Skip Navigation


American Journal of Epidemiology Advance Access originally published online on January 4, 2006
American Journal of Epidemiology 2006 163(5):421-432; doi:10.1093/aje/kwj058
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
163/5/421    most recent
kwj058v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (14)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Klungsøyr, O.
Right arrow Articles by Sandanger, I.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Klungsøyr, O.
Right arrow Articles by Sandanger, I.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

American Journal of Epidemiology Copyright © 2006 by the Johns Hopkins Bloomberg School of Public Health All rights reserved; printed in U.S.A.

Original Contribution

Cigarette Smoking and Incidence of First Depressive Episode: An 11-Year, Population-based Follow-up Study

Ole Klungsøyr1, Jan F. Nygård1,2, Tom Sørensen3 and Inger Sandanger1,4

1 Department of Behavioral Sciences in Medicine, Institute of Basic Medical Sciences, Faculty of Medicine, University of Oslo, Oslo, Norway
2 The Cancer Registry of Norway, Oslo, Norway
3 Institute Group of Psychiatry, University of Oslo, Oslo, Norway
4 Norwegian Health Services Research Centre, Quality Evaluation Department, University of Oslo, Oslo, Norway

Reprint requests to Dr. Ole Klungsøyr, Department of Behavioral Sciences in Medicine, Institute of Basic Medical Sciences, Faculty of Medicine, University of Oslo, P.O. Box 1111, Blindern, N-0317 Oslo, Norway (e-mail: ole.klungsoyr{at}medisin.uio.no).

Received for publication March 11, 2005. Accepted for publication September 23, 2005.


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
Smoking has been found to be associated with depression. Biologic hypotheses support causation in both directions. This study examined the association between cigarette smoking and a subsequent first depression. In 1990, 2,014 adults in Norway were interviewed about their lifestyle and mental health. A 2001 reinterview by trained interviewers defined the study cohort of 1,190 participants. The cases were those who experienced a first depression whose onset was estimated to occur during the follow-up period, based on retrospective assessment by the Composite International Diagnostic Interview (International Classification of Diseases, Tenth Revision). Cox regression was used to estimate the hazard rate of depression during follow-up. Alternative explanations for a direct causal influence from smoking on subsequent depression were assessed, and a sensitivity analysis was performed. The risk of depression was four times as high for heavy smokers compared with never smokers. A dose-response relation with an increasing hazard for past smokers and for an increasing number of cigarettes smoked per day for current smokers was found. Similarly, increasing smoking time was associated with increasing risk. Failure of other plausible alternatives to explain the observed association between smoking and depression might reflect a direct causal influence of smoking on depression.

causality; depression; prospective studies; smoking

Abbreviations: CI, confidence interval; CIDI, Composite International Diagnostic Interview; FUP, 11-year follow-up period between T1 (baseline) and T2 (reinterview); HR, hazard ratio; OR, odds ratio; pre-T1 depression, a first depressive episode whose estimated onset occurred before T1


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
By the year 2020, tobacco use is predicted to be the largest single health problem in the world in terms of both mortality and disability, whereas unipolar major depression is second regarding disability from disease (1Go). Numerous studies have found an association between cigarette smoking and depression, where depression is overrepresented among smokers and vice versa, although there are some reports without these findings (2Go, 3Go). The association has been found to be both scale and age robust, with results replicated in adolescents (4Go–9Go), adults (2Go, 10Go–15Go), and elderly (10Go, 11Go, 16Go, 17Go).

A key question regarding the smoking–depression association, which is still unresolved, is whether it reflects a causal influence in one or both directions. Lacking randomization, a careful prospective design can still study causality (18Go). Temporal sequence, plausible mechanisms, dose-response relations, and elimination of other alternatives are all indications of causation.

Changes in early depressive symptoms and smoking behavior are well described in studies of adolescents, although model variation and inconsistent findings make the causal direction unclear (5Go–9Go). Biologic hypotheses explaining causation in each direction are supported. Smoking as self-medication for depressive symptoms is plausible in view of the euphoriant effects of nicotine (19Go). A pharmacologic explanation for the influence of smoking on depression has been found (20Go). Years of smoking versus depression as a dose-response relation has received little attention, even though a pharmacologic explanation might be a process that needs time to evolve, as in long-time smokers. Some studies have concluded that the smoking–depression association probably is accounted for by confounding (4Go, 13Go, 15Go, 21Go).

This study examined the association between baseline smoking, emphasizing smoking time as well as number of cigarettes smoked per day, and first depressive episode during the 11-year follow-up period between T1 (baseline) and T2 (reinterview) (FUP) in a Norwegian adult population.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 References
 
Subjects
Data from a 1989–1991 population survey were used, in which risk factors for, incidence of, and prevalence of psychiatric illness in Norway were examined. Statistics Norway drew a random sample of individuals from Oslo (the capital of Norway) and Lofoten (a rural area dominated by the fishing industry), aged 18 years or older, of which 2,727 were approached about participating (1,356 from Oslo and 1,371 from Lofoten). Of these persons, 713 refused, leaving 2,014 participants: 1,009 in Oslo and 1,005 in Lofoten. This number corresponds to 74 percent of the 2,727 eligible, representative of age and gender for the source population, constituting our baseline at time T1. The interviews were conducted in person, mainly by medical students and nurses. More details of the initial sample are described elsewhere (22Go). The population sample was approached again 11 years later, and 1,286 individuals were reinterviewed (in person) in 2000–2002 (T2).

