Discussion (original) (raw)
In this report, we aimed to conduct a systematic review of the effectiveness of screening followed by behavioral counseling for alcohol misuse in primary care settings. In the introduction, we describe several categories of alcohol misuse (i.e., risky/hazardous drinking, harmful drinking, alcohol abuse, and alcohol dependence). It is important to note that the categories are not all discrete categories (i.e., an individual may meet the definition for more than one category for some of these categories). For example, one study estimated that 36 percent of men and 44 percent of women classified as hazardous drinkers also met the criteria for harmful drinking.18 It appears that the included trials of behavioral counseling generally enrolled subjects with risky/hazardous drinking, but the trials use varying terminology to describe the included populations and often enrolled heterogeneous populations (i.e., included subjects with various types of alcohol misuse). Nevertheless, the vast majority of trials excluded subjects with alcohol dependence or constructed inclusion/exclusion criteria to substantially limit the number of potential subjects with alcohol dependence. Just three studies reported that more than 10 percent of enrolled subjects had alcohol dependence.83, 90, 91 It is not clear how many trials enrolled subjects with alcohol abuse, because this was generally not mentioned in the publications.
Given the heterogeneity in terminology used by the included trials and the potential overlap of some categories of alcohol misuse, our best assessment is that our overall findings from behavioral counseling intervention trials are applicable to risky/hazardous drinkers; they are unlikely to be applicable to those with alcohol dependence (because very few subjects in the included trials had alcohol dependence—although that makes applicability to those with alcohol dependence somewhat uncertain). It is uncertain whether findings are applicable to harmful drinkers or people with alcohol abuse.
Although we did not systematically review the effectiveness of the recommended treatments for alcohol dependence (e.g., 12-step programs, specialized outpatient treatment programs, and pharmacotherapy) in this report, we summarize the evidence regarding such treatments below (the section titled Treatments for Alcohol Dependence) to provide some contextual information. Because screening for alcohol misuse will inevitably identify some individuals with alcohol dependence, providers and those making recommendations need some information about whether there are effective interventions available for such individuals.
Summary of Main Findings
Screening for Alcohol Misuse
We did not find any studies directly addressing Key Question 1 (What is the direct evidence that screening for alcohol misuse followed by a behavioral counseling intervention, with or without referral, leads to reduced morbidity, reduced mortality, or changes in other long-term outcomes?) or Key Question 3 (What adverse effects are associated with screening for alcohol misuse and screening-related assessment?). We searched for trials that randomized or assigned subjects to screening compared with another screening approach, no screening, or usual care, but none were found.
For Key Question 2 (How do specific screening modalities compare with one another for detecting alcohol misuse?), we found adequate evidence that several screening instruments can detect alcohol misuse in adults with acceptable sensitivity and specificity. A single-question screen (covering the past 12 months), AUDIT-C, and AUDIT appear to be the best overall instruments for screening adults for the full spectrum of alcohol misuse in primary care, considering sensitivity, specificity, and time burden. Several instruments require as little as 1 to 2 minutes to administer (e.g., single question screens, AUDIT-C). We present the main findings here by population.
All Adults (Age 18 or Older)
Single-question screens covering the past 12 months have reported sensitivities of 0.82 to 0.87 and specificities of 0.61 to 0.79 for detecting alcohol misuse in adults in primarycare17, 58—similar operating characteristics compared with longer questionnaires, supporting the use of the single-question screen endorsed by the NIAAA.26 Single-question screens typically ask people to report any occasions when they drank four (women) or five (men) drinks or more over a recent time period (past 12 months).
When focusing on adequately sized U.S. studies that reported sensitivity and specificity of screening for the full spectrum of alcohol misuse in primary care, data suggest that some often recommended cut-points for screening (i.e., AUDIT≥8) may need to be revised. The AUDIT had sensitivity of 0.44 to 0.51 for identifying the full spectrum of alcohol misuse in adults using a cut-point of ≥8; more optimal balance of sensitivity and specificity were seen at cutoffs of 4 or 5 (at a cutoff of ≥4: 0.84 to 0.85 and 0.77 to 0.84, respectively; and at a cutoff of ≥5: 0.70 to 0.92 and 0.73 to 0.94, respectively). Further, sex-specific cutoffs may be warranted as sensitivities for women at cutoffs of ≥4 and ≥5 were quite low (0.47 to 0.65 and 0.35 to 0.53, respectively), but improved at ≥3 (to 0.70 to 0.79 with specificity of 0.86 to 0.87).
