You Are Here:
A Systematic Review and Meta-Analysis for the U.S. Preventive Services Task Force
Release Date: September 25, 2012
By Daniel E. Jonas, MD, MPH; James C. Garbutt, MD; Halle R. Amick, MSPH; Janice M. Brown, PhD; Kimberly A. Brownley, PhD; Carol L. Council, MSPH; Anthony J. Viera, MD, MPH; Tania M. Wilkins, MS; Cody J. Schwartz, MPH; Emily M. Richmond, MPH; John Yeatts, MPH; Tammeka Swinson Evans, MOP; Sally D. Wood, BA; and Russell P. Harris, MD, MPH
The information in this article is intended to help clinicians, employers, policymakers, and others make informed decisions about the provision of health care services. This article is intended as a reference and not as a substitute for clinical judgment.
This article may be used, in whole or in part, as the basis for the development of clinical practice guidelines and other quality enhancement tools, or as a basis for reimbursement and coverage policies. AHRQ or U.S. Department of Health and Human Services endorsement of such derivative products may not be stated or implied.
This article was first published in Annals of Internal Medicine on September 25, 2012 (Ann Intern Med; http://www.annals.org).
Background: Alcohol misuse, which includes the full spectrum from risky drinking to alcohol dependence, is a leading cause of preventable death in the United States.
Purpose: To evaluate the benefits and harms of behavioral counseling interventions for adolescents and adults who misuse alcohol.
Data Sources: MEDLINE, EMBASE, the Cochrane Library, CINAHL, PsycINFO, and reference lists of published literature (January 1985 through January 2012, limited to English-language articles).
Study Selection: Controlled trials at least 6 months in duration that enrolled persons with alcohol misuse identified by screening in primary care settings and evaluated behavioral counseling interventions.
Data Extraction: One reviewer extracted data and a second checked accuracy. Two independent reviewers assigned quality ratings and graded the strength of the evidence.
Data Synthesis: The 23 included trials generally excluded persons with alcohol dependence. The best evidence was for brief (10- to 15-minute) multicontact interventions. Among adults receiving behavioral interventions, consumption decreased by 3.6 drinks per week from baseline (weighted mean difference, 3.6 drinks/wk [95% CI, 2.4 to 4.8 drinks/wk]; 10 trials; 4332 participants), 12% fewer adults reported heavy drinking episodes (risk difference, 0.12 [CI, 0.07 to 0.16]; 7 trials; 2737 participants), and 11% more adults reported drinking less than the recommended limits (risk difference, 0.11 [CI, 0.08 to 0.13]; 9 trials; 5973 participants) over 12 months compared with control participants (moderate strength of evidence). Evidence was insufficient to draw conclusions about accidents, injuries, or alcohol-related liver problems. Trials enrolling young adults or college students showed reduced consumption and fewer heavy drinking episodes (moderate strength of evidence). Little or no evidence of harms was found.
Limitations: Results may be biased to the null because the behavior of control participants could have been affected by alcohol misuse assessments. In addition, evidence is probably inapplicable to persons with alcohol dependence and selective reporting may have occurred.
Conclusions: Behavioral counseling interventions improve behavioral outcomes for adults with risky drinking.
Primary Funding Source: Agency for Healthcare Research and Quality.
Alcohol misuse, which includes the full spectrum from risky or hazardous drinking to alcohol dependence (1–3), is associated with numerous health and social problems and more than 85,000 deaths per year in the United States (4, 5). Alcohol misuse is the third leading cause of preventable death in the United States, after tobacco use and being overweight (6). It contributes to hypertension, cirrhosis, gastritis, gastric ulcers, pancreatitis, breast cancer, neuropathy, cardiomyopathy, anemia, osteoporosis, cognitive impairment, depression, insomnia, anxiety, suicide, injury, and violence (7–9). The definitions of the spectrum of alcohol misuse (that is, unhealthy alcohol use ) continue to evolve. For this review, we use the definitions in Table 1 (10–12).
About 30% of the U.S. population misuse alcohol, with most engaging in what is considered risky drinking (1). Recent U.S.-based data (13) revealed that 21.3% of primary care patients reported risky drinking.
