- Email: [email protected]
Abstinence effects as predictors of 28-day relapse in smokers
Addictive Behaviors. Vol. 21, No. 4, pp. 481--1YO.lYY6 Copyright 0 lYY6 Elsevier Science Ltd Printed in the USA. All rights reserved 0306-4603/Y6 $15.00 + .M)
Pergamon
SSDI 03~4603(95)00070-4
ABSTINENCE EFFECTS AS PREDICTORS RELAPSE IN SMOKERS
OF 2%DAY
GARY E. SWAN, MARCIA M. WARD, and LISA M. JACK Health
Sciences
and Policy Program,
SRI International,
Menlo Park, CA, USA
Abstract -The present analysis sought to determine the relationship between abstinence effects in 64 ex-smokers (mean age = 41.1 years) and the rate at which they relapsed over 4 weeks of biochemically confirmed follow-up. This analysis focused on six abstinence effects that play a central role in the DSM-III-R and DSM IV definitions of withdrawal from nicotine: anger, depression, craving, appetite, confusion, and tension. Significant increases were observed for all six symptoms following cessation, and, with the exception of craving, substantial intercorrelations among the abstinence effects were noted. Cox proportional hazards survival models identified increases in anger, depressed mood, and craving to be significantly associated with a shorter time to relapse (all p < .03). Stepwise Cox proportional hazards survival analysis identified increases in depressed mood and craving as the most significant combination of abstinence effects in relation to time to relapse. A more stringent test of the potency of the relationship between these abstinence effects and time to relapse was conducted in which two other risk factors in this sample, method of quitting and education level, were also included in the model testing sequence. Even after adjustment for these significant risk factors, the increase in craving remained a significant predictor of a higher rate of relapse. This result suggests a robustness to this particular abstinence effect as a determinant of the speed with which ex-smokers relapse over a l-month interval after cessation.
INTRODUCTION
Although several retrospective studies indicate that many ex-smokers cite withdrawal discomfort as a reason for relapse, surprisingly few prospective studies show a relationship between the effects of abstinence from nicotine following smoking cessation and the subsequent ability to remain an ex-smoker (Hughes. 1992; Hughes. Higgins, & Hatsukami, 1990). Of the many effects that have been examined. only depressed mood and craving have demonstrated at least some consistency across studies (Cummings, Jaen, & Giovino. 1985; Gritz, Carr, & Marcus, 1991; Killen, Fortmann, Newman, & Varady, 1991; U.S. Department of Health and Human Services [USDHHS], 1988; West, Hajek, & Belcher, 1989; Zelman, Brandon, Jorenby. & Baker, 1992) although the robustness of these results has recently been questioned (Hughes, 1992). Among the reasons for diminished consistency may be that no underlying model of withdrawal has been fully developed and tested that takes into account biochemical, physiological. and behavioral changes that are known to occur with smoking cessation. along with possible differences in the time course with which these changes occur. Thus, it may be that certain abstinence effects, such as depressed mood, have more immediate impact on the propensity to relapse, whereas others result from biochemical changes and environmental events that occur more slowly and exert their influence on relapse later in the maintenance of cessation phase. Requests for reprints should be sent to Gary E. Swan, PhD. Health Sciences and Policy Program, SRI International. 333 Ravenswood Ave.. Menlo Park, CA 94025. This work was supported by funds provided by the Cigarette and Tobacco Surtax Fund of the State of California through the Tobacco-Related Disease Research Program of the University of California. grant 1RT547. 451
482
G. E. SWAN
et al.
Another difficulty encountered by studies of abstinence effects as predictors of relapse is failure to take into account differences in subpopulations of smokers (e.g., self vs. program quitters, more educated vs. less educated, etc.) that may themselves be related to the likelihood of relapse (USDHHS, 1988). In the present investigation, we sought to determine the predictive association between abstinence effects (anger, depressed mood, craving, appetite, confusion, and tension) and the subsequent rate of relapse in 64 subjects who were followed for 4 weeks after smoking cessation, both before and after adjustment for subject characteristics also known to be related to the rate of relapse in this sample of smokers. METHODS
Subjects
Subjects for the study expressed interest in participating by either filling out a signup sheet after a presentation was given to their smoking cessation class or by responding to a flyer that was posted at various locations. Interested individuals were contacted by telephone and screened before their expected quit date. Subjects were excluded from participation if they were younger than 21 or older than 65 years or were planning to use nicotine gum or patch. Because this was a study of the effects of smoking cessation on withdrawal symptoms including cardiovascular responses (blood pressure and heart rate), subjects smoking fewer than 10 cigarettes per day were also excluded to keep response heterogeneity to a minimum due to small nicotine dose effects. A similar strategy to exclusion has been used in our previously published work (Swan, Ward, Carmelli, & Jack, 1993; Swan, Ward, Jack, & Javitz, 1993; Ward, Swan, Jack, & Javitz, 1994). One hundred and five subjects were considered eligible to participate and completed at least one assessment. Of this number, 18 subjects decided to discontinue participation before the planned quit date. Five subjects had reduced their smoking levels before cessation to below 10 cigarettes per day and were therefore ineligible to continue. One subject was dropped from analysis because the self-report of abstinence was disconfirmed by a high level of saliva cotinine. For the present study, 17 more subjects were excluded from analysis for the following reasons: 9 subjects had returned to smoking within 24 hours of quitting, 7 subjects had missing withdrawal or outcome data, and one subject did not have a valid quit. We compared the 41 people who were excluded for any reason to those who were entered into the study (n = 64) with respect to age, education, income, years smoked, cigarettes smoked per day, precessation cotinine, and Fagerstrom dependence (Fagerstrom, 1978). No significant differences were observed. Sample characteristics are shown in Table 1. The final sample consisted of 40 (63%) females and 24 (37%) males. Tests of means or proportions for males and females with respect to the variables presented in Table 1 revealed women to have a significantly lower level of precessation cotinine, 215.6 ng/ml vs 311.1 ng/ml, t (37) = -3.14, p < .Ol, and to be less likely to be self quitters, 38% vs 67%, x2 (1) = 5.11, p < .03. No other differences were observed. Approximately half of the subjects (n = 33) were participants in an organized smoking cessation program consisting of ten classes with cessation occurring at the fifth class. The class schedule was condensed into a two week period with two maintenance classes scheduled for weeks three and four. The content of the classes did not differ substantially from those offered by the American Lung Association and included
Abstinence effects as predictors
483
Table 1. Sample characteristics (n = 64) Characteristic Age (years) Education (l-8) Years smoked Precessationpercotinine Cigarettes day (@ml) Fagerstrom score Precessation symptoms Anger Depressed mood Craving Appetite Confusion Tension Abstinence effectsb Anger Depressed mood Craving Appetite Confusion Tension
Mean
Standard deviation
Range
41.1 4.9 22.7
9.5 1.4 10.5
22-63 3-8 546
251.4 24.7 6.1
116.8 11.1 1.5
:!ZZ0 39
4.9 4.4 o.oa 2.0 5.2 a.7
6.4 6.9
O-29 O-30
0.5 4.2 6.5
O-3 fL15 O-23
3 g*+* 2:4* g.4*+** 0.4*** 2.2*** 3 g****
7.6 7.4 2.5 0.9 4.6 6.5
- 18.0-30.7 - 17.7-28.7 2.312.0 -2.9-2.3 -8.7-17.3 - 12.3-26.0
aBecause subjects were smoking freely at this point, craving was assumed to be zero. bAbstinence effects are defined as the difference between the baseline symptom rating and the average of symptom ratings following the cessation of smoking in those who remained abstinent. *p < .05. ***p < ,001.
