A systematic review on the impact of alcohol warning labels

Abstract

Findings on the effects of alcohol warning labels (AWLs) as a harm reduction tool have been mixed. This systematic review synthesized extant literature on the impact of AWLs on proxies of alcohol use. PsycINFO, Web of Science, PubMED, and MEDLINE databases and reference lists of eligible articles. Following PRISMA guidelines, 1,589 articles published prior to July 2020 were retrieved via database and 45 were via reference lists (961 following duplicate removal). Article titles and abstracts were screened, leaving the full text of 96 for review. The full-text review identified 77 articles meeting inclusion/exclusion criteria which are included here. Risk of bias among included studies was examined using the Evidence Project risk of bias tool. Findings fell into five categories of alcohol use proxies including knowledge/awareness, perceptions, attention, recall/recognition, attitudes/beliefs, and intentions/behavior. Real-world studies highlighted an increase in AWL awareness, alcohol-related risk perceptions (limited findings), and AWL recall/recognition post-AWL implementation; these findings have decreased over time. Conversely, findings from experimental studies were mixed. AWL content/formatting and participant sociodemographic factors also appear to influence the effectiveness of AWLs. Findings suggest conclusions differ based on the study methodology used, favoring real-world versus experimental studies. Future research should consider AWL content/formatting and participant sociodemographic factors as moderators. AWLs appear to be a promising approach for supporting more informed alcohol consumption and should be considered as one component in a comprehensive alcohol control strategy.

Keywords:

• Alcohol
• health warnings
• bottles

Alcohol is the most consumed drug in the United States, with a lifetime prevalence rate of 87% among individuals >18 years-old.¹ Adults consume two-to-six liters of alcohol per year, on average,² and three-million deaths were attributed to alcohol misuse in 2016.³ The prevalence of alcohol use disorder is also rising (12.7%/8.5% in 2016/2015, respectively).⁴ To address detrimental alcohol-related concerns (e.g., car crashes, psychiatric comorbidities),⁵ policies are implemented to reduce alcohol’s impact on society (e.g., taxation, advertisement restrictions).⁶ There is growing evidence on alcohol warning labels (AWLs) as a policy to address alcohol concerns; however, research must be synthesized.

Warning labels appear on product packaging as text or text and pictorial formats, presenting facts and providing exposure to the potential physical/psychosocial harms of alcohol. AWLs are used in 49 countries with this under-utilization potentially result from industry pushback (e.g.,⁷) Of countries with AWLs, their content/format vary widely.⁸ Lack of consistent and global AWL implementation has resulted in few real-world studies testing AWLs particularly versus warnings on other addictive products (e.g., cigarettes).

Given the known harms of cigarettes, international guidelines (enforced under the World Health Organization Framework Convention on Tobacco Control).⁹ CWLs effectively reduce cigarette use and increase motivation to quit, helpline calls, and abstinence rates.^10,11 Similar guidelines do not exist for alcohol given conflicting evidence on their efficacy. Current AWLs are limited with unclear/abstract content, poor visibility, voluntary labeling, and no pictorial warnings/plain packaging.^12,13 We must determine the public health benefits of AWLs.¹²

Several reviews of the AWL literature have been conducted with mixed findings about the efficacy of AWLs.^14–20 Issues with these reviews include out-of-date literature searches, nonsystematic methods (e.g., rapid reviews), as well as a focus on one-to-two proxies of alcohol use, beverage warning labels (not alcohol specifically), nutritional information, and research-based AWLs. As such, an up-to-date systematic review of the AWL research, which addresses limitations of prior reviews, is critical. This review aimed to examine research on AWLs from inception. A focus was placed on comparing studies based on their methodology (i.e., real-world and experimental studies), while highlighting moderators to provide potential explanations for mixed findings across the field. Specific proxies of alcohol use were not pre-determined to expand the reviews breadth. It was predicted that AWLs would be more effective in the real-world versus experimental settings.

Methods

This systematic review adheres to the preferred reporting for systematic reviews and meta-analyses (PRISMA) guidelines.²¹ Study protocols were pre-developed and no deviations occurred. AWL articles were searched in PsycINFO, Web of Science, PubMED, and MEDLINE databases (Appendix A). All articles published from inception (i.e., 1979) to July 26^th, 2020 were included.

K. J. and M. A.-H. independently screened articles. Titles and abstracts were screened and excluded if they: were not original research; did not include an AWL; focused on alcohol ingredients/constituents and/or nutritional labels; provided general awareness/messaging; or were not written in English. No limits were set on participants or study design.

Thousand six hundred and thirty-four articles were found (1,589 from database searches and 45 from reference lists of eligible articles; Figure 1). After duplicate removal, 961 articles remained. Ninety-six articles underwent full-text screening. For full-text screening the following inclusion/exclusion criteria were added: clear purpose/research question/objective; clear participant description; clear analyses/findings. Excellent inter-rater reliability (κ = 0.84) and 82.85% agreement was obtained; disagreements were discussed to consensus. Seventy-seven articles were included. The themes of findings were subdivided into different proxies of alcohol use (described below).

Figure 1. PRISMA flow diagram depicting the selection process for journal articles in the current systematic review.

Article data were recorded in duplicate and compared to ensure accuracy. The Evidence Project risk of bias tool, which examines studies of varying research designs incompatible with traditional risk of bias measures,²² was completed individually by K.J. and E.M (Table 1).

Table 1. Summary of studies on the efficacy of alcohol warning labels.

