OUP user menu

Physical Activity and Risk of Alcohol Use Disorders: Results from a Prospective Cohort Study

Louise Kristiansen Ejsing, Ulrik Becker, Janne S. Tolstrup, Trine Flensborg-Madsen
DOI: http://dx.doi.org/10.1093/alcalc/agu097 agu097 First published online: 27 December 2014

Abstract

Aims: To examine the effect of physical activity on risk of developing alcohol use disorders in a large prospective cohort study with focus on leisure-time physical activity. Methods: Data came from the four examinations of the Copenhagen City Heart Study (CCHS), performed in 1976–1978, 1981–1983, 1991–1994 and 2001–2003. Information on physical activity (classified as Moderate/high, low or sedentary) and covariates was obtained through self-administered questionnaires, and information on alcohol use disorders was obtained from the Danish Hospital Discharge Register, the Danish Psychiatric Central Research Register and the Winalco database. In total, 18,359 people participated in the study, a mean follow-up time of 20.9 years. Cox proportional hazards model with delayed entry was used. Models were adjusted for available covariates (age, smoking habits, alcohol intake, education, income and cohabitation status) including updated time-dependent variables whenever possible. Results: A low or moderate/high leisure-time physical activity was associated with almost half the risk of developing alcohol use disorder compared with a sedentary leisure-time physical activity. This translates into a 1.5- to 2-fold increased risk of developing alcohol use disorder (Hazard ratios for men 1.64; 95% CI 1.29–2.10 and women 1.45; 1.01–2.09) in individuals with a sedentary leisure-time physical activity, compared with a moderate to high level. However, when stratifying by presence of other psychiatric disorders, no association was observed in women with psychiatric comorbidity. Residual confounding may have been present in this study, especially according to rough measures of income and education. Conclusions: In both men and women, being sedentary in leisure time was a risk factor for developing an alcohol use disorder.

INTRODUCTION

According to DSM, alcohol use disorders (AUD) is a psychiatric diagnosis consisting of alcohol abuse and alcohol dependence. Alcohol abuse is characterized by a pattern of alcohol intake causing damage to health, while alcohol dependence is characterized by physiological, behavioral and cognitive symptoms. It is estimated that 585,000, corresponding to approximately 13% of the adult Danish population, has a harmful use of alcohol, and that 140,000, corresponding to approximately 3% of the adult Danish population, fulfills criteria for alcohol dependence (Gottlieb Hansen et al., 2011). Numerous health consequences, both somatic and psychiatric as well as economic consequences, are associated with AUD (Harwood et al., 1998; Anderson and Baumberg, 2006).

Among the known risk factors for AUD are alcohol intake (Flensborg-Madsen et al., 2007), genetic factors (Mayfield et al., 2008) male gender (Nolen-Hoeksema, 2004), low alcohol reactivity (Nolen-Hoeksema, 2004) and a history of sexual assault (Nolen-Hoeksema, 2004). Other psychiatric diseases such as mood disorders and personality disorders as well as drug abuse are also known risk factors for AUD (Flensborg-Madsen et al., 2009).

Most studies have shown an association between sedentary life style and increased risk of a number of somatic diseases. Cross-sectional studies have previously linked physical activity to alcohol intake (Barrett et al., 1995; Smothers and Bertolucci, 2001; French et al., 2009; Liangpunsakul et al., 2010). Three studies have found a high alcohol intake to be associated with a high level of physical activity (Barrett et al., 1995; Smothers and Bertolucci, 2001; French et al., 2009), whereas one study has found a high alcohol intake to be associated with a lower level of physical activity (Liangpunsakul et al., 2010).

Physical activity has not yet been linked to the prevention of AUD, but previous studies indicate that physical activity may be useful in the treatment of AUD (Sinyor et al., 1982; Murphy et al., 1986; Brown et al., 2009). Also, physical activity has previously been shown to be associated with a lower risk of other psychiatric disorders such as depression (Mikkelsen et al., 2010).

