Key Question 1: Is there direct evidence that behavioral counseling interventions to reduce risky or harmful alcohol use reduce morbidity and/or mortality?

We identified four reports of health outcomes associated with long-term followup of adult interventions. ^{107 – 110} Fleming et al. reported outcomes at 48 months among 282 male and 292 female adults drinking more than 14 drinks/week (men) or 11 drinks/week (women). ¹⁰⁸ Drinkers were randomized as part of a trial in 17 community-based primary care practices to receive either usual care or 2 brief (15-minute) physician intervention visits about one month apart, each followed by 2 brief (5-minute) telephone calls by a nurse. ¹¹¹ The intervention visits used a workbook protocol that focused the interaction on drinking cues, a drinking agreement written as a prescription, and self-monitoring through drinking diary cards. Long-term (48 month) outcomes included self-reported alcohol use, self-reported health care utilization, self-reported accidents and injuries, and legal events and deaths ascertained from state agencies. Compared to the control group, the intervention group maintained statistically significant reductions in drinking outcomes (weekly alcohol use, number of binge episodes) seen at earlier time points. The intervention group self-reported fewer emergency department visits, fewer days in the hospital, and fewer motor vehicle accidents or moving violations, although only self-reported hospital days were statistically significant (429 vs. 664 days, p<.05). Intervention group participants incurred 32% fewer legal arrests (28 vs. 41), but only the subgroup of arrests for controlled substance/liquor violations were significantly different between intervention and control (2 vs. 11, p<0.05). Seven deaths were reported in the control group and 3 in the treatment group at 48-months follow-up (p>0.10).

Berglund ¹⁰⁹ reported on the long-term follow up of the Malmo Screening and Intervention Study (MSIS), a controlled study that conducted screenings for cardiovascular risk factors and high alcohol intake between 1974-1992. All middle-aged males and females in the community were identified by birth year, and cohorts in alternating birth years were invited to health screenings at the outpatient Department of Preventive Medicine, with about 70% of invited cohorts attending. Nurses performed examinations that involved physical measures, functional tests, risk factor identification, and laboratory tests. Those in the invited group who screened positive for increased serum gamma-glutamyl transpeptidase levels (GGT) and history of excessive alcohol consumption (about 11% of all those screened) were offered repeat MD office visits (1/2 hour) as often as every 3 months over at least 18 months, preceded by repeat lab and consumption measures. MDs gave feedback concerning lab results and consumption measures, advised patients to drink moderately, and encouraged self-monitoring for situations/cues to heavy use. A treatment goal of moderation rather than abstinence was agreed upon. ¹¹² Participants were also contacted monthly by an RN during active treatment. Over a 5-year period, active treatment was offered until lab tests and/or consumption normalized, and was then reduced to maintenance (2 MD visits and 2 RN contacts per year). Those in control cohorts received usual care. Subjects in both invited and non-invited cohorts were followed through 1995, beginning at the same age for individuals in both cohorts and using local registers to examine death rates and causes. Although there have been various followup reports over the years for this study, ^{109, 113 – 115} earlier reports are not included in the rest of this review by virtue of not collecting intermediate alcohol consumption measures on all participants, by high attrition rates, or by reporting results for subgroups. The 2000 report covers mortality measures on all individuals in both invited and non-invited cohorts, after individual followup periods from 3–21 years that were comparable between intervention and control cohorts. Using intention-to-treat analysis, total and cause-specific mortality rates did not differ between intervention and control groups among men. However, significantly fewer male intervention cohort members (18/10,000 person-years) than male control cohort members (23/10,000 person-years) had “other death mortality” (p<.02). “Other deaths” were those attributable to causes other than ischemic heart disease, stroke, other cardiovascular disease (CVD), total CVD, or cancer. Younger men at baseline (aged 32–37) had both lower total mortality rates (24/10,000 person-years compared to 30/10,000 person-years) (p<.02) and lower “other death” cause-specific mortality rates (p<.03) in intervention versus control cohorts, respectively. Mortality rates did not differ between cohorts for female participants, however the average length of followup was shorter than that for males.

