Laboratory of Epidemiology and Biometry, Division of Intramural Clinical and Biological Research, National Institute on Alcohol Abuse and Alcoholism, National Institutes of Health, Bethesda, MD, USA
* Author to whom correspondence should be addressed at: NIAAA/LEB Room 3083, 5635 Fishers Lane MSC 9304, Bethesda, MD 20892-9304, USA. Tel.: +1 301 435 2255; Fax: +1 301 443 1400; E-mail: ddawson{at}mail.nih.gov
(Received 14 March 2005; first review notified 2 May 2005; accepted in revised form 13 May 2005)
![]() |
ABSTRACT |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
INTRODUCTION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
A number of the more recent studies have examined stress as a continuous measure based on varying counts of negative life events. Two longitudinal studies that followed adolescents into adulthood found limited effects of stress on consumption. One found a positive relationship between negative events and consumption that was significant during high school years but diminished as individuals moved into late adolescence and adulthood (Aseltine and Gore, 2000), whereas the other found a positive effect only at the fourth follow-up (age 21) and only among men with relatively strong tension-reducing motivations for drinking (Rutledge and Sher, 2001
). In a regional sample of 6747 adult males, both the number of stressful life events and level of perceived stress increased significantly from abstainers to common drinkers to problem drinkers (Cole et al., 1990
). Similarly, a small (n = 154) sample of adults recruited from newspaper advertisement and alcohol treatment centers found a positive increase in the number of stressful life events from light drinkers to problem drinkers to alcoholics, among both men and women (King et al., 2003
). However, a 60-day diary study of the same individuals reported that consumption was higher on high-stress days than on low-stress days, but only among men with stronger positive expectancies or a greater sense of carelessness related to drinking (Armeli et al., 2000
). In an aggregate-level study that compared stressful life events with alcohol-related deaths and per capita consumption at the state level, 19 of the 20 correlations were positive, and most were significant, especially after adjusting for potential confounders (Linsky et al., 1985
).
Other studies have examined various domains of stressful events, the effects of which have often differed for men and women. In a study of more than 16 000 military personnel, work stress was positively associated with heavy alcohol in the past month among men, whereas the effect of family stress was not significant. Neither work nor family stress was a significant predictor of heavy drinking for women, although the estimated odds ratios were comparable in magnitude with those for men (Bray et al., 1999). A one-year follow-up study of 515 men and women who sought treatment for alcohol problems reported that acute life events at baseline (T1) were not associated with drinking at follow-up (T2) for either men or women, after adjusting for chronic stressors and functioning at T1. Among chronic sources of stress, problems with friends at T1 were positively associated with the volume of consumption and frequency of intoxication at T2 for women, but not for men (Skaff et al., 1999
). On the basis of data from four waves of the Health and Retirement Study, Perreira and Sloan (2001)
reported that an increase over time (Wave 1Wave 4) in usual daily consumption was positively associated with retirement and divorce. Widowhood was also positively associated with increased consumption, but only among men.
In a representative general population sample of 2802 Dutch adults 1574 years of age, heavy drinking was positively associated with any vs no negative life events, as well as specifically with crime victimization, decreased financial position, and divorce/breakup. However, all of these associations were significant only for men. In addition to acute stress, this study examined chronic stress and found a positive association between marital disruption and heavy drinking for both men and women, and a positive association between unemployment and heavy drinking for men only (San Jose et al., 2000). Hussong (2003)
found that gender differences in the relationship of alcohol involvement and various types of stress were modified by support seeking and coping mechanisms. Given a limited active coping style, men showed a stronger association than women between heavy drinking and social adjustment and school problems but a weaker association between heavy drinking and relationship stress. However, the effect of relationship stress was stronger for men among individuals with a limited support seeking style. In the 60-day diary study described previously, consumption was positively associated with negative non-work (i.e. social/interpersonal) events but negatively associated with both positive and negative health events, before and after adjusting for perceived level of stress (Carney et al., 2000
), with no gender differences reported.
Many of the studies already conducted of the association between stress and drinking are strong in terms of theoretical grounding and study design. However, apart from the Dutch study cited previously (San Jose et al., 2000), no study to date has examined the relationship of stress and alcohol consumption in a representative, general population sample of adults. Data from the 20012002 National Epidemiologic Survey on Alcohol and Related Conditions (NESARC) provide the opportunity to study this relationship in a US sample that is large enough to evaluate both gender differences and the moderating effects of economic and psychological vulnerability. In addition, the detailed consumption data contained in the NESARC permit comparison of the effects of life stress on various aspects of drinking pattern. This is an important advantage, as an impact of stress on risky patterns of drinking has far greater public health significance than an impact on non-risk drinking.
