Study design and setting
We conducted an individually randomised, two-arm parallel trial to test the effectiveness of the self-regulation intervention OPTIMISE (Online Programme to Tackle Individual Meat Intake through SElf-regulation) to reduce meat intake compared to a control condition. The study was delivered remotely through a bespoke website developed specifically for the intervention, through which all data collection took place between 15 March and 26 May 2021. The trial was granted ethical approval by the Central University Research Ethics Committee (CUREC) of the University of Oxford (REF: R71398/RE002).
We aimed to recruit 150 participants. The sample size was calculated with the aim to detect a medium-sized effect of d = 0.5 between conditions, at 80% power and 5% type 1 error rate, while allowing for a 15% drop-out rate . Participants were recruited through Prolific Academic  and completed a screening questionnaire on JISC online surveys . To be eligible, participants had to be aged 18 years or over, resident in the UK, eat meat at least five times per week, indicate they want to reduce their meat intake and be able to engage with the intervention content. Eligible participants who provided consent were invited to register with the OPTIMISE website through Prolific Academic, on a first-come first-served basis. The text from our Prolific Academic study advertisements is provided in Online Resource 1. Depending on adherence to the study procedures, participants were reimbursed with a payment of up to £32 (£0.50 for the screening questionnaire, and £1.50 or £0.75 per session for control group (21 sessions) and intervention group (42 sessions) participants, respectively).
After indicating their consent, participants were randomly assigned to intervention or control groups using a computer-generated list, with 1:1 randomisation. Participants were blinded to group allocation and informed they would be randomised to one of two groups, each following a different approach for reducing meat intake.
The study duration was 9 weeks (a baseline week of self-monitoring meat consumption, a 4-week active intervention phase, and a 4-week maintenance phase; Fig. 1). After registering with the website, all participants were presented with information regarding the health and environmental benefits of eating less meat (Online Resource 2). Participants then completed a baseline questionnaire that asked about their demographic characteristics, meat-free self-efficacy and meat-eating identity. To estimate meat-free self-efficacy, we used a scale adapted from Lacroix & Gifford’s self-efficacy scale , where participants were asked to rate the following statements on a scale from 1 (strongly disagree) to 7 (strongly agree): (1) “I lack the cooking skills to prepare meat-free meals”; (2) “I don’t know what to eat instead of meat” and (3) “I don’t have enough willpower to not eat meat”. To capture participants’ meat-eating identity, participants could self-identify as one of six identities (meat eater, omnivore, flexitarian, pescetarian, vegetarian or vegan). Despite our eligibility criteria for participants to eat meat regularly, participants were offered all meat-eating identity options at baseline and both follow-ups, as we know from our public engagement events that meat-eating identities do not always reflect actual consumption.
Meat consumption was measured daily during the baseline week (week 1), first follow-up week (week 5; FU1), and second follow-up week (week 9; FU2) using a meat frequency questionnaire, which has been found to be reliable and acceptable in a UK sample . This questionnaire combines data on food portion sizes from the UK Food Standards Agency with estimates of meat content in composite dishes from the UK National Diet and Nutrition Survey (NDNS). It asks participants to report how many servings of different meat and seafood products they consumed in the previous 24 h.
At each follow-up, participants repeated the meat-free self-efficacy and meat-eating identity questionnaires. At FU2, participants were asked what they thought the other study group had been doing to assess blinding. Participants received automated messages to their Prolific Academic accounts by the OPTIMISE website, prompting them to complete the sessions.
