This pilot study compares a healthy plus sustainable lifestyle intervention arm with a healthy lifestyle control intervention arm, hereafter referred to as the sustainable lifestyle (SL) and healthy lifestyle (HL) arm, respectively. After randomization, each arm received an explanation of the respective lifestyle concepts during workshop number 1. Participants in the HL arm were introduced to national recommendations that emphasize eating a varied diet rich in whole grains, fruits, and vegetables, along with limited consumption of red and processed meat, and engaging in 150-300 minutes of moderate-to-vigorous physical activity each week. Participants in the SL arm received the same health recommendations, framed within information on behaviors that also reduce environmental impact. This included choosing foods with lower greenhouse gas emissions, considering organically produced foods, and understanding the environmental benefits of active commuting such as walking or cycling. Because a pilot study may be considered a subset of feasibility studies, this study evaluates both the piloted trial procedures (eg, primary outcomes, delivery of workshops at the office locations, and intervention functions) and the feasibility aspects (eg, recruitment data, participant workshop attendance rate, and alignment of goals set by participants) to inform the design of a future full-scale SOFIA randomized controlled trial. A detailed study protocol of the SOFIA study as a full-scale cluster randomized, single-blind controlled trial has been described previously [34], and data have been reported according to the CONSORT (Consolidated Standards of Reporting Trials) checklist (Checklist 1).

This study was co-created with 2 offices in Sweden and part of a large body of research within the Concepts for the Sustainable Office of the Future (SOFCO) project. Office A was located in a medium-sized city and worked in real estate development. At the time of the investigation, the office included approximately 35 office workers. The workspace consisted of modern coworking areas, shared workspaces, meeting rooms, and social areas. Office B was located in a large city and delivered facility management and property management services. At the time of the intervention, the office contained approximately 400 office workers. Although larger, Office B featured workspaces similar to those in Office A, including modern offices designed for flexible work and remote work opportunities, in line with the new ways of working, that is, employees are free to work from anywhere and anytime, and the office lacks individual and assigned office space for each person [35]. Both offices were situated in urban environments on 2 or more floors in their respective buildings, surrounded by restaurants, shopping centers, and fitness facilities and within walking distance to public transport and parks. For additional details about the setting, see Ullberg et al [36].

Participants were recruited through a systematic multistep procedure. First, study information was communicated to managers at each office, who then shared the information internally. Second, recruitment advertisements were posted at the offices. Third, all office employees received an email invitation to participate. Finally, members of the research team visited the workplaces in person to provide information and recruit participants on site. Employees who expressed interest completed a short web-based screening questionnaire provided through the advertisement. Eligibility criteria included being at least 18 years of age, having access to a smartphone, and working 20 hours or more per week in one of the 2 offices. Upon confirmation of interest, participants were invited by email to attend a baseline assessment meeting at their workplace one week before the first educational workshop. The aim of the meeting was to provide information about the study, sign the informed consent form, and do baseline assessments.

After the baseline assessment, the 2 office sites were randomized as clusters using a computer-generated random number generator in the statistical software R (R Core Team). This allowed all participants to be assigned to either the sustainable or healthy arm based on their office’s geographical location [34]. Data collection was conducted in the fall of 2022 and repeated for a second cohort in the spring of 2023. Because the study included only 4 small clusters in total, the process functioned as 2 separate 2-arm allocations, each with equal probability of assignment. Note that the offices (A and B) received both interventions (SL and HL), but at different time points, as the randomization was repeated for each cohort. Participants were not aware of their allocation, as the different office-based clusters were geographically separated. However, the research team was not blinded and had access to files identifying each study arm allocation status, as they had to coordinate the intervention and data collection. We used purposive sampling, as only employees from the 2 companies with established agreements were eligible and invited to participate in the study. However, key outcome measures from this pilot study will be used to inform future power calculations and determine the required sample size of the future upscaled study, as outlined in the study protocol [34]. In total, 37 participants were recruited, of which 21 were included in the SL arm and 16 in the HL arm. The uneven distribution of participants in the SL and HL arms, respectively, was a result of the cluster randomization process, since different numbers of people signed up at the different offices. Given the small sample size and unequal group allocation, the study is not powered to detect between-group differences; therefore, all results should be regarded as exploratory.

