Perceived effectiveness of pictorial health warnings on changes in smoking behaviour in Asia: a literature review

Peer-reviewed articles were identified through multiple science databases: Sciencedirect, ProQuest, Oxford Journals, SpryngerLink, SAGE and Scopus which are indexed to Scopus, Medline or PubMed. We also included Google Scholar as the source of scientific papers to search for more articles that might not be published in the indexed journals. Electronic searches were conducted to identify relevant literature. The following keywords were used to identify relevant articles: (“pictorial warning”; “graphic warning”; “health warning”) with at least one of the following terms: smoking, tobacco, cigarette, product, package, and pack. All articles appearing in the search results were listed.

The review was limited to articles that reported original research findings and were published from January 2007 to Jathe nuary 2017. Studies which do not content aspects of warning, packaging and labelling regulation were excluded, as well as studies conducted in other Asian countries. Due to the diversity of research methods in this domain, the reviewer did not restrict studies to a particular design; however, each of the articles were reviewed for the following methodological criteria: (1) objectives and/or research questions were clearly explained, (2) sample and/or study population are described, (3) data collection method is consistent, (4) key measurements are adequate and valid, (5) results are clearly defined and measured (6) analysis of findings are clear and appropriate.

In conducting the data extraction, we firstly excluded articles that did not include Asian settings and/or did not sufficiently represent Asian countries as their study locations. A total of 141 articles were identified by titles which resulted in 14 articles excluded due to duplications. The 127 articles presenting empirical data were identified by titles and abstracts, 87 were excluded due to poor aspects of warning, packaging, and labelling. The 40 articles were identified by abstracts and methods which resulted in excluding 26 more articles due to insufficient methodological information. The 14 original articles included in the review consisted of 12 quantitative studies and 2 studies with both quantitative and qualitative components. The strategy of data extraction and identification used a PRISMA flow diagram [14].

No study met the criteria for inclusion prior to 2010. Meanwhile, the most recent study that was included in this study was published in 2016. During that period, the number of studies measuring the effectiveness of pictorial health warnings in Asian countries seems to fluctuate. There are two studies published in 2010, two studies published in 2011, four studies published in 2013, one study published in 2014, four studies published in 2015, and one study published in 2016. Most studies were published in credible international journals indexed by Scopus, MEDLINE, PubMed, ProQuest, and EBSCO. There was only one study published in the ResearchGate site. The online supplementary of the reviewed studies can be seen in Table 1.

Twelve studies used a quantitative method based upon the designs of cross-sectional (n = 5), cohort (n = 2), survey (n = 2), case-control (n = 1), or experiment (n = 2). The rest of the studies (n = 2) used a mix between quantitative and qualitative methods by conducting both a cross-sectional study and a focus group discussion. Regarding the number of the study population, the smallest sample was obtained by Wu et al [15]. who interviewed 202 people derived from a multistage-random sampling. Additionally, in Pakistan, the effectiveness of graphic health warning was measured among 388 high school students [16]. Meanwhile, the study in LAO PDR evaluated the implementation of pictorial health warnings both in average citizens and policy makers [17]. On the other hand, the study by Elton-Marhsall et al. [18] examined a big study population with 6513 adult smokers in China and 2883 adult smokers in Malaysia. Additionally, the study by Mutti et al. [19] had the highest response rate by 98.94% among 2061 respondents. Moreover, all studies performed univariate and bivariate analyses. However, only five studies were found to conduct a multivariate analysis. The operationalisation and statistical analysis of the reviewed studies can be seen in Table 2.

The studies obtained data from a variety of demographic characteristics of respondents, including age, sex, occupation, income, level of educations, ethnicity, and nationality. Three studies were conducted among adults (aged more than 18 years old), one study was carried out among teenagers (ranged from 15 to 18 years old), three studies were conducted among youths and young adults (ranged from 15 to 26 years old) and the rest five studies were carried out among both youths and adults (aged < 18 years old and ≥ 18 years old). Twelve studies obtained the data from both male and female respondents, while only one study conducted among male respondents only and the rest did not mention any specific gender characteristic.

