Using a whole-diet approach, we evaluated here how the partial replacement of animal-sourced proteins with plant-sourced proteins affected the estimated climate impact of the diet, and whether there is a cross-sectional relationship between the climate impact and some health indicators. The study was carried out using data from a 12-week randomised controlled trial specifically designed to study the nutritional and health effects of a dose-dependent switch from animal to plant-based protein sources [33]. Our main result is that replacing animal-sourced proteins with plant-sourced proteins reduces the climate impact of the diet, meanwhile the relationship between climate impact and health biomarkers is more ambiguous indicated by both beneficial and harmful consequences within lower climate impact.

In the PLANT diet, where 70% of total protein intake originated from plant-sourced proteins, the climate impact was 25–38% smaller than the climate impact of the ANIMAL diet with 70% of animal-source proteins, depending on the comparison basis, i.e., the functional unit (FU) used. By contrast, the climate impact of the 50/50 diet with a similar amount of plant and animal-sourced proteins was 7–20% smaller than the climate impact of the ANIMAL diet, and the difference was statistically significant only regarding the realised diets. The PLANT diet would therefore be the best in terms of climate impact alone.

As is typical in LCA studies, using different FUs produced different results. The choice of FU is crucial, and according to the LCA methodology [46, 47], it should be in line with the goal and scope of the study; different FUs fit into different situations and answer different questions [46]. In general, when assessing the climate impact of diets, the standardisation is usually done for energy content, because it is considered a good basis for comparability. Our result (− 38%) for an energy-adjusted PLANT diet with a reduction of animal-source protein of 57% (ANIMAL vs. PLANT) is well in line with the estimates presented in the literature which have shown that the climate impacts of vegetarian diets can be 20–35% lower than those of current Western diets, and the climate impacts of vegan diets can be 25–55% lower [27]. Also, diets composed in line with the nutrition recommendations but containing less meat and more plant-based foods than current (Western) diets can achieve a 20–50% reduction in the climate impact [47]. However, our result is somewhat lower than the maximum reduction potential based on a global analysis [20] where a reduction of up to 85% of the climate impact of diet in high-income countries has been reported if more plant-rich diets were adopted. In contrast, our results (− 15%) for an energy-adjusted 50/50 diet did not achieve any of these results. However, the dietary change also was quite moderate in this group, with only a 28% reduction in animal-sourced proteins. The differences between the studies may also be due to the different dietary references used. For example, the average food consumption data used a global analysis [20] for high-income countries differed largely from the current Finnish diet [36], as did the climate impact coefficients for foods. However, different research objectives and methodological differences between the studies may also explain the differing results.

The results for diets as such are especially in line with the results from a study of the nutritionally adequate dietary scenarios for the whole Finnish population [48], in which halving meat intake reduced the climate impact of an average diet by 14% and reducing meat intake to a third by 20%. In that study, the intake of dairy products was not reduced, except for a slight decrease in cheese consumption. This suggests that dairy consumption is not as crucial for the climate impact of the diet as meat consumption. Interestingly, the results of these two studies on Finnish diets are in excellent agreement, although food grouping and the compilation of the corresponding climate impacts data utilised different approaches. This suggests that the LCA-based results are robust with respect to climate impact data. More generally, it may also suggest that LCA-based assessment results of dietary climate impacts may not be highly sensitive to the consistency of source data for products’ climate impact although strict criteria have been proposed to data quality of LCA data in dietary assessment [50]. However, this robustness should be further validated in the future research.

While between-group variation in climate impacts could be expected, the large within-group variation in climate impacts was surprising. When meat consumption in personal diets was examined more closely, it was found that the groups were somewhat mixed at the extremes: for example, based on meat consumption, the individuals in the 50/50 group who consumed the most meat would in fact have belonged to the ANIMAL group, and those who consumed the least to the PLANT group. If the analysis had been done based on the actual consumption of animal- and plant-sourced proteins, the differences between groups would have been clearer than the differences obtained with mismatched individuals affecting the group measures. However, it is also interesting that such mixing between the intervention groups occurred even though the instructions given to the participants regarding how to follow the diet were extremely detailed. This may indicate that participants had difficulties in interpreting or following the instructions. This supports the claim that large-scale dietary changes are difficult in practice [50, 51]. Changes in consumption patterns can be particularly difficult when it comes to high-protein foods because plant-based proteins are less digestible and may cause gastrointestinal symptoms. These potentially hindering factors deserve attention in research on dietary change. However, mixing between the intervention groups might also be a consequence of the individual variation in food consumption. For example, the participants were supposed to consume the delivered foods within a week, but the food diaries were only collected for four days. Thus, it might be that they ate either more animal or plant products during the diary days than the rest of the week, especially in the 50/50 group.