Young Adults and College Students
The included systematic reviews identified only one study reporting the sensitivity and specificity of a screening instrument for this group, the full AUDIT (≥8), which reported a sensitivity of 0.82 and specificity of 0.78 for identifying risky/hazardous drinking.
Pregnant Women
The AUDIT-C performed better than other instruments with available data for detecting both at-risk drinking and abuse or dependence, demonstrating both high sensitivity (0.95 or higher) and high specificity (up to 0.85).
Adolescents
None of the included systematic reviews provided information about the use of screening instruments in adolescents. Note that our methods for identifying all potentially relevant studies for this Key Question have some limitations: we did not review all individual publications assessing screening instruments. Instead, we relied on previously published systematic reviews to find information and we filled in gaps with data from other sources (i.e., Technical Expert Panel members, peer and public reviewers, personal files).
Behavioral Counseling Interventions in Primary Care
This section summarizes the main findings regarding behavioral counseling interventions (Table 30) and their strength of evidence (SOE) from Key Questions (KQs) 4a, 4b, and 6 (KQ 4a: How do behavioral counseling interventions, with or without referral, compare with usual care for improving intermediate outcomes for people with alcohol misuse as identified by screening?; KQ 4b: How do specific behavioral counseling approaches, with or without referral, compare with one another for improving intermediate outcomes for people with alcohol misuse as identified by screening?; KQ 6: How do behavioral counseling interventions, with or without referral, compare with one another and with usual care for reducing morbidity (e.g., alcohol-related morbidity, alcohol-related accidents and injuries), reducing mortality, or changing other long-term (6 months or longer) outcomes (e.g., health care utilization, sick days, costs, legal issues, employment stability) for people with alcohol misuse as identified by screening?). The findings are presented by population and are summarized in Tables 31 and 33 through 35 below.
Table 30
What are brief behavioral counseling interventions delivered in primary care settings?
Table 31
Summary of effectiveness of behavioral interventions compared with controls for all adults: Intermediate, health, utilization, and other outcomes.
Table 33
Summary of effectiveness of behavioral interventions compared with controls for older adults: Intermediate, health, and utilization, and other outcomes.
Table 35
Summary of effectiveness of behavioral interventions compared with controls for pregnant women: Intermediate, health, utilization, and other outcomes.
All Adults (Age 18 or Older)
We found that behavioral counseling interventions improved self-reported alcohol consumption, heavy episodic drinking, and drinking above recommended amounts (moderate SOE). We found an average reduction of 3.6 drinks per week for adults receiving interventions compared with those in control groups and an 11 percent increase (absolute difference between intervention and controls) in the percentage of adults achieving recommended drinking limits over 12 months. This translates to a number needed to treat of 9 to get 1 person to change from risky/hazardous drinking to drinking beneath recommended limits over 12 months (Table 32).
Table 32
Projected range of outcomes of screening 1,000 adults in primary care and providing a behavioral counseling intervention for those identified with risky/hazardous drinking.
Behavioral counseling interventions also improved some health care utilization outcomes (including hospital days and costs: all low SOE). For most health outcomes, available evidence either found no difference between interventions and controls (e.g., mortality: low SOE) or was insufficient to draw conclusions about the effectiveness of behavioral interventions compared with controls (e.g., alcohol-related liver problems, alcohol-related accidents, and quality of life: insufficient SOE).
To assess the differential effects of interventions using more or less time and those using single or multiple contacts, we grouped interventions by intensity, as measured by duration and number of contacts: very brief (up to 5 minutes, single contact), brief (more than 5 and up to 15 minutes, single contact), extended (beyond 15 minutes, single contact), brief multicontact (each contact up to 15 minutes), and extended multicontact (some contacts beyond 15 minutes).