References	Sample size	AWL	Outcomes	Study design	Main results	Funded by alcohol industry
Al-Hamdani et al., 2017	440 (52% Female)	TW; PW Content: Liver cancer	Attention, recall, and recognition; perceptions	Cross-sectional design	TW + PW reduced positive alcohol-related perceptions; augmented with plain labels and large warnings More warning recognition with plain wine bottles vs. their branded counterparts	No
Al-Hamdani et al., 2015	92 (66% Female)	TW; PW Content: Liver cancer, liver cirrhosis	Attention, recall, recognition; perceptions	Cross-sectional design	TW-only and TW + PW reduced positive alcohol-related perceptions relative to no warnings More warning recognition for plain hard liquor bottles vs. their branded counterparts	No
Andrews et al., 1993	273 (53% Female)	TW Content: Birth defects, driving impairment; ARD, addiction, drug combination	Attitudes and beliefs	Cross-sectional design	Most favorable attitudes toward birth defect warnings	No
Andrews et al., 1991	273 (53% Female)	TW Content: Birth defects, driving impairment; ARD, addiction, drug combination	Attitudes and beliefs	Cross-sectional design	Birth defects and/or impaired driving warnings more believable and favorable for occasional vs. frequent drinkers	Not reported
Andrews et al., 1990	273 (53% Female)	TW Content: Birth defects, driving impairment; ARD, addiction, drug combination	Attitudes and beliefs	Cross-sectional design	Females believed birth defect AWL most Males believed impaired driving AWL most Those with favorable drinking attitudes disbelieved birth defect and disliked hypertension warnings vs. those with unfavorable attitudes toward drinking	No
Annunziata et al., 2016	300 (51% Female)	TW; PW Content: driving impairment	Attitudes and beliefs	Cross-sectional design	Slight majority wanted to learn more about health effects of excessive drinking Being young, female, educated, the primary grocery shopper, and having children increased the likelihood of warning interest	No
Armitage et al., 2016	Experiment 1 85 (62% Female) Experiment 2 58 (76% Female)	All Experiments TW Content: Standard government label, alcohol-related self-affirmation	All Experiments Intentions and behavior	Longitudinal design (2 waves)	Self-affirmation AWL reduced alcohol consumption vs. standard government AWL	No
Blume et al., 2007	99 (100% Female)	Label type NS Content: Birth defects	Awareness; attention, recall, and recognition	Cross-Sectional Design	More likely to recognize AWL when presented in a language in which the participant is proficient	No
Clarke et al., 2020	6,024 (50% Female)	TW; PW Content: Bowel/breast/liver cancer	Intentions and behavior	Cross-sectional between-subject design: 2 (image) × 2 (text)	Less likely to choose alcoholic drink when AWL present, strongest for PW-only AWL TW was most accepted followed by PW + TW, then PW	No
Coomber et al., 2015	561 (47% Female)	TW Content: Birth defects, ARH, underage drinking	Attention, recall, and recognition	Cross-sectional design	(Binge) Drinking from alcohol bottle and AWL-supportive attitudes increased AWL recall	No
Creyer et al., 2002	274 (52% Female)	TW Content: Birth defects, driving impairment, operating machinery, alcohol is a drug	Perceptions; attitudes and beliefs; intentions and behavior	Cross-sectional design	Perceived risk reduced to government vs. “alcohol is a drug” AWL Perceived risk lower among binge drinkers and those with lower levels of alcohol use For binge drinkers, the “alcohol is a drug” vs. government AWL is associated with more severe perceptions of drinking problems	Not reported
DeCarlo et al., 1997	150 (46% Female)	TW Content: driving impairment, birth defects, ARH	Awareness; intentions and behavior	Cross-sectional design	Slight majority think AWLs are informative Few change behaviors due to AWLs Alcohol use negatively related to reading AWLs	Not reported
Dossou et al., 2017	26 (50% Female)	PW Content: Birth defects	Awareness; intentions and behavior	Cross-sessional design	AWL lacked noticeability PW easier to understand AWL did not impact drinking behavior	No
Dumas et al., 2018	3,603 (100% Female)	PW Content: Birth defects	Awareness	Cross-sectional design	AWL awareness increased among younger and more educated women	No
Glock et al., 2013	40 (60% Female)	TW Content: Cardiovascular, liver damage, addiction, capacity, tension, ARH, stress, loneliness	Attitudes and beliefs; intentions and behavior	Two-factorial mixed design: 2 between-subjects factors (Warning label) × 2 within-subjects factors (Time)	Negative attitudes after seeing AWL Lower drinking intentions associated with AWL which contradict positive alcohol outcomes	Not reported
Graves et al., 1993	6,096 (57% Female)	TW Content: Birth defects, driving impairment, operating machinery	Awareness; attention, recall and recognition; perceptions	Longitudinal Design (2 waves) Samples in each wave were independent groups	Risk perceptions initially increased following AWL implementation, then declined AWL recall decreases 1-year post-implementation Being male, younger, a heavier drinker, and/or more educated increased likelihood of seeing the AWL	No
Greenfield et al., 2007	8,066 (58% Female)	Label type NS Content: NS	Attitudes and beliefs	Longitudinal Design (5 waves) Samples in each wave were independent groups	AWL policy implementation most supported policy in US after 10 years relative to other alcohol policies	Not reported
Greenfield et al., 1999 Title: Long-Term Effects of Alcohol Warning Labels: Findings from a Comparison of the United States and Ontario, Canada	10,182 (% Female NS)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness; attention, recall, and recognition; intentions and behaviors	Longitudinal quasi-experimental design (4 waves) Samples in each wave were independent groups	Awareness of AWL increased across time Being a heavy drinker and younger increased likelihood of seeing AWL in intervention vs. control site Recall differed between intervention and control sites Noticing AWLs increases likelihood of conversing with others about the AWL, specifically for women and younger individuals Positive relationship between seeing AWL and drinking and driving at intervention site, but also more likely to report deliberately not driving after alcohol use	No
Greenfield et al., 1998 Title: Five Years’ Exposure to Alcohol Warning Label Messages and their Impacts: Evidence from Diffusion Analysis	8,065 (% Female NS)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness; attention, recall, and recognition; attitudes and beliefs; intentions and behavior	Longitudinal Quasi-Experimental Design (5 waves) Samples were independent groups	AWL awareness and recall increased and then plateaued 4-5 years post-implementation Awareness and recall decreased with age Elevated AWL exposure, awareness, and recall in males and heavier drinkers Pregnancy and drinking impairment AWLs are recalled more by women and men, respectively Noticing vs. not noticing AWLs decreased drinking and driving	No