Assessments at T1 (baseline).
Selected covariates for the study participants are presented in table 1. The Hopkins Symptom Checklist 25 was chosen as a measure of psychological distress at T1 (23Go–25Go) because of good prediction of a depression diagnosis (26Go, 27Go). Items rated similarly, together with no references to change ("worse than before"), were also considered aspects in favor of the Hopkins Symptom Checklist 25 compared with, for example, the General Health Questionnaire (28Go).


View this table:
[in this window]
[in a new window]
  		TABLE 1. Selected covariates for 1,190 respondents from a Norwegian general population studied to determine an association between cigarette smoking and incidence of a first depressive episode, 1990–2001

 
Assessments at T2 (reinterview).
The Norwegian personal identification number made it possible to locate almost the whole sample after 11 years; only 23 of the 2,014 participants (1.1 percent) could not be located. At T2, all respondents were interviewed by using the Composite International Diagnostic Interview (CIDI-M 1.1, updated electronic version of World Health Organization CIDI version 1.2 in 2001) following International Classification of Diseases, Tenth Revision, Diagnostic Criteria for Research Diagnoses (29Go). The outcome was a first depressive episode, including dysthymia, recurrent and single episodes of a mild, moderate, and severe character (International Classification of Diseases, Tenth Revision, codes F31.3–F34.1) with estimated onset based on the retrospectively gathered information. Use of the CIDI algorithm returned a year of onset for all cases of depression. If possible, the date of a negative life event connected to the depression was used as the date of onset; if not, the date was estimated as the midpoint of the year (expected value of a uniformly distributed date). In the same way, the last time that the symptoms were present was estimated. To ensure validity of time-related symptom questions, memory aids using benchmarks such as before/after the war, marriage, having children, summer/winter, and so on, are used in CIDI-M 1.1. Smoking and somatic disease restraints were also assessed at T2 (table 1).

Study cohort.
The 728 persons not reinterviewed defines the population lost to follow-up; 384 were not eligible (260 had died, 68 were severely ill, 33 had emigrated, and 23 could not be found). Among the 1,630 eligible were 186 who refused and 158 who could not be reached, leaving 1,286 interviews at T2 (79 percent response rate). For 96 respondents, CIDI estimated the onset of a first depressive episode before T1 (pre-T1 depression), and these respondents were excluded. The study cohort (included population) consisted of the 1,190 remaining respondents with complete data on both interviews and no estimated depression onset prior to T1. The study was approved by the Norwegian ethical board, and informed consent was obtained from all participants at both T1 and T2.

Alternative explanations
Plausible alternatives to a direct causal influence of smoking on depression explaining an observed association included one or more of the following, described below.

Selection bias.
The smoking–depression association was evaluated across levels of both the Hopkins Symptom Checklist 25 and age at T1.

Recall bias.
Exclusion rates were compared across levels of smoking. Robustness of the smoking–depression association for measurement error regarding the estimated onset of depression was assessed.

Unmeasured confounding.
To assess the effect of a potential dichotomous unmeasured confounder (U), we performed a sensitivity analysis, described by Greenland (30Go). For simplicity, we used the crude odds ratio (OR) with corresponding cases/noncases and heavy smoking/no smoking. We assessed how strong the assumed simultaneous associations U–smoking (ORSU) and U–depression (ORDU) would have to be to account for a substantial part of the observed smoking–depression association (ORDS). The adjusted ORDS was calculated as a function of confounder prevalence among smokers (PU1), varying ORSU and ORDU. The minimum of this function represented the maximum possible reduction in ORDS due to confounding. We adjusted for Hopkins Symptom Checklist 25 at T1 to assess the influence of poor mental health at T1 and some unmeasured confounding.

Indirect causation.
The effect of decline in somatic health on the smoking–depression association was evaluated, for each disease separately and combined, by assuming that all decline occurred before depression onset. In addition, we assessed the influence of those who reported that they had stopped smoking after T1 and before the estimated onset of depression.

Reverse causation.
On the basis of recall from T2, smoking initiation before and after estimated depression onset was compared with respect to smoking at T1 and T2. We compared smoking at T1 between those with and without a pre-T1 depression. The pre-T1-, FUP- and no-depression groups were compared with respect to change in smoking behavior from T1 to T2. The change was also compared between the recent pre-T1 and the distant pre-T1 depressions.

Statistical analyses
The included and the lost-to-follow-up populations were compared with two-tailed t tests/F tests and Fisher's exact test. To estimate relative incidence rates (hazard ratios) adjusted for relevant baseline covariates, Cox proportional hazards regression was used. The time scale was observation time, and right censoring at the end of the study occurred for those without depression. In this paper, odds ratios and hazard ratios are presented with 95 percent confidence intervals in parentheses. When adjusting for covariates, model selection followed increasing likelihood with all combinations of covariates compared. For more flexible fitting, smoothing splines were used for selected continuous covariates. Both global and covariate-specific tests for assumptions of proportional hazards were performed (31Go).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 References
 
The groups lost to follow-up because of death and for other reasons were compared with the included population with respect to baseline covariates (table 2). Compared with the included population, both lost-to-follow-up populations were more likely to have psychological distress, loss of partner, low income levels, and low educational levels. The group lost to follow-up for other reasons was similar to the included group with respect to smoking, but it contained the most Oslo respondents and drank more alcohol. Those who died had smoked the least at T1 and had fewer relationship and external events but more somatic health problems.


