Identifying effective interventions to promote consumption of protein-rich foods from lower ecological footprint sources: A systematic literature review

Addressing overconsumption of protein-rich foods from high ecological footprint sources can have positive impacts on health such as reduction of non-communicable disease risk and protecting the natural environment. With the increased attention towards development of ecologically sustainable diets, this systematic review aimed to critically review literature on effectiveness of those interventions aiming to promote protein-rich foods from lower ecological footprint sources. Five electronic databases (Medline, Web of Science, Scopus, Embase and Global Health) were searched for articles published up to January 2021. Quantitative studies were eligible for inclusion if they reported on actual or intended consumption of protein-rich animal-derived and/or plant-based foods; purchase, or selection of meat/plant-based diet in real or virtual environments. We assessed 140 full-text articles for eligibility of which 51 were included in this review. The results were narratively synthesised. Included studies were categorised into individual level behaviour change interventions (n = 33) which included education, counselling and self-monitoring, and micro-environmental/structural behaviour change interventions (n = 18) which included menu manipulation, choice architecture and multicomponent approaches. Half of individual level interventions (52%) aimed to reduce red/processed meat intake among people with current/past chronic conditions which reduced meat intake in the short term. The majority of micro-environmental studies focused on increasing plant-based diet in dining facilities, leading to positive dietary changes. These findings point to a clear gap in the current evidence base for interventions that promote plant-based diet in the general population.

Introduction developed by the research team and with the help of research librarian. Two researchers (GS and GR) conducted the search independently through all databases. Then, potential articles were imported into Covidence (covidence.org) where duplicates were removed. The screening of search results was conducted and recorded using the PRISMA checklist, by two researchers (GS and GR) independently and in consultation with a third researcher (RR). First, two researchers (GS and GR) independently performed the title and abstract screening of all imported studies against inclusion and exclusion criteria. Where a consensus regarding the inclusion of a study between the first and a second researcher was not reached, it was resolved with two other researchers (RR and SG). Then, full-text versions were obtained for all studies identified to be suitable in the first stage of data screening and reviewed by two researchers (GS and GR) independently. The reference lists of all included studies were hand searched for relevant studies not identified in the first search strategy. Authors of identified studies and experts of the field were consulted, where further details were required. The PRISMA flow diagram was used to document the number of articles at each screening stage (see Fig 1).

Inclusion Exclusion
Population All except those ones listed in the exclusion criteria. People diagnosed with clinical condition(s) for which it is required to consume specific amounts of red meat.
Intervention Interventions aiming to reduce the demand for red/processed meat and to increase in plant-based proteins including micro-environmental structural (physical) changes.
Dietary interventions aiming to promote a general dietary pattern.
Interventions with structural (physical) environment changes but with no evaluation.
Comparator No-or minimal-intervention controls, pre-intervention baseline, or other eligible intervention(s). -

Outcome
Objective or self-reported measures of demand for red/processed meat and/or plantbased protein, defined as actual or intended consumption, purchase, or selection of meat in real or virtual environments. https://doi.org/10.1371/journal.pgph.0000209.t001

Data extraction and analysis
The following information was extracted from included articles: author(s), year of publication, country of study, title, location, study type (structural/individual), intervention year, intervention length, target audience, sample characteristics, aims, intervention design, behaviour change theory/framework used, eligibility, recruitment, demographic characteristics, measure/tool(s) used, outcomes measured, results, follow up period, follow up results. Three researchers (GS, GR and RR) tested the data extraction sheet by extracting data from 10% of articles; minor disagreements were identified and discussed. Then, two researchers (GS and GR) extracted data from all included studies independently and then cross-checked all extracted data. Any disagreements were resolved in consultation with the third researcher (RR). The risk and sources of potential bias of each included study were assessed by two researchers (GS and GR) independently by using the Effective Public Health Practice Project Quality Assessment Tool (EPHPP) [22]. This tool was developed to assess the quality of a diverse group of empirical studies. Each included study was assessed on study design, selection bias, confounders, blinding, data collection method, withdrawals and dropouts and assigned to either 'strong', 'moderate' or 'weak' category. Finally, the overall rating was determined based on these ratings as indicated in the assessment tool dictionary. Any disagreements were resolved with a third researcher (RR). Finally, all quantitative data were summarised.

