The association between depression and smoking is well established (1–4). Smoking cessation studies have reported a high proportion of smokers (median=32.7%, range=18.5%–60.6%) coming in for treatment with a lifetime history of major depression (4–13). Evidence, mostly from case studies, suggests some individuals are at a higher risk for developing major depressive episodes after smoking cessation, particularly those who have a history of depression (5, 14–16).
Reports of severe major depressive episodes after smoking cessation indicate that the onset of severe depressive symptoms ranges from 2 days to 6 weeks after the initial abstinence from smoking (3, 14, 16–21). In some cases, depression after smoking cessation was resolved with the use of nicotine replacement therapy (14, 19) or the use of antidepressants (14); in others, depressive symptoms dissipated after a relapse to smoking (16, 17). In a study examining the effectiveness of clonidine for smoking cessation (8), nine participants (3%) developed severe major depressive episodes during the 10-week treatment; seven patients (77.8%) had a history of depression. In another trial of smoking cessation using fluoxetine (5), five subjects (7%) were diagnosed with major depressive episodes during the 10-week treatment, and all had a history of depression.
Covey and colleagues (15) examined the predictors of major depressive episodes among 126 smokers who were abstinent at the end of the 10-week, placebo-controlled trial of smoking cessation with clonidine (8). During the 3-month follow-up, 7.1% (N=9) of the participants experienced major depressive episodes with remarkably higher rates of incidence among smokers with a history of depression (four [17%] of the 24 subjects with a history of single episodes and three [30%] of the 10 subjects with a history of recurrent episodes). The significant predictors of major depressive episodes were having a history of major depression and experiencing elevated withdrawal symptoms at the end of treatment. This study included only smokers who were abstinent; therefore, it was unclear whether cigarette abstinence induced major depression (15).
In summary, smoking cessation may be followed by new major depressive episodes, particularly among smokers with a history of depression. To better understand the role of abstinence status, history of depression, and other factors involved in major depression after treatment for smoking cessation, we included smokers who achieved abstinence at the end of treatment and those who relapsed or continued to smoke. The purpose of this study was to investigate the 12-month incidence of major depressive episodes after treatment for smoking cessation and the frequency of depressive symptoms and related impairments and to explore the predictors associated with major depressive episodes after treatment. In addition, two hypotheses were tested: 1) abstinence status at the end of treatment increased the likelihood of major depressive episodes after treatment; and 2) smokers with a history of depression were more likely to develop major depressive episodes after treatment.
Data analyses in this study were based on 304 participants (87.4%) of a total of 348 smokers enrolled in one of two separate smoking cessation trials. These 304 participants were assessed at the end of treatment and at one or more of three follow-up assessments for identifying the incidence of major depressive episodes. Demographic and baseline characteristics were similar between the participants who were included in the current analyses and those who were excluded (N=44), but there were proportionately more men in the excluded group (N=27 of 44, 61.4%, versus N=132 of 304, 43.4%, respectively) (χ2=4.99, df=1, p=0.02, N=348). Subject characteristics and treatment conditions are presented in t1.
The two clinical trials of smoking cessation were conducted at the same clinic in San Francisco. Trial 1 used nicotine gum in combination with psychological treatment for smoking cessation (10). Trial 2 compared the use of nortriptyline and placebo in combination with psychological treatment (12). Recruitment and treatment procedures were similar across trials, as were the incidence rates of major depressive episodes after treatment. We combined the data from these trials to increase statistical power.
Participants in both trials were recruited through newspaper ads and public service announcements. After a telephone screening, potential participants were invited to an orientation in which the study was described in detail and informed consent was obtained. The inclusion criteria were the same for both trials. Participants were between the ages of 18 and 65 and smoked more than 10 cigarettes per day. Individuals were excluded if they were taking prescribed psychoactive drugs or undergoing psychiatric treatment, experiencing alcohol or nonnicotine drug problems, or experiencing medical conditions that would interfere with treatment for smoking cessation. Individuals who met the criteria for a major depressive episode within 6 months before the baseline assessment in trial 1 and within 3 months in trial 2 were excluded and referred for treatment of depression.
