Background. Many studies have demonstrated co-morbidity of alcohol abuse}dependence with mood and anxiety disorders but relatively little is known about anxiety and depression across the full continua of alcohol consumption and problems associated with drinking.

Methods. Participants from a general population sample (N ¯ 2725) aged 18–80 years completed the Alcohol Use Disorders Identification Test (AUDIT) and four measures of negative affect (two depression and two anxiety symptom scales) included in a self-completion questionnaire.

Results. High consumption, AUDIT total score, and AUDIT problems score were associated with high negative affect scores in participants under 60 years old (ORs in the range 1±80–2±83). Graphical and statistical analyses using continuous measures of alcohol use}problems and negative affect identified non-linear relationships where abstainers and occasional drinkers, as well as heavy and problem drinkers, were at risk of high anxiety and depression levels. This pattern, however, was not found in those aged & 60 years. The U-shaped relationship was not an artefact of abstainers being typical of the general population in their distribution of negative affect.

Conclusions. Studies of co-morbidity should acknowledge the possibility of non-linear associations and employ both continuous and discrete measures. Abstainers, as well as heavy drinkers, are at increased risk of symptoms of depression and anxiety disorders. Psychosocial factors may play a role in the U-shaped relationship between alcohol consumption and mortality.

INTRODUCTION

Co-morbidity of alcohol use disorders with mood and anxiety disorders is well established. In general population and other non-clinical samples, the overlap between alcohol abuse} dependence and broadly classified mood and anxiety disorders is greater than expected by chance (Regier et al. 1990 ; Kessler et al. 1996 ; Australian Bureau of Statistics, 1998 ; Swendsen et al. 1998). This finding also holds for most of the more specific diagnoses within the classes of mood and anxiety disorders, including major depression, dysthymia, mania, panic disorder, agoraphobia, social phobia, obsessive–com-

" Address for correspondence : Dr Bryan Rodgers, Centre for Mental Health Research, The Australian National University, Canberra, ACT 0200, Australia.

pulsive disorder, post-traumatic stress disorder and generalized anxiety disorder, as well as for high neuroticism scores (Helzer & Pryzbeck, 1988 ; Dick et al. 1994 ; Grant & Harford, 1995 ; Kessler et al. 1996 ; Magee et al. 1996 ; Heath et al. 1997 ; Swendsen et al. 1998). The relationship with simple phobias is perhaps less certain (Dick et al. 1994 ; Magee et al. 1996 ; Swendsen et al. 1998). In clinical groups, high rates of alcohol abuse}dependence have been reported for patients with a range of mood and anxiety disorders (Kushner et al. 1990 ; Schuckit & Hesselbrook, 1994 ; Merikangas & Angst, 1995 ; Page & Andrews, 1996 ; Fones et al. 1998) and high rates of mood and anxiety disorders and high levels of neuroticism have been found in groups receiving treatment for alcohol use disorders (Beckman et al. 1980 ; Mullan et al. 1986 ; Kushner et al. 1990 ; Schuckit & Hessel-

421

422 B. Rodgers and others

brook, 1994 ; Walker et al. 1994 ; Schuckit et al. 1997).

The reasons behind such co-morbidity are not fully understood (Merikangas & Gelernter, 1990 ; Vaillaint, 1993 ; Merikangas et al. 1998), but three general mechanisms have been pro- posed, all having some support : (1) alcohol abuse leads to higher levels of depression and anxiety by contributing to the inception, dur- ation or recurrence of these disorders (Aneshen- sel & Huba, 1983 ; Hansell & White, 1991 ; Hartka et al. 1991 ; Schuckit & Hesselbrook, 1994) ; (2) depression and anxiety lead to increased levels of alcohol consumption (Anesh- ensel & Huba, 1983 ; Kushner et al. 1990 ; Hartka et al. 1991 ; Thorlindsson & Vilhjalms- son, 1991) and the persistence of alcohol dependence (Kessler et al. 1996) as a form of self-medication ; (3) there are common deter- minants, environmental or genetic, for alcohol abuse}dependence and other disorders (Kendler et al. 1993, 1995 ; Lin et al. 1996 ; Tambs et al. 1997). The relationship of alcohol use and dependence with depression and anxiety differs across population subgroups (Bell et al. 1977 ; Neff, 1986 ; Flint, 1994 ; Grant & Harford, 1995 ; Jones-Webb et al. 1996), and so it is possible that the contributions of these three general mechanisms also vary.

Given the abundance of evidence on such co- morbidity, relating to categorically defined al- cohol abuse and dependence, there is a surprising lack of information on disorders or symptoms of anxiety and depression over the full range of alcohol consumption, from abstinence to heavy drinking, or across continua of problems associ- ated with use. It may be that the latter relationships do not follow a linear or even a monotonic trend. This possibility can be seen in the literature on the relationship of alcohol consumption with mortality, physical health and blood pressure, usually described as a U- shaped curve, where moderate drinkers show advantages compared with both non-drinkers and heavy drinkers (Jackson et al. 1985 ; Shaper, 1990 ; Marmot & Brunner, 1991 ; Beaglehole & Jackson, 1992 ; Poikolainen, 1995 ; Poikolainen et al. 1996). A few investigations, some con- cerned with the putative stress-buffering role of alcohol use (Pearlin & Radaburgh, 1976), suggest that abstainers may have higher levels of depression and anxiety symptoms than do

moderate drinkers, but findings are mixed and inconclusive. Bell et al. (1977) found that among those who reported recent life events, symptoms of anxiety (but not depression) were more frequent in abstainers than light drinkers. Although a similar interaction between life events and alcohol consumption was suggested for the prediction of depression scores in a rural sample, the scores of abstainers and heavy drinkers were not substantially higher than those of moderate drinkers and the differences were not statistically significant (Neff & Husaini, 1982). A second study by Neff (1984) using a mixed urban}rural sample also found no dif- ference in measured affect between abstainers and moderate drinkers. High rates of depressive symptoms have been reported for abstainers in the Alameda County study but, curiously, there was no indication of an increased risk in the heavy drinkers in this sample (Camacho et al. 1987). In some of these earlier studies, the failure to analyse separately data for men and women could well have distorted findings, as women are more likely to be abstainers (Bell et al. 1977 ; Neff & Husaini, 1982) and to have higher levels of depression (Weissman & Klerman, 1977 ; Bebbington, 1996). Two further reports, one from a meta-analysis, have provided stronger evidence for a U-shaped relationship between depressive symptoms and alcohol consumption, in men if not in women, but the differences between abstainers and light drinkers were again small (Golding et al. 1990 ; Fillmore et al. 1998).