In this study, the climate impacts of diets were assessed as secondary outcomes of the nutritional intervention, in which the nutritional and health consequences of realised intervention diets were investigated. In previous work, we observed that some health benefits were achieved in both the 50/50 and PLANT groups, such as better lipid profile and higher fibre intake [33] and in the PLANT group lower concentrations of colorectal cancer risk markers, N-nitroso compounds [42]. However, some adverse health consequences such as lower vitamin B12 and iodine status and intake, as well as undesirable changes in bone turnover, were seen as a consequence of lower calcium and vitamin D intake, especially in the PLANT group [34, 35]. These changes in health indicators and nutrient intakes are a consequence of decreasing the amount dairy and meat dose-dependently in the diets. Calcium and vitamin D, usually from dairy products, are needed for proper bone function [52], whereas N-nitroso compounds are metabolites of red and processed meat [53].

We analysed correlations between climate impacts and health indicators that reflect biomarkers of risk for most important non-communicable diseases causing burden in the Western countries [39,40,41]. These indicators were risk of cardiovascular disease (blood lipids, blood pressure), type 2 diabetes (insulin, glucose), colorectal cancer (N-nitroso compounds), and osteoporosis (bone health; the ratio of markers for bone formation and resorption). Regarding the correlations, the results were mostly similar for the climate impact of diet as such and for energy-adjusted values (within an exception of total N-nitroso compounds where the correlation was not significant for the climate impact of diet as such). The dietary climate impact correlated positively with total and haem N-nitroso compounds and animal-sourced iron intake but negatively with fibre intake, indicating harmful consequences to gut health, particularly the risk of colorectal cancer. N-nitroso compounds and animal-sourced iron originate from red and processed meat, and especially beef is associated with a very high climate impact (per product mass) compared to other products. This supports the recommendation that the consumption of red and processed meat should be limited for both climate and health reasons (e.g [54]).

A significant correlation of climate impact with a positive status of bone turnover was observed, indicating that foods contributing to climate impact have beneficial effects on bone health. The climate impact and calcium intake had also a positive correlation. The intake of calcium is related to dairy products, which are widely consumed in Finland [36], particularly in liquid form. The climate impact of dairy products varies a lot (per product mass), with the climate impact of milk and other liquid dairy products being quite low, but those of cheese and butter being high (Supplement B). In the intervention, dairy product consumption was reduced in the 50/50 and PLANT diets along with meat. It is possible that the correlation between the climate impact of diet and calcium intake is also affected by meat consumption which is in turn related to the climate impact of the diet. This is supported by previous research, according to which a significant reduction in climate impact can be achieved with diets with little or no meat, but with relatively high milk consumption [48, 55, 56]. However, these analyses have not taken into account the interdependence between beef and milk production, on the basis of which the reduction of milk consumption to a moderate amount can be justified [55].

There was a positive correlation observed between climate impact and saturated fat. This reflects the fact that animal products, excluding low-fat dairy, are rich in saturated fat. Interestingly, however, many of the health indicators generally associated with high saturated fat intake, such as blood lipids and blood pressure [54], did not correlate with the climate impact (except the weak but significant correlation with non-HDL-cholesterol in an energy-adjusted analysis). The insignificant correlations with blood lipids or blood pressure may not be surprising, as not all protein sources with higher climate impact are necessarily high in saturated fat. A possible reason for not observing any significant correlations between climate impact and glucose metabolism markers, BMI, or blood pressure may be that our study subjects were generally healthy and not obese, as various factors in addition to dietary changes are reflected in these health indicators.

Based on these results, it is not obvious that the cross-sectionally assessed climate impact and health effects would go hand in hand. These results show that the climate impact of a diet can correlate with both positive or negative health biomarkers that are mediated through the eaten foods and their nutrients. To our knowledge, this has not been observed in previous studies.