Greenfield et al., 1993 Title: Early Impacts of Alcoholic Beverage Warning Labels—National Study Findings Relevant to Drinking and Drinking Behavior	2,000 (59% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness; attention, recall, and recognition; intentions and behavior	Cross-sectional design	AWL awareness increased with age Males and heavier drinkers noticed AWL more Those who drank and drove in the last year had increased AWL recall Heavy drinkers who never drank and drove most likely to recall AWL - Self-regulation to not drink and drive was high among those recalling the AWL	No
Hall et al., 2020	1,352 (47% Female)	TW; PW Content: Liver disease	Perceptions	Cross-sectional design	PW perceived as more effective than TW - PW led to lower believability, greater resistance, more thinking about harms, and lower product appeal	No
Hall et al., 2019	1,360 (47% Female)	TW Content: Liver disease	Perceptions; intentions and behavior	Cross-sectional design	AWLs with “cause” more likely to discourage use, while “may contribute to” is least likely to discourse use AWLs the “cause” had the most support Heavy drinkers were less likely to support AWLs with “cause”	No
Hankin, Firestone et al., 1996	17,456 (100% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Intentions and behavior	Pre/Post-law assessment Samples were independent groups	- AWL implementation decreased drinking in nulliparae pregnant women	No
Hankin, Sloan et al., 1996	7,334 (100% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness	Pre/Post Law Assessment - Samples were independent groups	AWL awareness increased for 3 years post policy implementation, leveling off thereafter − 1.25-fold increase in heavy drinkers’ knowledge of AWLs	No
Hankin, Sloan, Firestone, Ager, Sokol, & Martier, 1993 Title: A Time Series Analysis of the Impact of the Alcohol Warning Label on Antenatal Drinking	12,026 (100% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Intentions and behavior	Pre/Post Law Assessment - Samples were independent groups	- Decreased alcohol consumption in light-drinkers post-AWL implementation	No
Hankin, Firestone, Sloan, Ager, Goodman, Sokol, & Martier, 1993 Title: The Impact of the Alcohol Warning Label on Drinking During Pregnancy	4,397 (100% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Intentions and behaviors	Pre/Post Law Assessment - Samples were independent groups	- Reduction in alcohol consumption during pregnancy in light-drinking females post-AWL implementation	No
Hankin, Sloan, Firestone, Ager, Sokol, Martier, & Townsend,, 1993 Title: The Alcohol Beverage Warning Label: When Did Knowledge Increase?	5,169 (100% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness	Pre/Post Law Assessment - Samples were independent groups	- Women under age 30 had more awareness of a pregnancy AWL	No
Hankin et al., 1998	21,127 (100% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Intentions and behavior	Pre/Post Law Assessment Samples were independent groups	- Reduction in alcohol consumption after AWL implementation	No
Hilton et al., 1991	2,006 (52% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Attitudes and beliefs	Cross-sectional design	Individuals 30-39 showed less support for AWL vs. other age groups - Majority believe alcohol should have AWLs	No
Hobin Shokar et al., 2020	1,647 (49% Female)	TW Content: Cancer, national drinking guidelines, standard drinking information	Awareness; attention, recall, and recognition; intentions and behavior	Quasi-experimental design (2 waves)	AWL noticeability increased Recall of cancer AWL high in intervention vs. control site; elevated among more educated individuals Reading, thinking about, and talking about AWLs higher in intervention vs. control site; elevated among females - Intentions and alcohol use decreased following AWL implementation; elevated among females	No
Hobin, Weerasinghe et al., 2020	2,049 (49% Female)	TW Content: Cancer, national drinking guidelines, standard drinking information	Awareness; attention, recall, and recognition; attitudes and beliefs	Quasi-experimental design (3 waves)	Majority noticed AWLs Cancer AWL recall increased across waves Awareness of alcohol as a carcinogen increased in intervention vs. control across waves - Most strongly supported cancer AWL	No
Jarvis et al., 2013	300 (% Female NS)	TW Content: Brain effects, driving impairment, violence	Intentions and behavior	Mixed Methods Design: Focus groups and cross-sectional survey	AWLs influenced change when presented with specific attributes - Heavier drinking females were negatively affected by AWL on brain effects (negative framing)	Not reported
Jones et al., 2010	44 (45% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD, breast/liver cancer	Attitudes and beliefs; intentions and behavior	Qualitative focus group design	Large and plain warning increase AWL as a knowledge preference - Likely to change behavior after seeing warnings about short-term harms	Not reported
Kaskutas, 1993 Title: Differential Perceptions of Alcohol Policy Effectiveness	2,017 (% Female NS)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Perceptions	Cross-sectional design	AWL effectiveness increased for heavy drinkers’ post-AWL implementation - Drinkers believe AWL are effective for themselves, but not for moderate or heavy drinkers	No
Kaskutas, 1993 Title: Changes in Public Attitudes toward Alcohol Control Policies since the Warning Label Mandate in 1988	4,023 (% Female NS)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Attitudes and beliefs	Longitudinal design (2 waves)	Increased support 1989-1991 Awareness and abstinence associated with elevated support Being male, younger, a heavier drinker, and having lower education were associated with lower support - Decreased respondents thinking warnings are ineffective for heavy drinkers	No
Kaskutas et al., 1992	2,000 (59.2% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness	Longitudinal design (2 waves) Samples were independent groups	- Men, aged 18-29, and heavy drinkers had increased awareness of AWLs as well as caution about impaired driving and drinking during pregnancy post-AWL implementation	No
Kaskutas et al., 1991	4,006 (% Female NS)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Attention, recall, and recognition; attitudes and beliefs	Longitudinal design (2 waves) Samples were independent groups	Minority noticed AWL Being male and/or a heavy drinker increased likelihood of noticing AWL - Majority believed hazards on AWL were true	No
Kersbergen et al., 2017	Experiment 1 60 (63% Female) Experiment 2 120 (65% Female)	All Experiments Label type NS Content: NS	Attention, recall, and recognition; Intentions and behavior	All Experiments Cross-sectional eye tracking design	Experiment 1 Majority of time spent viewing product branding Those motivated to reduce alcohol use paid less attention to AWL Experiment 2 - Alcohol intervention decreased branding views but did not increase AWL views	No
Kozup et al., 2001	150 (66% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Attitudes and beliefs; perceptions	Between-Subjects Design: 2 (health effect claim) × 2 (wine drinking status)  × 2 (government warning statement)	When health claims were absent, drinkers’ attitudes toward wine products and their health benefits were more favorable when AWL was present vs. not When health claims were absent, AWLs had no influence or an unfavorable influence on nondrinkers’ attitudes, health benefits of wine and perceptions of wine	Not reported