****p < .OOOl.
identification of factors related to smoking, development of coping skills to deal with urges to smoke, and learning to substitute healthy behaviors for smoking. The other half of the sample (n = 31) quit on their own. No significant mean differences between self-quitters and class quitters on measures of dependence including Fagerstrom score, cigarettes smoked per day, or precessation cotinine were observed. The average age of the subjects at the time of the study was 41.1 ? 9.5 years (range = 22 to 63 years). Using an 8-point scale with 1 = less than eighth grade and 8 = advanced postgraduate degree, education averaged 4.9 + 1.4, with 78% of the subjects reporting an education level greater than high school. Subjects reported smoking an average of 24.7 +- 11.1 cigarettes per day in the month before quitting and had been smoking for an average of 22.7 -C 10.5 years. The average level of saliva cotinine before cessation was 251.4 _’ 116.8 ng/ml. Procedure
The study consisted of six assessments over the course of approximately 5 to 6 weeks. The first assessment was conducted, on average, 7.8 1?15.9 days before the expected quit date, while the subjects were still smoking 91% of their normal levels. The second assessment was conducted on the quit date, before the actual cessation of smoking. The third, fourth, fifth, and sixth assessments were conducted one per week for the four weeks after the quit date (averaging 7.7 -+ 1.1 days, 14.6 ? 1.2 days, 21.8 t 1.9 days, and 29.8 f 3.6 days, respectively).
484
G. E. SWAN et al.
At each of the six assessments, the subjects’ expired air was measured for the presence of carbon monoxide using a Vitalograph EC50 Carbon Monoxide Monitor. Levels of less than 10 ppm were considered to be nonsmoking. At each assessment subjects also provided saliva samples according to published guidelines (Butts, Kuehneman, & Widdowson, 1974). The airtight containers were frozen and later shipped to the Division of Physiologic Hygiene, University of Minnesota, where cotinine levels were determined by gas chromatography. All samples obtained at the precessation assessment were analyzed in order to determine precessation levels of cotinine, a useful indicator of exposure to nicotine (Fagerstrom, 1991). For subjects reporting continued abstinence throughout the course of the study, samples from the final assessment were analyzed to determine whether cotinine levels would confirm the report. Levels of saliva cotinine of less than 27 ng/ml were considered to be nonsmoking (Cummings & Richards, 1988). Background information collected before cessation included questionnaire measures of household annual income (rated on a 9-point scale ranging from 1 = less than $10,000 to 9 = greater than $lOO,OOO),education (rated on an &point scale ranging from 1 = less than 8th grade to 8 = advanced postgraduate degree), marital status, smoking and quitting history, and the Fagerstrom Tolerance Questionnaire (FagerStrom, 1978). Assessment of withdrawal symptoms focused on the symptoms listed in the DSMIII-R (American Psychiatric Association, 1987) and DSM-IV (American Psychiatric Association, 1994) definitions of withdrawal from nicotine: anger, depression, craving, appetite, confusion, and tension. These symptoms were assessed as follows: anger-hostility was defined as the sum of 12 items from the Profile of Mood States (POMS; McNair, Lorr, & Droppleman, 1981); depression-dejection was defined as the sum of 15 items from the POMS; confusion-bewilderment was defined as the sum of 7 items from the POMS; and tension-anxiety was defined as the sum of 9 items from the POMS. The other two symptoms were assessed from the Shiffman-Jarvik Withdrawal Symptom Scale (Shiffman & Jarvik, 1976; Shiffman, 1979): appetite was defined as the sum of 2 items, and craving was defined as the sum of 6 items. At each of the six assessments, the subjects filled out these questionnaires reporting how they felt in the previous 24 hours. To enable assessment of withdrawal symptoms in the early stages of cessation, the subjects were also given three sets of questionnaires at the second assessment to take home and fill out at the end of the first, second, and third days after stopping smoking. For the purposes of this analysis, an abstinence effect was defined as the difference between the precessation symptom rating and the average of the postcessation symptom ratings. For relapsers, symptoms reported after relapse were excluded. The subjects kept daily logs of any occurrence of smoking between the quit date and the final assessment (approximately 4 weeks). For each day after quitting smoking, the subjects recorded the number of cigarettes they smoked, or entered 0 if they remained abstinent. Daily logs were brought to each assessment, and a master log for that subject was updated by the interviewer. All subjects who smoked after quitting were asked to fill out a brief questionnaire regarding the exact date, time, and situation in which they had their first cigarette. Relapse measure
A time to relapse variable was created for each subject in the study. For those subjects who smoked during the course of the study, time to relapse was calculated as the days between quitting smoking and the first occurrence of smoking. Subjects who re-
Abstinence
485
effects as predictors
ported no smoking throughout the study and who had carbon monoxide and cotinine levels confirming that they were continuously abstinent were considered to be censored and were given a time to relapse value of 28 days. As mentioned previously, one subject had a disconfirming cotinine level at the third assessment and was subsequently dropped from the analysis. Of those individuals returning to smoking, median time to relapse was 5 days. Statistical methods
The analysis plan incorporated three important characteristics of this data set. First, because time to relapse was available, we were able to use Cox proportional hazards survival analysis (Statistics and Epidemiology Research Corporation, 1993) with time to relapse (measured in days) as the dependent variable. Cox proportional hazards survival analyses were run for each abstinence effect, as well as for method of quitting and education, the only baseline characteristics that were related to the rate of relapse. Second, because abstinence effects were intercorrelated (see Table 2) we used a multivariate, stepwise approach to identify the best combination of abstinence effects related to time to relapse. Using a forward stepping process, each abstinence effect was tested for its significance when entered in the model, the abstinence effect with the greatest statistical significance being entered first, followed by the abstinence effect with the next greatest statistical significance, and so on, until no more effects were found to be statistically significant below the required p = .05 level. Third, because method of quitting and education were also related to time to relapse in these data, we included these variables in the final stepwise prediction model. For all analyses involving Cox proportional hazards survival analysis, we report the hazard ratio as the index of the strength of the association between independent variables and time to relapse. The hazard ratio is defined as the multiplicative change in the hazard rate (the instantaneous rate of relapse at time t, conditional upon maintenance of abstinence to time t, that occurs when the predictor variable changes by one unit. RESULTS
Intercorrelations mood ratings
between sociodemographic,
nicotine dependence,
and baseline
The relationship between cotinine and the Fagerstrom dependence score and baseline characteristics including age, education, cigarettes smoked per day, years smoked, baseline anger, depressed mood, appetite, confusion, and tension were examined with Pearson correlation analysis. Results indicate that, as expected, the Fagerstrom total
Table 2. Intercorrelations Effect Anger Depressed Craving Appetite Confusion Tension
*p < .05. **p < .Ol
Anger
Depressed
.63** .lO .Ol .33** .54**
.06 -.16 .59** .45**
among Craving
.lO .07 .17
abstinence Appetite
-.30* -.12
effects Confusion
.57**
Tension
-
486
G. E. SWAN
et al.