The evidence for effectiveness in adults is strongest for brief multicontact interventions; these studies consistently found statistically significant improvements in consumption, heavy drinking episodes, and achieving recommended drinking limits. The brief multicontact interventions were generally 10-15 minutes per contact. The effect sizes for brief multicontact interventions were greater than for other intensities (although confidence intervals generally overlapped). In addition, the best studies show that the effect of brief multicontact interventions remains for several years of followup,97, 98, 107 and show improvement for some utilization outcomes (fewer hospital days97, 98) and costs (benefit-cost ratio of 39:1 over 48 months, 95% CI, 5.4, 72.598).
Brief single-contact interventions were effective for improving some intermediate outcomes in adult populations (i.e., achieving recommended drinking limits and reduction in drinks/week), but not others (i.e., heavy drinking episodes). Effect sizes were smaller than those for brief multicontact interventions for the outcomes showing benefit (e.g., 8% vs. 15% achieving recommended drinking limits and reduction of 3.7 vs. 4.4 drinks per week at 12 months). Although extended multicontact interventions appear to be effective for improving intermediate outcomes, we did not find evidence that they are more effective than brief multicontact interventions. Very brief interventions (up to 5 minutes, single contact) are likely not effective.
Long-term outcomes for consumption, heavy drinking episodes, and achievement of recommended drinking limits were available from two studies: Project TrEAT88, 98, 99 and Project Health.107 Both studies reported that participants in the intervention group maintained reductions in consumption or continued to reduce consumption further, but differences between intervention and control groups were no longer statistically significant by 48 months. These studies identified a relatively delayed reduction in consumption in control groups to levels achieved by the intervention group, which could reflect the natural history of alcohol consumption, the cumulative effect of yearly followups with the health care system, attrition (if more subjects lost to followup from the control group were risky drinkers than those lost to follow up from the intervention group), or (late) regression to the mean.
We conducted subgroup analyses to explore whether the effectiveness of interventions differed by sex, country, the person delivering the intervention, or setting. Our subgroup analyses found similar benefits for men and women and for studies conducted in the United States compared with those conducted in other countries. We found a trend toward a greater reduction in consumption for interventions delivered mostly by primary care providers [weighted mean difference (WMD) 4.0 drinks/week, 95% CI, 2.6 to 5.4] than for those delivered primarily by research personnel (3.0, 95% CI, 1.0 to 5.0). Similarly, we found a trend toward greater reduction in consumption for interventions delivered in academic/research-oriented settings (WMD, 5.0 drinks/week, 95% CI, 2.5 to 7.6) than for those delivered in community-based settings (3.2, 95% CI, 2.2 to 4.3).
Older Adults (Age 65 or Older)
Two studies112-115 enrolling only older adults provided evidence of the effectiveness of behavioral interventions for reducing consumption and improving the percentage of individuals drinking beneath recommended limits, but effect sizes were smaller than those found for all adults (Table 33). Evidence for health outcomes was insufficient to draw conclusions.
Young Adults and College Students
Trials conducted with college students provided evidence of the effectiveness of behavioral interventions for improving intermediate outcomes and some accident, utilization, and academic outcomes (Table 34). A subgroup analysis of young adults ages 18 to 30 enrolled in Project TrEAT reported fewer motor vehicle events, hospitalization days, emergency department visits for those in the intervention group compared with the control group (low SOE).88 Two studies of Web-based interventions from New Zealand reported some effectiveness for improving academic-related outcomes.116-118 Unlike studies in all adults, that generally found benefits to last for several years for intermediate outcomes, some positive outcomes of interventions for college students found at 6 months were no longer statistically significantly different between intervention and control groups at 12 months. This could be due to the natural history of drinking among college students or could indicate the need for additional booster sessions to maintain benefits.
Table 34
Summary of effectiveness of behavioral interventions compared with controls for young adults and college students: Intermediate, health, utilization, and other outcomes.