Krischler et al., 2015	122 (69% Female)	TW; PW Content: Vomiting, arrest, driving impairment	Perceptions; intentions and behavior	Mixed design: 3 (warning labels) between-subjects factor × 3 (expectancy category) within-subjects factor	Increased alcohol use related to depression vs. social- and tension-related alcohol expectancies Positive vs. negative expectancies associated with use Similar drinking intentions across AWL and control groups Majority would buy alcohol with AWL ∼37% believed AWLs are effective	No
Laughery et al., 1993	Experiment 1 75 (53% Female) Experiment 2 72 (51% Female) Experiment 3 24 (% Female NS)	TW; PW Content: Birth defects, driving impairment, operating machinery, ARD	Attention, recall and recognition	NS	Red PW without TW most noticeable	No
MacKinnon et al., 2001	649 (51% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Behavior	Longitudinal study (3 waves)	AWL implementation did not affect alcohol use	No
MacKinnon et al., 2000	32,517 (50% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness	Pre/Post-law assessment	AWL awareness increased for 3.7 years post-implementation, declining thereafter	No
MacKinnon et al., 1994	Experiment 1 128 (% Female NS) Experiment 2 813 (52% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD, liver damage, cancer, poison, toxic, hypertension, addiction drug combination	Attention, recall, and recognition	All Experiments Cross-sectional design	Specific risks were more important Qualifier words reduced avoidance Warnings with “poison,” “cancer,” or “health problems” were the most powerful	No
MacKinnon, 1993	Experiment 1 111 (% Female NS) Experiment 2 75 (% Female NS)	TW Content: Birth defects, driving impairment, operating machinery, ARD, cancer, poison, toxic	Behavior	All Experiments Cross-sectional design	Fictious warnings more effective in reducing self-reported drinking than US standard warnings	No
MacKinnon et al., 1993	3,217 (48% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Attention, recall, and recognition; awareness; intentions and behaviors	Pre/post-law assessment	Alcohol consumption positively correlated with AWL recall likelihood AWL awareness increased post-AWL implementation	No
MacKinnon & Faughtry 1993	Experiment 1 288 (38% Female) Experiment 2 243 (47% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD)	Attention, recall, and recognition; intentions and behavior	All Experiments Cross-sectional design	Alcohol consumption positively correlated with AWL recall likelihood	No
MacKinnon et al., 1992	288 (38% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Attention, recall, and recognition	Between-Subjects Design: 5 (label exposure) × 4 (level of tests)	Prompting of the risks on a n AWL improves recall	No
Malouff et al., 1993	Experiment 1 43 (51% Female) Experiment 2 50 (74% Female) Experiment 3 44 (84% Female) Experiment 4 75 (52% Female)	All Experiments TW Content: Birth defects, driving impairment, operating machinery, ARD	Attention, recall, and recognition; behavior	All Experiments Cross-sectional design Experiments 3 & 4 In-laboratory behavioral observation	Experiment 1 Recall of AWL components was high AWL should be larger for effectiveness Experiments 2 & 3 Rated horizontal AWL more prominent than vertical Experiment 4 Most recalled AWL component Those asked about AWL consumed less alcohol vs. control	Not reported
Marin et al., 1997	1,011 (63% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD, smoking, carbon monoxide, coffee, tea	Awareness	Pre/post-law assessment	Increased AWL awareness in accultured Hispanic women (vs. not)	No
Mayer et al., 1991	NS (% Female NS)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness; attention, recall, and recognition; perceptions; behavior	Pre/post-law assessment	AWL awareness increased pre- to post-AWL implementation Birth defects were the most recalled AWL	No
Mazis et al., 1996	5,032 (% Female NS)	TW Content: pregnancy, driving impairment and operation of machinery, health problems	Awareness; attention, recall, and recognition; behaviors	Longitudinal study (5 waves)	AWL awareness increased following implementation and remained relatively stable over time Being younger, less educated, black, and an alcohol drinker increased AWL awareness Being younger, less educated, white, and consuming 4+ drinks increased AWL recall Most believe effectiveness would increase by doubling size of AWL	No
Mazis et al., 1991	1,008 (56.5% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Perceptions	Pre/post-law assessment	Risk perceptions increased post-AWL implementation	No
Miller et al., 2016	1,545 (72% Female)	TW Content: Breast/mouth/throat/bowel cancer	Awareness; attitudes and beliefs; intentions and behavior	Longitudinal design	70% and 74% believed cancer-related AWL prompt conversations about cancer risk and cancer discussions with children, respectively 50% agreed AWL influence alcohol use	No
Monk et al., 2017	22 (68% Females)	TW; PW Content: ARD	Attention, recall, and recognition; attitudes and beliefs	Mixed factorial eye tracking design: 2 (stimuli) ×  2 (areas of interest) × 2 (positive expectancy change)	Dwell time was higher for PW vs. TW	Not reported
Nohre et al., 1999	6,391 (49% Females)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness	Pre/post-law assessment	Increased awareness post-AWL implementation	No
Parackal et al., 2010	1,129 (100% Female)	Label type NS Content: Birth defects	Awareness	Cross-sectional design	53% of pregnant women supported use of pregnancy-related AWL	No
Parker et al., 1994	2,402	TW Content: pregnancy, driving impairment and operation of machinery, health problems	Attention, recall, and recognition; awareness; perceptions	Pre/post-law assessment	Increased recall post-implementation Drinking and impaired drivers more likely to recall AWL vs. general public Increased awareness of alcohol-related harms on AWL Increased risk of drinking and driving among general public and drinking drivers pre- to post-implementation; not in impaired drivers AWL backlash from drinking and impaired drivers	No
	481 (37% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Awareness	Cross-sectional design	Increased awareness and caution about impaired drinking and drinking during pregnancy post-AWL implementation	No
Pechey et al., 2020	5,528 (49.1% Female)	PW Content: bowel cancer, breast cancer, liver cirrhosis, heart disease, liver cancer, not specified cancer	Attitude; intentions	Between-subjects design	Bowel cancer AWLs associated with highest negative arousal and lowest intentions to consume alcohol Low acceptability of AWL Some believe AWL would reduce alcohol consumption Majority viewed AWLs negatively	No
	2,168 (49.5% Female)	TW Content: Breast/mouth/throat/bowel cancer	Attitudes and beliefs	Mixed-methods design: qualitative and quantitative	Being young, a woman, or of higher educated increases AWL believability, personal relevance, and convincingness AWL more effective when causal language used	No