After completion of baseline measures, the participants in both trials were randomly assigned to receive either 10 sessions of cognitive behavioral mood management intervention or five sessions of health education over 8 weeks. Details of the interventions are described elsewhere (10). All participants were assigned a group quit date at the second week of the psychological intervention. In trial 1, all participants received 2 mg of nicotine gum from the quit date (week 2) through week 12. In trial 2, participants were randomly assigned to receive either nortriptyline or placebo from week 1 through week 12; the medication regimen started 3 weeks before the psychological intervention. The nortriptyline dose was titrated to the therapeutic blood level for depression (50–140 mg/ml) (12). F1 illustrates the assessment, treatment, and quit date schedules in both trials.
All measures used were administered in both trials. The National Institute of Mental Health Diagnostic Interview Schedule (DIS) (22), based on DSM-III-R criteria, was used to assess current depression and history of major depression. A computerized version of the DIS (23) was used in trial 2. The test-retest kappa values reported for the computerized version are similar to those reported for the original version. Baseline depression levels were assessed by the Beck Depression Inventory (24). The internal consistency of the 21-item Beck Depression Inventory, as measured by Cronbach’s coefficient alpha, was 0.87 in this group. The total mood disturbance score from the 65-item Profile of Mood States (25), with its six subscales (depression, tension, anger, confusion, fatigue, and vigor), was used to assess baseline negative affect. In this group, the internal consistency of each subscale was high (range=0.83–0.95). The State-Trait Anxiety Inventory (26), a 20-item instrument, was used to measure trait anxiety. The trait scale of the State-Trait Anxiety Inventory had a high internal consistency (Cronbach’s alpha=0.93) in this group. The Fagerstrom Tolerance Questionnaire (27), an eight-item measure assessing smoking behaviors, was used to measure nicotine dependence. Other information on subjects’ demographic characteristics and smoking history was obtained by means of a questionnaire developed for general use in the clinic.
Abstinence Status at End of Treatment
Participants were coded as abstinent if they reported no smoking during the last 7 days of the assessment. Abstinence was verified by expired-air carbon monoxide levels of 10 ppm or less in both trials; biochemical verification by urine cotinine levels of less than 60 ng/ml was conducted in trial 2.
The Inventory to Diagnose Depression (28), a self-report questionnaire, was used to identify major depressive episodes after treatment on the basis of DSM-III-R criteria. The sensitivity and specificity of the Inventory to Diagnose Depression were found to be 70.0% and 87.5%, respectively, among college students (29) and 90.0% and 92.0%, respectively, among psychiatric outpatients (30). Participants were asked to report symptoms at each follow-up assessment based on the worst depressive episode they had experienced since the previous assessment. The Inventory to Diagnose Depression also assessed psychosocial impairment from depression, including the length of the episode, effect on daily functioning, whether professional help was sought, use of medication, and hospitalization.
Multiple logistic regression was used to examine the predictors of major depressive episodes after treatment. The selection of variables included in the regression models was guided by the research questions and statistical significance. First, abstinence status and history of depression were included in the models to examine whether abstinence status or history of depression increased the likelihood of major depressive episodes after treatment of smoking cessation. Second, trial was included as a covariate to control for potential confounding effects due to differences in subject characteristics across trials. Third, given the purpose of exploring various predictors of major depressive episodes, other candidate variables were selected if they attained a p value of <0.10 in initial logistic regression analyses. The variables tested were the subject characteristics listed in t1. Treatment factors—psychological interventions (mood management versus health education) and medication (nortriptyline versus placebo, by using trial 2 data only)—were tested. After selecting the candidate variables, a manual backward stepwise elimination procedure based on model comparisons by means of likelihood ratio tests was performed. Variables that did not contribute significantly (p<0.05) to the model were eliminated one at a time, leaving abstinence status, history of depression, and covariate trial included in the model.
