In recent years, growing concerns regarding climate change, animal welfare and personal health have influenced the population’s dietary patterns(Reference Sabate and Soret1,Reference Willett, Rockstrom and Loken2) . For instance, exclusion of animal-based foods such as meat is likely to have various health and planetary benefits although potential negative health consequences have been pointed out(Reference Willett, Rockstrom and Loken2–Reference Green, Milner and Dangour5), including compensatory intake of discretionary food items with high sugar content(Reference Payne, Scarborough and Cobiac4,Reference Hendrie, Baird and Ridoutt6) . Nevertheless, diets based on nutritional recommendations are in general lower in greenhouse gas emissions (GHGE) than average consumption patterns in the population(Reference Sjors, Hedenus and Sjolander7–Reference Hyland, Henchion and McCarthy9). Despite increasing knowledge about diet-related climate impact, future improvements may be hindered by issues of affordability, lack of knowledge and resistance to change(Reference Masset, Vieux and Verger10–Reference Macdiarmid, Douglas and Campbell13). The present study describes trends in diet-related GHGE that have occurred during the millennium in western Sweden, with focus on potential characteristics of the population that may be associated with adoption of low-GHGE diets.
Methods
Two cross-sectional health examination surveys were conducted in 2001–2004 (T 1) and 2014–2018 (T 2). The average time between the two surveys was 13·7 years. Women and men aged 25–34, 35–44, 45–54, 55–64 and 65–75 years were randomly selected from the population. The examinations included physical measurements and self-administered questionnaires on health and lifestyle(Reference Mehlig, Berg and Bjorck14). The majority of participants in the two youngest groups were newly recruited at T 2, while the older participants had participated in the first survey and moved to a higher age group at T 2. The oldest group at T 1 was not included at T 2 because the participants exceeded the age limit for secular comparisons (see online supplementary material, Supplemental Fig. 1). Participation rates were comparable (approximately 40 %) at T 1 and T 2 (Reference Mehlig, Berg and Bjorck14).
At each time period, an eighty-six-item FFQ was embedded in the health survey. The FFQ was developed and validated at Karolinska Institute in Stockholm(Reference Messerer, Johansson and Wolk15–Reference Berg, Lappas and Wolk17). Food frequencies were combined with age- and sex-standardised portion-size estimates to calculate food specific and total food intake in kg/d(Reference Berg, Lappas and Wolk17). Incomplete FFQ with more than eight missing items were excluded (129 at T 1 and 19 at T 2), and the final analytic sample included 2569 individuals at T 1 and 2119 at T 2.
GHGE estimates in units of kg CO2 equivalents (CO2e) per kg consumed food were extracted from the RISE Food Climate Database(18), which is based on life cycle analyses of foods representing Swedish consumption patterns. Estimates were collected from consecutively updated studies, with the latest being the most reliable, and valid for both time points in this study. Around 70 % of the GHGE estimates applied in this study were based on production in Sweden and included GHGE from primary production to industry gate. GHGE values included transport to but not within Sweden and generally refer to the edible parts of foods. Specific GHGE estimates were derived for the eighty-six individual food items from the FFQ. The individual food items were then pooled into nineteen food groups (Fig. 1), and an average GHGE estimate was derived for each food group weighting estimates for individual items based on national consumption patterns. These nineteen food groups combined food items of the same origin (e.g. meat, vegetables and dairy), and with similar climate impact distinguishing for instance ruminants from other types of meat, and regular from low-fat dairy products.
Absolute changes in food intake (A) and in greenhouse gas emissions (GHGE) score (B) on food group level. Food groups are further divided into three categories: animal-based (top), plant-based (middle) and discretionary foods (bottom)
Statistical analyses
For each individual, we calculated the intake f j in kg/year for each food group (j = 1–19). These food intakes were multiplied with the conversion factor c j (= estimated kg CO2e per kg consumed food), which gives the yearly GHGE due to consumption of foods from group j, GHGE j = c j × f j (kg CO2e/year). Total CO2 emission is given by GHGE total = Σ GHGE j = Σ c j × f j . The ratio of total CO2 emission over total food intake Σ f j ,
$$ratio = {{\sum\nolimits_{j \ = \ 1}^{19} {{c_j}{f_j}} } \over {\sum\nolimits_{j \ = \ 1}^{19} {{f_j}} }}$$
gives an estimate for the climate impact in kg CO2e/kg consumed food in an individual. The mean value of individual ratios gives the diet-related climate impact per kg consumed food in this population. In addition, source-specific climate scores (animal-based, plant-based, discretionary foods) were divided by total food intake in order to investigate whether changes in source-specific climate scores were explained by secular changes in total food intake.