A more recent report from a nested randomized controlled trial within Malmo Screening and Intervention Study ¹⁰⁷ examined cause-specific mortality rates among a subgroup of men aged 45–49 with elevated serum gamma-glutamyl transferase levels (GGTs) at the initial screening examination who were followed a median of 13 years. Those with 2 elevated GGTs within 3 weeks were randomized to receive the active intervention (as described above) (n=365) versus letter notification and follow up at 2 years (n=302). A second control group was formed from those with a single elevated GGT who did not meet study inclusion criteria (n=311). Deaths were determined from a centralized registry, and a high proportion (88–100%) of deaths were autopsied during the followup time period. Overall death rates were 12.7% (38/365 in IG, 42/302 in CG1, and 44/311 in CG2), and were not reported as differing statistically between groups. Deaths were divided into four discrete causes: 1) alcohol-related deaths (ARD)—including accidents in an alcohol-intoxicated state, cause unknown in known alcoholics, death in known alcoholics with tissue evidence of alcoholic intoxication and no other specific demonstrable cause of death, deaths from organic complications of alcoholism, or suicide/pneumonia in known alcoholics; 2) coronary heart disease (CHD); 3) cancer; and 4) miscellaneous. Alcohol-related death rates were higher (52%) in both control groups than in the intervention group (37%) [risk ratio=2.0 (1.1–3.6), p=.026] and were higher than CHD death rates in control groups. CHD deaths exceeded alcohol-related deaths only in the intervention group. Among those with ARD in all groups, most had received treatment for alcoholism at some point.

Among risky/harmful non-dependent drinkers in the Australian arm of the WHO trial (n=554), 9-month reductions in average drinking consumption and unsafe drinking patterns in intervention vs. control subjects did not result in long-term differences in alcohol consumption (median alcohol consumption, mean reduction in consumption from baseline), in morbidity, problem drinking, or mortality rates, or in diagnoses of abuse/dependence at 10 year follow-up. ¹¹⁰ Younger drinkers (mean age 35 years vs. 41.6 years at baseline) were differentially lost to followup. The authors note that there is no evidence that brief advice and counseling without regular followup and reinforcement can sustain long-term reductions in drinking behavior, and therefore any associations with improved health outcomes.

Few fair-to-good quality intervention studies included in this review reported morbidity outcomes such as problem scores, ^{116 – 119} psychological scores, ^{116, 118} and other beneficial outcomes, such as improvements in lifestyle, and reduced accidents or injuries ^{111, 116, 120, 121} in the original time frame of the study. In 2 of 4 studies examining problem scores, both groups generally improved over the duration of the trials, but no differences were seen at followup between interventions and controls. ^{116, 117} The other 2 (male and female arms of the same study design) showed no changes from baseline to followup within or between groups. ^{118, 119} The 5 studies ^{111, 116, 118, 120, 121} measuring changes in outcomes such as general health, smoking depression, and accidents or injuries found no changes over the trials or differences between intervention and control groups.

