Abstract
The prevalence of smoking among psychiatric patients is higher than in the general population [1], with strong evidence of comorbidity between smoking and psychopathology being established by mid-adolescence [2–5]. While evidence of any relationship between smoking and anxiety disorders has been described as ‘modest’ and ‘inconsistent’ [6–8] epidemiological studies report a more consistent association between smoking and major depression [7,9–11]. Any increased risk of progression to nicotine dependence for those with a history of major depressive disorder [12] is of concern when nicotine dependence is linked to suicidal behaviour [13–16] and is considered by some as a ‘gateway’ step to precede problematic use of other addictive substances [17–19].
One interpretation of the association is that depression plays a causal role in both smoking initiation [8] and progression to regular and heavy smoking, a so-called ‘self-medication’ pathway, with the mood-altering effects of nicotine reinforcing its use [20–22]. While the role of depression in smoking initiation is still not certain, a history of depression has been demonstrated, however, to be associated with an increased risk of progression to more severe levels of dependence on nicotine [23]. Studies have identified smokers as experiencing both higher rates of major depressive order and depressive symptoms than experienced by non-smokers [9, 11], and heavier smokers experiencing more depression than do light smokers [9], which does pose a question as to any real therapeutic mood-altering benefit of nicotine.
A second explanation is that smoking is a causal determinant of depression via deleterious effects of long-term use of nicotine on relevant neurobiological systems [24, 25]. In support, there is evidence of an increased risk for first-onset major depression in young adults with prior smoking or pre-existing nicotine dependence [8, 12, 23].
A third explanation is that higher-order factors contribute to both depression and smoking. Social and behavioural factors [6, 26], genetic factors [9, 27, 28] and childhood environment influences [2, 4], including socioeconomic factors [3], physical abuse [29], having only one biological parent in the home [29, 30] and peer disapproval [3] have all been proposed. Behavioural problems, such as attention deficit hyperactivity disorder (ADHD) [31–33], conduct disorder [8, 23, 34, 35], dropping out of school [36], alcohol abuse [37] and personality style [38] have also been proposed as possible higher-order predisposing factors.
In this report, we pursue the third possibility, seeking to identify risk factors overrepresented in smokers (compared with non-smokers) in a sample of depressed patients.
Method
Assessment
The recruitment and assessment methodology has been detailed [39] and will be only outlined here. Depressed patients completed a self-report questionnaire providing sociodemographic information. A research psychologist obtained information on clinical features of the current episode, substance use (including current smoking status and number of cigarettes smoked per day) and family history of psychiatric disorder, and generated lifetime anxiety disorder diagnoses from the computerized Composite International Diagnostic Interview [40]. Further information was obtained by patients completing the Costello–Comrey trait anxiety measure [41], the Beck Depression Inventory or BDI [42] and checklists of descriptors of their current mood and anxiety symptoms rated on a four-point scale (from ‘absent’ to ‘severe’).
Information on more complex features of depression, self-injury and suicidal behaviour was obtained by an interviewing research psychiatrist who also completed the 17-item Hamilton [43] depression measure and allocated a Mood Disorders Unit (MDU) clinical diagnosis [i.e. psychotic (PD), endogenous (ED), neurotic (ND) and reactive (RD) depression], with diagnostic guidelines previously described [44]. The psychiatrist generated CORE scores of psychomotor disturbance [45] and returned a score on global assessment of functioning (GAF). The patient's childhood environment was assessed, including markers of later anxiety (i.e. behavioural inhibition and school phobia), while lifelong ‘trait anxiety’ was also assessed by asking patients if they viewed themselves as ‘nervy’, a ‘worrier’, ‘tense’, and ‘anxious’ when not depressed.
Several methods were used to assess personality style and disorder. The psychiatrist (assisted by information from the referrer or a corroborative witness) rated the degree to which 15 vignettes representing each DSM-IV [46] listed personality disorder ‘style’ approximated to the ‘individual's long-term personality’. ‘Inadequate personality’ was assessed by the relevant single item from the Newcastle Index [47], and the psychiatrist made a single overall judgement of ‘personality disturbance’ prior to depression onset. Millon's [48] framework for ‘disordered functioning'was used to assess eight ‘parameters’ of disordered personality (e.g. causing significant personal discomfort; reducing opportunities) and five ‘domains’ of dysfunctional relationships (i.e. intimate, family, peer, work and work relationships). Corroborative data on trait anxiety, domain and parameter measures were obtained (with patient's permission) from questionnaires completed by the referrer or a relative.
