Discussion

Summary of Research Findings

We found that good-quality brief, multi-contact behavioral counseling interventions reduced risky and harmful alcohol use by primary care patients for several alcohol outcomes. A recent meta-analysis that included 7 of the 12 trials we examined reported pooled estimates for the proportion drinking sensibly at followup, an absolute risk reduction of 10.5% (CI, 7.1% to 13.9%), with a number needed to treat for benefit of 10 (CI, 7 to 14).⁶⁷ We found similar results (ranging from 10% to 19% more intervention participants than controls achieving safe or recommended drinking levels) among studies providing brief, multi-contact interventions. We examined other equally relevant alcohol outcomes and found that good-quality brief multi-contact intervention trials also reduced weekly drinking 2.9 to 8.7 mean drinks per week more than in controls (13% to 34% net reductions) but had inconsistent effects on binge drinking. Very brief or brief single-contact interventions were less effective or ineffective in reducing risky/harmful alcohol use. This finding contrasts with the significant results seen for very brief and brief tobacco interventions among adults in primary care and other medical settings.⁶⁸ Effective interventions generally included advice, feedback, goal setting, and additional contacts for further assistance and support, although available evidence cannot clearly distinguish higher-intensity intervention effects from intervention components. The elements in effective interventions were generally consistent with the 5 A's (assess, advise, agree, assist, arrange) approach to behavioral counseling interventions adopted by the USPSTF.⁶⁹

Earlier intervention studies and reviews raised concerns that women either might not be as responsive to brief interventions as men or might be so responsive to screening alone that brief intervention would not confer much additional benefit. Our results are consistent with recent reviews that found no important sex differences in outcomes of brief interventions.^31,36,38 Primary care interventions also appear effective in older as well as younger adults, according to the results from a trial targeting older adults⁵¹ and inclusion of older adults in most trials reviewed.

Less is known about the direct effects of risky/harmful alcohol use interventions on morbidity and mortality than on alcohol intake. Mortality benefits were seen primarily in 1 extended intensive intervention (with repeated contacts up to 5 years) among more severely affected drinkers.⁶⁵ It is not clear whether mortality benefits will be seen with less severe drinkers undergoing the less intensive interventions typical of studies reviewed here. Since most favorable mortality outcomes were seen only in males or younger males, mortality benefits may accrue primarily to specific subgroups, and their demonstration may require 4 or more years of followup. Results were mixed for morbidity measures, and future research is clearly needed; primarily null findings may reflect lack of an effect, reduced power for secondary analyses, or insufficient measures.

Patients were identified for intervention by methods including standardized screening instruments such as AUDIT and CAGE (to detect alcoholism but not risky drinking) that have been shown to perform adequately in primary care populations.^2,44,45 The 2-step strategy used in trials approximates the NIAAA-recommended approach, in which all patients identified as alcohol drinkers are asked about usual quantity and frequency of drinking, maximum drinks per occasion in the past month, and the 4 CAGE screening questions (wanting to Cut down on drinking, people Annoying you by criticizing your drinking, feeling Guilty about your drinking, and having an "Eye-opener" drink upon arising in the morning).³⁰ The second step is a confirmatory clinical assessment that also considers specific alcohol problems and dependence.

If primary care clinicians appropriately use these validated screening instruments in conjunction with clinical assessment and judgment, they are likely to identify patients in their practices who are similar to trial participants. Screening and assessment steps were not tested as part of the clinical protocol in these studies, however, and most interventions involved contact with research personnel to determine study eligibility. We found that at least 8% to 18% of general primary care patients would be candidates for brief interventions (screen positives), with at least half remaining eligible after completing the assessment step; according to available data, active refusal rates should be fairly small.⁴⁰ In the recent meta-analysis of many of the same studies, a similar proportion (9% [range, 3% to 18%]) of patients screened positive, but estimates for the proportion remaining after the assessment step were much lower than ours.⁶⁷ The authors used their lower estimate of the final screening yield to calculate a benefit for screening and intervention of 2 to 3 per 1,000. They have been criticized, however, for such issues as equating the screening yields from recruitment for intervention efficacy trials with those that would result from usual care screening;⁷⁰ other concerns about this meta-analysis have also been discussed.^71-75