Regarding study locations, three studies obtain the data in urban and rural areas; one study obtained the data in urban, semi-urban and rural areas, six studies gathered the data in urban areas only, while the rest of the studies did not specify it. Three studies included ethnic groups in their respondents’ characteristics which were classified as han-chinese and non-han-chinese groups in China and the majority and minority in Thailand and Malaysia. Another study in Qatar also classified their respondents based on their nationality which was divided as Qatari and non-qatari [20]. One study also obtained information about respondents’ medical conditions. Additionally, two studies in Qatar and China also collected information about respondents’ marital status [18, 20]. Additionally, most of the studies classified their respondents based on their education levels which mostly were divided into four groups: illiterate, low (up to middle school), high (up to high school), and graduate (diploma or higher). Five studies obtained information about respondent’s occupations. Moreover, eight studies classified respondents’ incomes which were mostly divided into three groups: low, moderate and high.

The data in the studies reviewed were obtained from respondents with different smoking behaviours. Eight studies examine both smokers and non-smokers, while six studies examine smokers only. Adult smokers were classified as follow: non-daily smokers, daily smokers, weekly smokers, smokeless tobacco users, and mixed between smokeless and smoked tobacco users. While youth smokers were classified as follow: daily smokers, occasional smokers, smokeless tobacco users, and mixed users. On the other hand, non-smokers were mostly classified into never-smoker, former smoker, susceptible and non-susceptible (for youth). The majority of the studies found that the percentages of male smokers were higher than that of smoking women. It is relatively a result of the higher number of male respondents than female respondents in the studies.

Compared with China and Thailand, smokers in Malaysia smoked fewer cigarettes with the average of 0–10 cigarettes per day [13, 18]. However, another study found that the majority of smokers in Malaysia smoked 11–20 cigarettes per day (48.6% in the control group and 55.1% in the intervention group) [4]. Moreover, most of the smokers in China consumed 11–20 cigarettes per day [13, 18], while in Indonesia the average of cigarettes smoked per day was 15 [21]. Studies in China showed a different variation of time to the first cigarette after waking up on a day. A study by Elton-Marshall [18] found that they smoked their first cigarette in less than 5 min after waking up while another study by Fong [22] found that they started in between 5 to 30 min. Similarly, in Malaysia, Elton-Marshall [18] found that the smokers smoked their first cigarette in 6 to 30 min after waking up, while Fathelarahman [4] found that they started after 60 min they woke up in the morning both in control and intervention groups. The study in Indonesia shows that most of the adult smokers started their first smoke after 60 min as they woke up in the morning [21]. Further demographic characteristics and smoking behaviours measured in the reviewed studies can be seen in Table 3.

Generally, respondent’s reactions to pictorial warnings in the reviewed articles can be classified into three categories: salience (reading, looking at or noticing the warning), emotional reaction, and cognitive reaction as shown in Table 4. A study in Jordan found that all proposed pictorial warnings in the study had greater proportions of perception of salience among both non-smokers and smokers compared with the currently-implemented pictorial warning which was less shocking and has a smaller size [23]. In Malaysia, following the implementation of new graphic health warning, the study by Fathelrahman, et al. [4] found that the association between the pictorial warnings and the increase in avoiding looking at or thinking about the label was significant in both control and intervention groups (ps = 0.003 and < 0.001 respectively). However, the increase in the intervention group after being exposed to the pictorial warning was much higher than that in the control group after being exposed to the text-only warning (40.6% and 21.4% increases).

In line with that, in Malaysia, Elton-Marshall et al. [18] also found significant in the change in noticing, reading and avoiding looking at/thinking of the label following the new pictorial warning compared with China text-only warning (ps = 0.02, 0.04, 0.02 respectively). Prior to the new pictorial warning, Malaysia had implemented text-only warning until January 2009 which was not changed over a period of time [4, 13]. Meanwhile, here was no significant change in noticing, reading and avoiding looking at the label over the period of the implementation of text-only warning in Malaysia [13]. On the other hand, the study found significant in avoiding looking at the label following the new graphic health warning in Thailand compared with their Malaysian counterparts (p value < 0.01). The same pattern also applied to the change in noticing and reading the label in the comparison between Thai pictorial warning and Malaysian text-only warning (ps < 0.001 and < 0.05) [13]. On the other hand, In India, noticing the pictorial warning on tobacco product was not associated with smoking behaviour [24].

Several studies also measured the emotional effects of pictorial warning vs. text-only warning or new pictorial warning vs. the old ones. The study conducted in young Jordanian adult found that the picture of a coffin elicited more fear in both smokers and non-smokers compared with the current pictorial warning which showed a diseased lung (p value < 0.001) [23]. Similarly, the study in Turkish youth showed that fear and disgust were evoked by the exposure to pictorial warning label among smokers and non-smokers [25]. The fear elicitation effect was also shown among Qatari Adult after being exposed to the pictorial warning [20].