Accordingly, this study examined the effects of past-year acute stress, both on a continuous scale of negative life events and within the four dichotomous categories representing different types of stress, on volume of intake, drinking frequency (overall, heavy and moderate), and drinking quantity (usual and largest). It also examines the extent to which these associations are modified by gender, economic vulnerability, and psychological vulnerability. The following specific hypotheses were tested:
![]() |
METHODS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Measures
Stress. Respondents were asked whether they had experienced 12 different types of stressful life events in the 12 months preceding interview. Factor analysis indicated that these events were represented by four categories (data not shown):
Economic vulnerability. Economic vulnerability was represented by having a family income below the poverty line, using the official poverty level standards established by the Bureau of the Census. The poverty line varied as a function of the number and ages of related individuals in the household, from a low income of $8494 for a household containing a single adult 65 years or older to a high income of $39 413 for a household containing nine or more related members including only one child younger than 18 years of age.
Psychological vulnerability. Psychological vulnerability was defined as having a past-year mood or anxiety disorder: major depression, dysthymia, mania, hypomania, panic disorder, social phobia, specific phobia, or generalized anxiety. In order to be classified with a mood or anxiety disorder, respondents had to satisfy the requisite number of DSM-IV (American Psychiatric Association, 1994) criteria for at least one such disorder during the 12 months preceding interview, accompanied by social and/or occupational dysfunction. Major depression excluded cases owing to bereavement, and all mood and anxiety disorders excluded cases exclusively associated with physical illness, alcohol use, or drug use. The testretest reliability of the past-year mood and anxiety disorders was fair (Grant et al., 2003b
), and their validity was demonstrated by the fact that each disorder except for hypomania was associated with significant (P < 0.001) increases in disability and social and occupational dysfunctions (Grant et al., 2004
).
Alcohol consumption. All alcohol measures reflect past-year consumption, i.e. intake in the 12 months preceding the NESARC interview. The volume of intake was based on data summed across separate series of questions for coolers, beer, wine, and distilled spirits. For each beverage, the volume was estimated on the basis of overall frequency of drinking, usual and largest quantities consumed, frequency of consuming the largest quantity, frequency of consuming 5+ drinks, usual drink size, and ethanol content by volume of the brand usually consumed (NIAAA, 2004). All remaining consumption measures were based on a series of questions for all beverages combined. Overall frequency of drinking, usual quantity of drinks and largest quantity of drinks were asked directly. Frequency of heavy drinking reflects the frequency (days per year) of drinking 5+ drinks of any type for men and the frequency of drinking 4+ drinks of any type for women. Frequency of moderate drinking represents the difference between overall frequency of drinking and frequency of heavy drinking. Frequency responses were transformed into days per year using midpoints of response categories, e.g. 34 times a week = 182 days per year. The testretest reliability of the alcohol measures varied from good to excellent, from kappa = 0.69 for frequency of drinking 5+ drinks to kappa = 0.84 for the largest quantity of drinks consumed (Grant et al., 2003b
).
Analysis
All of the statistics used in this analysis were generated by SUDAAN (Research Triangle Institute, 2002), a software package that uses Taylor series linearization to adjust variance estimates for complex multistage sample designs. Multiple linear regression models were used to estimate the association between the number and the types of past-year stressors and various drinking measures, controlling for age, gender, race/ethnicity, marital status, education, poverty status, and the presence of past-year mood/anxiety disorders. These models also tested for significant interactions between stress and (i) gender, (ii) poverty, and (iii) mood/anxiety disorders.
Preliminary model fitting revealed that the best fit was obtained by applying a natural log transformation to all of the outcome measures of alcohol consumption. Thus, the beta coefficient for each predictor variable, when exponentiated, indicates the multiplicative (rather than additive) effect of a 1-unit increase in that variable on the consumption measure in question.
