|
Article Excerpt
Introduction
Overall, alcohol has been linked to substantial mortality and disability (WHO, 2002; Rehm et al., 2003a; 2004). However, there are marked regional differences in the disease burden associated with alcohol (Rehm et al., 2004). Alcohol-attributable burden of disease is especially high in the Americas and the Central and Eastern European regions (Rehm et al., 2003a; 2004). In the Americas, alcohol has found to be the most important single risk factor contributing to burden of disease (Rehm & Monteiro, 2005), surpassing smoking, obesity, and high blood pressure.
Two major dimensions of alcohol use have been linked to disease incidence: average volume of consumption and patterns of drinking, mainly irregular heavy drinking occasions (Greenfield, 2001; Rehm et al., 2003b; c; d). Investigating the impact of average volume of alcohol consumption has long been the dominant paradigm in alcohol epidemiology, and most epidemiological publications continue to focus on this aspect (Edwards et al., 1994; English et al., 1995; Corrao et al., 1999; 2004). However, it has become clear that other dimensions of alcohol consumption influence the occurrence of disease above and beyond those captured through average volume. A wide variety of measures along a number of dimensions are dealt with under the heading of "patterns of drinking" in alcohol epidemiology (Rehm et al., 1996), although research has focused especially on the role of heavy drinking occasions in the etiology of disease. It should be noted that even though the epidemiological research on drinking patterns and disease has been quite recent, the theoretical plea to conduct such research has been made repeatedly already decades ago (e.g., Room, 1979).
The present contribution tries to summarize research on drinking patterns and disease, focusing on the relationships between different dimensions of alcohol use and mortality. It starts with all-cause mortality, as all-cause mortality is the most widely used summary measure that combines different causes of death (Rehm & Gmel 2003). Using this measure, some of the complex relations between alcohol consumption measures and mortality rates can be illustrated. After discussing all-cause mortality we will proceed to summarize the research for different major categories of disease.
Methods
As illustrated below, the current research on patterns of drinking and mortality is characterized by diverse measures of alcohol intake. In addition to average volume assessed over varying reference periods (Rehm, 1998), among the most common "pattern" measures are presence or frequency of occasions with drinking exceeding various threshold numbers of standard drinks, like 5 or more (5+) drinks in any day (or in self-defined "occasion", or in as short a time as 2 hours) or 8+, 12+ etc., and maximum quantity consumed in any given period such as the last 12 months (Greenfield, 1998). Indeed, there are no standardized definitions or operationalizations of heavy drinking occasions (Gmel et al., 2003), and as a result, studies are hard to compare, so much so that no quantitative meta-analysis seems possible.
Instead, we undertake a qualitative overview of the evolving literature. Published articles were located through systematic computer-assisted searches in the PudMed and ETOH databases in August 2005 and January 2006 entering the following key words: "pattern of drinking" or "heavy drinking occasion" and "mortality." ETOH is the database of the U.S. National Institute for Alcohol Abuse and Alcoholism and covers relevant literature until the end of 2003.
Average volume of alcohol consumption and all-cause mortality
Meta-analyses on this large and diverse literature have found the relationships between average volume of consumption and all-cause mortality to be J-shaped curves, with women experiencing deleterious effects at lower levels of alcohol consumption (English et al., 1995; Rehm, Gutjahr & Gmel, 2001; Gmel, Gutjahr & Rehm, 2003). These curves can be interpreted as reflecting the beneficial effects of moderate alcohol consumption, in particular on ischaemic heart disease, together with the detrimental effects of alcohol on many other chronic diseases and acute outcomes (Rehm et al., 2003b; d). Some of the above mentioned meta-analyses have also demonstrated the importance of separating ex-drinkers from life-time abstainers to adjust for the effect of so-called "sick quitters" (Shaper et al., 1988; Gmel, Gutjahr & Rehm, 2003). The sick quitter effect results from people who have stopped drinking because of ill health, and who therefore obviously have a higher mortality risk than other abstainers (for quantification see Rehm, Gutjahr & Gmel, 2001). In numerous studies, where former drinkers are not distinguished, the mortality risk for abstention has likely been overestimated, making the overall risk curve look more U-shaped. Further, since abstainers are commonly used as the reference group, the protective effect of light to moderate drinking would be overestimated (Fillmore et al., in press).
This overall relationship between average volume of alcohol consumed and all-cause mortality has been found to differ across age-groups across countries and between ethnic groups. Most of the beneficial effects of alcohol on mortality concern older age groups only, as the respective disease categories do not really contribute to mortality for people younger than 45 years. Thus, the estimated relation between average volume of alcohol consumption and all-cause mortality is almost linear for individuals below age 45 (Rehm, Gutjahr & Gmel, 2001; Gmel, Gutjahr & Rehm, 2003).
Studies in some populations and subpopulations, including African Americans and a large study on Russian men and women, have not found evidence of a J-shaped curve in the relationship between average volume of alcohol consumption and mortality. Instead, no beneficial effect appeared and mortality risk actually increased with average consumption for more than one drink a day (Sempos et al., 2003; Nicholson et al., 2005).
In sum, the exact shape of the risk curve between average volume of consumption and all-cause mortality will vary. As we will argue below, much of the variation will depend on patterns of drinking, especially the extent of irregular heavy drinking occasions. However, other factors, including the distribution of diseases in a population as well as social and demographic factors also play a role.
Drinking patterns and risks curves of conditions that may contribute to mortality
The literature on the impact of drinking patterns over and above volume of consumption is better established for associations with specific, concurrent alcohol-related health harms than it is for subsequent all-cause and specific-cause mortality (Greenfield, 2001). One part of this literature that is relevant to this point deals with risk curves for nonfatal injuries and disease morbidity based on cross-sectional surveys. Risk curves by volume of drinking have been plotted for North American countries and a few other developed societies (e.g., Hauge & Irgens-Jensen, 1986; Greenfield et al., 1996; Midanik et al., 1996). A consistent finding is that reporting social and health problems with drinking, including nonfatal injuries, seems to be more heavily related to at least occasional intoxication than to volume or frequency of drinking (Midanik et al., 1996). The degree of intoxication has also been shown to relate to current self-reported problems. Substantially more problems were reported by drinkers who consumed 12+ drinks at least once a month than by those who had 5+ drinks at least once a week but not 12+ monthly (Cahalan & Room, 1974). Contrary to expectation, controlling for other factors, male and female risk relationships have been nearly the same in some studies of the relation between drinking pattern and alcohol dependence (e.g., Caetano et al., 1997). This suggests that although a given amount of alcohol may be expected to produce greater intoxication in women than in men (on average, due to smaller bodies and body water volumes), women may offset this by drinking in less hazardous ways than men (e.g., more slowly, or more often with meals) (Knibbe et al., 1993; Graham et al., 1998; Dorn et al., 2003; Stranges et al., 2004; York et al., 2003). This highlights the importance of carefully considering how variables chosen may affect gender-specific results.
Differences in operationalizing measures and other analytic choices affect findings (Greenfield, 1998), including gender-specific results. For example, in a risk curve analysis using the 1988 National Health Interview Survey Alcohol Supplement (NHIS; n = 22,102 adult drinkers), Caetano et al. (1997) found that while many fewer women than men drank large volumes, or frequently had 5+ drinks in a day, and Diagnostic and Statistical Manual of Mental disorder--fourth edition (DSM-IV) alcohol dependence rates were higher (6%) for men than for women (3%) drinkers, for given levels of the consumption variables (volume and heavy drinking), gender did not affect the odds of reporting alcohol dependence, after taking into...
|