Therefore, the objective of this study was to examine the effect of physical activity on the risk of developing an AUD using data from a large, Danish prospective cohort study.

METHODS

Population

Data used in this study came from the four examinations of the Copenhagen City Heart Study (CCHS), performed in 1976–1978, 1981–1983, 1991–1994 and 2001–2003. The CCHS is a prospective cohort study and the study procedures have been described elsewhere in detail (Appleyard et al., 1989; Schnohr et al., 2001). In 1976 a random sample of individuals above 20 years of age living in central Copenhagen was enrolled in the study. The sample was age stratified within 5-year age groups, with main emphasis on the age groups from 35 to 70 years. Individuals selected for the study were invited according to their date of birth, ensuring that subsets of the sample examined during any period of time within the examination period would also constitute a random subsample. Three weeks prior to the examination, the individuals received an invitation to participate in a health examination at the CCHS and if the invitation was not returned or the person did not show up at the examination, a new invitation was sent 6 months later. A self-administered questionnaire, concerning symptoms and diseases, familial disposition, education and socio-economic status, smoking and drinking habits, weight and weight changes, physical activity at work and leisure, medication and contact with the healthcare system, was sent along with the invitation and filled in by the participant before arriving for the examination. The questionnaires were checked by a trained staff member for missing information. At the first examination in 1976–1978, 14,223 persons aged 20–93 years participated (response rate 74%). In the next three examinations the sample were supplemented with younger persons. In 1981–1983, a total of 12,698 persons chose to participate (response rate 70%). In 1991–1994 10,135 persons participated (response rate 61%), and in the last examination conducted in 2001–2003, 6237 persons participated (response rate 50%).

Physical activity

In this study, information on leisure-time physical activity was based on the following question in the self-administered questionnaire: If you should state your LEISURE TIME PHYSICAL ACTIVITY including transport to and from work during the last year, in which group do you think you should be placed (one answer only)? The four possible answers were ‘Almost entirely inactive or light physical activity less than 2 hours per week’; ‘Light physical activity for 2–4 h per week’; ‘Light physical activity for more than 4 hours per week or more vigorous physical activity for 2–4 h per week’; and ‘More vigorous physical activity for more than 4 hours per week or regular heavy exercise or competitive sports several times per week’. This measure of physical activity has proven to be predictive in mortality, cancer and coronary heart disease (Schnohr et al., 2005, 2006; Schnohr, 2009). In the analyses, the first category was defined as a sedentary level of leisure-time physical activity, the second as a low level, and due to few respondents in the fourth category, the third and fourth categories were combined into one, defined as a moderate/high level of leisure-time physical activity.

Occupational physical activity was in the same way grouped into three categories: sedentary level, low level and moderate/high level. Both leisure-time and occupational physical activity was measured in all four examinations of the CCHS. The focus of the present paper is the relation between leisure-time physical activity and development of AUD and not a combined measure of leisure-time and occupational physical activity. Furthermore a combined measure would be difficult to interpret.

Follow-up

Information on AUD after inclusion was obtained by record linkage to three different registers using the unique person identification number: the National Patient Registry, the Danish Psychiatric Central Research Register and the Winalco database.

The National Patient Registry contains records of all individuals admitted to a Danish hospital, including psychiatric hospitals, since 1976. Subjects were followed in this register until the end of the study period on 3 May 2011. The Danish Psychiatric Central Research Register contains records of all individuals admitted to a psychiatric hospital in Denmark since 1969, and the Winalco database contains records of all individuals who have been treated for AUD in an outpatient clinic for AUD covering the greater Copenhagen and Frederiksberg municipalities since 1954. More than 90% of these patients fulfill dependence criteria. As a supplement to the record linkage to the National Patient Registry, subjects were followed in the Danish Psychiatric Central Research Register until 11 February 2002, and in the Winalco database until 7 April 2009. Individuals who were given an ICD diagnosis of an AUD (ICD-8: 303.0–303.2 and 303.9; ICD-10: F10.1 and F10.2) in either of the two registers and individuals who were registered in the Winalco database were considered to have an AUD at the given time.