Pham et al., 2018	Experiment 1 559 (% Female NS) Experiment 2 87 (% Female NS)	Label type NS Content: NS	Attention, recall, and recognition	Experiment 1 Cross-sectional design Experiment 2 Cross-sectional eye tracking design	Large and red AWL drew most attention	Not reported
Scammon et al., 1991	2,417 (57.3% Female)	TW Content: Pregnancy, driving impairment, operating machinery, health problems	Awareness; perceptions; behavior	Pre/post-law assessment Samples were independent groups	AWL awareness increased post- vs. pre-implementation Alcohol use did not change Perceptions of alcohol-related risks did not change Alcohol-related birth defects declined pre- to post-intervention Liver cirrhosis (not on AWL) slightly increased pre- to post-intervention	No
Schoueri-Mychasiw et al., 2020	2,049 (% Female NS)	TW Context: Cancer, national drinking guidelines, standard drinking information	Awareness, attention, recall, and recognition, perceptions	Quasi-experimental design	High percentage noticed AWL Unprompted recall increased 10.8-fold from wave 1 to 3 (intervention vs. control) Prompted recall increased 7-fold from wave 1 and 3 (intervention vs. control) Awareness of AWL increased 2.9-fold from wave 1 to 3 (intervention vs. control) Awareness of sex-specific drinking limits increased 1.5-fold from wave 1 to 3 (intervention vs. control) Support for AWLs were similar across sites	No
Sillero-Rejon et al., 2018	128	PW Context: Liver cirrhosis, brain damage, mental illness, cancer, road accidents, risk to unborn child	Attention, recall, and recognition; perceptions; intentions and behavior	Double-blind pseudorandomized experiment: between-subjects factor (self-affirmation) and within-subject factor (warning severity)	∼50% of time spent viewing PW Self-affirmation manipulation had no effect on fixations, avoidance, reactance, susceptibility, effectiveness, or motivation to reduce alcohol vs. control Warning severity did not influence fixations High severity AWLs associated with higher avoidance, effectiveness, and motivation to drink less	No
Snyder et al., 1992	159 (56.6% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Perceptions; intentions	Cross-sectional design	AWLs had no effect on product risk rating Nondrinkers reported lower alcohol benefits when viewing AWL Drinkers rated alcohol products as more beneficial when viewing an AWL Male drinkers had greater drinking intentions after viewing an AWL	Not reported
Stacy et al., 1993	813 (48% Female)	TW Content: Pregnancy, driving impairment, operating machinery, health problems	Attitudes and beliefs; behavior	Mixed Methods Design: 2 (gender) × 2 (expectancies)	In females, negative expectancies on AWLs were negatively correlated with driving under the influence In males, negative expectancies on AWLs were negatively associated with driving under the influence and alcohol use	No
Stafford et al., 2017	45 (100% Female)	TW; PW Content: Liver cancer	Intentions and behavior	Between-Subjects Design; Behavioral Lab Setting	PW, but not TW, reduced alcohol consumption	No
Tam et al., 2010	1,376 (47% Female)	TW Content: Birth defects, driving impairment, operating machinery, ARD	Attention, recall, and recognition	Cross-sectional design	Being male or aged <40 increase likelihood to recall warning	No
Vallance et al., 2020	45 (63.9% Female)	TW Content: sex-based drinking limits; low-risk drinking guidelines	Attitudes and beliefs	Qualitative focus groups	Believed larger labels will draw more attention AWL may impact behavior and is a key factor in informing drinkers about the risks of alcohol use Strong support for AWLs, specifically cancer risk and pregnancy AWLs	No
Vallance et al., 2018	836 (47.9% Female)	TW Content: Cancer, national drinking guidelines, standard drinking information	Attitudes and beliefs	Quasi-experimental design	Strong support for AWLs Preferred (a) large labels with standard drink information and national low-risk guidelines; (b) charts; and (c) cancer and pregnancy content	No
Weerasinghe et al., 2020	1,730 (49.3% Female)	TW Content: Cancer, national drinking guidelines, standard drinking information	Awareness; Attitudes and beliefs	Quasi-experimental design	Awareness of link between alcohol + cancer increased from wave 1-2, returning to baseline at wave 3 Increased awareness of alcohol as a carcinogen following AWL use Increased knowledge positively associated with support for alcohol pricing and control policies Support for alcohol-related policies associated with higher education, being female, and aged 45+; not supporting associated with risky levels of use	No
Weiss, 1997	3,065 (53% Female)	Label type NS Content: Operating machinery, driving impairment, cancer, hypertension, liver damage	Awareness	Cross-sectional design	Low knowledge about AWL pre-AWL implementation in Middle Eastern population	No
Wigg et al., 2016	60 (72% Female)	TW; PW Content: Liver cancer	Perceptions; Attention, recall, and recognition; Intentions and behavior	Between-subjects design	PW increased consumers’ fear and reduced drinking intentions vs. TW Perceptions of alcohol-related health risks increased for PW vs. TW	No
Winstock et al., 2020	75,969 (35.7% Female)	TW Content: heart disease, liver disease, cancer, calories, number of drinks to help reduce and control drinking, no benefit of alcohol, and risk of violence and abuse	Attitudes and beliefs; behavior	Cross-sectional design	Violence AWL was the most believable Highest proportion considers drinking less from cancer AWL Cancer, calories, health myth, and violence warnings increase consideration of drinking less in females vs. males Heart, liver, cancer, health myth, and violence labels increase < 25 age groups consideration of drinking less New, believed and personally relevant warnings positively associated with behavior change Less likely AWL would make men consider drinking less AWLs likely to make heavier drinkers consider drinking less	No
Zahra et al., 2015	Experiment 1 153 (90.8% Female) Experiment 2 58 (86.2% Female)	PW Content: NS	Intentions	Experiment 1 Repeated measures design: 2 (content type) × 2 (emotional valence Experiment 2 Repeated Measures Design: 2 (content type) × 4 (problem type)	Experiment 1 Emotional valence did not influence logical decision making when warning is alcohol-related Positively valanced items lead to more correct logical decisions vs. negatively valanced items when warning is not alcohol-related Experiment 2 Negatively framed AWLs lead to more accurate reasoning in situations most likely to be pertinent to drinking decisions	No
Zhao et al., 2020	N/A	TW Content: cancer, national drinking guidelines, standard drinking information	Behavior	Interrupted time series quasi-experimental design	Total per capital alcohol sales decreased at the intervention site Per capita samples of labeled products decreased, while sales of unlabeled products increased Consumption levels greater in males, younger adults, Indigenous persons, those in a greater population density, and those with a lower income	No