Cross-sectional and cohort studies have consistently related high average alcohol consumption and heavy per-occasion use to short- or long-term health consequences (14, 15). A meta-analysis examining the association between all-cause mortality and average alcohol consumption (16) found that men who drank an average of at least 4 drinks per day and women who drank an average of at least 2 drinks per day had increased mortality relative to nondrinkers. The National Institute on Alcohol Abuse and Alcoholism has proposed guidelines (17) to limit the risks for drinking-related consequences. The maximum recommended consumption is 3 or fewer standard drinks per day (≤7/wk) for adult women and anyone older than 65 years, and 4 or fewer standard drinks per day (≤14/week) for men (15, 17, 18). These guidelines do not apply to persons for whom alcohol intake is contraindicated, such as pregnant women, persons with alcohol dependence or medical conditions that can be worsened by drinking, or those receiving medications that interact with alcohol.
Behavioral counseling interventions include the range of personal counseling and related behavior-change interventions that are used to help patients change health-related behaviors (19). “Counseling” here denotes a cooperative method of work that demands active participation from both patient and clinician and aims to facilitate the patient's independent initiative (19). The goal of behavioral interventions for alcohol misuse is to eliminate risky drinking practices (for example, by encouraging fewer drinks per occasion or not drinking before driving) rather than to achieve abstinence.
For the Effective Health Care Program of the Agency for Healthcare Research and Quality (AHRQ) and to assist the U.S. Preventive Services Task Force (USPSTF) in updating its 2004 recommendation statement (20), we conducted a systematic review and meta-analysis of the effectiveness of screening followed by behavioral counseling, with or without referral, for alcohol misuse in primary care settings (21). The full report (21) addressed 7 questions (Appendix Table 1).
We developed and followed a standard protocol. A technical report that details methods and includes search strategies and additional evidence tables is available at www.effectivehealthcare.ahrq.gov/reports/final.cfm.
Key Questions and Analytic Framework
The USPSTF and AHRQ determined the focus of this review. Investigators developed key questions and created an analytic framework that incorporated the key questions and outlined patient populations, interventions, comparators, outcomes (including adverse effects), and settings (Appendix Figure 1). This report focuses on the key questions related to benefits and harms of behavioral interventions.
Data Sources and Searches
We searched MEDLINE, EMBASE, the Cochrane Library, CINAHL, PsycINFO, and the International Pharmaceutical Abstracts from 1 January 1985 to 31 January 2012, limited to English-language articles. The start date was selected on the basis of the earliest publication date found in previous reviews and expert opinion. We used Medical Subject Headings as search terms when available and key words when appropriate, focusing on terms to describe relevant populations, screening, and behavioral interventions.
We developed inclusion and exclusion criteria with respect to populations, interventions, comparators, outcomes, timing, settings, and study designs (22, 23). For the question related to behavioral interventions, we included randomized, controlled trials at least 6 months in duration, that enrolled adults or adolescents with alcohol misuse identified by screening in primary care settings, and that evaluated whether a counseling intervention improved behavioral or health outcomes.
Two investigators independently reviewed titles and abstracts, and then another 2 investigators independently reviewed the full text of all articles marked for possible inclusion during the initial review to determine final inclusion or exclusion. Disagreements were resolved with an experienced team member.
Data Extraction and Quality Assessment
We designed and used structured forms to extract pertinent information from each article, including information about the methods and populations, interventions, comparators, outcomes, timing, settings, and study designs. All data extractions were reviewed for completeness and accuracy by a second team member.
We assessed the quality (internal validity) of studies using predefined criteria based on those developed by the USPSTF (ratings of good, fair, or poor) (24) and the University of York Centre for Reviews and Dissemination (25). These included assessment of the adequacy of randomization, allocation concealment, similarity of groups at baseline, masking, attrition, and whether intention-to-treat analysis was used. Two independent reviewers assigned quality ratings for each study. Disagreements were resolved by an experienced member of the team.
Data Synthesis and Analysis
We stratified evidence by population (adults, older adults, young adults or college students, and pregnant women). Quantitative analyses were conducted of outcomes reported by a sufficient number of studies that were homogeneous enough to justify combining their results. We used random-effects models. For the outcome of alcohol consumption, the effect measure was mean difference between the intervention and control groups for change from baseline in drinks per week. The percentages of patients who had episodes of heavy drinking and those who achieved recommended drinking limits were compared (between intervention and control groups) with a risk difference. Because follow-up periods varied, the analysis for all-cause mortality was based on deaths per person-year and the comparison between intervention and control groups was calculated as a risk ratio. Analyses were conducted by using Comprehensive Meta Analysis, version 2.2.055 (BioStat, Englewood, New Jersey).
We used subgroup analyses to explore whether results differed by intensity, sex, country, deliverer of the intervention, or setting. The chi-square and I2 statistics were calculated to assess heterogeneity in effects between studies (26, 27). When quantitative analyses were not appropriate (for example, because of heterogeneity, insufficient number of similar studies, or insufficient or varied outcome reporting), we synthesized the data qualitatively.
