Also considered were a number of research areas that influence alcohol-related mortality and morbidity such as health services, advertising, public policy, minorities, elderly, nosology, measurement and genetic epidemiology. Further, we considered NIAAA initiatives lodged in the Division, or to which division members make a large contribution: college and underage drinking, AIDS, the Alcohol Policy Information System, and the Alcohol Epidemiology Data System.
As a next step, the Division undertook a review of its current research portfolio examining both the numbers of grants and dollars spent on grants in certain categories. The portfolio was evaluated according to: whether a grant is focused primarily on epidemiology or prevention; the population under study (e.g. age, gender, race/ethnic group); and the type of mortality, morbidity or social problem under study. This was done to assess whether there are important causes of mortality and morbidity and/or population subgroups not receiving sufficient research attention. (For a summary of the Division’s Research Portfolio according to these categories, see Appendix B.)
Finally, members of the division met to review the information accrued from the four steps outlined above. All division members attended the Institute wide Strategic Planning meeting in July 2005 and were asked during these deliberations to also consider the strategic plan priorities and rationales for those priorities posed by the various Institute Research Emphasis Teams and other Divisions. Division members discussed key aspects of the relationship of alcohol, and health and well-being in order to identify priorities for future alcohol research. Division members were then asked to prepare brief reviews of each of the areas identified. The reviews include a discussion of the magnitude and consequences of the particular alcohol-related problem for the nation, recent trends in the magnitude and consequences of that problem, which groups in the population are most affected, what is known about the etiology of the problem, how alcohol contributes to the problem, what is known from rigorous scientific research about what can be done to prevent the problem, and a summary of recent relevant research funding and findings from that research. Finally, Division staff was asked to articulate key research questions in each domain, the answers to which would better enable society to address the problem, and thereby reduce alcohol attributable mortality and morbidity. This report combines those reviews into a Division-wide statement of research progress and priorities.
Table 1
Source: Midanik et al. (2004). MMWR. 53(37): 866-870
III. ALCOHOL-RELATED MORTALITY AND MORBIDITY
A. OVERVIEWAlcohol consumption ranks among the leading causes of death and poor health in the United States. Adverse health consequences of alcohol consumption include a wide range of injuries, both unintentional and intentional, as well as numerous chronic health conditions. Because of the wide range of adverse health effects associated with alcohol consumption, no single measure can provide a complete representation of the burden to population health that results from alcohol consumption. Instead, a variety of measures may provide insights into the nature and extent of alcohol-attributable harms to health. Indicators of the health burden attributable to alcohol include direct measures of mortality and morbidity, indirect measures of adverse health consequences based on health care utilization, synthetic measures of population health that integrate mortality and morbidity in a single metric, and broader measures of the adverse economic effects of alcohol consumption.
Methodological and measurement factors affect indicators of the health burden attributable to alcohol, leading to significant variation across measures that might be expected to yield similar findings. Among the main sources of variation are differences in the specific conditions that may be caused by alcohol consumption that are included or excluded in a particular study, and the degree of causation attributed to alcohol consumption for specific conditions and outcomes. Despite the uncertainties arising from these considerations, the clear conclusion across the spectrum of indicators is that alcohol consumption causes or contributes to a significant share of death and poor health in the United States.
MORTALITY
Recent estimates of alcohol-related mortality range from 64,000 to 140,000 per year for 2000 or 2001 (Midanik et al., 2004; Mokdad et al., 2004; Rivara et al., 2004), with “best judgment” point estimates in the range of 76,000-85,000. Earlier estimates reported by NIAAA for 1979-1996 ranged from 103,000 to 113,000 (NIAAA QuickFacts). Analyses of external modifiable factors that contribute to death, labeled “actual causes of death,” have placed alcohol consumption as the third leading such cause for 1990 and 2000, after (1) tobacco use, and (2) poor diet and physical inactivity (Mokdad et al., 2004; McGinnis and Foege, 1993).