The intervention lasted for 8 weeks and consisted of 6 interactive educational workshops that occurred every week for the first 4 weeks and then every second week for the last 4 weeks, facilitated by study personnel (OHU and KB). Diet and physical activity data were collected at baseline and just before workshops 4 and 6 to make sure that data were assessed according to the timeline shown in Figure 1, and if needed, participants were reminded at the workshop. Data addressing the secondary aim—namely, evaluating the feasibility of the intervention functions—were collected at workshops 2, 4, and 6 as part of the behavioral change support activities. Additionally, participants in both study arms used the Stanford Healthy Neighborhood Discovery Tool (Discovery Tool) [37] to address the social-ecological aspects of behavior change; these data were collected between workshops 5 and 6. However, the results have been presented elsewhere [36]. Moreover, urine samples were collected for analysis of pesticide residues, as an indicator of consumption of organic foods, at baseline and workshop 6, but the result is beyond the scope of this paper.

Michie et al [32,38] Behavioral Change Wheel and techniques from the behavior change taxonomy served as the framework for this intervention, based on the COM-B (capability, opportunity, and motivation to support behavior change) model, which explains individual behavior change through 3 fundamental mechanisms: capability (C), opportunity (O), and motivation (M), which together support behavior change (B). The interactive educational workshops included discussions about how to change behavior at the individual level to achieve a more sustainable and healthier lifestyle, respectively. In the context of this study, capability (C) referred to knowledge, planning, and culinary skills related to a healthy and sustainable lifestyle; opportunity (O) referred to food options at work and areas for engaging in physical activity; motivation (M) referred to personal factors that motivate behavioral change (B), such as a desire to improve one’s health, increase work productivity, reduce one’s global carbon footprint, and contribute to reducing local food waste. In brief, workshops 1-6 used COM-B-based intervention functions to support behavior change. Education was delivered through framing and reframing, highlighting that sustainable lifestyle choices can also promote health in the SL arm. Environmental restructuring was introduced by using the workplace as a social arena and through the Our Voice Discovery tool (Stanford University); moreover, by serving lunch during the workshops to model recommended eating practices aligned with the nutritional recommendations of the Nordic Nutrition Recommendations (NNR) [39]. Workshops 2 and 4 emphasized incentivization through goal setting and used prompts and cues, including guidance on interpreting food labels such as the Swedish Keyhole symbol and KRAV (the Swedish label for organically produced foods). Throughout workshops 1-7, enablement and action planning were supported by education and the digital behavioral package (guidelines, food labels, and recipes to support adherence to the recommended advice about sustainable and healthy lifestyles, respectively). For educational content, see Table 1, and for an even more detailed description of program theory, supporting factors, and intervention functions, see published study protocol [34].

The participants in the SL arm were instructed to follow the dietary and physical activity recommendations of the World Health Organization (WHO), the Food and Agriculture Organization of the United Nations (FAO), and NNR [39-41]. Additionally, participants were instructed to adopt proenvironmental behaviors in both their work and private lives, defined as actions that minimize environmental harm and combat climate change [42]. For example, increase active transportation to and from work; use active transportation in and around the office workplace; break up prolonged sitting through walk-and-talk meetings; select food items with a low carbon footprint; and purchase organic food instead of conventionally produced food, as seen in Table 1.

The participants in the HL arm were instructed to follow the dietary recommendations by WHO and NNR 2014, as well as to follow recommendations about the “plate model” (ie, the combination of food on the plate intended to adhere to the recommended intake of nutrients) and to engage in physical activity corresponding to at least 150-300 minutes per week of moderate-vigorous intensity [39,40]; see Table 1.

Descriptive statistics were used to assess baseline characteristics, workshop attendance rate, goal fulfillment, and social support. Differences in baseline characteristics, goal fulfillment, and social support between the 2 arms were tested using the Student t test for numerical variables and a chi-square test for categorical variables. Linear mixed effects models (LMMs) were used to generate mean values and between-arm differences with 95% CIs, and the interaction effect was tested based on the II Wald chi-square test between the time points and arms, with degrees of freedom set to 1 (Multimedia Appendix 1). Finally, the Student t test was used to analyze the within-arm mean values and 95% CI between baseline and week 8 separately for each arm, respectively. All participants who were randomized and had answered baseline questionnaires were included in the analysis and analyzed according to intent-to-treat principles, regardless of drop-out, loss to follow-up, or adherence rate. All analyses assumed that numerical variables were normally distributed at the population level [51]. However, clustering was not considered in this study. All quantitative data were analyzed using R statistics version 4.3.2 (October 31, 2023).

