Study design.

We included systematic reviews (with or without meta-analyses) which we defined as syntheses that: 1) collate evidence that fits pre-specified eligibility criteria in order to address a specific research question; 2) report explicit systematic methods; and 3) include a comprehensive search strategy of at least two electronic databases to identify primary studies. Reviews of either experimental or observational studies, or both, were eligible, but reviews of qualitative studies only were excluded.

Intervention.

We defined alcohol control and public health interventions as those prevention interventions or policies which are implemented at a population or community level, and can be conceivably incorporated into legislation. Any of the following interventions included as one of the WHO recommendations for governments to address the harms due to alcohol consumption [5] were eligible:

• Drink-driving policies and countermeasures.
• Availability of alcohol.
• Marketing of alcoholic beverages (including online and social media platforms).
• Pricing policies.
• Reducing the public health impact of illicit alcohol and informally produced alcohol.
• Community action.
• Reducing the negative consequences of drinking and alcohol intoxication.

Where overlap existed in categorising an intervention (e.g. community-based programmes to reduce drink driving) we preferentially selected one category under which to report the review and report this in the text.

Prevention programmes focused within educational settings (universities, colleges or schools) and prevention and treatment programmes provided within the healthcare sector were excluded as delivery of interventions in these settings is less likely to be the focus of legislation. Interventions provided at an individual level only were also excluded.

Review selection and assessment of review quality.

Two investigators (NS and CP) independently selected potentially eligible systematic reviews from the records yielded in the search. The full-text article was obtained for selected reviews and for those where these was uncertainty. Both investigators independently applied a standardised form to each article to assess eligibility according to the inclusion criteria: 1) systematic review design; 2) included prospective studies; 3) general population (whether the intervention was amenable to legislation); and 4) alcohol consumption or related harms measured (see S1 File for eligibility form). Any disagreements were resolved through discussion. Both investigators independently evaluated the methodological quality of each eligible systematic review using the ROBIS tool which permits classification of the conduct of reviews as at high, low or uncertain risk of bias. Risk of bias is dependent on an evaluation of review validity in four domains: 1) study eligibility criteria; 2) identification and selection of studies; 3) data collection and study appraisal; and 4) synthesis and findings [4]. Any disagreements in the ratings were resolved by discussion, noting ambiguities in the ROBIS tool where this was present. As ROBIS is a new tool we contacted the development team on several occasions for clarifications and advice in addition to the available guidance. We calculated the inter-rater agreement for the ROBIS tool for each of the four domains and for the overall risk of bias assessment using Cohen’s kappa coefficient (for three categories: high, low and unclear risk of bias). We interpreted the agreement using the Altman scale [7]. To address a conflict of interest for a single review where both investigators of this umbrella review were also authors [8], ROBIS was conducted by a senior researcher with expertise in ROBIS but not involved in this umbrella review.

Intervention categories.

Thirty-four reviews evaluated a single alcohol control intervention category with the remaining eight reviews evaluating multiple intervention categories. Restricting availability of alcohol was evaluated in 15 reviews, drink-driving policies were evaluated in 10 reviews, pricing policies were evaluated in 10 reviews, interventions to reduce the negative consequences of drinking and alcohol intoxication were evaluated in seven reviews, marketing policies were evaluated in three reviews, community action was evaluated in three reviews, and a harm reduction approach and reducing the use of illicit alcohol were each evaluated in a single review respectively. Sub-categories of interventions within the broad WHO categories can be seen in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Included studies and ROBIS quality assessment.

https://doi.org/10.1371/journal.pone.0214865.t001

Publication.

The mean number of systematic reviews published per year was 4.2 (standard deviation (SD) = 1.6) with the greatest number of reviews (N = 6) published in each year of 2011, 2014 and 2017. No eligible reviews were published in 2007. The trend test for publication over time was R² = 0.4936. Almost all eligible reviews were published in English, except for one in German [9] and one in Portuguese [11].

Types of studies included in the reviews.