All participants were followed from the time of their first examination until first time diagnosed with AUD, death or emigration, or until the end of the study period (3 May 2011).

Covariates

The variables were selected on the basis of existing literature and own assumptions, as well as availability in the study. All analyses were adjusted for age (in days) and stratified by gender. Information on age and gender came from the Civil Registration System (CPR). Information on all other covariates came from the self-administered questionnaires.

Smoking was categorized as never smoker, ex-smoker and current smoker.

Alcohol intake was categorized as drinking less than 1 drink/week, 1–6 drinks/week, 7–13 drinks/week, 14–21 drinks/week, 21–34 drinks/week and more than 34 drinks/week. One Danish standard drink contains 12 g alcohol and sensible drinking limits in Denmark is 14 drinks per week for women and 21 drinks per week for men.

In the questionnaires, subjects were asked about their income in 3–7 categories. Based on these categories, income was categorized as low, medium and high in this study, depending on percentage. The amounts used as cut-offs varied over the four examinations due to inflation. Education was categorized as less than 8 years, between 8 and 11 years, and more than 11 years of education.

Cohabitation status was categorized as living alone and living with someone else (partner, children etc.).

Information on social network was only available in the third examination of the CCHS, where participants were asked how much contact they had with parents, children, other family members, coworkers (after working hours), neighbors, childhood friends, other friends and home carer or the like. A scale from 0 to 8 was constructed, where 0 corresponded to no contact, and 8 to some contact with all types of acquaintances. In the analyses, subjects were categorized as having a small network (0–2), a medium network (3–5) or a large network (6–8).

Information on children living at home was only available in the first, second and third examination of the CCHS. It was categorized as having children living at home or not.

Statistical methods

Data were analyzed by means of the Cox proportional hazards model with delayed entry (SAS version 9.1) implemented. Age (in days) was used as the underlying time axis to ensure maximal adjustment for confounding by age. Furthermore, all analyses were performed stratified by gender and adjusted for other relevant confounders. We included time-dependent covariates i.e. explanatory variables with updated information. That is, the most recent value of a time-dependent variable was used at each specific time in the model.

The main analyses were performed adjusted for age only, adjusted for all covariates and adjusted for only some covariates (social variables and lifestyle variables, respectively). Also, a backwards model selection was performed. The focus of this study was on leisure-time physical activity, rather than work-time physical activity. However, analyses were performed using work-time physical activity as the outcome, in order to see, whether the effects were similar. As reference group we chose moderate/high leisure-time physical activity throughout the analyses because sedentary leisure-time physical activity was the smallest group resulting in the most narrow confidence limit in estimates in the two other groups. Second, it may be easier to interpret hazard ratios higher than 1.00 and third it is meaningful to choose one of the extremes as reference, taking into account the natural order of the categories. In addition the main analyses were performed with sedentary leisure-time physical activity as reference group.

Additional analyses including only data from the third examination of the CCHS were performed in order to see the effect on the results of adjusting for social network and having children, since information on these variables were not available in all four examinations. Furthermore, analyses stratified by alcohol intake (above or below the sensible drinking limits), stratified by presence of psychiatric disorders other than AUD, and analyses stratified by presence of chronic diseases were performed. ‘Other psychiatric disorders’ was defined as having been given at least one psychiatric diagnosis other than AUD previous to or during the study period. ‘Chronic disease’ was defined as having been given at least one diagnosis of chronic disease previous to or up to three years after entry into the study.

Finally, sensitivity analyses introducing a time lag of two and three years were performed in order to examine whether latent disease at baseline caused some subjects to reduce their leisure-time physical activity level before entering the study.