Implications and Future Research Recommendations

Considerable work is needed to implement screening and brief intervention for risky/ harmful alcohol use as part of routine practice, and more research is needed on effective strategies and supports for adoption of these services by physicians and health plans. While brief or very brief interventions may be more easily incorporated into routine primary care, effectiveness of risky/harmful alcohol use interventions probably depends on multiple contacts over time. Most primary care physicians report asking about alcohol use, but far fewer use recommended screening protocols³³ or prefer physician counseling as the means to address risky/harmful users.⁷⁶ Current research points the way to persuading physicians to accept screening and intervention materials⁷⁷ and to providing training that increases screening and intervention activities.⁷⁸ Prompting untrained physicians with alcohol screening results and simple treatment recommendations yields mixed results in terms of alcohol advice and discussions or patient drinking behaviors.⁴⁸ Given the system supports provided for most trials, those seeking positive results from these interventions in real-world clinical practice will probably require similar support, such as:

1. Commitment to planning.
2. Allocation of resources and staff to consistently identify risky/harmful alcohol-using patients.
3. Delivery resources (such as clinician training, prompts, materials, reminders, and referral resources).

Trials are needed to examine the direct effects on alcohol use, mortality, and morbidity (including quality of life, mental health, and social functioning) of screening followed by interventions for risky/harmful alcohol use and to report possible harms associated with screening, assessment, and brief intervention. Future intervention research should more directly target screening, interventions, and outcome measures to address binge use. Future research is also needed to establish possible cost savings⁷⁹ or cost-effectiveness⁸⁰ for these interventions.

Limitations of Our Review

We did not quantitatively summarize study trial results; however, our findings are generally consistent with findings from meta-analyses of brief interventions on alcohol consumption in primary care.^36-38,67

Our review primarily addressed the effect of behavioral counseling interventions on patients identified as risky/harmful alcohol users and did not systematically address the performance of screening tests to identify these patients. We relied on the previous USPSTF recommendation and intervening systematic reviews by others for our conclusions about screening tests. We judged that methods to identify patients for the intervention trials and validated, feasible primary care screening tests (when coupled with clinical assessment) are sufficiently similar, after removing the burden imposed by research, although we did not test this assumption by this review.

The alcohol use outcomes relied on self-report, with occasional collateral verification, since there are no good objective measures of changes in alcohol use.⁸¹ Self-report of alcohol use has been found to be as accurate as or more accurate than other measures if collected carefully, such as when elicited as part of a general health assessment by non-clinical personnel outside the clinical setting.⁸² Given that these conditions were often met in the trials reviewed and that we relied on finding net improvements in alcohol consumption patterns, we believe that self-reported alcohol consumption is a reasonable basis for the findings in this report.

We did not address health care interventions in settings other than primary care. Other settings, such as the emergency department or trauma units, may offer other important health care opportunities to address problematic alcohol use in patients.

Publication bias may also have affected our results. Although we located many unpublished or pre-published studies, we cannot be certain that we located all negative studies.

Return to Contents

Acknowledgments

The assistance of Shannon Sanner, M.P.H., was invaluable in the conduct of this review and preparation of its results. Al Berg, M.D., M.P.H., and Al Siu, M.D., M.S.P.H., provided critical oversight and assistance for the USPSTF. David Atkins, M.D., M.P.H., and Eve Shapiro, Managing Editor of the USPSTF, provided insightful editorial assistance. Elizabeth Haney, M.D.; Jae Douglas, Ph.D.; and Nancy Perrin, Ph.D., provided outcome summary and statistical assistance, and Martha Swain provided editorial support.

This study was conducted by the Oregon Evidence-based Practice Center under contract to the Agency for Healthcare Research and Quality, Rockville, MD (contract 290-97-0018, task order No. 2).

Return to Contents
Proceed to Next Section