Table 2 Quality appraisal outcomes and study outcomes for each of the eligible studies (n = 19)

2009 Hammond [35]

High (cross-sectional)

PP; chi-square, linear regression

• Both brands with plain white packs were perceived as less attractive, non-preferred, and having a lower tar content compared to the branded packs.

• One pack brand was also considered as having a lower health risk, and one brand as having a less-smooth taste.

• The plain brown packs were less attractive and less smooth for one brand, and less attractive, less smooth, higher risk, and non-preferred for the other brand compared to branded packs. All p values for these stated differences are < .001.

2009 Vardavas [26]

High (cross-sectional)

GHW vs. text warnings; chi-square, multivariate logistic regression

• GHWs were considered more effective than text-only warnings for 71.6 to 96.1% of participants, both in preventing non-smoking participants from smoking and in describing the effects of smoking on health.

• Up to 84% of participants rated GHW as ‘effective’ or ‘very effective’ (4 or 5 out of 5) in preventing smoking initiation.

• The GHW depicting lung cancer was rated as the most effective, followed by the GHW depicting foetal damage caused when smoking whilst pregnant.

• Female participants had significantly higher effectiveness ratings of the GHWs depicting foetal damage, and protecting children from smoke (p < .05).

2010* Fong [27]

High (cross-sectional)

GHW vs. text warnings; chi-square, mixed-model ANOVA

• The four GHW packets were both rated and ranked as the most effective in motivating smokers to quit and preventing youth smoking, significantly higher than the six text warnings (p < .001), with the GHW depicting lung cancer rating the most effective, followed by the mouth disease, gangrene, and clogged arteries warnings (p < .05 between each warning).

• The four GHW (with lung cancer as the highest rated) were also the most effective in informing the public on the dangers of smoking, with 81.5% of adolescents stating that packaging within China should contain more health information and 78.9% stating that packaging should include pictures instead of text-only warnings.

2010 Germain [42]

High (RCT)

GHW/PP; chi-square, ANOVA, principal component analysis

• Mean ratings of all positive pack, smoker, and cigarette attributes significantly reduced as branding and colour were progressively removed from packaging (p < .001), with ‘lower class’ perceptions concurrently becoming stronger (p = .043).

• Smoking status was found to predict responses to pack ratings (p < .05), with established smokers having the most favourable perceptions of all packs. The addition of a larger GHW also had results dependent on smoker status, with experimenters and active smokers having the largest drop in perceptions of positive pack characteristics compared to susceptible and non-susceptible non-smokers (p < .01).

2011 Hammond [36]

High (cross-sectional)

PP; linear regression

• Compared to standard packs, of the eight brands used, plain packages were consistently the least appealing, were perceived as the worst tasting for six of the brands, had lower levels of tar for two of the brands, and were considered less harmful for two of the brands (all p < .05).

• Plain packs also received significantly fewer positive ratings for every smoker trait (glamour, femininity, slimness, coolness, popularity, attractiveness, and sophistication) compared to standard packs (p < .001).

• Significantly fewer participants preferred plain packs (p < .001).

2012a Hammond [28]

High (cross-sectional)

GHW; linear mixed effects models

• Text-only warnings were the lowest rated for all 15 health effects (p < .001), with the graphic warnings being rated as more effective than both the symbolic and lived experience warnings (p < .001), and those depicting external health effects perceived as more effective than those depicting internal health effects (p < .001).

• Lived experience warnings that depicted effects on others were rated as more effective than those that depicted effects on oneself (p < .001), and susceptible non-smokers had significantly higher ratings than non-susceptible non-smokers (p = .02).

2012b Hammond [37]

High (cross-sectional)

PP; linear regression

• Plain packs received the lowest appeal (p = .013), and taste ratings (p = .027), were less likely selected as a preferred pack (p = .026), and were considered to have higher tar compared to the fully branded packs (p = .024).

• Fully branded packs were also considered to have the lowest health risks compared to all other categories (p = .006).

• For perceived smoker traits, plain packs received the lowest ratings for all seven attributes: femininity, slimness, glamorous, coolness, popularity, attractiveness, and sophistication (all p < .05).

2012 Moodie [38]

High (cross-sectional)

PP; chi-square

• Half of the participants associated colour and strength of taste, and colour and perceived harm, with the red pack considered the strongest tasting and most harmful and the light blue pack and white packs as weaker tasting and being the least harmful.

• The brown plain pack was seen as largely unattractive, cheap, and uncool and used by boring, unfashionable, and older people. Smokers displayed less negativity towards the pack compared to non-smokers.

• Smokers were more likely (p < .001) to prefer a pack, with the slide pack being the most popular of the brown plain packs.

2013 Ford [39]

High (cross-sectional)

PP; principal components analysis

• The mean ratings for all 11 items for all packs (e.g. attractiveness, coolness, harmfulness) were generally negative (none > 3 out of 5), with the plain pack being the most negatively rated, with mean scores ranging from 1.24 to 1.99 (p < .01).

• The standard pack was also more negatively rated than the three novelty packs.

• Unlike the branded packs, the plain pack showed no association between the 11 rated aspects, and smoking susceptibility.

2013a* Hammond [29]

High (cross-sectional)

GHW; linear mixed effects models

• Full-colour warnings were rated more effective than black and white warnings (p = .004), as were real people over comic book-style (p < .001), and those featuring quitline information (p < .001), particularly for current over non-smokers (p = .046).