Another adult intervention study ¹²² has reported a post hoc analysis of the association between changes in drinking and quality of life and alcohol-related problems. ¹²³ A self-selected, statistically comparable subgroup (213/301) of those randomized to receive either usual care, brief, or extended brief intervention completed quality of life (SF-36, reported as the physical component summary, PCS, and the mental health component summary, MCS) and alcohol-related problem (Short Inventory of Problems, SIP) measures at baseline and 12 months. Since all groups equally improved their drinking from baseline to 12-month followup, all drinkers were combined and stratified by whether or not they had reduced their drinking by 30% or more over that time period. Seventy-seven (36%) of subjects sustained a 30% or more decrease in their drinking; these subjects averaged 85.5 drinks/month (SD 77.3) at baseline and 22.5 drinks/month (SD 29.0) at 12-month followup, while those that didn't decrease 30% or more had steady alcohol intakes of 66.0 (SD 61.2) and 65.3 (SD 59.1) drinks/months at baseline and 12 months respectively. Over 12 months, PCS, MCS, and SIP scores were relatively unchanged in those whose drinking was not reduced by at least 30%; compared to these, drinkers who reduced their average intake by 30% or more showed significantly improved MCS and SIP scores (p<.002) even after removing those with definite abuse/dependence. When the cut points were changed to include those with a 20% or greater reduction of alcohol consumption (n=98), MCS and SIP improvements remained statistically significant. Requiring a 40% or greater reduction of alcohol consumption (n=65) to define “improvers” retained the statistically significant comparative improvement in SIP scores, but eliminated the improvement in MCS scores. PCS scores appeared to improve primarily in those with definite abuse/dependence who reduced their drinking. These findings should be considered preliminary as they are limited by the post hoc nature of the analyses that addressed “responders” regardless of intervention condition and by virtue of including only two-thirds of the original participants who agreed to complete the additional outcome assessments analyzed in this study.

Key Question 2: What methods were used to identify the target populations for the behavioral counseling interventions?

We examined the recruitment, screening, and study participant identification methods used in the 12 studies included in our main evidence table for adults. Brief descriptions of these methods and procedures, inclusion and exclusion criteria related to alcohol consumption patterns and alcohol problems, and descriptive data on the numbers of participants included or excluded at various study stages are depicted in Table 11 of the Appendix.

Table 11

Screening and screening-related assessment procedures used in the brief intervention trials included in this review.

In all except 3 studies, ^{116, 118, 124} alcohol use was assessed as part of a battery of other lifestyle behaviors in an effort to maintain participant and/or provider blinding as to the intent of the study or to diffuse initial patient resistance to addressing alcohol consumption. Almost all screening and screening-related assessments were done by research personnel, prior to the intervention visit, by self-assessment in the waiting room, ^{117, 124} by mail, ¹²⁵ by telephone or face-to-face interview, ^{119, 126} or by a combination of these. ^{111, 116, 118, 120, 122, 127, 128} Studies used various instruments (AUDIT, CAGE, HSQ), and different subsets of questions to ascertain quantity, frequency, maximum use per occasion, and other alcohol-associated queries. In general, the types of patients identified for these intervention studies, particularly those conducted more recently in the United States, were less heavy, risky, or dependent drinkers than would be detected by use of validated screening tests, such as AUDIT, alone. Many studies also included patients if their quantity and frequency estimates exceeded recommended low-risk weekly thresholds for quantity (e.g., >12–14 drinks/week for men and > 7–9 drinks per week for women) or any indication of recent heavy episodic drinking (binge use).

Although the trials employed various recruitment and screening methods and procedures, several commonalities are apparent. First, several adapted the methods of Wallace et al. ¹²⁸ This study used a two-stage screening and screening-related assessment procedure. The first stage involved a self-administered questionnaire, which embedded drinking questions among other health behavior items and which included both drinking quantity and frequency questions, the CAGE instrument, and an item allowing self-identification of having a drinking problem. The second stage, a longer, in-person interview, asked detailed questions about each day's drinking in the past week. Inclusion criteria for the first stage were based on usual consumption that exceeded gender-specific limits, on indications of alcohol-related symptoms or consequences (from the CAGE), or on self-description of having a drinking problem. Inclusion criteria at the second stage were based on total consumption in the past week, and those whose consumption exceeded gender-specific limits were included in the trial. Close variants of these two-step identification procedures were used by 5 subsequent studies. ^{111, 116, 118, 120, 127} The U.S. studies ^{111, 120, 127} often used a time-line followback (TLFB) procedure, considered by many to be the gold standard for assessing alcohol consumption. ¹²⁹ This procedure is generally too time-consuming, however, for use in routine primary care.