Pregnant Women
We found just one study enrolling pregnant women (N=250)120 that met our inclusion criteria. The study did not find a significant difference for reduction in consumption, but found higher rates of abstinence maintained for subjects who were abstinent pre-assessment for the intervention group compared with the control group (Table 35).
A previously published Cochrane Review of psychological and/or educational interventions for reducing alcohol consumption among pregnant women121 included four studies (for a total of 715 pregnant women). The review found no significant differences between groups for most outcomes, and results related to abstaining or reducing alcohol consumption were mixed. Results from some individual studies suggested that interventions may encourage women to abstain during pregnancy. The authors concluded that the evidence suggests that interventions may result in increased abstinence from alcohol, and a reduction in alcohol consumption.121 In addition, they concluded that inconsistent results, the paucity of studies, the number of total participants, the high risk of bias in some of the studies, and the complexity of interventions limits the ability to determine the type of intervention that would be most effective for increasing abstinence or reducing consumption among pregnant women.
We included just one of the four studies from the Cochrane Review in our review. The other studies included in the Cochrane Review did not meet our inclusion criteria because the duration of follow up of subjects was too short (just 2 months) for some studies122, 123 or because the study was not conducted in a primary care setting.124
Our searches identified other studies focusing on pregnant women that did not meet our inclusion criteria.122-139 Several did not take place in a primary care setting, but instead were conducted in other settings, such as those that included jails and specialized drug and alcohol treatment centers; these included, for example, the Project CHOICES study.133 Others were excluded because they did not include a control group or because they followed participants after the intervention for less than 6 months.122, 131 Several of these studies reported benefits of behavioral interventions for pregnant women, including reduction of alcohol consumption,122, 131 reduced risk of an alcohol-exposed pregnancy,133 higher rates of abstinence,124 and better fetal and newborn outcomes (birthweights and birth lengths, and fetal mortality rates).124
Potential Adverse Effects of Behavioral Counseling Interventions
Potential adverse effects of screening and behavioral counseling interventions for alcohol misuse have received little attention in published studies. For Key Question 5 (What adverse effects are associated with behavioral counseling interventions, with or without referral, for people with alcohol misuse as identified by screening?), we found no studies reporting on illegal substance use, stigma, labeling, discrimination, or interference with the doctor-patient relationship. We found limited evidence reporting no difference between intervention and control groups for smoking rates and anxiety (low SOE). Studies reporting increased smoking or anxiety outcomes generally did not provide actual outcome data and often had little or no description of the procedures used for measuring the outcomes.
One study reported opportunity costs of $39 for enrolled subjects due to lost work time and travel related to the intervention.97
The time required for interventions used in the included studies ranged from a minimum of 5 minutes to a maximum of approximately 2 hours dispersed over multiple in-person and/or telephone visits (moderate SOE). The brief multicontact intervention used in Project TrEAT (which provides some of the best evidence of effectiveness of behavioral interventions for risky/hazardous drinking in primary care) required two 15-minute visits with the primary care physician 1 month apart and two followup phone calls from a nurse.
Although trial data are limited regarding adverse effects of screening and behavioral interventions for alcohol misuse in primary care settings, other types of studies may offer some insights. Among a group of 24 general practitioners in Denmark who were interviewed about their participation in a screening and brief intervention program for alcohol misuse, nearly all reported experiencing negative reactions from some patients.74 Such reactions ranged from feelings of uneasiness or embarrassment to finding another physician. The physicians themselves noted that the added work of screening and brief intervention was onerous and hampered the establishment of rapport with patients. They also expressed concerns that screening identified people for whom intervention was not necessary, yet took valuable time and resources, while at the same time failing to detect and help some for whom alcohol misuse was a real problem. However, other studies have found that patients view screening favorably, even perceiving higher quality of care when screening is followed by counseling.140 For example, one prospective cohort study found that communication and whole-person knowledge were perceived as better among patients who were counseled about their alcohol misuse compared with those who were not counseled.141
Treatments for Alcohol Dependence
Although we did not systematically examine the efficacy/effectiveness of various treatments for alcohol dependence (AD) (e.g., pharmacotherapy, 12-step programs, and specialized outpatient treatment programs), we provide contextual information regarding such treatments in this section. Because screening for alcohol misuse will inevitably identify some individuals with AD, providers and those making recommendations need some information about whether effective interventions are available for such individuals. However, a detailed review and comparison of treatments for alcohol dependence is beyond the scope of this review. We also summarize whether research demonstrates efficacy of pharmacotherapy for patients with AD who are identified by screening in the primary care setting) or treated in primary care settings (as opposed to treatment-seekers or those identified by other methods).