To assess the differential effects of using more or less time and single or multiple contacts, we grouped interventions by intensity of counseling, as measured by the duration and number of contacts: very brief (≤5 minutes, single-contact), brief (6 to 15 minutes, single-contact), extended (>15 minutes, single-contact), brief multicontact (each contact ≤15 minutes), or extended multicontact (some contacts >15 minutes).
We then graded the strength of evidence (SOE) as high, moderate, low, or insufficient on the basis of the guidance established for the Evidence-based Practice Center Program (Appendix Table 2) (28). Two reviewers assessed each domain for each key outcome, and differences were resolved by consensus.
Role of the Funding Source
This review was funded by AHRQ. Staff of AHRQ and members of the USPSTF participated in developing the scope of the work and reviewed draft manuscripts. Approval from AHRQ for copyright assertion was required before the manuscript could be submitted for publication, but the authors are solely responsible for the content and the decision to submit it for publication.
We included 38 articles reporting on 23 randomized, controlled trials (Appendix Figure 2). Sample sizes ranged from 72 to 1559, and study durations ranged from 6 to 48 months (Appendix Table 3). Eleven studies were done solely in the United States, 2 focused on older adults, 5 focused on young adults or college students, and 1 enrolled pregnant women. We identified no studies of adolescents.
Fourteen of the interventions (29–50) were delivered by a primary care physician alone or with a health educator or nurse. Three (51–54) were delivered by a nurse or physician assistant, 1 by a psychologist (55–57), 2 by a researcher (58–62), and 1 by unspecified interventionists (63). Two interventions in college students (64–66) were conducted via a computer. Most trials tested brief multicontact interventions (31–34, 42, 46, 50, 51, 53, 64, 65) or brief interventions (29, 49, 52, 58, 60, 62–66); fewer tested very brief (45, 63), extended (30), or extended multicontact interventions (38, 45, 48, 55, 60). Interventions were heterogeneous and included various counseling approaches, such as brief advice, feedback, or motivational interviews, and cognitive behavioral strategies, such as self-completed action plans, written health education or self-help materials, drinking diaries, or problem-solving exercises to complete at home (Appendix Table 4). Most comparator groups received screening or assessment followed by usual care or by provision of a general health pamphlet. A few studies included additional components in comparator groups that could have biased results toward the null, such as recording screening or assessment results on the chart (45) or forwarding them to physicians (60), advice from nurses on reducing drinking and a leaflet with benchmark alcohol guides (52), a pamphlet on the health effects of alcohol consumption (64–66), or a booklet about preventing alcohol problems (48). We summarize the main findings by population and outcome and report the SOE for each.
We found no studies meeting inclusion criteria that randomly assigned participants, practices, or providers to screening and a comparator (no studies addressing questions 1 or 3) (Appendix Table 1). We found adequate evidence that several screening instruments can detect alcohol misuse in adults with acceptable sensitivity and specificity (21). The full technical report includes additional details about the accuracy of screening tests.
Effectiveness for Improving Intermediate Outcomes
Table 2 summarizes the results of meta-analyses for consumption, heavy drinking, and recommended drinking limits, by population. The Figure shows the forest plots for 12-month outcomes from our meta-analyses for adults. Overall, evidence supports the effectiveness of behavioral interventions for improving several intermediate outcomes for adults, older adults, and young adults or college students. For pregnant women, the included study (250 participants) (30) did not provide evidence of effectiveness for improving intermediate outcomes over 6 months or longer (low or insufficient SOE, depending on the outcome). Subgroup analyses identified no significant differences between men and women. Brief multicontact interventions had the best evidence of effectiveness across populations and outcomes and had follow-up data spanning several years. Meta-analyses of studies in adults found very brief and brief single-contact interventions to be ineffective for some outcomes and less effective than brief multicontact interventions for others.
Effectiveness for Reducing Morbidity, Reducing Mortality, or Changing Other Outcomes
Table 3 summarizes results, by population. Our meta-analyses found no statistically significant reduction in all-cause mortality for adults (rate ratio, 0.64 [95% CI, 0.24 to 1.7]; 4 trials) or for all age groups combined (rate ratio, 0.52 [CI, 0.22 to 1.2]; 6 trials). Point estimates tended toward favoring interventions, but few studies reported mortality and few long-term data were available. No studies that enrolled pregnant women and reported these outcomes were found (insufficient SOE).
