Most studies that met inclusion criteria targeted university students (12/17 or 70.6%), although some recruited general company employees [38,39] or community members [40-42]. While the university-based studies generally involved participants aged 18 to 25, in other studies, the median reported age was 43.1 years. The size of recruitment pools and participation rates varied substantially. This appeared to be mediated in part by the study’s target population, the presence of incentives, the marketing and recruitment strategy, and whether participation was mandated (Table 1). In fact, Matano et al [38] distributed 8567 invitations to achieve 316 preintervention surveys.

Study sample sizes ranged from 40 to 3216 (median 196) with 64.7% (11/17) of the RCTs targeting at-risk, heavy, or binge drinkers. The percentage of females ranged from 27.6% to 77.9% (mean 54.5%, median 52%), which is substantially greater than in most AOD clinics (Table 1).

Of the studies that met criteria, 70.6% (12/17) evaluated the impact of brief personalized feedback, and 41.2% (7/17) examined an online multi-module information/education treatment (often incorporating personalized feedback). Control groups typically received psychoeducational resources (10/17 or 58.8%) or completed an online assessment.

Posttreatment assessments were conducted anywhere from 1 week to 12 months posttreatment, with several studies conducting assessments at multiple time points. Across the 17 studies, 7 (47.1%) had a maximum follow up period of a month, 4 (23.5%) had a maximum 3-month follow up, and 3 (17.6%) followed participants to 6-months post intervention. Only Kypri et al [35] employed a 12-month follow up.

Reported retention rates in the intervention groups ranged from 38.9% to 100%, and in controls, from 33.4% to 100%. Median reported retention in the treatment condition was 83.4% at 1 month, 74.5% 3 months, and 74.5% at 6 months. In control groups, the median retention rates at the same time points were 80%, 70.4%, and 74.9%. The Kypri et al study [35] reported 12-month retention rates of 83.5% for the intervention group and 86.3% for the control condition.

Several studies reported only combined retention data. The studies by Doumas and Hannah [39] and Doumas et al [43] reported 1-month whole sample retention rates of 63.3% and 88.2% respectively, with the Walters et al [23] study reporting retention rates of 71.7% at 2 months and 77.4% at 4 months. The Motano et al [38] study reported whole sample retention rates of 83.8% at 3 months.

A wide variety of outcome assessments were employed across the studies with all studies including some measure of alcohol consumption (eg, unit grams of alcohol, number of standard drinks, or blood alcohol concentrations) in relation to either a typical drinking occasion or when the greatest amount was consumed on a single occasion. In many cases, the measure of frequency of alcohol consumption used was either 4 or more or 6 or more drinks per occasion or drinking to intoxication. Quantity and frequency measures related to a designated assessment period (a typical week, the previous week, 2 or 6 weeks, or up to the last 12 months). Several studies assessed alcohol use in relation to specific events (eg, 21^st birthdays [44], homecomings, holidays, or pub nights [45]). A number of studies included measures of personal, social, sexual, or legal consequences of drinking [34,36,46,47], protective factors [45], alcohol-related knowledge [19,36], readiness to change [40,45,48], intention to seek help [48], drinking related self–efficacy [46], or outcome expectancies [36,46].

Based on 5 RCTs [41-43,47,49] where a measure of alcohol units per week or month could be extracted, differential effect sizes to posttreatment ranged from 0.02 to 0.81 (median 0.54) (Table 2). Using the full samples of participants, the mean differential effect size was 0.42. If only identified problem drinkers in the Cunningham study [41] are included (rather than the full sample dataset), the effect size rose to 0.47. The pre-post differential effect size for brief personalized feedback programs [41,43,47,49] ranged from 0.02 to 0.81 (mean 0.39, median 0.33), and for the multi-session modularized programs of Riper et al [42] and Hustad et al (AlcoholEdu) [47], a pre-post differential effect size of 0.56 was obtained in each case.