^Note : ARD, Alcohol-related diseases; ARH, Alcohol-related Harms; AWL, alcohol warning label; GW, graphic warning; NS, not specified; PW, pictorial warning; PDB, problem drinking behaviors; TW, text warning; US = United States.

Results

Risk of bias assessment

Moderate-to-substantial inter-rater reliability (κ = 0.54–0.79) and 76.62–94.81% agreement was obtained. Study descriptive information included: cohorts (9.09%), control/comparison (32.47%), pre–post intervention (33.77%; all interventions included participants viewing AWLs and the types of interventions can be characterized as policy-making or educational), and random participant selection (32.47%; Table 2). Fourteen (58.33%) of studies with 2+ groups (N = 24) randomly assigned participants. One of seven longitudinal studies had retention rates ≥ 80% at follow-up. Seven (28.00%) and two (8.00%) studies with an intervention and control group (N = 25), reported equivalent sociodemographic and outcome measures at baseline, respectively.

Table 2. Results from the evidence project risk of bias assessment for all eligible articles.

Study	Cohort	Control or Comparison Group	Pre/Post Intervention Data	Random Assignment of Participants to the Intervention	Random Selection of Participants for Assignment	Follow-up rate ≥80%	Comparison Groups Equivalent on Sociodemographics	Comparison Groups Equivalent at Baseline
Al-Hamdani et al., 2015	No	Yes	No	Yes	No	NA	NR	NR
Al-Hamdani et al., 2017	No	Yes	No	Yes	No	NA	NR	NR
Andrews et al., 1990	No	No	No	NA	No	NA	NA	NA
Andrews et al., 1991	No	No	No	NA	No	NA	NA	NA
Andrews et al., 1993	No	No	No	NA	No	NA	NA	NA
Annunziata et al., 2016	No	No	No	NA	No	NA	NA	NA
Armitage et al., 2016	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes
Blume et al., 2007	No	No	Yes	NA	No	NA	NA	NA
Clarke et al., 2020	No	Yes	No	Yes	No	NA	Yes	NR
Coomber et al., 2015	No	No	No	NA	No	NA	NA	NA
Creyer et al., 2002	No	Yes	No	Yes	No	NA	Yes	NR
DeCarlo et al., 1997	No	No	No	NA	No	NA	NA	NA
Dossou et al., 2017	No	No	No	NA	No	NA	NA	NA
Dumas et al., 2018	No	No	No	NA	No	NA	NA	NA
Glock et al., 2013	No	No	Yes	NA	No	NA	NA	NA
Graves, 1993	No	Yes	No	No	Yes	NA	No	NR
Greenfield et al., 1999	No	Yes	No	No	Yes	NA	NR	NR
Greenfield et al., 1998	No	No	No	NA	Yes	NA	NA	NA
Greenfield et al., 1993	No	No	No	NA	Yes	NA	NA	NA
Greenfield et al., 2007	No	No	No	NA	Yes	NA	NA	NA
Hall et al., 2020	No	No	No	Yes	No	NA	NA	NA
Hall et al., 2019	No	No	No	NA	No	NA	NA	NA
Hankin et al., 1993	No	No	Yes	NA	No	NA	NA	NA
Hankin et al., 1996	No	No	Yes	NA	No	NA	NA	NA
Hankin et al., 1993	No	No	Yes	NA	No	NA	NA	NA
Hankin et al., 1996	No	No	Yes	NA	No	NA	NA	NA
Hankin et al., 1993	No	No	Yes	NA	No	NA	NA	NA
Hankin et al., 1998	No	No	Yes	NA	No	NA	NA	NA
Hilton et al., 1991	No	No	No	NA	Yes	NA	NA	NA
Hobin et al., 2020	Yes	Yes	Yes	No	No	No	No	No
Hobin et al., 2020	Yes	Yes	Yes	No	No	No	No	No
Jarvis et al., 2013	No	No	No	NA	No	NA	NA	NA
Jones et al., 2010	No	No	No	NA	No	NA	NA	NA
Kaskutas, 1993	No	No	No	NA	Yes	NA	NA	NA
Kaskutas, 1993	No	No	Yes	NA	Yes	NA	NA	NA
Kaskutas et al., 1991	No	No	Yes	NA	Yes	NA	NA	NA
Kaskutas et al., 1992	No	Yes	Yes	NA	Yes	NA	Yes	NR
Kersbergen et al., 2017	No	Yes	No	Yes	No	NA	Yes	No
Kozup et al., 2001	No	Yes	No	NA	No	NA	NR	NR
Krischler et al., 2015	No	Yes	No	Yes	No	NA	NR	NR
Laughery et al., 1993	No	No	No	NA	Yes	NA	NA	NA
MacKinnon et al., 1994	No	No	No	NA	No	NA	NA	NA
MacKinnon, 1993	No	No	No	NA	No	NA	NA	NA
MacKinnon et al., 1993	No	No	No	NA	No	NA	NA	NA
MacKinnon et al., 2001	Yes	No	No	NA	No	No	NA	NA
MacKinnon et al., 2000	No	No	Yes	NA	Yes	NA	NA	NA
MacKinnon et al., 1993	No	No	Yes	NA	Yes	NA	NA	NA
MacKinnon et al., 1992	No	Yes	No	Yes	No	NA	NR	NR
Malouff et al., 1993	No	No	No	NA	No	NA	NA	NA
Marin et al., 1997	Yes	No	No	NA	Yes	No	NA	NA
Mayer et al., 1991	No	Yes	Yes	No	Yes	NA	NR	NR
Mazis et al., 1991	No	No	Yes	NA	Yes	NA	NA	NA
Mazis et al., 1996	No	No	Yes	NA	Yes	NA	NA	NA
Miller et al., 2016	No	No	No	NA	No	NA	NA	NA
Monk et al., 2017	No	Yes	Yes	Yes	No	NA	NR	Yes
Nohre et al., 1999	No	No	Yes	NA	Yes	NA	NA	NA
Parackal et al., 2010	No	No	No	NA	Yes	NA	NA	NA
Parker et al., 1994	No	No	Yes	NA	Yes	NA	NA	NA
Parsons et al., 1993	No	No	No	NA	Yes	NA	NA	NA
Pechey et al., 2020	No	No	No	NA	No	NA	NA	NA
Pettigrew et al., 2014	No	No	No	NA	Yes	NA	NA	NA
Pham et al., 2018	No	Yes	No	Yes	No	NA	NR	NR
Scammon et al., 1991	No	Yes	Yes	No	Yes	NA	NR	NR
Schoueri-Mychasiw et al., 2020	Yes	Yes	Yes	No	No	No	No	NR
Sillero-Rejon et al., 2018	No	Yes	No	Yes	No	NA	NR	NR
Snyder et al., 1992	No	No	No	NA	No	NA	NA	NA
Stacy et al., 1993	No	No	No	NA	Yes	NA	NA	NA
Stafford et al., 2017	No	Yes	No	Yes	No	NA	Yes	NR
Tam et al., 2010	No	No	No	NA	Yes	NA	NA	NA
Vallance et al., 2018	No	No	No	NA	No	NA	NA	NA
Vallance et al., 2020	No	Yes	No	No	No	NA	No	NR
Weerasinghe et al., 2020	Yes	Yes	Yes	No	No	No	NR	NR
Weiss, 1997	No	No	No	NA	No	NA	NA	NA
Wigg et al., 2016	No	Yes	No	Yes	No	NA	Yes	NR
Winstock et al., 2020	No	No	No	NA	No	NA	NA	NA
Zahra et al., 2015	No	No	No	NA	No	NA	NA	NA
Zhao et al., 2020	No	Yes	Yes	No	No	NA	NR	NR

Note: NA, not applicable; NR, not reported.

Publications by year

Publication of AWL research started in 1990; yearly publications peaked in 1993, declining thereafter. Yearly publications have recently increased (Figure 2).

Figure 2. The number of eligible AWL articles by year of publication.

Research funding

Sixty-five studies (84.42%) reported funding source(s); none reported funding by the alcohol industry.

Knowledge and awareness

Knowledge and awareness refer to awareness of, or knowledge about, AWL or policies pertaining to them.

Real-world studies

Increased awareness of ALWs in the general population and in grade 12 students’ post-AWL implementation was found which lasted up to 3.7 years in several studies.^23–31

When examining specific AWL content, awareness of (a) the national drinking guidelines increased two-to-three-fold and (b) knowledge of sex-specific drinking limits by one-and-a half-fold and (b) alcohol effects increased months to the years’ post-AWL implementation.^32–34

AWL awareness appears dependent on sociodemographic characteristics. Those who are male, black, young, less educated, heavy drinkers, and alcohol-dependent evidence increased AWL awareness.^{26,30,33,35–37} For wine products, being female, younger, having children, having a higher education, and being the primary household grocery shopper increased desire to further one’s knowledge on excessive wine-drinking based on AWLs.³⁸ Pregnancy-related AWL awareness was highest among those (a) under legal drinking age,³⁷ (b) females between 19-39 years old (versus other age groups),³⁹ and (c) pregnant and postpartum females who were drinkers and educated (versus nondrinkers and “uneducated,” respectively).⁴⁰ In a sample of pregnant African Americans, being <30 years old increased knowledge of pregnancy-related AWLs seven months post-AWL implementation; this increased awareness lasted for three years before declining.^41,42 Lower awareness of pregnancy-related AWLs was evident in older (>35 versus <25) and lower educated individuals.⁴⁰

Experimental studies

Contrary to real-work studies, findings from experimental studies are mixed. The minority of participants was unaware of AWLs pre-implementation,⁴³ cancer AWLs,⁴⁴ and alcohol drinkers preferred AWLs as a knowledge source.⁴⁵

AWL content/formatting also influences awareness and knowledge. High AWL awareness is evidenced among AWLs of horizontal orientation⁴⁶ and as text-warnings (versus pictorial warnings) in binge drinkers.⁴⁷ Higher preference of AWLs as a knowledge source are shown for (a) text warnings (versus pictorial AWLs)⁴⁵ and (b) large AWLs on plain packaging.^48,49 Blank labels were preferable to those containing poison and cancer warnings.⁵⁰

In summary, real-world studies suggest an increase in AWL awareness post-AWL implementation, with demographic characteristics and AWL content/formatting influencing awareness. Experimental study findings are mixed, suggesting most participants are unaware of AWLs or do not prefer them as a knowledge source; AWL formatting appear to influence awareness and knowledge.

Perceptions

Perception refers to any measure of how alcohol consumers positively or negatively interpret AWLs.

Real-world studies

Increased risk perceptions of alcohol were identified following AWL implementation.⁵¹ National surveys found initial increases in perceptions of alcohol use risk after AWL implementation, which decreased thereafter.^52–54 AWLs were effectively impacted perceptions in heavy drinker’s pre-to-post AWL implementation.⁵⁴ A longitudinal study also identified those with a history of drinking and driving increased their risk assessment of driving after two drinks post-AWL implementation; the same was not true for those with a history of impaired driving.³³

AWL content/formatting also influences risk perceptions, with less direct AWLs containing “may” reducing risk perceptions.⁵⁵

Experimental studies

AWLs appear to increase consumers alcohol-related risk perceptions (versus no AWL).⁵³ Similarly, combined text and pictorial AWLs decreases positive alcohol perceptions and elevates negative emotional arousal and perceptions of disease risk (versus no AWL).^13,14 Findings are mixed, as Snyder et al.⁵⁶ found AWLs did not influence alcohol risk perception.

