms, but does not support the idea that those with SUDs are at greater risk for PTSD (see Chilcoat & Breslau, 1998). Future research should explore this comorbidity longitudinally with older samples to evaluate whether age of onset of the first disorder (i.e., either SUD or PTSD) moderates this relation.
In addition, social anxiety disorder was associated with increased risk of AUDs, but AUDs did not predict social anxiety disorder, thus providing support for a unidirectional association. The social anxiety and alcohol association is well-documented in the literature on anxiety disorder/SUD comorbidity (e.g., Buckner et al., 2008; Kushner et al., 1990; Conger, 1956). Individuals with social anxiety disorder may begin misusing alcohol as a safety behavior to reduce short-term anxiety during social gatherings in which alcohol is present, thereby maintaining anxiety by preventing the gathering of information to disconfirm beliefs that a negative outcome would occur in social situations if alcohol is not consumed. These individuals may develop an AUD after frequent use during social gatherings and develop beliefs that they need the alcohol in order to cope with social situations.
Although SUDs generally did not predict emotional disorder onset, AUDs predicted OCD onset. Little research has explored SUD/OCD comorbidity. Possibly, these individuals self-medicated sub-threshold OC symptoms (Fatseas et al., 2010). Indeed, examination of the data revealed that over half of the participants with baseline AUDs who developed OCD over the follow-up had OC symptoms at baseline. These results should be interpreted cautiously, because there were few onsets of OCD over follow-up and results could be driven by a few outliers, as indicated by the large 95% confidence interval.
Although important conclusions can be drawn, the study did have limitations. The mean ages in this sample by the end of follow-up were in the early 20s, which is earlier than the mean age of onset for SUDs, panic disorder, and PTSD (Kessler et al., 2005b; Compton et al., 2000). Thus, although this sample represents an ideal age for examining first onsets of most emotional disorders, it is not ideal for identifying all emotional disorders or for SUD onsets. Future research should examine this question with older ages and longer follow-up periods. Also, consistent with epidemiological findings, onsets for some ADs had low frequencies, which may have resulted in low power to detect some effects.
The current data point to the importance of early detection of emotional disorders for identifying those who may be at risk for the development of SUDs later in adolescence or young adulthood. School counselors should assess for emotional disorders and be attuned to the increased risk this poses for later SUDs. SUD prevention programs should be developed for those with emotional disorders, and emotional disorder prevention programs should assess whether these programs also reduce risk for SUDs. Finally, treatment of emotional disorders in the context of SUD treatment may not only have a positive impact on mood and anxiety symptoms but on SUD outcomes as well. Thus, typical care for SUDs should include identifying and treating comorbid emotional disorders.
Statement 1: Role of Funding Sources
The research reported and the preparation of this article was supported by National Institute of Mental Health Grants R01 MH65651 and R01 MH65652 to Michelle G. Craske, Susan Mineka, and Richard E. Zinbarg.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Statement 2: ContributorsKate Wolitzky-Taylor took part in data collection, data analysis, and a major role in manuscript preparation. Lyuba Bobova took part in manuscript preparation. Richard Zinbarg took part in study design, assisting with data analysis, and substantial editing the manuscript. Susan Mineka and Michelle Craske took part in study design and substantial editing of the manuscript. All authors have approved the final version of the manuscript.
Statement 3: Conflict of Interest
There are no conflicts of interest.
1Because of the oversampling of high neuroticism and the gender imbalance in the sample, we conducted additional analyses examining neuroticism and gender as moderators of the predictive associations. Neither neuroticism nor gender moderated any of the associations between SUDs and emotional disorders and thus are not reported in detail. Also, including gender and neuroticism as main effects did not change the overall findings.
2Two cases had a full DSM diagnosis except with CSR = 2, indicating subclinical impairment/distress, and two cases did not meet the full DSM criteria for OCD and were thus categorized as Anxiety Disorder NOS (closest to the diagnosis of OCD), and also had a CSR = 3, indicating possible but not certain threshold for reaching clinical severity.
Kate Wolitzky-Taylor, University of California-Los Angeles.
Lyuba Bobova, Northwestern University, The Family Institute at Northwestern University.
Richard E. Zinbarg, Northwestern University, The Family Institute at Northwestern University.
Susan Mineka, Northwestern University.
Michelle G. Craske, University of California-Los Angeles.