RESULTS

The total number of participants in this study was 18,974. Most of these participated in more than one examination. 564 subjects were excluded due to existing AUD at baseline or missing or meaningless dates of entry into or exit from the study, and 51 were deleted due to missing information on leisure-time physical activity. This resulted in 18,359 subjects eligible for analyses. The mean follow-up time was 20.9 years. At the end of the study period, a total of 736 subjects had been diagnosed with AUD (corresponding to 4% of the total number of subjects eligible for analyses). Of these, 284 were women and 488 were men (Table 1). At all four examinations only 2–3% reported doing heavy physical activity in leisure time. Work-time physical activity was more evenly distributed across categories, but still only 2–6% reported doing heavy physical activity at work.

View this table:
Table 1.

Descriptive variables in the four examinations of the Copenhagen city heart study

1976–19781981–19831991–19942001–2003
Number of persons
N14,22312,69810,1356237
1st time participants
N (%)14,223 (100%)1563 (12.3%)2360 (23.3%)828 (13.3%)
Age
 Mean (range)53 (20–93)56 (20–98)58 (21–98)59 (20–97)
Sex
 % men46454443
Leisure-time physical activity
 % sedentary20171310
 % low54485351
 % moderate/high26353339
Work-time physical activity
 % sedentary28353436
 % low39403338
 % moderate/high33253226
Alcohol intake (drinks/week)
 % <12231232
 % 1–634293239
 % 7–1319192231
 % 14–201091114
 % 21–3477811
 % >347444
Smoking habits
 % never20212532
 % ex17212635
 % current64584833
Educational level (years)
 % <848463624
 % 8–1140414240
 % >1112132236
Cohabitation status
 % yes73696260
 % no27313840
Income
 % low28324039
 % middle53464445
 % high19221616

As shown in Table 1, the majority reported having a weekly alcohol intake between 1 and 6 drinks in the first, third and fourth examination, while the majority drank less than one drink per week in the second examination. The percentage of current smokers decreased from 64 to 33% from first to fourth examination, and the percentage of participants with less than 8 years of education decreased from 48 to 24% in this period, as well as the percentage living with other people decreased from 73 to 60%. At all four examinations, most subjects belonged to the middle-income category.

Leisure-time physical activity and risk of AUD

In all models, an increased risk of AUD was observed among both men and women with a sedentary level of leisure-time physical activity compared to a moderate/high level (Table 2). The associations were generally stronger in men. The risk associated with a low level of leisure-time physical activity did not differ significantly from the risk associated with a moderate/high level. For women the hazard ratio from the fully adjusted model was 1.45 (95% CI: 1.01–2.09) for the sedentary level and 0.88 (95% CI: 0.64–1.19) for the low level. For men, the corresponding hazard ratios were 1.64 (95% CI: 1.29–2.10) and 1.15 (95% CI: 0.93–1.42).

View this table:
Table 2.

Hazard ratios (95% confidence interval) for alcohol use disorders according to leisure-time physical activity

Leisure-time physical activity
Hazard ratio (95% CI)
N (cases)Moderate/highLowSedentary
Women
 Adjusted for age9974 (248)1.000.91 (0.68–1.23)1.77 (1.25–2.51)
 Fully adjusteda9670 (236)1.000.88 (0.64–1.19)1.45 (1.01–2.09)
 Adjusted for social variables9809 (237)1.000.89 (0.64–1.18)1.62 (1.13–2.33)
 Adjusted for lifestyle variables9828 (247)1.000.93 (0.69–1.25)1.65 (1.16–2.35)
 Backwards model selectionb9671 (236)1.000.88 (0.65–1.19)1.45 (1.01–2.09)
Men
 Adjusted for age8385 (488)1.001.24 (1.00–1.52)2.23 (1.75–2.83)
 Fully adjusteda8274 (484)1.001.15 (0.93–1.42)1.64 (1.29–2.10)
 Adjusted for social variables8349 (485)1.001.22 (0.99–1.50)1.99 (1.56–2.53)
 Adjusted for lifestyle variables8307 (487)1.001.19 (0.96–1.46)1.77 (1.39–2.26)
 Backwards model selectionc8275 (484)1.001.15 (0.93–1.42)1.65 (1.30–2.11)
  • Cox proportional hazards model with delayed entry stratified by gender and adjusted for confounders as noted in the footnote of the table. Time-dependent covariates i.e. explanatory variables with updated information were included.