View this table:
[in this window]
[in a new window]
  		TABLE 2. Comparison of baseline covariates for 1,895 respondents in the lost-to-follow-up and included groups from a Norwegian general population, 1990–2001

 
In the included population, most light smokers at T1 were women, while most heavy smokers were men (table 3). Among the never smokers, the majority had high levels of education, whereas low levels of education were most common among heavy smokers. Furthermore, alcohol index and external events were strongly associated with heavy smoking. The prevalence of daily cigarette smoking was 41.4 percent, whereas prevalences of daily cigar and pipe smoking were 0.9 percent and 1.2 percent, respectively.


View this table:
[in this window]
[in a new window]
  		TABLE 3. Baseline covariates versus categories of smoking for 1,190 respondents from a Norwegian general population, 1990–2001

 
Compared with never smokers, smokers had an increasing crude risk of depression with increasing amount of current smoking and a significant risk ratio for more than 10 cigarettes per day and more than 20 cigarettes per day (OR = 3.66, 95 percent confidence interval (CI): 1.63, 8.19) (table 4). The crude risk ratio for cumulative smoking was also significant, with an increase of 37 percent for an increase in smoking of 200 cigarette-years. Other significant crude risk ratios were observed for gender, age, education, loss of partner, alcohol index, moderate somatic disease restraint, network events, relationship events, and a Hopkins Symptom Checklist 25 score of 1.75 or more.


View this table:
[in this window]
[in a new window]
  		TABLE 4. Baseline covariates versus a first depressive episode occurring during follow-up for 1,190 respondents from a Norwegian general population, 1990–2001

 
The smoking–depression association adjusted for baseline covariates was evaluated with Cox regression, and the results are given in tables 5 and 6 (best fitted model shown). Assumptions of proportionality were found valid with the global test for zero slope in a generalized linear regression of Schoenfeld residuals not rejected (p = 0.48) and for individual covariates. Compared with never smoking, past and low current amounts of smoking were associated with a higher risk of depression, with significant risk ratios for 11–20 cigarettes per day (hazard ratio (HR) = 2.01, 95 percent CI: 1.17, 3.43) and for more than 20 cigarettes per day (HR = 4.34, 95 percent CI: 1.85, 10.18) (table 5). Current smokers had a higher risk of depression than past smokers and a significantly higher risk than never smokers (HR = 1.70, 95 percent CI: 1.08, 2.70). Other significant risk ratios were observed for gender, high educational level, high alcohol consumption, relationship events, and moderate and severe somatic disease restraint. The risk of depression associated with a high educational level was fairly homogeneous across levels of gender and residence, but it was highest for women in Lofoten. In contrast, high score on the Hopkins Symptom Checklist 25 and pre-T1 depression were associated with low educational level. No significant interactions were found.


View this table:
[in this window]
[in a new window]
  		TABLE 5. Hazard ratios for a first depressive episode occurring during follow-up within categories of smoking, adjusted for baseline covariates, for 1,190 respondents from a Norwegian general population, 1990–2001

 

View this table:
[in this window]
[in a new window]
  		TABLE 6. Hazard ratios for a first depressive episode occurring during follow-up within categories of smoking time, adjusted for baseline covariates and stratified by age at T1,* for 1,190 respondents from a Norwegian general population, 1990–2001

 
A dose-response relation between increasing smoking time and depression was observed, with significant risk ratios for more than 20 years of smoking and more than 20 cigarettes per day, also when stratified on age at T1 (table 6). Age differences at T1 showed a higher risk ratio for gender in the young group and for relationship events and moderate somatic disease restraint in the older group. Wide confidence intervals for the older group were due to having only 21 cases of depression. Cumulative smoking adjusted for covariates in table 5 was also associated with a significant risk ratio (HR = 1.66, 95 percent CI: 1.36, 2.03); otherwise, risk ratios were similar to those shown in table 5.

Estimates of cumulative hazard rates illustrate the dose responses for amount of current smoking and smoking time (not stratified by age at T1) (figure 1, top panels). The higher growth rate near T2 indicated increased hazard. The spline fit of cumulative smoking showed a significant drop in hazard for 100 cigarette-years and otherwise confirmed the assumed linearity in our model (as for alcohol index).


Figure 1
View larger version (26K):
[in this window]
[in a new window]
  		FIGURE 1. Estimates of the cumulative hazard for fixed values of covariates (women, no relationship events, medium educational level, and mean level of alcohol index) and spline fit of the cumulative smoking and alcohol effect for 1,190 respondents from a Norwegian general population, 1990–2001. In the bottom panels, each vertical tick along the x-axes represents one value (many ties for cumulative smoking). cig, cigarettes.

 
Alternative explanations
Selection bias.
High scores on the Hopkins Symptom Checklist 25 and high age at T1 were overrepresented in the lost-to-follow-up group (table 2). The smoking–depression association was homogeneous across levels of both the Hopkins Symptom Checklist 25 and age at T1.

Recall bias.
Among the 226 respondents with depression, the exclusion rate for those with symptoms within the month before T2 was higher than for those without symptoms (51 percent vs. 39 percent). This finding was also observed across levels of smoking. Excluding all respondents whose estimated onset of depression occurred during the first half of FUP increased the risk ratio for heavy smoking slightly; otherwise, the results were similar to those shown in table 5.

