Individual associations of alcohol consumption and physical activity with cardiovascular disease are relatively established, but the joint associations are not clear. Therefore, the aim of this study was to examine prospectively the joint associations between alcohol consumption and physical activity with cardiovascular mortality (CVM) and all-cause mortality. Four population-based studies in the United Kingdom were included, the 1997 and 1998 Health Surveys for England and the 1998 and 2003 Scottish Health Surveys. In men and women, respectively, low physical activity was defined as 0.1 to 5 and 0.1 to 4 MET-hours/week and high physical activity as ≥5 and ≥4 MET-hours/week. Moderate or moderately high alcohol intake was defined as >0 to 35 and >0 to 21 units/week and high levels of alcohol intake as >35 and >21 units/week. In total, there were 17,410 adults without prevalent cardiovascular diseases and complete data on alcohol and physical activity (43% men, median age 55 years). During a median follow-up period of 9.7 years, 2,204 adults (12.7%) died, 638 (3.7%) with CVM. Cox proportional-hazards models were adjusted for potential confounders such as marital status, social class, education, ethnicity, and longstanding illness. In the joint associations analysis, low activity combined with high levels of alcohol (CVM: hazard ratio [HR] 1.95, 95% confidence interval [CI] 1.28 to 2.96, p = 0.002; all-cause mortality: HR 1.64, 95% CI 1.32 to 2.03, p <0.001) and low activity combined with no alcohol (CVM: HR 1.93, 95% CI 1.35 to 2.76, p <0.001; all-cause mortality: HR 1.50, 95% CI 1.24 to 1.81, p <0.001) were linked to the highest risk, compared with moderate drinking and higher levels of physical activity. Within each given alcohol group, low activity was linked to increased CVM risk (e.g., HR 1.48, 95% CI 1.08 to 2.03, p = 0.014, for the moderate drinking group), but in the presence of high physical activity, high alcohol intake was not linked to increased CVM risk (HR 1.32, 95% CI 0.52 to 3.34, p = 0.555). In conclusion, high levels of drinking and low physical activity appear to increase the risk for cardiovascular and all-cause mortality, although these data suggest that physical activity levels are more important.
Several meta-analyses suggest a J-shaped association of alcohol consumption with all-cause and cardiovascular mortality (CVM) in older adults. The largest of these meta-analyses showed a decreased CVM risk of 14–25% with light to moderate alcohol consumption of 2.5–14.9 g/day (about ≤2 drink a day) compared with abstaining from alcohol. The inverse association between physical activity and cardiovascular disease is well established. A recent meta-analysis showed a 19% decreased risk in the relation between physical activity and all-cause mortality. A 50% reduction in all-cause mortality risk was found in older adults who met the recommended level of ≥30 minutes of moderate exercise 5 days/week compared with adults without any reported domestic physical activity. Although individual associations of alcohol consumption and physical activity on all-cause mortality and CVM are relatively established, the joint association has not been fully examined. One of the few studies investigating this joint association in fatal ischemic heart disease and all-cause mortality showed that physical activity and moderate alcohol consumption are important to decrease the risk for ischemic heart disease and all-cause mortality in comparison with inactive nondrinkers. Given the scarcity of information on the interactions of these 2 key health behaviors, the low prevalence of physical activity, and the increasing trend in alcohol consumption in the United Kingdom, we prospectively examined the joint associations of alcohol consumption and physical activity with all-cause mortality to determine whether engagement in physical activity modifies the association between alcohol and CVM.
Methods
We used data from 4 population-based studies (the 1997 and 1998 Health Surveys for England [HSE] and the 1998 and 2003 Scottish Health Surveys [SHS]), which were linked prospectively to cause-specific mortality records. Data collection was household based, and methodologies were almost identical across all 4 cohorts. Each baseline survey data collection featured nationally representative samples of adults living in households in England and Scotland. Samples were drawn using multistage stratified probability sampling with postal code sectors selected at the first stage and household addresses selected at the second stage. Informed consent was collected from all participants, and ethical approval was obtained from the North Thames Multicentre Research Ethics Committee for England, the Local Research Ethics Council in England, the Research Ethics Committee for all Area Health Boards in Scotland, and the Multicentre Research Ethics Committee for Scotland. This research was conducted in accordance with the Declaration of Helsinki. The surveys were linked to the patient-based database of deaths up to 2008 (HSE: February 15, 2008; SHS: December 31, 2008). Diagnoses for primary causes of death were recorded according to the International Classification of Diseases, Ninth Revision and Tenth Revision, with CVM codes 390 to 459 from the Ninth Revision and I01 to I99 from the Tenth Revision.