A recent study, conducted in parallel with the present investigation, reported consistent and prominent U-shaped relationships for three measures of self-reported health in relation to alcohol consumption in a large national sample of young men and women in Britain (Power et al. 1998). One of the instruments used, the Malaise Inventory, is a common measure of psychological distress, incorporating symptoms of depression and anxiety (Rutter et al. 1970 ; Rodgers et al. 1999). The limitations of the British study include the restriction of data to early adulthood (age 23 and 33), the use of a general measure of psychological distress with no separate assessment of anxiety and de- pression, and the reliance on consumption over the previous 7 days as the only measure relating to alcohol use. The present study addressed these issues using data from an Australian

Depression, anxiety and alcohol use 423

general population sample aged 18–80 years. It had three main aims : (1) to determine whether a U-shaped relationship of symptoms of anxiety and depression with alcohol use was present in a different population and across different age groups ; (2) to examine the relationship with respect to separate measures of alcohol con- sumption and problems associated with use ; and, (3) to establish whether these relationships were consistent for separate measures of de- pression symptoms and anxiety symptoms.

METHOD

Sample

The target population was community-dwelling adult residents of Canberra, Australia. Potential participants for the Stress and Well-Being Project (Henderson et al. 1998) were randomly selected from those indicated to be between 18 and 79-years-old on the electoral roll for Canberra. Registration for voting is compulsory for Australian citizens aged 18 years and over with very few exceptions (being of unsound mind, serving a prison sentence of 5 years or more and having been convicted of treason or treachery). Detailed information on the content and purpose of the investigation and invitations to participate were sent by post to 5226 individuals and were followed up by telephone or personal calls from interviewers. Contact could not be made with 1172 individuals and this was attributed in part to the interval since the roll had last been updated and to the level of migration out of the area during that period. People who had moved out of the area could not be included in data collection because the project also required collection of DNA from cheek swabs. Of those contacted, 67 % (1294 men and 1431 women) consented to participate. The mean age of the achieved sample at the time of data collection, based on their self reports, was 42±6 years with 44±3 % aged 18–39, 43±3 % aged 40–59 and 12±4 % aged 60–80. This is close to the age distribution of the Canberra adult popu- lation, but with some under-representation of the youngest group and over-representation of the intermediate group.

Data collection

Participants responded to a self-completion questionnaire in the presence of the interviewer.

The questionnaire reiterated the voluntary nature of the study and the confidential nature of all responses. Participants’ names did not appear on the questionnaire and they were identified only by a study number. They sealed their questionnaire in an envelope before handing it back to the interviewer. The Com- mittee on the Ethics of Human Experimentation at The Australian National University had approved the protocol for the fieldwork and the procedure complied with the National Health and Medical Research Council guidelines on human experimentation.

The questionnaire included four measures of negative affect used in the present investigation : the depression and anxiety scales developed by Goldberg et al. (1988) with nine items per scale, and the state of depression and anxiety scales of the Delusions-Symptoms-States Inventory (DSSI}sAD), also known as the Personal Dis- turbance Scale (Bedford et al. 1976 ; Bedford & Deary, 1997), with seven items per scale. The factor structure of these scales in the present sample has been described in detail elsewhere (Christensen et al. 1999). The questionnaire also contained the Australian version of the WHO Alcohol Use Disorders Identification Test (AUDIT), a 10-item instrument which covers recent alcohol intake, dependence, and con- sequences of drinking in respect of the reactions of others, amnesia and injury (Saunders et al. 1993). The AUDIT has been used extensively as a screening instrument for alcohol-related social and medical problems and shows good sen- sitivity and specificity in the prediction of harmful drinking and diagnosed alcohol abuse and dependence (Bohn et al. 1995 ; Conigrave et al. 1995). The Australian version has small modifications to the response categories of the quantity-frequency items and it omits mention of family members from the item on other people expressing concern or suggesting the respondent should cut down on drinking (Coni- grave & Elvy, 1998).

Derivation of variables

Summary scores for all scales were derived in accordance with specifications for the original measures. Scores from the two Goldberg scales have a possible range of 0 to 9, from the DSSI}sAD of 0 to 21, and from the AUDIT of 0 to 40. Two additional subscales were derived

424 B. Rodgers and others

for the AUDIT instrument, representing prob- lems associated with drinking and alcohol consumption. This followed preliminary item analysis suggesting a two-factor solution, albeit with strongly correlated factors. The measure of problems associated with drinking (range 0 to 28) was obtained by the conventional scoring of the AUDIT scale with the omission of the first three items covering quantity-frequency and binge drinking. The measure of consumption was constructed from questions on the frequency of alcohol intake and the number of standard drinks (i.e. containing 10 g ethanol) consumed on typical drinking days, following standard procedures used for quantity-frequency assess- ment (Redman et al. 1987 ; Shakeshaft et al. 1999). This enabled the classification of indi- viduals into five categories of : (i) non-drinkers (had not drunk alcohol in the past year) ; (ii) occasional drinkers (monthly or less) ; (iii) lower drinking levels (up to 14 standard drinks per week for men and seven per week for women) ; (iv) higher drinking levels (up to 28 standard drinks per week for men and 14 per week for women) ; and (v) those drinking at hazardous or harmful levels as defined by the National Health and Medical Research Council (over 28 and 14 standard drinks respectively). Although the upper limit for the category of occasional drinkers was not defined in terms of weekly consumption, it was known that 94 % of these individuals typically drank less than 1 unit per week and that most of the remaining 6 % drank less than 2 units per week.