Printed questionnaires with dietary goals, self-rated goal fulfillment, barriers, and facilitators were transferred to electronic format and analyzed deductively using framework analysis [52]. By reading the goals and narratives independently several times, an initial coding framework was developed, informed by the COM-B model. Thereafter, Excel (Microsoft Corp) spreadsheets were developed, incorporating codes, goals, barriers, and facilitators, resulting in thematic framework matrices. The matrices were filled with goals and participant-specific data by the first author (OHU). As data were compared, the framework was refined and reorganized to accommodate emerging codes and remove those that were no longer relevant. Each goal was coded into categories and analyzed separately for each arm. Moreover, if a goal included 2 dietary aspects, for example, “Eat more fruits and vegetables and increase the proportion of organically produced fruits and vegetables,” it was categorized into 2 different categories. Barriers to and facilitators of behavior change were initially analyzed separately but later combined due to overlap between the SL and HL arms. Thematic codes were then developed as analytic memos to support the creation of themes. Within these broad categories, COM-B mechanisms were identified inductively by interpreting the texts. The organization of the data into COM-B mechanisms and the identification of categories were discussed, refined, and agreed upon by OHU and KB. Final categories were discussed and agreed upon by all researchers, and tables were created to illustrate the categories, codes, COM-B mechanisms, and descriptive quotations.

The study was approved by the Swedish Ethical Review Authority located in Uppsala (#2021-02309) and conducted in accordance with the guidelines of the Declaration of Helsinki. All research activities strictly followed the regulations set by the Swedish Research Council concerning informed consent and confidentiality. Written informed consent was obtained from all participants prior to any research activities, and trained personnel from the research team were responsible for data management to ensure confidentiality. Moreover, participants were informed of their right to withdraw from the study at any time and to request removal of their data. Participation occurred during working hours and participants were covered by insurance through their respective employers. No financial compensation was provided.

Baseline characteristics of the study participants are shown in Table 2, and there were no statistically significant differences between the 2 arms. The majority of participants were women (28/37, 76%) and had a university degree (33/37, 89%), and all participants were nonsmokers (37/37, 100%). The mean age was 43 years, and the mean BMI was 24. Dropout after randomization was 2 participants in the SL arm and 2 participants in the HL arm; see Figure 2, and the main reason for dropping out was lack of time to participate in the workshops. The overall participant attendance rate at the educational workshops was high, 80% for the SL arm and 78% for the HL arm (Multimedia Appendix 2), and the general reasons for nonattendance were sickness or high workload.

In order to estimate changes in the quality of food intake, nutrients and diet-related carbon footprint (ie, kg CO₂e) were expressed per 1000 kcal per day in Table 3, and there were no statistically significant differences between the SL arm and the HL at baseline. When testing the change of daily carbon footprint between arms, a significant interaction effect was found between time and intervention arm (P=.05), suggesting that the various educational workshops in the SL and HL arms, respectively, had different impacts on carbon footprint over the 8 weeks of intervention, as illustrated in Figure 3. No other between-arm interactions were found for total energy or macro- and micronutrients in the energy-adjusted model (see Multimedia Appendix 1).

However, when analyzing within-arm changes over 8 weeks, the daily intake of vitamin C increased statistically significantly over time in the HL arm, with 12.5 mg per 1000 kcal (P=.003). Moreover, total energy decreased statistically significantly over time with 346 kcal per day for the HL arm (P=.01), but no other within-arm changes were found for macro- and micronutrients in the energy-adjusted model in Table 3.

In order to estimate changes for the absolute food intake, nutrients, and carbon footprint were expressed per gram or kilogram in Multimedia Appendices 1 and 3, but no between-arm interactions were found (Multimedia Appendix 1). Analyses of within-arm absolute changes show a statistically significant decrease in the HL arm for total carbohydrate intake by 40.5 grams per day (P=.03), total fat by 13 grams per day (P=.002), saturated fat by 5 grams (P=.03), and iron by 1.7 milligrams (P=.04).

Whereas total protein intake was decreased by 17 grams for the SL arm (P=.05) and 13.7 grams per day for the HL arm (P=.02), and total fiber intake decreased by 5 grams for the SL arm (P=.05) and 6 grams per day for the HL arm (P=.04) (Multimedia Appendix 3). Moreover, the same nonenergy adjusted model in Multimedia Appendix 3 also showed that total carbon footprint was statistically significantly decreased over time in the SL arm by 0.8 kilograms of CO₂e per day (P=.001). Overall, the SL arm shows more favorable outcomes related to ecological sustainability than the HL arm with respect to total diet-related carbon footprint and a reduction in protein intake over the 8-week intervention period.