When assessing AWL content, pictorial warnings lowered consumers positive perceptions and product appeal and led to greater reactance, fear, and thinking about alcohol-related harms;^45,57 this was particularly true when pictorial warnings were large and on plain backgrounds or depict bowel disease.^45,58 Direct/concise statements (“alcohol is a drug”) effectively increased alcohol-related risk perceptions in binge drinkers (versus standard US alcohol warning) and when paired with pictorial AWLs.^53,59

Although limited, real-life studies provide evidence for increased risk perceptions of alcohol post-AWL implementation; however, findings were not maintained across time. Experimental studies provided mixed findings on AWL-related perceptions; specific AWL content may increase/decrease effects.

Attention

Attention refers to objective measures of attention (e.g., eye tracking and fixation).

Real-world studies

Hobin et al.’s⁶⁰ findings suggest large AWLs impact on attention whether one reads, thinks about, or talks about AWLs, especially for females.

Experimental studies

Approximately, 50% of participant time is spent examining AWLs; the most time was spent focusing on product branding.^61,62

When examining AWL content/formatting, red and larger fonts⁶³ and pictorial AWLs had higher self-reported attention and longer dwell.⁶⁴ Equal attention was paid to cancer warnings, national drinking guidelines, and standard drinking information,³² while highly specific warnings containing “health problem,” “cancer” and/or “poison” are linked to increased attentional avoidance.⁶⁵

Research is lacking on consumers attention toward AWLs. Consumers appear to pay attention to AWLs; however, their attentional degree may be influenced by AWL content.

Recall and recognition

Recall and recognition refers to any indication of remembering AWLs after exposure in (recall) or shortly following exposure (recognition).

Real-world studies

Recall of AWLs increased post-AWL implementation across several studies and lasted one-to-five years, declining thereafter.^{25,28,33,39,52} Similarly, recall increased two- and six-months post-AWL implementation is reported.³⁴ Unprompted AWL recall increased three- and ten-fold post-AWL implementation versus the control (national drinking guidelines) and prompted recall between three- and seven-fold post-implementation.³²

AWL recall/recognition increased among those who (a) were <40-years-old, (b) have higher education and health literacy, (c) were binge drinkers who drink out of the bottle, (d) were white, (e) believed AWLs were effective, (f) have a history of drinking and driving, (g) have exposure to AWLs (e.g., drinking 4+ drinks), and (h) were male.^{27,30,33,36,60,66–69} Recall/recognition was greater with larger AWL texts.⁶⁰

Sociodemographic factors also influenced recall/recognition. Recall for pregnancy AWLs was greater among females and younger individuals but recall for impaired driving AWLs was greater among males, heavier drinkers, individuals with a history of impaired driving, as well as those holding beliefs that impaired driving is hazardous or perceiving themselves to be at risk for impaired driving.^36,39,68

Experimental studies

Regarding AWL content/formatting, pictorial AWL recognition increased when (a) printed in red with icons and/or adding icons to a black AWL font⁷⁰ and (b) on plain packaging.^13,45 Inconsistencies in findings were evident.^47,53

Recall of AWLs increased when (a) pictorial, (b) larger, (c) vertically oriented, (d) presenting birth defect information, and (e) in the consumers language.^23,46,47,71 Finally, laboratory-based AWL exposure (versus no exposure) increased real-world recall.⁷²

In summary, real-world studies suggest an increase in recall/recognition of AWLs post-implementation; however, recall/recognition decreases over time. Sociodemographic and AWL content/formatting may impact recall/recognition. Recall results are mixed for experimental studies which may be attributed to moderators (e.g., pictorial versus text AWL).

Attitudes and beliefs

Attitudes and beliefs include statements about AWLs that are regarded by participants to be true, even if they are not factual.

Real-world studies

AWLs were the most publicly supported alcohol control policy in the United States 10+ years post-AWL implementation,⁷³ with exposure to AWLs increase support for policies.⁷⁴ Support for standard health warnings,^49,75 national drinking guidelines,³² low-risk drinking guidelines,⁷⁵ warnings about drinking during pregnancy,^76,77 and cancer warnings³⁴ are high.

AWLs also influence alcohol-related beliefs, with most participants believing AWL content²⁶ and cancer-related AWLs prompt discussions and raise awareness of alcohol-related harms.⁷⁸ Beliefs also differ by sociodemograpic factors with (a) light drinkers believing AWLs are effective for them, but not moderate-heavy drinkers⁷⁹ and (b) adolescents reporting no changes in alcohol-related harm beliefs post-AWL implementation.²⁸

AWL content/formatting may influence consumer attitudes, with AWL reference to cancer causal language generating negative alcohol-related attitudes (versus the same content with non-causal language).⁸⁰ Attitudes also differ with respect to sociodemographic factors, where those with a history of drinking and driving and/or impaired driving not accepting alcohol policies, including AWLs (longitudinal study).³³

Experimental studies

Among young people there is evidence of AWL support post-AWL implementation.⁶⁴ The greatest level of public support included when “cause” was present⁸¹ and cancer/birth deficit AWLs.^49,82

AWLs considered highly credible,⁸³ more severe,⁶¹ and larger³⁰ are believed to be most effective. When negatively framed and content pertains to violence/birth defects/impaired driving, AWLs are more believable, favorable, and relatable.^44,84,85 Neither positive nor negative AWL framing is more effective in changing alcohol-related attitudes.^86,87

Negative AWL-related attitudes are common, with three-quarters of participants not viewing AWLs as acceptable and many viewing AWLs as lacking efficacy.⁵⁸ AWL content influences findings, with (a) AWLs containing health claims and warnings impacting alcohol-related attitudes⁸⁸ and (b) text-only AWLs being the most accepted.⁸⁹ Sociodemographic factors may also impact attitudes, with favorable attitudes toward drinking leading to disbelief of hazardous and long-term AWLs.⁸⁴

Collectively, real-world studies suggest high support for AWLs and attitudes/beliefs vary by AWL content and sociodemographic factors; experimental studies also found high support AWLs, but findings on attitudes/beliefs were mixed.

Intentions and behavior

Intentions to drink and drinking behavior includes measures of actual drinking behavior or likelihood of drinking in response to AWL exposure.