An important point, and one germane to the present review, is that very few studies have examined the efficacy of brief interventions for AD in a primary care setting. A systematic review of the literature concluded that there was no evidence of efficacy for brief behavioral interventions in patients with AD in a primary care setting.142 Similarly, our review did not find any studies demonstrating efficacy of behavioral interventions for people with AD in a primary care setting; studies included in our review that enrolled more than 10 percent of subjects with AD reported behavioral interventions to be less effective or ineffective compared with studies not enrolling subjects with AD. Thus, whereas the overall evidence for the effectiveness of treatment for AD is considerable,143 the same cannot be said for the effectiveness of brief interventions for AD in primary care settings.
Treatment for AD continues to evolve as research on the effectiveness of various treatments is published, and new treatments, including pharmacotherapy, are introduced and used more frequently. Treatment for AD can be quite effective, though no single best approach has yet proven superior among the variety of available treatment options. Treatment outcomes can be affected by many factors including the following: (1) AD is a heterogeneous illness with considerable variability in outcome and prognosis; (2) comorbidities: multiple physical and emotional illnesses can influence treatment outcomes; (3) there are many forms of treatment, including multiple varieties of psychosocial interventions and several pharmacological interventions; (4) patients have many pathways to treatment, ranging from voluntary care-seeking to legally mandated treatment. This complexity contributes to variance in treatment outcomes and does not permit a simple answer to the overall question--How Effective Are Treatments for Alcohol Dependence? Nevertheless, many individuals with AD, and other alcohol-use disorders, respond well to treatment and predictors of good or bad outcomes have been identified.9 Table 36 lists common treatments for alcohol dependence.
Table 36
Treatments for alcohol dependence.
When assessing the effectiveness of treatment for AD, the selection of the outcome measure is a key issue. Complete abstinence has long been viewed as the only meaningful indicator of treatment effectiveness, and abstinence remains the primary goal of treatment for AD given that continued low-level drinking may place the patient at risk for future problematic drinking.144
Using complete abstinence as an outcome, from 15 to 35 percent of patients have been reported to achieve 1 year of sobriety following a variety of treatment approaches.143 Treatment approaches reviewed have included clinical trials of disulfiram, motivational enhancement therapy, cognitive behavioral therapy, and 12-step facilitation, as well as treatment as usual within alcoholism-treatment centers. Sobriety outcomes at 3 to 5 years or longer have been reported to be in a similar range.9 However, the long-term efficacies of specific treatment approaches have not been systematically compared with one another in randomized trials, making interpretation and recommendations for specific interventions difficult.
Over the past 15 to 20 years, awareness has grown that treatment may still be beneficial even if complete abstinence is not achieved. As a result, research has used other outcomes to measure the effectiveness of treatment, which can be subsumed under the concept of harm reduction.145 These measures include significant increases in abstinent days or decreases in heavy drinking episodes, improved physical health, reductions in health care costs, and improvements in psychosocial functioning. Research using these nonabstinent outcomes provides additional evidence for the effectiveness of treatment for alcohol dependence. Miller et al. (2001)143 analyzed seven large multisite trials that tested the treatment approaches noted in the prior paragraph and found that whereas, in aggregate, about 25 percent of individuals maintained sobriety over 1 year, in the remaining nonabstinent individuals there were substantial decreases in drinking days, from 63 percent pretreatment to 25 percent post-treatment and a mean 57 percent decrease in drinks per drinking day.
