Publication Date


Document Type

Doctoral Dissertation

Academic Program

Clinical and Population Health Research



First Thesis Advisor

William H. Fisher, Ph.D.


Bipolar Disorder, Cannabis, Marijuana Abuse, Marijuana Smoking, Comorbidity, Risk Assessment


Bipolar disorders (BD) are characterized by symptoms of grandiosity, decreased need for sleep, pressure to keep talking, flight of ideas, distractibility, increased goal-directed activities, psychomotor agitation, and excessive involvement in pleasurable activities. Those with a bipolar disorder have a high degree of psychiatric comorbidity including substance use disorders, and they also experience increased mortality. Despite the widespread recognition of BD as an important psychiatric condition, available population-based estimates for BD prevalence differs across data sources.

Cannabis is one of the most widely-used illicit substances. Evidence supports it as a risk factor for psychotic symptoms and disorders. Because populations with psychotic disorders and populations with bipolar disorder share genetic characteristics, cannabis may increase risk for bipolar disorders through the same pathways as it does with psychotic disorders. Limited and conflicting evidence regarding the association of cannabis use and bipolar disorder is currently available. This dissertation investigates cannabis use as a risk factor for incident manic symptoms and bipolar disorders in a large nationally representative longitudinal cohort.

The first aim of this dissertation is to evaluate the implications for manic, hypomanic and major depressive episode prevalence estimates arising from the different approaches to assessing DSM-IV criterion between two national surveys. Differences in the assessment of impairment strongly influence manic or hypomanic classification within the NESARC. Compared to multiple imputation estimates (19.7% [95% CI: 19.3-20.1]) which treat depressed mood and anhedonia as separate symptoms, symptom assessment in the NESARC substantially underestimates major depressive episode prevalence (16.9% [95% CI: 16.1-17.6]).

The second research objective examined self-reported cannabis use as a risk factor for incident manic symptoms, bipolar spectrum disorders (including manic and hypomanic episodes) and SCID-based recalibrated BD I and II. Cannabis use risk was assessed in the population as a whole and in sub-populations defined by age, substance abuse/dependence status, and family history. Among those reporting no lifetime major depressive or manic symptoms at baseline, self-reported past-year cannabis use was associated with increased odds of an incident week of extremely elevated or irritable mood accompanied by at least two manic episode criterion B symptoms (adj. OR 1.69, 95% CI: 1.08-2.65, p=.02) over the three year follow-up period. Among adults (ages 26 to 45) >=1 reported use(s) of cannabis per week was associated with incident manic or hypomanic episodes (adjusted OR 2.52, 95% CI: 1.32-4.80, p=.006). Among those endorsing no major depressive symptoms, substance abuse/dependence, or anti-social traits in their first degree relatives, past year cannabis use is associated with increased risk for incident bipolar spectrum disorders (adjusted OR 2.27, 95% CI: 1.01-5.10, p=.05) and CIDI recalibrated BD I and II (adjusted OR 5.49, 95% CI: 1.38-21.9, p=.02). Past year cannabis use risk for DSM-IV manic or hypomanic episodes among those aged 26 to 45 is concentrated in those with a baseline history of a substance use disorder (adj. OR 2.00, 95% CI: 1.10-3.66, p=.02) as compared to those with no such history (adj. OR 1.87, 95% CI: 0.49-7.21, p=.36).

The third research objective of this dissertation was a sensitivity analysis using externally-predicted categorized exposures and continuous cannabis use propensities. The sensitivity analysis found evidence of exposure misclassification. Exposures defined by external propensity scores had improved cross-sectional association with bipolar spectrum disorders compared to reported use when both were compared to an external standard. No significant risk estimates were found for categorized predicted cannabis use among groups that were previously found to have significant risk from reported exposure. However, among adults 18 to 45 years of age with no manic or major depressive symptoms at baseline, past year cannabis use propensity (as a log transformed continuous measure) was associated with incident manic or hypomanic episodes (adj. OR 1.49, 95% CI: 1.10-2.03, p=.01). Elevated risk for high cannabis use propensity (>=1 use/week in the past year) was also found in this same group (adj. OR 1.33, 95% CI: 1.03-1.72, p=.03). Among those with no reported history of depression, substance abuse/dependence, or anti-social traits among their first-degree relatives, propensity for past year cannabis use (adj. OR 1.61, 95% CI: 1.11-2.32, p=.01) and propensity for >=1 use/week of cannabis in the past year (adj. OR 1.38, 95% CI: 1.03-1.85, p=.03) were associated with incident manic or hypomanic episodes. Among those without a substance use history at baseline, propensity for past year cannabis use (adj. OR 1.63, 95% CI: 1.33-1.55, p=1 use/week of cannabis in the past year (adj. OR 1.54, 95% CI: 1.26-1.88, p

The findings of the first aim support the conclusion that the AUDADIS substantially under-estimated lifetime major depressive episode prevalence compared to an imputed estimate that treated anhedonia and depressed mood as separate and concurrent MDE symptoms. The operationalization of impairment for manic disorders in both the AUDADIS and CIDI strongly influences case identification, with the CIDI having suppressed manic and hypomanic prevalence estimates. Evidence was found supporting the conclusion that self-reported cannabis use is a significant risk factor for incident bipolar spectrum outcomes within subpopulations in a nationally representative cohort. A sensitivity analysis finds evidence that supports the conclusion that increasing cannabis use propensity is associated with increased risk of bipolar spectrum outcomes within population subgroups, with the greatest increased risk among those with the lowest innate risk. Under-reporting of illicit substance use is a major limitation in this dissertation; further study is needed with improved exposure measures.



Rights and Permissions

Copyright is held by the author, with all rights reserved.