Available evidence indicates that total mortality attributable to alcohol has declined in recent years, although methodological changes may influence the estimates. The Quick Facts section of the NIAAA Web site reports that alcohol-related mortality for the United States declined from a high of 112,933 in 1980 to a low of 103,247 in 1983, followed by 12 years of steady increases to a peak of 111,290 in 1995 before leveling off at 110,640 in 1996.
Several recent studies have estimated alcohol-attributable mortality for 2000 or 2001 at levels substantially below those reported for earlier years. Midanik et al. (2004) estimated alcohol-attributable deaths at 75,766 for 2001 based on condition-specific alcohol-attributable fractions, most of which were calculated using estimates of relative risk drawn from Australian meta-analyses (English et al., 1995; Ridolfo and Stevenson, 2001). Alcohol was the third ranked preventable cause of death in the United States. Rivara et al. (2004) estimated 63,718 alcohol-attributable deaths for 2000 using generally similar methods but based their findings on relative risk estimates comparing heavier alcohol consumption levels to lower-level consumption (rather than to abstention). Mokdad et al. (2004) reported estimates of alcohol-attributable deaths for 2000 based on several alternative methods, including a condition-specific approach using the Australian relative risk estimates that yielded an overall estimate of 69,989 and a best-judgment estimate of approximately 85,000. The differences between recent estimates and earlier findings appear to be due in part to differences in methodology, as the specific conditions and associated attributable fractions included in the recent studies differ significantly from those that factored into the earlier estimates (Bloss, 2005). Taken collectively, however, the clear conclusion to be drawn from these recent studies is that overall alcohol-attributable mortality in the U.S. around the turn of the current century is somewhat lower than had been reported for the late 1980s and early 1990s. As will be discussed, there is strong evidence that alcohol-related traffic deaths, the largest single category of alcohol-related deaths, have shown marked declines since the early 1980s. One analysis indicates reductions in drinking and driving in the U.S. saved 153,000 lives between 1982 and 2001 – more than increased use of seat belts, airbags, bicycle helmets, and motorcycle helmets combined (Cummings et al., 2006)
Studies have also examined the years of potential life lost (YPLLs) due to alcohol, which incorporates information on the age of decedents in alcohol-attributable deaths. Relative to life expectancy, an estimated 2.3 million YPLLs were attributed to alcohol consumption in 2000 (Midanik et al., 2004), accounting for approximately half as many YPLLs as smoking (estimated for 1999) even though smoking accounts for nearly six times as many deaths. An earlier study estimated alcohol-attributable YPLLs for 1987 at 2.7 million (Shultz and Rice, 1990).
A number of studies have examined the risk relations between alcohol consumption and all-cause mortality. The central finding is a clear increase in risk of mortality with increasing consumption above a threshold level of moderate consumption. The majority of studies have found U-shaped or J-shaped relationships between consumption levels and overall mortality, suggesting that low-level alcohol consumption may confer a protective effect on some fatal outcomes relative to abstention (Poikolainen, 1995; Rehm, 2000; Gmel et al., 2003). Two recent studies have questioned the existence of a beneficial effect of low or moderate consumption on all-cause mortality (Fillmore et al., 2006) or cardiovascular mortality (Naimi et al., 2005) on methodological grounds.
MORBIDITY
Measuring the extent of poor health attributable to alcohol consumption is more difficult than measuring deaths caused by alcohol consumption. Few direct measures of alcohol-related morbidity are available. The National Epidemiologic Survey on Alcohol and Related Conditions found that 5.4 percent of men age 18 and older and 2.3 percent of women 18 and older met DSM-IV criteria for alcohol dependence within the past 12 months in 2001-2002. These figures represent modest but significant decreases compared to estimates based on the National Longitudinal Alcohol Epidemiologic Survey for 1991-1992. In contrast, the prevalence of DSM-IV alcohol abuse increased over the same period, from 4.7 percent to 6.9 percent for men, and from 1.5 percent to 2.6 percent for women (Grant et al., 2004).