![]() |
RESULTS |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
Table 3 presents the multivariate models estimating the association between the number of past-year stressors and consumption. As can be seen from the top panel of the table, the main effect of number of stressors was significant and positive for all outcomes except frequency of moderate drinking, for which it was significant and negative. As would be expected, the main effect of male gender was consistently positive. The main effects of economic and psychological vulnerability varied according to outcome and were often non-significant. There was a significant interaction between male gender and stress level for all outcomes except overall frequency of drinking. For outcomes where the main effect of stress was positive, male gender intensified this positive effect; for the outcome where the main effect of stress was negative, male gender intensified the negative effect.
|
Table 4 presents the multivariate models that assessed associations between specific types of stress and different measures of alcohol consumption. In general, the effects of specific types of stress mirrored the effect of the number of stressors, having the greatest impact on the frequency of heavy drinking and the volume of intake, followed by the largest quantity of drinks consumed, and a negative effect on the frequency of moderate drinking. However, there were exceptions to this rule, and effect modifiers varied according to the type of stress and outcome:
|
![]() |
DISCUSSION |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Job-related and legal sources of stress were more strongly associated with alcohol consumption than were social and health-related stress. They also were less commonly experienced by US drinkers and arguably represent more serious problems (e.g. job loss, arrest, major financial problems) than most of the items subsumed under social and health-related stressors. Thus, the stronger impact of job and legal stress would be consistent with the tension reduction hypothesis, in that they could be expected to result in greater tensionrequiring a greater increase in drinking for tension alleviationthan other types of stress. Consistent with their being less often reported than other types of stress, individuals who experienced legal and job-related stress reported a larger number of stressors than those who did not experience these types of stress (data not shown). To rule out the possibility that the differential impact of job and legal stress was no more than an indirect effect of overall stress level, the models that examined individual types of stressors were rerun to include number of stressors. Even with overall stress level controlled, job and legal stress continued to be more strongly associated with consumption than health or social stress (data not shown).
Although the associations between overall stress level and consumption generally were of greater magnitude for men than women, they were almost always statistically significant for both. The same was generally true for individual types of stress, although the effects of legal and job-related stress were male limited for some outcomes. Taken as a whole, these findings are not out of line with prior studies, in which smaller sample sizes often may have made it impossible to establish the statistical significance of the reduced effects observed among female drinkers, and they are consistent with our hypothesis that stronger stigmatization of heavy drinking among women would reduce their drinking response to stress. However, there was no evidence to support the hypothesis that women might compensate by increasing their frequency of drinking to a greater extent than men in response to stress.
With respect to the hypothesis that individuals with economic or psychological vulnerability would increase their drinking more in response to stress than less vulnerable individuals, this study's results were mixed. There was some evidence that poverty intensified the effects of job stress, which makes good sense in that the level of tension provoked by job loss or unemployment might well be greater for persons with fewer financial resources. Psychological vulnerability did not modify any of the associations between stress and consumption. It is possible, though, that any potential mediating effect of psychological vulnerability might have been masked by a tendency on the part of individuals with mood or anxiety disorders to have a lower threshold for stress and therefore to report more stressful life events. Also, both the main effects of poverty and psychological vulnerability, as well as their interactions with stressful life events, might have been stronger had continuous rather than dichotomous measures of these aspects of vulnerability been used.
As with any cross-sectional survey data, the NESARC data cannot establish causality in the association between stress and alcohol consumption. Although this study treated stress as the independent variable and alcohol consumption as the outcome, it is possible that the associations found in this analysis reflect an increase in negative life events as a result of drinking. In order to explore this possibility, we examined responses to two questions on problems caused by drinkingproblems with family or friends and job- or school-related problems. (These questions were asked as part of a larger series of indicators used for classification of alcohol use disorders.) Of all the drinkers who reported past-year social problems, only 2.4% endorsed having had problems with family or friends as a result of their drinking. Similarly, only 2.3% of those who reported past-year job-related problems claimed to have had trouble at school or on the job because of their drinking (data not shown). Although the attribution of problems to drinking increased with level of consumption, it remained rare enough to suggest that the findings of this study do not largely reflect causes rather than effects of stress. However, lack of attribution to alcohol does not completely rule out the possibility that the problems did in fact result from excessive drinking. Data from the second wave of the NESARC, currently in the field, will permit reevaluation of the relationship between stress and drinking using a prospective study design. Still, the results of this study suggest that treatment and brief interventions aimed at problem drinkersand indeed primary care visits for all drinkers reporting high levels of stressmight benefit from addressing the issue of tension alleviation and the building of alternative coping mechanisms.
![]() |
REFERENCES |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Armeli, S., Carney, M. A., Tennen, H. et al. (2000) Stress and alcohol use: A daily process examination of the stressor-vulnerability model. Journal of Personality and Social Psychology 78, 979994.[CrossRef][ISI][Medline]
Aseltine, R. H. and Gore, S. L. (2000) Variable effects of stress on alcohol use from adolescence to early adulthood. Substance Use and Misuse 35, 643668.
