These analyses sought to address whether nicotine withdrawal-induced symptoms of anxiety/depression are a function of liability to GAD/MD or are more strongly indexed by liability to ND. We further sought to address whether some portion of the liability—To psychopathology or ND—could be accounted for by genetic factors. To investigate these questions, we used co-twins’ GAD, MD, and ND phenotypes to predict the severity of their twins’ nicotine withdrawal-induced symptoms of anxiety or depression. Secondarily, we replaced co-twins’ GAD or MD history with co-twins’ mean neuroticism score since this measure is continuous and thus more statistically powerful than a binary diagnosis, and neuroticism is phenotypically and genetically correlated with both MD and GAD. We asked whether monozygotic co-twin's phenotypes (FTND, MD, or GAD) were more strongly predictive of their twin's depressive/anxious symptoms than were dizygotic twin's phenotypes to address whether genetic factors contribute to liability.
Age and sex were consistently associated with depressive symptoms: older individuals were more likely to experience more severe symptoms, as were women. Age was not associated with severity of anxiety symptoms, though sex was nearly always associated with this outcome, again with women experiencing a higher level of symptoms. Outcome was not significantly predicted by co-twin's mood-related psychopathology nor was there any indication that predictive value of these terms differed as a function of zygosity. Similarly, co-twin's neuroticism was largely not associated with severity of withdrawal-induced depressive or anxious symptoms nor was there a significant interaction between this term and zygosity. Critical to our analyses, co-twin's FTND score consistently predicted both outcomes; furthermore, when the interaction between co-twin's FTND and zygosity was included as a predictor, this term and the main effect of co-twin's FTND significantly predicted outcome, with monozygotic co-twin's FTND score more strongly predicting outcome than dizygotic co-twin's FTND score.
The results of these analyses strongly suggest that the severity of nicotine withdrawal-induced depressive and anxious symptoms is a pharmacological component of the nicotine withdrawal syndrome, which is indexed by ND. Furthermore, this association appears to be driven primarily by genetic influences on liability to ND and withdrawal, with environmental factors playing a less pronounced role: the main effect of co-twin's FTND score, which is strongly predictive of outcome in nearly every case, is sharply reduced when the co-twin's FTND × zygosity interaction term is included in the model (Equations 2, 4, 7, and 8
). Parameter estimates from these regressions further suggest that ND is slightly more predictive of withdrawal-induced anxious symptoms than depressive symptoms.
The main effects of co-twin's MD and GAD are not significantly associated with outcome in any regression. Co-twin's mean neuroticism score, which is positively and significantly correlated with both GAD and MD, is associated with withdrawal-induced depressive symptoms, but not with anxious symptoms. Nor was the interaction between zygosity and MD, GAD, or mean N
associated with either outcome (Equations 1
, and 7
, respectively). However, results from the within-individual regressions differ slightly in that a history of GAD was associated with withdrawal-induced anxious symptoms, and mean N
was associated with both outcomes. The results of within-individual regressions are largely consistent with previous studies that have reported that an individual's history of anxiety or depression is predictive of their withdrawal-induced anxious or depressive symptoms (Burgess et al., 2002
; Covey et al., 1990
; Pomerleau et al., 2000
). However, not all studies include ND as a covariate, which could be problematic: if ND is positively correlated with psychopathology (as it is in the current study), predictive value could be erroneously attributed to psychopathology. The use of co-twins’ phenotypes as instrumental variables in the current analyses enables us to distinguish between our two hypotheses regarding outcome.
The intraindividual analyses indicate that people with a lifetime history of MD or GAD have significantly higher levels of ND than do individuals without such a history. This could be due to common genetic/environmental influences underlying these traits and/or to a causal relationship: nicotine reduces negative affect and is anxiolytic (although the acute vs. chronic effects are complex) (Benowitz, 2008
; Kalman, 2002
; Picciotto, Brunzell, & Caldarone, 2002
). People suffering from depression or anxiety might use nicotine to alleviate these symptoms (Breslau, 1995
; Pomerleau et al., 2000
). Self-medicators could potentially increase their dependence on nicotine through increased consumption. Since our findings indicate that level of ND predicts severity of withdrawal symptoms related to negative affect, such a relationship could result in a positive association between psychopathology and withdrawal-induced depressive or anxious symptoms, though that association would likely disappear if ND were included in models of risk.