Questionnaire-based data were collected by trained interviewers at baseline (i.e., 1997, 1998, and 2003). Adults were asked about their drinking frequency during the past 12 months and the amount of units of different types of beverages (i.e., beer, cider, shandy, spirits, liqueurs, sherry, martinis, wine, champagne, and alcoholic soft drinks [“alcopops”]) consumed on a usual drinking occasion. One unit of alcohol is defined as 8 g ethanol and is the equivalent of a 125 ml glass of wine or half a pint of regular strength beer. Total weekly units were calculated by summing the units of each type of beverage multiplied by the frequency per week. Extreme values for alcohol consumption (i.e., on the basis of blood alcohol concentrations at postmortem in fatal cases of alcohol intoxication: weekly units equivalent to 355 mg% blood alcohol concentration per day with reference weights for men of 70 kg and for women of 60 kg ) were excluded.
Four alcohol consumption categories were constructed on the basis of the previous English Department of Health’s suggested weekly limits (i.e., 21 weekly units in men, 14 weekly units in women). These categories were none; moderate, defined as >0 to 21 weekly units for men and >0 to 14 weekly units for women; moderately high, defined as >21 to 35 weekly units for men and >14 to 21 weekly units for women; and high, defined as >35 weekly units for men and >21 weekly units for women. To examine joint associations with physical activity, the moderate and moderately high alcohol categories were merged into 1 moderate alcohol consumption category.
Physical activity interviews collected information on frequency, time spent, and intensity during the past 4 weeks of walking and 10 different sports (swimming, cycling, workout at a gym/exercise bike/weight training, aerobics/gymnastics, dancing, running/jogging, football/rugby, badminton/tennis, squash, and exercises such as press-ups and sit-ups). Domestic activity, including heavy housework, heavy do-it-yourself activities, and manual and gardening work, was also assessed. Physical activity was expressed as MET-hours per week. MET-hours per week for sports and walking were derived by multiplying time per occasion, frequency per week, and the assigned MET value that corresponds to the intensity of the activity (on the basis of the nature of the activity and the answer to the question of whether participants were out of breath or sweaty) or walking (i.e., slow, steady average, fairly brisk, and fast pace). MET-hours per week for the different sports and walking were summed to calculate the total MET-hours per week. MET-hours per week for domestic activity were derived in a similar way as for sports and walking. Extreme values of physical activity (on the basis of ≥5 SDs from the mean) were excluded. Three categories were derived on the basis of the mean MET-hours per week (i.e., 5 in men and 4 in women) combining sports and walking: no physical activity; low physical activity, defined as 0.1 to 5 MET-hours/week for men and 0.1 to 4 MET-hours/week for women; and high physical activity, defined as >5 MET-hours/week for men and >4 MET-hours/week for women. Although median MET-hours per week, excluding adults with no physical activity, provided roughly equally sized groups, using the mean as the cutoff was more comparable with the minimum recommended physical activity level of 7.5 MET-hours/week, and gender differences were captured.
Six joint categories were defined, combining nondrinking and low physical activity, moderate drinking and low physical activity, high drinking and low physical activity, nondrinking and high physical activity, moderate drinking and high physical activity, and high drinking and high physical activity.
In all analyses, we excluded prevalent cardiovascular diseases to minimize the possibility that changes in alcohol consumption and physical activity after the diagnosis would obscure the true association between physical activity and risk for CVM. We also excluded those aged <40 years, those who stopped drinking for health reasons, and those who reported doctor-diagnosed diabetes.
Information on potential confounders was assessed using questionnaire-based interviews, except for weight and height. Weight and height were measured by the interviewers using a Soehnle, Seca, or Tanita electric bathroom scale and a portable stadiometer. Body mass index was calculated as weight divided by the square of height.
Characteristics by alcohol consumption and physical activity categories are presented as medians and interquartile ranges (IQRs) for continuous variables and percentages for categorical variables. The included sample in the analyses was compared with the excluded sample in terms of key characteristics. Differences in continuous variables were tested using Kruskal-Wallis tests (>2 categories) or Mann-Whitney tests (2 categories) and in categorical variables using chi-square tests.