Unmeasured confounding.
Adjusting for Hopkins Symptom Checklist 25 at T1 as a continuous covariate showed significant risk ratios for 10–20 cigarettes per day (HR = 1.82, 95 percent CI: 1.07, 3.09) and for more than 20 cigarettes per day (HR = 3.65, 95 percent CI: 1.56, 8.52).

Figure 2 shows the results of the sensitivity analysis in which the unmeasured confounder was thought of as a dichotomous genetic susceptibility. The three curves show ORDS as functions of PU1 with fixed values of ORSU and ORDU and that this function has a unique minimum. The symmetric curve represents ORSU = ORDU = 7 (the strongest reported single-gene effect for depression (32Go)). The lowest possible ORDS in this instance is 1.6, corresponding to 12 percent in the population having the genetic susceptibility (increases with PU1). In the asymmetric situation in which ORSU or ORDU is set equal to 3.66 (the observed ORDS), the other odds ratio has to be 22 to obtain the same minimum. To make ORDS = 1 in the symmetric situation, ORSU and ORDU have to be higher than 12. With one of them at approximately 2, ORDS will always be higher than 1.6.


Figure 2
View larger version (18K):
[in this window]
[in a new window]
  		FIGURE 2. Variation in the odds ratio (OR) between smoking (S) and depression (D) (ORDS), with fixed values for confounding odds ratios (ORSU and ORDU) as a function of prevalence P of a dichotomous confounder U among smokers (PU1): 1,190 respondents from a Norwegian general population, 1990–2001.

 
Indirect causation.
Somatic decline during FUP was reported by 711 respondents, with prevalences of depression of 12.2 percent among those with a decline and 9.6 percent among those without one. When adjusting for decline and other potential confounders of the relation between decline and depression, such as somatic health, age, Hopkins Symptom Checklist 25, body mass index, and lifestyle factors at T1 (refer to table 1), we observed minor changes in the smoking–depression association together with a significant risk ratio for somatic decline. Worsening musculoskeletal disorders were associated with the highest risk ratio (HR = 1.79, 95 percent CI: 1.1, 2.89), followed by cardiovascular decline (myocardial infarction, angina pectoris, and high blood pressure) (HR = 1.7, 95 percent CI: 0.86, 3.36). However, many of those respondents with a serious heart condition at T1 were lost to follow-up. Of 37 respondents reporting severe restraint because of cardiovascular disease at T1, 23 were lost to follow-up; of these respondents, 10 men and nine women died, all older than age 55 years at T1. Six respondents reported that they had stopped smoking between T1 and the estimated onset of depression. Removing their data from the analysis had a negligible effect on the risk ratios.

Reverse causation.
For 35 of the 226 depression cases, estimated onset was prior to smoking initiation; for 129 cases, smoking was initiated first; and 58 cases of depression were observed among never smokers. The prevalences of ever smoking were 63 percent among the nondepressed population and 74 percent among those with depression. Those whose estimated onset occurred prior to smoking initiation smoked less than those who started smoking first, both at T1 (15 percent vs. 46 percent smoked >10 cigarettes per day) and at T2. The median time from smoking initiation to estimated FUP depression onset was 18 years. Those experiencing a pre-T1 depression did not smoke significantly more at T1 than the included population did (p = 0.31), but those who smoked more than 10 cigarettes per day at T1 had a significantly higher prevalence of FUP depression than the rest of the respondents (p = 0.005). When we regarded change in smoking behavior during FUP, no differences were found between the pre-T1-, FUP-, and no-depression groups (p = 0.51). For those experiencing a pre-T1 depression, those whose estimated onset occurred within 10 years before T1 (39 of 96 cases) reduced their smoking significantly more during FUP than those whose estimated onset occurred more than 10 years before T1 (p = 0.01).


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
References
 
The results of this analysis confirm a strong association between cigarette smoking and subsequent depression. The risk of depression was four times as high for heavy smokers compared with never smokers. We found dose-response relations with increasing risk for past smokers and for increasing number of cigarettes smoked per day for current smokers and also for increasing number of smoking-years. The effect of total cumulative smoking showed a significant increase in risk of 66 percent per 200 cigarette-years. Furthermore, the spline fit of cumulative smoking showed a decreased risk for less than 200 cigarette-years and a linear increase for more than 200 cigarette-years. This finding may reflect a short-term positive effect, supporting other findings (19Go), and a long-term negative effect. The same pattern, less pronounced, was observed for alcohol. Inhibition of monoamine oxidase B in the brains of smokers, probably due to pharmacologic factors, has been observed, which might be linked to depression through its effect on neurotransmitters such as serotonin (20Go, 32Go). An analysis similar to ours, one between nicotine dependence and depression, is a direction for future study. Daily cigar and pipe smoking were not analyzed because of low prevalences.

Alternative explanations
Selection bias.
Because the sample was randomly drawn, selections affecting the sample, and possibly the smoking–depression association, were nonresponse at T1, loss to follow-up at T2, and selection of smokers. Even if the lost-to-follow-up population experienced more psychological distress and was older than the included population, there was no support for a different smoking effect in these groups. No information regarding nonresponse at T1 was available, but similarity to the lost-to-follow-up group can be argued. Selection bias could also have occurred if various childhood and genetic factors led to recruitment of smokers from only those groups whose mental health was poor. If the mental health of those who were able to quit smoking was better than that of the rest of the smokers, this factor could explain the dose-response relation found between smoking time and depression, and also that former smokers had a lower risk than current smokers. Mean age at T1 for those with a FUP depression was 36 years, corresponding to smoking initiation in the early 1970s, because smoking is usually initiated before age 18 years (33Go). Prevalence of daily smoking has decreased steadily in the Norwegian adult population, from 42 percent (52 percent of men vs. 32 percent of women) in 1973 to 26 percent in 2004 (27 percent vs. 25 percent), accounted for by the decrease among those with a high level of education (34Go, 35Go). This finding implies higher prevalence, especially among those with high educational levels, in the early 1970s. The same is observed in the US population (33Go, 36Go, 37Go). Even distribution of smokers across levels of education and other measures of socioeconomic status (38Go), together with more smoking among men, who contributed only 28 percent of the cases of depression, makes recruiting mostly from groups of poor mental health unlikely.

