Twenty-year trajectories of alcohol consumption during midlife and atherosclerotic thickening in early old age: findings from two British population cohort studies

Alcohol consumption is one of the biggest public health challenges facing modern society and is ranked as the world’s third largest risk factor for disease burden [1]. There are also currently concerns about the long-term ill health effects of the drinking habits of middle-aged adults, a group who have been described as ‘hidden’ risky drinkers [2].

Epidemiological evidence indicates a protective effect of light-moderate alcohol consumption on cardiovascular morbidity and mortality compared to non-drinking and an increased risk of cardiovascular events amongst heavier drinkers [3‐5]. The endpoint of cardiovascular disease (CVD) often occurs later in life; the extent to which alcohol consumption earlier in the life course is implicated in the process is less clear. As yet, the biological mechanisms underlying the association between alcohol intake and CVD are not fully elucidated, and it has been argued that the protective effect observed in observational studies may, in part, be due to misclassification errors and reporting biases [6]. If the cardiovascular benefits of moderate alcohol consumption are genuine, they may be explained in part by alcohol’s influence on serum lipids [7] and other cardiovascular biomarkers [8], blood pressure, and atherosclerosis [9]. Nevertheless, the effect of alcohol on atherosclerotic changes in vessel walls is disputed. Carotid intima media thickness (cIMT) is an intermediate phenotype of early atherosclerosis or a marker of subclinical organ damage that independently predicts vascular events [10]. Some studies have found evidence of a reduced risk of atherosclerosis (as measured by cIMT) in moderate drinkers [11‐17], while others report no association [18‐21], or an increased risk [22‐24]. Most of the previous studies have only looked at the cross-sectional association between alcohol intake and cIMT. Single observations of alcohol consumption assume that drinking is a stable behaviour. However, there is evidence from descriptive studies that individuals change their alcohol consumption levels over time [25, 26]. Research on the health consequences of alcohol, therefore, needs to address the effects of changes in drinking behaviour over the life course [27]. Failure to include such dynamics can lead to incorrect inferences about the effects of alcohol on chronic disease risk [28].

In this paper we sought to (1) describe trajectories of alcohol consumption during midlife, (2) link these trajectories to risk of atherosclerosis, as measured by cIMT in early old-age, and (3) compare these associations with cross-sectional findings in the same cohorts.

Data were drawn from two UK population-based cohorts: the Whitehall II Cohort of British Civil Servants and the MRC National Survey of Health and Development (NSHD). In both cohorts, alcohol consumption was self-reported at four time points over a 20-year period and cIMT was measured when participants were in early old-age (50–74 years).

The most common derived alcohol trajectories were “stable moderate” (39.8 % Whitehall, 34.4 % NSHD) and “mostly moderate” (24.9 % Whitehall, 31.3 % NSHD) (Table 1). Smaller groups were “stable none” (5.3 % Whitehall II, 4.5 % NSHD) and “stable heavy” (6.8 % Whitehall II, 3.5 % NSHD). “Stable heavy” drinkers were more likely to be male, current smokers and of white ethnicity (in Whitehall II) (Table 1). “Stable none” drinkers were more likely to be female, non-smokers, of lower SEP and non-white. The median cIMT level measured in the overall Whitehall participant cohort was 0.77 mm (IQR 0.68–0.87), and 0.77 mm (IQR 0.68–0.88) for men and 0.77 mm (IQR 0.68–0.85) for women. The median lateral view cIMT was 0.67 mm (IQR 0.60–0.76) in the NSHD overall cohort, and 0.68 mm (IQR 0.61–0.79) for men and 0.66 mm (IQR 0.59–0.73) for women.

Cross-sectional analyses showed slightly elevated cIMT values amongst non-drinkers (pooled β = 0.012 mm, 95 % CI –0.001 to 0.026) compared to moderate drinkers (Fig. 1). When non-drinkers were separated into former drinkers and never drinkers, the cIMT was higher in former drinkers than never drinkers (available in Additional file 1: Figure S1). In cross-sectional analyses there was little difference in atherosclerotic thickening between heavy and moderate drinkers (pooled β = 0.006 mm, 95 % CI –0.005 to 0.017) (Fig. 1).

The association between trajectories of drinking during midlife and cIMT are shown in Fig. 2. Full results from regression analyses are shown in Additional file 2: Table S1 and S2. “Stable heavy” drinkers over the 20-year measurement period had elevated cIMT compared to “stable moderate” drinkers (pooled difference of 0.021 mm, 95 % CI 0.002 to 0.039), after adjustment for covariates. There were no appreciable differences between “stable non-drinkers” and “stable moderate” drinkers (pooled β = 0.008 mm, 95 % CI –0.003 to 0.018). “Former drinkers” had raised cIMT (pooled β = 0.021 mm, 95 % CI 0.005 to 0.037). Mostly moderate and mostly heavy were similar. Excluding those with prevalent CHD/diabetes in the Whitehall II cohort did not substantially alter the estimates (Additional file 1: Figure S2).

We derived 20-year drinking trajectories during midlife in two separate British population-based cohort studies and linked these to cIMT in early old age. We found evidence to suggest that former drinkers and those with sustained heavy drinking during midlife had greater cIMT values than stable moderate drinkers. We also observed that stable moderate drinkers did not have reduced cIMT values compared to stable non-drinkers. Combined, these findings indicate that midlife drinking habits affect the atherosclerotic process.