None of the systematic reviews limited inclusion criteria to randomised controlled trials. All sought to search for, and include, a broad range of study designs. However, most reviews identified only observational studies relevant to the focal intervention(s) (74%; 31/42). Of the nine reviews which also identified randomised controlled trials (RCTs), seven reviews identified three or less relevant RCTs. A review of interventions delivered within drinking environments identified seven relevant RCTs [17] as did a review of exposure to alcohol advertisements [18]. There was no reporting of the study design of included studies in two reviews.

Location of studies included in the review.

Thirty-five reviews adopted a global focus when searching for primary studies, with five reviews limiting inclusion to studies from high-income countries and two limiting inclusion to specific countries viz. USA and a combination of Denmark, Finland, Hong Kong, Sweden and Switzerland. Three and two primary studies were identified from the upper-middle income countries of China and Brazil respectively, and a single primary study was identified from each of the upper-middle-income countries of Colombia, South Africa, Thailand and Mexico. No studies were identified from low or lower-middle income countries.

Electronic searches and grey literature.

Together the 42 reviews searched for studies across 51 different electronic databases. The most common databases were Medline, PsychInfo and EMBase searched in 41, 27 and 24 reviews respectively. The Cochrane Library was searched in 16 reviews and the Web of Science in 13 reviews. The economic database, EconLit, was searched in eight reviews focused on pricing and taxation. The nursing database, CINAHL, and the social science database, Sociological abstracts and social services abstracts, were searched in eight and seven reviews respectively.

Nineteen reviews stated that the search was not limited by publication status and included additional means to identify studies from the grey literature. Four reviews did not report whether or not publication status was a limitation, of which two provided further details suggesting that additional unpublished studies were included. Methods to identify unpublished studies included conducting ancestry reviews of reference lists (N = 11), and searching relevant websites (N = 7), Google Scholar (N = 6), and trials registries (N = 3). Four reviews described conducting hand-searching which included 1) searching journals not indexed in the major databases and available only in specialist addiction libraries [19], 2) searching an extensive collection belonging to the review authors [20], 3) hand-searching the journal Traffic Injury Prevention, publications of the International Council on Alcohol, Drugs, and Traffic Safety’s Working Group on Alcohol Ignition Interlocks, and the proceedings of the International Symposia on Ignition Interlocks [21], and 4) hand-searching all issues of the journals Traffic Injury Prevention and Accident Analysis and Prevention [22].

Language limitations.

Eighteen reviews reported limiting inclusion to studies reported in English. Sixteen reviews reported that inclusion was not limited to studies reported in English: of these, three reviews stipulated that inclusion was limited to studies reported in English or German [9], English or Chinese [23], and English, Portuguese or Spanish [11]. It was not possible to determine language limitations in the remaining eight reviews.

ROBIS findings.

The four and overall ROBIS domains are presented in Table 1. Ten reviews (24%) were assessed as at low risk of bias, of which five were Cochrane systematic reviews and one reported applying Cochrane review methods but was not published as a Cochrane review.

We classified 75% (31/41) of reviews as at high risk of bias (we were unable to categorise one review due to the poor quality of reporting). Of these, eight conformed to the rigorous methods outlined by the United States non-federal Task Force on Community Preventive Services (Task Force) [24], but were classified as high risk due to limiting the search to English language publications and, in one such review [25], despite reporting limitations in the study quality, the interpretation of the overall evidence was categorised as ‘strong’. In the remaining 23 reviews which did not conform to Cochrane or Task Force methods, the reasons driving our decision to classify the risk of bias as high included alone or in combination: 1) publication and language limitations; 2) study selection and/or data extraction conducted by a single investigator; 3) lack of quality assessment of included studies and lack of integration of quality assessment into the results; and 4) contradictions between the presented data and the authors’ conclusions.

Agreement between investigators.

Inter-rater agreement for the ROBIS overall risk of bias was calculated for the 40 reviews which were rated by both investigators. The Cohen’s kappa coefficient for the overall risk of bias was 0.34 interpreted as fair agreement using the Altman benchmark scale [7]. For Domains 1 to 4 the coefficients were 0.51 (moderate), 0.38 (fair), 0.65 (good), and 0.51 (moderate) respectively.

