Explaining the association between anxiety disorders and alcohol use disorder: A twin study

Abstract

Background: It is unknown whether social anxiety disorder (SAD) has a unique association with alcohol use disorder (AUD) over and beyond that of other anxiety disorders, how the associations develop over time, and whether the associations are likely to be causal.

Methods: Diagnoses of AUD, SAD, generalized anxiety disorder, panic disorder, agoraphobia, and specific phobias were assessed twice using the Composite International Diagnostic Interview among 2,801 adult Norwegian twins. The data were analyzed using logistic regression analyses and multivariate biometric structural equation modeling.

Results: SAD had the strongest association with AUD, and SAD predicted AUD over and above the effect of other anxiety disorders. In addition, SAD was prospectively associated with AUD, whereas other anxiety disorders were not. AUD was associated with a slightly elevated risk of later anxiety disorders other than SAD. Biometric modeling favored a model where SAD influenced AUD compared to models where the relationship was reversed or due to correlated risk factors. Positive associations between AUD and other anxiety disorders were fully explained by shared genetic risk factors.

Conclusions: Unlike other anxiety disorders, SAD plausibly has a direct effect on AUD. Interventions aimed at prevention or treatment of SAD may have an additional beneficial effect of preventing AUD, whereas interventions aimed at other anxiety disorders are unlikely to have a similar sequential effect on AUD.

Keywords: agoraphobia; alcohol use disorder; anxiety disorders; social anxiety disorder; social phobia; specific phobia; twin studies.

Figure 1.

The full model (A) for longitudinal associations between alcohol use disorder (AUD) and anxiety disorders. Genetic and environmental influences on a disorder can influence later observations, but not earlier. Influences on disorders measured at the same point in time are allowed to correlate. In the more parsimonious model B, longitudinal associations across the disorders are modelled as direct effects. If shared genetic risk factors account for the longitudinal relationship, model A would fit best. If longitudinal associations result from causal effect between the disorders, model B would fit best. If the baseline (T1) associations fully explain the future (T2) associations, b₃₂ and b₄₁ would be estimated at zero. We separately test the associations from anxiety to AUD (a₄₁, c₄₁, and e₄₁ vs. b₄₁) and from AUD to anxiety (a₃₂, c₃₂, and e₃₂ vs. b₃₂).

Figure 2.

Longitudinal phenotypic associations among social anxiety disorder (SAD), other anxiety disorders (AnxIRT), and alcohol use disorder (AUD) occurring within 5-year intervals assessed 10 years apart, including 95% confidence intervals.

Figure 3.

Best fitting biometric models for the longitudinal relationship between social anxiety disorder (SAD) and alcohol use disorder (AUD) (left-hand side) and for the relationship between other anxiety disorders and AUD (right). Paths below 0.20 are drawn with dashed lines.