Employing Cohen’s effect size evaluation benchmarks [52], the effects on alcohol units consumed were generally in the small to medium range. A notable exception was the study by Hustad and colleagues [47] undertaken with a college sample. That study examined the effectiveness of 2 of the most commonly used electronic interventions for heavy drinking in college students: AlcoholEdu, an education style multi-session modularized Internet program, and Alcohol eCHECKUP TO GO (eCHUG), an intervention based on personalized normative feedback. In relation to peak drinks per occasion per month, the eCHUG personalized normative feedback intervention resulted in a large differential effect size (0.81).

Only 1 RCT allowed extraction of a follow-up effect size on alcohol units. The examination by Cunningham and colleagues [41] of the online CheckYourDrink.net website (10-minute online normative and alcohol severity feedback) resulted in a small differential effect size of 0.23 for the full sample at the 6-month follow up. When only the data from identified problem drinkers was included, a moderate differential effect size (0.43) was obtained.

Pre-post data on peak blood alcohol concentrations (BAC) were available from 2 RCTs [44,47] (Table 3) and resulted in a mean differential effect size of 0.66. The personalized normative feedback program of Hustad et al [47] for eCHUG and that of Neighbors et al [44] achieved differential effect sizes of 0.87 and 0.22 respectively. Interestingly, the eCHUG intervention produced a BAC effect size (0.88), comparable to that of the more extensive modularized 3-hour online alcohol program, AlcoholEdu tested in the same trial.

Differential effect sizes were extracted from 5 RCTs [39,43,44,49,51] in relation to number of drinks on 21^st birthday [44], units of alcohol per occasion [49], peak consumption [39,43], frequency of drinking to intoxication [39,43], or 30-day quantity of alcohol use [51] (Table 3). Most studies obtained small effect sizes on these measures. However, of note is the Moore et al [51] study that employed a control group that differed from the intervention group primarily in mode of delivery (via postal services) rather than in key content. In this study, there was a greater fall in 30-day quantity of alcohol use (number of drinks per occasion) for postal delivery than for Internet delivery (differential effect size = -0.26). This, however, reflected higher consumption by the postal control group at baseline (mean 3.15 vs mean 2.49): Posttreatment alcohol quantities were comparable across the groups (postal 2.51, Internet 2.53).

The use of online interventions for the treatment of alcohol-related problems requires more extensive research to establish the clinical appropriateness and usability of online health technologies [29], especially in nonstudent contexts. Given the potential benefit of these interventions for cost-effective delivery of interventions to large numbers of people, future research should incorporate economic analyses. As suggested by Copeland and Martin [24], the rigorous evaluation of online interventions would encourage their wider implementation and dissemination and increase their impact on public health and related service costs.

A significant challenge for this field is that advances in equipment, connectivity, and software capabilities are occurring much more rapidly than the evidence base can be fully established. In this context, recommendations for practice must necessarily rely to some extent on analogies from evidence that has been obtained on similar interventions using older forms of delivery. However, transfer of interventions to new modes of delivery run the risk of losing the key effective ingredients. It remains important that researchers respond rapidly to new technological advances, adapting treatments and routinely conducting trials to ensure that effects on alcohol use are retained.

As with all remotely delivered interventions, engagement of participants remains an issue. Internet-based interventions are likely to have greater reach if they are interfaced with targeted marketing campaigns or are embedded in routine primary care. Further research on the most effective marketing and widespread dissemination of these interventions is required.

While the current research evidence is fragmented and requires greater methodological rigor, it suggests that problematic or at-risk users may benefit from online alcohol interventions and that they may be a useful preventative and first step for groups such as women or young people who may not otherwise access more traditional AOD health services. Our confidence in these interventions is boosted by decades of research on bibliotherapies [55] and face-to-face interventions for alcohol use, including robust evidence in favour of brief interventions [56]. While further randomized controlled trials are required, there is sufficient evidence to suggest that standard health services and community campaigns evaluate and deploy online alcohol interventions to address alcohol-related problems.