Data analysis

Relationships between scores derived from the AUDIT and the four assessments of negative affect were examined using measures in both categorical and continuous forms. The logistic regression procedure of SPSS Version 8.0 was used to estimate odds ratios for the prediction of high negative affect scores from high alcohol use}problems scores, adjusted for gender and age. Continuous measures were used in graphical procedures, linear regression analyses and gen- eral linear modelling. In addition to untrans- formed scores, logarithmic transformations were used to compensate for skewed distributions of AUDIT scores and negative affect scales. Graphical descriptions of observed relationships were obtained using the LOWESS option of the

SPSS scatterplot procedure, an iterative curve- fitting technique using a locally weighted least- squares method. This descriptive approach was considered appropriate, given that previous studies have not investigated the shape of the curve. Formal statistical tests for nonlinearity were also applied by the inclusion of quadratic and cubic terms in regression analyses. Multi- variate and univariate analyses of variance were used to model the association between alcohol consumption levels and negative affect scores with adjustments for gender and age.

RESULTS

Age and sex differences

Alcohol consumption and problems associated with use differed by sex and age (Table 1). Men drank more than women and reported more problems associated with drinking. The categor- ization of consumption using different thresh- olds for men and women tends to compensate for the gender difference in higher levels of drinking (but clearly not in the proportion of non-drinkers or occasional drinkers). Younger people reported more problems associated with drinking than did older people and had higher AUDIT total scores. The pattern for overall consumption by age was less systematic, al- though this concealed a greater frequency of drinking in older people and larger quantities consumed by younger people on typical drinking days. Age and sex differences were also found for measures of negative affect. As in other studies (Weissman & Klerman, 1977 ; Hender- son, 1994 ; Ernst & Angst, 1995 ; Bebbington, 1996), women had significantly higher scores than men and younger people had higher scores than older cohorts. These findings have been reported in greater detail elsewhere (Henderson et al. 1998). In view of these differences, further analyses were either carried out with adjustments for gender and age, with a variable representing younger (18–39 years), intermediate (40– 59 years) and older age groups (60–80 years), or were applied to separate age}sex groups.

Associations between dichotomous measures

Although the data permitted analyses based on continuous measures of negative affect and problems associated with drinking, it was also possible to dichotomize these measures and

Depression, anxiety and alcohol use 425

Table. 1. Alcohol consumption levels, AUDIT scores, and negative affect scores by sex and age

Men (age, years) Women (age, years)

18–39 40–59 60–80 18–39 40–59 60–80

N 553 575 166 654 604 173 Alcohol consumption level

Non-drinkers (%) 7±1 9±6 9±8 12±2 13±9 22±1 Occasional drinkers (%) 21±7 15±1 12±8 32±1 21±9 22±7 Lower drinking levels (%) 55±6 56±2 50±0 42±8 46±1 37±2 Higher drinking levels (%) 13±4 12±7 20±7 7±5 12±8 14±5 Hazardous}harmful levels (%) 2±2 6±4 6±7 5±4 5±3 3±5

AUDIT total score Mean 6±82 5±82 5±46 4±35 3±48 2±65 .. 5±13 4±65 3±89 3±74 3±08 2±33

AUDIT problems score Mean 1±70 1±07 0±66 0±77 0±34 0±06 .. 2±96 2±49 1±50 1±84 1±37 0±41

Goldberg depression score Mean 2±36 2±23 1±79 3±15 2±53 2±41 .. 2±22 2±31 1±74 2±50 2±43 2±14

Goldberg anxiety score Mean 3±46 3±39 2±61 4±69 3±88 3±20 .. 2±64 2±80 2±58 2±76 2±76 2±55

DSSI}sAD depression score Mean 1±72 1±61 0±96 2±58 1±91 1±40 .. 2±57 2±64 1±55 3±34 2±91 2±68

DSSI}sAD anxiety score Mean 2±56 2±56 2±11 3±96 3±10 2±61 .. 2±72 2±82 2±21 3±41 2±98 2±54

examine the degree of overlap for high scorers on any two dimensions. Given the established sensitivity and specificity of the scales as indi- cators of clinical disorders, such cross-tabu- lations provide an approximation to the identifi- cation of co-morbidity. Cut-points for each continuous score were chosen to give a pro- portion of high scorers in the total sample as close as possible to 12±5 %. This value is of the order expected for the prevalence of anxiety and depressive disorders in the general population (Australian Bureau of Statistics, 1998) and it follows an ‘ in-house ’ convention of utilizing cut-points representing the median, upper quar- tile, upper octile, etc, in order to dichotomize scale scores (Rodgers, 1996). This strategy pre- empts the possibility of selecting a cut-point because it achieves more desirable results. The upper octile corresponded to cuts of 5}6 for Goldberg depression, 7}8 for Goldberg anxiety, 4}5 for DSSI}sAD depression, and 6}7 for DSSI}sAD anxiety. Equivalent cuts were 9}10 for the AUDIT total score and 2}3 for the AUDIT problems score. Alcohol consumption was also dichotomized to distinguish those drinking at hazardous or harmful levels from those who were not.

Separate logistic regression analyses were carried out for the prediction of each negative affect measure with each alcohol use measure, plus gender and age group, as independent variables. Small percentages of high negative affect scores in the older group and their low rate of problems associated with drinking (es- pecially in older women) resulted in empty cells, so analyses were restricted to participants aged 18–59 years. Odds ratios for the prediction of high negative affect scores from high alcohol use}problems, adjusted for gender and age group, are shown in Table 2 along with their 95 % confidence intervals. These odds ratios were in the range 1±80 to 2±83 and were significantly greater than unity at the P ! 0±001 level for all except the associations between hazardous}harmful consumption and high scores on the two Goldberg scales. There was no systematic variation across different measures of negative affect, however. As indicated by the confidence intervals, the lower odds ratios for the Goldberg scales when predicted by alcohol consumption were not significantly different from those for the equivalent DSSI}sAD scales. For the different measures of alcohol use, there were instances (in association with Goldberg

426 B. Rodgers and others

Table. 2. Odds ratios (adjusted for age and sex) for prediction of high negative affect scores by high alcohol use}problems (18–59 years only)

Goldberg depression

Goldberg anxiety

DSSI}sAD depression

DSSI}sAD anxiety

AUDIT total score 1±89 2±05 2±34 2±31 95 % CI 1±36–2±63 1±45–2±90 1±71–3±20 1±65–3±24

AUDIT problems score 2±16 2±20 2±52 2±54 95 % CI 1±57–2±97 1±57–3±09 1±85–3±42 1±83–3±52

Hazardous}harmful consumption 1±82 1±80 2±83 2±54 95 % CI 1±16–2±86 1±12–2±88 1±87–4±28 1±65–3±92

scores) where the odds ratios for AUDIT problems were noticeably greater than those for harmful}hazardous consumption, but these differences fell short of statistical significance. No significant interaction terms were found for either age or gender in conjunction with alcohol use measures in any of these analyses, implying some consistency in odds ratios across the younger and intermediate groups and across men and women.