Studies differed in the use of specific alcohol consumption measures (e.g., past week consumption vs. average consumption; or average consumption over different time frames), and in whether they specifically assessed for and included those with heavy episodic use (bingeing); when “bingers” were included, the definitions for these also varied. Studies also differed in whether or not they used established screening instruments (e.g., AUDIT, CAGE). The AUDIT was used as the primary identification tool in 2 studies, ^{122, 124} but even in these it was supplemented with additional items or criteria to more clearly identify heavy usual consumption or heavy episodic drinking. Most studies used established screening instruments as adjuncts to various quantity, frequency, and use-per-occasion measures. ¹²⁰ – ^{122, 126 – 128}

Two studies ^{124, 126} also used the AUDIT score to exclude patients and/or refer them for evaluation for specialty alcohol treatment. None of the studies relied primarily on the CAGE instrument to identify risky drinkers, using it instead as a supplement to quantity-frequency items.

Among more recent U.S. studies that identified those with risks due to heavy episodic drinking as well as recent heavy average use, the screening yields of potentially eligible patients were 11–18% for patients undergoing screening and screening-related assessment as part of a two-step process. ^{111, 120, 126, 127} Where data were available for the second step among these, about half of these patients were found to be eligible to receive primary-care based intervention after screening-related assessment. ^{111, 120, 127} This approximates the final screening yields reported (6.9–7.7%) for studies that recruited primary care patients from the waiting room, ^{119, 124} and suggests that about 7% of general primary care patients might be candidates for these brief interventions. Although many studies did not report active refusal rates among participants, those that did suggest only small proportions refused. ^{124, 127}

Key Question 3: What are adverse effects associated with alcohol use screening and screening-related assessment?

No data are available on any false-positive or negative effects that occur from generalized screening for alcohol misuse among patients. Another concern about potential harms related to screening for alcohol consumption revolves around uncertainties about whether patients may be offended. Greater dropout among those receiving brief interventions for risky alcohol use compared with controls in some studies ^{124, 126, 128} provides indirect evidence that such concerns may be warranted for small proportions of individuals. Alternatively, those receiving interventions that did not change their drinking may have dropped out due to embarrassment. However, we were unable to find research directly addressing potential adverse effects for either screening or brief interventions for alcohol problems.

Key Question 4: Does behavioral counseling intervention in primary care reduce risky or harmful alcohol use in the targeted subgroup?

We located 5 meta-analyses or systematic evidence reviews published between 1994 and April 2002, that addressed very brief to extended interventions to reduce risky/harmful alcohol use. ⁹⁶ – ^{99, 101} All of these were judged to meet the USPSTF criteria for good systematic reviews, although they differed in questions addressed, methodology used, or research base available at the time of the review (see Appendix: Detailed Review of Other Recent SERs/Meta-Analyses). Kahan ⁹⁶ reviewed the results of 11 randomized controlled trials (RCTs) on the effect of brief physician intervention with problem drinkers in a health care facility. The 4 with the highest validity scores ^{116, 118, 128, 130} were all in primary care settings and showed decreases in weekly alcohol consumption (5–20 standard drinks per week) and higher assumption of moderate drinking levels (7–19%) for men. The results for women were mixed. Wilk ⁹⁷ conducted a meta-analysis of 12 RCTs testing “brief” (less than 1 hour) alcohol interventions in heavy drinkers (heavy consumption, problem use, or those with mild alcohol abuse/dependence). Five of the 6 highest-quality trials ¹¹⁶ – ^{118, 128, 131} were all conducted in primary care settings; these 6 trials had a summary odds ratio for reduced or moderated drinking of 1.91 (95% 1.61–2.27) among heavy drinkers receiving “brief” intervention. In these trials, 44% of the intervention group achieved alcohol moderation compared to 28% of the controls. Poikolainen ⁹⁸ conducted a meta-analysis of 7 RCTs to examine the effectiveness of “very brief” (5–20 minutes) versus “extended brief” (up to several visits) physician interventions on average alcohol intake or serum GGT changes in problem drinkers (excluding alcoholics) from primary care populations. Outcomes were analyzed by intervention intensity (“brief” versus “extended brief”) for treatment arms comparing men and women separately. Three groups of comparisons (for very brief interventions in men and women, and extended brief interventions in women) were statistically homogeneous, but only extended brief interventions in women showed a statistically significant decrease in alcohol intake (-51 grams/week or ½ drink/day). Dinh-Zarr, DiGuiseppi, and others ¹⁰¹ published a systematic review of 18 interventions for preventing injuries in problem drinkers (those with alcohol dependence, alcohol abuse, and hazardous use of alcohol) for the Cochrane Injuries Group. ¹³² None of the studies that met our inclusion criteria also reported injury outcomes. Moyer ⁹⁹ completed a meta-analysis of 54 controlled clinical trials, 34 of which compared “brief” (4 or fewer sessions) interventions for alcohol use disorders with usual care controls or more extensive treatment. For studies with 6–12 months of followup, the average effect size for the composite of all drinking-related outcomes was .241 (95% C.I., .184, .299). The effect size for alcohol consumption in these same studies was .263 (95% C.I. .203, .323), but the results were statistically heterogeneous. As a convention, such effect sizes are generally interpreted as showing small-medium effects in favor of intervention. ^{133, 134} Very few studies (5/34) reported followup greater than 1 year.