Cox proportional-hazard models were used to examine the individual associations between (1) alcohol consumption (reference category: moderate drinking), and (2) physical activity (reference category: no physical activity) and all-cause mortality and CVM. Interaction terms of alcohol consumption (after excluding nondrinkers) by physical activity were added to the model to evaluate the joint association of these on CVM or all-cause mortality. In a sensitivity analysis, we entered combined categories as exposure in the models (reference category: combined moderate and moderately high drinking and high physical activity). The proportional-hazards assumptions were examined using a log-minus-log plot. Basic Cox models included adjustments for age (continuous) and gender. Further models were adjusted for survey year, marital status (single, married, separated, divorced, or widowed), registrar general’s social class (professional/managerial technical, skilled nonmanual, skilled manual, semiskilled/unskilled manual, or other), education (university or no university [on the basis of age finished full-time education]), ethnicity (white, black/South-Asian/other), cigarette smoking (never, past occasionally, past regularly, or current), self-reported longstanding illness (yes or no) and body mass index (<20.0, 20.0 to 24.9, 25.0 to 29.9, or ≥30.0 kg/m 2 ). Additionally, in the alcohol analyses, we also adjusted for physical activity (continuous) and current alcohol consumption compared with 5 years ago (less, similar, or more). In the physical activity analyses, we also adjusted for alcohol consumption (continuous) and domestic activity (continuous). Domestic activity was not included as part of the exposure variable, because in previous studies, it was not associated with cardiovascular risk factors and CVM. Only current alcohol drinkers were included in the analyses to test for a linear trend, because of the nonlinear association between alcohol and mortality. All statistical analyses were carried out using SPSS version 18.0 (SPSS, Inc., Chicago, Illinois).
Results
In the four surveys, data of 40,220 respondents were collected. After exclusion of those below the age of 40 years (n = 15,666), we had 24,554 adults left. Those with diagnosed diabetes mellitus (n = 1,126) and those who stopped drinking due to health conditions (n = 500) were excluded from analyses. Among the remaining 22,928 adults, 99.7% (n = 22,720) completed the alcohol consumption questionnaire. The sample included 20,005 participants after exclusion of adults with missing values for the variables of mortality (n = 2,478 not consented) and social class (n = 237). Prevalent cardiovascular diseases (coronary heart disease, stroke, angina; n = 2,595) at baseline were excluded resulting in a final sample of 17,410 participants.
Baseline key characteristics by categories of alcohol consumption are listed in Table 1 (and by categories of physical activity in Supplementary Table 1 ). Age ranged from 40 to 95 years (median 55), and 43% were men. Of the 17,410 adults at baseline, 1,624 (9%) reported no alcohol intake. Of the nondrinkers, 37% (n = 605) had stopped drinking within the past 5 years. In total, 3,574 adults (21%) exceeded the recommended weekly limits for alcohol (21 weekly units in men, 14 weekly units in women). Maximum amounts of weekly units of 146.0 (IQR 3.0 to 24.0) for men and 98.0 (IQR 0.3 to 8.5) for women were found. In total 3,719 adults (21%) reported no physical activity. The recommended level of physical activity (7.5 MET-hours/week ) was met by 2,702 adults (16%). Maximum levels of 53.5 MET-hours/wk (IQR 0.2 to 3.9) for men and 52.8 MET-hours/week (IQR 0.2 to 3.0) for women were found in this study population.
