This study examined substance use among youth with treatment-resistant MDD who did not meet criteria for a substance use disorder. More than half of the sample had used a substance at least once, and approximately one-quarter reported repeated experimentation with substances (≥3 times) at baseline. Baseline substance-related impairment was associated with objective (CDRS) and subjective (BDI) ratings of depression. Substance-related impairment was also associated with older age, history of physical or sexual abuse, greater family conflict, hopelessness, and comorbid ODD/CD. There was a trend toward greater substance-related impairment, but not frequency of substance use among non-responders to the MDD treatment protocol. Controlling for potential confounds, there was a significant improvement in substance-related impairment among MDD responders to the MDD treatment protocol. However, MDD response was not associated with changes in the frequency of substance use. Baseline suicidal ideation was higher among subjects who progressed to high substance-related impairment compared to those whose substance-related impairment remained low. In contrast, baseline parental BDI scores were higher among subjects whose substance-related impairment remained high compared to those who decreased from high to low substance-related impairment during the study. MDD response was greatest among subjects with low 12-week substance-related impairment regardless of whether they had high (68.0%) or low (55.8%) baseline substance-related impairment. MDD response was significantly lower among subjects with high 12-week substance-related impairment, regardless of whether they had high (36.7%) or low (36.8%) substance-related impairment at baseline. Finally, there were no significant differential effects of specific treatments, pharmacological or CBT, on substance use impairment or frequency.
These findings must be interpreted in the context of methodological limitations of this study. First, despite the longitudinal methodology, with only 2 assessments this study does not allow for determination of the direction of the observed associations. It is equally possible that decreasing substance use leads to improvement in mood, or that improvement in mood results in decreased substance use. Second, the DUSI category of “painkillers” does not explicitly distinguish between medications such as acetaminophen, opiates, or barbiturates, such that this category combines medications with lesser and greater propensity for misuse. Third, the DUSI is a self-report instrument. Several adolescents reported recurrent substance use along with endorsing an impairment item on the DUSI, despite not meeting SUD criteria via the KSADS. Some adolescents may be more likely to disclose substance use in a self-report instrument than in an interview.32
Although the KSADS includes information from adolescents as well as from parents, it is possible that adolescents with undetected SUD were included in TORDIA. Reasons that substance-using adolescents did not receive a DSM-IV SUD diagnosis via the KSADS are that they may not have met the severity threshold of clinical significance, or problems related to substance use may not have been recurrent. Relatedly, this study did not include urine toxicology. Fourth, DUSI data were not available for all subjects at 12-weeks, and subjects with missing data differed in some ways from those with both baseline and 12-week DUSI data. This may influence the reported findings; however it is not possible to determine in which direction the data may have been skewed. The finding that subjects who did not provide data at 12-weeks had greater baseline substance use than those with complete data is consistent with previous findings that a history of substance use predicts drop-out among adolescents in treatment for depression33
. Finally, despite the large TORDIA sample, the examination of substance use comprises a secondary analysis that is not sufficiently powered to take into account multiple potential covariates or confounds, and therefore a limited number of variables could be included in multivariable models.
A cut-off score of 30 for the DUSI impairment score was found to be sensitive and specific in a sample of adolescents without psychiatric illness. 30
Given the exclusion of adolescents with SUD from TORDIA, it is not surprising that the mean DUSI impairment score in the present study was 11.0 ± 18.8. Nonetheless, findings suggest that even substance use that falls short of a diagnosis of SUD via the KSADS is associated with similar clinical correlates as SUD and that subjects with DUSI impairment scores of at least 13.3 at 12-weeks showed poorer response to MDD treatment. These findings underscore the importance of characterizing substance-related impairment even among adolescents with MDD who do not suffer from SUD. Indeed, previous studies have found that substance use is associated with suicidality, legal problems, high-risk sexual behavior, injuries, and functional impairment.34-40
The presence of these indicators should increase the clinician's index of suspicion of substance use and vice versa.