  • aAdjusted for age, smoking habits, alcohol intake, education, income and cohabitation status.

  • bAdjusted for age, smoking habits, alcohol intake, education and income.

  • cAdjusted for age, smoking habits, alcohol intake, income and cohabitation status.

Using sedentary level of leisure-time physical activity as reference group we reached similar results. In women the hazard ratio from the fully adjusted model was 0.60 (95% CI: 0.44–0.84) for the low level and 0.69 (95% CI: 0.48–0.99) for the moderate/high level. For men, the corresponding hazard ratios were 0.70 (95% CI: 0.56–0.88) and 0.61 (95% CI: 0.48–0.78). In the model only adjusted for age women had hazard ratios 0.52 (95% CI: 0.38–0.79) and 0.57 (95% CI: 0.40–0.80) for low level and moderate/high levels respectively while men had hazard ratios of 0.56 (95% CI: 0.45–0.70) and 0.45 (95% CI: 0.35–0.57).

Adjustment for work-time physical activity did not change results markedly (data not shown).

Work-time physical activity and risk of AUD

No evidence of an association between work-time physical activity and AUD was found, neither for men nor women (data not shown).

Analyses including social network and children

When including only data from the third examination of the CCHS, 9631 subjects were eligible for analyses. Of these, 77 women and 145 men were diagnosed with AUD. In unadjusted analyses including only data from the third examination, the same tendency as previously was observed; i.e. a sedentary level of leisure-time physical activity was associated with higher risk of AUD than a moderate/high level for both men and women (Table 3). All results for both men and women became insignificant when further adjusting for social network and children living at home.

View this table:
Table 3.

Hazard ratios for alcohol use disorders according to leisure-time physical activity, data from the 3rd examination only

Leisure-time physical activity
Hazard ratio (95% CI)
N (cases)Moderate/highLowSedentary
Women
 Adjusted for age5490 (77)1.000.66 (0.39–1.12)2.31 (1.28–4.18)
 Fully adjusteda5298 (73)1.000.66 (0.38–1.13)1.97 (1.05–3.71)
 Fully adjusted + network and children5066 (73)1.000.64 (0.38–1.11)1.81 (0.96–3.44)
 Adjusted for network and children5243 (77)1.000.66 (0.39–1.13)2.16 (1.18–3.95)
 Backwards model selectionb5242 (74)1.000.66 (0.38–1.12)1.83 (0.98–3.43)
Men
 Adjusted for age4141 (145)1.001.52 (1.05–2.19)1.90 (1.14–3.16)
 Fully adjusteda4048 (140)1.001.36 (0.94–1.98)1.56 (0.92–2.62)
 Fully adjusted + network and children3905 (137)1.001.39 (0.96–2.02)1.40 (0.81–2.44)
 Adjusted for network and children3988 (142)1.001.56 (1.08–2.25)1.69 (0.98–2.89)
 Backwards model selectionc4073 (140)1.001.41 (0.97–2.05)1.63 (0.97–2.73)
  • Cox proportional hazards model with delayed entry stratified by gender and adjusted for confounders as noted in the footnote of the table.

  • aAdjusted for age, smoking habits, alcohol intake, education, income and cohabitation status.

  • bAdjusted for age, smoking habits, alcohol intake and social network.

  • cAdjusted for age, alcohol intake, income and cohabitation status.

Stratification by level of alcohol intake

Table 4 shows that the effect within drinking groups (i.e. above/below safe limits for men and women respectively) of leisure-time physical activity seems to be the same on risk of developing AUD: hence, a sedentary level of leisure-time physical activity is associated with a higher risk of AUD in both alcohol groups (Table 4).

View this table:
Table 4.