On the other hand, having a high level of education probably requires mental strength, maybe explaining our observed association between the pre-T1 depressions and low educational level. Mean age at estimated onset was 26 years for those with pre-T1 depressions and 43 years for those with FUP depressions. Removing data on the pre-T1 depressions resulted in removing most of those respondents whose education might have been ruined by early onset, thereby making our association between high educational level and FUP depression plausible.

Recall bias.
In an early version of CIDI, the validity of time-related items was found to be questionable in a clinical sample studied for nonpsychotic disorders including depression (39Go). However, this version lacked memory aids, included in CIDI-M 1.1. Another study of a clinical sample showed acceptable validity regarding diagnostic discrimination in CIDI (40Go). A review concluded with excellent interrater reliability, good test-retest reliability, and adequate validity, but almost no results came from general population samples (41Go). With the restricted information available, the validity of a CIDI lifetime diagnosis is acceptable, whereas age at onset might be less accurate than desired. Both remembering of earlier episodes better when having had a recent/current episode and "shrinkage" of elapsed time since various events occurred are characteristics of recall (42Go, 43Go), directly affecting our exclusion of pre-T1 depressions (estimated onset prior to T1). Even though these characteristics probably apply to our data, they did not seem to affect our smoking–depression association, causing recall bias. The exclusion rate for those with recent symptoms at T2 was higher than for those without them, but this finding was also observed across levels of smoking. Furthermore, the increasing incidence of depression during FUP (figure 1) was probably partly due to measurement error in placing the onsets too close to T2. However, even 5.5 years of misplacement for all estimated onsets barely changed the risk ratios, indicating robustness of the smoking–depression association regarding misplacement of onset.

Unmeasured confounding.
Many childhood, personality, and genetics factors may have represented unmeasured confounding. Adjusting for the Hopkins Symptom Checklist 25 at T1 should have reduced this confounding, but only a minor change in the smoking–depression association was observed.

Rebelliousness has been found to account for the smoking–depression association among adolescents (7Go). Adjustment in our model for external events similar to rebelliousness should have reduced the confounding from personality traits linked to risk behavior, but doing so had a negligible effect on the smoking–depression association. There is evidence of a genetic influence on smoking (44Go, 45Go), with single-gene relative risks of less than 2 (46Go), and on depression, with most single-gene relative risks at about 2 (32Go), and of simultaneous genetic influence (15Go, 47Go). In a study of female twins, genetic confounding accounted for most of the association between lifetime smoking and lifetime depression (15Go). However, when lifetime prevalences of smoking and depression were used, the association was a mixture of both causal directions, with one of them potentially weak, thus leading to attenuation. A high amount of baseline smoking also predicted new cases of depression, but this result was not analyzed in terms of zygosity. In contrast to these results, boys' smoking behavior has been found to be mostly mediated by environmental factors (48Go). The sensitivity analysis showed that our observed crude odds ratio was robust for a dichotomous unmeasured confounder, requiring confounding odds ratios of 12 to account for the observed smoking–depression association.

Indirect causation.
Decline in somatic health and smoking cessation were considered possible intermediate states in a causal pathway between smoking and depression. We found a significant risk ratio for somatic decline when we assumed that change occurred before depression onset. However, the minor change in the risk ratios for smoking after we also adjusted for potential confounders for the relation between somatic decline and depression supported a strong, direct effect of smoking. This relies on the assumption of no other strong, unmeasured confounder between somatic decline and depression (49Go, 50Go). Cardiovascular decline was expected to have a strong influence on depression (51Go), but 62 percent of those with severe heart problems at T1 and therefore with a high risk of becoming depressed were lost to follow-up.

Onset of depression during smoking cessation has been observed (52Go). However, we found no support for this result.

Reverse causation.
For our smoking–depression association to be explained by a causal influence of depression on smoking, latent pre-T1 depressions must have initiated pre-T1 smoking or increased smoking at T1 and then been manifested during FUP. Only 15.5 percent of the estimated depression onsets occurred prior to smoking initiation, and the prevalence of ever smoking in the nondepressed group was high (62.8 percent). In addition, those whose estimated depression onset occurred before smoking initiation smoked less at both T1 and T2 than those for whom smoking initiation occurred first. The median time from smoking initiation to estimated FUP depression onset was 18 years, which seems to be an unreasonably long latency period. The pre-T1 depressions were not strongly associated with smoking at T1, but smoking at T1 was strongly associated with the FUP depressions, so a mixture of associations from both causal directions (the former) seemed weaker than when smoking occurred first (the latter). The pre-T1-, FUP-, and no-depression groups seemed to exhibit the same pattern of change in smoking during FUP. In the pre-T1 depression group, those with a recent depression prior to T1 actually reduced their smoking during FUP more than those with a more distant estimated onset. In sum, no support was found for the manifested pre-T1 depressions (representing the latent pre-T1 depressions) to be an important reason to start smoking or having increased subsequent smoking.