score was correlated significantly with number of cigarettes smoked per day, r(59) = .46, p < .OOl, but not with any of the other demographic, smoking history, baseline mood ratings, or subsequent abstinent effects. The baseline cotinine measure was marginally associated with number of cigarettes smoked per day, r(63) = .21, p < .lO, and not with any of the other measured variables. Abstinence effects and their intercorrelations
As shown in Table 1, all measured symptoms showed significant increases from zero after smoking cessation. Table 2 reveals a pattern of intercorrelations among the abstinence effects, with increases in anger being substantially correlated with increases in depressed mood and tension and moderately associated with increases in confusion. In addition to this association with anger, increases in depressed mood were also associated with increases in confusion and tension. Increases in confusion were negatively associated with increases in appetite, in addition to previously noted associations, and were substantially related to increases in tension. Interestingly, increases in craving were unrelated to any of the other abstinence effects measured in this study. Association of time to relapse with sociodemographics, baseline mood ratings
nicotine dependence,
and
None of the baseline mood scores were associated significantly with time to relapse. No significant effects for gender, age, years smoked, cigarettes per day, precessation cotinine, or Fagerstrom score on the time to relapse were observed in this sample either. Those who quit on their own relapsed sooner than did those who quit as a result of participation in a smoking cessation program (hazard ratio = .124, 95% CI = .036.428, p < .OOl). Those with lower education relapsed sooner (hazard ratio = .643,95% CI = .44S-.924,p < .02). Associations with time to relapse before adjustment for method of quitting and education
As shown in Table 3, larger increases in anger, depressed mood, and craving were significantly associated with a shorter time to relapse whereas changes in appetite appeared unrelated to time to relapse in these analyses. Changes in confusion and tension were marginally related to the rate of relapse. Stepwise Cox proportional hazards analyses identified depressed mood and craving as the best combination of predictors of relapse.
Table 3. Survival models of abstinence effects and time to relapse adjustment for method of quitting and education Hazard
Effect Univariate models Anger Depressed mood Craving Appetite Confusion Tension Stepwise multivariate Depressed Craving
ratio
95% CI
before
p-value
1.067 1.069 1.307 0.829 1.082 1.067
1.00~1.130 1.012-1.128 1.044-1.636 0.507-l .3s5 0.99G1.182 0.9961.143
.026 ,016 ,019 .454 ,083 .066
1.073 1.329
1.019-1.131 1.057-1.671
.oos
model ,015
Abstinence
481
effects as predictors
Associations with time to relapse afier adjustment for method of quitting and education
After adjustment for method of quitting and level of education, the significance of the association between each abstinence effect and time to relapse was tested individually. As shown in Table 4, only craving remained as a significant predictor of time to relapse. The role of depressed mood as a predictor was diminished to marginal significance and did not enter the final model. In order to determine the confounding effect, if any, of indicators of nicotine dependence on the prediction of time to relapse, we conducted several additional Cox regression analyses. The results indicate (a) cotinine and Fagerstrom dependence were not associated significantly with time to relapse either individually or jointly; and, (b) when cotinine and Fagerstrom dependence were included in the model (without quit method or education), the stepwise procedure indicated that both craving and depressive mood abstinence effects were significantly associated with rate of relapse. However, consistent with our primary finding reported above, when education and quit method are included along with cotinine and the Fagerstrom dependence score in the model, only craving emerges as significant. DISCUSSION
This analysis identified several abstinence effects to be associated with the rate of relapse over a 2%day interval under different assumptions about the data. Under the assumption that the observed abstinence effects are not intercorrelated, we found that increases in three out of the six abstinence effects tested were associated with a shorter time to relapse. After taking into account their intercorrelation, we found that, consistent with previous research (Hughes, 1992) depressed mood and craving were the most effective combination of effects in predicting time to relapse. After taking into account the effects of two additional risk factors, method of quitting and education, we found that the role of depressed mood was reduced to marginal significance whereas that for craving remained significant. These results suggest a robustness to the role of craving as a predictor of relapse that is essentially unaffected by adjustment for other abstinence effects, the method of quitting, and level of education. The underlying causal model by which abstinence effects could influence the rate at which ex-smokers relapse is not well understood. The available models of withdrawal vary in the extent to which the abstinence effects could plausibly influence relapse over the short (within 1 week), intermediate (within 2 weeks), or long term (greater
Table 4. Survival models of abstinence effects and time to relapse adjustment for method of quitting and education Effect Univariate models Anger Depressed mood Craving Appetite Confusion Tension Stepwise multivariate Method of quitting Education Craving
Hazard
ratio
95% CI
after
p-value
1.038 1.054 I .286 0.883 I .086 1.053
.985-l .094 .999-1.111 1.037-1.594 .483-1.613 .995-1.185 .986-1.123
,159 ,022 .685 .066 .123
0.162 0.630 1.286
.046.567 .428-,926 1.037-l .594
.004 .019 .022
.05S
model
488
G. E. SWAN et al.
than 2 weeks). For example, Hughes et al. (1990) cite several models that might explain a high rate of relapse. These include sympathetic overactivity (Glassman, Jackson, Walsh, & Roose, 198.5) decreased cholinergic tone (McCarty, 1982) and hypoglycemia (Bourne, 1985). Other models, which incorporate homeostatic features, seem to suggest processes that might influence intermediate rates of relapse, including receptor induction, enzyme derepression, and neuronal redundancy (Martin, 1977). Finally, behavioral theories of relapse suggest processes that could influence relapse over longer intervals, such as conditioned withdrawal (Abrams, Monti, Pinto, Elder, & Brown, 1987; Niaura, Abrams, Demuth, Monti, & Pinto, 1989), overgeneralization (Solomon & Corbit, 1973) and a developmental coping model of smoking in which smokers gradually lose their ability to cope with negative affect, hunger, and lack of concentration, the effects of which remain unresolved and even heightened with cessation (Ashton & Stepney, 1982; Mangan & Golding, 1984). At this point in our understanding of withdrawal symptoms as a predictor of relapse, it is fair to say that no single model predominates. In fact, as Hughes et al. (1990) point out, many of the models remain completely or partially untested in humans. The results of the present study might give support to several models, including the positive reinforcement model, the homeostatic model, and the dopamine mediated reward/pleasure model of smoking (Wise & Bozarth, 1987). In particular, the homeostatic theory of withdrawal states that drug exposure triggers a negative feedback process that opposes the drug effect (Martin, 1977). With repeated exposure, this opposing effect becomes greater, thus decreasing the effect of the drug. Cessation of the drug leaves the opposing effect unopposed, and this is withdrawal. Homeostatic theories require withdrawal effects to be opposite those of the agonist effects of nicotine. This theory would apply to anger, concentration, and appetite. The reason dysphoria (e.g., depressed mood) increases on cessation is not clear (Hughes et al., 1990) although recent evidence that the consumption of nicotine can stimulate the production of dopaminergic neurons in the ventral tegmental area of the mesolimbic/mesocortical system suggests that the loss of reward and pleasure with cessation could certainly lead to a dysphoric or depressed mood (Balfour, 1991, 1993). Similarly, memory for the positive reinforcement (mediated through this neural pathway) associated with smoking has also been used as one definition of craving (Wise & Bozarth, 1987). In the present analysis, anger, depressed mood, and craving were associated individually with a higher rate of relapse over a l-month interval. These results are consistent with previous research identifying a role for these abstinence effects in relation to relapse over varying intervals (Cummings et al., 1985; Gritz et al., 1991; Hughes et al., 1990; Killen et al., 1991; USDHHS, 1988; West et al., 1989; Zelman et al., 1992). As in previous studies (Killen et al., 1991), most abstinence effects (with the exception of craving) are significantly intercorrelated and, thus, most likely represent some redundancy in the quantification of the construct “withdrawal discomfort.” In the stepwise Cox proportional hazards analysis before adjustment for other risk factors, both craving and depressed mood remained in the model as the best combination of abstinence effects in predicting the rate of relapse. The fact that both abstinence effects are implicated in theories involving the dopaminergic system may be significant. Our inclusion of quit method and education in the final Cox proportional hazards model should be viewed as a conservative approach to the study of withdrawal symptoms and relapse. The approach is similar to that taken in epidemiologic investigations in which the predictive power of a risk factor is determined after controlling for other
Abstinence
effects as predictors
489
known risk factors. Even with this conservative approach, however, craving remained as a significant predictor of the rate of relapse over a 2%day interval, a finding pointing to the robustness of the association and one that is in basic agreement with recent results reported by Killen et al. (1991; and Killen, Fortmann, Kraemer, Varady, & Newman, 1992). Because these analyses were conducted in a relatively small sample of smokers, the power to detect a significant association between depressed mood and the rate of relapse may have been limited. In order to assess this possibility, analyses were conducted on a sample that was twice the size (n = 128) of the present sample by creating a data set in which each observation counted twice. With this larger sample, we observe that, along with quit method and education, the final stepwise model included both craving and depressed mood (p = .04). Using the p-value for depressed mood from this simulation as a guide, we conclude that a sample size of approximately llO130 subjects would have been sufficient to detect the effect for depressed mood on time to relapse at p < .05 in the final stepwise multivariate model. In the present data set, significant change from baseline was observed for all mood ratings, suggesting the validity and robustness of the DSM III-R and DSM-N criteria as applied to these ex-smokers. As noted previously, with the exception of craving, all abstinence effects were significantly intercorrelated. Craving, on the other hand, appeared as essentially orthogonal to the other abstinence effects. This finding is consistent with the observation noted by Hughes et al. (1990) from other samples of smokers, and suggests at least a two-dimensional structure to abstinence effects, with craving standing as independent of the other symptoms.