This interpretation is also supported by a wider previous study of the nutrient composition of the same diets according to which the 50/50 group provided mostly an adequate amount of nutrients whereas in the PLANT diet the mean intakes did not meet the current Finnish reference values of some critical nutrients such as vitamin D and calcium [33,34,35, 37]. It is notable that the intake of iodine was fairly similar in 50/50 and PLANT groups but there were individuals below the reference values [35, 37]. Moreover, vitamin B12 intake exceeded the reference values in both groups, but there were individuals with inadequate intakes, especially in the PLANT group [35, 37]. Vitamin D is a common public health challenge, as the mean intakes were below the recommendations [34, 37] in all three study groups and 40% had inadequate vitamin D status [34, 52]. However, the intakes were lower in 50/50 and PLANT groups compared to the ANIMAL group [34]: if the amount of vitamin D fortified fluid milk in the diet is reduced, vitamin D should be received from other sources such as from fortified plant-based milk or supplements to ensure the adequate intake. The recommendations for vitamin B12 and calcium intake have increased in the latest Nordic Nutrition Recommendations 2023 [54]; if compared to these reference values, neither the 50/50 nor PLANT group reached the adequate intakes. Lower intakes of those critical nutrients in predominantly plant-based diets have also been observed previously [16, 57, 58]. Regarding the health outcomes, the PLANT diet provided both more health benefits such as a better lipid profile and lower concentrations of N-nitroso compounds [28]. On the other hand, the PLANT diet was also associated with some detrimental effects like accelerated bone turnover [29] which in the long run can increase fracture risk.

Even though these health impacts have also been observed in the previous studies with plant-based diets [6, 59,60,61,62], we cannot conclude that any of the diets is optimal from all three aspects: the environmental effects, dietary adequacy, and health effects. One also must take into account that the more we aim for sustainable diets by reducing the consumption of animal-sourced proteins, the more vulnerable our diets may become. This vulnerability arises from increased reliance on food fortification and dietary supplements to obtain critical nutrients, even though the overall health effects are likely to be positive. Overall, if a predominantly plant-based diet is to be followed, it is important to pay particular attention to the intakes of iodine, vitamins B12 and D, and calcium.

One strength of this study lay in its evaluation of realised instead of modelled diets. It was especially significant that the climate impacts and health biomarkers were studied from the same diets, so their relations could be analysed. However, a limitation of this study is that these were not self-selected diets, the evaluation of which would have yielded the most realistic results on the climate impact and health effects of a possible dietary transition to a more plant-based diet. Moreover, the analyses of health indicators are cross-sectional and relate to the climate impact, not the specific diets per se. Regarding health indicators, the markers were chosen based on their validity on the risk factor markers of several burdening non-communicable diseases. Besides well-established CVD and T2D risk markers, the markers of bone turnover describe acute or mid-term changes in bone metabolism, and they were chosen because in 12 weeks it is not possible to see changes in bone mineral density. Furthermore, the chosen markers were the ones recommended by the International Osteoporosis Foundation [63]. Similarly, N-nitroso compounds were used to estimate risk of colorectal cancer as they are carcinogenic compounds whose harmful effects have been established not only as molecular changes in CRC patients and laboratory animals but also in human cohort studies [53].

From a methodological perspective, one strength of this study was the use of different FUs in the assessment. In LCA, results are very sensitive with respect to the details of both the product system (diet composition) and the FU. The use of different FUs thus allows for a deeper understanding of the subject, which is why the use of several FUs is recommended [46]. In general, by using standardised energy content, e.g. 2,000 kcal or an amount of protein (e.g., one gram) as an FU, more accurate and comparable results can be obtained. These FUs also reflect at least some aspects of the nutritional quality of a diet, which is important in food LCA applications [46]. However, realised diets, such as the intervention diets in this study, differ qualitatively in many ways, which may be reflected in diet-specific nutrition and health biomarkers. Thus, the supposedly better accuracy of energy-adjusted or protein-based FUs can be misleading.

Finally, one limitation of this study is the scarcity of LCA results for a wide range of food products included in any realised diet. We used extensive data, but for many food products and dishes, we still had to use approximations. Furthermore, despite the additions we did to the source data, the final dataset was not fully harmonised. For example, we did not harmonise the methodological choices in the utilized literature such as emission models of agricultural production, impact assessment models, etc., because the data on imported products especially are expected to have large uncertainties due to the incompatibility of data sources. Thus, a harmonisation under these limitations would presumably not have resulted in increased accuracy. However, the choice of data was based on readily harmonised data sources as far as possible (Supplement B). For example, a database provided in [32 in Supplement B] was used, particularly for imported products. Data accuracy is always an issue in LCA studies, particularly in comparative product LCAs, but since this study focused on whole diets instead of single products, large uncertainties for individual products are unlikely to have outsized effects on the end result. However, this holds only under the assumption that the product grouping is sufficiently accurate and representative, taking into account the actual products sufficiently well [48, 49], as it is in this study. The shortcomings in data quality were therefore unlikely to be crucial.