Community action interventions.

Three reviews evaluated qualitatively different multicomponent interventions which were all delivered in the community [17,25,26]. None of the reviews conducted meta-analyses and all were rated as high risk of bias.

Community mobilisation: In 2009, Shults et al. [25] evaluated the effects of community mobilization on reducing alcohol-impaired driving specifically. Community mobilization was defined as organization and activation of a community to address local problems as part of multicomponent programmes which included a combination of responsible beverage service activities, enforcement of minimum legal drinking age laws, controlling alcohol outlet density, sobriety checkpoints, public education and media advocacy. The review authors concluded that evidence was ‘strong’ in favour of multicomponent programmes when implemented with community mobilization for reducing alcohol-related crashes, despite identifying several shortcomings within the studies and stating that there was a lack of unequivocal evidence. Given this contradiction, combined with a high risk rating on ROBIS, our assessment was that such programmes are possibly beneficial and that a full update and revision of the review is required. In the interim implementation of such programmes should include monitoring and evaluation.

Interventions implemented in the drinking environments: A 2011 review by Jones et al. [17] evaluated interventions implemented in drinking environments and was rated as at high risk of bias primarily due to a lack of clear reporting on duplicate and independent data extraction and publication limitations. The authors integrated the quality assessment of included studies with the numerical results and concluded that multicomponent intervention programmes combining community mobilisation, responsible beverage server training, house policies and stricter enforcement of licencing laws are possibly effective in reducing assaults, traffic crashes and underage sales, but there is less certainty for other interventions including patron-targeted interventions and stand-alone server training. Our assessment of their reported results confirmed their findings as possibly beneficial in favour of multicomponent interventions. The review requires updating with more recent studies and until such time, implementation of such programmes should be done within a monitoring context.

Indigenous community-led legal interventions to control alcohol: The 2017 review by Muhunthan and colleagues [26] evaluated 18 observational studies of indigenous community-led legal interventions to control alcohol and concluded that these can be effective in improving health and social outcomes. Given the high risk of bias in this review due to a lack of quality assessment of included studies and publication limitations, our assessment was that evidence is currently uncertain and that a full review revision with both quality assessment and integration of results using a GRADE approach is required prior to implementation and to inform future research studies.

Interventions to reduce unrecorded alcohol.

A single 2011 review without a meta-analysis evaluated the effects of policies to reduce the impact of unrecorded alcohol use and concluded that there is currently no clear evidence base on the effectiveness (or cost-effectiveness) of available policy options [20]. The review included less than 30 studies, described as ‘mostly observational’, and no quality assessment of included studies was undertaken. The ROBIS rating was high risk of bias. Our assessment concurred with the authors that evidence remains uncertain, and the review requires updating and revision with a quality assessment and integrated approach, to determine the need for, and design of, future controlled studies and implementation strategies within monitoring contexts.

Regulation of marketing beverages.

Two reviews with meta-analyses met the inclusion criteria to evaluate reducing or banning the marketing of alcohol and its effects on alcohol consumption through an interventional lens [8,18]. Both were rated as at low risk of bias, but assessed consumption at different time periods. The 2014 review by Siegfried et al. included three controlled interrupted time series studies of state- and provincial-wide banning of alcohol advertising in combinations of print, electronic and billboard media, and long-term consumption and concluded that due to methodological deficiencies in the dated studies (1976, 1980 and 1991), the current evidence base was uncertain. As co-authors of this review, we acknowledge that the review requires an updated search, and in the interim implementation of country-wide banning of alcohol advertising should be implemented with ongoing monitoring to allow for adequate analyses.