REFERENCES Abrams, D., Monti, P., Pinto, R., Elder, J., & Brown, P. (1987) Psychosocial stress and coping in smokers who relapse or quit. Health Psychology, 6,289-303. American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed.). Washington, DC: Author. American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author. Ashton, H., & Stepney, R. (1982). Smoking: Psychology andpharmacology. New York:Tavistock. Balfour, D. J. K. (1991). The neurochemical mechanisms underlying nicotine tolerance and dependence. In The biological bases of drug tolerance and dependence (pp. 121-157). London: Academic Press Balfour, D. J. K. (1993). The pharmacology of nicotine in the CNS and its bearing on nicotine replacement therapies Infernational Journal of Smoking Cessation, 2,3-8. Bourne, S. (198S).Treatment of cigarette smoking with short-term high-dosage corticotrophin therapy: Preliminary communication. Journal of the Royal Society of Medicine, 78.649-450. Butts, W. C., Kuehneman, M.. & Widdowson, G. M. (1974). Automated method for determining serum thiocyanate to distinguish smokers from nonsmokers. Clinical Chemistry, 2&1344-1348. Cummings, K. M., Jaen, C. R., & Giovino, G. (1985). Circumstances surrounding relapse in a group of recent exsmokers Preventi\se Medicine, 14,195-202. Cummings, S. R.. & Richard. R. J. (1988). Optimum cutoff points for biochemical validation of smoking status. American Journal of Public Health, 78,574575. Fagerstrom. K. 0. (1978). Measuring degree of physical dependence to tobacco smoking with reference to individualization of treatment. Addictive Behaviors, 3.2X-241. Fagerstrom. K. 0. (1991).Towards better diagnoses and more individual treatment of tobacco dependence. British Journal of Addiction, 86.543-547. Glassman, A. H., Jackson. W. K.. Walsh, B. T.. & Roose, S. P (1985). Cigarette craving, smoking withdrawal, and clonidine. Science, 226.864866. Gritz, E. R.. Carr. C. R.. & Marcus, A. C. (1991). The tobacco withdrawal svndrome in unaided ouitters British Journal of Addiction, 86.57-69. Hughes, J. R.. Higgins, S. T., & Hatsukami, D. (1990). Effects of abstinence from tobacco: A critical review. In L. T. Kozlowski, H. M. Annis. H. D. Cappell, F. B. Glaser, M. S. Goodstadt, et al. (Eds.), Research advances in alcohol and drug problems (Vol. 10, pp.317-398). New York: Plenum.
490
G. E. SWAN et al.
Hughes, J. R. (1992).The nicotine withdrawal syndrome: A brief review and update. Internarional Journal of Smoking Cessation, 1,21-26. Killen, J. D., Fortmann, S. P., Newman, B., & Varady,A. (1991). Prospective study of factors influencing the development of craving associated with smoking cessation. Psychopharmacology, 105,191-196. Killen, J. D., Fortmann, S. P., Kraemer, H. C., Varady, A., & Newman, B. (1992). Who will relapse? Symptoms of nicotine dependence predict long-term relapse after smoking cessation. Journal of Consulting and Clinical Psychology, 60,797-801.
Mangan, G. L., & Gelding, J. F. (1984). The psychopharmacology of smoking.New York: Cambridge University Press. Martin, W. (Ed.). (1977). Drug addiction I and II: Handbook of experimental pharmacology (Vol. 45). New York: Springer-Verlag. McCarty, M. E (1982). Nutritional support of central catecholaminergic tone may aid smoking withdrawal. Medical Hypotheses, 8,95-102.
McNair, D. M., Lorr, M., & Droppelman, L. F. (1981). Manual for the profile of mood states. San Diego, CA: Educational and Industrial Testing Service. Niaura, R., Abrams, D., Demuth, B., Monti, F!,& Pinto, R. (1989). Responses to smoking-related stimuli and early relapse to smoking. Addicfive Behaviors, 14,419428. Shiffman, S. (1979). The tobacco withdrawal syndrome. In N. A. Krasnegor (Ed.), Cigarette smoking as a dependence process (pp. 158184). NIDA Research Monograph #23. Washington, DC: U.S. Government Printing Office. Shiffman, S. M., & Jarvik, M. E. (1976). Smoking withdrawal symptoms in two weeks of abstincence. Psychopharmacology,
50,35-39.
Solomon, R. L., & Corbit, J. D. (1973). An opponent-process Journal of Abnormnl
theory of motivation: Cigarette addiction.
Psychology, 81,158171.
Statistics and Epidemiology Research Corporation. (1993). EGRET statistical software. Seattle, WA: Author. Swan, G. E., Ward, M. M., Jack, L. M. & Javitz, H. (1993). Cardiovascular reactivity as a predictor of relapse in male and female smokers Health Psychology, 12,451458. Swan, G. E., Ward, M. M., Carmelli, D., & Jack, L. M. (1993). Differential rates of relapse in subgroups of male and female ex-smokers Journal of Clinical Epidemiology, 46,1041-1053. U.S. Department of Health and Human Services (1988). The he&h consequences of smoking: Nicotine addiction. A report of the surgeon general. Washington, DC: U.S. Government Printing Office. Ward, M. M., Swan, G. E., Jack, L. M., & Javitz, H. S. (1994). Effect of smoking cessation and relapse on cardiovascular levels and reactivity. Psychopharmacology, 114,147-154. West, R. J., Hajek, P., & Belcher, M. (1989). Severity of withdrawal symptoms as a predictor of outcome of an attempt to quit smoking. Psychological Medicine, 19.981-985. Wise, R. A., & Bozarth, M. A. (1987). A psychomotor stimulant theory of addiction. Psychological Review, 94,469492.