Hazard ratios for AUDs according to leisure-time physical activity, stratified by level of alcohol intake

Leisure-time physical activity
Hazard ratio (95% CI)
N (cases)Moderate/highLowSedentary
Women
 ≤14 drinks per week
  Fully adjusteda8879 (165)1.000.81 (0.57–1.16)1.34 (0.86–2.07)
 >14 drinks per week
  Fully adjusteda782 (70)1.001.09 (0.61–1.96)1.80 (0.93–3.48)
Men
 ≤21 drinks per week
  Fully adjusteda6255 (215)1.001.32 (0.97–1.80)1.88 (1.30–2.73)
 >21 drinks per week
  Fully adjusteda2007 (266)1.001.03 (0.77–1.37)1.63 (1.18–2.25)
  • Cox proportional hazards model with delayed entry stratified by gender and adjusted for confounders as noted in the footnote of the table. Time-dependent covariates i.e. explanatory variables with updated information were included.

  • aAdjusted for age, smoking habits, education, income and cohabitation status.

Stratification by presence of other psychiatric disorders

When stratifying by presence of other psychiatric disorders, no association between leisure-time physical activity and AUD was observed among women with other psychiatric disorders. The hazard ratio from the fully adjusted model was 1.06 (95% CI: 0.66–1.70) for the sedentary level. Among women without other psychiatric disorders as well as both groups of men, a statistically significant increase in risk was observed for the sedentary level of leisure-time physical activity (Table 5). Based on these results it therefore seems that presence of other psychiatric disorders was an effect modificator in women only.

View this table:
Table 5.

Hazard ratios for AUDs according to leisure-time physical activity, stratified by presence of other psychiatric disorders

Leisure-time physical activity
Hazard ratio (95% CI)
N (cases)Moderate/highLowSedentary
Women
 With other psychiatric disorders
  Adjusted for age1635 (154)1.000.85 (0.59–1.22)0.94 (0.60–1.49)
  Fully adjusteda1598 (146)1.000.87 (0.59–1.28)1.06 (0.66–1.70)
 Without other psychiatric disorders
  Adjusted for age8338 (93)1.000.95 (0.56–1.59)2.65 (1.52–4.64)
  Fully adjusteda8071 (89)1.000.96 (0.57–1.62)2.30 (1.29–4.11)
Men
 With other psychiatric disorders
  Adjusted for age1022 (208)1.001.05 (0.77–1.45)1.65 (1.16–2.36)
  Fully adjusteda1012 (207)1.000.99 (0.72–1.37)1.49 (1.03–2.14)
 Without other psychiatric disorders
  Adjusted for age7360 (277)1.001.26 (0.96–1.67)2.19 (1.59–3.02)
  Fully adjusteda7259 (274)1.001.19 (0.90–1.58)1.59 (1.14–2.21)
  • Cox proportional hazards model with delayed entry stratified by gender and adjusted for confounders as noted in the footnote of the table. Time-dependent covariates i.e. explanatory variables with updated information were included.

  • aAdjusted for age, smoking habits, alcohol intake, education, income and cohabitation status.

Stratification by presence of chronic disease

A total of 3.6% of subjects were diagnosed with a chronic disease before or up to three years after entry into the study. The hazard ratios for men with chronic disease as well as both men and women without chronic disease were similar to the hazard ratios observed when not stratifying for chronic disease, though the fully adjusted hazard ratio for the sedentary level is not statistically significant for women without chronic disease (HR: 1.33, 95% CI: 0.91–1.95) and especially not for men with chronic disease (HR: 1.31, 95% CI: 0.50–3.42). The hazard ratio for women with chronic disease is 24.7 (95% CI: 2.41–253) for the sedentary level in the fully adjusted model.

Sensitivity analyses

Introducing a time lag of 2 and 3 years does not change the results markedly. In all models, the hazard ratio for the sedentary group was still significant. For women, the unadjusted hazard ratios were 1.70 (95% CI: 1.19–2.44) and 1.74 (95% CI: 1.21–2.48) for the sedentary level of leisure-time physical activity in the two models and for men 2.04 (95% CI: 1.57–2.66) and 1.80 (95% CI: 1.37–2.36), respectively.

