Tobacco smoking in adolescence or early adulthood may predict later major depression R1623BABCJEGF, an association not attributable to related variables such as alcohol use. However, not all studies have found this relationship, and some suggest a reverse association R1623BABICGAF. Among adults, smoking has been associated with suicidal behavior in psychiatric patients R1623BABFAAEG–R1623BABDJCAG and U.S. Army personnel R1623BABGGAIF.

We have reported a cross-sectional association between smoking and suicidal ideation among 15-year-olds R1623BABBFDBB. In this article we examine the predictive association between smoking in adolescence and suicidal ideation in early adulthood. We tested three hypotheses. First, tobacco smoking at age 15 predicts suicidal ideation at 18 and 21 years; second, smoking predicts later suicidal ideation, but only in the presence of depression at age 15; third, smoking predicts later depressive disorder, thereby placing the individual at risk of suicidal ideation.

The Dunedin Multidisciplinary Health and Development Study is a community-based investigation of 1,037 New Zealanders born in Dunedin in 1972–1973. The sample was first followed up at age 3 and most recently at age 26 years. Daily smoking was assessed at age 15 through self-report R1623BABBFDBB. Suicidal ideation with or without accompanying depressive disorder was assessed with the National Institute of Mental Health Diagnostic Interview Schedule R1623BABFBBDB based on DSM-III-R criteria at ages 18 and 21 R1623BABEDIDC. A complete description of the study, including the assessment of mental health and suicidal ideation, was provided to participants, who then gave written consent. Before the participants were 18 years old, their parents were provided with full details of the assessments and their written consent was obtained. The children provided verbal assent to the assessments and could refuse to take part in any assessments if they so wished.

We examined several other variables in modelling the association between smoking and suicidal ideation because these might independently account for any relationships between them. Childhood disadvantage to age 9 years was based on a composite index of the father’s occupation, age of the mother at first pregnancy, the mother’s educational qualifications, and sole parenting R1623BABEDIDC. Impulsiveness was assessed by 12 items based on DSM-III criteria for attention deficit hyperactivity disorder from parents’ and teachers’ reports at ages 9–11. At age 15, assessments included perceived attachment to parents R1623BABHEJCE, distress R1623BABGDFEA, significant depressed mood in the past year or DSM-III diagnosis of depressive disorder R1623BABFBCFC, and cannabis and alcohol use at school R1623BABFBCFC.

Data were available for 764 individuals from childhood to age 21; 100 (13.1%) reported daily smoking at age 15, and 130 (17.0%) reported suicidal ideation at ages 18–21. Logistic regression controlling for sex indicated that the odds ratio for later suicidal ideation associated with daily smoking at age 15 was 1.74 (95% confidence interval=1.06–2.87, p<0.03). Adolescents who were daily smokers at age 15 were nearly twice as likely to report later suicidal ideation as those who did not smoke daily.

t1 shows the multivariate logistic regression model of suicidal ideation in which the effect of each variable has been adjusted for all other variables. Later suicidal ideation was primarily a function of being female and experiencing depressed mood, high stress, and low parental attachment at age 15. After controlling for these, we found that there was no longer a significant relationship between daily smoking and suicidal ideation.

At age 15 there were 99 participants (13.0%) with depressed mood and 40 (5.2%) with a diagnosis of depressive disorder. To test our second hypothesis, we examined whether the rate of suicidal ideation was higher among adolescents who smoked daily and had depressed mood at age 15 or a diagnosis of depressive disorder at age 15, compared with those who had depressed mood or depressive disorder but did not smoke. This amounts to a test of a statistical interaction between daily smoking and depressed mood/depression at age 15. This interaction was not significant according to the likelihood ratio test for interaction for depressed mood (χ²=0.02, df=1, p>0.05) and likelihood ratio test for a diagnosis of depression (χ²=1.25, df=1, p>0.05).

Our third hypothesis was that tobacco smoking at age 15 predicts later depressive disorder. We examined this in a model similar to that shown in t1. The pattern of findings was the same as for suicidal ideation. Depressive disorder was predicted by being female, experiencing depressed mood and high levels of stress at age 15, and lower levels of parental attachment. Tobacco smoking at age 15 did not predict later depressive disorder.

Cigarette smoking in adolescence was significantly associated with suicidal ideation in early adulthood. This finding is similar to the finding of reports in psychiatric patients R1623BABFAAEG–R1623BABDJCAG and active-duty soldiers R1623BABGGAIF. However, our findings suggest that among young adults the association may be a function of dual associations with depressed mood, stress, and parental attachment in adolescence rather than a direct consequence of smoking.

Smoking among adolescents with depressed mood or with a diagnosis of depression was not associated with a greater risk of suicidal ideation. Furthermore, despite reports to the contrary R1623BABCJEGF, tobacco smoking at age 15 was not predictive of later depressive disorder. A recent study (2) suggested a small but significant predictive association between depression and tobacco smoking and argued for a self-medication hypothesis whereby depressed individuals smoke to reduce their symptoms. There are strong associations between anxiety and depression among young people R1623BABICGAF, R1623BABFBCFC and between smoking and anxiety disorders such as panic disorder R1623BABGBIHI. The key to clarifying the direction of causal influence may be a closer examination of the predictive significance of smoking for anxiety disorders with and without accompanying depression.

Received Aug. 26, 2003; revisions received Jan. 13 and May 25, 2004; accepted June 10, 2004. From the Department of Preventive and Social Medicine, University of Otago Medical School. Address correspondence and reprint requests to Dr. McGee, Department of Preventive and Social Medicine, University of Otago Medical School, Box 913, Dunedin, New Zealand; rob.mcgee@stonebow.otago.ac.nz (e-mail). The Dunedin Multidisciplinary Health and Development Study is supported by the Health Research Council of New Zealand. Collection of the self-harm data was funded by NIMH grants 1-23 MH-42723-01 and RO1 MH-45070. The authors thank the participants of the Dunedin Multidisciplinary Health and Development Study for their commitment to this research.