Background

Red and processed meat consumption is associated with an increased risk of developing some non-communicable chronic conditions, including cardiovascular disease [1,2,3], type-2-diabetes [3,4,5], and some forms of cancer [6,7,8]. Additionally, livestock negatively affects the natural environment and advances anthropogenic climate change [9,10,11]. These environmental changes might in turn affect human health by contributing – among other things – to the pollution of air and drinking water, the rise in antimicrobial resistance, and the spread of vector-borne diseases [12,13,14]. While the potential health and environmental benefits of reducing meat consumption are well established, concerns about a consumer backlash and the poor understanding of how to promote this behaviour change have contributed to a general state of inaction [15,16,17]. Interventions targeting conscious determinants of human behaviour, such as those providing information, are generally perceived to be acceptable approaches to promote health behaviours by the public in developed countries [18, 19] and might therefore help to overcome this state of inaction. Furthermore, interventions targeting conscious determinants of human behaviour might, over time, enhance the public’s support for more structural interventions aiming at reducing the demand for meat [15, 16, 20]. Accordingly, identifying effective interventions targeting conscious determinants of human behaviour to reduce the demand for meat is an important step towards promoting healthier and more environmentally sustainable diets. The aim of this systematic review is to synthesise the evidence from studies evaluating the effectiveness of interventions targeting conscious determinants of human behaviour to reduce the actual or intended consumption, purchase, and selection of meat and to identify combinations of intervention characteristics that effectively promote this behaviour change.

Methods

Protocol registration and eligibility criteria

A protocol for this systematic review was published on PROSPERO [21]. This review includes studies evaluating interventions targeting conscious determinants of human behaviour to reduce the consumption, purchase, or selection of meat, and that fulfilled the eligibility criteria outlined in Table 1.

Table 1 Eligibility Criteria
Full size table

We extracted data on participants’ demand for meat, defined as the actual or intended consumption, purchase, or selection of meat. We refer to meat purchases when the selection of meat involves a real or virtual monetary transaction, while we refer to meat selection when no form of monetary transaction is involved. Where reported, we also extracted data pertaining to attitudes, subjective social norms, and perceived behavioural control of eating, purchasing, or selecting meat, and on body weight, blood pressure, blood glucose, and blood lipids. When an outcome was assessed in multiple ways we only extracted data for the most granular measure (e.g. food diary > self-reported score of change) referring to the longest time-span (e.g. consumption over a month > consumption over a week). There were no exclusion criteria pertaining to the length of follow-up, publication status, publication year, or language.

Search strategy and data extraction

We searched six electronic databases on the 31st of August 2017 using a pre-specified search algorithm (Additional file 1: Table S1). We searched for further eligible records contacting researchers and experts, manually conducting forward and backward reference searches, and screening publicly accessible online resources following the methodology described by Stansfield et al. ( [22], Additional file 1: Table S2). Two members of the research team independently assessed the eligibility of the studies, extracted data from eligible records using a pre-piloted data extraction form, and evaluated the methodological quality of all eligible studies using the Quality Assessment Tool for Quantitative Studies [23]. Where additional information was required, we contacted authors and/or reviewed study protocols. Disagreements were resolved by discussion and by referral to a third member of the research team.