There were no between-arm differences at baseline for the SL versus HL arm in MET hours (41.5 vs 41.4) and total sedentary time (h/day; 9.1 vs 8.9). However, the arms differed in baseline active transportation, with the SL arm reporting more minutes per week (43.9 vs 16.8) and higher frequency of active commuting (1.2 vs 0.2 times per week). Additionally, no between-arm interactions were found for the mentioned variables, indicating that educational workshops in the SL and HL arms, respectively, had no impact on physical activity and active transportation (Multimedia Appendix 1). However, when analyzing within-arm changes over 8 weeks, sedentary time decreased statistically significantly in the SL arm (P=.03), where total sedentary time decreased by 0.4 hours per day between baseline and 8 weeks (Table 4).

The individual diet-related goals described above were followed up at workshops 4 and 6, and in general, the self-rated goal fulfillment was high, that is, close to 7, on a scale from 1 to 10; see Table 5. Moreover, the participants perceived support for behavioral change toward a more sustainable or healthy diet, respectively, at workshops 4 and 6, which was rated high, that is, 6.7 or above on a scale from 1 to 10, for the facilitating researcher team, colleagues, family members, and managers, with no differences in ratings between the SL and HL arms.

Participants in the SL and HL arms identified and formulated perceived barriers and facilitators for behavior change during workshop 4, and these were analyzed by the research team. Despite differences in goal settings (sustainable vs healthy lifestyle), there was considerable overlap, and therefore, all barriers and facilitators merged. In total, 2 themes, 13 categories, and 29 codes were identified, as shown in Table 6. In addition, each category was linked to a specific COM-B mechanism and illustrated by a quote from one of the participants. The first theme, “Barriers to adopt a sustainable or healthy lifestyle” included 8 categories. In brief, high prices of fruits, vegetables, and organic products, worsened by inflation, were a common barrier. Restaurant visits and takeaway food presented challenges due to the lack of transparency in ingredient sourcing and the presence of unhealthy options. Lack of time and planning, influenced by family dynamics and social events, were further complicated by holidays and associated gatherings. Moreover, personal barriers such as difficulty breaking old habits, lack of motivation, and insufficient culinary skills made it challenging for participants to adopt and maintain a sustainable or healthy lifestyle.

The second theme, “Facilitators to adopt a sustainable or healthy lifestyle” included 5 categories describing facilitators to adopt sustainable or healthy lifestyle habits. In brief, access to healthy options at work, such as salad bars and fruit, made it easier to maintain a healthy diet. Planning and preparation were essential, with participants describing routines like meal planning, grocery shopping for the whole week, and prepping food in advance as perceived facilitators. Moreover, increased knowledge about healthy food and its environmental impact was another important facilitator where the increased understanding of components of a healthy and sustainable diet empowered individuals to make informed food choices. Additionally, creating healthy lifestyle habits, such as exercising more and managing weight and high blood pressure, also supported progress toward personal dietary goals.

In this study of 37 generally healthy and normal-weight office workers, cluster-randomized to either a sustainable lifestyle or a healthy lifestyle arm, we found a statistically significant interaction for daily diet-related kg CO₂e per 1000 kcal from food, indicating that the sustainability education influenced health behaviors differently than the health education over the 8-week period.

Moreover, the total diet-related CO₂e decreased by 0.8 and 0.4 kg CO₂e per day in the SL and HL arms, respectively, corresponding to a potential reduction of 292 and 146 kg CO₂e per person per year. Given that red meat production generates a high carbon footprint whereas plant-based foods generate a low amount, and that high intake of red meat is associated with an increased risk of noncommunicable diseases like obesity, cardiovascular diseases, type 2 diabetes, and certain types of cancer [7-9,53,54], the present approach, which aims to promote both human and planetary health, is worth exploring further. Moreover, considering the office as an arena for promoting a sustainable lifestyle has the potential to reach a large portion of the population.