Although these findings indicate that genetic liability to ND—and by extension, to nicotine withdrawal—influences nicotine withdrawal-induced symptoms of anxiety and depression, the precise biological underpinnings of an affect-related response are beyond the scope of the analysis. Given that nicotine acts on neural pathways that are also involved in mood disorders, one might speculate that nicotine withdrawal and depressive or anxious episodes impact these pathways in similar manners and thus produce similar mood changes. If the relevant pathways are also involved in mood/affect, it is reasonable to expect a mood-related response to nicotine withdrawal. Gene products underlying liability to ND could influence the neural (and otherwise physiological) response to nicotine and likewise respond to the absence of nicotine once dependence has been established. The literature on genetic variants influencing ND is vast, leaving a detailed discussion of potential candidate genes beyond the scope of this report. However, we note that variants in genes encoding the mu-opioid receptor (Zhang, Kendler, & Chen, 2006
), the nicotinic acetylcholine receptors alpha three and alpha five (Chen et al., 2009
), and cannabinoid receptor 1 (Chen et al., 2008
) have been associated with ND in samples derived from, or quite similar to, the current sample. These variants could feasibly also be associated with nicotine withdrawal symptoms.
To our knowledge, this is the first reported analysis that investigates whether nicotine withdrawal-induced symptoms of anxiety or depression can be attributed to liability to psychopathology or are simply a pharmacological withdrawal response independent of liability to negative affect outside the context of withdrawal in a genetically informative sample. These results have a number of implications. Clinicians (and tobacco users themselves) should be aware that, the higher one's ND, the more likely one is to experience negative affect upon cessation of tobacco use even in the absence of a history of mood-related psychopathology. Importantly, an individual with relatively high genetic risk to ND could potentially experience severe withdrawal-induced negative affect, even if that individual were not highly nicotine dependent. Although a history of MD was not associated with severity of symptoms in the current study and a history of GAD was only weakly associated with symptoms, individuals with these disorders tend to be more nicotine dependent; thus, a history of psychopathology could potentially be indicative of risk. Furthermore, consideration of the current findings could improve phenotype refinement in gene finding efforts since individuals experiencing strong symptoms of negative affect upon smoking cessation might be more likely to harbor genetic variants influencing ND than are individuals who do not experience these symptoms.
These analyses should be considered in light of several limitations. The sample is restricted to twins born in Virginia and is primarily Caucasian; the findings might not be applicable to other ethnicities or populations. In addition, the items addressing affect-related symptoms of withdrawal were limited (see Methods section) and retrospective and are thus susceptible to recall bias. We emphasize that these variables potentially encompass a range of withdrawal constructs that are not necessarily mood related; for example, a respondent might endorse feeling anxious in cases where the symptom could be better described as psychomotor agitation or the anxiety was experienced somatically rather than psychologically. Furthermore, we did not have information on the duration of quit attempts, which could be informative given that the course of withdrawal varies across individuals. In addition, we did not control for baseline (prequit) measures of negative affect beyond MD and GAD diagnoses, which makes it difficult to interpret our assessment of cessation-induced negative affect; Shiffman, West, and Gilbertand (2004)
recommend that multiple prequit assessments be taken to account for baseline instability. Additional research that parallels the current analyses, but more thoroughly characterizes both baseline phenotypes and the withdrawal syndrome in terms of symptoms, duration, etc., is needed to confirm the findings reported herein.
We also note that these results might not be generalizable to subclinical anxiety or depression or to internalizing disorders other than MD and GAD, such as panic disorder, dysthymia, etc.; this limitation is particularly relevant given recent findings, suggesting that the relationships between smoking behaviors and anxiety differs across disorders (e.g., Cougle et al., 2010
). Potential differences in the relationships among other types of anxiety disorders, ND, and withdrawal-related symptoms of anxiety or depression should be the topic of future research. Finally, the diagnostic categories of MD and GAD are less statistically powerful than would be a continuous measure and this could have led to false negative results; however, our analyses including neuroticism scores suggest that these are accurate negative results, and the sample size is large enough to alleviate concerns about statistical power.