Data synthesis

We synthesised results narratively grouping interventions in five categories: (1) individual lifestyle counselling, (2) goal-setting and self-monitoring interventions, (3) non-tailored education (about meat consumption and one or more of health, the environment, animal welfare, socio-economic issues), (4) tailored education, and (5) interventions implicitly highlighting animal suffering. We augmented our narrative synthesis with an exploratory crisp-set Qualitative Comparative Analysis (QCA). In our QCA we used a binary coding system to categorise each intervention as presenting or not presenting specific behavioural strategies or implementation features to identify configurations of intervention characteristics associated with (and those not associated with) statistically significant (p < 0.05) reductions in the demand for meat at the follow-up closest to the intervention completion. Where studies reported results for multiple eligible outcomes we only considered the outcome closest to ‘actual behaviour’ (actual behaviour > behaviour in virtual environments > intention). The intervention characteristics included in QCA pertained to whether an intervention provided information about the (a) health, (b) environmental, (c) animal welfare, (d) socio-economic, (e) and multiple consequences of meat consumption, and whether it (e) provided tailored information, (f) employed self-monitoring strategies, (g) employed goal-setting strategies, (h) implicitly highlighted animal suffering, (i) implemented individual lifestyle counselling sessions, (j) targeted people with, or at increased risk of ill-health (k) provided information on ‘how to’ eat less meat, and whether (l) the outcome was actual as opposed to intentions or behaviour in virtual environments. In QCA we included comparisons between interventions and no-intervention controls or pre-intervention baselines, while comparisons between multiple interventions were excluded. A more detailed account on QCA in systematic reviews is outlined by Thomas et al. [24].

Results

Study selection

We screened 10,733 titles/abstracts, assessed the full-text of 60 papers, and included 24 papers reporting on 29 studies and 59 intervention conditions (Fig. 1). The majority of papers (54%) were published or written between 2015 and 2017.

Fig. 1
figure 1

Study flow diagram

Full size image

Study characteristics

This review includes 25,477 observations (i.e. individuals or individual meal purchases) at the follow-up closest to the intervention completion and 305 observations at the longest follow-up for the four studies reporting on multiple end-points. Where reported, mean age ranged from 19 to 74 years (median = 36), the proportion of female participants ranged from 49 to 100% (median = 61%), the proportion of white participants ranged from 58 to 83% (median = 75%), and participants’ education and/or household income were high in most studies. Forty-five percent of studies only included individuals who ate meat. Studies recruited individuals (N = 26) or food providers (N = 3) through active recruitment (N = 15), advertisement and passive recruitment (N = 8), or through online panels (N = 6). Studies were conducted in Europe (N = 12), the USA (N = 9), China (N = 2), New Zealand (N = 2), and Brazil (N = 1), and were implemented in various settings, including experimental settings (N = 18), canteens or food services (N = 3), healthcare settings (N = 2), and among free-living individuals (N = 6). Fourteen studies employed a randomised controlled trial (RCT), 5 employed a non-randomised controlled trial (CT), 1 study employed a crossover design, 7 employed a pre-post design, and 2 studies retrospectively evaluated the intervention effectiveness. Retention at the shortest follow-up ranged from 59 to 100% (median = 95%). The overall methodological quality was ‘strong’ for 6 studies, ‘medium’ for 10 studies, and ‘weak’ for 13 studies. Table 2 summarises the characteristics of all studies included.

Table 2 Study-level characteristics by study design underlying the main comparison reported in this review
Full size table

Interventions and outcomes

We included 37 interventions providing non-tailored information about meat consumption and health (N = 11), the environment (N = 8), animal welfare (N = 2), socio-economic issues (N = 2), or a combination thereof (N = 14). Ten interventions provided tailored education, four implicitly highlighted animal suffering, six delivered individual lifestyle counselling, and two delivered self-monitoring interventions. Of all interventions included, two interventions targeted food suppliers while the rest targeted consumers. Fifteen studies reported on participants’ actual meat consumption, 6 on meat purchase or selection, and 15 on intended consumption. Fifteen studies additionally reported on other pre-specified secondary psychosocial outcomes or biomarkers of health risk. Table 3 summarises each intervention and its impact on, or association with participants’ demand for meat at the shortest follow-up. The results for the longest follow-up and for other secondary psychosocial and biological outcomes are summarized in Additional file 1: Tables S3, S4, and S5.