Zelman, D. C., Brandon,T. H., Jorenby, D. E., & Baker,T. B. (1992). Measures of affect and nicotine dependence predict differential response to smoking cessation treatments Journal of Consulting and Clinical Psychology, 60,943-952.
Pergamon
SSDI 03~4603(95)00070-4
ABSTINENCE EFFECTS AS PREDICTORS RELAPSE IN SMOKERS
OF 2%DAY
GARY E. SWAN, MARCIA M. WARD, and LISA M. JACK Health
Sciences
and Policy Program,
SRI International,
Menlo Park, CA, USA
Abstract -The present analysis sought to determine the relationship between abstinence effects in 64 ex-smokers (mean age = 41.1 years) and the rate at which they relapsed over 4 weeks of biochemically confirmed follow-up. This analysis focused on six abstinence effects that play a central role in the DSM-III-R and DSM IV definitions of withdrawal from nicotine: anger, depression, craving, appetite, confusion, and tension. Significant increases were observed for all six symptoms following cessation, and, with the exception of craving, substantial intercorrelations among the abstinence effects were noted. Cox proportional hazards survival models identified increases in anger, depressed mood, and craving to be significantly associated with a shorter time to relapse (all p < .03). Stepwise Cox proportional hazards survival analysis identified increases in depressed mood and craving as the most significant combination of abstinence effects in relation to time to relapse. A more stringent test of the potency of the relationship between these abstinence effects and time to relapse was conducted in which two other risk factors in this sample, method of quitting and education level, were also included in the model testing sequence. Even after adjustment for these significant risk factors, the increase in craving remained a significant predictor of a higher rate of relapse. This result suggests a robustness to this particular abstinence effect as a determinant of the speed with which ex-smokers relapse over a l-month interval after cessation.
INTRODUCTION
Although several retrospective studies indicate that many ex-smokers cite withdrawal discomfort as a reason for relapse, surprisingly few prospective studies show a relationship between the effects of abstinence from nicotine following smoking cessation and the subsequent ability to remain an ex-smoker (Hughes. 1992; Hughes. Higgins, & Hatsukami, 1990). Of the many effects that have been examined. only depressed mood and craving have demonstrated at least some consistency across studies (Cummings, Jaen, & Giovino. 1985; Gritz, Carr, & Marcus, 1991; Killen, Fortmann, Newman, & Varady, 1991; U.S. Department of Health and Human Services [USDHHS], 1988; West, Hajek, & Belcher, 1989; Zelman, Brandon, Jorenby. & Baker, 1992) although the robustness of these results has recently been questioned (Hughes, 1992). Among the reasons for diminished consistency may be that no underlying model of withdrawal has been fully developed and tested that takes into account biochemical, physiological. and behavioral changes that are known to occur with smoking cessation. along with possible differences in the time course with which these changes occur. Thus, it may be that certain abstinence effects, such as depressed mood, have more immediate impact on the propensity to relapse, whereas others result from biochemical changes and environmental events that occur more slowly and exert their influence on relapse later in the maintenance of cessation phase. Requests for reprints should be sent to Gary E. Swan, PhD. Health Sciences and Policy Program, SRI International. 333 Ravenswood Ave.. Menlo Park, CA 94025. This work was supported by funds provided by the Cigarette and Tobacco Surtax Fund of the State of California through the Tobacco-Related Disease Research Program of the University of California. grant 1RT547. 451
482
G. E. SWAN
et al.
Another difficulty encountered by studies of abstinence effects as predictors of relapse is failure to take into account differences in subpopulations of smokers (e.g., self vs. program quitters, more educated vs. less educated, etc.) that may themselves be related to the likelihood of relapse (USDHHS, 1988). In the present investigation, we sought to determine the predictive association between abstinence effects (anger, depressed mood, craving, appetite, confusion, and tension) and the subsequent rate of relapse in 64 subjects who were followed for 4 weeks after smoking cessation, both before and after adjustment for subject characteristics also known to be related to the rate of relapse in this sample of smokers. METHODS
Subjects
Subjects for the study expressed interest in participating by either filling out a signup sheet after a presentation was given to their smoking cessation class or by responding to a flyer that was posted at various locations. Interested individuals were contacted by telephone and screened before their expected quit date. Subjects were excluded from participation if they were younger than 21 or older than 65 years or were planning to use nicotine gum or patch. Because this was a study of the effects of smoking cessation on withdrawal symptoms including cardiovascular responses (blood pressure and heart rate), subjects smoking fewer than 10 cigarettes per day were also excluded to keep response heterogeneity to a minimum due to small nicotine dose effects. A similar strategy to exclusion has been used in our previously published work (Swan, Ward, Carmelli, & Jack, 1993; Swan, Ward, Jack, & Javitz, 1993; Ward, Swan, Jack, & Javitz, 1994). One hundred and five subjects were considered eligible to participate and completed at least one assessment. Of this number, 18 subjects decided to discontinue participation before the planned quit date. Five subjects had reduced their smoking levels before cessation to below 10 cigarettes per day and were therefore ineligible to continue. One subject was dropped from analysis because the self-report of abstinence was disconfirmed by a high level of saliva cotinine. For the present study, 17 more subjects were excluded from analysis for the following reasons: 9 subjects had returned to smoking within 24 hours of quitting, 7 subjects had missing withdrawal or outcome data, and one subject did not have a valid quit. We compared the 41 people who were excluded for any reason to those who were entered into the study (n = 64) with respect to age, education, income, years smoked, cigarettes smoked per day, precessation cotinine, and Fagerstrom dependence (Fagerstrom, 1978). No significant differences were observed. Sample characteristics are shown in Table 1. The final sample consisted of 40 (63%) females and 24 (37%) males. Tests of means or proportions for males and females with respect to the variables presented in Table 1 revealed women to have a significantly lower level of precessation cotinine, 215.6 ng/ml vs 311.1 ng/ml, t (37) = -3.14, p < .Ol, and to be less likely to be self quitters, 38% vs 67%, x2 (1) = 5.11, p < .03. No other differences were observed. Approximately half of the subjects (n = 33) were participants in an organized smoking cessation program consisting of ten classes with cessation occurring at the fifth class. The class schedule was condensed into a two week period with two maintenance classes scheduled for weeks three and four. The content of the classes did not differ substantially from those offered by the American Lung Association and included
Abstinence effects as predictors
483
Table 1. Sample characteristics (n = 64) Characteristic Age (years) Education (l-8) Years smoked Precessationpercotinine Cigarettes day (@ml) Fagerstrom score Precessation symptoms Anger Depressed mood Craving Appetite Confusion Tension Abstinence effectsb Anger Depressed mood Craving Appetite Confusion Tension
Mean
Standard deviation
Range
41.1 4.9 22.7
9.5 1.4 10.5
22-63 3-8 546
251.4 24.7 6.1
116.8 11.1 1.5
:!ZZ0 39
4.9 4.4 o.oa 2.0 5.2 a.7
6.4 6.9
O-29 O-30
0.5 4.2 6.5
O-3 fL15 O-23
3 g*+* 2:4* g.4*+** 0.4*** 2.2*** 3 g****
7.6 7.4 2.5 0.9 4.6 6.5
- 18.0-30.7 - 17.7-28.7 2.312.0 -2.9-2.3 -8.7-17.3 - 12.3-26.0
aBecause subjects were smoking freely at this point, craving was assumed to be zero. bAbstinence effects are defined as the difference between the baseline symptom rating and the average of symptom ratings following the cessation of smoking in those who remained abstinent. *p < .05. ***p < ,001.