There have been suggestions that co-morbidity of anxiety disorders with alcohol use disorders is less evident in elderly people (Flint, 1994), therefore an attempt was made to include older participants in additional analyses in order to test for interactions with age. The problem of empty cells due to low base rates in the elderly was addressed by applying 12±5 % cut-points to each separate age}sex group rather than to the total sample. Logistic regression analyses using the resultant binary variables confirmed the general pattern of significant odds ratios re- ported in Table 2. Significant interactions with age were found in just two instances (Goldberg depression with AUDIT problems score and Goldberg anxiety with hazardous}harmful con- sumption), both reflecting lower odds ratios in the elderly compared to the younger and intermediate groups.

Relationships between continuous measures

The relationships between continuous measures of negative affect and alcohol use were initially investigated using the LOWESS procedure, where AUDIT total scores and problem scores were used as the independent variable (x axis) and negative affect scores as dependent variables (y axis). Non-linear relationships were revealed in many of the analyses carried out for separate age}sex groups (48 analyses in all) especially in

G o ld

b er

g d

ep re

ss io

n s

co re

( ln

)

2·5

2·0

1·5

1·0

0·5 3·53·02·52·01·51·00·50·0

AUDIT total score (ln)

F. 1. Fitted curves of Goldberg depression scores (ln) by AUDIT total scores (ln) derived from the LOWESS procedure. (Age 18–39 years : f f f f, men ; > E >, women. Age 40–59 years : - - - -, men ; ffff, women.)

relation to the AUDIT total score. Fig. 1 shows curves for the Goldberg depression score as the dependent variable (with α set at 0±5 for the LOWESS procedure), chosen as the most rep- resentative of the four negative affect variables utilized. The general pattern was one of a rise in depression and anxiety scores only at the upper end of the AUDIT scale. Many of these curves showed minimum values for measures of nega- tive affect at points around or above the median of the AUDIT total score distribution (a score of 4 on the untransformed scale). The non-linear component of the curves was confirmed by significant quadratic terms in regression analyses (including several instances where the initial linear regression was not significant) except those for the older age group. Fourteen of the 16

Depression, anxiety and alcohol use 427

Table 3. Estimated marginal means (adjusted for age) of negative affect scores by alcohol consumption level

Non-drinkers Occasional drinkers

Lower level

Higher level Hazardous}harmful

Goldberg depression Men 2±21 2±15 1±98 2±37* 2±77** Women 3±17*** 2±76* 2±38 2±52 3±79***

Goldberg anxiety Men 2±91 3±16 2±99 3±69** 3±72* Women 4±30** 3±99* 3±71 3±67 4±89***

DSSI}sAD depression Men 1±88* 1±35 1±31 1±43 2±17** Women 2±84*** 1±84 1±58 1±62 3±84***

DSSI}sAD anxiety Men 2±60 2±39 2±24 2±73* 2±96* Women 3±97*** 3±16 2±80 3±05 5±30***

* P ! 0±05 ; ** P ! 0±01 ; *** P ! 0±001 for contrasts with lower drinking level group.

analyses for the younger and intermediate age groups found significant quadratic terms. Some analyses also showed significant cubic terms, which typically provided a better fit to the steeper slopes at higher AUDIT scores and brought some flattening at the lower end. However, the pattern of an identifiable minimum point for each negative affect measure pervaded these analyses. Essentially, measures of negative affect showed a U-shaped or J-shaped relation- ship with AUDIT total score. There was no evidence to suggest that the quadratic com- ponent was any more or less important for anxiety measures by comparison with depression measures, or any indication of a difference between the younger and intermediate age groups or between men and women in this respect.

These non-linear relationships were investi- gated further by examining the separate com- ponents of AUDIT problem score and alcohol consumption derived from the quantity- frequency items. Using the problems score, two main differences were found by comparison with the AUDIT total score. First, with the exception of some analyses for the older group, all linear terms were significant and usually highly so. Secondly, quadratic terms were more often non- significant and, on the occasions when they were significant, they did not reflect minima for negative affect scores.

The importance of the quantity-frequency items to the U-shaped or J-shaped curves found for AUDIT total scores was confirmed by analyses using the five categories of alcohol

consumption ranging from non-drinkers to those drinking at hazardous or harmful levels. Initial multivariate analysis of variance, including all four negative affect measures as dependent variables, showed a significant effect for the consumption variable after adjustment for age group and sex (Wilk’s lambda : F

("',)#!() ¯ 5±09,

P ! 0±001) and an interaction of consumption with sex (Wilk’s lambda : F

("',)#!() ¯ 1±79, P !

0±027). The relationship of consumption level with each individual affect measure was then examined for men and women separately, using univariate analysis of variance. Additional paired contrasts between groups were carried out using the low drinking level group as the reference point for comparison. This followed the consideration of Shaper (1990) that the heterogeneity of non-drinkers makes them un- suitable for use as a baseline group, but also foreshadowed our subsequent investigation of different risk factors for negative affect in abstainers and heavy drinkers, requiring com- parison with moderate drinkers (Rodgers et al. 2000). Table 3 shows the estimated marginal means for each consumption group and the statistical significance of paired contrasts. All these analyses were carried out using both untransformed negative affect scores and log- arithmic transformations of these, and the pattern of results was found to be similar regardless of transformation. Results from the untransformed scores are shown in the table because they are easier to interpret intuitively. The most consistent and striking feature of these findings is the elevated negative affect scores of

428 B. Rodgers and others

N o n -d

ri n ke

O cc

as io

n al

L ow

er l

ev el

H ig

h er

l ev

H az

ar d o u s/

h ar

m fu

Alcohol consumption

M ea

n s

co re

F. 2. Mean Goldberg depression scores by level of alcohol consumption (younger and intermediate groups). (Age 18–39 years : E>E, men ; *>*, women. Age 40–59 years : V- - - -V, men ; X- - - -X, women.)

those drinking at hazardous}harmful levels. Non-drinkers also had high, but less extreme, mean scores and in five of the eight analyses these were exceeded only by the mean scores of those drinking at hazardous}harmful levels. Occasional drinkers and those drinking at the higher level also showed some tendency to increased negative affect, but differences com- pared to those drinking at the lower level were significant in only two and three instances respectively for these groups. The interactions with gender reflected more pronounced U- shaped relationships in women compared with men. This was most evident for the Goldberg anxiety scale, where non-drinking men were very similar to drinkers at the lower level of alcohol consumption.