Study quality
The overall methodological quality for all included studies was 'strong' for 12 studies, 'moderate' for 28 studies and 'weak' for 11 studies. The quality assessment for each individual study on each individual criterion is provided in Table 5.

Individual level studies
Educational interventions. Twelve studies (RCT = 8, Pre/post = 4) used educational approach to reduce red/processed meat intake and purchase behaviour which included tailored education, educational classes, workshops and courses. Among RCT studies, six found positive impacts on the reduction in red meat intake in IG in comparison to CG The intervention comprised of 4-week multicomponent intervention that aimed to reduce meat intake including 5 components: a social norm; an informational/educational; an appeal to fear (animal farming/ harm); a mind attribution induction; and a goal setting/selfmonitoring. It also included 3 inlab sessions for the intervention group and 2 for the control group.

Self-Determination Theory
Meat intake (red meat, white meat, fish and cold cuts; in grams) in total, during the week and weekend. Assessed using a dietary journal at baseline (T1), 2 weeks (T2) and 4 weeks (T3) postintervention.
No significant changes between the groups were for total meat intake; total, week and weekend white meat intake; weekend red meat intake; weekend fish intake; and weekend cold cuts intake.    The intervention targeted dietary intake, with an emphasis on food groups. Participants were randomized to 3 groups (lifestyle intervention, metformin or placebo) for an average of 3 years. The lifestyle intervention involved a 16-session core curriculum over the first 24 weeks, followed by an individualized counselling curriculum (at least monthly contact).
Red meat intake (serves). Assessed using a FFQ at baseline and at 1, 5, 6 and 9 years.
Participants in the lifestyle arm had significantly lower red meat intakes compared with the participants in the metformin and placebo arms. This change was statistically significant at 5 and 9 years (p<0.05).

Transtheoretical Model
Red meat intake (portions/ week). Assessed using FFQ at baseline and after the 7-month intervention.
The daily intake of red meat was reduced in the IG (P = 0.001) but it was insignificant when comparing to CG (p = 0.063).

James et al, 2015 Australia
Adults The intervention targeted diet including meat intake. IG participated in 12-month program by biweekly phone calls, personal meetings, and monthly handouts, while targeting the intake of at least 400 g/day of fruits and vegetables, and no more than 500 g/week of red or processed meats.
Red/processed meat (RPM) intake (<500 g/week). Assessed using FFQ at baseline and after the intervention.
The IG improved their adherence and intake of RPM intake to the guidelines (<500g/week). Change in meat intake before and after the intervention IG = -219.6 (-778.1;-77.0).  Meat intake (serves/day). Assessed using the dietary questionnaire administered at baseline, 6 weeks, and 6 months.

Social Cognitive Theory
Red/processed meat (RPM) intake (serves/week). Assessed using survey at baseline and post intervention (6 weeks follow up).
There was a significant decrease in RPM intake of from baseline (Median = 1) to post intervention (Median = 0, p = 0.01). No changes for red meat intake pre-and post-intervention.

Social Cognitive Theory
Red/processed meat intake (serves/week). Assessed using questionnaire at baseline and 6 weeks post intervention.
(Continued ) (Continued ) [25,29,36,43,45,46], follow up varied between 3 months to 6 years. Two RCTs found that daily consumption of red meat was reduced in the IG, but it was not significant in comparison to CG [38,39]. Among pre-post studies: one study found a significant decrease in processed meat intake from baseline to 6 weeks [49]; one study found decrease in meat servings per week by 86% over 3 weeks period and significant increase in legume intake (pre-intervention 4.43 servings to 12.13 servings post-intervention) [52]; one study showed that intakes of red meat and poultry decreased significantly post intervention from baseline to 6 weeks [53]. Finally, one study found a decrease in purchases of meat (average dollars/week spent on meat) at baseline to 6 weeks [51].