****p < .OOOl.
identification of factors related to smoking, development of coping skills to deal with urges to smoke, and learning to substitute healthy behaviors for smoking. The other half of the sample (n = 31) quit on their own. No significant mean differences between self-quitters and class quitters on measures of dependence including Fagerstrom score, cigarettes smoked per day, or precessation cotinine were observed. The average age of the subjects at the time of the study was 41.1 ? 9.5 years (range = 22 to 63 years). Using an 8-point scale with 1 = less than eighth grade and 8 = advanced postgraduate degree, education averaged 4.9 + 1.4, with 78% of the subjects reporting an education level greater than high school. Subjects reported smoking an average of 24.7 +- 11.1 cigarettes per day in the month before quitting and had been smoking for an average of 22.7 -C 10.5 years. The average level of saliva cotinine before cessation was 251.4 _’ 116.8 ng/ml. Procedure
The study consisted of six assessments over the course of approximately 5 to 6 weeks. The first assessment was conducted, on average, 7.8 1?15.9 days before the expected quit date, while the subjects were still smoking 91% of their normal levels. The second assessment was conducted on the quit date, before the actual cessation of smoking. The third, fourth, fifth, and sixth assessments were conducted one per week for the four weeks after the quit date (averaging 7.7 -+ 1.1 days, 14.6 ? 1.2 days, 21.8 t 1.9 days, and 29.8 f 3.6 days, respectively).
484
G. E. SWAN et al.
At each of the six assessments, the subjects’ expired air was measured for the presence of carbon monoxide using a Vitalograph EC50 Carbon Monoxide Monitor. Levels of less than 10 ppm were considered to be nonsmoking. At each assessment subjects also provided saliva samples according to published guidelines (Butts, Kuehneman, & Widdowson, 1974). The airtight containers were frozen and later shipped to the Division of Physiologic Hygiene, University of Minnesota, where cotinine levels were determined by gas chromatography. All samples obtained at the precessation assessment were analyzed in order to determine precessation levels of cotinine, a useful indicator of exposure to nicotine (Fagerstrom, 1991). For subjects reporting continued abstinence throughout the course of the study, samples from the final assessment were analyzed to determine whether cotinine levels would confirm the report. Levels of saliva cotinine of less than 27 ng/ml were considered to be nonsmoking (Cummings & Richards, 1988). Background information collected before cessation included questionnaire measures of household annual income (rated on a 9-point scale ranging from 1 = less than $10,000 to 9 = greater than $lOO,OOO),education (rated on an &point scale ranging from 1 = less than 8th grade to 8 = advanced postgraduate degree), marital status, smoking and quitting history, and the Fagerstrom Tolerance Questionnaire (FagerStrom, 1978). Assessment of withdrawal symptoms focused on the symptoms listed in the DSMIII-R (American Psychiatric Association, 1987) and DSM-IV (American Psychiatric Association, 1994) definitions of withdrawal from nicotine: anger, depression, craving, appetite, confusion, and tension. These symptoms were assessed as follows: anger-hostility was defined as the sum of 12 items from the Profile of Mood States (POMS; McNair, Lorr, & Droppleman, 1981); depression-dejection was defined as the sum of 15 items from the POMS; confusion-bewilderment was defined as the sum of 7 items from the POMS; and tension-anxiety was defined as the sum of 9 items from the POMS. The other two symptoms were assessed from the Shiffman-Jarvik Withdrawal Symptom Scale (Shiffman & Jarvik, 1976; Shiffman, 1979): appetite was defined as the sum of 2 items, and craving was defined as the sum of 6 items. At each of the six assessments, the subjects filled out these questionnaires reporting how they felt in the previous 24 hours. To enable assessment of withdrawal symptoms in the early stages of cessation, the subjects were also given three sets of questionnaires at the second assessment to take home and fill out at the end of the first, second, and third days after stopping smoking. For the purposes of this analysis, an abstinence effect was defined as the difference between the precessation symptom rating and the average of the postcessation symptom ratings. For relapsers, symptoms reported after relapse were excluded. The subjects kept daily logs of any occurrence of smoking between the quit date and the final assessment (approximately 4 weeks). For each day after quitting smoking, the subjects recorded the number of cigarettes they smoked, or entered 0 if they remained abstinent. Daily logs were brought to each assessment, and a master log for that subject was updated by the interviewer. All subjects who smoked after quitting were asked to fill out a brief questionnaire regarding the exact date, time, and situation in which they had their first cigarette. Relapse measure
A time to relapse variable was created for each subject in the study. For those subjects who smoked during the course of the study, time to relapse was calculated as the days between quitting smoking and the first occurrence of smoking. Subjects who re-
Abstinence
485
effects as predictors
ported no smoking throughout the study and who had carbon monoxide and cotinine levels confirming that they were continuously abstinent were considered to be censored and were given a time to relapse value of 28 days. As mentioned previously, one subject had a disconfirming cotinine level at the third assessment and was subsequently dropped from the analysis. Of those individuals returning to smoking, median time to relapse was 5 days. Statistical methods
The analysis plan incorporated three important characteristics of this data set. First, because time to relapse was available, we were able to use Cox proportional hazards survival analysis (Statistics and Epidemiology Research Corporation, 1993) with time to relapse (measured in days) as the dependent variable. Cox proportional hazards survival analyses were run for each abstinence effect, as well as for method of quitting and education, the only baseline characteristics that were related to the rate of relapse. Second, because abstinence effects were intercorrelated (see Table 2) we used a multivariate, stepwise approach to identify the best combination of abstinence effects related to time to relapse. Using a forward stepping process, each abstinence effect was tested for its significance when entered in the model, the abstinence effect with the greatest statistical significance being entered first, followed by the abstinence effect with the next greatest statistical significance, and so on, until no more effects were found to be statistically significant below the required p = .05 level. Third, because method of quitting and education were also related to time to relapse in these data, we included these variables in the final stepwise prediction model. For all analyses involving Cox proportional hazards survival analysis, we report the hazard ratio as the index of the strength of the association between independent variables and time to relapse. The hazard ratio is defined as the multiplicative change in the hazard rate (the instantaneous rate of relapse at time t, conditional upon maintenance of abstinence to time t, that occurs when the predictor variable changes by one unit. RESULTS
Intercorrelations mood ratings
between sociodemographic,
nicotine dependence,
and baseline
The relationship between cotinine and the Fagerstrom dependence score and baseline characteristics including age, education, cigarettes smoked per day, years smoked, baseline anger, depressed mood, appetite, confusion, and tension were examined with Pearson correlation analysis. Results indicate that, as expected, the Fagerstrom total
Table 2. Intercorrelations Effect Anger Depressed Craving Appetite Confusion Tension
*p < .05. **p < .Ol
Anger
Depressed
.63** .lO .Ol .33** .54**
.06 -.16 .59** .45**
among Craving
.lO .07 .17
abstinence Appetite
-.30* -.12
effects Confusion
.57**
Tension
-
486
G. E. SWAN
et al.