On the last day of the baseline week, participants received health and environmental feedback on their total meat and red meat consumption. They pre-selected strategies from a list of 26 meat consumption reduction actions (Online Resource 3) and set themselves a meat reduction goal. These strategies were created specifically for this study during a brainstorming session with a group of experts, including nutritionists, psychologists and health behaviour scientists in our department. They were then further refined during Patient and Public Involvement focus group sessions with meat eaters to ensure the actions were appropriate and understandable. The strategies offered a wide range of actions to cater to different stages of meat reduction. Participants were asked to pick actions they found challenging, thus creating their personal set of meat reduction tools. Throughout weeks 2–5, they planned a meat reduction action daily specifying when and how they would perform the action, as well as considering how to overcome barriers. Each subsequent morning they were asked whether they had managed to perform the action they had chosen on the previous day and if not they were asked to reflect on what they could do differently next time. Participants received weekly feedback on how their meat consumption compared to week 1 in terms of quantity and environmental and health impacts (Online Resource 4). Based on this information, they were asked to reflect on the usefulness of the actions they had attempted that week. Following week 5, participants entered a four-week maintenance phase (weeks 6–9), where they were asked to continue performing the actions they found useful during the intervention phase. Participants received access to downloadable materials, such as a detailed overview of the meat reduction actions and an action diary, to provide the opportunity to engage with the intervention offline. At FU1, participants completed an intervention evaluation questionnaire.
After the baseline week, participants were asked to try to reduce their meat consumption during the following eight weeks with no further guidance. At FU2, participants completed a questionnaire and were asked what strategies they had tried to reduce their meat consumption.
The primary outcome was the difference between groups in change in total daily meat consumption from baseline to FU1, based on 7-day self-reported meat intake in weeks 1 and 5.
The difference between groups in change in meat consumption from baseline to FU2, and from FU1 to FU2 was analysed as secondary outcomes. In addition, from baseline to both follow-ups, we compared the difference between groups in: (i) change in consumption of meat sub-types (red meat, processed meat, red and processed meat combined); (ii) change in meat-free self-efficacy; and (iii) change in meat-eating identity. We also assessed differences in the effect of the intervention on our primary outcome by meat-free self-efficacy level.
We explored the acceptability and feasibility of the intervention for reducing meat consumption through intervention evaluation questionnaire responses collected at FU1 and adherence throughout the study, respectively. Barriers to adherence to self-selected meat reduction actions were explored through the free-text responses to the daily action completion question, which was asked when respondents indicated they had not been able to perform their action during weeks 2–5 (active intervention period). Strategies reported by the control group participants in the strategy exploration questionnaire administered at FU2 were compared to actions taken by the intervention group from the suggested action list.
Quantitative analyses were conducted in Stata/IC Version 14.1. Qualitative data were coded and managed using NVivo 12 software. We published a statistical analysis plan on the Open Science Framework (28/04/2021) preceding the analyses .
For each participant and time point (baseline, FU1, and FU2), we calculated mean total daily intakes of meat, red meat, processed meat, and red and processed meat combined, and mean meat-free self-efficacy scores. Participants were grouped into tertiles of baseline self-efficacy (low-/medium-/high- self-efficacy). Meat-eating identities were grouped into three higher-level categories: (i) non-meat-eating identity; (ii) reduced meat-eating identity; (iii) meat-eating identity. We created a dummy variable for ‘positive meat-identity change’ for both follow-ups. This was coded as 1 if a positive meat-identity change had occurred or 0 if a positive meat-identity change had not occurred.
The member of the research team analysing the primary outcome was blinded to group allocation. A linear regression model was used to determine whether the change in mean daily meat intake from baseline to FU1 differed significantly between the intervention and control groups.
Linear regression models were also used to explore changes in: (i) meat intake from baseline to FU2; (ii) meat intake from FU1 to FU2; (iii) intake of meat sub-types (red meat, processed meat, and red and processed meat) from baseline to both follow-ups; and (iv) participants’ meat-free self-efficacy scores from baseline to both follow-ups. We also explored differences in the effect of the intervention on our primary outcome by baseline self-efficacy by introducing self-efficacy tertiles as a predictor in the model. We employed a logistic regression model to determine whether the intervention increased participants’ odds of making a positive meat-eating identity change at both follow-ups. Written feedback collected from participants as part of the evaluation questionnaire, strategy exploration questionnaire and daily action completion questions were analysed qualitatively using inductive thematic analysis , with all responses coded and then grouped into broader categories of shared meaning.
We carried out a sensitivity analysis excluding days on which participants’ total meat intake exceeded 1.5 kg to assess the effect of outliers. All participants in our final sample completed > 4 meat frequency questionnaires during the baseline week and FU1 so we did not carry out the second sensitivity analysis stated in our pre-published statistical analysis plan.