Fig. 2 shows, as examples, plots of mean scores on the Goldberg depression scale across the five consumption groups, separately for men and women, and for younger and intermediate participants. Similar U-shaped relationships were found for the other measures of negative affect in the younger and intermediate groups. In nine of these 16 analyses, non-drinkers were

exceeded in their mean negative affect score only by those drinking at hazardous}harmful levels. U-shaped or J-shaped curves were not found for the older participants, however.

Artefactual explanations for U-shaped relationships

One possible explanation for observed U-shaped relationships is that they are artefactual, with scatterplots of negative affect by alcohol con- sumption representing an aggregate of one subpopulation with monotonic associations (i.e. symptoms positively correlated with drinking level) and a second population of abstainers whose distribution of negative affect scores is similar to that for the general population. This would have the effect of raising the left arm of the curve. However, under this explanation, the mean scores for abstainers could never exceed the population mean. In the present study, however, non-drinkers had mean negative affect scores that were consistently and significantly higher than the means for their age}sex group, and occasional drinkers often had mean scores that were also above the general mean.

A second two-population account is that abstainers include a group (or groups) who do not drink for reasons related to cultural factors and who have, for whatever reasons, elevated symptoms of depression and anxiety. The only data available in the present study that could shed light on this issue were from questions on racial background and country of birth. There was evidence that non-Caucasians and indi- viduals born in certain countries had lower alcohol consumption than the average for the whole sample, but they represented only 5 % of the total sample. The exclusion of these indi- viduals from analyses did not change the relationships described above.

DISCUSSION

These findings from a study of a general population sample in Australia confirmed the often reported relationships of high levels of alcohol use and associated problems with symp- toms of anxiety and depression. The results were consistent for two measures each of depression and anxiety symptoms and applied to mean scores on these scales and to the prevalence of high scorers identified by applying cut-points.

Depression, anxiety and alcohol use 429

Significant associations were evident in both women and men and odds ratios were of a similar order to those reported by studies using diagnostic assessments in general population samples (Helzer & Pryzbeck, 1988 ; Regier et al. 1990 ; Kessler et al. 1996 ; Swendsen et al. 1998). The relationships appeared weaker in those aged 60 and over compared with younger participants, consistent with reports that anxiety disorders are not associated (or may even be inversely associated) with alcoholism in the elderly (Flint, 1994). Caution is needed in interpretation, however, as the power of analyses for the older group was diminished by their smaller number and also by their lower AUDIT scores and symptom levels. Interactions with age could only be tested by applying lower cut-points for high negative affect and alcohol use to the older group, and this weakens the relevance of these findings to disorders of clinical severity.

The second prominent feature of the results was that non-drinkers had higher depression and anxiety scores than those who drank at low consumption levels. No equivalent U-shaped relationship was found in respect of the scale of problems arising from drinking. The latter scale is of a different nature in that a large majority of individuals score zero (as indicated by the means and standard deviations shown in Table 1), giving no variation at the lower end. For alcohol consumption, the majority of the population are moderate drinkers and non-drinkers constitute a smaller proportion.

The observed U-shaped relationships of nega- tive affect scores with alcohol consumption level could not be accounted for by artefactual explanations nor can they be readily explained by selective non-response of participants (whether due to refusal or moving out of the study area). This would require the dispro- portionate loss of either non-drinkers with low negative affect or moderate drinkers with high negative affect (or both) and there is no obvious basis for such possibilities. Our findings were consistent with results from a British general population sample of young adults studied at ages 23 and 33 years (Power et al. 1998). That investigation found similar patterns for high scores on the Malaise Inventory, and also for self-reported health and limiting longstanding illness. The present study extends these findings in identifying the same relationship in adults

aged 40–59 years, but failed to demonstrate U- shaped or J-shaped curves in those aged 60 years and over. Again, caution is needed in interpreting the negative finding, because of the limitation of statistical power, and replication is required. Our results also show that the U- shaped relationship in young and intermediate age groups applied both to symptoms of depression and symptoms of anxiety.

As well as the importance of these findings for the study of mental health, and specifically the co-morbidity of mental disorders, they have the added significance of mirroring the established finding of a U-shaped relationship between mortality and alcohol consumption (Shaper, 1990 ; Marmot & Brunner, 1991 ; Poikolainen, 1995). It is possible that the link with negative affect and other self-reported measures of ill- health could provide a clue to the explanation for increased mortality in abstainers, as such factors are known to be associated with increased mortality (Coryell et al. 1982 ; Haines et al. 1987 ; Murphy et al. 1987 ; Allgulander & Lavori, 1991 ; Huppert & Whittington, 1995). It is pertinent, therefore, that the patterns observed in the Australian and British studies are evident in relatively young people, suggesting that factors in or before early adulthood may be implicated in the relationship between alcohol consumption and mortality.

A general shortcoming of the present study was the absence of information, especially longitudinal data, that could elucidate the origins of the U-shaped relationship for negative affect. Past research on mortality and alcohol consumption can provide some pointers in this respect, and one hypothesis explored was that ex-drinkers accounted for the high mortality of abstainers (Shaper et al. 1988 ; Wannamethee & Shaper, 1997). A specific limitation of the present study was that it was unable to determine directly whether ex-drinkers could account for the high symptom levels in current non-drinkers. However, there are three lines of evidence against this interpretation. First, this hypothesis predicts that the strength of the U-shaped relationship and, in particular, the difference between ab- stainers and moderate drinkers would increase with age. There was no support for this in our findings and, indeed, the trend was for the U- shape to be less prominent in the older group. Secondly, Power et al. (1998) did test this

430 B. Rodgers and others

hypothesis directly in the British study, using longitudinal data. The U-shaped relationship of psychological distress with alcohol consumption level was still evident after the exclusion of former heavy drinkers and those who reported ever having had drink problems (based on the CAGE screening questionnaire). Thirdly, data from the British sample and also from an early study in Boston (Goldman & Najman, 1984) indicate that the proportion of ex-problem or ex-heavy drinkers is as large, or even larger, among current moderate drinkers as among current abstainers. The absence of information on drinking history in the present study may, therefore, have led to an underestimation of the observed differences between moderate and non- drinkers.