Table 3 Interventions’ impact on or association with the demand for meat at the shortest follow-up
Full size table

Interventions’ impact on the demand for meat

Individual lifestyle counselling

Six studies (N = 2 RCT, N = 1 CT, N = 3 pre-post) evaluated the effectiveness of six lifestyle counselling interventions to reduce red and/or processed meat consumption [25,26,27,28,29,30]. All found evidence that lifestyle counselling led to [25, 26] or was associated with reduced meat consumption [27,28,29,30]. However, two studies measuring red and processed meat separately, only found significant reductions in the consumption of the latter [28, 29]. Lifestyle counselling interventions were delivered individually by a trained health professional through multiple face-to-face and/or phone sessions and lasted from 6 weeks [28, 29] to 1 year. All counselling interventions additionally comprised written supporting material, with two interventions tailoring this material to individuals [25, 26]. Lifestyle counselling promoted behavioural changes other than meat consumption, including one or more of fruit and vegetable consumption, multivitamin supplement usage, smoking cessation, physical activity, weight loss, and reduction in alcohol consumption. Most counselling interventions (N = 5) targeted individuals affected by, or at increased risk of chronic diseases [25, 27,28,29,30], and only one such intervention targeted healthy working-class individuals [26].

Self-monitoring interventions

Evidence from two RCTs suggested that two self-monitoring interventions reduced red [31] and processed meat consumption [32] during the intervention period and increased intentions to not exceed recommended levels of meat consumption over the week following the intervention. Both interventions lasted 1 week and comprised daily text-messages encouraging participants to self-monitor their red or processed meat consumption with the goal of not exceeding pre-specified recommendations. The intervention focussing on processed meat additionally encouraged participants to think about ‘the regret they could experience’ if they exceeded the recommendations [32].

Education on meat consumption and health

Five studies (N = 2 RCT, N = 1 CT, N = 1 CO, N = 1 Pre-post) evaluated seven interventions providing written information [33,34,35,36] or informational videos [37] about the health consequences of eating meat. Of these, five interventions led to [33, 37], or were associated with intended reductions in meat consumption directly post-intervention [35, 36]. Among these studies, one RCT found this effect to be sustained 1 week after the intervention [37]. Conversely, neither a cognitively framed nor an affectively framed message about the health consequences of eating meat were found to be associated with intended reductions in meat consumption directly post-intervention and/or 3 weeks later [34]. Two studies (N = 1 RCT, N = 1 CT) assessed the impact of two health focussed educational interventions on actual meat consumption but neither found evidence of an effect [34, 37]. Finally, one controlled trial including elderly people compared the effect of four interventions on the selection of meat in a virtual environment and found that messages on meat consumption and health more effectively reduced the selection of meat products when framed factually (i.e. describing the causal link between meat consumption and its consequences) rather than pre-factually (i.e. outlining hypothetical consequences of hypothetical present meat consumption) [38]. Conversely, messages on meat consumption and well-being more effectively reduced intended meat consumption when framed pre-factually rather than factually [38].

Education on meat consumption and the natural environment

Six studies (N = 3 RCT, N = 1 CO, N = 2 Pre-post) evaluated eight interventions providing written information about the environmental consequences of meat consumption [33, 35, 36, 39,40,41]. Six interventions led to [33, 39, 40] or were associated with increased intentions to eat less meat [35, 36] or with fewer meat products being selected in a simulated food choice experiment [40]. These interventions provided information in the form of paragraphs [35, 39], brief articles [36, 40], or websites [33] and two interventions also outlined practical strategies to aid reductions in meat intake [33, 35]. Conversely, one pre-post study found no evidence to suggest that two informational posters on livestock’s water footprint reduced meat purchases in a university canteen [41].

Education on meat consumption and animal welfare

Two studies (N = 1 crossover, N = 1 pre-post) evaluated two interventions providing written information about the animal welfare implications of consuming meat [35, 36]. Both were associated with significant reductions in intended meat consumption, and showed more promise than comparable messages on the impact of meat consumption on health or the environment. There was no study assessing the impact of such interventions on actual consumption, purchase, or selection of meat.