In addition to a main effects model, interaction effects were explored. Breslow-Day tests for all potential explanatory variables were conducted to test the homogeneity of odd ratios across each trial (31). The Breslow-Day tests indicated that the interactions of abstinence status by trial (χ2=4.08, df=1, p=0.04) and marital status by trial (χ2=4.15, df=1, p=0.04) were significant; therefore, these two interactions were further tested. The interaction effects of abstinence status by depression history and other potential interactions between each pair of variables in the main effects model were also examined. The interaction terms that contributed to the main effects model (p<0.15) were further tested with other interaction and main effects in the model (31).
Incidence of Major Depression After Treatment
Among the 304 participants, 14.1% (N=43; 95% confidence interval [CI]=10.2%–18.0%) reported a major depressive episode within 12 months after treatment for smoking cessation. The incidence rates across trials were similar (trial 1: N=16, 11.5%, versus trial 2: N=27, 16.4%) (χ2=1.46, df=1, p=0.23, N=304). About half (N=170, 55.9%) of the participants were abstinent from smoking at the end of treatment. Contrary to our hypothesis, there was no significant difference in the 12-month incidence rates of major depressive episodes among those who were abstinent at the end of treatment and those who were not (N=25 of 170, 14.7%, versus N=18 of 134, 13.4%) (χ2=0.10, df=1, p=0.75, N=304). The 12-month incidence rate of major depressive episodes was higher for participants who had a history of depression (23.7%, N=23 of 97; 95% CI=19.4%–28.0%) than for those who had no history of depression (9.7%, N=20 of 207; 95% CI=7.6%–11.8%) (χ2=10.74, df=1, p=0.001, N=304). No difference was found in the incidence rates of major depressive episodes among participants who had a history of a single depressive episode and those who had a history of recurrent episodes (N=4 of 18, 22.2%, versus N=19 of 79, 24.0%, respectively) (χ2=0.03, df=1, p=0.87, N=97). However, the results were limited because 81.4% (N=79) of those who were positive for a history of depression had recurrent depressive episodes.
Frequency of Depressive Symptoms and Related Impairment
Among the 43 individuals who met the criteria for major depressive episode on the basis of the Inventory to Diagnose Depression, five participants reported major depressive episodes at more than one assessment. For these participants, we used the first reported episode for the description of symptom frequency. Twenty-seven participants (62.8%) reported that their depressive episodes lasted for at least 1 month. All 43 participants reported depressed mood, and 27 (62.8%) reported loss of interest or pleasure. Twenty-one participants (48.8%) reported suicidal ideation for at least 2 weeks. Other depressive symptoms were frequently reported: poor concentration (N=40, 93.0%), insomnia or hypersomnia (N=37, 86.0%), fatigue (N=36, 83.7%), significant weight change (N=34, 79.1%), feelings of worthlessness (N=27, 62.8%), and psychomotor rigidity or agitation (N=24, 55.8%). The mean number of depressive symptoms reported was 6.72 (SD=1.32, range=5–9). Most participants (N=37, 86.0%) who experienced major depressive episodes reported a significant impact on their normal activities due to major depressive episodes. Eighteen participants (41.9%) reported seeking professional help for their posttreatment depression; 10 participants (23.3%) reported taking medication (both over-the-counter and prescribed), and one participant (2.3%) reported being hospitalized because of depression.
Predictors of Major Depressive Episodes After Treatment
Significant predictors of the main effects model were the Beck Depression Inventory score, history of depression, college education, and age at smoking initiation (t2). The logits of the continuous scaled predictors (baseline Beck Depression Inventory score and age at smoking initiation) were found to have a linear association with the risk of major depressive episodes after treatment. The log likelihood of the main effects model was 107.52 (χ2=31.78, df=7, p<0.001). The Hosmer and Lemeshow goodness-of-fit statistic (31) was 6.77 (df=8, p=0.56), indicating an acceptable fit for the data.