The 2016 review by Stautz et al. considered immediate alcohol consumption following exposure to television or movie alcohol advertising in a meta-analysis of seven laboratory-based RCTs. The review authors conclude that viewing alcohol advertisements may increase immediate alcohol consumption by a standard mean deviation of 0.20 (95% CI: 0.05, 0.34), interpreted as an increase of 1.57 (95% CI: 0.39, 2.67) alcohol units [18]. Given the low risk of bias and the integration of the results with a thorough quality assessment, we agree that restricting alcohol advertising is possibly beneficial in the short-term (limited to a few hours). Additional controlled studies are required in non-student populations to evaluate generalizability prior to scale-up.

Availability.

Fifteen reviews evaluated availability interventions including licencing restrictions on trading times, minimum drinking age, outlet density and distance, retail privatization and dram shop liability with significant overlap between reviews [9,11,17,27–38]. Two reviews were rated as low risk of bias (both published before 2015) and of these, we report the more recent review preferentially.

Bryden et al. published a well-conducted review in 2012 and state that overall the narrative results for restricting availability were inconclusive but that higher outlet density may lead to increased alcohol use. Our assessment concurs that the evidence is uncertain for licencing restrictions (including banning sales, and making changes to the hours, days and volumes of alcohol sales) but that reducing outlet density is possibly beneficial. The review requires updating with more recent studies and additional longitudinal primary research is required. Implementation of policies should undergo monitoring and evaluation.

Drink-driving interventions.

Nine reviews evaluated six interventions aimed at addressing drunk driving [11,21–23,25,39–43].

Blood alcohol concentration limits: Three reviews evaluated the effects of blood alcohol concentration (BAC) limits on traffic injuries [11], mortality from motorcycle crashes [42], and levels of BAC in Chinese drivers respectively [23]. All were assessed as at high risk of bias. None of the reviews included overlapping primary studies or outcomes and we therefore report on all three review findings briefly.

The 2014 review focused on all traffic injuries and reported that no beneficial effects were found unless the BAC limit was combined with other interventions [11]. The 2016 review specific to the effects of BAC limits on motorcycle crash mortality included two longitudinal studies and reported the results to be inconsistent but concluded that it was a potentially effective intervention [42]. The China-specific review included a very large 2009 controlled-before-after study of enhanced BAC enforcement among 32,101 drivers which reportedly found that the intervention was successful in decreasing driver BAC (but traffic injury was not measured as an outcome) [23].

Given the above inconsistencies in review results, the differences between the included studies and the high risk of bias observed in all three reviews, we assessed the overall effect of BAC limits (when not combined with other interventions) as uncertain. A single over-arching review is required with an updated search for primary studies, an overall quality assessment, and integration of this with the results. In the interim implementation should be conducted in a monitoring environment and additional controlled studies are required.

Police patrols: Two reviews, published in 2008 and 2009, evaluated the effectiveness of police patrols in reducing alcohol-related crashes [39,40]. Given that both reviews are out of date, we selected the 2008 Cochrane review as the primary data source given its low risk of bias rating. The authors included 1 RCT and 31 observational studies and concluded that there is consistent evidence that police patrol programmes reduce traffic crashes and fatalities, but noted that due to limitations in study quality and data analysis, the evidence was supportive but not unequivocal [40]. Our assessment was that police patrols are possibly beneficial. The review requires updating and we agree with the authors that further well-designed controlled studies are required. Implementation should occur in a monitoring environment.

Mass media campaigns: Two reviews focused on mass media campaigns, a 2014 review evaluated the effects on increasing awareness of publicized sobriety check-points [43] and a 2015 review assessed the effects on alcohol-impaired driving and alcohol-related crashes [41]. The 2015 review, an update of a previous review [44], was rated as at low risk of bias so we selected to summarise these results. The review conducted a meta-analysis which did not show any improved risk of alcohol-related injuries or fatalities from the intervention (RR = 1.00, 95% CI = 0.94–1.06). The review authors note the heterogeneity among the media campaigns, in the methods used and in the outcome measurements, as well as the observed statistical heterogeneity in the pooled results. Given these concerns, the authors state that they cannot conclude that media campaigns reduce the risk of related alcohol-related crashes or injury. Our assessment is that the evidence is uncertain and we concur with their recommendation that better controlled studies should be conducted, especially with respect to newer media methods.