Conclusion
We observed a fourfold higher risk of depression for heavy smokers compared with never smokers. Failure to explain the smoking–depression association with other plausible alternatives might reflect a direct causal influence of smoking on depression.


   ACKNOWLEDGMENTS
 
This work was supported by grant 2001/2/0012 from The Norwegian Council for Mental Health, The Norwegian Research Council, The Norwegian Women's Public Health Association, Dr. Trygve Gythfeldt and Wife Research Foundation, Haldis and Josef Andresens Legacy, Proprietor Jonn Nilsen and wife Maja Jonn-Nilsens Legacy for Promotion of Norwegian Psychiatric Research, Per Risteigens Legacy, and Sommers Legacy.

The authors thank Dr. Odd Aalen for helpful comments.

Conflict of interest: none declared.


   References
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 

  1. Murray CJ, Lopez AD. Alternative projections of mortality and disability by cause 1990–2020: Global Burden of Disease Study. Lancet 1997;349:1498–504.[CrossRef][Web of Science][Medline]
  2. Haines AP, Imeson JD, Meade TW. Psychoneurotic profiles of smokers and non-smokers. Br Med J 1980;280:1422.[Free Full Text]
  3. Frerichs RR, Aneshensel CS, Clark VA, et al. Smoking and depression: a community survey. Am J Public Health 1981;71:637–40.[Abstract/Free Full Text]
  4. Fergusson DM, Lynskey MT, Horwood LJ. Comorbidity between depressive disorders and nicotine dependence in a cohort of 16-year-olds. Arch Gen Psychiatry 1996;53:1043–7.[Abstract/Free Full Text]
  5. Fergusson DM, Goodwin RD, Horwood LJ. Major depression and cigarette smoking: results of a 21-year longitudinal study. Psychol Med 2003;33:1357–67.[CrossRef][Web of Science][Medline]
  6. Goodman E, Capitman J. Depressive symptoms and cigarette smoking among teens. Pediatrics 2000;106:748–55.[Abstract/Free Full Text]
  7. Albers AB, Biener L. The role of smoking and rebelliousness in the development of depressive symptoms among a cohort of Massachusetts adolescents. Prev Med 2002;34:625–31.[CrossRef][Web of Science][Medline]
  8. Patton GC, Carlin JB, Coffey C, et al. Depression, anxiety, and smoking initiation: a prospective study over 3 years. Am J Public Health 1998;88:1518–22.[Abstract/Free Full Text]
  9. Choi WS, Patten CA, Gillin JC, et al. Cigarette smoking predicts development of depressive symptoms among U.S. adolescents. Ann Behav Med 1997;19:42–50.[Web of Science][Medline]
  10. Ismail K, Sloggett A, De Stavola B. Do common mental disorders increase cigarette smoking? Results from five waves of a population-based panel cohort study. Am J Epidemiol 2000;152:651–7.[Abstract/Free Full Text]
  11. Jorm AF, Rodgers B, Jacomb PA, et al. Smoking and mental health: results from a community survey. Med J Aust 1999;170:74–7.[Web of Science][Medline]
  12. Breslau N, Kilbey M, Andreski P. Nicotine dependence, major depression, and anxiety in young adults. Arch Gen Psychiatry 1991;48:1069–74.[Abstract/Free Full Text]
  13. Breslau N, Kilbey MM, Andreski P. Nicotine dependence and major depression. New evidence from a prospective investigation. Arch Gen Psychiatry 1993;50:31–5.[Abstract/Free Full Text]
  14. Breslau N, Peterson EL, Schultz LR, et al. Major depression and stages of smoking. A longitudinal investigation. Arch Gen Psychiatry 1998;55:161–6.[Abstract/Free Full Text]
  15. Kendler KS, Neale MC, MacLean CJ, et al. Smoking and major depression. A causal analysis. Arch Gen Psychiatry 1993;50:36–43.[Abstract/Free Full Text]
  16. Cervilla J, Prince M, Joels S, et al. Genes related to vascular disease (APOE, VLDL-R, DCP-1) and other vascular factors in late-life depression. Am J Geriatr Psychiatry 2004;12:202–10.[CrossRef][Web of Science][Medline]
  17. Lam TH, Li ZB, Ho SY, et al. Smoking and depressive symptoms in Chinese elderly in Hong Kong. Acta Psychiatr Scand 2004;110:195–200.[CrossRef][Web of Science][Medline]
  18. Hernan MA, Brumback B, Robins JM. Marginal structural models to estimate the joint causal effect of nonrandomized treatments. J Am Stat Assoc 2001;96:440–8.[CrossRef][Web of Science]
  19. Pomerleau CS, Pomerleau OF. Euphoriant effects of nicotine in smokers. Psychopharmacology 1992;108:460–5.[CrossRef][Medline]
  20. Fowler JS, Volkow ND, Wang GJ, et al. Inhibition of monoamine oxidase B in the brains of smokers. Nature 1996;379:733–6.[CrossRef][Medline]
  21. Roy K, Parker G, Mitchell P, et al. Depression and smoking: examining correlates in a subset of depressed patients. Aust N Z J Psychiatry 2001;35:329–35.[CrossRef][Web of Science][Medline]
  22. Sandanger I, Nygard JF, Ingebrigtsen G, et al. Prevalence, incidence and age at onset of psychiatric disorders in Norway. Soc Psychiatry Psychiatr Epidemiol 1999;34:570–9.[CrossRef][Web of Science][Medline]
  23. Derogatis LR, Lipman RS, Rickels K, et al. The Hopkins Symptom Checklist (HSCL): a self-report symptom inventory. Behav Sci 1974;19:1–15.[Medline]
  24. Rickels K, Garcia CR, Lipman RS, et al. The Hopkins Symptom Checklist. Assessing emotional distress in obstetric-gynecologic practice. Prim Care 1976;3:751–64.[Medline]
  25. Winokur A, Winokur DF, Rickels K, et al. Symptoms of emotional distress in a family planning service: stability over a four-week period. Br J Psychiatry 1984;144:395–9.[Abstract/Free Full Text]
  26. Sandanger I, Moum T, Ingebrigtsen G, et al. Concordance between symptom screening and diagnostic procedure: the Hopkins Symptom Checklist-25 and the Composite International Diagnostic Interview I. Soc Psychiatry Psychiatr Epidemiol 1998;33:345–54.[CrossRef][Web of Science][Medline]