Potential Adverse Effects
We found no evidence of direct harms, aside from opportunity costs associated with interventions, which ranged from 5 minutes to 2 hours dispersed over several in-person or telephone visits (moderate SOE). We searched for evidence of potential adverse effects, such as illegal substance use, increased smoking, anxiety, stigma, labeling, discrimination, or interference with the physician–patient relationship. We found no evidence for most of these potential harms and very limited evidence reporting no difference between groups for smoking rates and anxiety (low SOE). Other than the results for opportunity costs, our results are limited by the few trials that reported any information; 5 of 23 reported smoking (29, 33, 34, 39, 41, 49, 50), and 2 reported anxiety (29, 49).
Health Care System Influences
Where the study was conducted (United States vs. non–United States) had no impact on the effectiveness of interventions for consumption outcomes. Data showed a tendency toward greater reduction in consumption for interventions delivered in academic- or research-oriented settings than for those delivered in community-based settings (weighted mean difference, −5.0 drinks/wk [CI, −7.6 to −2.5 drinks/wk] vs. −3.2 drinks/wk [CI, −4.3 to −2.2 drinks/wk]; 3 vs. 7 trials). Interventions delivered mostly by primary care providers showed a tendency toward greater reduction in consumption than did those delivered primarily by research personnel (weighted mean difference, −4.0 drinks/wk [CI, −5.4 to −2.6 drinks/wk] vs. −3.0 drinks/wk [CI, −5.0 to −1.0 drinks/wk]; 7 vs. 2 trials). Our consumption meta-analysis included only 1 intervention delivered by a nurse (52), and the reduction was not statistically significant in that study (weighted mean difference, −0.2 drinks/wk [CI, −8.9 to 8.6 drinks/wk]). Two other studies, each of which provided insufficient data for our consumption meta-analysis, reported benefits of interventions delivered primarily by nurses (51) or by nurses and physician assistants (53) for some consumption outcomes. In addition, 2 interventions conducted by computer reported some evidence of effectiveness for reduced consumption in college students (64–66).
We found no studies that directly addressed our overarching question (Key Question 1)—no studies randomly assigned patients, practices, or providers to screening and comparator groups and subsequently provided interventions for those with positive screening results. All of the included studies randomly assigned patients after they had received positive screening results.
We found that behavioral counseling interventions improved drinking behavior outcomes (moderate SOE) and reduced hospital days (low SOE) for adults with risky drinking. For most health outcomes, available evidence either found no difference between intervention and control groups, such as for mortality (low SOE), or was insufficient to draw conclusions, such as for alcohol-related liver problems (insufficient SOE). Long-term outcomes from 2 studies (33, 35–37, 39, 42, 43) revealed that participants in the intervention groups maintained reductions in consumption or continued to reduce consumption, but differences between intervention and control groups were no longer statistically significant by 48 months. Studies identified delayed reduction in consumption in control groups that could reflect the natural history of alcohol consumption, the cumulative effect of follow-up with the health care system, differential attrition (if more participants lost to follow-up in the control group were risky drinkers), or (late) regression to the mean.
The evidence for effectiveness in adults is strongest for brief multicontact interventions. The effect sizes for these interventions were greater than those for other intensities (although CIs often overlapped). In addition, the best studies show that the effect of brief multicontact interventions remains for several years (35, 36, 43) and also show improvement for some utilization outcomes, such as fewer hospital days (35, 36) and costs (benefit–cost ratio of 39:1 over 48 months [CI, 5.4 to 72.5]) (36).
The brief multicontact interventions generally lasted 10 to 15 minutes per contact. All of the brief multicontact interventions in our meta-analyses of behavioral outcomes at 12 months were delivered by primary care providers, sometimes with additional intervention from a nurse or health educator. For example, the intervention in Project TrEAT (Trial for Early Alcohol Treatment) (33) included two 15-minute visits with a primary care provider 1 month apart and two 5-minute follow-up phone calls from a nurse 2 weeks after each visit. The intervention also included feedback about health behaviors, a review of problem drinking prevalence, a list of the adverse effects of alcohol, a worksheet on drinking cues, a drinking agreement or prescription, and drinking diary cards. Of note, 2 studies of brief multicontact interventions in adults, both of which provided insufficient data for our meta-analyses, reported benefits of interventions delivered primarily by nurses (51) or by nurses and physician assistants (53) for some consumption outcomes.
Evidence suggests that very brief interventions (up to 5 minutes, single-contact) and brief interventions (up to 15 minutes, single-contact) are less effective or ineffective, depending on the outcome. Although extended multicontact interventions seem to be effective for improving intermediate outcomes, we found no evidence that they are more effective than brief multicontact interventions.