Screening of occupational drivers: Two reviews evaluated testing occupational drivers for alcohol to prevent injury or work-related effects. A 2009 Cochrane review, rated as at low risk of bias, concluded that there was insufficient evidence for or against alcohol testing [22]. A 2014 review, rated as high risk of bias, searched for studies published after 2006 and identified a single uncontrolled before-after study and reported that fatalities were reduced by 80% in truck drivers and by 41% in other drivers involved after compulsory alcohol testing of drivers [11]. Given the inconsistency in these results and the need to update and combine the data synthesis, we assessed the current evidence as uncertain and advocate for an updated and revised review to establish the evidence base and to determine if there is a need for conduct of further controlled studies or ongoing monitoring.

Administrative licence revocation: A 2016 review, rated as at high risk of bias, included two longitudinal studies (see also under BAC limits) that considered administrative licence revocation (e.g. if individuals refuse a blood alcohol test) to reduce motorcycle crashes specifically [42]. The studies found inconsistent results and the authors conclude that the intervention is potentially effective, despite reporting that there is doubt whether the intervention has a significant effect on crash and injury rates. Given the risk of bias, the limited number of studies and the inconsistency reported in results, we assess the evidence as uncertain and recommend the review be revised with integration of the study quality into the conclusions. Implementation should be done with ongoing monitoring.

Ignition Locks: A 2011 review updated a 2004 Cochrane review [45] with grey literature searching (no revised electronic database searching was conducted) and evaluated the effects of ignition interlocks for preventing alcohol-impaired driving and crashes in drivers convicted of driving under the influence of alcohol [21]. The review was rated as at high risk of bias and the authors concluded that there is ‘strong’ evidence that installation is associated with large reductions in re-arrest rates for alcohol-impaired driving but evidence was insufficient regarding the effect on alcohol-related crashes. We noted the large effect reported consistently across primary studies but given that limitations of the observational studies were not adequately integrated into the results and the high risk of bias rating, we assessed the evidence as possibly beneficial and a revised, and updated review is required.

Reducing the negative consequences of drinking and alcohol intoxication.

Seven reviews appraised five interventions aimed at reducing the negative consequences of drinking and alcohol intoxication. None included meta-analyses.

Interventions for disorder and severe intoxication in and around licenced premises: A 2011 review was rated at high risk of bias (primarily due to a lack of clear reporting regarding review methodology) [46]. Interventions included responsible beverage server training, server violence prevention training, enhanced enforcement of licencing regulations, multi-level interventions, licensee accords, and a risk-focused consultation. The review authors found that server training courses designed to reduce disorder show potential, but lack evidence to support reducing intoxication. We concur with the authors’ conclusions that the evidence base is uncertain overall and possibly beneficial for server training courses. However, the review requires updating with recent studies and revision incorporating a GRADE-like approach.

Alcoholic beverage labels: Two reviews evaluated labels on alcoholic beverages: a 2012 review focused on the effects of alcohol warning labels in adolescents [47] and a 2017 review evaluated the effects of labels stating ‘low alcohol’ [48].

The review of warning labels was rated at high risk of bias. No quality appraisal was conducted, although the authors did consider the limitations of the cross-sectional nature of the studies on the overall evidence. They concluded that warning labels may show increased awareness among adolescents, but have little effect on individual beliefs of risk or on alcohol-related behaviours. Our assessment is that the evidence is uncertain and the review requires updating and a revision with a validated quality assessment of studies and integrated GRADE approach. In the interim, the effects of implementation should be monitored.

The 2017 review of ‘low alcohol’ labelling employed Cochrane methods and included a single non-randomised cross-over controlled trial [48]. The review concluded that evidence is uncertain. We noted that the evidence is current and of a high quality and agree with the authors’ conclusions that the evidence is uncertain and that there is an urgent need to conduct well-designed controlled studies of this intervention.