For our review, we identified 44 potentially eligible trials. We eliminated 27 studies that were poor-quality (n=17) ^{115, 131, 135 – 149} or were conducted in a non-primary care setting (inpatient setting or emergency department) (n=10) ^{150 – 159} (see Appendix: Table 10). As a result, we included 16 trials (12 targeting adults; 3 targeting pregnant women; 1 targeting adolescents) that met all inclusion/exclusion criteria for review. One adult study ¹⁶⁰ met criteria but could not be included because data were not published at the time of this report (see Appendix: Table 12). We report on study results for each population group separately.

Table 12

Studies Awaiting Assessment and Ongoing Studies.

Adults

The 12 fair- to good-quality trials of behavioral counseling interventions in adults are summarized in Table 8 of the Appendix. ^{111, 116} – ^{120, 122, 124 – 128} All of the studies were randomized controlled trials except for one population-based study from Norway. ¹²⁵ All but one study ¹²⁵ were conducted in multiple primary care practices (ranging from 3–47 practices per study). Eight were conducted in community-based primary care settings, one in an academic medical center, ¹²⁷ 2 in an HMO setting ^{124, 126} and 1 in a range of outpatient settings in 8 countries. ¹¹⁹ Most studies involved a large number of participants (n>500). All except 2 ^{119, 125} clearly targeted participants who were seeking or had recently received routine primary care. Women represented a third or more of study participants in all US studies but were less well represented in international studies that did not target females specifically. ^{119, 125, 128} Rates of non-White participation were not reported in many older international studies, and were low where reported in more recent US studies (4–27% non-White participation). ^{111, 124, 126, 127}

Table 8

Evidence for Adults.

The trials generally targeted high-average and/or heavy occasional (binge) users. Alcohol consumption limits for risky/harmful use varied among studies but were lower for women in most studies. ^{111, 116} – ^{120, 122, 127, 128}

Table 13 of the Appendix, shows the baseline consumption for participants across studies by gender when available. “Corrected” columns reflect consumption levels adjusted for whether a standard drink is defined as containing 12 grams or 14 grams of alcohol, the usual range of standard drink sizes in the US. Even after “correction” it is clear that earlier non-US studies ¹¹⁶ – ^{118, 128} tended to recruit heavier baseline alcohol users among both men and women (men 37–44 mean DR/week at baseline; women 25 mean DR/week at baseline) than later US studies; these earlier studies also included some drinkers with “less severe” but definite dependence symptoms compared with more recent US studies, which tended to exclude heavier or dependent drinkers.

Table 13

Baseline Consumption of Alcohol Among Risky/Harmful Drinkers Recruited for Interventions in Primary Care (Mean DR/week) *.