Bray, R. M., Fairbank, J. A. and Marsden, M. E. (1999) Stress and substance use among military women and men. American Journal of Drug and Alcohol Abuse 25, 239256.[CrossRef][ISI][Medline]
Carney, M. A., Armeli, S., Tennen, H. et al. (2000) Positive and negative daily events, perceived stress, and alcohol use: diary study. Journal of Consulting and Clinical Psychology 68, 788798.[CrossRef][ISI][Medline]
Cherpitel, C. J., Tam, T., Midanik, L. et al. (1995) Alcohol and non-fatal injury in the U.S. general population: a risk function analysis. Accident Analysis and Prevention 27, 651661.[CrossRef][ISI]
Cole, G., Tucker, L. and Griedman, G. M. (1990) Relationships among measures of alcohol drinking, behavior, life-events and perceived stress. Psychological Reports 67, 587591.[ISI][Medline]
Conger, J. J. (1956) Alcoholism: Theory, problem and challenge. II. Reinforcement theory and the dynamics of alcoholism. Quarterly Journal of Studies on Alcohol 17, 296305.[ISI][Medline]
Dawson, D. A. (2001) Alcohol and mortality from external causes. Journal of Studies on Alcohol 62, 790797.[ISI][Medline]
Dawson, D. A., Grant, B. F. and Li, T. K. (2005) Quantifying the risks associated with exceeding recommended drinking limits. Alcoholism, Clinical and Experimental Research 29, 902908.[CrossRef][ISI][Medline]
Grant, B. F., Moore, T. C. and Kaplan, K. (2003a) Source and accuracy statement: wave 1 National Epidemiologic Survey on Alcohol and Related Conditions (NESARC). National Institute on Alcohol Abuse and Alcoholism, Bethesda MD.
Grant, B. F., Dawson, D. A., Stinson, F. S. et al. (2003b) The Alcohol Use Disorder and Associated Disabilities Schedule (AUDADIS): reliability of alcohol and drug modules in a general population sample. Drug and Alcohol Dependence 39, 716.
Grant, B. F., Stinson, F. S., Hasin, D. S. et al. (2004) Prevalence and comorbidity of substance use disorders and independent mood and anxiety disorders: results from the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC). Archives of General Psychiatry 61, 807816.
Greeley, J. and Oei, T. (1999) Alcohol and tension reduction. In: Psychological Theories of Drinking and Alcoholism: Second Edition, Leonard, K. E. and Blane, H. T., eds., pp. 1453, Guilford Press, New York.
Hussong, A. M. (2003) Further refining the stress-coping model of alcohol involvement. Addictive Behaviors: An International Journal 28, 15151522.[CrossRef]
King, A. C., Bernardy, N. C. and Hauner, K. (2003) Stressful events, personality, and mood disturbance: gender differences in alcoholics and problem drinkers. Addictive Behaviors 28, 171187.[CrossRef][ISI][Medline]
Linsky, A. S., Straus, M. A. and Colby, J.P. Jr.(1985) Stressful events, stressful conditions and alcohol problems in the United States: a partial test of Bales's theory. Journal of Studies on Alcohol 46, 7280.[ISI][Medline]
National Institute on Alcohol Abuse and Alcoholism (2004) Data Notes on Alcohol Consumption Variables (Section 2A). Available at http://niaaa.census.gov/pdfs/datanotes.pdf
Perreira, K. M. and Sloan, F. A. (2001) Life events and alcohol consumption among mature adults: a longitudinal analysis. Journal of Studies on Alcohol 62, 501508.[ISI][Medline]
Ramstedt, M. (2002) Alcohol consumption and the experience of adverse consequences: a comparison of six European countries. Contemporary Drug Problems 29, 549575.
Research Triangle Institute (2002) SUDAAN User's Manual, Release 8.0. Research Triangle Institute, Research Triangle Park, NC.
Rutledge, P. C. and Sher, K. J. (2001) Heavy drinking from freshman year into early young adulthood: the roles of stress, tension-reduction drinking motives, gender and personality. Journal of Studies on Alcohol 62, 457466.[ISI][Medline]
San Jose, B., van Oers, H. A. M., van de Mheen, H. D. et al. (2000) Stressors and alcohol consumption. Alcohol and Alcoholism 35, 307312.
Skaff, M. M., Finney, J. W. and Moos, R. H. (1999) Gender differences in problem drinking and depression: different vulnerabilities? American Journal of Community Psychology 27, 2554.[CrossRef][ISI][Medline]
|