For conditions other than alcohol use disorders, the role of alcohol in generating poor health is usually inferred from data on health care utilization. A report on alcohol-related hospital discharges (Chen et al., 2005) based on data from the National Hospital Discharge Survey finds that 424,000 hospital discharge episodes in 2003 for persons 15 and older had a principal (first-listed) alcohol-related diagnosis, and approximately 1.6 million discharges had any alcohol-related diagnosis. Discharges with any alcohol-related diagnosis accounted for between 5 and 6 percent of all hospital discharges in 2003 (excluding discharges for females with delivery of a baby). The alcohol-related diagnoses that formed the basis of these findings were limited to alcoholic psychoses, alcohol dependence syndrome, nondependent abuse of alcohol, and chronic liver disease and cirrhosis. While the final category included some discharges for liver disease that were not caused by alcohol, the overall estimates are surely very conservative because they omit a wide range of conditions known to be partially caused by alcohol consumption, including various cancers, cardiac conditions, stroke, gastro-intestinal conditions, various unintentional and intentional injuries, and fetal effects caused by alcohol. Males accounted for slightly more than twice as many alcohol-related discharges as females, and the rate of alcohol-related hospitalization per 10,000 population was greatest among the 45-64 age group, slightly lower for 65 and older and 25-44 year olds, and considerably lower for 15-24 year olds. The highest rates of hospitalization were for alcohol dependence syndrome and alcoholic psychoses, with lower hospitalization rates for non-dependent abuse of alcohol and for chronic liver disease and cirrhosis. The study also found increasing frequency of alcohol-related diagnoses from 1979-2003, although the frequency of alcohol-related conditions appearing as the first-listed diagnosis declined over the same period. These trends are shown in Figure 1.
Another report examined hospitalizations for alcohol abuse disorders for 2003 based on a different data source and considering only diagnoses of alcoholic psychoses, alcohol dependence syndrome, and nondependent abuse of alcohol. This study found that these alcohol abuse disorders were principally responsible for almost 210,000 hospitalizations and were listed as concomitant conditions for nearly 1.1 million additional hospital stays (Russo and Elixhauser, 2006). An earlier study using a broader set of alcohol-related conditions estimated the total number of hospital days attributable to alcohol at 5.3 million for the United States for 1992 (Harwood et al., 1998).
Figure 1
Trends in percent of discharges with principal (first-listed) or any (all-listed) mention of an alcohol-related diagnosis among all discharges, 1979–2003.
*Note: Shaded are represents the period before implementation of a new sample design which may affect the trend data. Caution should be taken when making comparisons between the new and old sample design periods.
Source: Chen et al., 2005
Further indicators of the contributions of alcohol consumption to acute health problems are available from studies of alcohol-related visits to emergency departments. McDonald et al. (2004) estimated that the number of emergency department (ED) visits in the United States attributable to alcohol averaged 7.6 million per year over from 1992 to 2000, accounting for 7.9 percent of all ED visits over that period, with a rising trend showing an overall 18 percent increase in alcohol-related ED visits over the 9-year study period. Elder et al. (2004) considered alcohol-related ED visits among 13-25 year-olds, and found 244,331 visits to U.S EDs during 2001 by patients age 13-25, with half of these patients under the legal drinking age. Cherpital and colleagues (2005) used meta-analytic methods to estimate risks attributable to drinking before the injury (evidenced by positive BAC or self-report) and to a pattern of binge drinking (5+ drinks at least once per month) for both all-cause and violence-related injuries. They found that only between 2 percent (positive BAC) and 6 percent (self-report) of all-cause injuries were attributable to drinking before the occurrence of the injury, and 6 percent were attributable to binge drinking patterns. Violence-related injuries, however, showed a much stronger relationship to alcohol consumption, with between 28 percent (positive BAC) and 43 percent (self-report) of such injuries attributable to drinking before the occurrence of the injury, and 27 percent attributable to binge-drinking patterns. For each category, attributable risks were at least three times as large for men as for women.
The clear conclusion that emerges from studies of morbidity attributable to alcohol is that alcohol consumption accounts for a significant share of poor health in the United States.