score was correlated significantly with number of cigarettes smoked per day, r(59) = .46, p < .OOl, but not with any of the other demographic, smoking history, baseline mood ratings, or subsequent abstinent effects. The baseline cotinine measure was marginally associated with number of cigarettes smoked per day, r(63) = .21, p < .lO, and not with any of the other measured variables. Abstinence effects and their intercorrelations
As shown in Table 1, all measured symptoms showed significant increases from zero after smoking cessation. Table 2 reveals a pattern of intercorrelations among the abstinence effects, with increases in anger being substantially correlated with increases in depressed mood and tension and moderately associated with increases in confusion. In addition to this association with anger, increases in depressed mood were also associated with increases in confusion and tension. Increases in confusion were negatively associated with increases in appetite, in addition to previously noted associations, and were substantially related to increases in tension. Interestingly, increases in craving were unrelated to any of the other abstinence effects measured in this study. Association of time to relapse with sociodemographics, baseline mood ratings
nicotine dependence,
and
None of the baseline mood scores were associated significantly with time to relapse. No significant effects for gender, age, years smoked, cigarettes per day, precessation cotinine, or Fagerstrom score on the time to relapse were observed in this sample either. Those who quit on their own relapsed sooner than did those who quit as a result of participation in a smoking cessation program (hazard ratio = .124, 95% CI = .036.428, p < .OOl). Those with lower education relapsed sooner (hazard ratio = .643,95% CI = .44S-.924,p < .02). Associations with time to relapse before adjustment for method of quitting and education
As shown in Table 3, larger increases in anger, depressed mood, and craving were significantly associated with a shorter time to relapse whereas changes in appetite appeared unrelated to time to relapse in these analyses. Changes in confusion and tension were marginally related to the rate of relapse. Stepwise Cox proportional hazards analyses identified depressed mood and craving as the best combination of predictors of relapse.
Table 3. Survival models of abstinence effects and time to relapse adjustment for method of quitting and education Hazard
Effect Univariate models Anger Depressed mood Craving Appetite Confusion Tension Stepwise multivariate Depressed Craving
ratio
95% CI
before
p-value
1.067 1.069 1.307 0.829 1.082 1.067
1.00~1.130 1.012-1.128 1.044-1.636 0.507-l .3s5 0.99G1.182 0.9961.143
.026 ,016 ,019 .454 ,083 .066
1.073 1.329
1.019-1.131 1.057-1.671
.oos
model ,015
Abstinence
481
effects as predictors
Associations with time to relapse afier adjustment for method of quitting and education
After adjustment for method of quitting and level of education, the significance of the association between each abstinence effect and time to relapse was tested individually. As shown in Table 4, only craving remained as a significant predictor of time to relapse. The role of depressed mood as a predictor was diminished to marginal significance and did not enter the final model. In order to determine the confounding effect, if any, of indicators of nicotine dependence on the prediction of time to relapse, we conducted several additional Cox regression analyses. The results indicate (a) cotinine and Fagerstrom dependence were not associated significantly with time to relapse either individually or jointly; and, (b) when cotinine and Fagerstrom dependence were included in the model (without quit method or education), the stepwise procedure indicated that both craving and depressive mood abstinence effects were significantly associated with rate of relapse. However, consistent with our primary finding reported above, when education and quit method are included along with cotinine and the Fagerstrom dependence score in the model, only craving emerges as significant. DISCUSSION
This analysis identified several abstinence effects to be associated with the rate of relapse over a 2%day interval under different assumptions about the data. Under the assumption that the observed abstinence effects are not intercorrelated, we found that increases in three out of the six abstinence effects tested were associated with a shorter time to relapse. After taking into account their intercorrelation, we found that, consistent with previous research (Hughes, 1992) depressed mood and craving were the most effective combination of effects in predicting time to relapse. After taking into account the effects of two additional risk factors, method of quitting and education, we found that the role of depressed mood was reduced to marginal significance whereas that for craving remained significant. These results suggest a robustness to the role of craving as a predictor of relapse that is essentially unaffected by adjustment for other abstinence effects, the method of quitting, and level of education. The underlying causal model by which abstinence effects could influence the rate at which ex-smokers relapse is not well understood. The available models of withdrawal vary in the extent to which the abstinence effects could plausibly influence relapse over the short (within 1 week), intermediate (within 2 weeks), or long term (greater
Table 4. Survival models of abstinence effects and time to relapse adjustment for method of quitting and education Effect Univariate models Anger Depressed mood Craving Appetite Confusion Tension Stepwise multivariate Method of quitting Education Craving
Hazard
ratio
95% CI
after
p-value
1.038 1.054 I .286 0.883 I .086 1.053
.985-l .094 .999-1.111 1.037-1.594 .483-1.613 .995-1.185 .986-1.123
,159 ,022 .685 .066 .123
0.162 0.630 1.286
.046.567 .428-,926 1.037-l .594
.004 .019 .022
.05S
model
488
G. E. SWAN et al.
than 2 weeks). For example, Hughes et al. (1990) cite several models that might explain a high rate of relapse. These include sympathetic overactivity (Glassman, Jackson, Walsh, & Roose, 198.5) decreased cholinergic tone (McCarty, 1982) and hypoglycemia (Bourne, 1985). Other models, which incorporate homeostatic features, seem to suggest processes that might influence intermediate rates of relapse, including receptor induction, enzyme derepression, and neuronal redundancy (Martin, 1977). Finally, behavioral theories of relapse suggest processes that could influence relapse over longer intervals, such as conditioned withdrawal (Abrams, Monti, Pinto, Elder, & Brown, 1987; Niaura, Abrams, Demuth, Monti, & Pinto, 1989), overgeneralization (Solomon & Corbit, 1973) and a developmental coping model of smoking in which smokers gradually lose their ability to cope with negative affect, hunger, and lack of concentration, the effects of which remain unresolved and even heightened with cessation (Ashton & Stepney, 1982; Mangan & Golding, 1984). At this point in our understanding of withdrawal symptoms as a predictor of relapse, it is fair to say that no single model predominates. In fact, as Hughes et al. (1990) point out, many of the models remain completely or partially untested in humans. The results of the present study might give support to several models, including the positive reinforcement model, the homeostatic model, and the dopamine mediated reward/pleasure model of smoking (Wise & Bozarth, 1987). In particular, the homeostatic theory of withdrawal states that drug exposure triggers a negative feedback process that opposes the drug effect (Martin, 1977). With repeated exposure, this opposing effect becomes greater, thus decreasing the effect of the drug. Cessation of the drug leaves the opposing effect unopposed, and this is withdrawal. Homeostatic theories require withdrawal effects to be opposite those of the agonist effects of nicotine. This theory would apply to anger, concentration, and appetite. The reason dysphoria (e.g., depressed mood) increases on cessation is not clear (Hughes et al., 1990) although recent evidence that the consumption of nicotine can stimulate the production of dopaminergic neurons in the ventral tegmental area of the mesolimbic/mesocortical system suggests that the loss of reward and pleasure with cessation could certainly lead to a dysphoric or depressed mood (Balfour, 1991, 1993). Similarly, memory for the positive reinforcement (mediated through this neural pathway) associated with smoking has also been used as one definition of craving (Wise & Bozarth, 1987). In the present analysis, anger, depressed mood, and craving were associated individually with a higher rate of relapse over a l-month interval. These results are consistent with previous research identifying a role for these abstinence effects in relation to relapse over varying intervals (Cummings et al., 1985; Gritz et al., 1991; Hughes et al., 1990; Killen et al., 1991; USDHHS, 1988; West et al., 1989; Zelman et al., 1992). As in previous studies (Killen et al., 1991), most abstinence effects (with the exception of craving) are significantly intercorrelated and, thus, most likely represent some redundancy in the quantification of the construct “withdrawal discomfort.” In the stepwise Cox proportional hazards analysis before adjustment for other risk factors, both craving and depressed mood remained in the model as the best combination of abstinence effects in predicting the rate of relapse. The fact that both abstinence effects are implicated in theories involving the dopaminergic system may be significant. Our inclusion of quit method and education in the final Cox proportional hazards model should be viewed as a conservative approach to the study of withdrawal symptoms and relapse. The approach is similar to that taken in epidemiologic investigations in which the predictive power of a risk factor is determined after controlling for other
Abstinence
effects as predictors
489
known risk factors. Even with this conservative approach, however, craving remained as a significant predictor of the rate of relapse over a 2%day interval, a finding pointing to the robustness of the association and one that is in basic agreement with recent results reported by Killen et al. (1991; and Killen, Fortmann, Kraemer, Varady, & Newman, 1992). Because these analyses were conducted in a relatively small sample of smokers, the power to detect a significant association between depressed mood and the rate of relapse may have been limited. In order to assess this possibility, analyses were conducted on a sample that was twice the size (n = 128) of the present sample by creating a data set in which each observation counted twice. With this larger sample, we observe that, along with quit method and education, the final stepwise model included both craving and depressed mood (p = .04). Using the p-value for depressed mood from this simulation as a guide, we conclude that a sample size of approximately llO130 subjects would have been sufficient to detect the effect for depressed mood on time to relapse at p < .05 in the final stepwise multivariate model. In the present data set, significant change from baseline was observed for all mood ratings, suggesting the validity and robustness of the DSM III-R and DSM-N criteria as applied to these ex-smokers. As noted previously, with the exception of craving, all abstinence effects were significantly intercorrelated. Craving, on the other hand, appeared as essentially orthogonal to the other abstinence effects. This finding is consistent with the observation noted by Hughes et al. (1990) from other samples of smokers, and suggests at least a two-dimensional structure to abstinence effects, with craving standing as independent of the other symptoms.