The interaction between abstinence status and history of depression was not significant (χ2=0.11, df=1, p=0.74). Three interaction terms were found to have potential significant contribution (p<0.15): depression history by marital status (χ2=3.79, df=1, p=0.05), abstinence status by trial (χ2=3.40, df=1, p=0.06), and marital status by trial (χ2=3.98, df=1, p=0.05). The marital status and trial interaction was later removed from the final logistic regression because of a lack of significant contribution (p<0.05) to the model in the presence of other predictors and interaction terms (χ2=3.15, df=1, p=0.08). t3 shows the final logistic regression model, containing both main and interaction effects. The log likelihood of the final model was 103.29 (χ2=40.31, df=9, p<0.001). The Hosmer and Lemeshow goodness-of-fit statistic (31) was 3.25 (df=8, p=0.92), indicating that the logistic model for major depressive episodes after treatment fit the data adequately.
Contrary to our first hypothesis, abstinence status at the end of treatment did not increase the risk of major depressive episodes. Unexpectedly, a similar rate of incidence of major depressive episodes was observed among those who were not abstinent. The cause of these depressive episodes was not clear. On the basis of previous reports (5, 21), it might be argued that reduction in smoking, as opposed to abstinence, precipitated major depressive episodes. However, reduction in cigarettes smoked at the end of treatment was not found to be associated with the incidence of major depressive episodes when tested instead of abstinence in logistic regression analyses. Psychological factors might precipitate major depressive episodes, such as the fear of relapse among participants who were abstinent (14) or the experience of failure among those who were motivated to quit but were unsuccessful. However, these hypotheses could not be empirically tested by using the current data because these data were not collected.
The risk of major depressive episodes after treatment associated with abstinence was significantly higher in the trial with nicotine gum (trial 1) than in the trial with nortriptyline and placebo (trial 2), independent of whether the participants were taking active medication or placebo. The reason for the interaction of abstinence status and trial was not clear. The participants differed significantly on several variables, including nicotine dependence (Fagerstrom Tolerance Questionnaire score), age first smoked, depressed symptoms (Beck Depression Inventory score), and mood measures (total mood disturbance and trait anxiety scores). However, the direction of the differences was not consistent and was not logically related to the interaction.
Consistent with our second hypothesis, the risk of experiencing major depressive episodes was two times higher among smokers with a history of depression. Preliminary results showed that marital status might serve to moderate the effects of a history of depression. The risk of major depressive episodes after treatment in those who had a history of depression was four times higher among participants who had no partners, but the risk was similar among those who had partners. This finding should be interpreted with caution since the estimate was based on exploratory analyses and a small group of participants (N=23) who had a partner and a history of depression.
The incidence of major depressive episodes reported in the first 4 months after treatment (N=25, 8.5%; 95% CI=6.8%–10.0%) was similar to the 3-month incidence of major depressive episodes (N=9, 7.1%; 95% CI=2.8%–11.6%) reported in a previous study that used interview data from abstinent smokers (15). (Calculations for the rate of incidence and confidence intervals were based on the total of 295 subjects who provided follow-up data during the first 4 months.) We observed a markedly high 12-month incidence of major depressive episodes after treatment compared to population estimates (1.6%–3.2%) of the 12-month incidence of major depressive disorders (32–34) and to an estimate of the first onset of major depressive episodes among young smokers (1.7%) (35). Without a comparison group of smokers who did not attempt to abstain from smoking, it is unclear whether the elevated incidence of major depressive episodes in our group was related to the process of quitting, participants’ self-selection bias in seeking intensive treatment, assessment issues, or other factors.