The risks associated with heavy drinking in midlife were only appreciable when considering a 20-year trajectory of drinking and not in cross-sectional analyses. This highlights the need to take a life course approach and our study should be replicated with other outcomes. Our data start from around age 35 years. Others have suggested that an adverse effect on atherogenesis from drinking may occur even earlier in life. In a cohort of young Finns, aged 24–39 years, there was a dose–response relationship between alcohol consumption and increased cIMT [23].

However, there is a risk of misclassification if only one measure of alcohol is considered. For example, we found that, when current non-drinkers were separated into former drinkers and never drinkers, the median cIMT levels were substantially higher in the former drinkers. Former drinkers may include individuals who experienced ill-health (including increased vascular problems as indicated by higher cIMT) and subsequently quit. This concurs with the well-known ‘sick-quitter’ phenomenon whereby former drinkers who become ill and cease drinking are classified as non-drinkers and this can erroneously lead to suggested protective effects of drinking compared to not drinking [35]. Alcohol consumption is not a stable phenomenon over the life course. Exposure varies, as shown in this paper by the large proportion in both cohorts who did not have stable drinking trajectories over the 20 years, and, elsewhere, we have shown how mean consumption and frequency changes in these cohorts and others [26].

Much of the existing literature uses data with alcohol exposure measured at a single time-point and takes a mixed group of “non-drinkers” as the reference group. Ours is the first study to look at trajectories of alcohol and risk of cIMT, which makes drawing comparisons between our work and others difficult. Our cross-sectional findings concur with those from the Atherosclerosis Risk in Communities study and the National Heart, Lung, and Blood Institute Family Heart Study, which indicate no association between current alcohol intake and carotid artery wall thickness [18, 19]. Likewise, data from over 6000 men and women in three French cities showed no marked relationship of alcohol and cIMT [20]. Others have found a U-shaped relationship between alcohol and cIMT in cross-sectional analyses. In the Cardiovascular Health Study investigating subjects over 65 years, consumers of 1–6 drinks per week (equalling < 15 g/d) had a cIMT 0.07 mm lower than abstainers, whereas consumers of 14 or more drinks (equalling > 30 g/d) had an IMT 0.07 mm higher than abstainers. Like in the present study, they found that former drinkers had an increased cIMT [16].

Our trajectory work suggests that clinicians need to place emphasis on drinking histories as well as current drinking behaviour. Attempting to quantify the effect of heavy drinking on atherosclerosis is challenging. Elsewhere, it has been shown, using data from the Whitehall II study, that the mean progression rate of cIMT is estimated to be 0.012 ± 0.028 mm per year [36]. Therefore, the effects of consistent heavy drinking roughly equate to 19 months expedited vascular ageing.

Measuring alcohol use over time is likely to be more aetiologically relevant than a snap shot of alcohol at one point in time. For example, it is reasonable to hypothesise that sustained high levels of drinking increase the formulation of atheromas on the vessel wall over time [23, 24].

A major strength of this study is our ability to use repeated measures of alcohol consumption on the same individuals over two decades before the measurement of cIMT. The derived trajectories have policy relevance as they are defined by UK government guidelines [33]. Treating cIMT as a surrogate endpoint allowed for us to investigate midlife drinking as a risk factor for early atherosclerosis. This enabled us to determine how drinking during this period might set the stage for the development of CVD, before it is necessarily symptomatic.

Our study is limited in that alcohol consumption was self-reported and therefore at risk of over and under-reporting [37, 38]. We were only able to capture snap shots of drinking over the past week/5 days and have assumed that these are a general representation of levels consumed over that period. This may introduce error, but we utilised the repetition of these snap shops to more accurately create trajectories than simple use of a baseline measure. Our sample is made up of individuals who have remained in an epidemiological study for decades. This is a form of selection bias [39] and may mean that our sample is no longer representative of the original population from which it was drawn. We found those remaining in the studies and attending clinical research facilities to be a healthier subsample than those who dropped out or who did not participate fully. Furthermore, population-based studies do not capture the extremes of drinking and therefore may be underpowered to look at the effects of very heavy drinking. At the time of cIMT measurement, the proportion drinking in excess of guidelines was 19.3 % in Whitehall II (21.9 % men and 12.8 % women) and 13.0 % in NSHD (17.8 % men and 7.9 % women), which is considerably lower than recent estimates from Health Survey for England (31 % men and 20 % women aged 55–64 years) [40]. It may be that NSHD, in particular, was underpowered to detect effects among sustained heavy drinkers (who constituted only 3.5 % of the sample). Furthermore, we were unable to assess the effects of binge drinking as these data were not adequately captured in both studies at that time. Although we included several covariates in the analysis, as with most observational studies, we cannot rule out that possibility of residual confounding.

Future work should consider the progression of cIMT [18] into older age and whether changes in atherosclerosis, and ultimately CHD cases, are affected by changes in drinking levels. Furthermore, it would be interesting to extend the work on pattern of consumption [24] to see whether sustained ‘binge’ drinking confers an increased risk of higher cIMT compared to regular heavy drinking [41], as is found for CHD overall [42].

These findings indicate that the drinking habits adopted by adults during midlife affect the atherosclerotic process, and that sustained heavy drinking is associated with an increased risk of poor cardiovascular health compared to stable moderate drinkers. This finding was not seen when using cross-sectional analyses only, thus highlighting the importance of taking a longitudinal approach. Furthermore, there was no evidence of a favourable atherosclerotic profile among stable moderate drinkers compared to non-drinkers.

cIMT, carotid intima media thickness; CVD, cardiovascular disease; NSHD, National Survey of Health and Development; SEP, socioeconomic position