Two online studies were conducted with separate sets of participants, one focusing on alcoholic drinks (Alcohol study) and one on energy-dense snack foods (Food study). These are described in turn.
Alcohol study
Preregistration
The study protocol was preregistered on the Open Science Framework, prior to data collection: https://osf.io/pr8zu/.
Design
A between-subject design was used, with participants randomised to view one of 21 image-and-text HWLs on an alcoholic drink.
Setting
The study was conducted online using the Qualtrics survey platform.
Participants
5376 participants, recruited through a market research agency (Dynata) were purposefully sampled from the general UK population to include a range of age, gender, and social grades. Eligible participants were those aged 18 or more, fluent in English, with access to a computer, who self-reported consuming either beer or wine at least once a week. Potential participants were registered with the research agency. All registered participants over 18 years old were invited to take the survey via email or could access the survey platform directly through the research agency website. Eligibility to participate was then determined via screening questions at the beginning of the survey. Only participants who consumed the target products were included, to reflect the likely principal target population for any comparable intervention applied in a real world setting.
Sample size determination
The current study was powered conservatively to detect a small difference between any pair of the conditions (Cohen’s d = 0.25) with power = 0.8, and alpha = 0.05, requiring 256 participants for each of 21 conditions, giving a total minimum sample size of 5376 participants.
Materials
Each image-and-text HWL comprised an image depicting a health outcome i.e. various cancers or heart disease, accompanied by text describing that outcome. A single graphic image was used for each HWL, defined as a photographic representation of the human body’s structure, anatomy or pathology (such as damaged organs or scenes of surgery). The full range of HWLs are included in Additional file 1.
Images used for the HWLs were chosen from a pool of 47 possible images sourced from previous studies (31) and online image databases such as Science Photo Library and Shutterstock. These comprised images of seven health consequences linked to alcohol consumption including bowel cancer, breast cancer, liver cancer, cancer (non-specified), heart disease, liver cirrhosis and liver disease. A convenience sample of 15 colleagues, working in the fields of public health, psychology and statistics, ranked the 47 different images based on the image’s perceived impact in deterring individuals from consuming excess alcohol. The three images ranked most highly for each of the seven health consequences were selected. Accompanying text was developed based on text used in previous studies and evidence about designing effective messages (2, 20, 32-35). See Box 1 for the different health consequences used for the Alcohol and Food studies.
|
Health consequences1
|
Alcohol Study
“Alcohol causes…”
|
Food Study
“Excess calories cause obesity, which causes)…”
|
Cancer
|
|
X
|
Bowel cancer
|
X
|
X
|
Breast cancer
|
X
|
|
Liver cancer
|
X
|
|
7 types of cancer
|
X
|
|
13 types of cancer
|
|
X
|
Heart disease
|
X
|
X
|
Liver disease
|
X
|
|
Liver cirrhosis
|
X
|
|
Type 2 diabetes
|
|
X
|
Obesity2
|
|
X
|
Box 1. Health consequences depicted in HWLs in the Alcohol and Food studies
|
1 Three HWLs with different images were used for each health consequence.
2 The text read ‘Excess calories cause obesity’ without the additional clause.
Measures
Primary outcome measure
Negative emotional arousal. This was assessed using a four-item measure (Cronbach’s α = 0.89) previously used to assess the impact of HWLs on cigarette packages (36) and on sugar-sweetened beverages (20). Responses to each of the four items are rated on a seven-point scale: How [afraid/worried/ uncomfortable/disgusted] does the label on this drink make you feel?’ (0 Not at all [afraid/worried/uncomfortable/disgusted] to 7 very [afraid/worried/uncomfortable/ disgusted])
Secondary outcome measures
Desire to consume the labelled product. This was assessed using a single-item seven-point measure: ‘How much do you want to drink this (wine or beer) right now?’ (0 Not at all to 7 very much).
Acceptability of health warning labels. This was assessed using a single-item seven-point measure adapted from previous studies (37) ‘Do you support or oppose putting this label on alcoholic drinks?’ (Strongly oppose – neither oppose nor support – strongly support). Ratings past the midpoint, i.e. above 4 on the scale, were taken to indicate that the label was acceptable.
Free-text responses. A free-text box was provided at the end of the study into which participants were invited to write comments (‘Do you have any further thoughts or comments that you would like to add?’). See Data Analysis section for details.
Additional measures
Age, gender, ethnicity, education level, household income, height and weight were all self-reported. Participants also reported their typical alcohol consumption. These measures were collected for the purpose of describing the sample only.