Consistent with previous studies, substance-related impairment was associated with depressive severity, history of physical or sexual abuse, and comorbid ODD/CD. In contrast, comorbid anxiety was not associated with substance use at baseline. Previous studies found that anxiety is associated with SUD among youth41
, and predicts SUD among youth with MDD.19
Because the present study examines sub-threshold substance use, as opposed to SUD, this finding is not directly comparable to previous studies and replication is warranted. The association between substance use and family conflict could have been anticipated on the basis of prior research.42, 43
This study examined whether baseline substance-related impairment predicts outcome of treatment for depression. For the overall sample, there was a trend toward greater baseline substance-related impairment, but not substance use frequency, among MDD treatment non-responders. For subjects with available DUSI scores at both time points, substance-related impairment at 12-weeks was associated with MDD treatment response at 12-weeks. High baseline substance-related impairment was associated with lower MDD treatment response only if it remained high. Similarly, low baseline substance-related impairment was only associated with higher MDD treatment response if there was also low substance-related impairment at 12-weeks. Controlling for other variables that were associated with MDD treatment response in TORDIA, MDD treatment responders had significantly greater reduction in substance-related impairment compared to non-responders. We are not aware of any studies regarding the association of substance use with treatment response among adolescents with MDD. However, a study of 94 adults with MDD openly treated with fluoxetine indicated that even when consumed in moderation, alcohol is associated with decreased treatment response among adults with MDD. These findings converge with those of a recent study of 126 adolescents with comorbid MDD and SUD in which subjects whose depression remitted had a greater proportion of negative urine drug screens and greater reduction in past-month self-reported days of drug use17
. In that study, substance use did not decrease significantly among non-remitters. An earlier pilot study also noted an association between changes in depressive symptoms and frequency of substance use.14
A meta-analysis of pharmacological treatment of adults with MDD and comorbid SUD found that studies with larger effect sizes for depression were more likely to have a favorable impact on quantity of substance use13
. Nonetheless, as acknowledged above, present findings do not inform our understanding regarding the direction of these associations. Future studies with multiple time-points are required to test for mediation.
Finally, this study examined whether treatment of MDD is associated with changes in substance use, and whether this association depends on the specific type of MDD treatment. The majority of subjects, 84% (230/274), did not demonstrate changes in substance-related impairment. For the overall sample, 9% decreased from high to low DUSI impairment, and 7% increased from low to high DUSI impairment. Within the 200 subjects with low baseline DUSI impairment, 19 (10%) demonstrated high DUSI impairment at 12 weeks. Within the 74 subjects with high baseline DUSI impairment, 25 (34%) demonstrated low DUSI impairment at 12 weeks. No between-treatment differences in changes in substance use or impairment were observed. The study by Riggs and colleagues indicated that treatment with fluoxetine demonstrated superiority over placebo in terms of proportion of negative urine drug screens, whereas no significant between-group difference in self-reported substance use was detected.17
Previous smaller placebo-controlled trials had failed to demonstrate superiority of active treatment with sertraline or fluoxetine over placebo with respect to the reduction of substance use.14, 16
This is the first study to our knowledge that specifically examines the association of substance use with treatment of treatment-resistant MDD among adolescents. Present findings suggest that even adolescents without SUD experience substance-related impairment and that substance-related impairment is potentially more strongly associated with predicting MDD treatment response than is the frequency of substance use. Since higher substance-related impairment at 12 weeks is associated with decreased probability of MDD treatment response, it is important to recognize that suicidal ideation and parental depressive symptoms are associated with initiation and persistence, respectively, of substance-related impairment. Although present findings do not allow for definitive conclusions, they suggest that substance use is clinically noteworthy even if symptoms of SUD are denied during interviews with the adolescent or his/her parents. Clinicians treating adolescents with MDD should be aware of the potential effects of even relatively low substance use and impairment on depression treatment. When substance use is detected, psychoeducation should be emphasized throughout MDD treatment and specific strategies to address substance use, such as motivational interviewing and refusal skills, should be employed as needed.