Dietary information was studied in relation to time period. Because some participants were measured at both T 1 and T 2, the main analyses were stratified into five age bands between ages 25 and 75 years. In this way, statistical comparisons between time periods were performed between independent samples, and no longitudinal changes were considered at the individual level (see online supplementary material, Supplemental Fig. 1). Non-parametric tests examined time period differences in dietary and background characteristics (χ 2 test for categorical variables and Wilcoxon rank sum test for continuous variables). Linear regression was used to analyse the logarithmically transformed GHGE score as a function of time, with adjustment for sex, exact age, BMI and education, giving the relative difference in GHGE score at T 2 relative to T 1 in percent. Effect modification by sex, overweight (BMI ≥ 25 kg/m2) and university education was examined by introducing product terms with time period into the age-specific regression models (see online supplementary material, Supplemental Fig. 2). Analyses were performed using SAS (version 9·4; SAS Institute) and MATLAB (R2016b; The Math Works, Inc.). Statistical significance was set at P-value < 0·05 (two-sided tests).
Results
Descriptive background data on the population are shown in Table 1. The prevalence of overweight was stable between time periods, whereas significant period differences in university education were observed. These increases may be attributed to secular trends in educational standards in the underlying population and to self-selection among both newly recruited and returning participants. Additional analyses (not shown) confirmed that the lack of trend in overweight was independent of increasing educational attainment.
Age-specific characteristics of the population including background covariates as percentage, followed by dietary outcomes as median and interquartile range. Dietary characteristics are divided into total, animal-based, plant-based and discretionary foods. Statistically significant period differences are indicated (ref = T 1)
GHGE, greenhouse gas emissions.
*** P < 0·001; **P < 0·01; *P < 0·05.
† Period differences by χ 2 test.
‡ Period differences by Wilcoxon rank sum test.
§ Climate score divided by total food intake.
Dietary characteristics within each 10-year age band were compared at T 2 v. T 1. Significant decreases in total climate scores were observed in all five age groups and were largest (−374 kg CO2e/year) in the youngest group (Table 1). Comparing source-specific scores, the largest differences in GHGE were seen for animal-based foods suggesting that improvements in total GHGE were mostly due to lower consumption of animal products. This trend was accompanied by some increases in plant-based food consumption in the two younger age groups. Finally, GHGE from the discretionary category decreased significantly in all age groups. Time period differences in absolute GHGE score were generally confirmed when considering its ratio to the total amount of food consumed, an indicator of changed dietary GHGE pattern rather than amount, adjusting for period differences in total food intakes.
Multivariable regression models (Table 2) confirmed the significant reductions in GHGE in all five age groups. The largest differences were consistently seen in the youngest age group and the smallest differences in the 45–54-year-old group. The magnitude of the crude effects (model A) hardly changed after adjustments for age, sex, education and BMI (models B and C). The secular differences were slightly attenuated but remained statistically significant in all age groups after further adjustment for total food intake (model D). Results from models C and D also implied that decreases in GHGE could not be attributed to the increasing educational level between the two periods. Considering education per se, no differences in GHGE were observed between individuals with university v. lesser education, at either time period (not shown). In contrast, BMI was positively associated with GHGE scores at T 2, with and without adjustment for the total amount of food consumed: GHGE in overweight individuals was 3 % higher compared with those with lower BMI (P = 0·01, adjusted for age, sex, education and total intake, not shown). Furthermore, the magnitude of GHGE differences over time tended to be smaller in overweight individuals, with significant time by overweight interaction in age group 35–44 (see online supplementary material, Supplemental Fig. 2 middle panel). There were no interactions of time period with education or sex (see online supplementary material, Supplemental Fig. 2).
Secular trends in climate impact by age group, with effect sizes expressed as percent change in greenhouse gas emissions (GHGE) score in T 2 v. T 1 †
† Regression of log (GHGE) on time point and covariates, with results expressed as (exp (b)–1) × 100 = % GHGE change in T 2 v. T 1.