In recent years, with the Food and Drug Administration (FDA) approval of additional medications for AD, pharmacotherapy has received increasing attention. From the 1950s until the 1990s the pharmacotherapy for AD consisted of disulfiram—an aversive deterrent that produces significant physical symptoms, such as nausea, when alcohol is consumed. Disulfiram can be an effective adjunct to psychosocial treatment for AD, though its effectiveness seems to require a high degree of patient motivation, thereby limiting its overall usefulness. Since the 1990s two oral medications, naltrexone and acamprosate, and a long-acting intramuscular formulation of naltrexone have been approved by the FDA for AD. These medications target neurobiological systems thought to be involved in the pathophysiology of alcoholism (e.g., naltrexone blocks the alcohol-induced “high” in some patients presumably by blocking the action of β-endorphin, which is released by alcohol consumption). In clinical trials these medications have shown evidence for efficacy in enhancing abstinence, reducing relapse to heavy drinking and reducing overall drinking behavior.146 The average effect sizes for these medications are considered low to moderate (from 0.11 to 0.16 for effects on abstinence or heavy drinking for naltrexone and acamprosate) when heterogeneous populations of patients with AD are studied,146 which has led to efforts to identify individual predictors of response to both naltrexone and acamprosate, with some signs of success. For example, Anton et al. (2008) found that alcoholic individuals who were carriers of the Asp40 allele of the μ-opioid receptor had an 87.1 percent good outcome with naltrexone compared with only a 48.6 percent good outcome for those who received placebo, whereas noncarriers demonstrated no naltrexone/placebo difference. Kim et al. (2009) and Oslin et al. (2003) also reported that the Asp40 allele was predictive of improved naltrexone response in alcohol dependence whereas Gerlenter et al. (2007) did not find this relationship. Mitchell et al. (2007) and Arias et al. (2008) failed to find an association of the Asp40 allele with treatment response to naltrexone or nalmefene in heavy drinkers. While clearly requiring additional confirmation and extension, these findings suggest that individual characteristics such as genetic polymorphisms may eventually prove of value to choosing a particular pharmacotherapy for a specific patient. The NIAAA recommends that medications be considered as part of the overall treatment approach to patients with AD along with psychosocial treatment.
Studies of pharmacotherapy for patients with AD have generally enrolled subjects responding to advertisements or those being treated in specialty alcohol treatment centers. We were unable to identify any double-blind randomized controlled trials (RCTs) of pharmacotherapy that identified subjects by screening in a primary care setting or that assessed the efficacy or comparative effectiveness of pharmacotherapy in a primary care setting. Further, we were unable to identify any studies of pharmacotherapy for people with risky/hazardous drinking.
Applicability
The findings are generally applicable to people with risky/hazardous drinking identified by screening in primary care settings (see beginning of Discussion). It is uncertain whether findings are applicable to harmful drinkers or people with alcohol abuse. Most studies excluded all or most potential subjects with alcohol dependence; thus, our findings for behavioral interventions in primary care settings likely do not apply to people with alcohol dependence, who probably require other treatments (e.g., referred for specialty treatment; see section on Treatments for Alcohol Dependence). Compared with the results of studies that enrolled few or no subjects with alcohol dependence, our subgroup analyses found that studies enrolling 10 percent or more subjects with alcohol dependence found behavioral interventions to be ineffective or less effective. This supports the theory that people with alcohol dependence are not likely to respond to the types of interventions evaluated in this report. Most studies enrolled some subjects with heavy episodic drinking patterns of consumption, and one study focused only on those with binge drinking.89 Overall findings and those from the one study focused on binge drinking were consistent in finding interventions to be efficacious for reducing heavy episodic drinking.
We did not identify any studies in adolescent populations or any conducted exclusively in veterans, and the results thus have uncertain applicability to these populations. We did, however, identify a sufficient number of studies of young adults/college students and older adults to draw conclusions (of low to moderate strength) for several intermediate outcomes for these populations. Although we searched for studies conducted in settings with primary care–like relationships (e.g., nontraditional primary care settings such as infectious disease clinics for people with HIV), we did not find any, and our results have uncertain applicability to such settings.