Counselling interventions.
Eleven studies (RCT = 9, CT = 1, Pre/post = 1) used counselling approach to reduce red/processed meat intake (n = 10) and increase in soya intake (n = 1). These interventions included telephone and in person counselling sessions providing dietary advice. Among RCT studies, nine studies found positive impact on reduction in red/processed meat intake in IG in comparison to CG [24,28,30,[33][34][35]40,42,44] and increase intake in soyabean products [24], follow up period varied between 4 weeks to 24 months. One CT study showed that the IG improved their adherence of red/processed meat intake to the guidelines over 12-month program [48]. Pre-post study found that processed meat intake decreased preto post-intervention (6 weeks) but no changes for red meat intake were observed [50].
Self-monitoring interventions. Two studies used self-monitoring approach to reduce red meat intake. They were both RCTs and used daily text-messaging (SMS) approach to reduce red and processed meat intake. These studies urged participants to self-monitor meat intake and measured attitudes, intentions and anticipated regret [26,27]. It found positive impact on reduction in processed meat intake in IG in comparison to CG after one week [26], and red meat intake in IG in comparison to CG after two weeks [27].
Multicomponent interventions. Eight studies (RCT = 5, CT = 1, Pre/post = 2) used multicomponent approach to reduce red/processed meat intake such as education and self-monitoring [23], education and counselling approaches [31,32,37,41,47,54,55]. Among RCTs, three studies found no significant intervention effect on red meat/processed meat [31,32,41] and two studies showed significant reductions in red meat intake at 4 weeks [23] and over 9 years [37]. Also, a CT study showed significant reductions in red/processed meat intake among In study 2 across 7 menu conditions and testing a more expensive decoy also showed no effect of the intervention decoy absent vs. decoy present: OR 0.68 (95% CI 0.41 to 1.12). Ground and processed meat intake (serves) assessed using self-administered health habits questionnaire (HHR) including FFQ measured at baseline and after intervention. A significant decrease in ground and processed meats were observed among intervention companies (p = 0.05). (1) students were given menus upon entry into the dining hall to help them decide their meal; (2) a "dimsum" style vegetable cart was pushed throughout the dining hall; (3) prepared balanced meals were placed on display at the dining hall's entrance; (4) a Chef's Pick of the Day that included a portrait of the chef and a plated vegetarian meal were placed on display at the front entrance of the dining hall. CG: 4 weeks of business as usual.

Social Cognitive Theory Transtheoretical
Model Intake of fruits, vegetables and high-fat meat measured as servings per week assessed at baseline and after intervention using FFQ. Students in the intervention dining halls consumed significantly less junk food and high-fat meat and increased their perceived importance of eating a healthful diet (more fruits and vegetable servings per week) relative to the CG. In the IG, high-fat meat intake reduced by 0.9 servings per week and in CG increased by 0.9 servings per week; p = 0.04.
(Continued ) Health and environment Three restaurants were randomly assigned to a sequence of an intervention and control condition. In the intervention period, the vegetable portion sizes on the plates of main dishes were doubled (150g of vegetables instead of 75g) and the portion sizes of meat on the plates were reduced by an average of 12.5% for 6 weeks. In the control period, the portion sizes of the main dishes were maintained as usual for 6 weeks.
Vegetable and meat intake measured by subtracting the grams of meat returned to the kitchen from the average grams of meat in each of the targeted dishes. Vegetable consumption from plates was significantly higher during the intervention period (Mean = 115.5g) than during the control period (Mean = 61.7g). Total vegetable intake (including side dishes) was significantly higher during the intervention period (Mean = 178.0 g) than during the control period (Mean = 137.0g). Meat intake was significantly lower during the intervention period (Mean = 183.1g) than during the control period (Mean = 211.1g) p<0.001. The worksite intervention was 18-month multicomponent intervention aiming to increase fruit and vegetable intake, physical activity and reduce red meat intake and smoking. It included policies aimed at offering healthful food options at company meetings, interactive activities, and education.
Fruit, vegetable, and red meat intake measured in servings per week using FFQ at baseline and directly after intervention. At follow-up, 22% of workers at intervention worksites were eating 5 servings of fruits and vegetables per day, compared with between 12% and 15% of workers and managers at control worksites. The change in percentage of participants eating �3 servings per week of red meat did not differ between intervention (+4.1%) and control group (+3%), p = 0.72.