DISCUSSION

We found that a sedentary level of leisure-time physical activity was associated with an approximately 1.5- to2-fold increased risk of developing AUD, compared with a moderate to high level of physical activity. The increased risk of AUD associated with a sedentary level was present in unadjusted and fully adjusted analyses, as well as in analyses adjusted for only some confounders. If we used sedentary level of physical activity in leisure time as reference we found that that even low level leisure-time physical activity was associated with almost half the risk of developing AUD and we found no difference between the effect of low or moderate/high activity.

In both men and women, it was present in both strata of alcohol intake. Performing analyses using data from only the third examination, where social network and children living at home could be included, estimates did not change. Introducing a time lag did not change the tendencies either, indicating that reverse causality is not a problem in this study. Stratification on presence of other psychiatric disorders resulted in no association in women with other psychiatric disorders.

Until now, no other studies have investigated the association between physical activity and AUD. Four cross-sectional studies have, however, been conducted to evaluate the association between physical activity and alcohol intake, and all four found significant associations between these factors, though disagreement existed with regard to the direction of the association. Three studies found that a high alcohol intake was associated with a high level of physical activity (Barrett et al., 1995; Smothers and Bertolucci, 2001; French et al., 2009), whereas one study found that a high alcohol intake was associated with a lower level of physical activity (Liangpunsakul et al., 2010). In the first study of 2072 individuals (Barrett et al., 1995), Barrett et al. found that moderate and heavy drinkers who did not smoke, were more likely than abstainers to report engaging in regular leisure-time physical activity. Smothers and Bertolucci (2001), found that both moderate and heavy drinkers were more likely than abstainers to have a physically active lifestyle among 41,104 Americans. In the largest study comprising 230,856 Americans (French et al., 2009), a linear relationship between alcohol intake and mean time spent on physical activity was observed with a clear ‘dose-response’ association. For women, when compared to abstainers, light, moderate and heavy drinkers spent approximately 8, 14 and 27% more time per week on exercise, respectively. For men, the corresponding numbers were 2, 10 and 23%, respectively. In the last study (Liangpunsakul et al., 2010), 10,550 Americans aged 20 years or more were included, and those with a high alcohol intake were significantly less active than non-drinkers or moderate social drinkers. The above-mentioned cross-sectional studies can, however, not be directly compared to the present prospective study as they may very well reflect consequences of a high alcohol intake rather than risk factors of developing this high intake. Until now a physiologic explanation for the association between leisure-time physical activity and AUD has not been established, but in cross-sectional studies increased physical activity has been associated with lower prevalence of anxiety and depression disorders (Weinstock, 2010). This could contribute to the effect on risk of developing AUD as depression and anxiety disorders often co-occurring with AUD. Furthermore physical exercises seem to reduce craving, withdrawal symptoms and relapse rates in AUD patients. Hence, there is overlap between the neurobiological mechanisms causing exercised-induced rewards and drug- and alcohol-induced rewards and this may explain some of the reduction for example craving (Olsen, 2011).

The interpretation of the results independent of choice of reference group, and for all analyses the conclusion would still be that there is no statistically significant difference between low level and moderate/high level leisure-time PA and sedentary life style result in an increased risk of AUD. The choice of reference group comes down to which contrast you want to focus on and our focus was on the impact of sedentary lifestyle, we chose moderate/high leisure-time PA as reference group.