Education on meat consumption and social issues

Two studies (N = 1 CT, N = 1 pre-post) evaluated two interventions focussing on the social consequences or antecedents of eating meat [36, 42]. Providing information about the association between pursuing high meat diets and holding social dominance values was not found to be associated with reductions in intended consumption directly post-intervention or with actual meat consumption 3 weeks later [42]. Conversely, reading an article on the adverse social consequences of high meat diets was associated with increased intentions to eat less meat, though it offered less promise than similar articles on meat consumption and health, the environment, or animal welfare [36].

Education on multiple consequences of meat consumption

Nine studies (N = 4 RCT, N = 1 CT, N = 2 Pre-post, N = 2 retrospective evaluations) assessed 14 interventions providing written information about multiple consequences of meat consumption [34, 43,44,45,46,47]. These interventions provided printed or online information about two or more of health, environmental, animal welfare, social, personal appearance, and economic consequences of eating meat. The impact on actual meat consumption was evaluated in seven interventions: two were associated with lower meat intakes directly and 4 weeks post-intervention in pre-post studies [45], four were not found to effectively reduce meat intakes in RCTs [44], and there was insufficient data to make inferences about the effectiveness of the last intervention [34]. There was no evidence that any of five interventions reduced intended consumption directly post-intervention [43] or 3 weeks later [34], while a pre-post study suggested that providing information about the health and environmental consequences of meat consumption was associated with reduced meat selection in a virtual food choice experiment [46]. Actual food purchases were measured following two interventions providing the Meatless Monday toolkit to the US government/corporate accounts or the US healthcare accounts of a large food service company. Only the latter intervention was associated with significant declines in meat purchases [47]. In summary, four of 14 interventions providing written information about multiple consequences of eating meat were associated with significant reductions in meat consumption, purchases or selection, though none of these interventions was assessed in a RCT.

Tailored education

Four RCTs evaluated ten tailored educational interventions, none of which was found to effectively reduce actual [44] or intended meat consumption [43]. Specifically, providing information that was matched to participants’ stage of change (for an account on the stages of change model refer to [48]) for eating less beef was not found to reduce beef consumption in two RCTs, and there was no robust evidence that this tailored intervention outperformed its stage-mismatched equivalent [44]. Two RCTs found no evidence that messages tailored to participants’ most valued consequence of meat consumption, and/or to participants’ self-schema (being responsible, or adventurous, or compassionate, or logical), and/or participants’ personal levels of meat intakes reduced their intended meat consumption [43].

Interventions implicitly highlighting animal suffering

Two RCTs found no evidence that any of four interventions implicitly highlighting animal suffering reduced intended meat consumption [49]. All such interventions employed a combination of visual and written material leading recipients to reflect upon the animal suffering involved in the production of meat [49]. For example, participants were shown a picture of a cow accompanied by the statement that the cow will be send to a slaughterhouse, and were asked to think about what would happen to the animal. None of the aforementioned studies assessed actual meat consumption or purchases.

Qualitative comparative analysis

Fifty-five comparisons were included in QCA. Four configurations of intervention characteristics were consistently associated with reductions in actual or intended meat consumption in real or virtual environments among two or more intervention evaluations (Table 4), while six configurations were consistently not found to be associated with these outcomes (Table 5). QCA supported the findings of the narrative synthesis suggesting that the approaches that were consistently associated with reductions in actual meat consumption were self-monitoring and lifestyle counselling interventions. Non-tailored information provision about the detrimental health or environmental consequences of eating meat was consistently associated with reductions in intended consumption or purchases/selection of meat in virtual environments but was not found to be associated with changes in actual behaviour. Tailored and non-tailored interventions elaborating on several consequences of eating less meat and interventions implicitly highlighting animal suffering were not found to be associated with actual or intended consumption, purchase, or selection of meat in real or virtual environment.