What other explanations are there for the non-monotonic associations between negative affect and alcohol use ? A general framework for approaching this question can be derived by adapting the three general accounts of co- morbidity outlined in the Introduction : (1) the influence of drinking on depression and anxiety ; (2) causal influences in the reverse direction ; and (3) third-factor models with common risk factors for negative affect and alcohol use. From the first of these, a parsimonious explanation of the U-shaped relationship is that moderate drinking is somehow protective in relation to anxiety and depression. This could encompass direct effects of alcohol ingestion or associated social benefits. The second mechanism, that negative affect influences drinking, suggests that high de- pression and}or anxiety can lead to either abstinence or to increased consumption. A full account would need, therefore, to identify moderating factors that determine whether a decrease or increase in consumption occurs. These could be personal characteristics of individuals or features of their circumstances. The final general model can be subdivided into the role of third factors that are implicated in the negative affect of both abstainers and heavy users and factors that influence one or other of these groups. As an example of the former, financial hardship could lead to either heavy drinking or abstinence (because of increased stress and economic constraint respectively) and is a known risk factor for depression and anxiety. For the latter, non-drinkers may have per- sonality characteristics that place them at risk

for depression and anxiety whereas heavy drinkers may be at risk because of greater psychosocial adversity, including life events. These three general models take on a greater complexity when possible combinations are considered. A model of one type could apply to the overlap of heavy drinking and negative affect while a model of a different type could account for the association between abstinence and negative affect. It is evident that longitudinal studies have a particularly important role to play in disentangling these possible contributory mechanisms.

There are additional important research questions concerning the nature of the observed U-shaped relationship, some of which can be addressed through cross-sectional studies. It needs to be established whether the associations found for negative affect scales also apply to specific mood and anxiety disorders, defined by standard diagnostic criteria. It would also be valuable to determine whether U-shaped re- lationships apply to a greater diversity of populations, either across geographical locations or across cultural groups and population sub- groups within regions. It is possible, for example, that the elevated psychological distress found in non-drinkers may only occur in populations or cultures where non-drinkers constitute a min- ority.

Conclusions

This study, in conjunction with a parallel British investigation, has identified U-shaped relation- ships in associations of depression and anxiety symptoms with alcohol use, such that both non- drinkers and heavy drinkers report more symp- toms than those drinking at moderate levels. These findings have methodological and sub- stantive implications. They indicate that a full account of the co-morbidity of alcohol use disorders and other mental disorders should acknowledge the possibility of non-linear associ- ations and employ continuous as well as discrete measures. They further signal that more at- tention should be given to abstainers as a group at risk for mood and anxiety disorders. And, they hint at the possibility that psychosocial factors evident in early adulthood (or before) play a role in the established U-shaped re- lationship between alcohol consumption and mortality.

Depression, anxiety and alcohol use 431

This work was supported by a Unit Grant from the National Health and Medical Research Council (No 973302).

REFERENCES

Allgulander, C. & Lavori, P. W. (1991). Excess mortality among 3302 patients with ‘ pure ’ anxiety neurosis. Archives of General Psychiatry 48, 599–602.

Aneshensel, C. S. & Huba, G. J. (1983). Depression, alcohol use, and smoking over one year : a four-wave longitudinal causal model. Journal of Abnormal Psychology 92, 134–150.

Australian Bureau of Statistics (1998). Mental Health and Wellbeing : Profile of Adults, Australia. Cat. No. 4326.0. ABS : Canberra.

Beaglehole, R. & Jackson, R. (1992). Alcohol, cardiovascular diseases and all causes of death : a review of the epidemiological evidence. Drug and Alcohol Review 11, 275–290.

Bebbington, P. (1996). The origins of sex differences in depressive disorder : bridging the gap. International Review of Psychiatry 8, 295–332.

Beckman, L. J., Day, T., Bardsley, P. & Seeman, A. Z. (1980). The personality characteristics and family backgrounds of women alcoholics. International Journal of the Addictions 15, 147–154.

Bedford, A. & Deary, I. J. (1997). The Personal Disturbance Scale (DSSI}sAD) : development, use and structure. Personality and Individual Differences 22, 493–510.

Bedford, A., Foulds, G. A. & Sheffield, B. F. (1976). A new personal disturbance scale (DSSI}sAD). British Journal of Social and Clinical Psychology 15, 387–394.

Bell, R. A., Keeley, K. A. & Buhl, J. M. (1977). Psychopathology and life events among alcohol users and nonusers. In Currents in Alcoholism Vol. II (ed. F. A. Seixas), pp. 103–123. Grune & Stratton : New York.

Bohn, M. J., Babor, T. F. & Kranzler, H. R. (1995). The Alcohol Use Disorders Identification Test (AUDIT) : validation of a screening instrument for use in medical settings. Journal of Studies on Alcohol 56, 423–432.

Camacho, T. C., Kaplan, G. A. & Cohen, R. D. (1987). Alcohol consumption and mortality in Alameda County. Journal of Chronic Diseases 40, 229–236.

Christensen, H., Jorm, A. F., Mackinnon, A. J., Korten, A. E., Jacomb, P. A., Henderson, A. S. & Rodgers, B. (1999). Age differences in depression and anxiety symptoms : a structural equation modelling analysis of data from a general population sample. Psychological Medicine 29, 325–339.

Conigrave, K. M. & Elvy, G. (1998). An Australian modification of the popular AUDIT questionnaire. Drug and Alcohol Review 17, 223–224.

Conigrave, K. M., Hall, W. D. & Saunders, J. B. (1995). The AUDIT questionnaire : choosing a cut-off score. Addiction 90, 1349–1356.