Interventions

Studies primarily tested interventions delivered all or part by the patient's usual primary care physician. Exceptions to this were the use of research personnel, ¹²² “health advisors” (46% RNs, 18% MDs, 35% other), ¹¹⁹ or unspecified intervention personnel. ¹²⁵ Among the interventions involving a patient's usual primary care physician, some used the physicians to deliver initial and any repeated intervention contacts ^{117, 118, 127} while others also used health educators/counselors, ^{124, 126} or clinic nurses. ^{111, 120, 128} However, most interventions also involved contact with a research person to determine study eligibility through screening and screening-related assessment prior to intervention.

According to our intensity definition (see Appendix, Table 2), 2 studies evaluated very brief interventions, ^{117, 119} 6 studies ^{116, 118, 119, 122, 124, 125} evaluated brief interventions, and 7 evaluated brief multi-contact interventions. ^{111, 120, 122, 125 – 128} Fifteen intervention arms were contributed by the 12 studies (i.e., since three studies adequately tested more than one arm of varying intervention intensities against controls). ^{119, 122, 125} Control conditions involved screening/assessment only, ^{116, 118, 119, 125, 126} screening/assessment followed by usual care, ¹²⁴ provision of a general health pamphlet after screening/assessment, ^{111, 120, 127, 128} or screening/assessment results recorded on the chart ¹¹⁷ or forwarded to a physician. ¹²² We considered that very brief or brief interventions (n=8) were those that could more feasibly be incorporated into current routine primary care with relatively minimal changes, while brief multi-contact interventions (n=7) were those likely to need more substantial primary care reconfiguration and/or systems support in addition to provider behavior change for incorporation into routine care.

Rates of intervention delivery provide an estimate of whether the tested intervention was delivered adequately enough to fairly evaluate it. Most fair- to good-quality studies reported to what degree the intervention was delivered as intended, although several studies did not, ^{116, 118, 119, 125} and 1 study with multiple intervention arms reported intervention delivery only for the most intensive intervention. ¹¹⁷ For the extended brief intervention arm of this study, only 49% of patients got anything beyond the very brief initial intervention contact, diluting the possibility of adequately evaluating the more intensive approach. Some studies reported reasonable completion rates for the entire protocol ^{111, 122, 124, 127} varying from 60–76%, while others reported fewer participants completing the entire protocol. ^{126, 128} Interpreting the lower delivery rates for these 2 studies is complicated because their protocols involve “up to” a certain number of contacts at the discretion of the provider and/or patient. Completion rates for the first 2 contacts of these protocols were 60% or above. Some studies reported much higher completion rates (91–94%), ^{120, 122, 126, 127} but these rates usually reflected only the initial contact of an intervention.

Outcomes. All studies included estimates of intervention impact on average alcohol consumption at 6–12 months followup. One study ¹¹¹ reported longer-term (48-month) followup in 2 separate reports. ^{108, 121} Intervention groups had significantly lower average alcohol consumption in six of the 12 studies, ^{111, 119, 120, 125, 127, 128} and the rest consistently reported minimal to modest decreases in average drinks/week in both interventions and controls that were not statistically significantly different. ¹¹⁶ – ^{118, 122, 124, 126}

Among the very brief and brief single contact interventions, ¹¹⁶ – ^{119, 122, 124, 125} one study testing both very brief and brief interventions ¹¹⁹ (with optional followup at some centers) reported significantly reduced average alcohol consumption among those receiving either intervention compared to controls at 12-month followup. Another study testing both brief and brief multi-contact interventions ¹²⁵ found that both types of interventions significantly reduced alcohol consumption compared to controls, although the potential for contamination between conditions wasn't clear. Most of the very brief or brief intervention studies did not report intervention delivery except 2 ^{122, 124} with adequate to good intervention delivery for the single component intervention. Therefore lack of intervention delivery may confound the failure to find treatment effects for most brief or very brief interventions. Among seven adequately tested brief multi-contact interventions, ^{111, 120, 122, 125 – 128} five ^{111, 120, 125, 127, 128} reported that participants had significantly lower average alcohol consumption than controls at 12 months ^{111, 125, 128 120} or 6 ¹²⁷ months. The other 2 showed no difference in mean drinks (DR)/week among intervention and control participants. ^{122, 126}