Kurz (2018), Sweden
Adults Setting: University restaurant Sample size: 2 university restaurants (IG = 1 and CG = 1) Health and environment The intervention tested if nudging can increase the consumption of vegetarian food. At the treated restaurant, the salience of the vegetarian option was increased by changing the menu order and enhancing the visibility of the vegetarian dish. The other restaurant served as a control.
Nudge theory Daily sales on three main dishes (one of which vegetarian) at the baseline (first 9 weeks) and intervention period of 17 weeks. The nudge increased the share of vegetarian lunches sold by on average 6% point, and that the treatment effect increased over time. The change in behaviour was partly persistent, as the share of vegetarian lunches sold remained 4% point higher after the intervention ended than before the experiment.

Polak et al. (2019), Israel
Adults and children Sales of plant-based breakfast and lunch dishes assessed at baseline and 8 weeks after dish names were changed. Changing the breakfast dish "Meat-free Breakfast" name to the alternatives of "Garden Breakfast" (OR = 1.  Table 5. Quality assessment of included studies.

Withdrawals and drop-outs
women with breast cancer after 12 months [47]. Two pre-post studies showed significant changes in decreasing red/processed meat intake among students and cancer survivors [54,55].

Micro-environment level studies
Menu manipulation interventions. Of 18 micro-environmental level studies, seven studies (RCT = 4, CT = 1, Exp = 3) used menu manipulation approach in order to reduce meat options or increase choice/sale of plant-based meal options [56][57][58][59]65,67,69,73]. Menu manipulation included different approaches such as adding attractive meat free choices on the menu, adding specific symbols specifying that less meat intake can save the environment, increasing the visibility of plant-based options, and describing the plant-based option as a 'Dish of the day'. Of four RCTs, two did not show a difference on the choice of plant-based options in IG compared to CG [56,57], and two showed a positive impact on meat reduction behaviour by choosing more plant-based options in restaurants in IG compared to CG [58,59]. Also, a CT study showed increase in sales of plant-based lunches [65]. One experimental study revealed a significant changes on plant-based dish sales by changing the language to explain plant-based options on café/restaurant menus (e.g. replacing Meat-Free Breakfast with Garden Breakfast) [67]. However, the other experimental study found that the nudging strategy a 'Dish of the day' did not show a difference on the choice of the plant-based option among adolescents [69].
Choice architecture interventions. Three studies used choice architecture approach (RCT = 2, Exp = 1) which included dining environment manipulation such as altering the serving sequence of plant-based and meat-based dishes [61], altering portion sizes of plant and meat-based foods [63] and using priming (environmental changes-adding green plants, herbs and green colour bowls), default (pre-portioned salad bowls) and perceived variety options (pre-mixed salad to increase the visual variety of vegetables) [68]. The manipulation of altering the serving sequence and default approach found no significant difference in selection of meat dishes between IG and CG [61,68]. However, the manipulation of altering portion sizes resulted in significant higher vegetable intake and lower meat intake in IG than CG [63] as well as using the priming and perceived variety conditions showed decrease in choosing meatbased options [68].
Multicomponent interventions. Seven studies (RCT = 3, CT = 1, Pre/post = 3) used multicomponent approach to reduce red meat intake which included combination of education, labelling, policy, counselling and choice architecture. Of three RCTs, two studies found significant decrease in ground and processed meat intake [60] and high-fat meat intake [62]. One RCT showed that percentage of participants eating �3 servings per week of red meat did not differ between IG and CG over 18-months [64]. One CT study and three pre-/post-studies showed significant reduction in processed meat intake [66] and red meat intake [70][71][72].