• Those with personal information were higher rated over those that did not (p < .004), as were those with graphic content compared to those that did not (p < .001), particularly for females over males. Mean scores were higher for ‘minority race respondents’ compared to ‘white respondents’ (p = .002).

2013b Hammond [43]

High (cross-sectional)

GHW/PP; chi-square, generalised estimating equation model

• Compared to branded packs, plain packs were considered less attractive, less likely to encourage smoking uptake, and had higher impact health warnings. Brown packs and those with graphic health warnings were also less likely perceived to have a smooth taste, present a lower health risk, or contain a lower amount of tar (all p < .001).

• Larger GHWs were rated as the least attractive compared to moderate-size GHWs (p = .001) and text warnings (p < .001), were the least smooth tasting (p < .001 and p < .001 respectively), the least likely perceived to have a lower health risk (p < .001 compared to text warnings), the least likely perceived to have lower levels of tar (p < .001 and p < .001 respectively), and were perceived as having the highest impact on health (p < .001 and p < .001 respectively).

2013 Pepper [30]

Moderate (cross-sectional)

GHW; linear regression, ANOVA

• The lung cancer warnings (both text-only and text plus image) received higher ratings than the addiction warnings, with 60% of assigned participants rating them 5 out of 5 for discouraging smoking, compared to 34% for addiction warnings (p < .001).

• There were no significant differences in deterring smoking or perceived risk for text vs. text plus image for either category.

• Over half of assigned participants believed they would develop lung cancer if they smoked regularly, and over two thirds held this belief for developing nicotine addiction, with both categories also generally being considered as very severe.

2015* Alaouie [31]

High (cross-sectional)

GHW; McNemar test

• Participants perceived all GHWs as significantly more effective for all items compared to the text-only warning (p < .001).

• Overall, compared to the text warnings, the lung cancer GHW received significantly higher effectiveness rating, followed by tooth decay, and death (all p < .01) except for female smokers due to low participant numbers.

• All warnings were significantly more effective than text warnings (all p < .001) in preventing non-smokers from smoking.

2015 Babineau [40]

High (cross-sectional)

PP; chi-square, generalised estimating equation

• Two of the branded packs were perceived to be more attractive and healthier and used by ‘popular’ individuals, and were chosen twice as frequently compared to plain packs (all p < .001).

• One pack brand (with pink and purple colouring) had a lower margin for choice (p < .001) and did not experience differences in attractiveness (p = .08), between the two packs, though the branded pack was perceived as healthier (p < .001).

• Female participants were significantly more likely than males to associate this brand with popularity (p = .03).

2016 Adebiyi [32]

Moderate (cross-sectional)

GHW; bivariate analysis

• Responses to the four GHWs included fear in 37.3–56.4%, shock in 23.3–37.3%, anxiety in 2.9–21.1%, and indifference in 3.3–20.0% of participants. The GHW suggesting that smoking causes impotence had the highest indifference rating.

• The GHW depicting airway cancer had the highest fear and shock ratings, and the lowest ratings for anxiety and indifference, and perceived as the most effective in preventing adolescents from smoking, especially those < 15 years (p < .05).

• The GHW stating cigarette smoke harming children received the highest frequency of anxiety.

2016 Andrews [44]

High (RCT)

GHW/PP; multivariate analysis

• The two most graphic health warnings significantly increased thoughts of quitting, evoked fear, and reduced feelings towards the pack and cigarette cravings compared to the control and low-graphic health warning (all p < .05).

• Plain packaging led to significant reductions in cigarette craving and feelings towards the pack (p < .05) and increased evoked fear (p < .05), but had no effect in increasing thoughts of quitting.

• There were no combined effects overall for PP and GHWs, though there were some combined effects in France and Spain in reducing cravings and pack feelings respectively, though there were smaller cell sizes and reduced statistical power.

2016 Mutti [41]

High (RCT)

PP; chi-square, linear regression models

• Plain (with descriptor) packages received significantly lower ratings for appeal and taste (both p < .001) compared to branded packs, though there was no significant difference in perceptions of harm.

• Female participants were more likely to give higher appeal and taste scores and rate packs as less harmful compared to males (p < .001, < .001, = .02 respectively).

• Smokers were more likely to give higher taste ratings and consider packs as less harmful compared to non-smokers (p < .05).

• Non-smokers rated branded packs significantly higher for all positive smoker-image traits (all p < .05), whilst smokers only rated two traits higher from branded compared to plain packs (stylish and sophistication, both p < .05).

• Older adolescent participants also rated positive smoker-image traits higher than younger participants.

2016 Netemeyer [33]

High (cross-sectional)

GHW; linear regression models

• Perceived graphicness was associated with an increase in evoked fear and guilt (p < .01) for smokers and non-smokers.

• Smokers had lower levels of disgust with increased graphicness compared to non-smokers.

• Increased graphicness also led to increased hesitance (reduced personal consideration) towards smoking.

• Stronger emotions in response to higher levels of perceived graphicness were more significant in smokers compared to non-smokers.

2017 Reid [34]

High (cross-sectional)

GHW; chi-square, ANOVA, logistic regression

• Perceptions of the health effects of smoking significantly increased for those who viewed the mouth cancer, heart disease, emphysema, and stroke (China and Korea), throat cancer (Bangladesh and Korea), skin ageing (India), impotence (India, China, and Korea), and gangrene (Bangladesh, India, and Korea) warnings (all p < .05).

• Three quarters of participants in China, Bangladesh, and Korea and half in India also believed that cigarette packages should include more health-related information than the current packaging warnings were displaying in their respective country.