Long-term alcohol consumption was measured in 1 study ¹¹¹ for the entire cohort ¹⁰⁸ and for a subset of women of child-bearing age. ¹²¹ Men and women in the intervention group maintained the reduction in alcohol consumption (mean DR/week) achieved by 12-months through 48-month follow up. By 48 months, differences between intervention and control were no longer statistically significant because of late onset (between 36 and 48 months) reductions in control group usage primarily among men. The relatively delayed reduction in control consumption to levels achieved by the intervention group at 12 months could reflect (late) regression to the mean, the natural history of alcohol consumption, or the cumulative effect of yearly followups. ¹⁰⁸ For a subset of women of childbearing age (18–40), ¹¹¹ average alcohol consumption significantly decreased in interventions compared with controls by 12 months. Reduced levels were maintained among intervention patients over 48 months, but were no longer statistically significantly different at 48 months because of late onset (between 24–28) reductions among controls.

To further summarize the effect sizes and consistency of the alcohol consumption results, Figure 4 of the Appendix shows the net change in mean drinks/week (DR/week) for brief interventions (listed first) and brief multi-contact interventions that reported intention-to-treat baseline and followup levels of consumption for intervention subjects compared with controls. ^{111, 116, 118, 120, 122, 126 – 128} Two studies were rated fair quality ^{116, 122} while the rest were rated good quality. The effects range from a net reduction of about 1 to 10 mean DR/week. Among brief multi-contact trials with statistically significant effects, ^{111, 120, 127, 128} patients receiving the intervention consumed 2.9 to 8.7 fewer drinks per week than those in the control conditions. Among effective trials that report results by gender, 2 ^{111, 127} of 3 ^{111, 127, 128} show similar or better effects among women.

Figure 5 of the Appendix shows the net percentage-point reduction in mean DR/week between interventions and controls at followup. The 3 brief intervention arms are listed first, followed by the brief multi-contact interventions. Drinkers receiving effective brief multi-contact interventions reduced their drinking between 13 and 34 percentage points compared with controls. The largest effect sizes (30–34%) were seen among women in the Fleming 1997 study and older adults (65+). ¹²⁰

Figure

Figure 5. Net percentage point reduction in mean drinks per week .

Binge/heavy episodic drinking was reported in a subset of good studies ^{111, 116, 120, 124, 126, 127} and 1 fair-rated study ¹¹⁸ (n=7) (see Appendix: Figure 6). Three good quality studies ^{111, 116, 120} that also reported reducing average alcohol consumption showed reduced numbers of binge users in intervention compared with control, while the rest showed no differences. Across studies, binge use remained fairly common. For example, in trials with at least 49% binge users at baseline, ^{111, 120, 127} binge drinkers remained fairly common (31–69%) among intervention participants.

Figure

Figure 6. Percent Not Binge Drinking at Follow-up.

Three fair-quality ^{117 – 119} and 7 good-quality studies ^{111, 116, 120, 124, 126 – 128} reported some measure of the proportion that benefited from the intervention compared to controls (see Appendix: Figures 7 & 8). These were defined differently across studies, but addressed the proportion drinking within safe/recommended limits, ¹¹⁶ – ^{120, 124} the proportion not drinking excessively, ^{111, 128} the proportion with safe weekly intake and no bingeing, ¹²⁷ and the proportion not reporting any at-risk drinking pattern (including driving after drinking 3 or more drinks). ¹²⁶ Among 6 ^{111, 116, 120, 126 – 128} of the 7 good quality studies, significantly more intervention participants than controls achieved recommended drinking patterns. Across these studies, 10–19% more intervention participants than controls achieved recommended or safe drinking limits. Three studies that did not show a beneficial effect ^{117, 118, 124} were similarly ineffective as measured by average drinking consumption and bingeing outcomes where available. One study ¹¹⁹ showed significant results for males only.