Finally, Fig. 1 shows the secular trends for specific food groups within the broader categories of animal, plant and other sources. Contrasting patterns may be observed regarding the two measures of secular change, i.e. period differences in foods consumed and in food-related GHGE scores. For instance, an apparent replacement of light dairy products with a smaller amount of full fat ones (panel A) produced a net pattern of increasing GHGE for these two items considered together (panel B). Moreover, trends in consumption of the mixed red meat group (mainly processed meat items) dominate the decrease in GHGE scores compared with all other items (panel B). Much smaller changes were observed in both food intake and GHGE from ruminant animals (beef, veal and lamb).
Discussion
The current study showed that Swedish men and women in all age groups decreased their dietary GHGE over approximately 14 years. In particular, the younger age groups (25–44 years) consumed less animal-based and more plant-based foods. Decreases in discretionary foods were seen in all age groups. There was no accompanying difference in overweight prevalence over time, in contrast to earlier trends of increasing BMI and waist-to-hip ratio in this population in the late 20th century(Reference Berg, Rosengren and Aires19). However, the most recent examination (2014–2018) showed that participants with overweight had higher diet-related GHGE than non-overweight participants, independent of amount of food consumed. In this context, it is noted that total food consumption may be considered a proxy for energy consumption. Although energy intake was not estimated for the second time period, the high correlation between total food and energy intake in 2001–2004 (Pearson’s correlation coefficient 0·79, P < 0·001) motivated our decision to treat food intake as an indicator of energy intake.
Food production causes around one-third of global GHGE, and dietary changes hold great potential for reducing these emissions(Reference Hallstrom, Carlsson-Kanyama and Borjesson20). Changes in dietary patterns of the younger age groups studied here, with major shifts in both animal- and plant-based foods, are promising, but improvements appear to be smaller in all other age groups, particularly in 45–54-year-olds. Decreases in GHGE from discretionary foods occurred in parallel with a stable prevalence of overweight and obesity. The association between overweight status and higher dietary GHGE in this study is consistent with results from a less urbanised Northern Swedish cohort(Reference Strid, Hallstrom and Hjorth21). Our observation that the youngest age groups showed highest GHGE in 2001–2004 (3·4 kg CO2e/d) and lowest levels in 2014–2018 (2·4 kg CO2e/d) may be an indication that food products with lower carbon footprint have become more socially desirable, available and affordable, especially to younger adults.
While longitudinal decrease in dietary GHGE was reported for cohort studies in the Netherlands(Reference Biesbroek, Bueno-de-Mesquita and Peeters22) and Northern Sweden(Reference Hjorth, Huseinovic and Hallstrom23), to our knowledge, this is the first study to document decreasing secular trends of dietary GHGE in same-aged adults compared 2001–2004 and 2014–2018. Strengths of this study include the population-based recruitment and the repeated cross-sectional design based on similar survey methodologies, together with derivation of GHGE estimates specific to the Swedish diet. Among the limitations are consistently low participation rates, probable dietary reporting biases and numerous assumptions involved in GHGE estimation. Moreover, the FFQ method does not reflect complete dietary intake, but relatively broad-ranged definitions allowed to aggregate few items newly introduced at T 2 into existing food item categories, which were comprehensive regarding, e.g. seasonal variations.
Conclusion
In conclusion, the magnitude of the secular differences in the younger age groups was promising, but the lesser effects in other age groups underscore the need for effective policies to improve climate impact of diets. The consistent decreases in discretionary foods indicate a healthy trend with a small but favourable climate impact, whereas lack of changes in consumption of meat from ruminant animals suggests a potential for greater improvements. Finally, the positive association between BMI and GHGE in the recent survey is consistent with potential health benefits of a dietary shift, while at the same time suggesting that the climate message might not be reaching individuals with overweight.
Acknowledgements
Acknowledgements: Not applicable. Financial support: This study was supported by grants from the Swedish Council for Health, Working Life, and Welfare (Forte) and the Swedish Research Council for Environment, Agricultural Sciences, and Spatial Planning (Formas) and by grants from the Swedish state under the agreement between the Swedish government and the country councils, the ALF-agreement (30411). Conflict of interest: There are no conflicts of interest. Authorship: I.B., K.M. and L.L. formulated the research question and prepared the manuscript; K.M. and I.B. conducted the data analysis; S.K., L.L., K.M. and M.H. supervised the research; M.B. and J.S. assisted with methods pertaining to the use of the RISE Food Climate Database. All authors have participated in writing the manuscript and approved the final version. None of the authors has any commercial association that would pose a conflict of interest. There have been no previous publications of this work. Ethics of human subject participation: This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving research study participants were approved by the regional ethics review board (237/2000). Written informed consent was obtained from all participants.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980020004073
Open access
