Both in the UK and internationally, similar alcohol consumption levels have been associated with greater impacts on the health of more deprived individuals. In a systematic review of socio-economic differences in alcohol-attributable mortality Probst et al. suggest that the poorer diet of individuals living in deprivation (e.g. more high fat and salt foods and less fruit and vegetable consumption) may interact with alcohol consumption to alter protein and vitamin absorption and increase risks of health harms [31]. They also acknowledge that interactions between higher smoking prevalence in deprived areas and alcohol consumption may contribute to an increased risk of some cancers. Moreover, they suggest that poorer access to primary care may contribute to greater harms from alcohol in deprived groups. Consequently, individuals with lower socio-economic status may face cost, transport, availability and stigma-related issues that restrict their access to services which might help with alcohol- related problems [31–33]. Along with an increasing literature describing the disproportionate impact of alcohol on more deprived communities advocacy to address such health inequalities is also developing at national and international levels [34]. Generally, however, alcohol consumption surveys typically do not collect sufficient current and historic data to test competing explanations for this alcohol harm paradox. Consequently, using a bespoke national survey we have examined how combined health challenges, current drinking patterns and historical drinking behaviours differed with deprivation between individuals who currently consume similar quantities of alcohol.

Results confirm strong associations between drinking and smoking (Table 2) and specifically identify a disproportionate concentration of smokers in increased and higher risk alcohol consumers from deprived communities (Table 3). In addition such deprived individuals are more likely to be overweight and have unhealthier lifestyles. Consideration of alcohol-related health harms often focuses on higher risk drinkers and results here suggest such individuals (males >50 units or 400 g of pure alcohol/week; females >35 units or 280 g of pure alcohol/week) in deprived communities face combined health challenges likely to have a multiplicative impact on health. Such impacts include increased risks of conditions specifically associated with alcohol (e.g. alcohol-related liver disease) and those where alcohol is one of many multifactorial causes (e.g. oesophageal cancer, breast cancer, hypertension and macular degeneration) [3, 35–37]. Arguably, less attention is paid to increased risk drinkers (here males >21–50 units or >168–400 g of pure alcohol/week; females >15–35 units, >112–280 g of pure alcohol/week). However, across England 18 % of men and 13 % of women report drinking at increased risk levels (vs. 5 and 3 % at higher risk levels respectively) [29] and the contribution of increased and even lower risk drinkers to overall harms is substantive. For instance only around a fifth of alcohol-related breast cancer deaths are in women drinking ≥35 units (≥280 g of pure alcohol) a day with the rest in those drinking at lower consumption levels (England [38]). For many conditions multiplicative impacts are still relatively poorly defined. However, the combined risk from smoking with alcohol consumption may be more than double that expected from the summed risks from smoking and alcohol in the absence of synergies [39]. In this study increased risk drinkers in deprived communities were 10.9 times more likely to carry the additional burden of not just smoking but also unhealthy lifestyle and excess weight (Fig. 1).

As well as combined health challenges, results suggest that individuals in deprived groups may differ from those in non-deprived groups (with similar current total weekly alcohol consumption) in their choice of alcohol types and both current binge and historical binge drinking (Tables 2 and 3). Deprived drinkers are less likely to typically consume wine and more likely to consume beer or spirits (Table 3). Following extensive debate over the Mediterranean diet and relative benefits of wine consumption compared with other alcoholic drinks [40], more recent epidemiological analyses appear to offer some support for health challenges differing by drink type [41–44]. Further, we found individuals from deprived groups also appear to consume alcohol in fewer but heavier drinking sessions (Tables 2 and 3). Moreover, based on retrospective estimates of frequency of heavy drinking (at ages 18 and 30 years) deprived individuals are also more likely to have previously been drunk/bingeing, although differences from more affluent individuals are limited mainly to lower and increased risk drinkers (Tables 2 and 3). Consuming similar amounts of alcohol in fewer sessions increases risks of alcohol-related injuries (including unintentional and violent [35, 45]) and critically can also eradicate any potential protection moderate drinking might offer from IHD. Consequently, higher bingeing in deprived groups (both historically and currently) is consistent with these populations suffering more injury and IHD than more affluent drinkers despite current total alcohol consumption being the same. The long-term impact of a history of more frequent binge drinking is still poorly understood but any impact on life-time risks of IHD, cancers or other alcohol-related conditions is likely to be exacerbated by ex-drinkers (but not never drinkers) in deprived communities continuing to carry higher combined health challenges (smoking, poor diet and excess weight; Table 2) than ex-drinkers from more affluent groups.

There are a number of important limitations to this study. Response rate was 23.3 % and we cannot quantify any bias introduced by differences between individuals who agreed or declined to participate. We could not distinguish unoccupied properties from those where individuals chose not to answer their phones. Individuals who chose not to answer calls also represent a potential source of bias in the final sample which we cannot quantity. Typical response rates for telephone surveys are declining with one major US provider tracking falls from 28 % compliance in 2000 to rates well below those achieved here (i.e. around 9 %) in 2012 [46] Consequently, while response related bias remains a potential confounder, compliance here is well within the range experienced elsewhere. Inevitably alcohol harm paradox variables were proxy measures. Thus, our history of drunkenness/bingeing and of frequent drinking was limited to retrospective measures for ages 18 and 30 years. We cannot establish how well they correlate with all consumption over this 12 year period or any other period in respondents’ drinking histories and how any recall error may have impacted results. Accuracy of recall is also a potential issue for age at which regular drinking began and age when first drunk. Our measure of deprivation was ecological and relied on assigning individuals an average level of deprivation according to their area of residence. This methodology has the benefit of using a composite of multiple quality-assured measures of deprivation. However, ecological categorisation inevitably means individuals with different personal deprivation characteristics can be classified within the same category. Such classification may have hidden significant relationships between deprivation and some of the main variables of interest. Consideration of questionnaire length and compliance precluded us incorporating a comprehensive set of deprivation measures in this survey. However, future studies of the alcohol harm paradox would benefit from examining both individual and ecological measures. We also used a measure of atypical/special occasion drinking to identify unreported alcohol consumption. However, deprived and non-deprived individuals may have differed in recall of these occasions or assessed sizes and strengths of drinks differently [22]. Finally, we could not assess some alternative hypotheses for the alcohol harm paradox. Other competing theories include: individuals who become ill as a result of alcohol being drawn into more deprived communities through long-term disability and unemployment; genetic predisposition to suffering harms from alcohol in deprived populations; lower survey completion rates amongst heavy drinkers in poorer areas; and poorer access to and use of health and social support systems in deprived communities resulting in later or less treatment and support to avoid or tackle alcohol-related ill health [1, 9, 10].