In the reviewed studies, the cognitive effects such as the gain of knowledge or information and thinking of harm of smoking had been found significantly associated with the exposure to the pictorial warnings. The study in young Jordanian adult yielded a significantly greater proportion of gained information after being exposed to a child using inhaler compared with the current-lung-diseased-warning in both smokers and non-smokers (ps = 0.05 and < 0.001 respectively) [23]. Similarly, Malaysian-adult smokers thought of harm (p value = 0.004) and showed the change of knowledge (p value = < 0.001) after being exposed to the pictorial warning [16]. In line with that, a study in Qatari adult yielded gained information after the exposure to pictorial warning vs. exposure to text only warning (0.03). However, the thought of health risk was not found significantly associated with the exposure of pictorial warning nor significantly different between the change of text-only warning to the pictorial warning in Malaysia and the change of the old text-only warning to the new one in China (p value = 0.13) [13, 18, 23].

In the reviewed articles, the perceived effectiveness of pictorial warning as a deterrent of smoking intention was mostly assessed in both smokers and non-smokers. In non-smokers, the perceived effects of pictorial warnings were evaluated in deterring smoking initiation among youth. Meanwhile, in smokers, the perceived effects of pictorial warnings were examined in restraining them to start smoking. Aumaneekul et al. [26] compared pictorial warning and the plain packaging which resulted in greater intention not to smoke in non-smokers and ex-smokers after exposure to plain packaging compared with the current smokers (p value < 0.05). Similarly, in Indonesia, the level of confidence to avoid smoking in the future in youth non-smokers was found significantly different between before and after exposure to pictorial warning (p value < 0.05) [21]. Moreover, a study in India and Bangladesh also showed that the perceived effectiveness of graphic warning was rated higher than the symbolic and testimonial warnings (p value < 0.01) [19]. In contrary, the study in Jordan did not find any significant difference between the proposed pictorial warnings compared with the current warning [21]. In line with that, Tugrul et al. [25] did not find any differences between female and male respondents in perceiving the effectiveness of pictorial warning in motivating not to smoke among non-smokers and among those who ever consider smoking (ps = 0.561 and 0.424 respectively).

The study by Hawari et al. [23] yielded a significant difference between female and male smokers in perceiving the effectiveness of pictorial warning in motivating not to start smoking and to quit smoking (ps = 0.019 and 0.002, respectively). The perception was significantly indicated by fear rather than disgust [23]. In Indonesia, the study by Dien et al. [21] revealed that the differences of the level of confidence to stop smoking in the future between before and after the exposure to pictorial warnings were found significant with p value < 0.001 in both youth and adult smokers. Similarly, Mutti et al. [19] found that the perceived effectiveness of graphic warning on smoking cessation compared with symbolic and testimonial warning was significantly different (ps = < 0.001).

In Malaysia, fathelarahman et al. [4] examine the perceived effectiveness of pictorial warnings in three different variables: think to quit smoking, no interest in quitting smoking; and interested in quitting smoking within the next month. All of the three variables were found significantly different between before and after exposure to pictorial warnings [4]. Fong et al. compared the old Chinese text-only warning with the new one and with several different types of warnings from different countries. The study found that pictorial warnings were rated higher in motivating to quit smoking compared to text-only warning (p value < 0.001). Similarly, Yong et al. [13] found that there were significant differences between the likely to quit smoking in wave 1 and wave 3 (after the implementation of new warning) with p values < 0.001 in Thailand and < 0.05 in Malaysia. Moreover, the study yielded the significant differences between the avoidance of cigarette in wave 1 and wave 3 with the p value < 0.001 in Thailand and Malaysia.

Awaisu et al. [26] compared non-smokers and smokers in perceiving pictorial warnings to alter smoking cessation behaviours, and the result found significantly different (p value < 0.001). Additionally, when comparing abstract and real pictures in influencing the intention to quit smoking, Wu, et al. [15] found a significant difference between those types (p value = 0.025). Moreover, the authors found that there was a significant difference between a picture with less graphic and that with more graphic (p value = 0.001). Similarly, Elton-Marshall, et al. [18] found that there was a difference after the changes of health warnings on cigarette packs (p value < 0.001) in China compared to that in Malaysia. The perceived effectiveness of pictorial health warnings measured by the studies can be seen in Table 5.