Coryell, W., Noyes, R. & Clancy, J. (1982). Excess mortality in panic disorder. Archives of General Psychiatry 39, 701–703.

Dick, C. L., Sowa, B., Bland, R. C. & Newman, S. C. (1994). Epidemiology of psychiatric disorders in Edmonton. Phobic disorders. Acta Psychiatrica Scandinavica (Suppl. 376), 36–44.

Ernst, C. & Angst, J. (1995). Depression in old age. Is there a real decrease in prevalence ? A review. European Archives of Psychiatry and Clinical Neuroscience 245, 272–287.

Fillmore, K. M., Golding, J. M., Graves, K. L., Kniep, S., Leino, E. V., Romelsjo$ , A., Shoemaker, C., Ager, C. R., Allebeck, P. & Ferrer, H. P. (1998). Alcohol consumption and mortality. I. Characteristics of drinking groups. Addiction 93, 183–203.

Flint, A. J. (1994). Epidemiology and comorbidity of anxiety disorders in the elderly. American Journal of Psychiatry 151, 640–649.

Fones, C. S. L., Manfro, G. G. & Pollack, M. H. (1998). Social phobia : an update. Harvard Review of Psychiatry 5, 247–259.

Goldberg, D., Bridges, K., Duncan-Jones, P. & Grayson, D. (1988). Detecting anxiety and depression in general medical settings. British Medical Journal 297, 897–899.

Golding, J. M., Burnam, M. A. & Wells, K. B. (1990). Alcohol use and depressive symptoms among Mexican Americans and non- Hispanic Whites. Alcohol and Alcoholism 25, 421–432.

Goldman, E. & Najman, J. M. (1984). Lifetime abstainers, current abstainers and imbibers : a methodological note. British Journal of Addiction 79, 309–314.

Grant, B. F. & Harford, T. C. (1995). Comorbidity between DSM-IV alcohol use disorders and major depression : results of a national survey. Drug and Alcohol Dependence 39, 197–206.

Haines, A. P., Imeson, J. D. & Meade, T. W. (1987). Phobic anxiety and ischaemic heart disease. British Medical Journal 295, 297–299.

Hansell, S. & White, H. R. (1991). Adolescent drug use, psychological distress, and physical symptoms. Journal of Health and Social Behavior 32, 288–301.

Hartka, E., Johnstone, B., Leino, E. V., Motoyoshi, M., Temple, M. T. & Fillmore, K. M. (1991). A meta-analysis of depressive symptomatology and alcohol consumption over time. British Journal of Addiction 86, 1283–1298.

Heath, A. C., Bucholz, K. K., Madden, P. A., Dinwiddie, S. H., Slutske, W. S., Bierut, L. J., Statham, D. J., Dunne, M. P., Whitfield, J. B. & Martin, N. G. (1997). Genetic and environmental contributions to alcohol dependence risk in a national twin sample : consistency of findings in women and men. Psychological Medicine 27, 1381–1396.

Helzer, J. E. & Pryzbeck, T. R. (1988). The co-occurrence of alcoholism with other psychiatric disorders in the general population and its impact on treatment. Journal of Studies on Alcohol 49, 219–224.

Henderson, A. S. (1994). Does ageing protect against depression ? Social Psychiatry and Psychiatric Epidemiology 29, 107–109.

Henderson, A. S., Jorm, A. F., Korten, A. E., Jacomb, P., Christen- sen, H. & Rodgers, B. (1998). Symptoms of depression and anxiety during adult life : evidence for a decline in prevalence with age. Psychological Medicine 28, 1321–1328.

Huppert, F. A. & Whittington, J. E. (1995). Symptoms of psycho- logical distress predict 7-year mortality. Psychological Medicine 25, 1073–1086.

Jackson, R., Stewart, A., Beaglehole, R. & Scragg, R. (1985). Alcohol consumption and blood pressure. American Journal of Epidemiology 122, 1037–1044.

Jones-Webb, R., Jacobs, D. R. Jr., Flack, J. M. & Liu, K. (1996). Relationships between depressive symptoms, anxiety, alcohol consumption, and blood pressure : results from the CARDIA study. Alcoholism : Clinical and Experimental Research 20, 420–427.

Kendler, K. S., Heath, A. C., Neale, M. C., Kessler, R. C. & Eaves, L. J. (1993). Alcoholism and major depression in women. A twin study of the causes of comorbidity. Archives of General Psychiatry 50, 690–698.

Kendler, K. S., Walters, E. E., Neale, M. C., Kessler, R. C., Heath, A. C. & Eaves, L. J. (1995). The structure of the genetic and environmental risk factors for six major psychiatric disorders in women. Phobia, generalized anxiety disorder, panic disorder, bulimia, major depression and alcoholism. Archives of General Psychiatry 52, 374–383.

Kessler, R. C., Nelson, C. B., McGonagle, K. A., Edlund, M. J., Frank, R. G. & Leaf, P. J. (1996). The epidemiology of co- occurring addictive and mental disorders : implications for pre- vention and service utilization. American Journal of Ortho- psychiatry 66, 17–31.

Kushner, M. G., Sher, K. J. & Beitman, B. D. (1990). The relation between alcohol problems and the anxiety disorders. American Journal of Psychiatry 147, 685–695.

Lin, N., Eisen, S. A., Scherrer, J. F., Goldberg, J., True, W. R., Lyons, M. J. & Tsuang, M. T. (1996). The influence of familial and non-familial factors on the association between major depression and substance abuse}dependence in 1874 monozygotic male twin pairs. Drug and Alcohol Dependence 43, 49–55.

Magee, W. J., Eaton, W. W., Wittchen, H.-U., McGonagle, K. A. & Kessler, R. C. (1996). Agoraphobia, simple phobia, and social phobia in the National Comorbidity Survey. Archives of General Psychiatry 53, 159–168.

432 B. Rodgers and others

Marmot, M. & Brunner, E. (1991). Alcohol and cardiovascular disease : the status of the U shaped curve. British Medical Journal 303, 565–568.

Merikangas, K. R. & Angst, J. (1995). Comorbidity and social phobia : evidence from clinical, epidemiologic, and genetic studies. European Archives of Psychiatry and Clinical Neuroscience 244, 297–303.