A 1-year follow-up assessment was undertaken, allowing state depression levels to be quantified, as well as episode duration data to be collected.
Subjects
The pool from which the study sample was drawn consisted of 270 patients (weighted towards more chronic and treatment disorders) assessed at our MDU. Smoking rates for the diagnostic groups were 7.6% for PDs, 25.0% for both the EDs and RDs and 42.4% for the NDs, suggesting some spread by diagnostic group. Preliminary analyses revealed that, on average, smokers were younger than non-smokers at both that assessment and at the time of their first depressive episode.
Ninety-two (34%) of the sample were designated as ‘depressed smokers’ (DSs) on the basis of their smoking status at baseline assessment. We elected to derive a similar-sized control group of ‘depressed non-smoker’ (DNS) subjects, matched by age within a 6-year range (with 83% being within 3 years), gender, clinical diagnosis and on CORE scores, with the latter two variables seeking to control for diagnostic ‘type’.
Statistical analyses
Analyses comparing the two groups used paired t-tests and the nonparametric χ2 (McNemar's test), with Pearson's χ2 being used for variables with more than two categories. The DS and DNS groups were treated as two independent cells for logistic regression analyses, and for reported prevalence rates and odds ratios.
Results
Matched group comparisons
Table 1 compares sociodemographic, family history of depression and childhood stressor variables. On average, the DSs had completed fewer years of education, scored higher on socioeconomic status (where high score means low status), and, at baseline assessment, the DSs were less likely to be employed. They were more likely to have been separated or divorced or to have never been married, and were less likely to be currently married, in a de facto relationship or in an intimate relationship. There was no significant difference in family or parental prevalence of depression or in likelihood of parents receiving psychiatric treatment. The DSs were more likely to have experienced physical abuse, as well as parental separation or divorce in childhood. They rated one or both of their parents as more critical, unpredictable and as making them feel unsafe and unsupported and failing to protect them.
Comparisons of demographic items, family psychopathology and childhood stressors, between matched depressed ‘smokers’ and ‘non-smokers’
Table 2 considers data on anxiety, personality functioning, depression and psychopathological behaviour. There was no difference on the Costello–Comrey trait anxiety score and, although rates of affirming other trait anxiety items were generally high, the DSs returned a higher rate on the trait ‘nervy’ item only. DS subjects were more likely to have engaged in school truancy, but did not have higher rates of school phobia or of behavioural inhibition. They tended to have higher rates of social phobia, and were more likely to have used anxiolytic medication for longer than 12 months and to have become anxiolyticdependent. The DS subjects scored higher on ‘eccentric’ personality style, higher on the total score of personality ‘domains’ and on the total score for personality ‘parameters’, were more likely to be rated as having a personality disturbance prior to the current episode and to rate positively on the Newcastle ‘inadequate personality’ item.
Comparisons of anxiety, personality, depression and psychopathological behaviour, between matched depressed ‘smokers’ and ‘non-smokers’
Examining only the matched groups, the DS and NDS subjects did not differ by age at first episode of depression, but were more likely to have taken time off work in recent years, and to have received social welfare for their mood state. In terms of depression severity, although DSs scored lower on GAF, there were no differences on either of the two state depression measures (i.e. BDI, Hamilton) or CORE scores. Sixty-one per cent of the DSs and 63% of the DNSs were assessed at the 1-year review (revealing no differences in duration of index episode, weeks depressed over that follow-up year or on their followup BDI state depression scores).
The DS subjects were more likely to have used illicit drugs (marihuana, amphetamines, LSD, cocaine), to have abused alcohol in the past and to have alcohol-related problems, to have a history of having made at least one suicide attempt and tended to be more likely to engage in self-injurious behaviours.
Multivariate analyses
Logistic regression analyses were undertaken, with significant items from the previous analyses entered as predictors of DS assignment. Since there were numerous significant items a reduced set of representative items was selected from separate analyses for the various domains (i.e. childhood, personality, anxiety, psychopathological behaviour). High rates of intercorrelations resulted in redundant items, leaving a representative set which included personality ‘domain’, anxiolytics ever used for more than 12 months, physical violence from parents and illicit drug use, to be entered in the final analysis. This analysis produced improvement in the model (χ2 = 49.4, df = 4, p < 0.001). Statistics reported for significant items remaining in the equation include the coefficient (B), standard error (SE), Wald statistic (W), odds ratio (OR) and 95% confidence interval (CI). Remaining items were: personality ‘domain’ (B = 0.41, SE = 0.2, W = 3.4, OR = 1.51, 95% CI = 1.0–2.3), anxiolytic use (B = 0.95, SE = 0.4, W = 5.2, OR = 2.58, 95% CI = 1.1–5.8), physical violence (W = 3.8, OR = 1.54, 95% CI = 1.0–2.4) and illicit drug use (B = 1.61, SE = 0.3, Wald = 22.0, 95% CI = 2.6–9.8).