The severity of baseline depression increased the risk of major depressive episodes after treatment by 9% with each 1-point increase in Beck Depression Inventory score. With other factors controlled, participants who were college educated were three times more likely than those who were less educated to develop major depressive episodes. The earlier one started smoking, the higher the risk that one might experience major depressive episodes. In particular, those who started smoking before age 15 (N=148, 48.7% of the group) were twice as likely to report major depressive episodes than those who started smoking later (odds ratio=2.21, 95% CI=1.05–4.62; Wald χ2=4.41, df=1, p=0.04). (The odds ratio was estimated from the multiple logistic regression model by using the same predictors shown in t2). The presence of depression symptoms in adolescence predicted the duration of smoking in adulthood (36) and the presence of nicotine dependence (37). If some smokers rely mostly on smoking to cope with depression at an early age, they might be at a higher risk for severe depression when they attempt to quit smoking later in life. This was purely speculative, however, since there was no comparison group of participants who were not trying to quit smoking.
The use of cognitive behavioral mood management and nortriptyline for the treatment of smoking cessation was not related to the risk of major depressive episodes. The lack of association between the risks of major depressive episodes and the psychological treatment conditions might be explained by the intervention’s focus on teaching mood management skills related to smoking cessation rather than on preventing major depression itself. The observed lack of association between nortriptyline administration and the recurrence of major depressive episodes should be viewed as preliminary evidence. Treatment guidelines have recommended at least 4–9 months of antidepressant treatment for the prevention of the recurrence of major depressive episodes (38). However, because the primary focus of the original design was on the acute effects of antidepressant medication on smoking cessation, participants in the current study received only 11 weeks of nortriptyline treatment. The follow-up assessments of major depressive episodes took place when the subjects were no longer receiving medication. These factors might account for the lack of difference in the risk of major depressive episodes between the medication and placebo groups.
These findings should be interpreted taking the following limitations into account. First, the incidence estimate in this study was based on self-report questionnaire data. Second, this study lacked a comparison group of smokers who did not attempt to quit smoking. Third, no detailed data were available on the onset of major depressive episodes in relation to quit dates or to precipitating events such as psychological and situational stress. Furthermore, the results were confined to smokers who sought intensive treatment for smoking cessation and who tended to be more nicotine dependent.
Using prospective data, we found no relation between cigarette abstinence or reduction and depressive episodes after treatment for smoking cessation. Nor did we find an interaction of abstinence or reduction with history of depression. We conclude that smoking reduction or abstinence, whether with a history of depression or not, does not increase the risk of major depressive episodes. Participants with a history of depression were more likely to have a recurrence of major depressive episodes during the study period. Independent of depression history, elevated baseline depression level, smoking initiation at an early age, and higher education appear to increase the risk of major depressive episodes. Whether the rate was any different than it would have been for participants who did not enroll in a smoking cessation trial was unclear. Given the limitations previously discussed, we were unable to test the effect of smoking cessation on the development of major depression. Practitioners who continue to care for patients with a history of major depression must be careful of the possibility of recurring depression, no matter the context.
Presented in part at the 60th annual scientific meeting of the College on Problems of Drug Dependence, Scottsdale, AZ, June 16, 1998. Received Feb. 22, 1999; revisions received June 25 and Aug. 2, 1999; accepted Aug. 24, 1999From the Department of Psychiatry, University of California, San Francisco. Address correspondence to Dr. Tsoh, Department of Psychiatry (0984-TRC), University of California, 401 Parnassus Ave., San Francisco, CA 94143; email@example.com (e-mail). Supported by grants DA-02538, DA-23625, DA-09253, and DA-07250 from the National Institute on Drug Abuse. The authors thank Kevin Delucchi, Ph.D., and John Neuhaus, Ph.D., for statistical consultation and the members of the Treatment Research Center Writers’ Tasks Force at the University of California, San Francisco, for reviewing early versions of this article.
Treatment Schedules for 304 Participants Recruited From Two Smoking Cessation Trialsa
aBaseline assessment was conducted at week 0 and treatment was begun at week 1 in both trials. Quit dates were at week 2 of psychological treatment in trial 1 and at week 6 in trial 2. Assessment was conducted at the end of the psychological intervention. Follow-up assessments were conducted at weeks 12, 26, and 52 in trial 1 and at weeks 24, 38, and 64 in trial 2.