Procedure
Ethical approval was granted by the Cambridge Psychology Research Ethics Committee (PRE.2018.071). Participants were informed at the start of the study that they could withdraw at any point. After providing written consent to take part via an online consent form at the beginning of the study, participants completed screening questions relating to their typical consumption of alcohol and their age. Participants who reported drinking less than once a week on average, or were under 18 years old, were screened out (see ‘Participants’). Those screened out were diverted to a webpage that explained that they were ineligible for taking part in this study, and then redirected to the research agency home page. Eligible participants were then asked questions regarding their demographic characteristics (gender, ethnicity, education level, household income, height and weight). In order to ensure that images were clearly visible, the study was conducted on computer-size screens. Participants accessing the study on mobile phones were identified by the Qualtrics software and automatically screened out as soon as they clicked on the study link. Inattentive participants were screened out via an embedded attention check (asking participants, ‘when was the last time you travelled to Mars?’ [months ago/weeks ago/a few days ago/ never] (20, 37), with any participant who responded anything other than ‘never’ being screened out).
Participants viewed an image of beer or wine (depending on their stated preference) displaying, at random, one of 21 different image-and-text HWLs illustrating the adverse health consequences of alcohol consumption (see Figure 1 for example). Randomisation to the different HWLs was completed by an algorithm within the Qualtrics online platform. After viewing the HWL, participants completed the measures of negative emotional arousal, desire to consume the labelled product and acceptability. Finally, participants had the opportunity to leave a comment in a free-text box. Upon completion, participants were debriefed, which included providing information about the health consequences of consuming excess alcohol, and were reimbursed for their participation. Data were collected from October to December 2018.
[Figure 1]
Analysis
Descriptive analysis
Following normality checks, the mean and standard deviation (SD) and 95% confidence intervals (CIs) were calculated for each outcome and each HWL. An analysis plan was pre-registered before the data were inspected (https://osf.io/pr8zu/).
Content analysis
Comments provided by participants were manually coded and two emergent themes were identified. Responses were coded into themes, and as being either positive, neutral or negative. For full details of the analytic procedure see Additional file 2.
Food study
Preregistration
The study protocol was preregistered on the Open Science Framework, prior to data collection: https://osf.io/k7tw5/.
Design
Participants were randomised to view one of 18 image-and-text HWLs on an energy-dense snack.
Setting
The study was conducted online using the Qualtrics survey platform.
Participants
Recruitment process, sample and eligibility criteria were identical to that of the Alcohol Study (see above). The sole difference was the target product consumed - participants were self-reported regular consumers of biscuits, cake, crisps or chocolate (i.e. consumed at least once a week), and liked chocolate.
Sample size
The same sample size information was used as for the Alcohol study (with 256 participants needed per condition). With 18 conditions, a sample size of at least 4608 participants was required.
Materials
HWL images were selected from a pool of 33 images depicting six different health consequences, including bowel cancer, cancer (non-specified), heart disease, obesity and type 2 diabetes. The form comprised a graphic image plus text statement. For each of the six health consequences, three image-and-text HWLs were developed. The process of developing the HWLs and piloting them was the same as that described for the Alcohol study.
Measures
The measures used were identical to those used in the Alcohol Study, with the only differences being on three measures that named the product on which the label was placed, which was changed from alcohol to snack:
Primary outcome measure
Negative emotional arousal: ‘How [afraid/worried/uncomfortable/disgusted] does the label on this snack make you feel?’ ’ (Cronbach’s α = 0.91)
Secondary outcome measures
Desire to consume the product: ‘How much do you want to eat this snack right now?’
Acceptability: ‘Do you support or oppose putting this label on high calorie snacks?’
In an identical method as in the Alcohol study, free-text responses were also collected at the end of the study.
Additional measures
Age, gender, ethnicity, education level, household income, height and weight were all self-reported. Participants also reported their typical consumption of energy-dense snacks. These measures were collected for the purpose of describing the sample only.
Procedure
Ethical approval was granted by the Cambridge Psychology Research Ethics Committee (PRE.2018.072). As in the Alcohol study, participants were informed at the start of the study that they could withdraw at any point and gave written consent. Participants used computer-size screens for enhanced visibility of the HWLs. Participants completing the study on mobile phones were screened out. After consenting, participants completed screening questions relating to age, their typical consumption of energy-dense snacks, and whether they liked chocolate. If they did not eat an energy-dense snack at least once a week, did not like chocolate, or were under 18 years old, they were screened out. Participants then answered demographic questions, with an embedded attention check (identical to Alcohol study procedure - see above).
Participants viewed an image of a chocolate bar illustrating 18 different image-and-text HWLs depicting the adverse health consequences of obesity and related conditions, caused by excess calorie consumption (see Figure 2 for example) and rated them on negative emotional arousal, desire to consume the snack and acceptability of the HWL. Randomisation to the different HWLs was completed by an algorithm within the Qualtrics online platform. Finally, participants had the opportunity to leave a comment in a free text box. Upon completion, participants were debriefed, which included providing information about the health consequences of consuming excess calories, and reimbursed. Data were collected from October to December 2018.
[Figure 2]
Descriptive and content analysis
Both descriptive and content analysis procedures were identical to the Alcohol study (see above). The analysis plan was preregistered: https://osf.io/k7tw5/.