Table 4 Configurations of intervention components associated with reductions in meat consumption, purchase, or selection in QCA
Full size table
Table 5 Configurations of intervention components not found to be associated with reductions in meat consumption, purchase, or selection in QCA
Full size table

Secondary psychosocial outcomes

Self-monitoring interventions led to more favourable instrumental attitudes (beliefs about the impact of eating meat) but not affective attitudes (feeling about the hedonic aspects of eating meat) towards eating less meat [31, 32]. One self-monitoring intervention additionally enhanced perceived behavioural control [31], while there was no evidence that either influenced subjective social norms [31, 32]. In all five interventions evaluating the impact of providing information about the health consequences of eating meat and measuring attitudes, there was evidence that the intervention led to or was associated with more favourable attitudes towards eating less meat [33,34,35, 37]. Two out of four interventions providing information on the environmental impact of meat consumption led to or were associated with more favourable attitudes towards eating less meat [33, 35, 39]. One non-tailored educational intervention focussing on animal welfare [35] and another focussing on social antecedents of meat consumption were associated with less favourable attitudes towards meat consumption [42]. Conversely, none of the eight tailored educational interventions or five non-tailored interventions focussing on multiple consequences of meat consumption were found to influence attitudes [34, 43].

Secondary biological outcomes

The only biomarker of health on which studies reported was weight. Of four studies (N = 1 CT, N = 3 pre-post) evaluating the impact of lifestyle counselling interventions on weight [27,28,29,30] only one intervention, which explicitly focussed on weight management, was associated with a significant reduction in BMI in a pre-post comparison [29].

Discussion

Evidence from experimental intervention studies suggests that some interventions targeting conscious determinants of human behaviour could contribute towards reducing the demand for meat. Six lifestyle counselling interventions designed to reduce red and processed meat consumption led to [25, 26] or were associated with reduced meat consumption [27,28,29,30]. Evidence from RCTs suggested that two self-monitoring interventions reduced red [31] or processed meat consumption [32] during the intervention period and enhanced intentions to not exceed recommended levels of consumption. Providing non-tailored information on the health, or environmental, or socio-political, or animal welfare consequences of eating meat was associated with reduced intended consumption and virtual selection of meat in 16 out of 17 interventions [33,34,35,36,37, 39,40,41,42]. However, there was no evidence that any of six such interventions influenced actual behaviour [34, 37, 41, 42]. Only four of 14 interventions providing information about multiple consequences of eating meat were associated with reductions in meat consumption, purchases, or selection [34, 43,44,45,46,47]. None of the ten tailored educational interventions or of the four interventions implicitly highlighting animal welfare issues was found to reduce actual [44] or intended meat consumption [43, 49]. Participants’ attitudes towards eating (less) meat generally mirrored their intentions to do so. There was no robust evidence pertaining to any of the other secondary psychosocial outcomes or biomarkers of health risk.