Figure

Figure 7. Percent Achieving Recommended Drinking Levels/Patterns at Followup (Total).

Few fair- to good-quality studies in adults addressed intervention effects besides alcohol consumption and alcohol-related behaviors. These are addressed in Key Question 5.

Key Question 4a: What are the essential elements of efficacious interventions?

Within the adult intervention trials reviewed for KQ4, intervention intensity appeared to explain some of the heterogeneity in the results of the adult trials: 2 of 6 very brief or brief interventions had statistically significant effects on average alcohol consumption, while 5 of 7 trials of more intensive interventions did. Similarly, other alcohol outcomes appeared more likely in brief multi-contact interventions compared with the less intensive interventions. To identify other factors that may be associated with heterogeneity in the results of these trials, we examined how well other intervention elements that have been cited as important to these interventions ^{40, 55, 64, 105} distinguished between studies with positive and negative effects on average alcohol consumption.

Key Question 5: Are there other positive outcomes from behavior counseling interventions to reduce risky/harmful alcohol use?

Few fair- to good-quality studies in adults addressed intervention effects such as health care utilization, ^{111, 116, 118, 121, 124, 166} or cost effectiveness. Of all the studies that examined health care utilization, only 1 found reduced self-reported hospital days among males in the intervention group compared with controls at 12 months of followup. ¹¹¹ At longer-term followup (48 months) in the same study, males and females in the intervention group reported significantly fewer hospital days compared with controls. ¹⁰⁸ These data reflected self-reported estimates of use rather than analyses of administrative data, as did 1 other study. ^{124, 166}

Out of 100 articles identified in the literature search for economic studies addressing counseling interventions to reduce risky/harmful alcohol use, 14 were retrieved, and 12 of these were excluded: seven were not original economic analyses, ^{167 – 173} two included only costs and no measures of effectiveness ^{174, 175} two examined non-primary care interventions ^{176, 177} and one reported measures of effectiveness for interventions that could not be disaggregated. ¹⁷⁸ In addition to the two studies identified through the literature search, ^{179, 180} another article relating to one of these studies was identified through expert contact ¹⁰⁸ and a third study from examining reference lists. ¹⁸¹ These three original cost-effectiveness studies were abstracted and reviewed by a separate cost working group team, but interpretation of their results was limited because of noncomparability between studies in definitions of drinking misuse/problems and intervention designs, and lack of inclusion of non-medical (e.g., quality of life) as well as medical outcomes. Additionally, most studies lacked a societal perspective in terms of both benefits and effects. Despite these limitations, these studies tended to show that brief interventions are either cost savings or have cost-effectiveness ratios that compare favorably to other commonly endorsed interventions. Firm conclusions from these studies are premature, however, given the limitations cited, particularly the variability in inputs, definitions, and differences in assumptions between studies.

Key Question 6: What are adverse effects associated with behavioral counseling interventions for risky/harmful alcohol use?

As discussed above, higher attrition rates among intervention compared with control groups in some studies suggests that, for some individuals, alcohol misuse interventions may be embarrassing or even offensive. There are no data on adverse effects of treatment, including theoretical reduction in “benefits” from moderate use among patients who may be benefiting from non-risky use. Also, some drinkers with more severe abuse or dependence might be undertreated and harmed if guided towards moderation rather than abstinence. However, literature searches targeted to potential harms retrieved no articles, and no information to address these questions was reported in the screening reviews or the intervention studies included in our review.

Key Question 7: What healthcare system influences are present in effective screening, and screening-related assessments and interventions to reduce risky/harmful alcohol use and/or its outcomes?

Interventions function in a “context” of the network of resources, services, and conditions in the larger health care system and community. ¹⁸² To evaluate generalizability (of individual studies on brief alcohol interventions in primary care) and feasibility (of this approach in the current and future health care system), some level of context evaluation considers factors related to the health care team, the service system and environment, and the broader community.

Within the adult intervention studies reviewed for Key Question 4, the little information reported about contextual factors concerned the health care team and, to some extent, the health care system and environment.