Portion size: A 2015 Cochrane review evaluated the effect of portion size on alcohol consumption [49]. This review was rated as at low risk of bias. No eligible studies were identified and as such, the authors identify a lack of evidence. We assessed the evidence as current and concur with the authors’ conclusion that well-designed controlled studies are required prior to policy implementation on this issue.

Interventions in sport settings: Two reviews evaluated policy interventions delivered in sports settings, both were rated as at low risk of bias and no meta-analyses were conducted [50,51]. The 2016 review was conducted according to Cochrane methodology and identified two cluster RCTs of policy interventions [51]. The review authors conclude that the effects are inconsistent and that given the paucity of studies in this area, more well-conducted controlled studies are required. We assessed the evidence as current and concur with the authors’ conclusion of the evidence as uncertain.

Public health interventions specific to pregnant women: A 2015 review of public health interventions aimed at reducing alcohol consumption and/or increasing knowledge among pregnant women was rated as at high risk of bias. The authors considered the impact of the quality of studies on the results, concluding that there is little evidence available on the effectiveness of pregnant-specific interventions [52]. We assessed the current evidence as uncertain, and concur with the authors’ conclusions that well-conducted controlled studies are required.

Taxation and pricing.

Ten reviews reported in 11 articles and one abstract evaluated the effects of taxation and pricing, or both, on alcohol consumption or related harms [9,23,36,38,53–59]. Several reviews had specific foci including youth [9], China [23], intimate partner violence [36] and binge drinking [55]. All, except the review on youth, were rated at high risk of bias.

Increased pricing and taxation: We selected to report in full on the four reviews conducted since 2015 given that all were at high risk of bias. Three reviews, all authored by Nelson et al., were conducted in 2015 [55], 2016 [54] and 2017 [38] and utilised data from the same database supplemented with discrete searches to evaluate the effects of price and taxation. The reviews were limited by including only English publications, a lack of duplication of data extraction, and no assessment of study quality or integration thereof into the results.

The 2015 review included 56 econometric studies and evaluated the consistency and significance of findings across studies, concluding that binge drinkers are not highly responsive to taxation or pricing. The responsiveness was not quantified and we conclude that for binge drinking specifically the evidence remains uncertain.

For the 2016 and 2017 reviews conducted to assess harms in nine countries and consumption in five countries, 45 and 29 studies were included respectively. The authors found considerable diversity and inconsistency in the research base for price and taxation to reduce alcohol consumption and harms. They state that they do not argue that prices have no effect on alcohol consumption, but that the effects depend on temporality, location, and population factors. We assessed the evidence as possibly beneficial, and recommend revision with adequate quality assessment, meta-analysis and regression where possible to account for the potential confounders, and integration of the overall quality into the interpretation of the results.

Price only (minimum unit pricing): A 2017 review focused on minimum alcohol unit pricing (MUP) excluding studies on taxation and elasticity [58]. The review appraised the data using Bradford-Hill causality criteria and integrated the quality assessment into the final conclusions. Meta-analysis was not conducted. It was rated at high risk of bias due to uncertainty regarding duplication of screening and data collection and limitations to English language. The authors conclude that given the lack of controlled studies, it is highly probable, but not definite, that MUP reduces alcohol consumption and alcohol-related harms. We determined the evidence to be possibly beneficial and recommend that further implementation of MUP occur within a monitoring environment.

Tax on retailer: A 2017 review investigated taxation on retailers and manufacturers of unhealthy food products. The review identified 102 studies for inclusion [59], but only a single case study of taxation of large alcohol and tobacco retailers based on the value of their premises. The study reported that revenue raised was predictable and above government expectations, but the impact on alcohol-related consumption was not evaluated. The review authors conclude that retail taxation above 20% on unhealthy food products in general reduces consumption, but the evidence base for alcohol is a single study with no evaluative component. We assessed the evidence for alcohol as uncertain. The review requires revision with conduct and integration of quality assessment into the final results, and additional controlled studies are required with implementation within a research/monitoring context.