Strengths and limitations

In this systematic review we employed gold standard methods striving to achieve robust and unbiased results. The scope of our review was kept broad to allow capturing a wide range of intervention approaches, which is important to inform this novel field of research. However, the results should be interpreted in the context of limitations of this review and of the studies included, most of which were of medium or low methodological quality. Having employed extensive methods to identify literature beyond the databases searches, over 40% of the papers included were unpublished records. This potentially prevented publication bias affecting the review unduly, but it also increased the proportion of studies with methodological limitations. Given this field is developing, we elected to include non-randomised studies, which increases the risk of bias in the findings. Additionally, the main outcome measures of this review presented challenges, with self-reported measures of consumption being prone to bias [50], food selection in virtual settings potentially lacking external validity [51, 52], and behavioural intentions being only moderately related to future behaviour [53]. Most studies only measured outcomes during or shortly after the intervention, so conclusions cannot be drawn on lasting effects. Many interventions were either underpowered, making potentially effective interventions difficult to detect, or were evaluated in non-randomised designs, precluding direct causal inference of effectiveness. Future studies should employ well-powered RCTs with longer follow-up to allow for causal inferences to be drawn on the interventions’ effectiveness and to better understand their longer term impact. The studies reviewed included predominantly white and well-educated volunteers, limiting the generalizability of the data to other population groups. Particular caution should be exercised when interpreting results of lifestyle counselling, as our search algorithm might have failed to identify all such interventions. Titles and abstracts of these studies often referred to ‘multiple health behaviours’ as the intervention target, and it is possible that papers specifically mentioning ‘meat consumption’, and therefore identified by our search algorithm, were more likely to be those finding significant results for this outcome. None of the interventions directly targeted gender-related barriers to reduce the demand for meat. Considering the importance of gender as a determinant of meat consumption, future studies should explore whether interventions targeting gender-related barriers can effectively reduce the demand for meat. In the absence of a taxonomy to classify different interventions targeting conscious determinants of human behaviour we grouped interventions according to some of their key behavioural strategies and implementation features. Our synthesis was partly based on an exploratory crisp-set QCA, which represents a sophisticated technique to descriptively identify patterns within data, but not to infer causality of effects. In particular, combinations associated with no evidence of effectiveness should not be thought as ‘ineffective’, as ‘absence of evidence does not equal evidence of absence’ [54]. Similarly, since we did not exclusively include RCTs, combinations associated with effectiveness do not necessarily support an underlying causal mechanism. It must also be considered that QCA did not differentiate between studies of different quality and size. Finally, while the quality assessment tool employed in this review allowed assessing different study designs, caution must be exercised when comparing quality scores between different study designs.

Findings in the context of existing evidence

The results of this review were consistent with past research on behaviour change interventions targeting other health behaviours. Lifestyle counselling and self-monitoring interventions have previously emerged as promising approaches to change different eating behaviours [55,56,57]. However, the resources needed to implement lifestyle counselling pose a barrier to its scalability. Moreover, existing evidence on lifestyle counselling targeting the general population rather than people suffering from ill-health is less clear [55], suggesting that our findings on lifestyle counselling may not apply to the general population. Educational interventions providing rational reasons for eating less meat may have underpinned the observed changes in conscious intentions. Nevertheless, unconscious psychosocial and environmental cues that influence behaviour in real-life settings [20, 58] may have prevented these effects from translating into actual behaviour outside experimental settings. Interventions targeting unconscious determinants of human behaviour [20, 58, 59] may therefore play an important role to help overcome this intention-behaviour gap. We plan to review the effectiveness of these approaches elsewhere [60]. The non-significant findings on tailored education are in contrast with the literature highlighting the importance of tailoring information [61, 62]. Nevertheless our findings were based on only two papers that found no evidence of effectiveness for any of the interventions they tested, regardless of tailoring [43, 44]. It is therefore possible, that intervention characteristics other than tailoring or methodological limitations of the aforementioned studies contributed to these non-significant results. However, it is also possible that individuals do not substantially benefit from receiving information about issues that they already value as important consequences of meat consumption, as these arguments might have naturally exhausted their potential for prompting behavioural change. Finally, while providing information was not found to directly influence behaviour, future research should explore whether this approach might contribute towards reducing population-wide demand for meat in other ways. For example, providing information on the benefits of eating less meat might increase the public’s acceptability for more structural interventions to reduce meat consumption.

Conclusion

This review is the first systematic synthesis of evidence about the effectiveness of interventions targeting conscious determinants of human behaviour to reduce the demand for meat. Some interventions targeting conscious determinants of human behaviour have the potential to reduce the demand for meat. In particular, self-monitoring and individual lifestyle counselling interventions showed promise in reducing actual consumption of meat. Education about health, environmental, socio-political, or animal welfare consequences of eating meat can reduce intended meat consumption and selection of meat in virtual environments, but there was little evidence on whether this approach influenced actual behaviour and the few studies examining this found no evidence that it did. Interventions providing information on several consequences of meat consumption, those implicitly highlighting animal suffering in the context of meat production or consumption, and those providing tailored information offered less promise. While the impact of interventions targeting conscious determinants of human behaviour was modest, if delivered at scale these interventions could contribute towards reducing the demand for meat at population-level.