The only included study that enrolled pregnant women (250 participants) (30) found no difference in reduced consumption between groups but did find higher rates of continued abstinence among women who were abstinent before the assessment in the intervention group than among those in the control group. Our searches identified other studies focusing on pregnant women that did not meet our inclusion criteria (67–84). Several took place in such settings as jails or specialized drug and alcohol treatment centers (75), and others lacked a control group or followed participants for fewer than 6 months (73, 84). Several of these studies reported benefits of interventions, including reduced consumption (73, 84), reduced risk for an alcohol-exposed pregnancy (75), higher rates of abstinence (79), and better fetal and newborn outcomes (higher birth weights and lengths and reduced fetal mortality rates ).
We have described several categories of alcohol misuse (such as risky or hazardous use and alcohol dependence). These categories are not all discrete (an individual may meet the definition of more than one). Included trials generally enrolled participants with risky or hazardous drinking, but the trials used varying terminology to describe the populations and often enrolled heterogeneous samples. Nevertheless, most investigators excluded participants with alcohol dependence or constructed their inclusion and exclusion criteria to substantially limit the number of such participants. Our best assessment is that our overall findings apply to risky or hazardous drinkers but not to persons with alcohol dependence. It is uncertain whether our findings apply to harmful drinkers or persons with alcohol abuse.
All interventions required support systems to provide screening; screening-related assessment; and in some cases, provider prompting. Screening assessments were often multistep processes that included interviews with research personnel that lasted up to 30 minutes. Less time would be required for screening and screening-related assessments in primary care practice. We estimate that 5 to 10 minutes would be required for persons who had positive screening results, with most of the time used to assess whether such persons have alcohol abuse or dependence (and should probably be referred for specialized treatment) as opposed to risky or hazardous drinking (for which behavioral counseling interventions in primary care may be effective). Nevertheless, support systems are probably required for effective screening and intervention. In addition, most interventions required training providers or staff.
It is unclear whether our findings apply to persons with certain comorbid conditions, and some researchers have suggested that brief behavioral interventions may be ineffective or less effective in people with comorbid psychiatric conditions. A subgroup analysis from a German study (56) found that brief interventions did not reduce drinking among 88 participants with comorbid anxiety or depression. Although most trials in our review did not exclude persons with depression, anxiety, or chronic pain, it is unclear how many participants with these conditions were included in most trials.
A previous systematic review (85) found no evidence of efficacy for brief behavioral interventions in patients with alcohol dependence in primary care settings. Our review also found no such evidence. Included studies that enrolled more than 10% of participants with alcohol dependence reported interventions to be less effective or ineffective than studies that did not enroll alcohol-dependent participants.
Screening for alcohol misuse will inevitably identify some alcohol-dependent individuals; thus, providers and those making recommendations need information about whether effective interventions are available for alcohol dependence. If complete abstinence is used as an outcome, 15% to 35% of patients have been reported to achieve 1 year of sobriety after such treatment approaches (86) as pharmacotherapy, motivational enhancement therapy, cognitive behavioral therapy, 12-step facilitation, and therapy at alcoholism-treatment centers. Similar sobriety outcomes at 3 to 5 years or longer have been reported (9).
Our review has limitations. First, the scope of our review was limited to primary care settings. Second, most evidence involved self-report of alcohol use. Investigators in some trials verified self-reported use with other persons (such as family members). Self-report of alcohol use has been found to be accurate if collected carefully (87, 88). Third, the assessments conducted in the included trials could have concealed benefits of interventions (and biased results toward the null) by causing behavior changes. Control participants generally reduced alcohol consumption. Possible explanations include increased awareness of drinking, discussions with their provider about drinking that were prompted by the screening questions, receipt of some minimal intervention (control groups in the included studies often received some printed educational materials), or regression to the mean. A recent systematic review (89) concluded that answering questions on drinking in brief intervention trials seems to alter subsequent self-reported behavior, potentially generating bias by exposing nonintervention control groups to an integral component of the intervention. Finally, publication bias and selective reporting may be present.
In conclusion, behavioral counseling interventions improve intermediate outcomes, such as alcohol consumption, heavy drinking episodes, and drinking above recommended amounts (moderate SOE) and may reduce hospital days (low SOE) for adults with risky or hazardous drinking. For most health outcomes, available evidence found no difference between intervention and control groups, such as for mortality (low SOE), or was insufficient to draw conclusions about the effectiveness of behavioral interventions, such as for alcohol-related accidents or quality of life (insufficient SOE). Brief multicontact interventions (about 10 to 15 minutes per contact) have the best evidence of effectiveness for adults.