We found support for pathways in both directions between nonmedical opioid use and onset of psychiatric disorders. All of the psychiatric disorders tested were estimated to be three times as likely to precede nonmedical opioid use (HR ranging from 2.8 to 3.1). . Nonmedical use of opioids was linked to subsequent increased risk of mood disorders (MDD and bipolar I disorders) and anxiety disorders (panic and GAD), with hazard ratios ranging from 2.8-3.6, even in the presence of preexisting other illegal drug use. In sum, these data provide support for the presence of a general vulnerability to both psychiatric disorders and nonmedical opioid use (hazard ratios in opposite directions were of similar magnitude), for a self-medication pathway and for a precipitational model pathway.
The conclusions for the associations between psychiatric disorders and dependence resulting from nonmedical opioid use are more complex. Individuals with preexisting psychiatric disorders were estimated to be at increased risk for the onset of dependence resulting from the nonmedical opioid use, with particularly strong hazard ratios documented for GAD and bipolar I disorder (HR~10).. The estimated associations between preexisting dependence resulting from nonmedical opioid use and the onset of psychiatric disorders were also increased at a strong but generally lower magnitude than the reverse ordering (HR ranging from 4.1 to 8.5) This again indicates support for pathways in both directions (self-medication and precipitational model) and a general vulnerability model for dependence resulting from nonmedical opioid use and psychiatric disorders, but more strongly for a self-medication model for the onset of dependence from nonmedical use of opioids after preexisting bipolar I disorder and GAD.
An underlying general vulnerability to both opioid use and mood/anxiety disorders may be the explanatory factor in the observed associations between opioid use and psychiatric disorders. While a robust literature has suggested a shared etiology for illegal drug use and externalizing psychopathology with a large heritable component (Hicks et al., 2004
; Krueger, 1999
; Krueger et al., 1998
; Vollebergh et al., 2001
; Sullivan & Kendler, 1998
), the underlying pathways linking drug use to mood and anxiety disorders remains debated in the literature. Certainly, the etiologic pathways between drug use and psychopathology are complex and multifactorial; as such, the aggregation of illegal drug use and dependence into one class may obscure rather than illuminate relevant mediational models. Due to the large sample size of the NESARC, we can examine drug-specific pathways to motivate research questions and refine hypothesis testing. While broad phenotype definition of “internalizing” and “externalizing” psychopathology are common in genetic studies (Young et al.,2000
; Jorm et al., 2000
; Rowe et al., 1998
), these results suggest that productive genetic linkage and association studies could test relevant shared vulnerability hypotheses by aggregating nonmedical opioid use and dependence with mood and anxiety disorders for phenotype definition. Additionally, subgroups of individuals with distinct pathways might exist; for instance, some individuals who self-medicate, some who develop psychopathology in response to use, and some with shared vulnerability. Further research in this area and other large datasets using growth modeling and latent class analysis to disaggregate pathways of users might be productive.
As previously described, the self-medication pathway is predicated on two necessary assumptions: the psychiatric disorder preceded the onset of the drug use and/or disorder, and the use of the drug alleviates symptoms of the specific psychiatric disorder (Khantzian, 1997
; Khantzian, 2003
). We found that bipolar I disorder and GAD were strongly associated to the subsequent onset of dependence resulting from nonmedical opioid use with weaker evidence for the converse. Additionally, evidence from clinical and psychopharmacologic studies indicate that the acute effect of opioids include antidepressant, anti-panic, and anti-anxiety symptom reduction (Gold et al., 1979
; Emrich et al., 1982
). As such, the use of opioids amounts leading to opioid dependence to relieve these symptoms in individuals with bipolar disorder and GAD could be a reasonable hypothesis. Additionally, while we found evidence for a shared vulnerability between opioid use/dependence and other psychiatric disorders, shared vulnerability does not rule out the possibility that some individuals use opioids for symptom reduction during acute phases of psychiatric illness. We also conducted analyses for the temporal sequencing of dependence resulting from nonmedical opioid use and psychiatric disorders among the subset of individuals who used opioids nonmedically (available upon request) to shed light on whether unique genetic components may be implicated in use versus dependence (Heiman et al., 2007). Results were similar in direction as those obtained in the overall population, but of somewhat weaker magnitude, indicating further support for a generalized non-specific vulnerability model for the association between dependence resulting from nonmedical opioid use and psychiatric disorders.
Case reports and small clinical studies have suggested that opioid use can induce manic episodes in subjects with pre-existing affective disorders (Shaffer et al., 2007
; Watts & Grady, 1997
; Gonzalez-Pinto et al., 2001
; Orr et al., 1998
). While we do not have information on reasons for opioid use among individuals with bipolar disorder (e.g., to self-medicate back pain symptoms-analgesia or simply to “get high”-euphoria), we did find strong and significant associations for an increased risk of opioid use and dependence in the presence of bipolar disorder. This provides some epidemiologic support for a link between opioids and mania in the general population.
Several study limitations merit mention. First, all information is based on self-report, as in all large-scale epidemiologic surveys. As such, the validity of these results is predicated on the accuracy of the age of onset information provided by respondents. Further, as the wave 1 NESARC survey is a cross-sectional design relying on retrospective reports, older respondents may be reporting age of onset information that occurred several decades previously, which might lead to recall bias. However, we do not think recall bias would be differential across any relevant subpopulations, thus having limited effect on our hazard ratios and other effect estimates. We are currently planning follow-up analyses on temporal ordering of substance use and psychiatric disorders as the 3-year follow-up of NESARC participants becomes available, however, these analyses will be limited to the time-frame of the NESARC study. Moreover, due to the nature of the NESARC data we relied on information on age of onset of psychiatric disorders and not on age of onset of subthreshold psychiatric symptoms, respondents might have been experiencing sub-threshold symptoms before the onset of the psychiatric disorder. Further, we do not have information on whether nonmedical opioid users initiate opioid seeking euphoria (via medical prescription or not) or analgesia (via medical prescription or not) and where the first significant exposure occurred. Subtypes of opioid users may be unique in many aspects of comorbidity and demographics. Because the NESARC does not focus on opioid use specifically, this level of detail in a large-scale survey is untenable. Future research specifically focusing on opioid use and dependence may be able to provide more information on subtypes of opioid users in the general population. In addition, the use of time-varying covariates assumes that once the disorder appears, it remains indefinitely, however, substance use disorders and other psychiatric disorders often remit, or remit and relapse. With NESARC data it is not possible to take into account age of remission and relapse in these models.
Despite these limitations, the present study adds substantial information to the literature on opioid use and dependence and psychopathology. The large sample size of the NESARC allows for statistical power to test temporal ordering of not only opioid use but also the less common condition of dependence that had resulted from nonmedical opioid use with psychiatric disorders. Further, the AUDADIS-IV has documented reliability and validity in assessing drug use disorders as well as psychiatric disorders. Thus, the NESARC study is well suited to provide much-needed information about the temporal ordering of use and dependence on drugs such as the relationship between nonmedical opioid use with psychiatric disorders. As both the incidence and prevalence of opioid use and dependence in the population continue to show increases over the past decade, information of causal pathways is an important public health goal to better understand the specific comorbidities of opioid users as they may be distinct from other drug users. Our findings support a general vulnerability to opioid use and major psychopathologies, as well as evidence for a ‘self-medication’ model for dependence resulting from nonmedical opioid use with bipolar disorder and GAD. Future research with the longitudinal aspects of the NESARC study may be able to refine and extend these findings but using age of onset information less distal from some individuals in the study.