Discussion
This systematic literature review explored the effectiveness of interventions which aimed to promote protein consumption from low ecological footprint sources (reduction in animalderived proteins and increase in plant-based proteins). Most of individual level studies demonstrated reduction of animal-derived protein intake, mainly measured by reduction in red and/ or processed meat intake with only a few studies measured increase intake in plant-based proteins such as legumes and soyabeans. Furthermore, 52% of these studies (n = 17) targeted people with current or past chronic conditions such as cancer, diabetes and cardiovascular diseases. The majority of micro-environmental/structural level studies found positive dietary changes in reducing animal-derived protein intake mainly red meat and increase in plantbased protein (e.g. plant-based vegetarian dishes). The majority of these studies were conducted in dining facilities such as cafes, restaurants and canteens. Similar findings have been observed in a several systematic literature reviews which evaluated the effectiveness of interventions targeting to reduce demand for meat [74,75]. Educational, counselling and self-monitoring interventions are promising approaches to dietary behaviour change such as increase in environmentally sustainable protein intake which have been observed in other systematic literature reviews [76,77]. However, these approaches have been often used among at risk or highly motivated populations such as people with obesity or other chronic condition, cancer survivors, and may have limited success in changing behaviour among general population. In addition, research indicates that educational interventions on its own may not be sufficient to behaviour change in long term [77,78]. There is a need for further longitudinal studies to confirm that the reduction in animal-derived protein intake from high ecological footprint sources among healthy general population sustain over prolonged period of time.
The majority of our included interventions focused on reduction of red/processed meat intake mainly from a health perspective with only three studies emphasising the reduction of animal-derived foods due to negative impact on the environment. This is not a surprising finding as recent studies reported general population having low food literacy and limited understanding of food impact on the environment and often focus on changing their dietary behaviours such as reducing red/processed meat intake due to health reasons [79][80][81][82]. Therefore, there is a need to develop interventions aiming to educate general public on sustainable and healthy diets and evaluate how feasible these interventions are in reducing overconsumption of protein from high ecological footprint sources in order to reduce the negative impact it has on health and environment.
The findings from included micro-environmental behaviour change interventions showed that altering food environments using nudges or choice architecture can lead to positive dietary behaviour changes such as reduction in unsustainable protein intake/purchase and increase in plant-based meals which aligns with findings from previous studies [75,83]. Most promising approaches included altering portion sizes, menu manipulation by adding plantbased meal options and policy implementation; similar environmental approaches have been identified as successful in changing dietary behaviours among young adults [84]. In order to reduce the human behaviour towards more environmentally sustainable protein intake, it is important not only to change the supply but also the demand of unsustainable foods [78]. Interestingly, the most recent qualitative study indicated that young Australians were open to or interested in affordable meat alternatives such as plant-based meals and reported that often these options were not available or very limited when dining out (unpublished work). This indicates that people may be interested in changing their dietary behaviours to more sustainable and healthy but food environments need to be supportive in helping them to make informed food choices. Choice architecture, nudging strategies and policy implementation can be promising approaches to create enabling food environments and for changing dietary behaviour towards more sustainable diets. However, there is still a need to develop and test different strategies among more general population and settings to determine what motivates them in choosing more environmentally sustainable food options and if it leads to sustained behaviour change. In addition, there is a need to explore what would motivate food retailers to offer plant-based meal options and what the impact it may have on the food purchasing behaviour.
The main strength of this review is that a systematic approach was used and reported following the PRISMA guidelines to synthesise the evidence on the interventions aiming to promote protein intake from low ecological footprint sources. We included individual and microenvironmental level behaviour change interventions, which provides a more comprehensive picture on the effectiveness of interventions in changing people behaviour towards increase in environmentally sustainable protein intake. One limitation of this review is that most of the included studies have been conducted in high income countries and only a few studies were conducted in low-and middle-income countries (LMIC). This might be due to the fact that plant-based diet concept in high income countries has received increased attention in the last five years and LMICs have not prioritised it as a significant nutrition and environmental issue due to dealing with other diet related issues such as undernutrition and nutrient deficiencies. Research indicates that meat intake in LMIC has been associated with wealth as the rise in income has resulted in significant animal-derived food consumption in these countries [85]. Furthermore, most studies used self-reported measures to measure dietary behaviours which may increase biases [86]. Also, this review was limited to the literature published in English language and did not included articles published in grey literature, therefore it may be we missed some important research written in other languages. Finally, the majority of individual level behaviour change interventions included people who may be highly motivated to change their dietary behaviour such as cancer survivors, people at risk of developing chronic conditions, limiting the generalizability of the data to general population.

Conclusions
The present review identified effective individual and micro-environmental behaviour change interventions which showed promising results in reducing protein intake from high ecological footprint sources. The findings suggest that individual behaviour change interventions such as education, counselling and self-monitoring interventions might be useful strategies to educate people to change their dietary behaviours to more sustainable ones. However, there is a need to test these strategies among the general population longitudinally. In addition, our findings showed that altering food environments using nudging and choice architecture approaches can achieve positive dietary changes but there is a need for development and evaluation of interventions in general settings (macro-environments) and explore motivations in sustainable food purchasing behaviours. Our findings inform future research for development and evaluation of interventions and strategies to encourage greater adoption of sustainable and healthy diets.