The reduction of CO₂e may be the result of the lower intake of energy by 223 and 346 kcal per day for the SL and HL arms, respectively. This is the result of lower intake of total carbohydrate, fat, and, in particular, protein intake between baseline and week 8. However, at the same time, vitamin C increased when expressed as mg per 1000 kcal, indicating that the nutrient density increased, a marker for improved dietary quality and consumption of fruits and vegetables. The energy restriction was not part of the educational workshops and therefore unexpected, but it indicates that the intervention tools and functions were feasible in a society characterized by an unhealthy food environment and overwhelming amounts of foods available consistently. Moreover, a reduction of sedentary time by 0.4 hours per day, corresponding to 2.8 hours per week, was observed in the SL arm. Together, these findings possibly indicate that sustainable and healthy lifestyle practices can both support healthier behaviors in office workers, and be of great relevance, as this population is known to be sedentary at work [55]. This reduction in sedentary time may be partly explained by contextual factors identified in the Our Voice data. Participants in the sustainable lifestyle arm noted barriers to everyday movement, such as limited access to stairwells, and showed stronger motivation to create opportunities for light physical activity during the workday. Between the 2 iterations in data collection, that is, the fall of 2022 and spring of 2023, Office B introduced the option of booking walk-and-talk meetings directly in the digital calendar system. This organizational change likely supported active meetings in line with the message in the sustainable lifestyle to break up prolonged sitting time. Together, these environmental and organizational factors may have contributed to the observed decrease in sedentary time in the sustainable lifestyle arm. However, no changes were found in the time or frequency of active travel, even though the sustainable lifestyle arm took more photos of these elements and initiated a bike-sharing program; see Ullberg et al [36].

This workplace-based educational intervention was grounded in theoretical frameworks targeting dietary and physical activity behaviors [34]. Given that a substantial portion of the adult population spends considerable time at work daily, office-based interventions allow many individuals to participate simultaneously [17,56]. From the employees’ perspective, these interventions are convenient, easily accessible, and foster interaction and mutual support for behavior change efforts [57]. By targeting capabilities, opportunities, and motivation based on the Behavioral Change Wheel and COM-B [58], we enhance the intervention’s effectiveness in the office context. Moreover, implementing the COM-B mechanisms in combination with the socioecological model increases the understanding of behaviors in relation to the workplace environment [31,32].

The goals set by participants at the beginning of the intervention aligned well with each intervention arm. Although there was some overlap, such as the goal of consuming 500 grams of fruits and vegetables per day and 500 grams of red meat per week, the SL arm included more goals related to organic foods and increasing the number of vegetarian meals per week, as compared to the HL arm. Despite these differences, the barriers and facilitators described were quite similar, likely due to the inherent complexity of changing one’s diet, whether the change is for health or sustainability reasons [59,60].

Both arms had approximately an 80% attendance rate at the educational workshops, and participants reported overall strong social support from colleagues, researchers, and managers. In addition, the goals formulated within each arm aligned well with the content of the corresponding educational workshops. Taken together, these findings indicate that key components were implemented as intended within this pilot. However, the intervention is time-consuming and places a burden on the company’s already overbooked conference rooms and participants’ schedules, making it hard to be compatible with the existing system and to allocate enough infrastructure support. Therefore, future iterations of the SOFIA study may incorporate digital education sessions, use interactive online course modules, and involve the human resources department to a greater extent. Additionally, the study participants identified potential barriers and facilitators to healthy and sustainable lifestyles outside or beyond the context of the workplace, such as planning, motivation, takeaway food, and the complexities of family dinners. These factors should be highlighted in future interventions in connection to COM-B mechanisms, such as social and physical opportunities, automatic and reflective motivation, and psychological capabilities to further support behavioral change based on the Behavioral Change Wheel [58].

This is a pilot study with a small sample size with limited statistical power to detect the intervention’s effects on nutritional and physical activity variables and may be subjected to random bias. However, we did find an intervention effect for daily diet-related kg CO₂e per 1000 kcal and a reduction in sedentary time in the SL arm, indicating that a future scaling of the present study is reasonable. However, the small sample size of predominantly women indicates that a more effective recruitment strategy is necessary for a full-scale trial. The use of a validated food and physical activity questionnaire (Meal-Q and Active-Q) to assess dietary intake, carbon footprint, and physical activity was a strength [44,49], but self-reported data may still introduce social desirability and recall bias, resulting in potential over- or underestimation of the true dietary intake and level of physical activity. To overcome these issues, CO₂e and nutrients were expressed per 1000 kcal and MET-hours adjusted to per 24 hours to highlight the quality and not only the absolute quantity.

The use of cluster randomization offered several strengths and limitations. By targeting geographically separated groups, the intervention minimizes cross-group influence. However, this method has drawbacks, as individuals within a cluster often resemble each other more than those in different clusters, requiring a larger sample size to account for the lack of variation as well as imbalances between groups. Despite these challenges, the SOFIA study design takes advantage of social influence within groups, as colleagues can support each other, potentially enhancing the intervention’s effects. Also, the upscaled version of the SOFIA study should incorporate intracluster correlation coefficients to provide a more accurate assessment by allowing for appropriate adjustments in the statistical models.