All interventions required support systems to provide screening and screening-related assessment, and, in some cases, provider prompting. Screenings to identify subjects for the included studies were often extensive, multistep processes that included face-to-face interviews lasting up to 30 minutes by research personnel. Less time would be required for screening and screening-related assessments in primary care practice; we estimate less than 2 minutes for negative screens and 5 to10 minutes for positive screens, with most of the time for screening-related assessment to determine whether the patient has an alcohol use disorder as opposed to risky/hazardous drinking. Nevertheless, supports are likely required for effective screening and intervention. In addition, most interventions required training of providers and/or staff. Such training may be required to ensure that practices conduct effective screening and interventions for alcohol misuse.
Effective interventions were generally delivered either completely in person or also included phone followups. However, one study of adults in Germany demonstrated some benefits resulting from a telephone-based intervention,90 and two studies conducted in college student populations demonstrated benefits resulting from Web-based interventions delivered via computer.116-118
It is unclear whether our findings are applicable to people with comorbid medical or psychiatric conditions, including those with multiple substance use disorders, and some researchers have suggested that brief behavioral interventions may be ineffective or less effective in people with comorbid psychiatric conditions. A subgroup analysis (N=88) from a study conducted in Germany found that brief interventions did not significantly reduce drinking for subjects with comorbid anxiety and/or depression.93
We did not find any evidence that would inform decisions about the appropriate frequency of screening (i.e., whether it should be done annually, every 5 years, or another interval).
Limitations
The scope of this report is limited to primary care settings. Emergency departments or other health care settings may also provide opportunities to provide behavioral interventions to reduce alcohol misuse.
For Key Question 2 (“How do specific screening modalities compare with one another for detecting alcohol misuse?”), we did not review all individual publications assessing screening instruments. Instead, we relied on previously published systematic reviews to find information on their sensitivity and specificity and filled gaps with data from other sources. In addition, our review did not attempt to systematically evaluate biomarkers for screening [e.g., gamma-glutamyl transferase (GGT) or carbohydrate deficient transferrin (CDT)].
Studies were generally not designed to assess the impact of the interventions on morbidity and mortality; their focus was primarily on behavioral outcomes. In addition, most of the evidence we identified in this report was in the form of intermediate outcomes that rely on self-report of alcohol use. Some studies verified self-report using collaterals, such as a family member. Although no biomarkers are accurate enough to be widely accepted to measure changes in alcohol use, self-report of alcohol use has been found to be accurate if collected carefully.73, 147 Nevertheless, it remains a concern that social desirability bias could play a role in the results of the included studies (i.e., although self-report is from both randomized groups in these studies, the group that gets more attention and advice to decrease their drinking may be more likely to report that they decreased their drinking). When grading the strength of evidence, we considered self-reported measures of alcohol use to be indirect (i.e., not the direct health or utilization outcomes that we are most interested in improving); thus, for situations when evidence had a low risk of bias and was consistent and precise, we graded the strength of evidence for intermediate outcomes as moderate rather than high.
It is possible that the assessments of alcohol misuse conducted in the included trials conceal therapeutic benefits of the behavioral interventions (i.e., bias results toward the null). Many studies included extensive assessment of alcohol-related behaviors, which could directly result in behavior changes. The control groups in the included studies generally reduced alcohol consumption. Some possible explanations for changes in behavior as a result of the screening and screening-related assessment include (1) increased awareness of the extent of their drinking; (2) the screening questions prompted them to discuss drinking with their primary care provider at a subsequent visit; (3) receipt of some minimal intervention, such as printed educational materials about general health or about alcohol specifically (control groups in the included studies often received some printed materials); or (4) regression to the mean. One study empirically tested whether brief assessment (without a behavioral intervention) reduces hazardous drinking by comparing brief assessment with a control that did not include assessment. The study concluded that assessment appears to reduce hazardous drinking but noted a potential limitation of measurement artifact due to social desirability bias.117
Key Question 7 was confined to examining RCTs that were included in the other questions in this report (RCTs primarily examining the efficacy or effectiveness of screening and brief intervention). This report does not address dissemination and implementation literature that may shed further light on health care system influences that promote or hinder effective screening and interventions for alcohol misuse.