REFERENCES Abrams, D., Monti, P., Pinto, R., Elder, J., & Brown, P. (1987) Psychosocial stress and coping in smokers who relapse or quit. Health Psychology, 6,289-303. American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed.). Washington, DC: Author. American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author. Ashton, H., & Stepney, R. (1982). Smoking: Psychology andpharmacology. New York:Tavistock. Balfour, D. J. K. (1991). The neurochemical mechanisms underlying nicotine tolerance and dependence. In The biological bases of drug tolerance and dependence (pp. 121-157). London: Academic Press Balfour, D. J. K. (1993). The pharmacology of nicotine in the CNS and its bearing on nicotine replacement therapies Infernational Journal of Smoking Cessation, 2,3-8. Bourne, S. (198S).Treatment of cigarette smoking with short-term high-dosage corticotrophin therapy: Preliminary communication. Journal of the Royal Society of Medicine, 78.649-450. Butts, W. C., Kuehneman, M.. & Widdowson, G. M. (1974). Automated method for determining serum thiocyanate to distinguish smokers from nonsmokers. Clinical Chemistry, 2&1344-1348. Cummings, K. M., Jaen, C. R., & Giovino, G. (1985). Circumstances surrounding relapse in a group of recent exsmokers Preventi\se Medicine, 14,195-202. Cummings, S. R.. & Richard. R. J. (1988). Optimum cutoff points for biochemical validation of smoking status. American Journal of Public Health, 78,574575. Fagerstrom. K. 0. (1978). Measuring degree of physical dependence to tobacco smoking with reference to individualization of treatment. Addictive Behaviors, 3.2X-241. Fagerstrom. K. 0. (1991).Towards better diagnoses and more individual treatment of tobacco dependence. British Journal of Addiction, 86.543-547. Glassman, A. H., Jackson. W. K.. Walsh, B. T.. & Roose, S. P (1985). Cigarette craving, smoking withdrawal, and clonidine. Science, 226.864866. Gritz, E. R.. Carr. C. R.. & Marcus, A. C. (1991). The tobacco withdrawal svndrome in unaided ouitters British Journal of Addiction, 86.57-69. Hughes, J. R.. Higgins, S. T., & Hatsukami, D. (1990). Effects of abstinence from tobacco: A critical review. In L. T. Kozlowski, H. M. Annis. H. D. Cappell, F. B. Glaser, M. S. Goodstadt, et al. (Eds.), Research advances in alcohol and drug problems (Vol. 10, pp.317-398). New York: Plenum.
490
G. E. SWAN et al.
Hughes, J. R. (1992).The nicotine withdrawal syndrome: A brief review and update. Internarional Journal of Smoking Cessation, 1,21-26. Killen, J. D., Fortmann, S. P., Newman, B., & Varady,A. (1991). Prospective study of factors influencing the development of craving associated with smoking cessation. Psychopharmacology, 105,191-196. Killen, J. D., Fortmann, S. P., Kraemer, H. C., Varady, A., & Newman, B. (1992). Who will relapse? Symptoms of nicotine dependence predict long-term relapse after smoking cessation. Journal of Consulting and Clinical Psychology, 60,797-801.
Mangan, G. L., & Gelding, J. F. (1984). The psychopharmacology of smoking.New York: Cambridge University Press. Martin, W. (Ed.). (1977). Drug addiction I and II: Handbook of experimental pharmacology (Vol. 45). New York: Springer-Verlag. McCarty, M. E (1982). Nutritional support of central catecholaminergic tone may aid smoking withdrawal. Medical Hypotheses, 8,95-102.
McNair, D. M., Lorr, M., & Droppelman, L. F. (1981). Manual for the profile of mood states. San Diego, CA: Educational and Industrial Testing Service. Niaura, R., Abrams, D., Demuth, B., Monti, F!,& Pinto, R. (1989). Responses to smoking-related stimuli and early relapse to smoking. Addicfive Behaviors, 14,419428. Shiffman, S. (1979). The tobacco withdrawal syndrome. In N. A. Krasnegor (Ed.), Cigarette smoking as a dependence process (pp. 158184). NIDA Research Monograph #23. Washington, DC: U.S. Government Printing Office. Shiffman, S. M., & Jarvik, M. E. (1976). Smoking withdrawal symptoms in two weeks of abstincence. Psychopharmacology,
50,35-39.
Solomon, R. L., & Corbit, J. D. (1973). An opponent-process Journal of Abnormnl
theory of motivation: Cigarette addiction.
Psychology, 81,158171.
Statistics and Epidemiology Research Corporation. (1993). EGRET statistical software. Seattle, WA: Author. Swan, G. E., Ward, M. M., Jack, L. M. & Javitz, H. (1993). Cardiovascular reactivity as a predictor of relapse in male and female smokers Health Psychology, 12,451458. Swan, G. E., Ward, M. M., Carmelli, D., & Jack, L. M. (1993). Differential rates of relapse in subgroups of male and female ex-smokers Journal of Clinical Epidemiology, 46,1041-1053. U.S. Department of Health and Human Services (1988). The he&h consequences of smoking: Nicotine addiction. A report of the surgeon general. Washington, DC: U.S. Government Printing Office. Ward, M. M., Swan, G. E., Jack, L. M., & Javitz, H. S. (1994). Effect of smoking cessation and relapse on cardiovascular levels and reactivity. Psychopharmacology, 114,147-154. West, R. J., Hajek, P., & Belcher, M. (1989). Severity of withdrawal symptoms as a predictor of outcome of an attempt to quit smoking. Psychological Medicine, 19.981-985. Wise, R. A., & Bozarth, M. A. (1987). A psychomotor stimulant theory of addiction. Psychological Review, 94,469492.
Zelman, D. C., Brandon,T. H., Jorenby, D. E., & Baker,T. B. (1992). Measures of affect and nicotine dependence predict differential response to smoking cessation treatments Journal of Consulting and Clinical Psychology, 60,943-952.