This study has been retrospectively registered as a full-scale clinical trial on ClinicalTrials.gov (ID no. NCT06698094) and reports data according to the CONSORT checklist (Checklist 1). By registering and publishing a study protocol and additional findings from the SOFIA pilot phase [36], we have maintained transparency about our hypotheses and planned outcomes, which guides and strengthens the development of the SOFIA study [34]. Moreover, it addresses issues such as p-hacking and selective reporting of variables [61].

A previous Swedish study using Meal-Q, including 5364 well-educated adults aged 18‐45 years, showed that the carbon footprint from their food consumption was, on average, 4.7 kg CO₂e per person per day, which corresponds to 1.7 tons of CO₂e per person and per year [62]. At the end of the present intervention, the SL arm had a mean CO₂e per person and day of 3.0 kg, corresponding to 1 ton per year, while the HL arm had a mean CO₂e per person and per day of 3.6 kg, corresponding to 1.3 tons of CO₂e per person per year, in line with lacto-ovo-vegetarian and pescetarian diets in other studies [63]. This suggests that relative minor dietary changes can result in CO₂e emissions from food close to 1 ton per person per year, a target number set by the Swedish Environmental Protection Agency back in 2015 [64].

However, this study is one of the few that have investigated CO₂e emissions from food as the primary outcome in an intervention study and the only workplace-based intervention known to the authors. A recent randomized controlled trial in Sweden assessed the effects of an app-based educational intervention on the dietary intake of individuals with type 2 diabetes. The 12-week intervention provided participants with recommendations from the Swedish Food Agency, hypothesizing that healthier food choices would also reduce CO₂e emissions per person per day. However, the ad hoc analyses showed no difference in CO₂e emissions between the intervention and control groups [65], and this may be explained by the lack of information about diet-related carbon footprint during the intervention.

A 2-armed multicomponent behavioral randomized controlled trial in the United Kingdom used meat substitutes, information, and social support among adults and resulted in statistically significant reductions in CO₂e after 8 weeks. The intervention group reported emissions of 31.7 kg per week (4.5 kg/d), compared to 40.3 kg/wk (5.7 kg/d) in the control group at the end of the intervention [66]. Moreover, in a pilot study from the United States, seminars in academic campus courses were used to explore food systems and environmental sustainability. Students in the intervention group lowered their daily dietary carbon footprint by an estimated 0.5 kg CO₂e per 2000 kcal per day [67]. Another study compared a sustainable education cluster with those in a control cluster, and at the 6-month follow-up, the dietary footprints for students in the sustainable education cluster were 4.15 kg CO₂e per day for female students and 5.72 kg CO₂e per day for male students, compared to 4.94 kg CO₂e per day for females and 6.95 kg CO₂e per day for males in the control cluster [68].

Comparisons with previous work conclude that interventions should specifically target sustainable lifestyle aspects, such as reducing meat intake through substitutes or increased plant-based foods, compared to general health recommendations when aiming at reducing carbon footprint from food. Interventions should also include social support, goal-setting, self-monitoring, action planning, prompts and cues, recipes, culinary skill tips, and education focused on the environmental impact of food production. Compared to the present study, participants in previous studies have generally higher levels of diet-related CO₂e per day at the end of each intervention, levels that were closer to those of our participants’ levels at baseline. This could be due to the use of a different LCA calculation and allocation, making it challenging to quantify and compare reduction in CO₂e, or it may reflect that our study population had different food habits due to high health literacy since they were well educated, had normal BMI, were nonsmokers, and were physically active, potentially resulting in lower meat consumption at baseline. However, protein intake, particularly meat consumption, is high in a typical Western diet [69] and substantially contributes to diet-related CO₂e. Therefore, our findings are likely generalizable to office workers in countries where a Western diet is common.

This study suggests that healthy lifestyle recommendations embedded in sustainable development can reduce participants’ food-related CO₂e in a workplace context. Additionally, the present study indicates that adopting a sustainable lifestyle has a small but significant effect on the intake of key nutrients such as vitamin C, and reduction of total sitting time, indicating that a sustainable lifestyle is a healthy lifestyle. Moreover, the goals set by participants aligned well with the content of the corresponding educational workshops, and the workshop attendance rate was high (80%), indicating good feasibility of study procedures. Given the complex global challenges of climate change, environmental degradation, and the epidemic of metabolic diseases, future intervention studies about healthy behavior should therefore be integrated with sustainable development.