  27. Sandanger I, Moum T, Ingebrigtsen G, et al. The meaning and significance of caseness: the Hopkins Symptom Checklist-25 and the Composite International Diagnostic Interview. II. Soc Psychiatry Psychiatr Epidemiol 1999;34:53–9.[Medline]
  28. Goldberg DP. The detection of psychiatric illness by questionnaire: a technique for the identification and assessment of non-psychotic psychiatric illness. Oxford, United Kingdom: Oxford University Press, 1972.
  29. Robins LN, Wing J, Wittchen HU, et al. The Composite International Diagnostic Interview. An epidemiologic instrument suitable for use in conjunction with different diagnostic systems and in different cultures. Arch Gen Psychiatry 1988;45:1069–77.[Abstract/Free Full Text]
  30. Greenland S. Basic methods for sensitivity analysis of biases. Int J Epidemiol 1996;25:1107–16.[Free Full Text]
  31. Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994;81:515–26.[Abstract/Free Full Text]
  32. Morley KI, Hall WD, Carter L. Genetic screening for susceptibility to depression: can we and should we? Aust N Z J Psychiatry 2004;38:73–80.[CrossRef][Web of Science][Medline]
  33. Giovino GA, Henningfield JE, Tomar SL, et al. Epidemiology of tobacco use and dependence. Epidemiol Rev 1995;17:48–65.[Free Full Text]
  34. StatBank Norway. Health, social conditions, social services and crime. Health conditions. Smoking habits. Statistics Norway, 2005 (http://statbank.ssb.no).
  35. Lindbak R, Lund M. Tobacco statistics in Norway 1973–2003. (In Norwegian). Directorate for Health and Social Affairs, Department for Tobacco Control, 2005 (http://www.shdir.no/tobakk/english/).
  36. Fiore MC, Novotny TE, Pierce JP, et al. Trends in cigarette smoking in the United States. The changing influence of gender and race. JAMA 1989;261:49–55.[Abstract/Free Full Text]
  37. Pierce JP, Fiore MC, Novotny TE, et al. Trends in cigarette smoking in the United States. Educational differences are increasing. JAMA 1989;261:56–60.[Abstract/Free Full Text]
  38. Lund KE, Lund M. Smoking and inequality in Norway 1998–2000. (In Norwegian). Tidsskr Nor Laegeforen 2005;125:560–3.[Medline]
  39. Wittchen HU, Burke JD, Semler G, et al. Recall and dating of psychiatric symptoms. Test-retest reliability of time-related symptom questions in a standardized psychiatric interview. Arch Gen Psychiatry 1989;46:437–43.[Abstract/Free Full Text]
  40. Peters L, Andrews G. Procedural validity of the computerized version of the Composite International Diagnostic Interview (CIDI-Auto) in the anxiety disorders. Psychol Med 1995;25:1269–80.[Web of Science][Medline]
  41. Andrews G, Peters L. The psychometric properties of the Composite International Diagnostic Interview. Soc Psychiatry Psychiatr Epidemiol 1998;33:80–8.[CrossRef][Web of Science][Medline]
  42. Eaker S, Adami HO, Sparen P. Reasons women do not attend screening for cervical cancer: a population-based study in Sweden. Prev Med 2001;32:482–91.[CrossRef][Web of Science][Medline]
  43. Thompson R, Bogner HR, Coyne JC, et al. Personal characteristics associated with consistency of recall of depressed or anhedonic mood in the 13-year follow-up of the Baltimore Epidemiologic Catchment Area survey. Acta Psychiatr Scand 2004;109:345–54.[CrossRef][Web of Science][Medline]
  44. Madden PA, Heath AC, Pedersen NL, et al. The genetics of smoking persistence in men and women: a multicultural study. Behav Genet 1999;29:423–31.[CrossRef][Web of Science][Medline]
  45. Kendler KS, Neale MC, Sullivan P, et al. A population-based twin study in women of smoking initiation and nicotine dependence. Psychol Med 1999;29:299–308.[CrossRef][Web of Science][Medline]
  46. Lerman C, Berrettini W. Elucidating the role of genetic factors in smoking behavior and nicotine dependence. Am J Med Genet 2003;118B:48–54.
  47. Audrain-McGovern J, Lerman C, Wileyto EP, et al. Interacting effects of genetic predisposition and depression on adolescent smoking progression. Am J Psychiatry 2004;161:1224–30.[Abstract/Free Full Text]
  48. Silberg J, Rutter M, D'Onofrio B, et al. Genetic and environmental risk factors in adolescent substance use. J Child Psychol Psychiatry 2003;44:664–76.[CrossRef][Web of Science][Medline]
  49. Cole SR, Hernan MA. Fallibility in estimating direct effects. Int J Epidemiol 2002;31:163–5.[Abstract/Free Full Text]
  50. Hernan MA, Hernandez-Diaz S, Robins JM. A structural approach to selection bias. Epidemiology 2004;15:615–25.[CrossRef][Web of Science][Medline]
  51. Everson-Rose SA, Lewis TT. Psychosocial factors and cardiovascular diseases. Annu Rev Public Health 2005;26:469–500.[CrossRef][Web of Science][Medline]
  52. Killen JD, Fortmann SP, Schatzberg A, et al. Onset of major depression during treatment for nicotine dependence. Addict Behav 2003;28:461–70.[CrossRef][Web of Science][Medline]
  53. Sandanger I, Nygard JF, Sorensen T, et al. Is women's mental health more susceptible than men's to the influence of surrounding stress? Soc Psychiatry Psychiatr Epidemiol 2004;39:177–84.[CrossRef][Web of Science][Medline]
  54. Avison WR, Turner RJ. Stressful life events and depressive symptoms: disaggregating the effects of acute stressors and chronic strains. J Health Soc Behav 1988;29:253–64.[CrossRef][Web of Science][Medline]
  55. Paykel ES. The Interview for Recent Life Events. Psychol Med 1997;27:301–10.[CrossRef][Web of Science][Medline]
  56. Saunders JB, Aasland OG. WHO collaborative on identification and treatment of persons with harmful alcohol consumption. Oslo, Norway: World Health Organization, 1987:97.