Discussion
Methodological limitations to this study must be conceded. The sample was restricted to depressed patients at a tertiary referral Mood Disorders Unit, we did not have a non-depressed smoker comparison group, and absence of data on age of smoking initiation disallowed examination of temporal and causal relationships. Limited child temperament data also restricts consideration of the nature of personality predisposition.
Results failed to identify any longer history of depression or greater severity of depression at baseline or at follow up in the DSs. If cigarette smokers experience intrinsically more severe depression, it is possible that the nicotine was effective to some degree as a selfmedication for the depressed mood. However, we noted that the DSs reported more irritability (symptom score 12.1 vs 10.3, t = 2.3, p < 0.05), were more likely to have a lifetime anxiety disorder and had higher rates of substance abuse as well as of anxiolytic use and dependency, so arguing against nicotine being a successful therapeutic self-medication.
The third broad explanation concedes higher-order variables disposing to comorbid depression and cigarette smoking. High anxiety is one possibility, but any such influence was not clearly identified. Smokers were significantly more likely to score higher on some anxiety variables, but not on others. Rates of anxiolytic use and dependence were significantly higher than for nonsmokers, as was a general trend for greater use of illicit drugs, perhaps suggesting that differences might emerge from management of anxiety and other dysphoric states (rather than from differences in dysphoric states themselves), when we identified a general trend for greater use of illicit drugs as well.
Thus, although the cigarette smokers may not have been inherently more predisposed to early anxiety, the demonstrated adverse and non-supportive childhood environment may have resulted in increased insecurity and general anxiety, which was dealt with by smoking, medication or illicit drugs.
A speculative proposal is that dysfunctional or disordered personality may play an important role in comorbid smoking and depression, possibly via impaired coping capacity or style. Coping was not assessed, although it could be informative to consider in future studies.
An overview of our broad variable set indicated that smokers were most clearly distinguished from nonsmokers by being more likely to report exposure to a range of adverse developmental stressors and to show considerable evidence of disordered personality functioning.
It could be that the latter is genetically determined, influencing both dysfunctional parenting in the parents and disordered personality functioning in the children. Alternatively, intrinsically defective parenting, here flagged by characteristics such as emotional, physical and sexual abuse, may create certain vulnerabilities and coping limitations in the children. Our smokers described a greater chance of exposure to parenting that was unpredictable, lacking in support, failing to protect the child when under threat, and also making the child feel unsafe. It would be hardly surprising that such exposure would increase the chance of ‘insecurity’ in the child and of ineffective or dysfunctional coping as a consequence. Development has been held to impair at the neurobiological level [49], with a ‘punishing’ hostile or emotionally insecure childhood environment possibly resulting in poor development of coping ability and/or exacerbation of any predisposition for dysfunctional personality.
At the structural level, just as our smokers were more likely to have been exposed to parental separation/ divorce, they themselves were more likely to be separated or divorced, less likely to be employed, and more likely to be in receipt of social welfare, all structural indicators of impaired functioning.
A dysfunctional predisposition and impaired development of coping capacities may result in coping strategies that are at best likely to exacerbate problems and at worst to cause engagement in potentially self-destructive behaviours (such as illicit drug use, alcohol, cigarettes and suicide attempts).
Our logistic regression analysis results support the hypothesized trajectory speculated on here. The DS subjects were more likely to report physical violence from parents, long-term anxiolytic use, illicit drug use and to return a higher disordered personality domain score.
Factors linked here to increased chance of smoking are also factors linked to an increased chance of nonmelancholic depression [50, 51]. The early adverse environmental factors may then lead to a set of sequelae, including disordered personality functioning, drug and alcohol dependence and depression (especially of the non-melancholic type) so creating associations and formal comorbidity of axis I and axis II states. We conclude that the increased chance of cigarette smoking in those with depressive disorders is more likely to emerge from such shared higher-order variables as identified here rather than from simple direct links whereby smoking causes depression, or the converse, although this would need to be replicated in a study with non-smokers and non-depressed subjects.
Footnotes
Acknowledgements
This study was supported by NHMRC Program Grant 993208. We acknowledge the key support of Kerrie Eyers and Christine Taylor.