Real-world studies

Reduced intentions to drink and consumption may be linked to larger AWLs.⁶⁰ Regarding drinking behavior, light drinkers appear most strongly impacted by AWLs.^39,78,90 There were fewer sales of products with AWLs (versus non-labeled products) post-AWL implementation⁶⁶ Some findings suggest AWLs influence drinking, short-term warnings are more effective in changing behaviors,⁴⁸ and AWLs with “cause” effectively discourage consumption.⁵⁷ AWLs content also impacts drinking behavior, especially in females and when pregnancy-related.^46,91,92 Conversely, some studies find no effect of AWLs on drinking behavior.^24,93

Experimental studies

AWL content/formatting influences drinking intentions. AWLs containing-specific health consequences/rebut positive social expectancies, emotionally framed messages, pictorial warnings, self-affirming intention messaging, and which depict bowel cancer reduce drinking intentions.^{44,53,58,89,94,95} Drinking intentions reduce when attention is spent toward AWLs (versus branding).⁶² Despite these findings, male drinkers who viewed AWLs had increased drinking intentions⁵⁶ and branding may have a stronger impact on alcoholic beverage choice versus AWL content.⁸⁶

Reductions in alcohol use following AWL exposure has been found across studies.^58,83,94 AWL content/formatting also influences results, with (a) AWL containing self-affirming reducing drinking intentions;⁹⁴ (b) pictorial AWLs (versus text) with liver cancer or negative alcohol expectancies reducing drinking and fewer impaired driving instances;^96,97 (c) AWLs with novel, believed, and personally relevant information reducing drinking behaviors;⁴⁴ and (d) specific health consequences/rebut positive social expectancies on AWLs reducing drinking.^14,44 Contrarily, drinking behaviors were not influenced post-AWL implementation in binge drinkers.⁴⁷

In summary, findings on intentions to drink and drinking behavior have been mixed. AWL content and sociodemographic factors may help explain conflicting results.

Discussion

This review highlights important findings of the impact of AWL on proxies of alcohol use which were divided into knowledge/awareness, perceptions, attention, recall/recognition, attitudes/beliefs, and intention/behavior. Conclusions differ based on the study methodology, with real-world studies being more favorable than experimental studies, on average. Findings also highlight the importance of considering AWL content/formatting and sociodemographic factors, as abundant findings suggest different findings based on the same. Results of this systematic review highlight the need for future studies on AWLs to consider moderators when analyzing data, and doing so within proper theoretical frameworks (e.g., basing the full text of manuscripts on race differences or genuine gender-based analysis).

Real-world studies suggested an increase in (a) AWL awareness, (b) risk perceptions of alcohol use (although findings are limited), and (c) AWL recall/recognition, post-AWL implementation which has decreased over time. High levels of support for AWLs were also identified. Findings on attitudes/beliefs and drinking intentions/behavior were mixed, however. One explanation for these findings is that AWLs have more universal effects on distal proxies of alcohol use (i.e., awareness, risk perceptions, recall/recognition), while their effects on attitudes/beliefs and drinking intentions and behaviors are more variable. For example, a recent systematic review by Hobin et al.¹⁷ proposed a conceptual model of the expected causal pathway which explains how AWLs work. Hobin et al.’s¹⁷ conceptual model suggests AWLs function by drawing in one’s attention and then increasing their knowledge of and attitudes/beliefs toward AWLs. Following this, it has been hypothesized that individuals will experience changes to their intentions to drink and then their drinking behaviors.¹⁷ Although Hobin et al.¹⁷ did not include all alcohol-related proxies identified in our review within their model, awareness, risk perceptions of alcohol use and recall/recognition would occur early in this model. As such, findings appear to be effective when assessing more distal proxies of alcohol use.

Findings from experimental studies were rather mixed, with some findings suggesting positive effects of AWLs, while other reported negative or no effects. One plausible explanation for the inconsistency in experimental studies stem from studies using a wide variety of warnings in terms of size, placement relative to brand imagery, and the types of warnings, all of which contribute to varied results. Contrarily, real-world studies mostly referred to the same AWLs, the ones in the United States, and therefore yielded similar results.

Across real-world and experimental studies, finding strength differed based on AWL content/formatting and/or consumer sociodemographic factors. Results suggesting AWL contact/formatting and/or consumer sociodemographic factors act as a moderator on AWL efficacy may explain mixed/conflicting findings across the field in prior literature reviews.^14–20 As such, it will be critical for future studies on the efficacy of AWLs to consider moderators, described throughout this review, into account.

When examining the publication timeline of articles on AWLs, there first articles were published in 1990 which coincides with the initial AWL enactment in the United States via the Alcoholic Beverage Labeling Act of 1988,⁹⁸ peaking during 1993. The increase in publications from 2015 onward is likely related to the concern of governments regarding increasing rates of alcohol use and related harms.⁴

The findings of this review should be considered with the following limitations and strengths in mind. Firstly, our risk of bias assessment²² highlighted the presence of potentially bias results across the most studies identified. Future research on AWLs must work to address sources of potential bias to better inform conclusions drawn from the literature. Secondly, only studies published in English were included which may limit generalizability. The number of AWL studies conducted in non-English languages is thought to be low; the future studies may wish provide insight into the effect of AWLs across non-English languages. Thirdly, we did not include unpublished studies, theses, or dissertations to control the quality of evidence being included (i.e., studies which have been peer-reviewed); however, this exclusion of unpublished studies may have biased our results in favor of positive AWL effects, also known as publication bias.⁹⁹ Fourthly, proxies of alcohol use identified in this review lack consistency and are differentially defined across the literature. As such, our definitions and categorization of proxies were broad. Future research should attempt to agree upon and standardize the definitions of alcohol use proxies to allow for study comparisons. Fifth, studies varied in design—where some were published pre-AWL implementation while others were post-AWL implementation.

Finally, hypotheses and methodology were not always informed by strong empirical, logical, or theoretical underpinnings. Researchers must use reviews of the literature to inform future studies on AWLs. With respect to strengths, included articles were systematically screened from inception and a risk of bias assessment was conducted. Issues with prior reviews (as outlined in the introduction) were also addressed. Proxies of alcohol use were also identified based on identified articles versus limiting results to pre-identified proxies. In sum, this review included a breadth of alcohol use proxies to provide a more comprehensive overview of extant literature than prior reviews.^14–20

This review highlights the evolution of AWL literature concluding that findings vary based on the methodology used (i.e., real-world versus experimental studies). Real-world studies highlighted an increase in AWL awareness, risk perceptions of alcohol use (limited findings), and AWL recall/recognition, post-AWL implementation which has decreased over time. Conversely, findings from experimental studies were mixed. Results underscore the importance of considering AWL content/formatting and sociodemographic factors in future AWL studies.

Additional information

Funding

Open Access funding provided by the Qatar National Library. At the time this review paper was written, Ms. Joyce’s graduate studies were supported by a Nova Scotia Graduate Scholarship, a Scotia Scholars Award from the Nova Scotia Health Research Foundation (NSHRF), a Joseph Armand Bombardier Canada Graduate Scholarship-Masters from the Social Sciences and Humanities Research Council of Canada (SSHRC), and a Vanier Canadian Graduate Scholarship. Dr. Stewart was supported through a Canadian Institutes of Health Research (CIHR) Tier 1 Canada Research Chair in Addictions and Mental Health at Dalhousie University. Dr. Al-Hamdani’s post-doctoral work were supported by a Mitacs Elevate Fellowship. At the time this review was completed and finalized for submission, Dr. Al-Hamdani was affiliated with Qatar University and the open access fees for publishing this article are funded by Qatar National Library.

Appendix A.

Search criteria

"Text warning" or "Textual warning" or "Graphic warning" or "Graphical warning" or "Health warning" or "Health message" or "Health communication" or "Warning message" or "Warning label" and Alcohol* or Drink* or Beer or Wine or Spirits