Future Research
We identified numerous gaps in the evidence, which future research could address. We identified no studies that randomized subjects, providers, or practices to screening compared with no screening to answer Key Questions 1 or 3. A cluster RCT of practices/health centers could perhaps address this gap in the literature. We found insufficient evidence to draw conclusions about the impact of screening and behavioral interventions on followup with referrals. Future studies could assess referral to treatment for alcohol dependence for people identified by screening in primary care, evaluating whether they follow up with referrals and whether it works when they get there. We also found very few studies that measured health or utilization outcomes, with overall insufficient or low strength of evidence for the impact of behavioral interventions on mortality, morbidity, utilization, costs, and quality of life. We found very limited data on potentially harmful effects of behavioral interventions, making it difficult to determine whether interventions to reduce alcohol use lead to increases in smoking, illegal drug use, or anxiety. Also, none of the included studies reported on stigma, labeling, discrimination, or potential interference with the doctor-patient relationship.
Although we concluded that brief multicontact interventions have the best evidence of effectiveness, direct comparative evidence (i.e., studies directly comparing various behavioral intervention approaches) was generally insufficient to make firm conclusions about which intensity of intervention is most effective (i.e., how many visits are needed? how long do they need to be? what specific components must be included?). We found no studies evaluating a very brief (each contact 5 minutes or less) multicontact intervention, and it is unknown whether very brief multicontact interventions would be as effective as the brief multicontact interventions identified in this report (generally 10 to15 minutes per contact). Knowing the minimum amount of time needed for an intervention to be effective is very important for busy primary care practices, where a positive screen triggering a brief intervention could take up the entire allotted time for the visit to discuss alcohol misuse—and might mean postponing the original purpose of the visit. Future studies could possibly compare the intervention delivered in Project TrEAT (two 15-minute visits with the primary care physician and followup calls by a nurse) that provides some of the best available long-term evidence for the effectiveness of behavioral interventions with a shorter version of the same intervention (using interventions of 5 minutes or less).
Future studies could provide more guidance for individualizing therapy for various populations. The included studies generally did not provide information to determine the characteristics of individuals who responded positively to interventions as opposed to those who did not. Future studies could explore whether the individuals who are reducing consumption are those who have a low risk of developing adverse health or social outcomes, a high risk, or both. Long-term studies and a better understanding of the natural history of alcohol misuse would be needed to address this question. Future studies could also explore whether people meeting criteria for alcohol abuse are more or less likely than those with risky/hazardous drinking to respond to interventions, or whether people with alcohol abuse or those with alcohol dependence receive any benefit from behavioral interventions delivered in primary care settings. Future research could also determine whether our findings are applicable to people with comorbid medical or psychiatric conditions—and could explore whether people with comorbid psychiatric conditions (e.g., anxiety, depression, or serious mental illness) respond to behavioral interventions delivered in primary care settings.
Finally, we found no double-blind RCTs of pharmacotherapy for alcohol dependence that identified subjects by screening in a primary care setting or that assessed the efficacy or comparative effectiveness of pharmacotherapy in a primary care setting. Future studies could fill this void in the literature.
Conclusions
Behavioral counseling interventions improve intermediate outcomes (i.e., alcohol consumption, heavy episodic drinking, drinking above recommended amounts: moderate SOE) and some health care utilization outcomes (including hospital days and costs: low SOE) for adults with risky/hazardous drinking. For most health outcomes, available evidence either found no difference between interventions and controls (e.g., mortality: low SOE) or was insufficient to draw conclusions about the effectiveness of behavioral interventions compared with controls (e.g., alcohol-related liver problems, alcohol-related accidents, quality of life: insufficient SOE). Brief multicontact interventions (generally 10 to 15 minutes per contact) have the best evidence of effectiveness for adults (compared with single-contact interventions or very brief 5-minute interventions).