Strengths and limitations

The present study was based on a large population followed prospectively for more than 20 years. This is generally recognized as a strong study design. However, some misclassification of exposure and/or outcome in a cohort study is inevitable as well as those with unhealthy life style will be underrepresented in the cohort. It is likely that participants may have overestimated their level of physical activity when completing the questionnaires, as is often seen when using self-reported data (Rzewnicki et al., 2003). However, due to the prospective design of the study, participants were unaware of whether or not they would later be diagnosed with AUD, and recall bias is therefore not a problem and the misclassification most likely to be non-differential. The data are representative of the Copenhagen capital but there is no reason why the results on the association between leisure-time physical activity and AUD cannot be generalized to the whole country. The decreasing response rate over the years may have tended to select the more healthy group of the general population favoring those with better life style, but again this may not necessarily affect the association between leisure-time physical activity and risk of AUD. The questionnaire based measure of leisure-time physical activity has previously been shown to be inversely associated with mortality (Schnohr, 2009) as well as the measure is associated with self-rated health and muscle strength (Hansen et al., 2013).

Early symptoms of AUD present at baseline (previous to receiving the diagnosis) could have caused subjects to lower their level of leisure-time physical activity. This would have been a differential misclassification, but was probably not the case, as the association did not change when introducing time lags of two or three years.

We have no information in any of the registers used in this study on individuals with AUD who do not seek medical help or are only seen by their general practitioner. Therefore the number of individuals with AUD is underestimated. The misclassification of AUD will be differential if admittance to a hospital or referral to the outpatient clinic is related to the patients’ level of physical activity, and this cannot be ruled out. Furthermore, level of physical activity—both leisure time and work time—can affect the risk of a range of other diseases, and thereby also the risk of hospitalization for other reasons than AUD. In this case, the AUD can be registered as a secondary diagnosis. Another source of misclassification of the outcome comes from people being diagnosed with AUD without actually having it, but this is probably a minor problem. The prevalence of AUD may have changed over the 27-year recruiting period but this not likely to affect the association between leisure-time physical activity and AUD.

Selection bias have occurred as non-participation was associated with a higher risk of AUD. However, the level of physical activity among non-participants is unknown, and therefore the size of selection bias cannot be determined. Participants were followed in national registers, and less than one percent of participants were lost to follow-up in the CCHS. Loss to follow-up is therefore only a minor cause of selection bias.

Risk of residual confounding or confounding due to unmeasured covariates will always be possible. Information on psychological factors or personality traits would have been relevant to include in the models as several studies have shown an association between these and physical activity or the risk of AUD. For example, conscientiousness was associated with both a higher level of physical activity (Raynor and Levine, 2009) as well as with a lower risk of AUD (Martin and Sher, 1994; Hopwood et al., 2007). Awareness of health and low self-esteem are other potential confounders that were not included in this study (Fox, 1999; Nolen-Hoeksema, 2004; Biddle and Asare, 2011). Another important potential confounder is AUD in one or both parents as genetic factors have been found to influence the risk of AUD (Mayfield et al., 2008). However, information on these factors was not available. Residual confounding can be a problem with regard to all included variables, but it is especially likely that the rough measures of income and education could lead to residual confounding.

In this study, a sedentary life style in leisure time was associated with a 1.5- to 2-fold increase in risk of AUD i.e. mainly alcohol dependency in patients admitted to alcohol outpatient clinics and hospitals compared to a moderate/high level of leisure-time physical activity. In analyses stratified by gender only, this was true for both men and women. A low level of leisure-time physical activity was not associated with a statistically significant increase in risk, though in men there seemed to be a tendency that a low level of leisure-time physical activity could be associated with a small increase in risk. Regarding work-time physical activity, no association with risk of AUD was found. A causal link between physical activity and AUD is possible but has not been proven yet although physical activity have been shown to reduce the risk of other psychiatric disorders, and since physical activity may be effective in the treatment of AUD.

This is one of the first studies to suggest a positive effect of leisure-time physical activity in the prevention of AUD, and may strengthen the general recommendations of increased leisure-time physical activity. The data warrants further research into this association.

Conflict of interest statement

None of the Authors have any connection to tobacco, alcohol, pharmaceutical or gaming industries, nor has the present work been funded by any of these organizations. None of the authors have any financial conflict of interest with organizations that seek to provide help with or promote recovery from addiction.

REFERENCES

View Abstract