Merikangas, K. R. & Gelernter, C. S. (1990). Comorbidity for alcoholism and depression. Psychiatric Clinics of North America 13, 613–632.

Merikangas, K. R., Stevens, D. E., Fenton, B., Stolar, M., O’Malley, S., Woods, S. W. & Risch, N. (1998). Co-morbidity and familial aggregation of alcoholism and anxiety disorders. Psychological Medicine 28, 773–788.

Mullan, M. J., Gurling, H. M. D., Oppenheim, B. E. & Murray, R. M. (1986). The relationship between alcoholism and neurosis : evidence from a twin study. British Journal of Psychiatry 148, 435–441.

Murphy, J. M., Monson, R. R., Olivier, D. C., Sobol, A. M. & Leighton, A. H. (1987). Affective disorders and mortality. Archives of General Psychiatry 44, 473–480.

Neff, J. A. (1984). The stress-buffering role of alcohol consumption : the importance of symptom dimension. Journal of Human Stress 10, 35–42.

Neff, J. A. (1986). Alcohol consumption and psychological distress among US Anglos, Hispanics and Blacks. Alcohol and Alcoholism 21, 111–119.

Neff, J. A. & Husaini, B. A. (1982). Life events, drinking patterns and depressive symptomatology : the stress buffering role of alcohol consumption. Journal of Studies on Alcohol 43, 301–318.

Page, A. C. & Andrews, G. (1996). Do specific anxiety disorders show specific drug problems ? Australian and New Zealand Journal of Psychiatry 30, 410–414.

Pearlin, L. I. & Radaburgh, C. W. (1976). Economic strains and the coping functions of alcohol. American Journal of Sociology 82, 652–663.

Poikolainen, K. (1995). Alcohol and mortality : a review. Journal of Clinical Epidemiology 48, 455–465.

Poikolainen, K., Vartiainen, E. & Korhonen, H. J. (1996). Alcohol intake and subjective health. American Journal of Epidemiology 144, 346–350.

Power, C., Rodgers, B. & Hope, S. (1998). U-shaped relation for alcohol consumption and health in early adulthood and impli- cations for mortality (research letter). Lancet 352, 877.

Redman, S., Sanson-Fisher, R. W., Wilkinson, C., Fahey, P. P. & Gibberd, R. W. (1987). Agreement between two measures of alcohol consumption. Journal of Studies on Alcohol 48, 104–108.

Regier, D. A., Farmer, M. E., Rae, D. S., Locke, B. Z., Keith, S. J., Judd, L. L. & Goodwin, F. K. (1990). Comorbidity of mental disorders with alcohol and other drug abuse : results from the Epidemiologic Catchment Area (ECA) Study. Journal of the American Medical Association 264, 2511–2518.

Rodgers, B. (1996). Reported parental behaviour and adult affective symptoms. 1. Associations and moderating factors. Psychological Medicine 26, 51–61.

Rodgers, B., Pickles, A., Power, C., Collishaw, S. & Maughan, B. (1999). Validity of the Malaise Inventory in general population samples. Social Psychiatry and Psychiatric Epidemiology 34, 333–341.

Rodgers, B., Korten, A., Jorm, A. F., Christensen, H., Henderson, S. & Jacomb, P. A. (2000). Risk factors for depression and anxiety in abstainers, moderate drinkers and heavy drinkers. (Submitted).

Rutter, M., Tizard, J. & Whitmore, K. (1970). Education, Health and Behaviour. Longmans : London.

Saunders, J. B., Aasland, O. G., Babor, T. F., De La Fuente, J. R. & Grant, M. (1993). Development of the Alcohol Use Disorders Identification Test (AUDIT) : WHO collaborative project on early detection of persons with harmful alcohol consumption – II. Addiction 88, 791–804.

Schuckit, M. A. & Hesselbrook, V. (1994). Alcohol dependence and anxiety disorders : what is the relationship ? American Journal of Psychiatry 151, 1723–1734.

Schuckit, M. A., Tipp, J. E., Bucholz, K. K., Nurnberger, J. I., Hesselbrock, V. M., Crowe, R. R. & Kramer, J. (1997). The life- time rates of three major mood disorders and four major anxiety disorders in alcoholics and controls. Addiction 92, 1289–1304.

Shakeshaft, A. P., Bowman, J. A. & Sanson-Fisher, R. W. (1999). A comparison of two retrospective measures of weekly alcohol consumption : diary and quantity}frequency index. Alcohol and Alcoholism 34, 636–645.

Shaper, A. G. (1990). Alcohol and mortality : a review of prospective studies. British Journal of Addiction 85, 837–847.

Shaper, A. G., Wannamethee, G. & Walker, M. (1988). Alcohol and mortality in men : explaining the U-shaped curve. Lancet ii, 1267–1273.

Swendsen, J. D., Merikangas, K. R., Canino, G. J., Kessler, R. C., Rubio-Stipec, M. & Angst, J. (1998). The comorbidity of alcoholism with anxiety and depressive disorders in four geographic communities. Comprehensive Psychiatry 39, 176–184.

Tambs, K., Harris, J. R. & Magnus, P. (1997). Genetic and environmental contributions to the correlation between alcohol consumption and symptoms of anxiety and depression. Results from a bivariate analysis of Norwegian twin data. Behavior Genetics 27, 241–250.

Thorlindsson, T. & Vilhjalmsson, R. (1991). Factors related to cigarette smoking and alcohol use among adolescents. Adolescence 26, 399–418.

Vaillaint, G. E. (1993). Is alcoholism more often the cause or the result of depression ? Harvard Review of Psychiatry 1, 94–99.

Walker, R. D., Howard, M. O., Lambert, M. D. & Suchinsky, R. (1994). Psychiatric and medical comorbidities of veterans with substance use disorders. Hospital and Community Psychiatry 45, 232–237.

Wannamethee, S. G. & Shaper, A. G. (1997). Lifelong teetotallers, ex-drinkers and drinkers : mortality and the incidence of major coronary heart disease events in middle-aged men. International Journal of Epidemiology 26, 523–531.

Weissman, M. M. & Klerman, G. L. (1977). Sex differences and the epidemiology of depression. Archives of General Psychiatry 34, 98–111.

View publication statsView publication stats

https://www.researchgate.net/publication/12492744