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 Psychosom. Med.Home page
K. J. Anstey, R. Burns, P. Butterworth, T. D. Windsor, H. Christensen, and P. Sachdev
Cardiovascular Risk Factors and Life Events as Antecedents of Depressive Symptoms in Middle and Early-Old Age: Path Through Life Study
Psychosom Med, November 1, 2009; 71(9): 937 - 943.
[Abstract] [Full Text] [PDF]

Home page
 IBMS BoneKEyHome page
I. Bab and R. Yirmiya
Depression, Selective Serotonin Re-Uptake Inhibitors and the Regulation of Bone Mass
IBMS BoneKEy, January 1, 2009; 6(1): 8 - 15.
[Abstract] [Full Text] [PDF]

Home page
 Br. J. PsychiatryHome page
J. A. Pasco, L. J. Williams, F. N. Jacka, F. Ng, M. J. Henry, G. C. Nicholson, M. A. Kotowicz, and M. Berk
Tobacco smoking as a risk factor for major depressive disorder: population-based study
The British Journal of Psychiatry, October 1, 2008; 193(4): 322 - 326.
[Abstract] [Full Text] [PDF]

Home page
 Schizophr BullHome page
R. Kotov, L. T. Guey, E. J. Bromet, and J. E. Schwartz
Smoking in Schizophrenia: Diagnostic Specificity, Symptom Correlates, and Illness Severity
Schizophr Bull, June 17, 2008; (2008) sbn066v1.
[Abstract] [Full Text] [PDF]

Home page
 Am J EpidemiolHome page
A. Kouvonen, T. Oksanen, J. Vahtera, M. Stafford, R. Wilkinson, J. Schneider, A. Vaananen, M. Virtanen, S. J. Cox, J. Pentti, et al.
Low Workplace Social Capital as a Predictor of Depression: The Finnish Public Sector Study
Am. J. Epidemiol., May 15, 2008; 167(10): 1143 - 1151.
[Abstract] [Full Text] [PDF]

Home page
 Adv. Psychiatr. Treat.Home page
J. Campion, K. Checinski, J. Nurse, and A. McNeill
Smoking by people with mental illness and benefits of smoke-free mental health services
Adv. Psychiatr. Treat., May 1, 2008; 14(3): 217 - 228.
[Abstract] [Full Text] [PDF]

Home page
 Am J EpidemiolHome page
K. Sheikh
RE: "CIGARETTE SMOKING AND INCIDENCE OF FIRST DEPRESSIVE EPISODE: AN 11-YEAR, POPULATION-BASED FOLLOW-UP STUDY"
Am. J. Epidemiol., November 1, 2006; 164(9): 918 - 919.
[Full Text] [PDF]

Home page
 Am J EpidemiolHome page
O. Klungsoyr and J. F. Nygard
TWO OF THE AUTHORS REPLY
Am. J. Epidemiol., November 1, 2006; 164(9): 919 - 920.
[Full Text] [PDF]

Home page
 Am J EpidemiolHome page
C. M. Ross
RE: "CIGARETTE SMOKING AND INCIDENCE OF FIRST DEPRESSIVE EPISODE: AN 11-YEAR, POPULATION-BASED FOLLOW-UP STUDY"
Am. J. Epidemiol., November 1, 2006; 164(9): 917 - 918.
[Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
163/5/421    most recent
kwj058v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (14)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Klungsøyr, O.
Right arrow Articles by Sandanger, I.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Klungsøyr, O.
Right arrow Articles by Sandanger, I.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?