| Variable | Alcohol Consumption ∗ | p Value | |||
|---|---|---|---|---|---|
| None (n = 1,624) | Moderate (n = 12,212) | Moderately High (n = 1,921) | High (n = 1,653) | ||
| Alcohol consumption (units/week) | N/A | 3.8 (0.8–9.0) | 22.8 (17.3–28.2) | 42.0 (31.5–56.0) | <0.001 § |
| Physical activity | |||||
| Sports and walking (MET-hours/week) | 0.2 (0.0–0.9) | 0.5 (0.2–3.5) | 0.7 (0.2–5.0) | 0.7 (0.2–4.3) | <0.001 § |
| Domestic activity (MET-hours/week) | 0.2 (0.0–1.9) | 0.5 (0.0–2.3) | 1.0 (0.0–2.8) | 0.7 (0.0–2.7) | <0.001 § |
| Age (yrs) | 64 (51–73) | 56 (47–66) | 53 (46–62) | 51 (46–60) | <0.001 § |
| White † | 91% | 98% | 99% | 99% | <0.001 ⋮ |
| Body mass index (kg/m 2 ) ‡ | 27.0 (23.9–30.4) | 26.7 (24.1–29.8) | 26.7 (24.3–29.3) | 26.5 (24.0–29.3) | 0.052 § |
| Education, university † | 8% | 14% | 18% | 17% | <0.001 ⋮ |
| Social class, professional/managerial technical † | 19% | 32% | 42% | 37% | <0.001 ⋮ |
| Marital status, single † | 56% | 68% | 72% | 68% | <0.001 ⋮ |
| Current cigarette smoking status † | 23% | 24% | 27% | 38% | <0.001 ⋮ |
| Reduced alcohol consumption | 37% | 43% | 35% | 32% | <0.001 ⋮ |
| Longstanding illness | 54% | 48% | 43% | 46% | <0.001 ⋮ |
∗ Four groups of alcohol consumption were defined: none; moderate, defined as >0 to 21 units/week for men and >0 to 14 units/week for women; moderately high, defined as >21 to 35 units/week for men and >14 to 21 units/week for women; and high, defined as >35 units/week for men and >21 units/week for women.
† Categories were as follows: for ethnicity, white, black/South-Asian/other; for education, university or no university; social class, professional/managerial technical, skilled nonmanual, skilled manual, semiskilled/unskilled manual, or other; for marital status, single, married, separated, divorced, or widowed; and for cigarette smoking status, never, past occasionally, past regularly, or current.
‡ Only adults with values for body mass index were included.
During a median follow-up period of 9.7 years (IQR 6.6 to 10.2, 148,502 total person-years), 2,204 adults (12.7%) died, of whom 638 (3.7%) had CVM. A J-shaped association was found between alcohol consumption and all-cause mortality ( Figure 1 ). Those reporting high levels of drinking had a 26% increased all-cause mortality risk compared with moderate drinkers ( Figure 1 ). Nondrinkers had a 14% increased all-cause mortality risk compared with moderate drinkers. A linear trend was found between alcohol consumption (after excluding nondrinkers) and all-cause mortality (p for trend = 0.003). For alcohol consumption and CVM, a J-shaped association was also seen ( Figure 1 ). Nondrinkers had 30% increased risk compared with moderate drinkers ( Figure 1 ). We found evidence of an inverse association between physical activity and all-cause mortality and CVM in the multivariate model ( Supplementary Table 2 ). In the association with CVM, inverse associations were found for low physical activity and for high physical activity compared with no physical activity. The strongest inverse association was found for high physical activity compared with no physical activity. A linear trend was found in the multivariate model for physical activity and all-cause mortality and CVM.
The joint associations of alcohol consumption and physical activity with all-cause mortality are listed in Table 2 and shown in Supplementary Figure 1 . The results suggest evidence of joint associations with all-cause mortality. High levels of drinking in combination with low physical activity had the highest all-cause mortality risk (i.e., 64% increased risk) in comparison with the reference category (moderate to moderately high drinking and high physical activity) ( Table 2 ). Moderate to moderately high drinking and low physical activity was associated with a 35% increased all-cause mortality risk compared with the reference group. The joint associations of alcohol consumption and physical activity with CVM are listed in Table 3 and shown in Supplementary Figure 1 . In the joint associations analysis, low physical activity combined with high levels of alcohol (CVM: hazard ratio [HR] 1.95; all-cause mortality: HR 1.64) and low physical activity combined with no alcohol (CVM: HR 1.93, all-cause mortality: HR 1.50) was linked to the highest risk, compared with moderate or moderately high drinking and higher levels of physical activity. Within each given alcohol group, low physical activity was linked to increased CVM risk (e.g., HR 1.48, 95% confidence interval 1.08 to 2.03, p = 0.014, for the moderate or moderately high drinking group), but in the presence of high physical activity, high levels of drinking were not linked to increased CVM risk (HR 1.32, 95% confidence interval 0.52 to 3.34, p = 0.555) ( Table 3 , Supplementary Figure 1 ).