OTHER MEASURES OF HEALTH BURDEN
Recent research has emphasized summary measures of population health that incorporate mortality and morbidity effects into a single measure (Gold et al., 2002; Murray et al., 2000). A leading measure of the health burden of specific diseases or risk factors is disability-adjusted life years (DALYs). Conceptually, DALYs represent the sum of life-years lost to a particular condition plus a weighted fraction of years lived with disability as a result of the condition, with the fractional weights increasing with greater degrees of disability. A recent study of the contribution of various diseases and risk factors to the overall burden of disease in the United States found that alcohol use was the 5th-leading source of DALYs for men and the 11th-leading source of DALYs for women in 1996 (McKenna et al., 2005). For both groups, the burden of disability attributed to alcohol far outweighed the years of life lost (by a factor of 8 for men and a factor of more than 16 for women). This estimate may err on the conservative side, however, as there is no indication of the range of health consequences attributed to alcohol consumption. In addition, an earlier study using DALYs to assess the burden attributable to various risk factors in an international framework reported deaths attributable to alcohol use net of deaths averted as a result of alcohol use through the hypothesized protective effect of alcohol consumption on cardiovascular disease (Murray and Lopez, 1997). The low number of deaths (6,000) attributed to alcohol use in the more recent study suggests that a similar procedure may have been used. Thus, the burden of the adverse consequences of alcohol use may be greater than this study suggests.
POPULATIONS AT RISK
Numerous studies have documented increasing risk of all-cause mortality with heavier average consumption levels, indicating that heavy drinkers face increased alcohol-related mortality risks. For some conditions, particularly various outcomes associated with acute intoxication, evidence suggests that concentrating a given amount of consumption in fewer drinking occasions leads to increases in risk (Gutjahr et al., 2001). These considerations indicate, not surprisingly, that risks of alcohol-related death and illness may be greatest among drinkers with heavy average consumption levels and those who drink heavily on an episodic basis. A study of U.S. data found that increases in risk among heavy drinkers were found among those who met criteria for alcohol dependence, while protective effects associated with moderate alcohol consumption were limited to those without alcohol dependence (Dawson, 2000). Alcohol-related mortality is greater among males than among females, reflecting males’ greater consumption levels. Because alcohol-related motor vehicle crash injuries represent the largest single category of alcohol-related deaths, drinking drivers and their passengers are also identifiable as an at-risk group.
RESEARCH PROGRESS
Estimation of alcohol-related mortality may be accomplished through cause-specific or all-cause estimation methods. In either case, alcohol-attributable fractions are generally derived from estimates of relative risk (RR) for various exposure levels combined with prevalence rates for those exposure levels. The current state-of-the-art involves meta-analyses of individual studies to arrive at pooled estimates of relative risk. This approach was advanced by English et al. (1995) with refinements by Ridolfo and Stevenson (2001) and is the subject of a NIAAA-contracted study conducted by Harvard University School of Public Health.
Use of alcohol-attributable fractions raises a number of methodological and conceptual issues (Rockhill et al., 1998). Based on the detail available in the underlying studies, relative risk estimates may differentiate by population groups (e.g., sex, age groups). Different relative risks for specific conditions may also be associated with different patterns of consumption (e.g., consumption parameters beyond measures of average volume of ethanol) and the extent to which potential confounders such as diet, physical activity, and smoking have been controlled. To the extent that consumption patterns may vary across population groups, these two categorizations may be confounded (Bloss, 2005; Rehm and Gmel, 2003; Rehm et al., 2002).
There is now substantial evidence of beneficial effects of moderate alcohol consumption for cardiovascular disease (Mukamal and Rimm, 2001; Corrao et al., 2000). Additional beneficial effects have been suggested, but not conclusively determined, for ischemic stroke, diabetes, and cholelithiasis (gallstones) and may be present for other conditions as well (Ashley et al., 2000). Particularly given the apparent strength of the beneficial effect and the aggregate number of deaths attributable to cardiovascular disease, it is important to report such effects appropriately. This may not mean, however, offsetting beneficial effects against adverse consequences, but instead may require separate reporting of beneficial and adverse effects at various levels of consumption.
RESEARCH QUESTIONS/ DIRECTIONS
